1 2018-08-21 Tom de Vries <tdevries@suse.de>
3 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5 (debuginfo_early_stop): Declare.
6 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8 (debuginfo_early_stop): New function.
9 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
10 and debuginfo_early_stop.
11 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
12 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
13 (general_init): Call debuginfo_early_init.
14 (finalize): Call debuginfo_fini.
15 (do_compile): Call debuginfo_init.
16 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
18 (@item -fdump-debug, @item -fdump-earlydebug): Add.
20 2018-08-21 Tom de Vries <tdevries@suse.de>
22 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
23 flag_dump_noaddr and flag_dump_unnumbered.
25 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
27 * wide-int-range.cc (wide_int_range_abs): New.
28 (wide_int_range_order_set): Rename from wide_int_range_min_max.
29 * wide-int-range.h (wide_int_range_abs): New.
30 (wide_int_range_min_max): New.
31 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
32 case to call wide_int_range_abs.
33 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
34 (extract_range_from_abs_expr): Delete.
36 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
39 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
40 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
43 2018-08-20 Nathan Sidwell <nathan@acm.org>
44 Jeff Law <law@redhat.com>
46 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
47 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
49 2018-08-20 David Malcolm <dmalcolm@redhat.com>
52 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
53 (decl_attributes): Likewise.
54 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
56 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
57 * diagnostic-core.h (class auto_diagnostic_group): New class.
58 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
59 (diagnostic_report_diagnostic): Handle the first diagnostics within
61 (emit_diagnostic): Add auto_diagnostic_group instance.
65 (warning_at): Likewise.
66 (warning_n): Likewise.
68 (permerror): Likewise.
73 (fatal_error): Likewise.
74 (internal_error): Likewise.
75 (internal_error_no_backtrace): Likewise.
76 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
77 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
78 * diagnostic.h (struct diagnostic_context): Add fields
79 "diagnostic_group_nesting_depth",
80 "diagnostic_group_emission_count", "begin_group_cb",
82 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
83 Add auto_diagnostic_group instance(s).
84 (find_explicit_erroneous_behavior): Likewise.
85 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
86 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
87 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
88 (gimplify_va_arg_expr): Likewise.
89 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
90 (HSA_SORRY_AT): Likewise.
91 * ipa-devirt.c (compare_virtual_tables): Likewise.
93 * multiple_target.c (expand_target_clones): Likewise.
94 * opts-common.c (cmdline_handle_error): Likewise.
95 * reginfo.c (globalize_reg): Likewise.
96 * substring-locations.c (format_warning_n_va): Likewise.
97 * tree-inline.c (expand_call_inline): Likewise.
98 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
99 * tree-ssa-loop-niter.c
100 (do_warn_aggressive_loop_optimizations): Likewise.
101 * tree-ssa-uninit.c (warn_uninit): Likewise.
102 * tree.c (warn_deprecated_use): Likewise.
104 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
107 * config/i386/i386.md (eh_return): Always update EH return
108 address in word_mode.
110 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
112 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
113 TARGET_SPLIT_COMPLEX_ARG is defined.
115 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
117 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
119 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
122 * expr.c (expand_assignment): Assert that bitpos is positive.
123 (store_field): Likewise
124 (expand_expr_real_1): Make sure that bitpos is positive.
125 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
128 2018-08-20 Nathan Sidwell <nathan@acm.org>
130 * Makefile.in (CPP_ID_DATA_H): Delete.
131 (CPP_INTERNAL_H): Don't add it.
132 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
133 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
135 2018-08-20 Richard Biener <rguenther@suse.de>
137 PR tree-optimization/78655
138 * tree-vrp.c (extract_range_from_binary_expr_1): Make
139 pointer + offset nonnull if either operand is nonnull work.
141 2018-08-20 Tom de Vries <tdevries@suse.de>
143 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
144 unless the referenced die describes the added property using
145 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
146 Otherwise, add a DW_AT_location to the referenced die.
148 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
151 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
152 register_operand when calling ix86_set_reg_reg_cost.
153 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
154 Set *total to 0 for operands that satisfy x86_64_immediate_operand
155 predicate and to 1 otherwise.
157 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
159 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
160 emit a diagnostic that it is not supported and reset the option.
161 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
162 supported and consume the option. (ASM_FINAL_SPEC): New.
164 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
166 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
169 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
171 C-SKY port: Documentation
173 * doc/extend.texi (C-SKY Function Attributes): New section.
174 * doc/invoke.texi (Option Summary): Add C-SKY options.
175 (C-SKY Options): New section.
176 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
178 2018-08-17 Jojo <jijie_rong@c-sky.com>
179 Huibin Wang <huibin_wang@c-sky.com>
180 Sandra Loosemore <sandra@codesourcery.com>
181 Chung-Lin Tang <cltang@codesourcery.com>
183 C-SKY port: Backend implementation
185 * config/csky/*: New.
186 * common/config/csky/*: New.
188 2018-08-17 Jojo <jijie_rong@c-sky.com>
189 Huibin Wang <huibin_wang@c-sky.com>
190 Sandra Loosemore <sandra@codesourcery.com>
191 Chung-Lin Tang <cltang@codesourcery.com>
192 Andrew Jenner <andrew@codesourcery.com>
194 C-SKY port: Configury
196 * config.gcc (csky-*-*): New.
197 * configure.ac: Add csky to targets for dwarf2 debug_line support.
198 * configure: Regenerated.
200 2018-08-17 David Malcolm <dmalcolm@redhat.com>
202 * dump-context.h: Include "dumpfile.h".
203 (dump_context::dump_printf_va): Convert final param from va_list
204 to va_list *. Convert from ATTRIBUTE_PRINTF to
205 ATTRIBUTE_GCC_DUMP_PRINTF.
206 (dump_context::dump_printf_loc_va): Likewise.
207 * dumpfile.c: Include "stringpool.h".
208 (make_item_for_dump_printf_va): Delete.
209 (make_item_for_dump_printf): Delete.
210 (class dump_pretty_printer): New class.
211 (dump_pretty_printer::dump_pretty_printer): New ctor.
212 (dump_pretty_printer::emit_items): New member function.
213 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
215 (dump_pretty_printer::emit_item): New member function.
216 (dump_pretty_printer::stash_item): New member function.
217 (dump_pretty_printer::format_decoder_cb): New member function.
218 (dump_pretty_printer::decode_format): New member function.
219 (dump_context::dump_printf_va): Reimplement in terms of
221 (dump_context::dump_printf_loc_va): Convert final param from va_list
223 (dump_context::begin_scope): Reimplement call to
224 make_item_for_dump_printf.
225 (dump_printf): Update for change to dump_printf_va.
226 (dump_printf_loc): Likewise.
227 (selftest::test_capture_of_dump_calls): Convert "stmt" from
228 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
230 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
231 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
232 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
233 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
234 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
235 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
236 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
237 within a dump_printf_loc call to "%wu".
238 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
239 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
240 missing space after "=".
241 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
242 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
243 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
244 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
245 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
246 duplicate "vectorized" from message.
248 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
250 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
251 polyNxK_t element's TYPE_STRING_FLAG.
253 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
255 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
256 (they were unnamed before). Fix comments.
258 2018-08-17 Nathan Sidwell <nathan@acm.org>
260 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
262 2018-08-17 Richard Biener <rguenther@suse.de>
264 PR tree-optimization/86841
265 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
267 2018-08-17 Martin Liska <mliska@suse.cz>
269 * common.opt: Remove Warn, Init and Report for options with
270 Ignore/Deprecated flag. Warning is done automatically for
272 * config/i386/i386.opt: Likewise.
273 * config/ia64/ia64.opt: Likewise.
274 * config/rs6000/rs6000.opt: Likewise.
275 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
276 Remove usage of flag_check_pointer_bounds.
277 * lto-wrapper.c (merge_and_complain): Do not handle
278 OPT_fcheck_pointer_bounds.
279 (append_compiler_options): Likewise.
280 * opt-functions.awk: Do not handle Deprecated.
281 * optc-gen.awk: Check that Var, Report and Init are not
282 used for an option with Ignore/Deprecated flag.
283 * opts-common.c (decode_cmdline_option): Do not report
285 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
287 * opts.h (struct cl_option): Remove cl_deprecated flag.
288 (CL_ERR_DEPRECATED): Remove error enum value.
290 2018-08-17 Richard Biener <rguenther@suse.de>
293 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
294 across a va-arg-pack using call adjust its return value accordingly.
296 2018-08-16 Martin Sebor <msebor@redhat.com>
298 PR tree-optimization/86853
299 * gimple-ssa-sprintf.c (struct format_result): Rename member.
300 (struct fmtresult): Add member and initialize it in ctors.
301 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
302 (format_string): Handle %S the same as %ls. Set MAYFAIL.
303 (format_directive): Set POSUNDER4K when MAYFAIL is set.
304 (parse_directive): Handle %C same as %c.
305 (sprintf_dom_walker::compute_format_length): Adjust.
306 (is_call_safe): Adjust.
308 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
310 * builtins.c (c_strlen): Add new parameter eltsize. Use it
311 for determining how to count the elements.
312 * builtins.h (c_strlen): Adjust prototype.
313 * expr.c (string_constant): Add new parameter mem_size.
314 Set *mem_size appropriately.
315 * expr.h (string_constant): Adjust protoype.
316 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
317 * gimple-fold.h (get_range_strlen): Adjust prototype.
318 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
319 (format_string): Call get_string_length with eltsize.
321 2018-08-16 David Malcolm <dmalcolm@redhat.com>
323 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
324 to emit the span, rather than setting it as the prefix.
326 2018-08-16 David Malcolm <dmalcolm@redhat.com>
328 * diagnostic-show-locus.c (layout::start_annotation_line): Add
329 "margin_char" parameter, defaulting to space. Use it in place
330 of pp_space for the initial part of the margin.
331 (layout::print_leading_fixits): Use '+' when filling the margin
332 of line-insertion fix-it hints.
334 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
336 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
339 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
341 * config/rs6000/altivec.md: Don't set length attribute to the default
343 * config/rs6000/darwin.md: Ditto.
344 * config/rs6000/dfp.md: Ditto.
345 * config/rs6000/htm.md: Ditto.
346 * config/rs6000/rs6000.md: Ditto.
347 * config/rs6000/sync.md: Ditto.
348 * config/rs6000/vsx.md: Ditto.
350 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
352 * config/rs6000/altivec.md: Don't set length attribute to the default
353 value, for branch instructions.
354 * config/rs6000/darwin.md: Ditto.
355 * config/rs6000/rs6000.md: Ditto.
357 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
359 * config/rs6000/rs6000.md (length): Always define as const_int 4.
360 (unnamed conditional branch define_insn): Set length to 4 or 8
362 (<bd>_<mode>): Similar, for alternative 0.
363 (<bd>tf_<mode>): Ditto.
365 2018-08-16 Tamar Christina <tamar.christina@arm.com>
367 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
369 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
371 * doc/rtl.texi: Replace old RTX class names with new names.
374 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
376 * expmed.h (canonicalize_comparison): New declaration.
377 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
378 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
379 * optabs.c (prepare_cmp_insn): Likewise.
380 * rtl.h (unsigned_condition_p): New function which checks if a
381 comparison operator is unsigned.
383 2018-08-16 Nathan Sidwell <nathan@acm.org>
385 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
386 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
388 2018-08-16 Tamar Christina <tamar.christina@arm.com>
391 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
392 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
393 (mov<mov>): ..this and enable unconditionally.
395 2018-08-16 Tamar Christina <tamar.christina@arm.com>
397 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
399 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
401 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
402 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
403 "Common" with "Target".
405 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
407 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
408 * doc/invoke.texi (mmitigate-rop): Remove.
409 * config/i386/i386.c: Do not include "regrename.h".
410 (ix86_rop_should_change_byte_p, reg_encoded_number)
411 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
413 (ix86_reorg): Remove call to ix86_mitigate_rop.
414 * config/i386/i386.md (attr "modrm_class"): Remove.
415 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
416 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
417 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
419 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
421 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
422 allow folding of mergeh() and mergel() for the float and double types.
423 (fold_mergehl_helper): Rework to handle building a permute tree
426 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
428 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
431 2018-08-15 David Malcolm <dmalcolm@redhat.com>
433 * common.opt (fdiagnostics-show-labels): New option.
434 * diagnostic-show-locus.c (class layout_range): Add field
436 (class layout): Add field "m_show_labels_p".
437 (layout_range::layout_range): Add param "label" and use it to
439 (make_range): Pass in NULL for new "label" param of layout_range's
441 (layout::layout): Initialize m_show_labels_p.
442 (layout::maybe_add_location_range): Pass in loc_range->m_label
443 when constructing layout_range instances.
444 (struct line_label): New struct.
445 (layout::print_any_labels): New member function.
446 (layout::print_line): Call it if label-printing is enabled.
447 (selftest::test_one_liner_labels): New test.
448 (selftest::test_diagnostic_show_locus_one_liner): Call it.
449 * diagnostic.c (diagnostic_initialize): Initialize
450 context->show_labels_p.
451 * diagnostic.h (struct diagnostic_context): Add field
453 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
454 -fno-diagnostics-show-labels.
455 * dwarf2out.c (gen_producer_string): Add
456 OPT_fdiagnostics_show_labels to the ignored options.
457 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
459 (gcc_rich_location::maybe_add_expr): Likewise.
460 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
461 label" param, defaulting to NULL.
462 (gcc_rich_location::add_expr): Add "label" param.
463 (gcc_rich_location::maybe_add_expr): Likewise.
464 (class text_range_label): New class.
465 (class range_label_for_type_mismatch): New class.
466 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
467 of format_warning_va.
468 (fmtwarn_n): Likewise for new params of format_warning_n_va.
469 * lto-wrapper.c (merge_and_complain): Add
470 OPT_fdiagnostics_show_labels to the "pick one setting" options.
471 (append_compiler_options): Likewise to the dropped options.
472 (append_diag_options): Likewise to the passed-on options.
473 * opts.c (common_handle_option): Handle the new option.
474 * selftest-diagnostic.c
475 (test_diagnostic_context::test_diagnostic_context): Enable
477 * substring-locations.c: Include "gcc-rich-location.h".
478 (format_warning_n_va): Add "fmt_label" and "param_label" params
479 and use them as appropriate.
480 (format_warning_va): Add "fmt_label" and "param_label" params,
481 passing them on to format_warning_n_va.
482 (format_warning_at_substring): Likewise.
483 (format_warning_at_substring_n): Likewise.
484 * substring-locations.h (format_warning_va): Add "fmt_label" and
485 "param_label" params.
486 (format_warning_n_va): Likewise.
487 (format_warning_at_substring): Likewise.
488 (format_warning_at_substring_n): Likewise.
489 * toplev.c (general_init): Initialize global_dc->show_labels_p.
491 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
494 * builtins.c (expand_builtin_memcmp): Do not expand the call
495 when overflow is detected.
497 2018-08-15 Martin Sebor <msebor@redhat.com>
499 PR tree-optimization/71625
500 * config/aarch64/aarch64-builtins.c
501 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
503 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
505 * config/s390/s390.c (s390_reorg): Remove loop.
507 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
510 (darwin_function_switched_text_sections): Delete.
511 * gcc/config/darwin.h
512 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
514 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
517 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
518 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
519 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
521 2018-08-15 Martin Liska <mliska@suse.cz>
523 PR tree-optimization/86925
524 * predict.c (expr_expected_value_1): When taking
525 later predictor, assign also probability.
526 Use fold_build2_initializer_loc in order to fold
527 the expression in -frounding-math.
529 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
531 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
533 (expand_vec_perm_1): Try the new method.
535 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
538 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
539 Check whether lra_live_max_point is 0 before dividing.
541 2018-08-14 Martin Sebor <msebor@redhat.com>
543 PR tree-optimization/86650
544 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
545 (vrp_prop::check_mem_ref): Same.
547 2018-08-13 Liu Hao <lh_mouse@126.com>
549 * pretty-print.c (eat_esc_sequence): Swap the foreground and
550 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
551 and clear it thereafter, as it only works for DBCS.
553 2018-08-13 Liu Hao <lh_mouse@126.com>
555 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
556 handle returned by _get_osf_handle().
558 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
560 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
561 for folding vec_perm.
563 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
565 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
566 Add support for gimple-folding of vec_pack() and vec_unpack()
569 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
571 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
572 vec_xst variants to the list.
573 (rs6000_gimple_fold_builtin): Add support for folding unaligned
574 vector loads and stores.
576 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
578 * config.gcc (rs6000-ibm-aix4.x): Delete.
579 (rs6000-ibm-aix5.1): Delete.
580 (rs6000-ibm-aix5.2): Delete.
581 (rs6000-ibm-aix5.3): Delete.
582 * config/rs6000/aix43.h: Delete.
583 * config/rs6000/aix51.h: Delete.
584 * config/rs6000/aix52.h: Delete.
585 * config/rs6000/t-aix43: Delete.
587 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
589 * config/s390/s390.c (s390_decompose_constant_pool_ref):
591 (s390_decompose_address): Factor out constant pool ref
594 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
596 * config/nds32/nds32-predicates.c
597 (nds32_can_use_bclr_p): Change return type as bool.
598 (nds32_can_use_bset_p): Ditto.
599 (nds32_can_use_btgl_p): Ditto.
600 (nds32_can_use_bitci_p): Ditto.
601 * config/nds32/nds32-protos.h
602 (nds32_can_use_bclr_p): Change declaration.
603 (nds32_can_use_bset_p): Ditto.
604 (nds32_can_use_btgl_p): Ditto.
605 (nds32_can_use_bitci_p): Ditto.
607 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
610 Support -msched-prolog-epilog option.
611 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
613 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
615 * common/config/nds32/nds32-common.c
616 (nds32_option_optimization_table): Enalbe -malways-align.
618 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
620 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
622 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
623 OPT_misr_secure_ case.
624 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
625 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
626 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
628 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
629 (nds32_isr_info): New field security_level.
630 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
631 * config/nds32/nds32.md (return_internal): Consider critical attribute.
632 * config/nds32/nds32.opt (misr-secure): New option.
633 * config/nds32/nds32_init.inc: New file.
634 * config/nds32/nds32_isr.h: New file.
636 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
638 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
639 Update comment for atomic instructions.
640 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
641 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
643 (atomic_loaddi): Revise fence expansion to only emit fence prior to
644 load for __ATOMIC_SEQ_CST model.
645 (atomic_loaddi_1): Remove float register target.
646 (atomic_storedi): Handle CONST_INT values.
647 (atomic_storedi_1): Remove float register source. Add special case
649 (memory_barrier): New expander and insn.
651 2018-08-11 Jakub Jelinek <jakub@redhat.com>
653 PR tree-optimization/86835
654 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
655 new_stmt after def_gsi, make sure to insert new_square_stmt after
656 that stmt, not 2 stmts before it.
658 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
661 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
664 2018-08-10 Martin Liska <mliska@suse.cz>
667 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
669 * builtins.c (expand_builtin_expect_with_probability):
671 (expand_builtin_expect_with_probability): New function.
672 (build_builtin_expect_predicate): Add new argumnet probability
673 for BUILT_IN_EXPECT_WITH_PROBABILITY.
674 (fold_builtin_expect):
677 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
678 * builtins.h (fold_builtin_expect): Set new argument.
679 * doc/extend.texi: Document __builtin_expect_with_probability.
680 * doc/invoke.texi: Likewise.
681 * gimple-fold.c (gimple_fold_call): Pass new argument.
682 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
683 also BUILT_IN_EXPECT_WITH_PROBABILITY.
684 * predict.c (get_predictor_value): New function.
685 (expr_expected_value): Add new argument probability. Assume
686 that predictor and probability are always non-null.
687 (expr_expected_value_1): Likewise. For __builtin_expect and
688 __builtin_expect_with_probability set probability. Handle
689 combination in binary expressions.
690 (tree_predict_by_opcode): Simplify code by simply calling
692 (pass_strip_predict_hints::execute): Add handling of
693 BUILT_IN_EXPECT_WITH_PROBABILITY.
694 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
696 * tree.h (DECL_BUILT_IN_P): New function.
698 2018-08-10 Martin Liska <mliska@suse.cz>
700 PR tree-optimization/85799
701 * passes.def: Add argument for pass_strip_predict_hints.
702 * predict.c (class pass_strip_predict_hints): Add new argument
704 (strip_predictor_early): New function.
705 (pass_strip_predict_hints::execute): Call the function to
707 (strip_predict_hints): New function.
708 * predict.def: Fix comment.
710 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
712 * Makefile.in: Clarify which tm.texi to copy over to assert the
713 right to grant a GFDL license for all.
715 2018-08-09 Jeff Law <law@redhat.com>
717 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
720 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
722 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
725 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
727 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
728 reductions for variable-length vectors.
730 2018-08-09 David Malcolm <dmalcolm@redhat.com>
733 * common.opt (fdiagnostics-show-line-numbers): New option.
734 * diagnostic-show-locus.c (class layout): Add fields
735 "m_show_line_numbers_p" and "m_linenum_width";
736 (num_digits): New function.
737 (test_num_digits): New function.
738 (layout::layout): Initialize new fields. Update m_x_offset
739 logic to handle any left margin.
740 (layout::print_source_line): Print line number when requested.
741 (layout::start_annotation_line): New member function.
742 (layout::print_annotation_line): Call it.
743 (layout::print_leading_fixits): Likewise.
744 (layout::print_trailing_fixits): Likewise. Update calls to
745 move_to_column for new parameter.
746 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
747 it to potentially call start_annotation_line.
748 (layout::show_ruler): Call start_annotation_line.
749 (selftest::test_line_numbers_multiline_range): New selftest.
750 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
751 and selftest::test_line_numbers_multiline_range.
752 * diagnostic.c (diagnostic_initialize): Initialize
754 * diagnostic.h (struct diagnostic_context): Add field
755 "show_line_numbers_p".
756 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
757 -fno-diagnostics-show-line-numbers.
758 * dwarf2out.c (gen_producer_string): Add
759 OPT_fdiagnostics_show_line_numbers to the ignored options.
760 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
761 one setting" options.
762 (append_compiler_options): Likewise to the dropped options.
763 (append_diag_options): Likewise to the passed-on options.
764 * opts.c (common_handle_option): Handle the new option.
765 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
767 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
769 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
770 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
771 third argument of type "const signed char" to descriptions of
772 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
773 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
774 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
775 __builtin_bcdsub_ov functions.
777 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
779 PR tree-optimization/86858
780 * tree-vect-loop.c (vect_is_simple_reduction): Restore
781 flow_bb_inside_loop_p calls.
783 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
785 PR tree-optimization/86871
786 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
787 instead of gimple_assign_lhs.
789 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
792 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
793 register constraint to operand 0.
794 (add<mode>3_carryinC): Likewise.
795 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
797 2018-08-09 Martin Liska <mliska@suse.cz>
800 * common.opt: Remove extra line.
802 2018-08-09 Martin Liska <mliska@suse.cz>
804 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
805 at the end of a line, make first letter capital and end up
806 a sentence with a dot.
807 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
808 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
809 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
810 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
811 (PARAM_MAX_ISL_OPERATIONS): Likewise.
812 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
813 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
814 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
815 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
816 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
817 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
818 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
819 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
820 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
821 (PARAM_TREE_REASSOC_WIDTH): Likewise.
822 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
823 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
824 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
826 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
829 * config/s390/s390.c (s390_option_override_internal): Reduce the
830 stack-clash-protection-probe-interval param if it would be too big
833 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
836 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
837 * config/m68k/m68k.c (handle_move_double): Don't call
838 m68k_final_prescan_insn.
839 (m68k_adjust_decorated_operand): Renamed from
840 m68k_final_prescan_insn, remove first and third operand and
842 (print_operand): Call it.
843 (print_operand_address): Call it.
845 2018-08-08 Nathan Sidwell <nathan@acm.org>
847 * diagnostic.c (diagnostic_report_current_module): Use
848 linemap_included_from & linemap_included_from_linemap.
850 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
852 * config/aarch64/aarch64-cores.def: Add phecda core.
853 * config/aarch64/aarch64-tune.md: Regenerate.
854 * doc/invoke.texi: Add phecda core.
856 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
859 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
861 * config/s390/s390.md ("movti"): Add more alternatives for
862 constant to GPR copies.
864 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
866 * config/s390/s390.c: Fix whitespace damage throughout the file.
867 * config/s390/s390.h: Likewise.
868 * config/s390/tpf.h: Likewise.
870 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
872 * config/s390/s390.c (s390_loadrelative_operand_p):
873 Remove TARGET_CPU_ZARCH usages.
874 (s390_rtx_costs): Likewise.
875 (s390_legitimate_constant_p): Likewise.
876 (s390_cannot_force_const_mem): Likewise.
877 (legitimate_reload_constant_p): Likewise.
878 (s390_preferred_reload_class): Likewise.
879 (legitimize_pic_address): Likewise.
880 (legitimize_tls_address): Likewise.
881 (s390_split_branches): Removed.
882 (s390_add_execute): Removed.
883 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
884 (s390_mainpool_start): Likewise.
885 (s390_mainpool_finish): Likewise.
886 (s390_mainpool_cancel): Removed.
887 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
888 (s390_chunkify_cancel): Likewise.
889 (s390_return_addr_rtx): Likewise.
890 (s390_register_info): Remove split_branches_pending_p uages.
891 (s390_optimize_register_info): Likewise.
892 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
893 split_branches_pending_p usages.
894 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
895 (s390_load_got): Likewise.
896 (s390_expand_split_stack_prologue): Likewise.
897 (output_asm_nops): Likewise.
898 (s390_function_profiler): Likewise.
899 (s390_emit_call): Likewise.
900 (s390_conditional_register_usage): Likewise.
901 (s390_optimize_prologue): Likewise.
902 (s390_reorg): Remove TARGET_CPU_ZARCH and
903 split_branches_pending_p usages.
904 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
906 (s390_output_indirect_thunk_function): Likewise.
907 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
908 (TARGET_CPU_ZARCH_P): Removed.
909 (struct machine_function): Remove split_branches_pending_p.
910 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
912 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
914 * common/config/s390/s390-common.c (processor_flags_table):
916 * config.gcc: Remove with_arch/with_tune support.
917 * config/s390/2064.md: Remove cpu attribute comparisons.
918 * config/s390/driver-native.c (s390_host_detect_local_cpu):
920 * config/s390/linux.h (ASM_SPEC):
921 Remove -march support.
922 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
923 Use a table to get an arch level.
924 * config/s390/s390-opts.h (enum processor_type):
927 (processor_table): Remove entries, add arch_level values.
928 (s390_issue_rate): Remove cases.
929 (s390_option_override): Adjust
930 s390_option_override_internal() call.
931 (s390_option_override_internal): Remove deprecation warning.
932 (s390_valid_target_attribute_tree): Adjust
933 s390_option_override_internal() call.
934 * config/s390/s390.h (struct s390_processor):
935 Share with s390-c.c, add arch_level field.
936 * config/s390/s390.md:
937 Remove occurrences in cpu attribute.
938 * config/s390/s390.opt: Remove -march/-mtune support.
939 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
940 * doc/invoke.texi: Remove deprecation warning.
942 2018-08-08 Luis Machado <luis.machado@linaro.org>
944 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
946 (qdf24xx_tunings): Set vector cost structure to
949 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
950 <register_sextend>: Set to 3.
952 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
955 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
956 * config/aarch64/aarch64-simd.md
957 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
958 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
959 (aarch64_frecpx<mode>): ...this new pattern.
960 * config/aarch64/aarch64-simd-builtins.def: Remove comment
961 about aarch64_frecp<FRECP:frecp_suffix><mode>.
963 2018-08-07 Martin Liska <mliska@suse.cz>
966 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
967 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
968 * predict.def (PRED_MALLOC_NONNULL): New predictor.
969 * doc/extend.texi: Document that malloc attribute adds
972 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
975 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
976 Define to speculation_safe_value_not_needed.
978 2018-08-06 Jeff Law <law@redhat.com>
980 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
981 the vr_values instance to cprop_into_stmt.
982 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
983 (cprop_operand): Also query EVRP to determine if OP is a constant.
985 2018-08-06 Nathan Sidwell <nathan@acm.org>
987 * diagnostic.c (diagnostic_report_current_module): Reroll
988 included-at loop. Translate text.
990 2018-08-06 David Malcolm <dmalcolm@redhat.com>
992 * function-tests.c (selftest::test_expansion_to_rtl): Call
993 free_after_compilation.
995 2018-08-06 Alan Hayward <alan.hayward@arm.com>
997 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
999 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
1001 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
1002 loops with memory block operations from getting unrolled.
1004 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
1007 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1008 Define to speculation_safe_value_not_needed.
1010 2018-08-06 Jeff Law <law@redhat.com>
1012 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
1015 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1018 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
1019 with all enabled __intN types.
1021 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
1023 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1025 * alias.c (record_set): Check for clobber high.
1026 * cfgexpand.c (expand_gimple_stmt): Likewise.
1027 * combine-stack-adj.c (single_set_for_csa): Likewise.
1028 * combine.c (find_single_use_1): Likewise.
1029 (set_nonzero_bits_and_sign_copies): Likewise.
1030 (get_combine_src_dest): Likewise.
1031 (is_parallel_of_n_reg_sets): Likewise.
1032 (try_combine): Likewise.
1033 (record_dead_and_set_regs_1): Likewise.
1034 (reg_dead_at_p_1): Likewise.
1035 (reg_dead_at_p): Likewise.
1036 * dce.c (deletable_insn_p): Likewise.
1037 (mark_nonreg_stores_1): Likewise.
1038 (mark_nonreg_stores_2): Likewise.
1039 * df-scan.c (df_find_hard_reg_defs): Likewise.
1040 (df_uses_record): Likewise.
1041 (df_get_call_refs): Likewise.
1042 * dwarf2out.c (mem_loc_descriptor): Likewise.
1043 * haifa-sched.c (haifa_classify_rtx): Likewise.
1044 * ira-build.c (create_insn_allocnos): Likewise.
1045 * ira-costs.c (scan_one_insn): Likewise.
1046 * ira.c (equiv_init_movable_p): Likewise.
1047 (rtx_moveable_p): Likewise.
1048 (interesting_dest_for_shprep): Likewise.
1049 * jump.c (mark_jump_label_1): Likewise.
1050 * postreload-gcse.c (record_opr_changes): Likewise.
1051 * postreload.c (reload_cse_simplify): Likewise.
1052 (struct reg_use): Add source expr.
1053 (reload_combine): Check for clobber high.
1054 (reload_combine_note_use): Likewise.
1055 (reload_cse_move2add): Likewise.
1056 (move2add_note_store): Likewise.
1057 * print-rtl.c (print_pattern): Likewise.
1058 * recog.c (decode_asm_operands): Likewise.
1059 (store_data_bypass_p): Likewise.
1060 (if_test_bypass_p): Likewise.
1061 * regcprop.c (kill_clobbered_value): Likewise.
1062 (kill_set_value): Likewise.
1063 * reginfo.c (reg_scan_mark_refs): Likewise.
1064 * reload1.c (maybe_fix_stack_asms): Likewise.
1065 (eliminate_regs_1): Likewise.
1066 (elimination_effects): Likewise.
1067 (mark_not_eliminable): Likewise.
1068 (scan_paradoxical_subregs): Likewise.
1069 (forget_old_reloads_1): Likewise.
1070 * reorg.c (find_end_label): Likewise.
1071 (try_merge_delay_insns): Likewise.
1072 (redundant_insn): Likewise.
1073 (own_thread_p): Likewise.
1074 (fill_simple_delay_slots): Likewise.
1075 (fill_slots_from_thread): Likewise.
1076 (dbr_schedule): Likewise.
1077 * resource.c (update_live_status): Likewise.
1078 (mark_referenced_resources): Likewise.
1079 (mark_set_resources): Likewise.
1080 * rtl.c (copy_rtx): Likewise.
1081 * rtlanal.c (reg_referenced_p): Likewise.
1082 (single_set_2): Likewise.
1083 (noop_move_p): Likewise.
1084 (note_stores): Likewise.
1085 * sched-deps.c (sched_analyze_reg): Likewise.
1086 (sched_analyze_insn): Likewise.
1088 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1090 * cse.c (invalidate_reg): New function extracted from...
1091 (invalidate): ...here.
1092 (canonicalize_insn): Check for clobber high.
1093 (invalidate_from_clobbers): invalidate clobber highs.
1094 (invalidate_from_sets_and_clobbers): Likewise.
1095 (count_reg_usage): Check for clobber high.
1096 (insn_live_p): Likewise.
1097 * cselib.c (cselib_expand_value_rtx_1):Likewise.
1098 (cselib_invalidate_regno): Check for clobber in setter.
1099 (cselib_invalidate_rtx): Pass through setter.
1100 (cselib_invalidate_rtx_note_stores):
1101 (cselib_process_insn): Check for clobber high.
1102 * cselib.h (cselib_invalidate_rtx): Add operand.
1104 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1106 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
1107 (mark_not_eliminable): Likewise.
1108 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
1109 * lra-lives.c (process_bb_lives): Check for clobber high.
1110 * lra.c (new_insn_reg): Remember clobber highs.
1111 (collect_non_operand_hard_regs): Check for clobber high.
1112 (lra_set_insn_recog_data): Likewise.
1113 (add_regs_to_insn_regno_info): Likewise.
1114 (lra_update_insn_regno_info): Likewise.
1116 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1118 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
1119 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
1121 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1123 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
1124 (copy_insn_1): Likewise.
1125 (gen_hard_reg_clobber_high): New gen function.
1126 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
1127 * genemit.c (gen_exp): Likewise.
1128 (gen_emit_seq): Pass through info.
1129 (gen_insn): Check for CLOBBER_HIGH.
1130 (gen_expand): Pass through info.
1131 (gen_split): Likewise.
1132 (output_add_clobbers): Likewise.
1133 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
1134 (remove_clobbers): Likewise.
1135 * rtl.h (gen_hard_reg_clobber_high): New declaration.
1137 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1139 * doc/rtl.texi (clobber_high): Add.
1140 (parallel): Add in clobber high
1141 * rtl.c (rtl_check_failed_code3): Add function.
1142 * rtl.def (CLOBBER_HIGH): Add expression.
1143 * rtl.h (RTL_CHECKC3): Add macro.
1144 (rtl_check_failed_code3): Add declaration.
1145 (XC3EXP): Add macro.
1147 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
1150 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
1151 cfun->machine->max_used_stack_alignment if needed.
1153 2018-08-04 Martin Sebor <msebor@redhat.com>
1155 PR tree-optimization/86571
1156 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
1159 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
1162 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1165 2018-08-03 Jeff Law <law@redhat.com>
1168 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1169 Define to speculation_safe_value_not_needed.
1171 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1173 * doc/gcov.texi (-x): Remove duplicate "to".
1174 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1175 (-Wif-not-aligned): Remove duplicate "is".
1176 (-flto): Remove duplicate "the".
1177 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1178 duplicate "v5.00.b".
1179 (MSP430 Options): Remove duplicate "and" from the description
1180 of "-mgprel-sec=regexp".
1181 (x86 Options): Remove duplicate copies of "vmldLog102" and
1182 vmlsLog104 from description of "-mveclibabi=type".
1184 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
1186 * internal-fn.h (first_commutative_argument): Declare.
1187 * internal-fn.c (first_commutative_argument): New function.
1188 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
1189 restrictions for pattern statements. Use first_commutative_argument
1190 to look for commutative operands in calls to internal functions.
1192 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
1194 * Makefile.in (wide-int-range.o): New.
1195 * tree-vrp.c: Move all the wide_int_* functions to...
1196 * wide-int-range.cc: ...here.
1197 * tree-vrp.h: Move all the wide_int_* prototypes to...
1198 * wide-int-range.h: ...here.
1200 2018-08-03 Tom de Vries <tdevries@suse.de>
1202 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
1204 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
1205 * except.c (output_function_exception_table): Do early exit if
1206 targetm_common.except_unwind_info (&global_options) == UI_NONE.
1208 2018-08-03 Martin Liska <mliska@suse.cz>
1210 * predict.c (dump_prediction): Change to 2 digits
1213 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
1215 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
1217 (falkor_am_1_gtov_gtov): ... a new insn reservation.
1219 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
1221 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
1222 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
1223 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
1224 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
1225 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
1227 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1229 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1230 pp_take_prefix when saving the existing prefix.
1231 * diagnostic.c (diagnostic_append_note): Likewise.
1232 * langhooks.c (lhd_print_error_function): Likewise.
1233 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1234 param's type. Free the existing prefix.
1235 (pp_take_prefix): New function.
1236 (pretty_printer::pretty_printer): Drop the prefix parameter.
1237 Rename the length parameter to match the comment.
1238 (pretty_printer::~pretty_printer): Free the prefix.
1239 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1241 (struct pretty_printer): Drop the "const" from "prefix" field's
1242 type and clarify memory management.
1243 (pp_set_prefix): Drop the "const" from the 2nd param.
1244 (pp_take_prefix): New decl.
1246 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1248 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
1249 for word_mode_ok here instead of passing as argument.
1250 (expand_block_compare): Change select_block_compare_mode() call.
1251 (expand_strncmp_gpr_sequence): New function.
1252 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
1254 2018-08-02 Jeff Law <law@redhat.com>
1257 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1258 Define to speculation_safe_value_not_needed.
1261 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1262 Define to speculation_safe_value_not_needed.
1264 2018-08-02 Tom de Vries <tdevries@suse.de>
1267 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1268 function. Return UI_TARGET unconditionally.
1269 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1270 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1272 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1274 * genemit.c (print_overload_test): Fix typo.
1276 2018-08-02 Richard Biener <rguenther@suse.de>
1278 PR tree-optimization/86816
1279 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1280 which checks for value availability before querying it.
1281 (gvn_uses_equal): Use it.
1282 (same_succ_hash): Likewise.
1283 (gimple_equal_p): Likewise.
1285 2018-08-02 Nick Clifton <nickc@redhat.com>
1288 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1289 Define to speculation_safe_value_not_needed.
1292 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1293 Define to speculation_safe_value_not_needed.
1296 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1297 Define to speculation_safe_value_not_needed.
1300 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1301 Define to speculation_safe_value_not_needed.
1304 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1305 Define to speculation_safe_value_not_needed.
1308 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1309 Define to speculation_safe_value_not_needed.
1312 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1313 Define to speculation_safe_value_not_needed.
1316 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1317 Define to speculation_safe_value_not_needed.
1320 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1321 speculation_safe_value_not_needed.
1324 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1325 to speculation_safe_value_not_needed.
1327 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1329 * doc/md.texi: Expand the documentation of instruction names
1330 to mention port-local uses. Document '@' in pattern names.
1331 * read-md.h (overloaded_instance, overloaded_name): New structs.
1333 (md_reader::handle_overloaded_name): New member function.
1334 (md_reader::get_overloads): Likewise.
1335 (md_reader::m_first_overload): New member variable.
1336 (md_reader::m_next_overload_ptr): Likewise.
1337 (md_reader::m_overloads_htab): Likewise.
1338 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1339 m_next_overload_ptr and m_overloads_htab.
1340 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1341 (get_mode_token, get_code_token, get_int_token): New functions.
1342 (map_attr_string): Add an optional argument that passes back
1343 the associated iterator.
1344 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1345 (md_reader::handle_overloaded_name, add_overload_instance): New
1347 (apply_iterators): Handle '@' names. Report an error if '@'
1348 is used without iterators.
1349 (initialize_iterators): Initialize the new iterator_group fields.
1350 * genopinit.c (handle_overloaded_code_for)
1351 (handle_overloaded_gen): New functions.
1352 (main): Use them to print declarations of maybe_code_for_* and
1353 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1354 * genemit.c (print_overload_arguments, print_overload_test)
1355 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1356 (main): Use it to print definitions of maybe_code_for_* and
1357 maybe_gen_* functions.
1358 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1359 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1360 instead of explicit mode checks.
1361 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1362 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1363 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1364 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1365 (aarch64_expand_compare_and_swap): Likewise
1366 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1367 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1368 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1369 (aarch64_constant_pool_reload_icode): Delete.
1370 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1371 instead of aarch64_constant_pool_reload_icode. Use
1372 code_for_aarch64_reload_mov instead of explicit mode checks.
1373 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1374 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1375 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1376 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1377 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1378 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1379 (aarch64_atomic_load_op_code): Delete.
1380 (aarch64_emit_atomic_load_op): Likewise.
1381 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1382 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1383 instead of aarch64_emit_atomic_load_op.
1384 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1385 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1386 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1387 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1388 character before the pattern name.
1389 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1390 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1391 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1392 (aarch64_frecps<mode>): Likewise.
1393 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1394 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1395 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1396 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1397 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1399 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1401 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1402 Allow HFmode constants if TARGET_FP_F16INST.
1404 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1407 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1408 No longer check last store for clobber of address register.
1410 2018-08-02 Martin Liska <mliska@suse.cz>
1412 PR gcov-profile/86817
1413 * gcov.c (process_all_functions): New function.
1415 (process_file): Move functions processing to
1416 process_all_functions.
1418 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1420 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1421 "const" to the "gimple *" and "rtx_insn *" parameters.
1422 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1424 (dump_location_t::dump_location_t): Likewise.
1426 2018-08-01 Martin Sebor <msebor@redhat.com>
1428 PR tree-optimization/86650
1429 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1430 rather than a "gcall *". Directly pass the data of interest
1431 to percent_K_format, rather than building a temporary CALL_EXPR
1433 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1434 (gimple_fold_builtin_strncat): Adjust.
1435 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1436 gcall* argument with gimple*.
1437 * gimple-ssa-warn-restrict.c (check_call): Same.
1438 (wrestrict_dom_walker::before_dom_children): Same.
1439 (builtin_access::builtin_access): Same.
1440 (check_bounds_or_overlap): Same
1441 (maybe_diag_overlap): Same.
1442 (maybe_diag_offset_bounds): Same.
1443 * tree-diagnostic.c (default_tree_printer): Move usage of
1444 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1446 * tree-pretty-print.c (percent_K_format): Add argument.
1447 * tree-pretty-print.h: Add argument.
1448 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1449 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1450 (maybe_diag_stxncpy_trunc): Same.
1451 (handle_builtin_stxncpy): Same.
1452 (handle_builtin_strcat): Same.
1454 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1456 * match.pd: Optimise pointer range checks.
1458 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1460 PR tree-optimization/86758
1461 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1462 to remove pattern statements.
1464 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1466 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1467 result of dfs_enumerate_from when constructing stmt_vec_infos,
1468 instead of additionally calling get_loop_body.
1470 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1472 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1474 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1475 When creating an iv, assert that the step is not known to be zero.
1476 (vect_setup_realignment): Update call accordingly.
1477 * tree-vect-stmts.c (vectorizable_store): Likewise.
1478 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1480 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1482 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1483 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1484 (vectorizable_reduction): Likewise.
1485 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1486 (vect_detect_hybrid_slp_stmts): Likewise.
1487 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1489 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1491 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1492 (wide_int_set_zero_nonzero_bits): ...this.
1493 (zero_nonzero_bits_from_vr): Rename to...
1494 (vrp_set_zero_nonzero_bits): ...this.
1495 (extract_range_from_multiplicative_op_1): Abstract wide int
1497 (wide_int_range_multiplicative_op): ...here.
1498 (extract_range_from_binary_expr_1): Extract wide int binary
1499 operations into their own functions.
1500 (wide_int_range_lshift): New.
1501 (wide_int_range_can_optimize_bit_op): New.
1502 (wide_int_range_shift_undefined_p): New.
1503 (wide_int_range_bit_xor): New.
1504 (wide_int_range_bit_ior): New.
1505 (wide_int_range_bit_and): New.
1506 (wide_int_range_trunc_mod): New.
1507 (extract_range_into_wide_ints): New.
1508 (vrp_shift_undefined_p): New.
1509 (extract_range_from_multiplicative_op): New.
1510 (vrp_can_optimize_bit_op): New.
1511 * tree-vrp.h (value_range::dump): New.
1512 (wide_int_range_multiplicative_op): New.
1513 (wide_int_range_lshift):New.
1514 (wide_int_range_shift_undefined_p): New.
1515 (wide_int_range_bit_xor): New.
1516 (wide_int_range_bit_ior): New.
1517 (wide_int_range_bit_and): New.
1518 (wide_int_range_trunc_mod): New.
1519 (zero_nonzero_bits_from_bounds): Rename to...
1520 (wide_int_set_zero_nonzero_bits): ...this.
1521 (zero_nonzero_bits_from_vr): Rename to...
1522 (vrp_set_zero_nonzero_bits): ...this.
1523 (range_easy_mask_min_max): Rename to...
1524 (wide_int_range_can_optimize_bit_op): this.
1525 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1526 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1528 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1530 * tree-vectorizer.h (vect_orig_stmt): New function.
1531 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1532 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1533 (vect_create_epilog_for_reduction): Likewise.
1534 (vectorizable_live_operation): Likewise.
1535 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1536 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1537 * tree-vect-stmts.c (vectorizable_call): Likewise.
1538 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1540 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1542 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1544 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1545 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1546 (vect_transform_loop): Likewise.
1547 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1549 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1551 * tree-vectorizer.h (vect_schedule_slp): Return void.
1552 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1553 (vect_schedule_slp): Likewise.
1555 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1557 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1559 (vect_transform_loop): Update calls accordingly. Schedule SLP
1560 instances before the main loop, if any exist.
1562 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1564 PR tree-optimization/86749
1565 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1566 If the lhs is used in a COND_EXPR, check that it is being used
1567 as the "then" or "else" value.
1569 2018-08-01 Tom de Vries <tdevries@suse.de>
1572 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1573 speculation_safe_value_not_needed.
1575 2018-08-01 Richard Biener <rguenther@suse.de>
1577 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1580 2018-08-01 Martin Liska <mliska@suse.cz>
1582 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1583 format how successful transformation is dumped.
1584 (gimple_mod_pow2_value_transform): Likewise.
1585 (gimple_mod_subtract_transform): Likewise.
1586 (gimple_stringops_transform): Likewise.
1588 2018-08-01 Martin Liska <mliska@suse.cz>
1591 * value-prof.c (interesting_stringop_to_profile_p):
1592 Simplify the code and add BUILT_IN_MEMMOVE.
1593 (gimple_stringops_transform): Likewise.
1595 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1597 * config/aarch64/aarch64-simd.md
1598 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1599 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1600 use GPI iterator instead of SI mode.
1602 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1604 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1605 rs6000_speculation_barrier.
1606 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1607 new barrier pattern name.
1609 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1611 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1612 (speculation_barrier): New insn.
1614 2018-08-01 Richard Biener <rguenther@suse.de>
1617 * graphite.h: Include isl/id.h and isl/space.h to allow build
1620 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1623 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1624 mode for COFF targets.
1625 * defaults.h (TARGET_COFF): Define.
1626 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1627 TARGET_COFF): Define.
1628 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1629 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1630 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1632 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1634 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1635 * gimple-streamer-out.c (output_bb): Save it.
1636 * lto-streamer-in.c (input_struct_function_base): Restore
1637 instance discriminator if available. Create map on demand.
1638 * lto-streamer-out.c (output_struct_function_base): Save it if
1640 * final.c (decl_to_instance_map): Document LTO strategy.
1642 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1643 Olivier Hainque <hainque@adacore.com>
1645 * debug.h (decl_to_instance_map_t): New type.
1646 (decl_to_instance_map): Declare.
1647 (maybe_create_decl_to_instance_map): New inline function.
1648 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1649 to track basic block discriminators.
1650 (final_start_function_1): Initialize them.
1651 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1653 (decl_to_instance_map): New variable.
1654 (map_decl_to_instance, maybe_set_discriminator): New functions.
1655 (notice_source_line): Set discriminator.
1657 2018-07-31 Ian Lance Taylor <iant@golang.org>
1659 * targhooks.c (default_have_speculation_safe_value): Add
1662 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1664 * dump-context.h: Include "pretty-print.h".
1665 (dump_context::refresh_dumps_are_enabled): New decl.
1666 (dump_context::emit_item): New decl.
1667 (class dump_context): Add fields "m_test_pp" and
1669 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1670 (temp_dump_context::get_dumped_text): New decl.
1671 (class temp_dump_context): Add field "m_pp".
1672 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1673 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1675 (set_dump_file): Update for above change.
1676 (set_alt_dump_file): Likewise.
1677 (dump_loc): New overload, taking a pretty_printer *.
1678 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1679 to any test pretty-printer.
1680 (make_item_for_dump_gimple_stmt): New function, adapted from
1681 optinfo::add_gimple_stmt.
1682 (dump_context::dump_gimple_stmt): Call it, and use the result,
1683 eliminating the direct usage of dump_file and alt_dump_file in
1684 favor of indirectly using them via emit_item.
1685 (make_item_for_dump_gimple_expr): New function, adapted from
1686 optinfo::add_gimple_expr.
1687 (dump_context::dump_gimple_expr): Call it, and use the result,
1688 eliminating the direct usage of dump_file and alt_dump_file in
1689 favor of indirectly using them via emit_item.
1690 (make_item_for_dump_generic_expr): New function, adapted from
1692 (dump_context::dump_generic_expr): Call it, and use the result,
1693 eliminating the direct usage of dump_file and alt_dump_file in
1694 favor of indirectly using them via emit_item.
1695 (make_item_for_dump_printf_va): New function, adapted from
1696 optinfo::add_printf_va.
1697 (make_item_for_dump_printf): New function.
1698 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1699 and use the result, eliminating the direct usage of dump_file and
1700 alt_dump_file in favor of indirectly using them via emit_item.
1701 (make_item_for_dump_dec): New function.
1702 (dump_context::dump_dec): Call it, and use the result,
1703 eliminating the direct usage of dump_file and alt_dump_file in
1704 favor of indirectly using them via emit_item.
1705 (make_item_for_dump_symtab_node): New function, adapted from
1706 optinfo::add_symtab_node.
1707 (dump_context::dump_symtab_node): Call it, and use the result,
1708 eliminating the direct usage of dump_file and alt_dump_file in
1709 favor of indirectly using them via emit_item.
1710 (dump_context::begin_scope): Reimplement, avoiding direct usage
1711 of dump_file and alt_dump_file in favor of indirectly using them
1713 (dump_context::emit_item): New member function.
1714 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1715 Set up test pretty-printer on the underlying context. Call
1716 refresh_dumps_are_enabled.
1717 (temp_dump_context::~temp_dump_context): Call
1718 refresh_dumps_are_enabled.
1719 (temp_dump_context::get_dumped_text): New member function.
1720 (selftest::verify_dumped_text): New function.
1721 (ASSERT_DUMPED_TEXT_EQ): New macro.
1722 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1723 and then without optinfo enabled. Add uses of
1724 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1725 * dumpfile.h: Update comment for the dump_* API.
1726 * optinfo-emit-json.cc
1727 (selftest::test_building_json_from_dump_calls): Update for new
1728 param for temp_dump_context ctor.
1729 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1730 and "m_owned" field.
1731 (optinfo_item::~optinfo_item): Likewise.
1732 (optinfo::add_item): New member function.
1733 (optinfo::emit): Update comment.
1734 (optinfo::add_string): Delete.
1735 (optinfo::add_printf): Delete.
1736 (optinfo::add_printf_va): Delete.
1737 (optinfo::add_gimple_stmt): Delete.
1738 (optinfo::add_gimple_expr): Delete.
1739 (optinfo::add_tree): Delete.
1740 (optinfo::add_symtab_node): Delete.
1741 (optinfo::add_dec): Delete.
1742 * optinfo.h (class dump_context): New forward decl.
1743 (optinfo::add_item): New decl.
1744 (optinfo::add_string): Delete.
1745 (optinfo::add_printf): Delete.
1746 (optinfo::add_printf_va): Delete.
1747 (optinfo::add_gimple_stmt): Delete.
1748 (optinfo::add_gimple_expr): Delete.
1749 (optinfo::add_tree): Delete.
1750 (optinfo::add_symtab_node): Delete.
1751 (optinfo::add_dec): Delete.
1752 (optinfo::add_poly_int): Delete.
1753 (optinfo_item::optinfo_item): Remove "owned" param.
1754 (class optinfo_item): Remove field "m_owned".
1756 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1759 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1760 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1761 requested variable alignment.
1762 (expand_one_ssa_partition): Likewise.
1763 (expand_one_var): Likewise.
1765 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1767 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1768 to speculation_safe_value_not_needed.
1770 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1772 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1773 * targhooks.c (speculation_safe_value_not_needed): New function.
1774 * target.def (have_speculation_safe_value): Update documentation.
1775 * doc/tm.texi: Regenerated.
1777 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1779 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1780 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1782 (despeculate_copy<ALLI:mode>_insn): New insn.
1783 (despeculate_copyti_insn): New insn.
1784 (despeculate_simple<ALLI:mode>): New insn
1785 (despeculate_simpleti): New insn.
1786 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1788 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1789 aarch64_speculation_safe_value.
1790 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1792 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1794 * config/aarch64/aarch64-speculation.cc: New file.
1795 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1796 before pass_reorder_blocks.
1797 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1799 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1800 X14 and X15 when tracking speculation.
1801 * config/aarch64/aarch64.md (register name constants): Add
1802 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1803 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1804 (speculation_barrier): New insn attribute.
1805 (cmp<mode>): Allow SP in comparisons.
1806 (speculation_tracker): New insn.
1807 (speculation_barrier): Add speculation_barrier attribute.
1808 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1809 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1810 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1812 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1814 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1815 aarch64_track_speculation is true.
1816 (tb<optab><mode>1): Likewise.
1817 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1818 generate CB[N]Z when tracking speculation.
1819 (aarch64_split_compare_and_swap): Likewise.
1820 (aarch64_split_atomic_op): Likewise.
1822 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1824 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1826 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1828 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1829 (speculation_barrier): New insn.
1831 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1833 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1834 * config/arm/arm.md (speculation_barrier): New expand.
1835 (speculation_barrier_insn): New pattern.
1837 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1839 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1840 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1841 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1842 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1844 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1845 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1846 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1847 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1848 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1849 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1850 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1851 * builtins.c (expand_speculation_safe_value): New function.
1852 (expand_builtin): Call it.
1853 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1854 * doc/extend.texi: Document __builtin_speculation_safe_value.
1855 * doc/md.texi: Document "speculation_barrier" pattern.
1856 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1857 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1858 * doc/tm.texi: Regenerated.
1859 * target.def (have_speculation_safe_value, speculation_safe_value): New
1861 * targhooks.c (default_have_speculation_safe_value): New function.
1862 (default_speculation_safe_value): New function.
1863 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1864 (default_speculation_safe_value): Add prototype.
1866 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1868 * dump-context.h (dump_context::dump_loc): New decl.
1869 * dumpfile.c (dump_context::dump_loc): New member function.
1870 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1871 and dump_gimple_stmt.
1872 (dump_context::dump_gimple_expr_loc): Likewise, using
1874 (dump_context::dump_generic_expr_loc): Likewise, using
1876 (dump_context::dump_printf_loc_va): Likewise, using
1878 (dump_context::begin_scope): Explicitly using the global function
1879 "dump_loc", rather than the member function.
1881 2018-07-31 Martin Sebor <msebor@redhat.com>
1883 PR tree-optimization/86741
1884 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1886 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1888 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1889 depend on whether prefetch instructions will be emitted or not.
1890 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1891 will be emitted or not.
1892 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1893 (TARGET_SETMEM_PFD): New macros.
1895 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1897 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1898 (NULL_STMT_VEC_INFO): Delete.
1899 (stmt_vec_info::operator*): Likewise.
1900 (stmt_vec_info::operator gimple *): Likewise.
1901 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1902 of NULL_STMT_VEC_INFO.
1903 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1904 (vect_reassociating_reduction_p): Likewise.
1905 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1906 (vectorizable_store): Likewise.
1907 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1908 (vec_info::free_stmt_vec_infos): Likewise.
1910 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1912 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1913 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1915 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1917 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1918 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1919 (vec_info::free_stmt_vec_info): New private member functions.
1920 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1921 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1922 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1923 set_stmt_vec_info_vec and free_stmt_vec_infos.
1924 * tree-vect-loop.c (_loop_vec_info): Remove call to
1925 set_stmt_vec_info_vec.
1926 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1927 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1928 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1929 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1930 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1931 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1932 with the clearing in the xcalloc.
1933 (stmt_vec_info_vec): Delete.
1934 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1935 (vectorize_loops): Likewise.
1936 (vec_info::~vec_info): Remove argument from call to
1937 free_stmt_vec_infos.
1938 (vec_info::add_stmt): Remove vinfo argument from call to
1941 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1943 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1944 rather than a gimple stmt.
1945 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1946 information for pattern statements when passed the original
1947 statement; instead wait to be passed the pattern statement itself.
1948 Don't call set_vinfo_for_stmt here.
1949 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1950 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1951 stmt_vec_infos here.
1952 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1953 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1954 stmt_vec_infos entry.
1956 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1958 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1959 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1960 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1961 * tree-vect-stmts.c (vectorizable_call): Likewise.
1962 (vectorizable_simd_clone_call): Likewise.
1964 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1966 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1967 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1968 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1969 * tree-vect-loop.c (vect_transform_loop): Likewise.
1970 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1971 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1973 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1975 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1976 (vect_dr_stmt): Delete.
1977 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1978 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1980 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1981 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1982 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1983 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1984 (vect_analyze_data_refs): Likewise.
1985 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1987 (vect_find_same_alignment_drs): Likewise.
1988 (vect_slp_analyze_node_dependences): Update calls accordingly.
1989 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1990 instead of DR_VECT_AUX.
1991 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1992 of a vector data references. Use vec_info::lookup_dr instead of
1994 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1995 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1996 instead of DR_VECT_AUX.
1998 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2000 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
2002 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2004 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2005 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2006 (vect_gen_prolog_loop_niters): Likewise.
2008 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2010 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
2011 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
2012 (vect_known_alignment_in_bytes, vect_dr_behavior)
2013 (vect_get_scalar_dr_size): Take references as dr_vec_infos
2014 instead of data_references. Update calls to other routines for
2015 which the same change has been made.
2016 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
2017 dr_vec_infos instead of stmt_vec_infos.
2018 (vect_analyze_data_ref_dependence): Update call accordingly.
2019 (vect_slp_analyze_data_ref_dependence)
2020 (vect_record_base_alignments): Use DR_VECT_AUX.
2021 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
2022 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
2023 (vector_alignment_reachable_p, vect_get_data_access_cost)
2024 (vect_peeling_supportable, vect_analyze_group_access_1)
2025 (vect_analyze_group_access, vect_analyze_data_ref_access)
2026 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
2027 (vect_compile_time_alias, vect_small_gap_p)
2028 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
2029 (vect_supportable_dr_alignment): Take references as dr_vec_infos
2030 instead of data_references. Update calls to other routines for
2031 which the same change has been made.
2032 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
2033 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2034 (vect_slp_analyze_and_verify_node_alignment)
2035 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2036 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2037 (vect_setup_realignment): Use dr_vec_infos. Update calls after
2039 (_vect_peel_info::dr): Replace with...
2040 (_vect_peel_info::dr_info): ...this new field.
2041 (vect_peeling_hash_get_most_frequent)
2042 (vect_peeling_hash_choose_best_peeling): Update accordingly.
2043 (vect_peeling_hash_get_lowest_cost):
2044 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
2045 routines for which the same change has been made.
2046 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
2048 * tree-vect-loop-manip.c (get_misalign_in_elems)
2049 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
2051 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2052 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
2053 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
2054 (get_group_load_store_type, get_negative_load_store_type)
2055 (vect_get_data_ptr_increment, vectorizable_store)
2056 (vectorizable_load): Likewise.
2057 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
2058 Update calls to other routines for which the same change has been made.
2060 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2062 * tree-vectorizer.h (vec_info::move_dr): New member function.
2063 (dataref_aux): Rename to...
2064 (dr_vec_info): ...this and add "dr" and "stmt" fields.
2065 (_stmt_vec_info::dr_aux): Update accordingly.
2066 (_stmt_vec_info::data_ref_info): Delete.
2067 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
2068 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
2069 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
2070 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
2072 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
2073 (STMT_VINFO_DR_INFO): New macro.
2074 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
2075 (set_dr_misalignment): Update after rename of dataref_aux.
2076 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
2077 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
2078 initialization of STMT_VINFO_DATA_REF.
2079 * tree-vectorizer.c (vec_info::move_dr): New function.
2080 * tree-vect-patterns.c (vect_recog_bool_pattern)
2081 (vect_recog_mask_conversion_pattern)
2082 (vect_recog_gather_scatter_pattern): Use it.
2083 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
2084 the "dr" and "stmt" fields of dr_vec_info instead of
2085 STMT_VINFO_DATA_REF.
2087 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2089 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
2090 (is_pattern_stmt_p): Use it.
2091 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
2092 on pattern statements.
2094 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2096 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
2097 original stmt as a stmt_vec_info rather than a gimple stmt.
2098 (vect_pattern_recog_1): Take the statement directly as a
2099 stmt_vec_info, rather than via a gimple_stmt_iterator.
2100 Update call to vect_mark_pattern_stmts.
2101 (vect_pattern_recog): Update calls accordingly.
2103 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2105 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
2106 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
2107 a vect_def_type for the first argument.
2108 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
2109 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
2110 operand if it isn't defined by a vectorized statement.
2111 (vect_build_gather_load_calls): Remove the mask_dt argument and
2112 update calls to vect_get_vec_def_for_stmt_copy.
2113 (vectorizable_bswap): Likewise the dt argument.
2114 (vectorizable_call): Update calls to vectorizable_bswap and
2115 vect_get_vec_def_for_stmt_copy.
2116 (vectorizable_simd_clone_call, vectorizable_assignment)
2117 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
2118 (vectorizable_comparison): Update calls to
2119 vect_get_vec_def_for_stmt_copy.
2120 (vectorizable_store): Likewise. Remove now-unnecessary calls to
2122 (vect_get_loop_based_defs): Remove dt argument and update call
2123 to vect_get_vec_def_for_stmt_copy.
2124 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
2125 and vect_get_vec_def_for_stmt_copy.
2126 (vectorizable_load): Update calls to vect_build_gather_load_calls
2127 and vect_get_vec_def_for_stmt_copy.
2128 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2129 (vectorizable_reduction, vectorizable_live_operation): Update calls
2130 to vect_get_vec_def_for_stmt_copy.
2132 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2134 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
2135 and gimple stmt arguments with a stmt_vec_info.
2136 (vect_record_base_alignments): Update calls accordingly.
2137 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
2138 and gimple stmt arguments with a stmt_vec_info.
2139 (vect_build_slp_tree_1): Remove vinfo argument and update call
2140 to vect_record_max_nunits.
2141 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
2142 and vect_record_max_nunits.
2144 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2146 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
2147 file and take a stmt_vec_info instead of a gimple stmt.
2148 (supportable_widening_operation, vect_finish_replace_stmt)
2149 (vect_finish_stmt_generation, vect_get_store_rhs)
2150 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
2151 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
2152 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
2153 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
2154 (vect_create_data_ref_ptr, bump_vector_ptr)
2155 (vect_permute_store_chain, vect_setup_realignment)
2156 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
2157 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
2158 (vectorizable_reduction, vectorizable_induction)
2159 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
2160 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
2161 than gimple stmts as arguments.
2162 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
2163 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
2164 (can_group_stmts_p, vect_check_gather_scatter)
2165 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2166 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
2167 (vect_permute_load_chain, vect_shift_permute_load_chain)
2168 (vect_transform_grouped_load)
2169 (vect_record_grouped_load_vectors): Likewise.
2170 * tree-vect-loop.c (vect_fixup_reduc_chain)
2171 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
2172 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
2173 (vectorizable_reduction, vectorizable_induction)
2174 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
2175 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
2176 (vect_get_load_store_mask): Likewise.
2177 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
2178 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
2179 * tree-vect-stmts.c (vect_mark_relevant)
2180 (is_simple_and_all_uses_invariant)
2181 (exist_non_indexing_operands_for_use_p, process_use)
2182 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
2183 (vect_get_vec_def_for_operand, vect_get_vec_defs)
2184 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
2185 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
2186 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
2187 (get_negative_load_store_type, get_load_store_type)
2188 (vect_check_load_store_mask, vect_check_store_rhs)
2189 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
2190 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2191 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2192 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2193 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
2194 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
2195 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
2196 (supportable_widening_operation): Likewise.
2198 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2200 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
2201 a stmt_vec_info instead of a gcall.
2202 (vect_check_gather_scatter): Update call accordingly.
2203 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
2205 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
2206 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
2207 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
2208 instead of a gimple stmt.
2209 (vect_transform_loop): Update calls accordingly.
2210 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
2211 stmt_vec_infos instead of gimple stmts.
2212 (vect_analyze_slp_instance): Update use accordingly.
2213 * tree-vect-stmts.c (read_vector_array, write_vector_array)
2214 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
2215 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
2216 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
2217 (vect_gen_widened_results_half, vect_get_loop_based_defs)
2218 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
2219 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
2220 down to subroutines.
2222 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2224 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
2225 of the worklist from a vector of gimple stmts to a vector of
2227 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2228 (vect_mark_stmts_to_be_vectorized): Likewise
2230 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2232 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
2233 statement before passing it to vect_analyze_stmt.
2234 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
2235 the chain of phi vector definitions. Track the exit phi via its
2237 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
2238 STMT_VINFO_REDUC_DEF.
2239 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
2240 stmt_vec_infos to handle the statement chains.
2241 (vect_get_slp_defs): Record the first statement in the node
2242 using a stmt_vec_info.
2243 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
2244 statements here and pass their stmt_vec_info down to subroutines.
2245 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
2246 down to vect_finish_stmt_generation.
2247 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
2248 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
2249 stmt_vec_infos to subroutines.
2250 (vect_remove_stores): Use stmt_vec_infos to handle the statement
2253 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2255 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
2256 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
2257 (vect_permute_store_chain, vect_setup_realignment)
2258 (vect_permute_load_chain, vect_shift_permute_load_chain)
2259 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
2260 stmts internally, and when passing values to other vectorizer routines.
2261 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2262 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2263 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2264 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2265 (vectorizable_reduction, vectorizable_induction)
2266 (vectorizable_live_operation, vect_transform_loop_stmt)
2267 (vect_transform_loop): Likewise.
2268 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2269 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2270 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2271 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2272 (vect_slp_analyze_node_operations_1): Likewise.
2273 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2274 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2275 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2276 (vect_finish_stmt_generation_1, get_group_load_store_type)
2277 (get_load_store_type, vect_build_gather_load_calls)
2278 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2279 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2280 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2281 (vectorizable_store, vectorizable_load, vectorizable_condition)
2282 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2283 (supportable_widening_operation): Likewise.
2284 (vect_get_vector_types_for_stmt): Likewise.
2285 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2287 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2289 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2290 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2291 (vect_permute_store_chain, vect_permute_load_chain)
2292 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2293 repeated stmt_vec_info lookups.
2294 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2295 (vect_update_ivs_after_vectorizer): Likewise.
2296 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2297 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2298 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2299 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2300 (vect_bb_slp_scalar_cost): Likewise.
2301 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2303 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2305 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2306 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2307 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2308 to get gassigns and gcalls, rather than operating on generc gimple
2310 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2311 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2312 (vectorizable_load, vect_analyze_stmt): Likewise.
2313 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2315 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2317 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2318 return stmt_vec_infos rather than gimple stmts. Do not accept
2320 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2322 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2323 Update use of get_later_stmt.
2324 (vect_get_constant_vectors): Update call accordingly.
2325 (vect_schedule_slp_instance): Likewise
2326 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2327 (vect_slp_analyze_instance_dependence): Likewise.
2328 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2330 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2332 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2333 (stmt_info_for_cost::stmt_info): ...this new field.
2334 (add_stmt_costs): Update accordingly.
2335 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2336 (vect_get_known_peeling_cost): Likewise.
2337 (vect_estimate_min_profitable_iters): Likewise.
2338 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2340 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2342 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2343 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2344 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2346 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2348 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2350 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2351 a gimple stmt to a stmt_vec_info.
2352 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2356 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2357 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2358 (_loop_vec_info::reduction_chains): Likewise.
2359 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2361 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2363 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2365 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2366 a gimple stmt to a stmt_vec_info.
2367 (_stmt_vec_info::next_element): Likewise.
2368 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2369 (vect_slp_analyze_and_verify_node_alignment)
2370 (vect_analyze_group_access_1, vect_analyze_group_access)
2371 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2372 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2373 (vect_supportable_dr_alignment): Update accordingly.
2374 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2375 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2376 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2377 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2378 * tree-vect-slp.c (vect_build_slp_tree_1)
2379 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2380 (vect_split_slp_store_group, vect_analyze_slp_instance)
2381 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2382 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2383 (get_group_load_store_type, get_load_store_type)
2384 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2385 (vect_transform_stmt, vect_remove_stores): Likewise.
2387 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2389 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2391 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2392 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2393 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2394 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2395 (vector_alignment_reachable_p, vect_get_data_access_cost)
2396 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2397 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2398 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2399 (vect_analyze_group_access_1, vect_analyze_group_access)
2400 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2401 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2402 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2403 result of vect_dr_stmt and use the stmt_vec_info instead of
2404 the associated gimple stmt.
2405 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2406 (vect_gen_prolog_loop_niters): Likewise.
2407 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2409 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2411 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2412 to a vec<stmt_vec_info>.
2413 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2414 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2416 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2417 to a vec<stmt_vec_info>.
2418 (bst_traits::value_type, bst_traits::value_type): Likewise.
2419 (bst_traits::hash): Update accordingly.
2420 (vect_get_and_check_slp_defs): Change the stmts parameter from
2421 a vec<gimple *> to a vec<stmt_vec_info>.
2422 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2423 (vect_build_slp_tree): Likewise.
2424 (vect_build_slp_tree_2): Likewise. Update uses of
2425 SLP_TREE_SCALAR_STMTS.
2426 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2427 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2428 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2429 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2430 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2431 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2432 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2433 (vect_get_constant_vectors, vect_get_slp_defs)
2434 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2435 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2436 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2437 instead of gimple stmts.
2438 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2439 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2440 (vect_slp_analyze_instance_dependence): Update uses of
2441 SLP_TREE_SCALAR_STMTS.
2442 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2443 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2444 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2445 (get_initial_defs_for_reduction): Likewise.
2446 (vect_create_epilog_for_reduction): Likewise.
2447 (vectorize_fold_left_reduction): Likewise.
2448 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2449 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2450 (can_vectorize_live_stmts): Likewise.
2452 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2454 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2455 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2456 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2458 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2460 (gather_scalar_reductions): Update after above interface changes.
2461 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2462 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2464 (vect_force_simple_reduction): Likewise.
2465 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2466 LOOP_VINFO_REDUCTIONS.
2467 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2469 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2471 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2472 a gimple stmt to a stmt_vec_info.
2473 * tree-vect-loop.c (vect_active_double_reduction_p)
2474 (vect_force_simple_reduction, vectorizable_reduction): Update
2477 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2479 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2480 vec<gimple *> to a vec<stmt_vec_info>.
2481 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2482 the reduction_phis argument from a vec<gimple *> to a
2484 (vectorizable_reduction): Likewise the phis local variable that
2485 is passed to vect_create_epilog_for_reduction. Update for new type
2486 of SLP_TREE_VEC_STMTS.
2487 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2488 (vectorizable_live_operation): Likewise.
2489 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2490 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2492 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2494 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2495 a gimple stmt to a stmt_vec_info.
2496 (vectorizable_condition, vectorizable_live_operation)
2497 (vectorizable_reduction, vectorizable_induction): Pass back the
2498 vectorized statement as a stmt_vec_info.
2499 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2500 use of STMT_VINFO_VEC_STMT.
2501 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2502 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2503 as stmt_vec_infos rather than gimple stmts.
2504 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2506 (vectorizable_live_operation): Likewise.
2507 (vectorizable_reduction, vectorizable_induction): Likewise,
2508 updating use of STMT_VINFO_VEC_STMT.
2509 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2510 of STMT_VINFO_VEC_STMT.
2511 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2512 (vectorizable_simd_clone_call, vectorizable_conversion)
2513 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2514 (vectorizable_store, vectorizable_load, vectorizable_condition)
2515 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2516 from a gimple stmt to a stmt_vec_info.
2517 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2518 pointer to a stmt_vec_info to the vectorizable_* routines.
2520 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2522 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2523 a gimple stmt to a stmt_vec_info.
2524 (is_pattern_stmt_p): Update accordingly.
2525 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2526 (vect_record_grouped_load_vectors): Likewise.
2527 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2528 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2529 (vect_model_reduction_cost): Likewise.
2530 (vect_create_epilog_for_reduction): Likewise.
2531 (vectorizable_reduction, vectorizable_induction): Likewise.
2532 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2533 Return the stmt_vec_info for the pattern statement.
2534 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2535 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2536 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2537 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2538 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2539 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2540 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2541 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2543 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2545 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2546 (vect_finish_stmt_generation): Likewise.
2547 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2548 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2549 (vect_build_gather_load_calls): Use the return value of the above
2550 functions instead of a separate call to vinfo_for_stmt. Use narrow
2551 scopes for the input gimple stmt and wider scopes for the associated
2552 stmt_vec_info. Use vec_info::lookup_def when setting these
2553 stmt_vec_infos from an SSA_NAME definition.
2554 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2555 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2556 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2557 (vectorizable_store, vectorizable_load, vectorizable_condition)
2558 (vectorizable_comparison): Likewise.
2559 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2560 (vectorizable_reduction): Likewise.
2562 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2564 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2565 stmt_vec_info * parameter before the optional gimple **.
2566 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2567 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2568 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2569 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2570 (vectorizable_live_operation): Likewise.
2571 * tree-vect-patterns.c (type_conversion_p): Likewise.
2572 (vect_look_through_possible_promotion): Likewise.
2573 (vect_recog_rotate_pattern): Likewise.
2574 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2578 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2579 a typedef to a wrapper class.
2580 (NULL_STMT_VEC_INFO): New macro.
2581 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2582 (stmt_vec_info::operator*): New function.
2583 (stmt_vec_info::operator gimple *): Likewise.
2584 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2585 (add_stmt_costs): Likewise.
2586 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2587 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2588 (vect_get_known_peeling_cost): Likewise.
2589 (vect_estimate_min_profitable_iters): Likewise.
2590 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2591 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2592 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2593 (vectorizable_store, free_stmt_vec_infos): Likewise.
2594 (new_stmt_vec_info): Change return type of xcalloc to
2597 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2599 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2600 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2601 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2602 a single_imm_use-based sequence.
2603 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2605 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2607 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2608 * tree-vectorizer.c (vec_info::lookup_def): New function.
2609 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2610 (vect_widened_op_tree): Likewise.
2611 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2612 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2613 (vectorizable_reduction): Likewise.
2614 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2616 (vect_is_slp_reduction): Update calls accordingly. Use
2617 vec_info::lookup_def.
2618 (vect_is_simple_reduction): Likewise
2619 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2621 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2623 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2624 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2625 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2627 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2628 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2629 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2630 (vect_is_slp_reduction, vectorizable_induction)
2631 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2632 * tree-vect-patterns.c (vect_init_pattern_stmt):
2633 (vect_determine_min_output_precision_1, vect_determine_precisions)
2634 (vect_pattern_recog): Likewise.
2635 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2636 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2637 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2638 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2639 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2640 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2641 info field from a loop to a loop_vec_info.
2643 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2645 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2646 (vec_info::add_stmt): Declare.
2647 * tree-vectorizer.c (vec_info::add_stmt): New function.
2648 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2649 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2650 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2651 (vectorizable_induction): Likewise.
2652 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2653 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2654 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2655 (vectorizable_load): Likewise.
2656 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2657 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2658 (vect_recog_gather_scatter_pattern): Likewise.
2659 (append_pattern_def_seq): Likewise. Remove a check that is
2660 performed by add_stmt itself.
2662 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2664 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2665 which make_ssa_name was called with new_stmt before new_stmt
2668 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2670 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2672 (vect_is_slp_reduction): ...here...
2673 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2674 that are already known to be false.
2676 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2678 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2679 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2680 STMT_VINFO_NUM_SLP_USES when it's true.
2681 (vect_free_slp_instance): Add a final_p parameter and pass it to
2683 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2684 (vect_analyze_slp_instance): Likewise.
2685 (vect_slp_analyze_operations): Likewise.
2686 (vect_slp_analyze_bb_1): Likewise.
2687 * tree-vectorizer.c (vec_info): Likewise.
2688 * tree-vect-loop.c (vect_transform_loop): Likewise.
2690 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2692 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2693 function is not called for second and subsequent members of
2696 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2698 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2699 cases for nested loops from here to ...
2700 (vect_create_epilog_for_reduction): ...here. Only call
2701 vect_is_simple_use for inner-loop reductions.
2703 2018-07-31 Martin Liska <mliska@suse.cz>
2705 PR gcov-profile/85338
2706 PR gcov-profile/85350
2707 PR gcov-profile/85372
2708 * profile.c (struct location_triplet): New.
2709 (struct location_triplet_hash): Likewise.
2710 (output_location): Do not output a BB that
2711 is already recorded for a line.
2712 (branch_prob): Use streamed_locations.
2714 2018-07-31 Martin Liska <mliska@suse.cz>
2716 PR gcov-profile/85370
2717 * coverage.c (coverage_begin_function): Do not mark target
2718 clones as artificial functions.
2720 2018-07-31 Martin Liska <mliska@suse.cz>
2722 PR gcov-profile/83813
2723 PR gcov-profile/84758
2724 PR gcov-profile/85217
2725 PR gcov-profile/85332
2726 * profile.c (branch_prob): Do not record GOTO expressions
2727 for GIMPLE statements which locations are already streamed.
2729 2018-07-31 Olivier Hainque <hainque@adacore.com>
2731 * gcc.c (handle_spec_function): Accept a soft_matched_part
2732 argument, as do_spec_1. Pass it down to ...
2733 (eval_spec_function): Accept a soft_matched_part argument,
2734 and pass it down to ...
2735 (do_spec_2): Accept a soft_matched_part argument, and pass
2736 it down to do_spec_1.
2737 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2738 (handle_braces): Update call to handle_spec_function.
2739 (driver::set_up_specs): Update calls to do_spec_2.
2740 (compare_debug_dump_opt_spec_function): Likewise.
2741 (compare_debug_self_opt_spec_function): Likewise.
2743 2018-07-31 Olivier Hainque <hainque@adacore.com>
2745 * common.opt (nolibc): New option.
2746 * doc/invoke.texi (Link Options): Document it.
2747 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2748 * config/alpha/linux.h: Likewise.
2749 * config/arc/elf.h: Likewise.
2750 * config/arm/uclinux-elf.h: Likewise.
2751 * config/arm/unknown-elf.h: Likewise.
2752 * config/avr/avrlibc.h: Likewise.
2753 * config/bfin/bfin.h: Likewise.
2754 * config/bfin/linux.h: Likewise.
2755 * config/bfin/uclinux.h: Likewise.
2756 * config/darwin.h: Likewise.
2757 * config/darwin10.h: Likewise.
2758 * config/darwin12.h: Likewise.
2759 * config/gnu-user.h: Likewise.
2760 * config/lm32/uclinux-elf.h: Likewise.
2761 * config/pa/pa-hpux11.h: Likewise.
2762 * config/pa/pa64-hpux.h: Likewise.
2763 * config/sparc/sparc.h: Likewise.
2765 2018-07-31 Olivier Hainque <hainque@adacore.com>
2767 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2768 undefined variables.
2770 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2773 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2776 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2778 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2779 terminated string literal.
2781 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2783 PR rtl-optimization/85160
2784 * combine.c (is_just_move): New function.
2785 (try_combine): Allow combining two instructions into two if neither of
2786 the original instructions was a move.
2788 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2791 * doc/extend.texi (Global Register Variables): Discourage use of type
2793 (Local Register Variables): Likewise.
2795 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2797 PR tree-optimization/86506
2798 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2800 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2803 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2804 hard_regno, make sure no insn between `from` and `to` clobbers it.
2806 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2807 Tom de Vries <tdevries@suse.de>
2809 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2810 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2811 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2812 PTX_DEFAULT_RUNTIME_DIM.
2813 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2815 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2817 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2818 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2819 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2821 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2823 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2825 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2826 predicate to reflect toc fusion has been deleted.
2827 (toc_fusion_mem_raw): Delete, no longer used.
2828 (toc_fusion_mem_wrapped): Likewise.
2829 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2831 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2832 Delete, no longer used.
2833 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2834 meant to be used for toc fusion.
2835 (rs6000_debug_print_mode): Delete toc fusion debugging.
2836 (rs6000_debug_reg_global): Likewise.
2837 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2838 fusion and secondary reload support that were never used.
2839 (rs6000_option_override_internal): Delete TOC fusion, that was only
2840 partially defined, and it did not work unless you also used the
2842 (rs6000_legitimate_address_p): Delete TOC fusion support.
2843 (rs6000_opt_masks): Likewise.
2844 (fusion_wrap_memory_address): Delete function, no longer used.
2845 (fusion_split_address); Delete TOC fusion support.
2846 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2847 longer used with toc fusion being deleted.
2848 (TARGET_TOC_FUSION_FP): Likewise.
2849 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2851 (toc fusion spliter): Delete TOC fusion support.
2852 (toc_fusionload_<mode>): Likewise.
2853 (toc_fusionload_di): Likewise.
2854 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2855 longer needs to be named. Rename predicate to delete TOC fusion.
2856 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2857 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2858 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2859 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2860 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2862 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2864 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2865 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2866 __int128 in built-in function prototypes.
2867 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2868 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2870 2018-07-27 Martin Sebor <msebor@redhat.com>
2872 PR tree-optimization/86696
2873 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2874 types, including enums.
2875 (handle_char_store): Be prepared for the above function to fail.
2877 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2879 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2880 when optimization level is lower than 2 or optimize for size.
2882 2018-07-26 Martin Sebor <msebor@redhat.com>
2884 PR tree-optimization/86043
2885 PR tree-optimization/86042
2886 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2887 (get_string_cst_length): Rename...
2888 (get_min_string_length): ...to this. Add argument.
2889 (handle_char_store): Extend to handle multi-character stores by
2891 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2892 * tree.h (initializer_zerop): Add argument.
2894 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2897 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2898 declare target to variables if they have always,{to,from,tofrom} map
2901 2018-07-26 Martin Liska <mliska@suse.cz>
2904 * lto-wrapper.c: Add linker_output as prefix
2905 for ltrans_output_file.
2907 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2909 PR rtl-optimization/85805
2910 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2911 value for hard registers if that was written in the same mode.
2913 2018-07-26 Martin Liska <mliska@suse.cz>
2915 PR gcov-profile/86536
2916 * gcov.c (format_gcov): Use printf format %.*f directly
2917 and do not handle special values.
2919 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2921 * common/config/arc/arc-common.c (arc_option_optimization_table):
2922 Update default optimizations for size.
2924 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2926 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2927 * config/arc/constraints.md (CfZ): New constraint.
2928 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2929 (subsf3_fpu): Likewise.
2930 (cmpsf_fpu): Likewise.
2931 (cmpsf_fpu_uneq): Likewise.
2933 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2935 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2937 (arc_is_uncached_mem_p): uncached applies to both the variable and
2940 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2942 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2945 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2947 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2948 field "m_scopes" from vec to auto_vec.
2950 2018-07-25 Martin Liska <mliska@suse.cz>
2952 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2955 2018-07-25 Richard Biener <rguenther@suse.de>
2958 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2959 special wrt context_die late.
2960 (gen_subprogram_die): Re-use DIEs in local scope.
2962 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2964 PR tree-optimization/86644
2965 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2967 2018-07-25 Martin Liska <mliska@suse.cz>
2970 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2971 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2973 2018-07-25 Martin Liska <mliska@suse.cz>
2976 * params.def: Explain ASan abbreviation and provide
2977 a documentation link.
2979 2018-07-24 Martin Sebor <msebor@redhat.com>
2981 PR tree-optimization/86622
2982 PR tree-optimization/86532
2983 * builtins.h (string_length): Declare.
2984 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2985 (check_access): Be prepared for non-constant length ranges.
2986 (string_length): Make extern.
2987 * expr.c (string_constant): Only handle the minor non-constant
2988 array index. Use string_constant to compute the length of
2989 a generic string constant.
2991 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2993 PR tree-optimization/86618
2994 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2995 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2997 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2999 PR tree-optimization/86636
3000 * json.cc (json::object::set): Fix comment. Add assertions.
3001 (json::array::append): Move here from json.h. Add comment and an
3003 (json::string::string): Likewise.
3004 * json.h (json::array::append): Move to json.cc.
3005 (json::string::string): Likewise.
3006 * optinfo-emit-json.cc
3007 (optrecord_json_writer::impl_location_to_json): Assert that we
3008 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
3009 wrapper around it. Expand the location once, rather than three
3011 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
3012 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
3014 (optrecord_json_writer::optinfo_to_json): Likewise, in four
3015 places. Fix some overlong lines.
3017 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3019 * config/aarch64/aarch64-simd.md
3020 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
3021 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
3022 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
3023 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
3024 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
3025 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
3026 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
3027 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
3028 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
3030 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3033 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3034 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3035 and abs_d == d, do the power of two handling if profitable.
3037 2018-07-24 Richard Biener <rguenther@suse.de>
3039 * match.pd: Add BIT_FIELD_REF canonicalizations.
3041 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3044 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3046 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3048 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
3049 terminated STRING_CST object.
3051 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3053 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
3055 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3057 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3058 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
3059 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
3060 the elements into a register.
3061 (rs6000_split_v4si_init_di_reg): Delete.
3062 (rs6000_split_v4si_init): Delete.
3063 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
3064 (vsx_init_v4si): Rewrite as a define_expand.
3066 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3068 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
3069 zero_extend argument from memory): New.
3071 2018-07-22 Martin Sebor <msebor@redhat.com>
3074 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
3075 diagnosing calls with unknown arguments unless -Walloca-larger-than
3076 is restricted to less than PTRDIFF_MAX bytes.
3078 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
3080 * doc/gcov.texi (Invoking Gcov): Editorial changes.
3082 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3084 * pretty-print.c (text_info::set_location): Remove redundant
3085 "line_table" parameter from call to rich_location::set_range.
3087 2018-07-20 Martin Sebor <msebor@redhat.com>
3090 * builtins.c (expand_builtin_alloca): Adjust.
3091 * calls.c (alloc_max_size): Simplify.
3092 * cgraphunit.c (cgraph_node::expand): Adjust.
3093 * common.opt (larger_than_size, warn_frame_larger_than): Remove
3095 (frame_larger_than_size): Same.
3096 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
3097 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
3099 * doc/invoke.texi (GCC Command Options): Document option arguments.
3100 Explain byte-size arguments and suffixes.
3101 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
3102 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
3103 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
3104 * doc/options.texi (UInteger): Expand.
3105 (Host_Wide_Int, ByteSize): Document new properties.
3106 * final.c (final_start_function_1): Include sizes in an error message.
3107 * function.c (frame_offset_overflow): Same.
3108 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
3109 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
3110 Diagnose unbounded alloca calls only for limits of less than
3112 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
3113 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
3115 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
3116 only for limits of less than PTRDIFF_MAX.
3117 * langhooks-def.h (lhd_handle_option): Change function argument
3119 * langhooks.c (lhd_handle_option): Same.
3120 * langhooks.h (handle_option): Same.
3121 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
3123 (var_type, var_type_struct): Same.
3124 (var_set): Handle ByteSize flag.
3125 * optc-gen.awk: Add comments to output to ease debugging. Make
3126 use of HOST_WIDE_INT where appropriate.
3127 * opts-gen-save.awk: Use %lx to format unsigned long.
3128 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
3129 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
3130 arguments. Parse bytes-size suffixes.
3131 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
3132 (enum_value_to_arg): Same.
3133 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
3134 (handle_option): Adjust.
3135 (generate_option): Change function argument to HOST_WIDE_INT.
3136 (cmdline_handle_error): Adjust.
3137 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
3138 (set_option): Change function argument to HOST_WIDE_INT.
3139 (option_enabled): Handle cl_host_wide_int.
3140 (get_option_state): Handle CLVC_SIZE.
3141 (control_warning_option): Same.
3142 * opts.c (common_handle_option): Change function argument to
3143 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
3144 OPT_Wvla_larger_than_.
3145 * opts.h (enum cl_var_type): Add an enumerator.
3146 * stor-layout.c (layout_decl): Print a more meaningful warning.
3147 * toplev.c (output_stack_usage): Adjust.
3149 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
3151 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
3152 call to inline_expand_builtin_string_cmp.
3153 (expand_builtin_strcmp): Likewise.
3154 (expand_builtin_strncmp): Likewise.
3155 (inline_string_cmp): Delete the last parameter, change char_type_node
3156 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
3158 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
3159 the inlining expansion on target where the type of the call has same or
3160 narrower precision than unsigned char.
3162 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3164 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
3165 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
3166 * common.opt (fsave-optimization-record): New option.
3167 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
3168 * doc/invoke.texi (-fsave-optimization-record): New option.
3169 * dumpfile.c: Include "optinfo-emit-json.h".
3170 (struct kv_pair): Move to coretypes.h.
3171 (optgroup_options): Make non-static.
3172 (dump_context::end_scope): Call
3173 optimization_records_maybe_pop_dump_scope.
3174 * dumpfile.h (optgroup_options): New decl.
3175 * json.cc: New file.
3177 * optinfo-emit-json.cc: New file.
3178 * optinfo-emit-json.h: New file.
3179 * optinfo.cc: Include "optinfo-emit-json.h".
3180 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
3181 (optinfo_enabled_p): Check optimization_records_enabled_p.
3182 (optinfo_wants_inlining_info_p): Likewise.
3183 * optinfo.h: Update comment.
3184 * profile-count.c (profile_quality_as_string): New function.
3185 * profile-count.h (profile_quality_as_string): New decl.
3186 (profile_count::quality): New accessor.
3187 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
3188 and optinfo_emit_json_cc_tests.
3189 * selftest.h (selftest::json_cc_tests): New decl.
3190 (selftest::optinfo_emit_json_cc_tests): New decl.
3191 * toplev.c: Include "optinfo-emit-json.h".
3192 (compile_file): Call optimization_records_finish.
3193 (do_compile): Call optimization_records_start.
3194 * tree-ssa-live.c: Include optinfo.h.
3195 (remove_unused_scope_block_p): Retain inlining information if
3196 optinfo_wants_inlining_info_p returns true.
3198 2018-07-20 Richard Biener <rguenther@suse.de>
3201 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
3202 to cover -flto-partition=none.
3204 2018-07-20 Martin Liska <mliska@suse.cz>
3206 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
3207 (get_decl_source_range): Remove unused function.
3209 2018-07-20 Richard Biener <rguenther@suse.de>
3211 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
3212 (struct vn_phi_s): Likewise.
3213 (struct vn_reference_s): Likewise.
3214 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
3215 for searching the slot of an entry known to be in the hash itself.
3216 (vn_phi_hasher::equal): Likewise.
3217 (vn_reference_hasher::equal): Likewise.
3218 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
3220 (optimistic_info, current_info): Remove, keeping only valid_info.
3221 (vn_reference_lookup_1): Remove fallback lookup.
3222 (vn_reference_lookup_2): Likewise.
3223 (vn_nary_op_lookup_1): Likewise.
3224 (vn_phi_lookup): Likewise.
3225 (vn_nary_build_or_lookup_1): Make sure to not chain the built
3227 (vn_reference_insert): Adjust, chain the inserted hash element
3228 at last_inserted_ref.
3229 (vn_reference_insert_pieces): Likewise.
3230 (visit_reference_op_call): Likewise.
3231 (vn_nary_op_insert_into): Chain the inserted hash element at
3233 (vn_nary_op_insert_pieces): Adjust.
3234 (vn_nary_op_insert): Likewise.
3235 (vn_nary_op_insert_stmt): Likewise.
3236 (vn_phi_insert): Adjust, chain the inserted hash element at
3238 (process_scc): Remove clearing and copying the optimistic
3239 table. Instead remove elements inserted during an optimistic
3240 iteration from the single table we maintain.
3241 (init_scc_vn): Adjust.
3242 (free_scc_vn): Likewise.
3243 (sccvn_dom_walker::record_cond): Likewise.
3244 (sccvn_dom_walker::after_dom_children): Likewise.
3246 2018-07-19 Martin Sebor <msebor@redhat.com>
3248 PR tree-optimization/84047
3249 PR tree-optimization/83776
3250 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
3251 (check_array_bounds): Call it.
3253 2018-07-19 Martin Sebor <msebor@redhat.com>
3255 * align.h (align_flags): Use member initialization.
3257 2018-07-19 David Malcolm <dmalcolm@redhat.com>
3259 * Makefile.in (OBJS): Add optinfo.o.
3260 * coretypes.h (class symtab_node): New forward decl.
3261 (struct cgraph_node): New forward decl.
3262 (class varpool_node): New forward decl.
3263 * dump-context.h: New file.
3264 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3266 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3267 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3268 (set_alt_dump_file): Likewise.
3269 (dump_context::~dump_context): New dtor.
3270 (dump_gimple_stmt): Move implementation to...
3271 (dump_context::dump_gimple_stmt): ...this new member function.
3272 Add the stmt to any pending optinfo, creating one if need be.
3273 (dump_gimple_stmt_loc): Move implementation to...
3274 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3275 Start a new optinfo and add the stmt to it.
3276 (dump_gimple_expr): Move implementation to...
3277 (dump_context::dump_gimple_expr): ...this new member function.
3278 Add the stmt to any pending optinfo, creating one if need be.
3279 (dump_gimple_expr_loc): Move implementation to...
3280 (dump_context::dump_gimple_expr_loc): ...this new member function.
3281 Start a new optinfo and add the stmt to it.
3282 (dump_generic_expr): Move implementation to...
3283 (dump_context::dump_generic_expr): ...this new member function.
3284 Add the tree to any pending optinfo, creating one if need be.
3285 (dump_generic_expr_loc): Move implementation to...
3286 (dump_context::dump_generic_expr_loc): ...this new member
3287 function. Add the tree to any pending optinfo, creating one if
3289 (dump_printf): Move implementation to...
3290 (dump_context::dump_printf_va): ...this new member function. Add
3291 the text to any pending optinfo, creating one if need be.
3292 (dump_printf_loc): Move implementation to...
3293 (dump_context::dump_printf_loc_va): ...this new member function.
3294 Start a new optinfo and add the stmt to it.
3295 (dump_dec): Move implementation to...
3296 (dump_context::dump_dec): ...this new member function. Add the
3297 value to any pending optinfo, creating one if need be.
3298 (dump_context::dump_symtab_node): New member function.
3299 (dump_context::get_scope_depth): New member function.
3300 (dump_context::begin_scope): New member function.
3301 (dump_context::end_scope): New member function.
3302 (dump_context::ensure_pending_optinfo): New member function.
3303 (dump_context::begin_next_optinfo): New member function.
3304 (dump_context::end_any_optinfo): New member function.
3305 (dump_context::s_current): New global.
3306 (dump_context::s_default): New global.
3307 (dump_scope_depth): Delete global.
3308 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3309 (dump_symtab_node): New function.
3310 (get_dump_scope_depth): Reimplement in terms of dump_context.
3311 (dump_begin_scope): Likewise.
3312 (dump_end_scope): Likewise.
3313 (selftest::temp_dump_context::temp_dump_context): New ctor.
3314 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3315 (selftest::verify_item): New function.
3316 (ASSERT_IS_TEXT): New macro.
3317 (ASSERT_IS_TREE): New macro.
3318 (ASSERT_IS_GIMPLE): New macro.
3319 (selftest::test_capture_of_dump_calls): New test.
3320 (selftest::dumpfile_c_tests): Call it.
3321 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3322 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3323 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3324 descriptive comment.
3325 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3326 (dump_node, dump_bb): Move these unrelated decls.
3327 (class dump_manager): Add leading comment.
3328 * optinfo.cc: New file.
3329 * optinfo.h: New file.
3331 2018-07-19 Michael Collison <michael.collison@arm.com>
3332 Richard Henderson <rth@redhat.com>
3334 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3335 (subti): Handle op1 zero.
3336 (subvti4, usub4ti4): New.
3337 (*sub<GPI>3_compare1_imm): New.
3338 (sub<GPI>3_carryinCV): New.
3339 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3340 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3342 2018-07-19 Michael Collison <michael.collison@arm.com>
3343 Richard Henderson <rth@redhat.com>
3345 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3346 (addti3): Create simpler code if low part is already known to be 0.
3347 (addvti4, uaddvti4): New.
3348 (*add<GPI>3_compareC_cconly_imm): New.
3349 (*add<GPI>3_compareC_cconly): New.
3350 (*add<GPI>3_compareC_imm): New.
3351 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3352 handle constants within this pattern..
3353 (*add<GPI>3_compareV_cconly_imm): New.
3354 (*add<GPI>3_compareV_cconly): New.
3355 (*add<GPI>3_compareV_imm): New.
3356 (add<GPI>3_compareV): New.
3357 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3358 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3359 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3360 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3361 with 'comparison' operator.
3362 (*add<GPI>3_compareV_cconly_imm): Ditto.
3363 (*add<GPI>3_compareV_cconly): Ditto.
3364 (*add<GPI>3_compareV_imm): Ditto.
3365 (add<GPI>3_compareV): Ditto.
3366 (add<mode>3_carryinC): Ditto.
3367 (*add<mode>3_carryinC_zero): Ditto.
3368 (*add<mode>3_carryinC): Ditto.
3369 (add<mode>3_carryinV): Ditto.
3370 (*add<mode>3_carryinV_zero): Ditto.
3371 (*add<mode>3_carryinV): Ditto.
3373 2018-07-19 Michael Collison <michael.collison@arm.com>
3374 Richard Henderson <rth@redhat.com>
3376 * config/aarch64/aarch64-modes.def (CC_V): New.
3377 * config/aarch64/aarch64-protos.h
3378 (aarch64_addti_scratch_regs): Declare
3379 (aarch64_subvti_scratch_regs): Declare.
3380 (aarch64_expand_subvti): Declare.
3381 (aarch64_gen_unlikely_cbranch): Declare
3382 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3383 for signed overflow using CC_Vmode.
3384 (aarch64_get_condition_code_1): Handle CC_Vmode.
3385 (aarch64_gen_unlikely_cbranch): New function.
3386 (aarch64_addti_scratch_regs): New function.
3387 (aarch64_subvti_scratch_regs): New function.
3388 (aarch64_expand_subvti): New function.
3390 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3392 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3394 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3395 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3398 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3401 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3402 address check not strict.
3404 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3406 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3408 (store_pair_lanes<mode>): Likewise.
3409 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3410 enum value 'ADDR_QUERY_LDP_STP_N'.
3411 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3412 (aarch64_print_address_internal): Add declaration.
3413 (aarch64_print_ldpstp_address): Remove.
3414 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3415 (aarch64_print_operand): Change printing of 'y'.
3416 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3417 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3418 'true' rather than '1'.
3419 * config/aarch64/constraints.md (Uml): Likewise.
3420 (Uml): Rename to Umn.
3423 2018-07-19 Richard Biener <rguenther@suse.de>
3425 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3427 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3428 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3429 (vn_reference_hasher): Likewise.
3430 (struct vn_tables_s): Remove obstack and alloc-pool members.
3431 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3432 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3433 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3434 (vn_reference_insert_pieces): Likewise.
3435 (alloc_vn_nary_op_noinit): Adjust.
3436 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3437 (vn_phi_eq): Adjust.
3438 (shared_lookup_phiargs): Remove.
3439 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3440 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3441 (visit_reference_op_call): Likewise.
3442 (copy_nary, copy_phi, copy_reference): Remove.
3443 (process_scc): Rewind the obstack when iterating. Do not
3444 copy the elements to valid_info but just move them from one
3445 hashtable to the other.
3446 (allocate_vn_table): Adjust.
3447 (free_vn_table): Likewise.
3448 (init_scc_vn): Likewise.
3449 (free_scc_vn): Likewise.
3451 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3454 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3455 indirect_return as function type attribute.
3456 (ix86_attribute_table): Change indirect_return to function
3458 * doc/extend.texi: Update indirect_return attribute.
3460 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3462 * wide-int.h (widest2_int): New.
3463 * gimple-fold.c (arith_overflowed_p): Use it.
3464 * tree.h (widest2_int_cst): New.
3465 * tree-vrp.c (wide_int_binop_overflow): Rename from
3466 vrp_int_const_binop.
3467 Rewrite to work on trees.
3468 (extract_range_from_multiplicative_op_1): Abstract code to...
3469 (wide_int_range_min_max): ...here.
3470 (wide_int_range_cross_product): ...and here.
3471 (extract_range_from_binary_expr_1): Abstract overflow code to...
3472 (wide_int_range_mult_wrapping): ...here.
3473 * tree-vrp.h (wide_int_range_cross_product): New.
3474 (wide_int_range_mult_wrapping): New.
3476 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3477 Julia Koval <julia.koval@intel.com>
3479 * config/i386/x86-tune-costs.h (skylake_memcpy,
3480 skylake_memset): Replace rep_prefix with unrolling for size 512.
3482 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3485 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3486 comparision with EQ_EXPR in last stmt.
3488 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3490 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3491 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3492 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3493 previously known as "PowerPC AltiVec Built-in Functions". Move
3494 some material to new subsubsections "PowerPC AltiVec Built-in
3495 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3497 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3498 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3499 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3500 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3502 2018-07-18 Richard Biener <rguenther@suse.de>
3504 PR tree-optimization/86557
3505 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3508 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3510 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3512 2018-07-17 Jeff Law <law@redhat.com>
3514 * config/arm/arm.c (get_label_padding): Update for recent
3515 changes to label_to_alignment.
3517 PR tree-optimization/86010
3518 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3520 * config/mips/mips.c (vr4130_align_insns): Update for recent
3521 changes to label_to_alignment.
3523 * config/frv/frv.c (frv_label_align): Update for recent changes
3524 to label_to_alignment.
3526 * config/nios2/nios2.c (nios2_label_align): Update for recent
3527 changes which dropped ALIGN_LABELS_LOG.
3529 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3531 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3532 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3534 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3536 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3537 deprecated align_labels_log.
3539 2018-07-17 Richard Biener <rguenther@suse.de>
3542 * dwarf2out.c (init_sections_and_labels): Always generate
3543 a debug_line_str_section for early LTO debug.
3544 (dwarf2out_finish): Reset debug_line_str_hash output early.
3545 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3547 (dwarf2out_early_finish): Output debug_line_str.
3549 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3551 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3552 index register on z196 or later.
3554 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3556 * config/s390/s390.c (s390_default_align): Set default function
3558 (s390_override_options_after_change): Call s390_default align.
3559 (s390_option_override_internal): Call s390_default_align.
3560 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3562 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3565 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3566 to _looptemp_ clauses, other than the first two.
3568 2018-07-17 Martin Liska <mliska@suse.cz>
3570 * opts.c: Do not enable OPT_falign_* for -Os.
3572 2018-07-17 Martin Liska <mliska@suse.cz>
3574 * align.h (MAX_CODE_ALIGN): New.
3575 (MAX_CODE_ALIGN_VALUE): New.
3576 * common/config/i386/i386-common.c (ix86_handle_option):
3577 (MAX_CODE_ALIGN): Moved to align.h.
3578 * final.c (MAX_CODE_ALIGN): Likewise.
3579 * opts.c (parse_and_check_align_values):
3580 (MAX_CODE_ALIGN): Likewise.
3581 (MAX_CODE_ALIGN_VALUE): Likewise.
3583 2018-07-17 Martin Liska <mliska@suse.cz>
3585 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3586 in order to fulfil coding style.
3587 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3588 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3589 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3590 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3591 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3592 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3593 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3594 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3595 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3597 2018-07-17 Martin Liska <mliska@suse.cz>
3599 * align.h: New file.
3600 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3602 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3603 align_flags of label_to_alignment.
3604 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3606 * config/m68k/m68k.c: Do not use removed align_labels_value and
3608 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3609 (LOOP_ALIGN): Likewise.
3610 (LABEL_ALIGN): Likewise.
3611 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3612 Remove not used macro.
3613 (rs6000_loop_align): Change return type to align_flags.
3614 (rs6000_loop_align_max_skip): Remove.
3615 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3616 Change return type to align_flags.
3617 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3618 Remove not used macro.
3619 (rs6000_loop_align): Change return type to align_flags.
3620 (rs6000_loop_align_max_skip): Remove.
3621 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3622 * config/rx/rx-protos.h (rx_align_for_label): Make it
3624 * config/rx/rx.c (rx_align_for_label): Change return type
3626 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3628 into align_flags class.
3629 (LABEL_ALIGN): Likewise.
3630 (LOOP_ALIGN): Likewise.
3631 * config/s390/s390.c (s390_label_align): Use align_flags
3633 (s390_asm_output_function_label): Likewise.
3634 * config/sh/sh.c (sh_override_options_after_change):
3635 Use align_flags class directly without macros.
3636 (find_barrier): Likewise.
3637 (barrier_align): Likewise.
3638 (sh_loop_align): Likewise.
3639 * config/spu/spu.c (spu_option_override):
3640 Use align_flags_tuple::get_value instead of removed macros.
3641 (spu_sched_init): Likewise.
3642 * config/spu/spu.h (GTY): Likewise.
3643 * config/visium/visium.c (visium_option_override):
3644 Set "8" as default secondary alignment.
3645 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3646 in order to guarantee secondary alignment of 8.
3647 * coretypes.h: Include align.h header file.
3648 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3649 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3650 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3651 * doc/tm.texi.in: Likewise.
3652 * final.c (struct label_alignment): Remove not used structure.
3653 (LABEL_ALIGN): Change type to align_flags.
3654 (LOOP_ALIGN): Likewise.
3655 (JUMP_ALIGN): Likewise.
3656 (default_loop_align_max_skip): Remove.
3657 (default_label_align_max_skip): Likewise.
3658 (default_jump_align_max_skip): Likewise.
3659 (default_label_align_after_barrier_max_skip):
3660 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3661 (LABEL_TO_MAX_SKIP): Remove.
3662 (label_to_alignment): Return align_flags type instead of integer.
3663 (label_to_max_skip): Remove.
3664 (align_fuzz): Use align_flags type.
3665 (compute_alignments): Use align_flags type and use align_flags::max
3666 to combine multiple alignments.
3667 (grow_label_align): Grow vec instead of C array.
3668 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3669 (shorten_branches): Use align_flags type and use align_flags::max
3670 to combine multiple alignments.
3671 (final_scan_insn_1): Remove usage of secondary alignment that comes
3672 from label alignment, but instead use proper secondary alignment
3673 which is computed in grow_label_align.
3674 * flags.h (struct align_flags_tuple): Move to align.h.
3675 (struct align_flags): Likewise.
3676 (state_align_loops): Rename to align_loops.
3677 (state_align_jumps): Rename to align_jumps.
3678 (state_align_labels): Rename to align_labels.
3679 (state_align_functions): Rename to align_functions.
3680 (align_loops_log): Remove.
3681 (align_jumps_log): Remove.
3682 (align_labels_log): Remove.
3683 (align_functions_log): Remove.
3684 (align_loops_max_skip): Remove.
3685 (align_jumps_max_skip): Remove.
3686 (align_labels_max_skip): Remove.
3687 (align_functions_max_skip): Remove.
3688 (align_loops_value): Remove.
3689 (align_jumps_value): Remove.
3690 (align_labels_value): Remove.
3691 (align_functions_value): Remove.
3692 * output.h (label_to_alignment): Change return type to align_flags.
3693 (label_to_max_skip): Remove.
3694 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3695 jump_align_max_skip macros.
3696 * targhooks.h (default_loop_align_max_skip): Remove.
3697 (default_label_align_max_skip): Likewise.
3698 (default_jump_align_max_skip): Likewise.
3699 (default_label_align_after_barrier_max_skip): Remove.
3700 * toplev.c (read_log_maxskip): Use ::normalize function.
3701 (parse_N_M): Remove not used argument and also call ::normalize.
3702 (parse_alignment_opts): Do not pass unused arguments.
3703 * varasm.c (assemble_start_function): Use directly align_functions
3704 instead of removed macros.
3705 * system.h: Do not poison removed macros.
3707 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3710 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3711 and cond temporaries don't have reference type if iterator has
3712 pointer type. For init use &for_pre_body instead of pre_p if
3713 for_pre_body is non-empty.
3715 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3717 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3718 double-double modes to SFmode directly directly.
3719 (trunc<mode>sf2_fprs): Delete.
3721 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3723 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3724 for conversions between IFmode and the decimal floating point modes.
3725 (init_float128_ieee): Use the correct names for conversions between
3726 KFmode and the decimal floating point modes.
3728 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3730 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3731 for the conversions between TDmode and IFmode.
3732 (init_float128_ieee): Use more correct names for the conversions
3733 between TDmode and KFmode.
3735 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3737 PR tree-optimization/86526
3738 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3739 (inline_expand_builtin_string_cmp): Likewise.
3740 (inline_string_cmp): Likewise. Use c_readstr instead of
3741 builtin_memcpy_read_str. Add unit_mode temporary.
3743 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3746 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3747 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3749 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3751 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3752 Alphabetize prototypes of built-in functions, separating out
3753 built-in functions that are listed in this section but should be
3754 described elsewhere.
3756 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3759 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3760 conditional move when trapping comparison was split to a
3761 non-trapping one (and vice versa).
3763 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3765 * config/s390/s390.c (s390_function_profiler): Generate nops
3766 instead of profiler call sequences.
3767 * config/s390/s390.opt: Add the new option.
3769 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3771 * config/s390/s390.c (s390_function_profiler): Generate
3772 __mcount_loc section.
3773 * config/s390/s390.opt: Add the new option.
3775 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3777 * common.opt: Add the new warning.
3778 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3779 %r0,__fentry__" when -mfentry is specified.
3780 (s390_option_override_internal): Disallow -mfentry for 31-bit
3782 * config/s390/s390.opt: Add the new option.
3784 2018-07-16 Richard Biener <rguenther@suse.de>
3787 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3788 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3790 2018-07-16 Martin Liska <mliska@suse.cz>
3793 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3796 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3798 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3800 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3802 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3804 * config/arc/arc-cpus.def (hs4x): New cpu.
3806 * config/arc/arc-tables.opt: Regenerate.
3807 * config/arc/arc.c (arc_sched_issue_rate): New function.
3808 (TARGET_SCHED_ISSUE_RATE): Define.
3809 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3810 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3812 (attr tune): Add ARCHS4x tune values.
3813 (attr tune_dspmpy): Define.
3814 (*tst): Correct instruction type.
3815 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3816 * config/arc/arcHS4x.md: New file.
3817 * config/arc/fpu.md: Update instruction type attributes.
3818 * config/arc/t-multilib: Regenerate.
3820 2018-07-16 Tom de Vries <tdevries@suse.de>
3823 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3825 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3827 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3830 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3832 PR tree-optimization/86514
3833 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3834 conversion to a boolean type from a type with greater precision.
3836 2018-07-16 Tom de Vries <tdevries@suse.de>
3838 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3839 dump_flags request TDF_SLIM.
3841 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3843 * fold-const.c (int_const_binop_1): Abstract...
3844 (wide_int_binop): ...wide int code here.
3845 (poly_int_binop): ...poly int code here.
3846 Abstract the rest of int_const_binop_1 into int_const_binop.
3847 * fold-const.h (wide_int_binop): New.
3848 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3849 Remove useless PLUS/MINUS_EXPR case.
3850 (zero_nonzero_bits_from_vr): Move wide int code...
3851 (zero_nonzero_bits_from_bounds): ...here.
3852 (extract_range_from_binary_expr_1): Move mask optimization code...
3853 (range_easy_mask_min_max): ...here.
3854 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3855 (range_easy_mask_min_max): New.
3857 2018-07-15 Jeff Law <law@redhat.com>
3860 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3863 2018-07-14 Jim Wilson <jimw@sifive.com>
3865 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3867 2018-07-14 Paul Koning <ni1d@arrl.net>
3869 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3871 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3873 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3876 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3877 Steve Munroe <munroesj52@gmail.com>
3879 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3880 (_mm_andnot_si128): Likewise.
3881 (_mm_or_si128): Likewise.
3882 (_mm_xor_si128): Likewise.
3884 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3887 * builtins.c (expand_builtin_memcmp): Inline the calls first
3888 when result_eq is false.
3889 (expand_builtin_strcmp): Inline the calls first.
3890 (expand_builtin_strncmp): Likewise.
3891 (inline_string_cmp): New routine. Expand a string compare
3892 call by using a sequence of char comparison.
3893 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3894 a call to str(n)cmp/memcmp.
3895 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3897 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3899 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3901 * config/arm/driver-arm.c: Include arm-native.h.
3902 (host_detect_local_cpu): Use auto-generated data tables.
3903 (vendors, arm_cpu_table): Delete. Move part information to ...
3904 * config/arm/arm-cpus.in: ... here.
3905 * config/arm/parsecpu.awk (gen_native): New function.
3906 (vendor, part): New CPU fields.
3907 (END): Add support for building the native CPU detection tables.
3908 * config/arm/t-arm (arm-native.h): Add build rule.
3909 (driver-arm.o): Add dependency on arm-native.h.
3911 2018-07-13 Richard Biener <rguenther@suse.de>
3914 * match.pd (addr1 - addr2): Allow either of the operand to
3917 2018-07-13 Tom de Vries <tdevries@suse.de>
3919 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3922 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3924 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3925 arrays instead of numbered variables.
3927 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3929 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3930 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3932 2018-07-13 Richard Biener <rguenther@suse.de>
3935 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3936 instead of get_context_die.
3938 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3939 Richard Biener <rguenther@suse.de>
3942 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3943 that the loop latch destination where phi is defined.
3945 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3947 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3948 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3949 (riscv_set_current_function): Extract getting interrupt type to new
3951 (riscv_get_interrupt_type): New function.
3952 (riscv_merge_decl_attributes): New function, checking interrupt type is
3954 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3956 2018-07-12 Paul Koning <ni1d@arrl.net>
3958 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3961 2018-07-12 Paul Koning <ni1d@arrl.net>
3963 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3964 zero reference, add doloop_end instead.
3965 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3966 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3967 detail for doloop_end.
3969 2018-07-12 Martin Sebor <msebor@redhat.com>
3972 * attribs.c (decl_attributes): Reject conflicting attributes before
3973 calling attribute handlers.
3975 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3977 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3979 (gcc::dump_manager::get_dump_file_name): likewise.
3980 (dump_begin): Likewise.
3981 * dumpfile.h (dump_begin): Update prototype.
3982 (gcc::dump_manager::get_dump_file_name,
3983 gcc::dump_manager::get_dump_file_name): Update prototype.
3985 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3987 * internal-fn.h (vectorizable_internal_fn_p): New function.
3988 * tree-vect-slp.c (compatible_calls_p): Likewise.
3989 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3990 to internal functions.
3991 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3993 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3995 * fold-const.h (inverse_conditions_p): Declare.
3996 * fold-const.c (inverse_conditions_p): New function.
3997 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3998 that test the inverse condition of a conditional internal function.
3999 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
4000 * internal-fn.c (internal_fn_mask_index): Handle conditional
4002 (vectorized_internal_fn_supported_p): New function.
4003 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
4004 (any_pred_load_store): Replace with...
4005 (need_to_predicate): ...this new variable.
4006 (redundant_ssa_names): New variable.
4007 (ifcvt_can_use_mask_load_store): Move initial checks to...
4008 (ifcvt_can_predicate): ...this new function. Handle tree codes
4009 for which a conditional internal function exists.
4010 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
4011 instead of ifcvt_can_use_mask_load_store. Update after variable
4013 (predicate_load_or_store): New function, split out from
4014 predicate_mem_writes.
4015 (check_redundant_cond_expr): New function.
4016 (value_available_p): Likewise.
4017 (predicate_rhs_code): Likewise.
4018 (predicate_mem_writes): Rename to...
4019 (predicate_statements): ...this. Use predicate_load_or_store
4020 and predicate_rhs_code.
4021 (combine_blocks, tree_if_conversion): Update after above name changes.
4022 (ifcvt_local_dce): Handle redundant_ssa_names.
4023 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
4024 general conditional functions.
4025 * tree-vect-stmts.c (vectorizable_call): Likewise.
4027 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4028 Alan Hayward <alan.hayward@arm.com>
4029 David Sherwood <david.sherwood@arm.com>
4031 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
4032 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
4033 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
4034 plus and minus and convert them into IFN_COND_FMA-based sequences.
4035 (convert_mult_to_fma): Handle conditional plus and minus.
4037 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4039 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
4040 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
4041 (cond_fnms_optab): New optabs.
4042 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
4044 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
4045 * internal-fn.h (get_conditional_internal_fn): Declare.
4046 (get_unconditional_internal_fn): Likewise.
4047 * internal-fn.c (cond_ternary_direct): New macro.
4048 (expand_cond_ternary_optab_fn): Likewise.
4049 (direct_cond_ternary_optab_supported_p): Likewise.
4050 (FOR_EACH_COND_FN_PAIR): Likewise.
4051 (get_conditional_internal_fn): New function.
4052 (get_unconditional_internal_fn): Likewise.
4053 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
4054 (gimple_match_op::gimple_match_op): Add a new overload for 5
4056 (gimple_match_op::set_op): Likewise.
4057 (gimple_resimplify5): Declare.
4058 * genmatch.c (decision_tree::gen): Generate simplifications for
4060 * gimple-match-head.c (gimple_simplify): Define an overload for
4061 5 operands. Handle calls with 5 arguments in the top-level overload.
4062 (convert_conditional_op): Handle conversions from unconditional
4063 internal functions to conditional ones.
4064 (gimple_resimplify5): New function.
4065 (build_call_internal): Pass a fifth operand.
4066 (maybe_push_res_to_seq): Likewise.
4067 (try_conditional_simplification): Try converting conditional
4068 internal functions to unconditional internal functions.
4069 Handle 3-operand unconditional forms.
4070 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
4071 Define ternary equivalents of the current rules for binary conditional
4073 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
4075 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
4076 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
4077 (optab): Handle them.
4078 (SVE_COND_FP_TERNARY): New int iterator.
4079 (sve_fmla_op, sve_fmad_op): New int attributes.
4080 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
4081 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
4082 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
4084 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4086 * target.def (preferred_else_value): New target hook.
4087 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
4088 * doc/tm.texi: Regenerate.
4089 * targhooks.h (default_preferred_else_value): Declare.
4090 * targhooks.c (default_preferred_else_value): New function.
4091 * internal-fn.h (conditional_internal_fn_code): Declare.
4092 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
4093 (get_conditional_internal_fn): Use it.
4094 (conditional_internal_fn_code): New function.
4095 * gimple-match.h (gimple_match_cond): New struct.
4096 (gimple_match_op): Add a cond member function.
4097 (gimple_match_op::gimple_match_op): Update all forms to take a
4099 * genmatch.c (expr::gen_transform): Use the same condition as res_op
4100 for the suboperation, but don't specify a particular else_value.
4101 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
4102 (visit_nary_op, visit_reference_op_load): Pass
4103 gimple_match_cond::UNCOND to the gimple_match_op constructor.
4104 * gimple-match-head.c: Include tree-eh.h
4105 (convert_conditional_op): New function.
4106 (maybe_resimplify_conditional_op): Likewise.
4107 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
4108 (gimple_resimplify2): Likewise.
4109 (gimple_resimplify3): Likewise.
4110 (gimple_resimplify4): Likewise.
4111 (maybe_push_res_to_seq): Return null for conditional operations.
4112 (try_conditional_simplification): New function.
4113 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
4115 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
4117 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
4119 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
4121 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4123 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4125 (hash_tree): Do not hash DECL_FCONTEXT
4126 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
4127 Do not stream DECL_FCONTEXT.
4128 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
4129 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
4131 2018-07-12 Richard Biener <rguenther@suse.de>
4134 * dwarf2out.c (gen_block_die): Only output blocks when they have
4135 at least one !DECL_IGNORED_P variable.
4137 2018-07-12 Richard Biener <rguenther@suse.de>
4140 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
4141 Remove -mieee-fp handling.
4143 2018-07-12 Richard Biener <rguenther@suse.de>
4145 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
4146 left-over from last patch.
4148 2018-07-12 Jakub Jelinek <jakub@redhat.com>
4150 PR tree-optimization/86492
4151 * gimple-ssa-store-merging.c
4152 (imm_store_chain_info::coalesce_immediate_stores): Call
4153 check_no_overlap even for the merge_overlapping case. Formatting fix.
4155 2018-07-12 Richard Biener <rguenther@suse.de>
4158 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
4159 move possibly trapping operations into the conditional.
4161 2018-07-12 Richard Biener <rguenther@suse.de>
4163 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4164 (vn_lookup_simplify_result): Remove recursion limit applied
4166 (vn_nary_build_or_lookup_1): Adjust.
4167 (try_to_simplify): Likewise.
4168 * gimple-match-head.c (gimple_resimplify1): Instead apply one
4170 (gimple_resimplify2): Likewise.
4171 (gimple_resimplify3): Likewise.
4172 (gimple_resimplify4): Likewise.
4174 2018-07-11 Jakub Jelinek <jakub@redhat.com>
4176 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4177 Use __mmask64 type instead of __mmask8 for __M argument.
4178 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
4179 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
4181 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
4182 __mmask16 for __M argument.
4183 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
4184 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
4185 to __mmask16 instead of __mmask8.
4186 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
4187 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
4188 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
4189 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
4190 instead of __mmask16 for __U argument.
4191 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
4192 __mmask16 instead of __mmask8 for __U argument.
4193 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
4195 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
4197 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
4199 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
4201 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
4203 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
4205 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
4207 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
4209 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
4211 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
4213 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
4214 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
4215 return type as well as __M argument type and all casts from __mmask8
4217 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
4218 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
4219 return type as well as __M argument type and all casts from __mmask8
4221 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
4222 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
4223 return type as well as __M argument type and all casts from __mmask8
4225 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
4226 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
4227 return type as well as __M argument type and all casts from __mmask8
4229 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
4230 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
4233 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
4235 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
4236 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
4239 2018-07-11 Paul Koning <ni1d@arrl.net>
4241 * doc/md.texi (define_subst): Document how multiple occurrences of
4242 the same argument in the replacement pattern are handled.
4244 2018-07-11 Paul Koning <ni1d@arrl.net>
4246 * doc/extend.texi (Common Variable Attributes): Move "mode" into
4248 (Common Type Attributes): Add "mode" attribute.
4250 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
4252 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
4253 stream DECL_ORIGINAL_TYPE.
4254 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
4255 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
4256 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
4257 Do not walk original type.
4258 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
4260 (write_ts_decl_non_common_tree_pointers): Do not stream
4262 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4263 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4265 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4267 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4268 thread twice from the same starting edge.
4270 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4272 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4273 * gimple.c (gimple_call_nonnull_result_p): ...here...
4274 (gimple_call_nonnull_arg): ...and here.
4275 * gimple.h (gimple_call_nonnull_result_p): New.
4276 (gimple_call_nonnull_arg): New.
4278 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4280 * config/arm/arm-cpus.in: Move information from fpu field of each
4281 cpu definition to the isa field.
4282 * config/arm/parsecpu.awk (fpu): Delete match rule.
4283 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4285 2018-07-11 Richard Biener <rguenther@suse.de>
4288 * dwarf2out.c (init_sections_and_labels): Use
4289 output_asm_line_debug_info consistently.
4290 (dwarf2out_early_finish): Likewise.
4291 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4294 2018-07-11 Richard Biener <rguenther@suse.de>
4296 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4297 Rework father_bb setting in a way to avoid propagating constants
4298 multiple times on a loop body.
4300 2018-07-10 Mark Wielaard <mark@klomp.org>
4303 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4305 2018-07-10 Richard Biener <rguenther@suse.de>
4307 * hash-map.h (hash_map::iterator::operator*): Return
4308 references to key and value.
4310 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4313 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4314 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4315 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4316 (gimplify_omp_for): For composite loops, move outer
4317 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4318 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4319 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4320 TREE_LIST for both the original class iterator and the "last" helper
4321 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4322 loop, remember has_decl_expr from outer composite loops for the
4323 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4325 2018-07-09 Martin Sebor <msebor@redhat.com>
4329 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4330 accessing implicitly initialized array elements.
4331 * expr.c (string_constant): Handle string initializers of
4332 character arrays within aggregates.
4333 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4334 Store element offset. As a special case, handle zero size.
4335 (fold_nonarray_ctor_reference): Same.
4336 (fold_ctor_reference): Add argument. Store subobject offset.
4337 * gimple-fold.h (fold_ctor_reference): Add argument.
4339 2018-07-09 Paul Koning <ni1d@arrl.net>
4341 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4342 (pdp11_insn_cost): New function.
4343 (pdp11_md_asm_adjust): New function.
4344 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4345 (pdp11_rtx_costs): Update to match machine better.
4346 (output_addr_const_pdp11): Correct format mismatch warnings.
4347 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4348 * config/pdp11/pdp11.md: General change to add base_cost and/or
4349 length attributes for use by new pdp11_insn_cost function.
4350 (MIN_BRANCH): Correct definition.
4352 (doloop_end): Use standard pattern name for looping pattern.
4353 (doloop_end_nocc): New.
4354 (movsf): Add another constraint alternative.
4355 (zero_extendqihi2): Add constraint alternatives for not in place
4357 (zero_extendhisi2): Remove.
4358 (shift patterns): Add CC handling variants.
4362 (define_peephole2): New peephole to recognize mov that sets CC for
4365 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4367 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4368 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4371 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4374 * real.c (real_nextafter): Return true if result is denormal.
4376 2018-07-09 Martin Liska <mliska@suse.cz>
4378 * common.opt: Add back wrongly removed attribute.
4380 2018-07-09 Richard Biener <rguenther@suse.de>
4383 * dwarf2out.c (gen_block_die): For an early generated DIE
4384 always output high/low PC attributes.
4386 2018-07-09 Tom de Vries <tdevries@suse.de>
4388 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4389 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4390 onto VAR_DECL with abstract origin.
4392 2018-07-07 Jim Wilson <jimw@sifive.com>
4394 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4396 2018-07-07 Tom de Vries <tdevries@suse.de>
4398 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4400 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4402 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4404 (combine_bound): Use wide-int overflow calculation instead of
4406 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4408 * fold-const.c (int_const_binop_2): Same.
4409 (extract_muldiv_1): Same.
4410 (fold_div_compare): Same.
4411 (fold_abs_const): Same.
4413 * poly-int.h (add): Same.
4417 * predict.c (predict_iv_comparison): Same.
4418 * profile-count.c (slow_safe_scale_64bit): Same.
4419 * simplify-rtx.c (simplify_const_binary_operation): Same.
4420 * tree-chrec.c (tree_fold_binomial): Same.
4421 * tree-data-ref.c (split_constant_offset_1): Same.
4422 * tree-if-conv.c (idx_within_array_bound): Same.
4423 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4424 * tree-ssa-phiopt.c (minmax_replacement): Same.
4425 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4426 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4427 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4428 * wide-int.cc (wi::add_large): Same.
4429 (wi::mul_internal): Same.
4430 (wi::sub_large): Same.
4431 (wi::divmod_internal): Same.
4432 * wide-int.h: Change overflow type to overflow_type for neg, add,
4433 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4434 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4435 mul_internal, divmod_internal.
4436 (overflow_type): New enum.
4437 (accumulate_overflow): New.
4439 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4441 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4442 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4444 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4446 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4447 argument is checked for zero before entering loop, avoid checking again.
4449 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4451 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4452 * gimplify.h (generic_expr_could_trap_p): Likwise.
4453 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4455 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4457 PR tree-optimization/86401
4458 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4459 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4460 (fold_bit_and_mask): ... here. New helper function for match.pd.
4461 * fold-const.h (fold_bit_and_mask): Declare.
4462 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4464 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4467 * target.def (translate_mode_attribute): New hook.
4468 * targhooks.h (default_translate_mode_attribute): Declare.
4469 * targhooks.c (default_translate_mode_attribute): New function.
4470 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4471 * doc/tm.texi: Regenerate.
4472 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4473 (rs6000_translate_mode_attribute): New function.
4475 2018-07-06 Paul Koning <ni1d@arrl.net>
4477 * doc/md.texi (define_split): Document DONE and FAIL.
4478 (define_peephole2): Ditto.
4480 2018-07-05 Jeff Law <law@redhat.com>
4482 PR tree-optimization/86010
4483 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4484 both the head and tail of mem* and str* calls.
4486 2018-07-05 Jim Wilson <jimw@sifive.com>
4488 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4490 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4492 * config/aarch64/aarch64-simd.md: correct flags text for
4493 MIN_EXPR replacement.
4495 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4497 * configure: Regenerated.
4499 2018-07-05 Carl Love <cel@us.ibm.com>
4501 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4502 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4503 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4504 VSX_BUILTIN_DOUBLEL_V4SF.
4506 2018-07-05 Martin Sebor <msebor@redhat.com>
4509 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4510 than its domain to compute its the upper bound of a char array.
4512 2018-07-05 Nathan Sidwell <nathan@acm.org>
4514 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4515 * doc/cpp.texi: Update comment.
4516 * doc/tm.texi: Rebuilt.
4517 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4518 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4519 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4520 headers affected by extern "C".
4521 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4522 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4523 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4524 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4525 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4526 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4527 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4528 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4529 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4530 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4531 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4532 config/sparc/sp64-elf.h, config/spu/spu.h,
4533 config/stormy16/stormy16.h, config/v850/v850.h,
4534 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4535 define NO_IMPLICIT_EXTERN_C.
4536 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4538 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4541 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4542 instead of GET_MODE_SIZE when comparing Units.
4544 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4547 * rtlanal.c (set_noop_p): Constrain on mode change,
4548 include hard-reg-set.h
4550 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4552 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4554 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4557 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4560 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4562 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4564 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4567 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4569 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4571 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4574 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4576 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4578 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4579 neg_*_op* variables.
4581 2018-07-04 Martin Liska <mliska@suse.cz>
4583 * tree-switch-conversion.c: Define
4584 max_ratio_for_speed and max_ratio_for_size constants.
4586 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4587 Martin Liska <mliska@suse.cz>
4592 * common.opt: Rename align options with 'str_' prefix.
4593 * common/config/i386/i386-common.c (set_malign_value): New
4595 (ix86_handle_option): Use it to set -falign-* options/
4596 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4597 type from int to string.
4598 * config/aarch64/aarch64.c: Update default values from int
4600 * config/alpha/alpha.c (alpha_override_options_after_change):
4602 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4603 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4604 max skip conditionally.
4605 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4606 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4607 max skip conditionally.
4608 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4609 max skip conditionally.
4610 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4611 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4612 max skip conditionally.
4613 * config/i386/i386.c (struct ptt): Change type from int to
4615 (ix86_default_align): Set default values.
4616 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4617 max skip conditionally.
4618 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4619 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4620 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4621 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4622 max skip conditionally.
4623 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4624 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4625 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4626 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4627 max skip conditionally.
4628 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4629 * config/ia64/ia64.c (ia64_option_override): Set default values
4630 for alignment options.
4631 * config/m68k/m68k.c: Handle new str_align_* options.
4632 * config/mips/mips.c (mips_set_compression_mode): Change
4634 (mips_option_override): Set default values for options.
4635 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4637 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4639 * config/rx/rx.c (rx_option_override): Likewise.
4640 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4641 (LABEL_ALIGN): Use align_labels_log.
4642 (LOOP_ALIGN): Use align_loops_align.
4643 * config/s390/s390.c (s390_asm_output_function_label): Use new
4645 * config/sh/sh.c (sh_override_options_after_change):
4646 Change type of constants.
4647 * config/spu/spu.c (spu_sched_init): Likewise.
4648 * config/sparc/sparc.c (sparc_option_override): Set default
4650 * config/visium/visium.c (visium_option_override): Likewise.
4651 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4652 emit p2align format with last argument if it's not needed.
4653 * doc/invoke.texi: Document extended format of -falign-*.
4654 * final.c: Use align_labels alignment.
4655 * flags.h (struct target_flag_state): Change type to use
4657 (struct align_flags_tuple): New.
4658 (struct align_flags): Likewise.
4659 (align_loops_log): Redefine macro to use new types.
4660 (align_loops_max_skip): Redefine macro to use new types.
4661 (align_jumps_log): Redefine macro to use new types.
4662 (align_jumps_max_skip): Redefine macro to use new types.
4663 (align_labels_log): Redefine macro to use new types.
4664 (align_labels_max_skip): Redefine macro to use new types.
4665 (align_functions_log): Redefine macro to use new types.
4666 (align_loops): Redefine macro to use new types.
4667 (align_jumps): Redefine macro to use new types.
4668 (align_labels): Redefine macro to use new types.
4669 (align_functions): Redefine macro to use new types.
4670 (align_functions_max_skip): Redefine macro to use new types.
4671 (align_loops_value): New macro.
4672 (align_jumps_value): New macro.
4673 (align_labels_value): New macro.
4674 (align_functions_value): New macro.
4675 * function.c (invoke_set_current_function_hook): Propagate
4676 alignment values from flags to global variables default in
4678 * ipa-icf.c (sem_function::equals_wpa): Use
4679 cl_optimization_option_eq instead of memcmp.
4680 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4682 (cl_optimization_stream_in): Likewise.
4683 * optc-save-gen.awk: Support strings in cl_optimization.
4684 * opth-gen.awk: Likewise.
4685 * opts.c (finish_options): Remove error checking of invalid
4687 (MAX_CODE_ALIGN): Remove.
4688 (MAX_CODE_ALIGN_VALUE): Likewise.
4689 (parse_and_check_align_values): New function.
4690 (check_alignment_argument): Likewise.
4691 (common_handle_option): Use check_alignment_argument.
4692 * opts.h (parse_and_check_align_values): Declare.
4693 * toplev.c (init_alignments): Remove.
4694 (read_log_maxskip): New.
4695 (parse_N_M): Likewise.
4696 (parse_alignment_opts): Likewise.
4697 (backend_init_target): Remove usage of init_alignments.
4698 * toplev.h (parse_alignment_opts): Declare.
4699 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4701 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4702 * tree.c (cl_option_hasher::equal): New.
4703 * varasm.c: Use new global macros.
4705 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4707 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4708 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4709 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4710 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4711 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4712 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4713 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4714 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4715 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4716 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4717 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4718 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4719 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4721 2018-07-04 Martin Liska <mliska@suse.cz>
4722 Jonathan Wakely <jwakely@redhat.com>
4724 * coverage.c: Use correct type.
4725 * doc/invoke.texi: Language correction.
4727 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4730 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4731 ENDBRANCH for non-tail call which may return via indirect branch.
4732 * doc/extend.texi: Document indirect_return attribute.
4734 2018-07-03 Martin Sebor <msebor@redhat.com>
4736 PR tree-optimization/86274
4737 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4739 (format_floating): Correct handling of infinities and NaNs.
4741 2018-07-03 Martin Sebor <msebor@redhat.com>
4743 * print-tree.c (print_real_cst): New function.
4744 (print_node_brief): Call it.
4745 (print_node): Ditto.
4747 2018-07-03 Jeff Law <law@redhat.com>
4749 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4750 into a single pattern.
4752 * config/h8300/h8300.md (ors code_iterator): New.
4753 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4754 a single pattern and single splitter.
4755 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4756 (iorqi3_1, xorqi3_1): Likewise.
4757 (iorqi3, xorqi3 expanders): Similarly.
4759 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4760 (movmd_internal) into a single pattern using the P mode iterator.
4761 (movmd splitters): Similarly.
4762 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4763 (movsd splitters): Similarly.
4765 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4766 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4767 right table based on the size of the operand.
4768 * config/h8300/h8300.md (length_table): Corresponding changes. All
4769 references to "addb", "addw" and "addl" changed to "add".
4770 (btst patterns): Merge two variants into a single pattern.
4771 (tstqi, tsthi): Likewise.
4772 (addhi3_incdec, addsi3_incdec): Likewise.
4773 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4774 (mulhi3, mulsi3): Likewise.
4775 (udivhi3, udivsi3): Likewise.
4776 (divhi3, divsi3): Likewise.
4777 (andorqi3, andorhi3, andorsi3): Likewise.
4779 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4782 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4783 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4785 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4787 PR tree-optimization/85694
4788 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4789 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4791 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4792 (<u>avg<mode>3_ceil): New patterns.
4794 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4796 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4797 scan-tree-dump directive.
4798 * gcc.dg/vect/slp-perm-2.c: Likewise.
4799 * gcc.dg/vect/slp-perm-3.c: Likewise.
4800 * gcc.dg/vect/slp-perm-5.c: Likewise.
4801 * gcc.dg/vect/slp-perm-6.c: Likewise.
4802 * gcc.dg/vect/slp-perm-7.c: Likewise.
4803 * gcc.dg/vect/slp-perm-8.c: Likewise.
4805 2018-07-03 Marek Polacek <polacek@redhat.com>
4808 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4810 2018-07-03 Richard Biener <rguenther@suse.de>
4813 * tree-ssa-structalias.c (find_func_clobbers): Properly
4814 handle indirect calls.
4816 2018-07-03 Jeff Law <law@redhat.com>
4818 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4819 (shifts): New code iterator.
4820 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4821 expander. Fix HImode handling on H8/SX.
4822 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4823 (subqi3, subhi3, subsi3 expanders): Likewise.
4824 (andqi3, andhi3, andsi3 expanders): Likewise.
4825 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4826 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4827 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4828 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4829 (zero_extendqihi2, zero_extendqisi2): Likewise.
4830 (extendqihi2, extendqisi2): Likewise.
4831 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4832 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4833 (rotlqi3_1, rotlhi3_1): Likewise.
4834 (logicalhi3_sn, logicalsi3_sn): Likewise.
4835 (logicalhi3, logicalsi3): Likewise.
4837 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4839 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4840 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4841 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4842 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4843 type to append_pattern_def_seq instead of creating a stmt_vec_info
4845 (build_mask_conversion): Likewise. Remove vinfo argument.
4846 (vect_add_conversion_to_patterm): Likewise, renaming to...
4847 (vect_add_conversion_to_pattern): ...this.
4848 (vect_recog_mask_conversion_pattern): Update call to
4849 build_mask_conversion. Pass the vector type to
4850 append_pattern_def_seq here too.
4851 (vect_recog_gather_scatter_pattern): Update call to
4852 vect_add_conversion_to_pattern.
4854 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4856 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4857 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4858 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4859 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4860 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4861 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4862 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4863 append_pattern_def_seq instead of new_pattern_def_seq.
4864 (vect_recog_divmod_pattern): Do both of the above.
4865 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4868 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4870 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4871 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4872 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4873 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4874 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4875 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4876 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4877 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4878 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4879 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4880 parameter with a single stmt_vec_info.
4881 (vect_recog_func_ptr): Likewise.
4882 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4883 (vect_try_gather_scatter_pattern): ...this.
4884 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4885 the stmt_vec_info of the statement to be matched. Don't clear
4886 STMT_VINFO_RELATED_STMT.
4887 (vect_pattern_recog): Update call accordingly.
4889 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4891 PR tree-optimization/85694
4892 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4893 (uavgM3_ceil): Document new optabs.
4894 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4895 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4897 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4898 (savg_ceil_optab): New optabs.
4899 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4900 (vect_vect_recog_func_ptrs): Add it.
4901 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4902 constant directly from the associated lhs.
4904 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4906 * tree-vect-patterns.c (vect_split_statement): New function.
4907 (vect_convert_input): Use it to try to split an existing cast.
4909 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4911 * poly-int.h (print_hex): New function.
4912 * dumpfile.h (dump_dec, dump_hex): Declare.
4913 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4914 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4915 min_input_precision, operation_precision and operation_sign.
4916 * tree-vect-patterns.c (vect_get_range_info): New function.
4917 (vect_same_loop_or_bb_p, vect_single_imm_use)
4918 (vect_operation_fits_smaller_type): Delete.
4919 (vect_look_through_possible_promotion): Add an optional
4920 single_use_p parameter.
4921 (vect_recog_over_widening_pattern): Rewrite to use new
4922 stmt_vec_info infomration. Handle one operation at a time.
4923 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4924 (vect_truncatable_operation_p, vect_set_operation_type)
4925 (vect_set_min_input_precision): New functions.
4926 (vect_determine_min_output_precision_1): Likewise.
4927 (vect_determine_min_output_precision): Likewise.
4928 (vect_determine_precisions_from_range): Likewise.
4929 (vect_determine_precisions_from_users): Likewise.
4930 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4931 (vect_vect_recog_func_ptrs): Put over_widening first.
4933 (vect_pattern_recog): Call vect_determine_precisions.
4935 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4937 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4938 statements that have been replaced by further pattern statements.
4939 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4941 2018-07-03 Richard Biener <rguenther@suse.de>
4943 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4944 always set *dt. Dump vectype in vectype overload.
4945 * dumpfile.h (dump_gimple_expr): New function.
4946 (dump_gimple_expr_loc): Likewise.
4947 * dumpfile.c (dump_gimple_expr): New function.
4948 (dump_gimple_expr_loc): Likewise.
4950 2018-07-02 Jeff Law <law@redhat.com>
4952 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4953 the H8/300, H8/300H and H8/S variants into a single pattern.
4954 (movhi_h8300, movqi_h8300hs): Similarly.
4955 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4956 (QHI mode iterator): New.
4958 * config/h8300/h8300.md: Remove trailing whitespace.
4960 2018-07-02 Jim Wilson <jimw@sifive.com>
4962 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4963 instead of emit_insn for interrupt returns.
4964 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4965 (riscv_sret, riscv_uret): Likewise.
4967 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4969 * pretty-print.c (selftest::test_pp_format): Move save and restore
4970 of quotes to class auto_fix_quotes, and add an instance.
4971 * selftest.c: Include "intl.h".
4972 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4973 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4974 * selftest.h (selftest::auto_fix_quotes): New class.
4976 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4978 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4979 (aarch64_sve_prepare_conditional_op): Remove.
4980 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4981 Allow aarch64_simd_reg_or_zero as select operand; remove
4982 the aarch64_sve_prepare_conditional_op call.
4983 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4984 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4985 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4986 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4987 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4988 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4989 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4990 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4991 and a splitters to match all of the *_any patterns.
4992 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4994 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4995 (SVE_COND_FP_BINARY_REV): Remove.
4996 (sve_int_op_rev, sve_fp_op_rev): New.
4997 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4998 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4999 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
5000 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
5001 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
5002 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
5003 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
5004 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
5005 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
5007 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
5008 Remove match_dup 1 from the inner unspec.
5009 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5011 * config/aarch64/aarch64.md (movprfx): New attr.
5012 (length): Default movprfx to 8.
5013 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
5014 (*madd<SVE_I>, *msub<SVE_I): Likewise.
5015 (*<su>mul<SVE_I>3_highpart): Likewise.
5016 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
5017 (*v<ASHIFT><SVE_I>3): Likewise.
5018 (*<su><MAXMIN><SVE_I>3): Likewise.
5019 (*<su><MAXMIN><SVE_F>3): Likewise.
5020 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
5021 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
5022 (*div<SVE_F>4): Likewise.
5024 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
5026 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
5029 2018-07-02 Richard Biener <rguenther@suse.de>
5031 PR tree-optimization/86363
5032 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
5033 memset argument refers to a non-variable address.
5035 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5037 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
5038 {PLUS,MINUS}_EXPR code to...
5039 (adjust_symbolic_bound): ...here,
5040 (combine_bound): ...here,
5041 (set_value_range_with_overflow): ...and here.
5043 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5045 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
5047 (extract_range_from_abs_expr): ...here.
5049 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
5051 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
5052 -fno-omit-frame-pointer when not optimizing.
5054 2018-07-02 Martin Liska <mliska@suse.cz>
5057 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5058 (propagate_nothrow): Likewise.
5060 2018-07-02 Martin Liska <mliska@suse.cz>
5063 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
5065 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5067 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
5068 function in r262149, changing "loc" param from source_location to
5069 const dump_location_t &.
5070 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
5071 declaration, as above.
5073 2018-07-01 Paul Koning <ni1d@arrl.net>
5075 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5076 -munit-asm, -mgnu-asm, -mdec-asm.
5077 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
5078 (pdp11_output_labelref): New.
5079 (pdp11_output_def): New.
5080 (pdp11_output_addr_vec_elt): New.
5081 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
5082 %# and %@ format codes.
5083 (pdp11_option_override): New.
5084 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
5085 (pdp11_output_ident): New.
5086 (pdp11_asm_named_section): New.
5087 (pdp11_asm_init_sections): New.
5088 (pdp11_file_start): New.
5089 (pdp11_file_end): New.
5090 (output_ascii): Use .ascii/.asciz for -mdec-asm.
5091 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
5092 %o, like %c but octal.
5093 (pdp11_option_override): New.
5094 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
5096 (DATA_SECTION_ASM_OP): Ditto.
5097 (READONLY_DATA_SECTION_ASM_OP): New.
5098 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
5099 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
5100 (ASM_OUTPUT_LABELREF): Ditto.
5101 (ASM_OUTPUT_DEF): Ditto.
5102 (ASM_OUTPUT_EXTERNAL): New.
5103 (ASM_OUTPUT_SOURCE_FILENAME): New.
5104 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
5105 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
5106 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
5107 %# and %@ format codes.
5108 * config/pdp11/pdp11.opt (mgnu-asm): New.
5109 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
5110 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
5111 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
5113 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
5115 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
5116 dereferencing path[] beyond its length.
5118 (debug_all_paths): New.
5119 (rewire_first_differing_edge): New.
5120 (adjust_paths_after_duplication): New.
5121 (duplicate_thread_path): Call adjust_paths_after_duplication.
5123 (thread_through_all_blocks): Add new argument to
5124 duplicate_thread_path.
5126 2018-06-30 Jim Wilson <jimw@sifive.com>
5128 * config/riscv/predicates.md (p2m1_shift_operand): New.
5129 (high_mask_shift_operand): New.
5130 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
5131 pattern using p2m1_shift_operand.
5132 (lshsi3_zero_extend_3+2): New combiner pattern using
5133 high_mask_shift_operand.
5135 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5137 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
5139 (vect_recog_rotate_pattern): ...here.
5140 (vect_convert_input): Try to insert casts of invariants in the
5142 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
5143 preheader to be empty.
5145 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5147 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
5148 vector type. If given, install it in the new statement's
5150 (vect_element_precision): New function.
5151 (vect_unpromoted_value): New struct.
5152 (vect_unpromoted_value::vect_unpromoted_value): New function.
5153 (vect_unpromoted_value::set_op): Likewise.
5154 (vect_look_through_possible_promotion): Likewise.
5155 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
5156 (vect_widened_op_tree, vect_convert_input): Likewise.
5157 (vect_convert_inputs, vect_convert_output): Likewise.
5158 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
5159 to handle the optional cast of the multiplication result and
5160 vect_widened_op_tree to detect the widened multiplication itself.
5161 Do not require the input and output of promotion casts to have
5162 the same sign, but base the signedness of the operation on the
5163 input rather than the result. If the pattern includes two
5164 promotions, check that those promotions have the same sign.
5165 Do not restrict the MULT_EXPR handling to a double-width result;
5166 handle quadruple-width results and wider. Use vect_convert_inputs
5167 to convert the inputs to the common type.
5168 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
5169 to handle the optional cast of the ABS result. Also allow a sign
5170 change or a sign extension between the ABS and MINUS.
5171 Use vect_widened_op_tree to detect the widened subtraction and use
5172 vect_convert_inputs to convert the inputs to the common type.
5173 (vect_handle_widen_op_by_const): Delete.
5174 (vect_recog_widen_op_pattern): New function.
5175 (vect_recog_widen_mult_pattern): Use it.
5176 (vect_recog_widen_shift_pattern): Likewise.
5177 (vect_recog_widen_sum_pattern): Use
5178 vect_look_through_possible_promotion to handle the promoted
5181 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5183 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
5184 the containing gimple_seq *.
5185 * gimple-iterator.h (gsi_for_stmt): Declare it.
5186 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
5187 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
5188 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5189 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5190 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
5192 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
5194 (vect_mark_pattern_stmts): ...here. Handle cases in which the
5195 statement being replaced is part of an existing pattern
5196 definition sequence, inserting the new pattern statements before
5198 (vect_pattern_recog_1): Don't return a bool. If the statement
5199 is already part of a pattern, instead apply pattern matching
5200 to the pattern definition statements. Don't clear the
5201 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
5202 (vect_pattern_recog): Don't break after the first match;
5203 continue processing the pattern definition statements instead.
5204 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
5206 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5208 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5209 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5210 (vect_recog_widen_sum_pattern): Use it.
5212 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5214 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5215 phi is not a pattern statement and has not been replaced by
5216 a pattern statement.
5217 * tree-vect-patterns.c (type_conversion_p): Don't check
5218 STMT_VINFO_IN_PATTERN_P.
5219 (vect_recog_vector_vector_shift_pattern): Likewise.
5220 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
5221 the pattern statement rather than the original statement; check
5222 directly for a WIDEN_MULT_EXPR here.
5223 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
5224 vect_is_simple_use to return the pattern statement rather
5225 than the original statement; use is_pattern_stmt_p to check
5226 for such a pattern statement.
5227 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
5228 to return the pattern statement rather than the original statement;
5229 don't do the same transformation here.
5230 (vect_is_simple_use): If the defining statement has been replaced
5231 by a pattern statement, return the pattern statement instead.
5232 Remove the corresponding (local) transformation from the vectype
5235 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5237 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
5238 end and default to null.
5239 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5240 (vectorizable_reduction): Update calls accordingly, dropping the
5241 gimple ** argument if the passed-back statement isn't needed.
5242 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
5243 (vect_recog_rotate_pattern): Likewise.
5244 (vect_recog_mask_conversion_pattern): Likewise.
5245 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5246 (vect_mask_constant_operand_p): Likewise.
5247 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
5248 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
5249 (get_group_load_store_type, get_load_store_type): Likewise.
5250 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
5251 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
5252 (vectorizable_conversion, vectorizable_assignment): Likewise.
5253 (vectorizable_shift, vectorizable_operation): Likewise.
5254 (vectorizable_store, vect_is_simple_cond): Likewise.
5255 (vectorizable_condition, vectorizable_comparison): Likewise.
5256 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
5257 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
5258 and move it to the end. Cope with null def_stmt_outs.
5260 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5262 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5264 2018-06-29 Jeff Law <law@redhat.com>
5266 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5267 displacements for TARGET_V850E2V3 and newer.
5268 (TARGET_LRA_P): Remove. Defaults to LRA now.
5269 * config/v850/v850.md (sign23byte_load): Remove.
5270 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5271 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5273 2018-06-29 Martin Liska <mliska@suse.cz>
5276 * coverage.c (coverage_init): Mangle full path name.
5277 * doc/invoke.texi: Document the change.
5278 * gcov-io.c (mangle_path): New.
5279 * gcov-io.h (mangle_path): Likewise.
5280 * gcov.c (mangle_name): Use mangle_path for path mangling.
5282 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5284 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5285 if starting source register is not even.
5287 2018-06-29 Martin Liska <mliska@suse.cz>
5289 PR tree-optimization/86263
5290 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5291 Make edge redirection.
5293 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5295 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5296 (dump_scope_depth): New variable.
5297 (get_dump_scope_depth): New function.
5298 (dump_begin_scope): New function.
5299 (dump_end_scope): New function.
5300 * dumpfile.h (get_dump_scope_depth): New declaration.
5301 (dump_begin_scope): New declaration.
5302 (dump_end_scope): New declaration.
5303 (class auto_dump_scope): New class.
5304 (AUTO_DUMP_SCOPE): New macro.
5305 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5308 2018-06-29 Richard Biener <rguenther@suse.de>
5310 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5311 compute_all_dependences succeeds.
5312 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5313 exceed --param loop-max-datarefs-for-datadeps.
5315 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5317 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5319 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5322 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5323 alternative 0 in preferred_for_speed attribute.
5325 2018-06-28 Paul Koning <ni1d@arrl.net>
5327 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5328 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5329 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5330 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5331 for shift insn patterns.
5333 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5335 * cgraph.c (cgraph_node::get_body): Replace assignments to
5336 "dump_file" with calls to set_dump_file.
5337 * dumpfile.c (alt_dump_file): Make static, and group with...
5338 (alt_flags): ...this definition.
5339 (dumps_are_enabled): New variable.
5340 (refresh_dumps_are_enabled): New function.
5341 (set_dump_file): New function.
5342 (set_alt_dump_file): New function.
5343 (gcc::dump_manager::dump_start): Replace assignments to
5344 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5346 (gcc::dump_manager::dump_finish): Likewise.
5347 * dumpfile.h (alt_dump_file): Delete decl.
5348 (dumps_are_enabled): New variable decl.
5349 (set_dump_file): New function decl.
5350 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5352 * tree-nested.c (lower_nested_functions): Replace assignments to
5353 "dump_file" with calls to set_dump_file.
5355 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5357 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5358 goto_locus of each outgoing edge of each basic block.
5360 2018-06-28 Richard Biener <rguenther@suse.de>
5362 * dwarf2out.c (decl_scope_table): Remove.
5363 (push_decl_scope): Likewise.
5364 (pop_decl_scope): Likewise.
5365 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5366 (gen_struct_or_union_type_die): Likewise.
5367 (gen_tagged_type_die): Likewise.
5368 (dwarf2out_init): Do not initialize decl_scope_table.
5369 (dwarf2out_c_finalize): Do not free it.
5371 2018-06-28 Richard Biener <rguenther@suse.de>
5373 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5374 deciding whether to not re-use a DIE.
5376 2018-06-28 Richard Biener <rguenther@suse.de>
5378 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5379 DW_AT_abstract_origin attribute.
5381 2018-06-28 Martin Liska <mliska@suse.cz>
5383 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5384 Use newly introduced constants.
5385 * tree-switch-conversion.h (struct jump_table_cluster):
5386 Define max_ratio_for_size and max_ratio_for_speed.
5388 2018-06-28 Martin Liska <mliska@suse.cz>
5390 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5391 Add new checking assert to catch invalid state.
5392 (jump_table_cluster::can_be_handled): Handle single case
5394 (jump_table_cluster::is_beneficial): Bail out for such case.
5395 (bit_test_cluster::find_bit_tests):
5396 Add new checking assert to catch invalid state.
5397 (bit_test_cluster::can_be_handled): Handle single case
5399 (bit_test_cluster::is_beneficial): Bail out for such case.
5400 (switch_decision_tree::analyze_switch_statement):
5403 2018-06-28 Martin Liska <mliska@suse.cz>
5405 * common.opt: Introduce -completion option.
5406 * gcc.c (driver_handle_option): Handle it.
5407 (driver::main): Print completions if completion
5409 * opt-suggestions.c (option_proposer::get_completions):
5411 (option_proposer::suggest_completion): Likewise.
5412 (option_proposer::find_param_completions): Likewise.
5413 (verify_autocompletions): Likewise.
5414 (test_completion_valid_options): Likewise.
5415 (test_completion_valid_params): Likewise.
5416 (in_completion_p): Likewise.
5417 (empty_completion_p): Likewise.
5418 (test_completion_partial_match): Likewise.
5419 (test_completion_garbage): Likewise.
5420 (opt_proposer_c_tests): Likewise.
5421 * opt-suggestions.h: Declare new functions.
5422 * opts.c (common_handle_option): Handle OPT__completion_.
5423 * selftest-run-tests.c (selftest::run_tests): Add
5424 opt_proposer_c_tests.
5425 * selftest.c (assert_str_startswith): New.
5426 * selftest.h (assert_str_startswith): Likewise.
5427 (opt_proposer_c_tests): New.
5428 (ASSERT_STR_STARTSWITH): Likewise.
5430 2018-06-28 Martin Liska <mliska@suse.cz>
5432 * Makefile.in: Add opt-suggestions.o.
5433 * gcc-main.c: Include opt-suggestions.h.
5434 * gcc.c (driver::driver): Likewise.
5435 (driver::~driver): Remove m_option_suggestions.
5436 (driver::build_option_suggestions): Moved to option_proposer.
5437 (driver::suggest_option): Likewise.
5438 (driver::handle_unrecognized_options): Use option_proposer.
5439 * gcc.h (class driver): Add new memver m_option_proposer.
5440 * opt-suggestions.c: New file.
5441 * opt-suggestions.h: New file.
5443 2018-06-28 Martin Liska <mliska@suse.cz>
5445 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5446 (auto_string_vec::~auto_string_vec): Likewise.
5448 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5450 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5451 prevent_decl_creation_for_types fields up and add reset_location field.
5452 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5453 statement if id->reset_location is true.
5454 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5455 id->reset_location is true.
5456 (copy_phis_for_bb): Force input_location on the arguments if
5457 id->reset_location is true.
5458 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5459 is set on the function to be inlined.
5461 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5463 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5465 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5467 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5468 registers for Pmode.
5469 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5470 hard registers for the clobbered pseudo.
5472 2018-06-27 Paul Koning <ni1d@arrl.net>
5474 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5475 mutually exclusive options.
5476 * config/pdp11/constraints.md (h): New constraint.
5477 (O): Update definition to match shift code generation.
5478 (D): New constraint.
5479 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5481 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5483 (output_jump): Change arguments.
5484 (pdp11_fixed_cc_regs): New function.
5485 (pdp11_cc_mode): Ditto.
5486 (pdp11_expand_shift): Ditto.
5487 (pdp11_assemble_shift): Ditto.
5488 (pdp11_small_shift): Ditto.
5489 (pdp11_branch_cost): Remove.
5490 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5492 (pdp11_register_move_cost): Update for CC registers.
5493 (pdp11_rtx_costs): Add case for LSHIFTRT.
5494 (pdp11_output_jump): Add CCNZ mode conditional branches.
5495 (notice_update_cc_on_set): Remove.
5496 (pdp11_cc_mode): New function.
5497 (simple_memory_operand): Correct pre/post decrement case.
5498 (no_side_effect_operand): New function.
5499 (pdp11_regno_reg_class): Add CC_REGS class.
5500 (pdp11_fixed_cc_regs): New function.
5501 (pdp11_small_shift): New function.
5502 (pdp11_expand_shift): New function to expand shift insns.
5503 (pdp11_assemble_shift): New function to output shifts.
5504 (pdp11_branch_cost): Remove.
5505 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5506 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5507 (WCHAR_TYPE): Ditto.
5508 (PTRDIFF_TYPE): Ditto.
5509 (ADJUST_INSN_LENGTH): New macro.
5510 (FIXED_REGISTERS): Add CC registers.
5511 (CALL_USED_REGISTERS): Ditto.
5513 (REG_CLASS_NAMES): Ditto.
5514 (REG_CLASS_CONTENTS): Ditto.
5515 (SELECT_CC_MODE): Use new function.
5516 (TARGET_FLAGS_REGNUM): New macro.
5517 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5518 (cc0_reg_rtx): Remove.
5519 (CC_STATUS_MDEP): Remove.
5520 (CC_STATUS_MDEFP_INIT): Remove.
5521 (CC_IN_FPU): Remove.
5522 (NOTICE_UPDATE_CC): Remove.
5523 (REGISTER_NAMES): Add CC registers.
5524 (BRANCH_COST): Change to constant 1.
5525 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5527 * config/pdp11/pdp11.opt (mbcopy): Remove.
5528 (mbcopy-builtin): Remove.
5529 (mbranch-cheap): Remove.
5530 (mbranch-expensive): Remove.
5531 * config/pdp11/predicates.md (expand_shift_operand): Update to
5532 match shift code generation.
5533 (ccnz_operator): New predicate.
5534 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5535 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5536 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5538 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5539 description of O constraint.
5541 2018-06-27 Jeff Law <law@redhat.com>
5542 Austin Law <austinklaw@gmail.com>
5544 * config/v850/v850.md (addsi3_set_flags): New pattern.
5545 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5546 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5547 (zero_extendhisi2_v850_set_flags): Likewise.
5548 (zero_extendqisi2_v850_set_flags): Likewise.
5549 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5550 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5551 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5553 * config/v850/v850-protos.h (notice_update_cc): Remove.
5554 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5555 (v850_print_operand): Handle 'D' and "d".
5556 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5557 Add handling of arithmetic/logical operations compared against zero.
5558 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5559 Do not look at v850_compare_op, instead get mode from last argument.
5560 (v850_gen_compare): Remove
5561 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5562 after reload for prologue insns.
5563 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5565 (construct_save_jarl): Likewise.
5566 (TARGET_FLAGS_REGNUM): Define.
5567 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5568 (NOTICE_UPDATE_CC): Remove.
5569 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5570 than cc0. Conditionalize on reload_completed.
5571 (cmpsi_insn, setfcc_insn): Likewise.
5572 (tst1 splitter): Turn into define_and_split which sets the flags
5574 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5575 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5576 (cstoresf4, cstoredf4): Clobber the flags.
5577 (cmpsi, cmpsf, cmpdf): Remove expanders.
5578 (setf_insn): Remove pattern.
5579 (addsi3): Turn into define_and_split which clobbers the flags after
5580 reload and a suitable pattern (addsi3_clobber_flags) for use after
5582 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5583 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5584 (ashrsi3, ashrsi3_v850e2): Likewise.
5585 (bins): Clobber the flags.
5586 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5587 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5588 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5589 (call_value_internal_short, call_value_internal_long): Likewise.
5590 (callt_save_interrupt, callt_return_interrupt): Likewise.
5591 (save_interrupt, return_interrupt): Likewise.
5592 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5593 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5594 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5595 (All FP comparisons): Only allow after reload has completed.
5597 (divh, divhu): Tweak output template.
5598 (branch_z_normal, branch_z_invert): Remove
5599 (branch_nz_normal, branch_nz_invert): Likewise.
5600 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5602 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5603 * config/v850/v850.c (notice_update_cc): Remove.
5604 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5605 (CC_NO_CARRY): Likewise.
5606 (NOTICE_UPDATE_CC): Define to nothing.
5607 * config/v850/v850.md: Remove block comment on cc0 handling
5608 Remove "cc" attribute from all patterns. Remove cc_status handling
5609 from all patterns. Minor formatting fixes.
5611 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5613 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5614 (cortex-a76.cortex-a55): Likewise.
5615 * config/aarch64/aarch64-tune.md: Regenerate.
5616 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5617 cortex-a76.cortex-a55.
5619 2018-06-27 Jeff Law <law@redhat.com>
5621 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5622 (MULTILIB_DIRNAMES): Similarly.
5624 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5626 * gimple.h (gimple_return_retbnd): Delete.
5627 (gimple_return_set_retbnd): Likewise.
5628 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5629 gimple_return_set_retbnd.
5630 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5631 gimple_return_retbnd and adjust.
5632 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5633 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5634 Explicitly return NULL in a couple more cases. Move assertion
5635 on debug statements and remove unreachable code.
5636 (reset_debug_binding): Do not test id->retbnd.
5637 (expand_call_inline): Do not set it.
5639 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5641 * configure.ac: Add --disable-gcov option.
5642 * configure: Regenerate.
5643 * Makefile.in: Honour @enable_gcov@.
5644 * doc/install.texi: Document --disable-gcov.
5646 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5648 * config/arm/arm-cpus.in (cortex-a76): New entry.
5649 (cortex-a76.cortex-a55): Likewise.
5650 * config/arm/arm-tables.opt: Regenerate.
5651 * config/arm/arm-tune.md: Likewise.
5652 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5653 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5654 cortex-a76.cortex-a55.
5656 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5659 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5661 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5663 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5665 (EPILOGUE_USES): Likewise.
5667 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5669 * tree-inline.c (remap_location): New function extracted from...
5670 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5671 (copy_phis_for_bb): ...here. Call remap_location.
5672 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5674 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5676 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5677 unaligned vsx for 16B memset.
5679 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5682 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5683 ieee128_float_type_node to long_double_type_node unless
5684 TARGET_LONG_DOUBLE_128 is set.
5686 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5688 * cfgloop.c (get_loop_location): Convert return type from
5689 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5690 by implicit construction from rtx_insn *, and using
5691 dump_user_location_t::from_function_decl for the fallback case.
5692 * cfgloop.h (get_loop_location): Convert return type from
5693 location_t to dump_user_location_t.
5694 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5695 dump_printf_loc to pass in a dump_location_t rather than a
5696 location_t, via the gimple stmt.
5697 * coverage.c (get_coverage_counts): Update calls to
5698 dump_printf_loc to pass in dump_location_t rather than a
5700 * doc/optinfo.texi (Dump types): Convert example of
5701 dump_printf_loc from taking "locus" to taking "insn". Update
5702 description of the "_loc" calls to cover dump_location_t.
5703 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5705 (dump_user_location_t::dump_user_location_t): New constructors,
5706 from gimple *stmt and rtx_insn *.
5707 (dump_user_location_t::from_function_decl): New function.
5708 (dump_loc): Make static.
5709 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5710 const dump_location_t &.
5711 (dump_generic_expr_loc): Delete.
5712 (dump_printf_loc): Convert param "loc" from location_t to
5713 const dump_location_t &.
5714 (selftest::test_impl_location): New function.
5715 (selftest::dumpfile_c_tests): New function.
5716 * dumpfile.h: Include "profile-count.h".
5717 (class dump_user_location_t): New class.
5718 (struct dump_impl_location_t): New struct.
5719 (class dump_location_t): New class.
5720 (dump_printf_loc): Convert 2nd param from source_location to
5721 const dump_location_t &.
5722 (dump_generic_expr_loc): Delete.
5723 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5724 const dump_location_t &.
5725 * gimple-fold.c (fold_gimple_assign): Update call to
5726 dump_printf_loc to pass in a dump_location_t rather than a
5727 location_t, via the gimple stmt.
5728 (gimple_fold_call): Likewise.
5729 * gimple-loop-interchange.cc
5730 (loop_cand::analyze_iloop_reduction_var): Update for change to
5731 check_reduction_path.
5732 (tree_loop_interchange::interchange): Update for change to
5734 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5735 change in return-type of find_loop_location.
5736 (graphite_regenerate_ast_isl): Likewise.
5737 * graphite-optimize-isl.c (optimize_isl): Likewise.
5738 * graphite.c (graphite_transform_loops): Likewise.
5739 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5740 pass in a dump_location_t rather than a location_t, via the
5742 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5743 * ipa.c (walk_polymorphic_call_targets): Likewise.
5744 * loop-unroll.c (report_unroll): Convert "locus" param from
5745 location_t to dump_location_t.
5746 (decide_unrolling): Update for change to get_loop_location's
5748 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5749 location_t to dump_user_location_t.
5750 (grid_find_single_omp_among_assignments_1): Updates calls to
5751 dump_printf_loc to pass in a dump_location_t rather than a
5752 location_t, via the gimple stmt.
5753 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5754 location_t to dump_location_t. Updates calls to dump_printf_loc
5755 to pass in a dump_location_t rather than a location_t, via the
5757 (grid_inner_loop_gridifiable_p): Likewise.
5758 (grid_dist_follows_simple_pattern): Likewise.
5759 (grid_gfor_follows_tiling_pattern): Likewise.
5760 (grid_target_follows_gridifiable_pattern): Likewise.
5761 (grid_attempt_target_gridification): Convert initialization
5762 of local "grid" from memset to zero-initialization; FIXME: does
5763 this require C++11? Update call to dump_printf_loc to pass in a
5764 optinfo_location rather than a location_t, via the gimple stmt.
5765 * profile.c (read_profile_edge_counts): Updates call to
5766 dump_printf_loc to pass in a dump_location_t rather than a
5768 (compute_branch_probabilities): Likewise.
5769 * selftest-run-tests.c (selftest::run_tests): Call
5771 * selftest.h (dumpfile_c_tests): New decl.
5772 * tree-loop-distribution.c (pass_loop_distribution::execute):
5773 Update for change in return type of find_loop_location.
5774 * tree-parloops.c (parallelize_loops): Likewise.
5775 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5776 "locus" from location_t to dump_user_location_t.
5777 (canonicalize_loop_induction_variables): Likewise.
5778 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5779 for change in return type of find_loop_location.
5780 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5781 to dump_printf_loc to pass in a dump_location_t rather than a
5782 location_t, via the stmt.
5783 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5785 * tree-vect-loop-manip.c (find_loop_location): Convert return
5786 type from source_location to dump_user_location_t.
5787 (vect_do_peeling): Update for above change.
5788 (vect_loop_versioning): Update for change in type of
5790 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5791 from location_t to dump_user_location_t.
5792 (vect_estimate_min_profitable_iters): Update for change in type
5794 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5795 location_t to dump_location_t.
5796 (vect_slp_bb): Update for change in type of vect_location.
5797 * tree-vectorizer.c (vect_location): Convert from source_location
5798 to dump_user_location_t.
5799 (try_vectorize_loop_1): Update for change in vect_location's type.
5800 (vectorize_loops): Likewise.
5801 (increase_alignment): Likewise.
5802 * tree-vectorizer.h (vect_location): Convert from source_location
5803 to dump_user_location_t.
5804 (find_loop_location): Convert return type from source_location to
5805 dump_user_location_t.
5806 (check_reduction_path): Convert 1st param from location_t to
5807 dump_user_location_t.
5808 * value-prof.c (check_counter): Update call to dump_printf_loc to
5809 pass in a dump_user_location_t rather than a location_t; update
5810 call to error_at for change in type of "locus".
5811 (check_ic_target): Update call to dump_printf_loc to
5812 pass in a dump_user_location_t rather than a location_t, via the
5815 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5817 * config/s390/s390.h (enum processor_flags): Do not use
5818 default tune parameter when -march was specified.
5820 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5823 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5824 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5827 2018-06-26 Richard Biener <rguenther@suse.de>
5829 PR tree-optimization/86287
5831 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5832 (vect_analyze_loop): Initialize n_stmts.
5834 2018-06-26 Richard Biener <rguenther@suse.de>
5837 * fold-const.c (fold_convertible_p): Pointer extension
5840 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5843 * dwarf2out.c (loc_list_has_views): Adjust comments.
5844 (dw_loc_list): Split single cross-partition range with
5847 2018-06-25 Jeff Law <law@redhat.com>
5849 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5850 on -mbig-switch by default.
5852 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5854 (const_float_0_operand): Remove unused predicate.
5855 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5856 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5857 (recipsf2): New expander. Original pattern now called
5859 (recipdf2, recipdf2_insn): Similarly.
5860 (rsqrtsf2, rsqrtsf2_insn): Similarly
5861 (rsqrtdf2, rsqrtdf2_insn): Similarly
5863 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5865 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5866 Simplify logic for FreeBSD (twice).
5868 2018-06-25 Martin Sebor <msebor@redhat.com>
5870 PR tree-optimization/86204
5871 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5872 a strnlen result if it's less than the length of the string.
5874 2018-06-25 Martin Sebor <msebor@redhat.com>
5876 PR tree-optimization/85700
5877 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5878 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5879 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5881 2018-06-25 Martin Sebor <msebor@redhat.com>
5883 * doc/extend.texi (Zero-length arrays): Update and clarify.
5885 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5887 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5888 added IEEE/IBM long double multilib support on PowerPC little
5889 endian Linux systems.
5890 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5891 (MULTILIB_DEFAULTS): Likewise.
5892 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5894 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5895 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5896 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5898 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5901 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5902 (REORDER_45): Likewise.
5904 2018-06-25 Jeff Law <law@redhat.com>
5906 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5907 dividend to 32 bits. Adjust length.
5908 (udivmodhi4): Cleanup output template. Fix length.
5910 2018-06-25 Carl Love <cel@us.ibm.com>
5912 * config/rs6000/vsx.md: Change word selector to prefered location.
5914 2018-06-25 Richard Biener <rguenther@suse.de>
5916 PR tree-optimization/86304
5917 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5918 epilogue-if-converted loops as well.
5920 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5922 * lto-section-out.c (lto_begin_section): Do not print section
5923 name for noaddr and unnumbered dumps.
5925 2018-06-25 Richard Biener <rguenther@suse.de>
5927 * tree-vectorizer.h (struct vec_info_shared): New structure
5928 with parts split out from struct vec_info and loop_nest from
5929 struct _loop_vec_info.
5930 (struct vec_info): Adjust accordingly.
5931 (struct _loop_vec_info): Likewise.
5932 (LOOP_VINFO_LOOP_NEST): Adjust.
5933 (LOOP_VINFO_DATAREFS): Likewise.
5934 (LOOP_VINFO_DDRS): Likewise.
5935 (struct _bb_vec_info): Likewise.
5936 (BB_VINFO_DATAREFS): Likewise.
5937 (BB_VINFO_DDRS): Likewise.
5938 (struct _stmt_vec_info): Add dr_aux member.
5939 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5940 (DR_MISALIGNMENT_UNINITIALIZED): New.
5941 (set_dr_misalignment): Adjust.
5942 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5943 (vect_analyze_loop): Adjust prototype.
5944 (vect_analyze_loop_form): Likewise.
5945 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5946 Compute dependences lazily.
5947 (vect_record_base_alignments): Use shared datarefs/ddrs.
5948 (vect_verify_datarefs_alignment): Likewise.
5949 (vect_analyze_data_refs_alignment): Likewise.
5950 (vect_analyze_data_ref_accesses): Likewise.
5951 (vect_analyze_data_refs): Likewise.
5952 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5953 constructor parameter for shared part.
5954 (vect_analyze_loop_form): Pass in shared part and adjust.
5955 (vect_analyze_loop_2): Pass in storage for the number of
5956 stmts. Move loop nest finding to the caller. Compute
5958 (vect_analyze_loop): Pass in shared part.
5959 (vect_transform_loop): Verify shared datarefs are unchanged.
5960 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5961 constructor parameter for shared part.
5962 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5963 (vect_slp_bb): Verify shared datarefs are unchanged before
5965 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5967 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5968 DR_MISALIGNMENT_UNINITIALIZED.
5969 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5970 parameter for shared part.
5971 (vec_info::~vec_info): Adjust.
5972 (vec_info_shared::vec_info_shared): New.
5973 (vec_info_shared::~vec_info_shared): Likewise.
5974 (vec_info_shared::save_datarefs): Likewise.
5975 (vec_info_shared::check_datarefs): Likewise.
5976 (try_vectorize_loop_1): Construct shared part live for analyses
5977 of a single loop for multiple vector sizes.
5978 * tree-parloops.c (gather_scalar_reductions): Adjust.
5980 2018-06-25 Richard Biener <rguenther@suse.de>
5982 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5983 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5984 (vect_analyze_data_refs): Remove similar code from here and
5985 simplify accordingly.
5987 2018-06-25 Richard Biener <rguenther@suse.de>
5989 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5990 for reverse storage order accesses rather than asserting
5991 they cannot happen here.
5993 2018-06-25 Tom de Vries <tdevries@suse.de>
5996 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5997 Use data16 instead of .byte for insn prefix.
5999 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
6002 * parser.c (make_char_string_pack): Pass this literal chars
6003 through cpp_interpret_string.
6004 (cp_parser_userdef_numeric_literal): Check the result of
6005 make_char_string_pack.
6007 2018-06-24 Maya Rashish <coypu@sdf.org>
6009 * ginclude/stddef.h: Simplify conditions around avoiding
6010 re-definition of __size_t.
6012 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6014 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
6015 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6017 2018-06-22 Maya Rashish <coypu@sdf.org>
6019 * doc/invoke.texi (mno-fancy-math-387): Update for changes
6020 made to OpenBSD and NetBSD through the years.
6022 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6024 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6025 behavior of vec_pack (vector double, vector double) to match
6026 behavior of vec_float2 (vector double, vector double).
6028 2018-06-22 Olivier Hainque <hainque@adacore.com>
6030 * gimplify.c (gimplify_function_tree): Prevent creation
6031 of a trampoline for the address of the current function
6032 passed to entry/exit instrumentation hooks.
6034 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
6037 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
6040 2018-06-22 Martin Liska <mliska@suse.cz>
6042 PR tree-optimization/86263
6043 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6044 Bail out if is_enabled is false.
6045 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
6047 (jump_table_cluster::is_enabled): New function.
6049 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6051 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6052 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
6053 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
6054 (lto_input_ts_binfo_tree_pointers): Likewise.
6055 * tree-streamer-out.c (streamer_write_tree_bitfields,
6056 write_ts_binfo_tree_pointers): Likewise.
6057 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
6059 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6061 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
6063 2018-06-22 Martin Liska <mliska@suse.cz>
6065 * symbol-summary.h (get): Make it pure and inline move
6066 functionality from ::get function.
6067 (get): Remove and inline into ::get and ::get_create.
6068 (get_create): Move code from ::get function.
6070 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6073 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
6074 -x assembler-with-cpp.
6076 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6078 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
6079 _FILE_OFFSET_BITS=64 for C++.
6081 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
6083 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
6084 conversion insn that shows up when pr85657-3.c is compiled using
6085 IEEE 128-bit long double.
6086 (neg<mode>2_internal): Use the correct mode to check whether the
6087 mode is IBM extended.
6088 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
6089 multiply and divide external functions from being created more
6092 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6094 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
6096 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
6097 the edge can be forwarded.
6098 (cfg_layout_merge_blocks): Likewise.
6100 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6102 * except.c (finish_eh_generation): Commit edge insertions only after
6103 the EH edges have been redirected from post-landing to landing pads.
6105 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6107 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
6108 create_tmp_var_for to create the FRAME decl.
6109 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
6111 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6113 * tree-inline.c (copy_edges_for_bb): Minor tweak.
6114 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
6115 debug statement when resetting its value.
6116 (expand_call_inline): Copy the locus of the call onto the assignment
6117 of the return value, if any. Use local variable in more cases.
6119 2018-06-21 Martin Liska <mliska@suse.cz>
6121 * ipa-pure-const.c (propagate_nothrow): Use
6122 funct_state_summaries->get.
6123 (dump_malloc_lattice): Likewise.
6124 (propagate_malloc): Likewise.
6126 2018-06-21 Richard Biener <rguenther@suse.de>
6128 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
6129 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
6130 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
6132 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
6133 BLOCK_ABSTRACT_ORIGIN unconditionally.
6135 2018-06-21 David Malcolm <dmalcolm@redhat.com>
6137 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
6139 * ipa-reference.c (ipa_reference_c_finalize): Delete
6140 ipa_ref_opt_sum_summaries and set it to NULL.
6142 2018-06-21 Tom de Vries <tdevries@suse.de>
6144 PR tree-optimization/85859
6145 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
6146 test with comment from bb_no_side_effects_p.
6148 2018-06-21 Richard Biener <rguenther@suse.de>
6150 PR tree-optimization/86232
6151 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
6152 max for constant niter.
6154 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6156 * config/aarch64/aarch64-simd.md
6157 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
6159 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6161 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
6162 Make opernads of the unspec commutative.
6164 2018-06-21 Richard Biener <rguenther@suse.de>
6166 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
6167 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
6168 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6169 (vect_analyze_data_ref_dependence): Re-order checks to deal with
6171 (vect_record_base_alignments): Do not record base alignment
6172 for gathers or scatters.
6173 (vect_compute_data_ref_alignment): Drop return value that is always
6174 true. Bail out early for gathers or scatters.
6175 (vect_enhance_data_refs_alignment): Bail out early for gathers
6177 (vect_find_same_alignment_drs): Likewise.
6178 (vect_analyze_data_refs_alignment): Remove dead code.
6179 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6180 (vect_analyze_data_refs): For possible gathers or scatters do
6181 not create an alternate DR, just check their possible validity
6182 and mark them. Adjust DECL_NONALIASED handling to not rely
6184 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
6185 update inits of gathers or scatters.
6186 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
6187 Also copy gather/scatter flag to pattern vinfo.
6189 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
6191 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6192 behavior of vec_packsu (vector unsigned long long, vector unsigned
6193 long long) to match behavior of vec_packs with same signature.
6195 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
6196 Thomas Schwinge <thomas@codesourcery.com>
6197 Cesar Philippidis <cesar@codesourcery.com>
6199 * gimplify.c (gimplify_scan_omp_clauses): Add support for
6200 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6201 (gimplify_adjust_omp_clauses): Likewise.
6202 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
6203 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
6204 (gimplify_omp_target_update): Update handling of acc update and
6206 * omp-low.c (install_var_field): Remove unused parameter
6207 base_pointers_restrict.
6208 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
6209 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
6211 (omp_target_base_pointers_restrict_p): Delete.
6212 (scan_omp_target): Update call to scan_sharing_clauses.
6213 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
6215 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
6216 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6217 (convert_local_omp_clauses): Likewise.
6218 * tree-pretty-print.c (dump_omp_clause): Likewise.
6219 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
6221 (omp_clause_code_name): Likewise.
6223 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6226 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
6229 PR tree-optimization/86231
6230 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
6231 anti-range don't overwrite *vr0min before using it to compute *vr0max.
6233 2018-06-20 Tom de Vries <tdevries@suse.de>
6235 PR tree-optimization/86097
6236 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
6237 iv type if signedness of iv type is not the same as that of *nit.
6239 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6241 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
6242 EDGE_EH edges, verify they are all EDGE_EH.
6244 2018-06-20 Maya Rashish <coypu@sdf.org>
6246 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
6248 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6250 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
6251 * config/aarch64/aarch64.c (xgene1_tunings): Add
6252 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
6253 (aarch64_mode_valid_for_sched_fusion_p):
6254 Allow 16-byte modes.
6255 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
6256 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
6258 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
6260 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6261 * config/aarch64/iterators.md (VQ2): New mode iterator.
6263 2018-06-20 Martin Liska <mliska@suse.cz>
6265 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6266 Change default ratio from 10 to 8.
6268 2018-06-20 Martin Liska <mliska@suse.cz>
6270 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6272 (bit_test_cluster::find_bit_tests): Likewise.
6273 (switch_decision_tree::analyze_switch_statement): Find clusters.
6274 * tree-switch-conversion.h (struct jump_table_cluster): Document
6277 2018-06-20 Martin Liska <mliska@suse.cz>
6279 * tree-switch-conversion.c (switch_conversion::collect):
6280 Record m_uniq property.
6281 (switch_conversion::expand): Bail out for special conditions.
6282 (group_cluster::~group_cluster): New.
6283 (group_cluster::group_cluster): Likewise.
6284 (group_cluster::dump): Likewise.
6285 (jump_table_cluster::emit): New.
6286 (switch_decision_tree::fix_phi_operands_for_edges): New.
6287 (struct case_node): Remove struct.
6288 (jump_table_cluster::can_be_handled): New.
6289 (case_values_threshold): Moved to header.
6290 (reset_out_edges_aux): Likewise.
6291 (jump_table_cluster::is_beneficial): New.
6292 (bit_test_cluster::can_be_handled): Likewise.
6293 (add_case_node): Remove.
6294 (bit_test_cluster::is_beneficial): New.
6295 (case_bit_test::cmp): New.
6296 (bit_test_cluster::emit): New.
6297 (expand_switch_as_decision_tree_p): Remove.
6298 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6299 (fix_phi_operands_for_edge): Likewise.
6300 (switch_decision_tree::analyze_switch_statement): New.
6301 (compute_cases_per_edge): Move ...
6302 (switch_decision_tree::compute_cases_per_edge): ... here.
6303 (try_switch_expansion): Likewise.
6304 (switch_decision_tree::try_switch_expansion): Likewise.
6305 (record_phi_operand_mapping): Likewise.
6306 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6307 (emit_case_decision_tree): Likewise.
6308 (switch_decision_tree::emit): Likewise.
6309 (balance_case_nodes): Likewise.
6310 (switch_decision_tree::balance_case_nodes): Likewise.
6311 (dump_case_nodes): Likewise.
6312 (switch_decision_tree::dump_case_nodes): Likewise.
6313 (emit_jump): Likewise.
6314 (switch_decision_tree::emit_jump): Likewise.
6315 (emit_cmp_and_jump_insns): Likewise.
6316 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6317 (emit_case_nodes): Likewise.
6318 (switch_decision_tree::emit_case_nodes): Likewise.
6319 (conditional_probability): Remove.
6320 * tree-switch-conversion.h (enum cluster_type): New.
6322 (struct cluster): Likewise.
6323 (cluster::cluster): Likewise.
6324 (struct simple_cluster): Likewise.
6325 (simple_cluster::simple_cluster): Likewise.
6326 (struct group_cluster): Likewise.
6327 (struct jump_table_cluster): Likewise.
6328 (struct bit_test_cluster): Likewise.
6329 (struct min_cluster_item): Likewise.
6330 (struct case_tree_node): Likewise.
6331 (case_tree_node::case_tree_node): Likewise.
6332 (jump_table_cluster::case_values_threshold): Likewise.
6333 (struct case_bit_test): Likewise.
6334 (struct switch_decision_tree): Likewise.
6335 (struct switch_conversion): Likewise.
6336 (switch_decision_tree::reset_out_edges_aux): Likewise.
6338 2018-06-20 Martin Liska <mliska@suse.cz>
6340 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6341 (hoist_edge_and_branch_if_true): Likewise.
6342 (expand_switch_using_bit_tests_p): Likewise.
6343 (struct case_bit_test): Likewise.
6344 (case_bit_test_cmp): Likewise.
6345 (emit_case_bit_tests): Likewise.
6346 (switch_conversion::switch_conversion): New class.
6347 (struct switch_conv_info): Remove old struct.
6348 (collect_switch_conv_info): More to ...
6349 (switch_conversion::collect): ... this.
6350 (check_range): Likewise.
6351 (switch_conversion::check_range): Likewise.
6352 (check_all_empty_except_final): Likewise.
6353 (switch_conversion::check_all_empty_except_final): Likewise.
6354 (check_final_bb): Likewise.
6355 (switch_conversion::check_final_bb): Likewise.
6356 (create_temp_arrays): Likewise.
6357 (switch_conversion::create_temp_arrays): Likewise.
6358 (free_temp_arrays): Likewise.
6359 (gather_default_values): Likewise.
6360 (switch_conversion::gather_default_values): Likewise.
6361 (build_constructors): Likewise.
6362 (switch_conversion::build_constructors): Likewise.
6363 (constructor_contains_same_values_p): Likewise.
6364 (switch_conversion::contains_same_values_p): Likewise.
6365 (array_value_type): Likewise.
6366 (switch_conversion::array_value_type): Likewise.
6367 (build_one_array): Likewise.
6368 (switch_conversion::build_one_array): Likewise.
6369 (build_arrays): Likewise.
6370 (switch_conversion::build_arrays): Likewise.
6371 (gen_def_assigns): Likewise.
6372 (switch_conversion::gen_def_assigns): Likewise.
6373 (prune_bbs): Likewise.
6374 (switch_conversion::prune_bbs): Likewise.
6375 (fix_phi_nodes): Likewise.
6376 (switch_conversion::fix_phi_nodes): Likewise.
6377 (gen_inbound_check): Likewise.
6378 (switch_conversion::gen_inbound_check): Likewise.
6379 (process_switch): Use the newly created class.
6380 (switch_conversion::expand): New.
6381 (switch_conversion::~switch_conversion): New.
6382 * tree-switch-conversion.h: New file.
6384 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6386 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6387 tree-vect-patterns.c.
6388 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6389 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6390 (vect_recog_sad_pattern): Likewise.
6391 (vect_recog_widen_sum_pattern): Likewise.
6392 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6393 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6394 (vect_recog_rotate_pattern): Likewise.
6395 (vect_recog_mult_pattern): Likewise.
6396 (vect_recog_vector_vector_shift_pattern): Likewise.
6397 (vect_recog_divmod_pattern): Likewise.
6398 (vect_recog_mixed_size_cond_pattern): Likewise.
6399 (vect_recog_bool_pattern): Likewise.
6400 (vect_recog_mask_conversion_pattern): Likewise.
6401 (vect_try_gather_scatter_pattern): Likewise.
6402 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6403 (vect_recog_over_widening_pattern): Likewise.
6404 (vect_recog_gather_scatter_pattern): Likewise.
6405 (vect_recog_func_ptr): Move from tree-vectorizer.h
6406 (vect_vect_recog_func_ptrs): Move further down the file.
6407 (vect_recog_func): Likewise. Remove the third argument.
6408 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6409 (vect_pattern_recog_1): Expect the pattern function to do any
6410 necessary target tests. Also expect it to provide a vector type.
6411 Remove the type_in handling.
6413 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6415 * tree-vect-patterns.c (vect_pattern_detected): New function.
6416 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6417 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6418 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6419 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6420 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6421 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6422 (vect_recog_mask_conversion_pattern)
6423 (vect_try_gather_scatter_pattern): Likewise.
6425 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6427 * tree-vect-patterns.c (vect_get_internal_def): New function.
6428 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6429 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6430 (search_type_for_mask_1): Use it.
6432 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6434 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6435 redundant WIDEN_SUM_EXPR handling.
6436 (vect_recog_sad_pattern): Likewise.
6438 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6440 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6441 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6442 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6443 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6444 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6446 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6448 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6449 use the stmt_vec_info of the original bb statement for the
6450 new zero assignment, even if the call is part of a pattern.
6452 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6454 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6455 that the sequence is attached to the original statement rather
6456 than the pattern statement.
6457 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6458 PATTERN_DEF_SEQ from the original statement rather than
6459 the main pattern statement.
6460 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6461 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6462 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6464 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6466 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6467 definition statements before the early exit for statements that aren't
6469 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6471 (vect_transform_loop): ...here. Process pattern definition
6472 statements without first checking whether the main pattern
6473 statement is live or relevant.
6475 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6477 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6478 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6480 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6482 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6483 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6484 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6485 (expand_block_compare): Change select_block_compare_mode call.
6486 (expand_strncmp_align_check): Use new functions, fix comment.
6487 (emit_final_str_compare_gpr): New function.
6488 (expand_strn_compare): Refactor and clean up code.
6489 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6491 2018-06-19 Tony Reix <tony.reix@atos.com>
6492 Damien Bergamini <damien.bergamini@atos.com>
6493 David Edelsohn <dje.gcc@gmail.com>
6495 * collect2.c (static_obj): New variable.
6496 (static_libs): New variable.
6497 (is_in_list): Uncomment declaration.
6498 (main): Track AIX libraries linked statically.
6499 (is_in_list): Uncomment definition.
6500 (scan_prog_file): Don't add AIX shared libraries initializer
6501 to constructor list if linking statically.
6503 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6505 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6507 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6509 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6511 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6514 2018-06-19 Martin Liska <mliska@suse.cz>
6516 * config/i386/i386.c (ix86_can_inline_p): Do not use
6517 ipa_fn_summaries::get_create.
6518 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6520 (devirtualization_time_bonus): Likewise.
6521 (ipcp_propagate_stage): Likewise.
6522 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6523 (edge_set_predicate): Likewise.
6524 (evaluate_conditions_for_known_args): Likewise.
6525 (evaluate_properties_for_edge): Likewise.
6526 (ipa_call_summary::reset): Tranform to ...
6527 (ipa_call_summary::~ipa_call_summary): ... this.
6528 (ipa_fn_summary::reset): Transform to ...
6529 (ipa_fn_summary::~ipa_fn_summary): ... this.
6530 (ipa_fn_summary_t::remove): Rename to ...
6531 (ipa_fn_summary_t::remove_callees): ... this.
6532 (ipa_fn_summary_t::duplicate): Use placement new
6533 instead of memory copy.
6534 (ipa_call_summary_t::duplicate): Likewise.
6535 (ipa_call_summary_t::remove): Remove.
6536 (dump_ipa_call_summary): Change get_create to get.
6537 (ipa_dump_fn_summary): Dump only when summary exists.
6538 (analyze_function_body): Use symbol_summary::get instead
6540 (compute_fn_summary): Likewise.
6541 (estimate_edge_devirt_benefit): Likewise.
6542 (estimate_edge_size_and_time): Likewise.
6543 (inline_update_callee_summaries): Likewise.
6544 (remap_edge_change_prob): Likewise.
6545 (remap_edge_summaries): Likewise.
6546 (ipa_merge_fn_summary_after_inlining): Likewise.
6547 (write_ipa_call_summary): Likewise.
6548 (ipa_fn_summary_write): Likewise.
6549 (ipa_free_fn_summary): Likewise.
6550 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6551 (struct ipa_call_summary): Likewise.
6552 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6554 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6555 (estimate_size_after_inlining): Likewise.
6556 (estimate_growth): Likewise.
6557 (growth_likely_positive): Likewise.
6558 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6559 (inline_call): Likewise.
6560 * ipa-inline.c (caller_growth_limits): Likewise.
6561 (can_inline_edge_p): Likewise.
6562 (can_inline_edge_by_limits_p): Likewise.
6563 (compute_uninlined_call_time): Likewise.
6564 (compute_inlined_call_time): Likewise.
6565 (want_inline_small_function_p): Likewise.
6566 (edge_badness): Likewise.
6567 (update_caller_keys): Likewise.
6568 (update_callee_keys): Likewise.
6569 (inline_small_functions): Likewise.
6570 (inline_to_all_callers_1): Likewise.
6571 (dump_overall_stats): Likewise.
6572 (early_inline_small_functions): Likewise.
6573 (early_inliner): Likewise.
6574 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6575 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6576 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6577 * ipa-split.c (execute_split_functions): Likewise.
6578 * symbol-summary.h: Likewise.
6579 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6581 2018-06-19 Richard Biener <rguenther@suse.de>
6583 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6584 (vectorize_loops): ... here. Fix dbgcnt handling.
6585 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6587 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6590 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6591 ieee128 argument takes up only one (vector) register, not two (floating
6594 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6596 * gimplify.c (gimplify_init_constructor): Really never clear for an
6597 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6599 2018-06-19 Richard Biener <rguenther@suse.de>
6601 PR tree-optimization/86179
6602 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6603 after failed recognition.
6605 2018-06-18 Martin Sebor <msebor@redhat.com>
6608 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6609 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6610 Handle integer subtraction.
6611 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6612 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6614 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6616 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6617 param from rtx to rtx_insn *.
6618 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6620 (frv_ifcvt_modify_insn): Likwise.
6621 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6622 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6623 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6624 as_a <rtx_insn *> cast to local "unprotected_region" once
6625 it's been established that it's not NULL or pc_rtx.
6626 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6627 param "sethi" from rtx to rtx_insn *.
6628 (nds32_group_float_insns): Likewise for param "insn".
6629 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6631 (vax_output_int_subtract): Likewise.
6632 * config/vax/vax.c (vax_output_int_add): Likewise for param
6634 (vax_output_int_subtract): Likewise.
6635 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6636 (emit_pattern_after): Likewise for param "after".
6637 (emit_insn_after): Likewise.
6638 (emit_jump_insn_after): Likewise.
6639 (emit_call_insn_after): Likewise.
6640 (emit_debug_insn_after): Likewise.
6641 (emit_pattern_before): Likewise for param "before".
6642 (emit_insn_before): Likewise.
6643 (emit_jump_insn_before): Likewise.
6644 * final.c (get_insn_template): Likewise for param "insn", removing
6646 * output.h (get_insn_template): Likewise for 2nd param.
6647 * rtl.h (emit_insn_before): Likewise.
6648 (emit_jump_insn_before): Likewise.
6649 (emit_debug_insn_before_noloc): Likewise.
6650 (emit_insn_after): Likewise.
6651 (emit_jump_insn_after): Likewise.
6652 (emit_call_insn_after): Likewise.
6653 (emit_debug_insn_after): Likewise.
6654 (set_insn_deleted): Likewise for param.
6656 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6659 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6660 floating point modes, so that IFmode is numerically greater than
6661 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6662 to declare the ordering. This prevents IFmode from being
6663 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6664 machine. Include rs6000-modes.h to share the fractional values
6665 between genmodes* and the rest of the compiler.
6669 * config/rs6000/rs6000-modes.h: New file.
6670 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6671 meaning of rs6000_long_double_size so that 126..128 selects an
6672 appropriate 128-bit floating point type.
6673 (rs6000_option_override_internal): Likewise.
6674 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6675 (TARGET_LONG_DOUBLE_128): Change the meaning of
6676 rs6000_long_double_size so that 126..128 selects an appropriate
6677 128-bit floating point type.
6678 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6679 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6680 source and destination to match the standard usage.
6681 (truncifkf2): Likewise.
6682 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6683 ISA 2.07 to use an explicit clobber, instead of passing in a
6685 (copysign<mode>3_soft): Likewise.
6687 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6689 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6690 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6691 (vect_slp_analyze_instance_dependence): Likewise.
6692 (vect_enhance_data_refs_alignment): Likewise.
6693 (vect_analyze_data_refs_alignment): Likewise.
6694 (vect_slp_analyze_and_verify_instance_alignment
6695 (vect_analyze_data_ref_accesses): Likewise.
6696 (vect_prune_runtime_alias_test_list): Likewise.
6697 (vect_analyze_data_refs): Likewise.
6698 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6699 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6700 (vect_analyze_scalar_cycles_1): Likewise.
6701 (vect_get_loop_niters): Likewise.
6702 (vect_analyze_loop_form_1): Likewise.
6703 (vect_update_vf_for_slp): Likewise.
6704 (vect_analyze_loop_operations): Likewise.
6705 (vect_analyze_loop): Likewise.
6706 (vectorizable_induction): Likewise.
6707 (vect_transform_loop): Likewise.
6708 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6709 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6710 (vect_make_slp_decision): Likewise.
6711 (vect_detect_hybrid_slp): Likewise.
6712 (vect_slp_analyze_operations): Likewise.
6713 (vect_slp_bb): Likewise.
6714 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6715 (vectorizable_bswap): Likewise.
6716 (vectorizable_call): Likewise.
6717 (vectorizable_simd_clone_call): Likewise.
6718 (vectorizable_conversion): Likewise.
6719 (vectorizable_assignment): Likewise.
6720 (vectorizable_shift): Likewise.
6721 (vectorizable_operation): Likewise.
6722 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6724 2018-06-18 Martin Sebor <msebor@redhat.com>
6726 PR tree-optimization/81384
6727 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6728 * builtins.c (expand_builtin_strnlen): New function.
6729 (expand_builtin): Call it.
6730 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6731 * builtins.def (BUILT_IN_STRNLEN): New.
6732 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6733 Warn for bounds in excess of maximum object size.
6734 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6735 single-value ranges. Handle strnlen.
6736 (handle_builtin_strlen): Handle strnlen.
6737 (strlen_check_and_optimize_stmt): Same.
6738 * doc/extend.texi (Other Builtins): Document strnlen.
6740 2018-06-18 Maya Rashish <coypu@sdf.org>
6742 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6743 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6744 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6746 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6748 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6750 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6752 * tree.c (escaped_string::escape): Replace cast to char * by
6753 const_cast<char *> (unescaped).
6755 2018-06-18 Nick Clifton <nickc@redhat.com>
6758 * tree.c (escaped_string): New class. Converts an unescaped
6759 string into its escaped equivalent.
6760 (warn_deprecated_use): Use the new class to convert the
6761 deprecation message, if present.
6762 (test_escaped_strings): New self test.
6763 (test_c_tests): Add test_escaped_strings.
6764 * doc/extend.texi (deprecated): Add a note that the
6765 deprecation message is affected by the -fmessage-length
6766 option, and that control characters will be escaped.
6767 (#pragma GCC error): Document this pragma.
6768 (#pragma GCC warning): Likewise.
6769 * doc/invoke.texi (-fmessage-length): Document this option's
6770 effect on the #warning and #error preprocessor directives and
6771 the deprecated attribute.
6773 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6775 * tree.c (decl_value_expr_lookup): Revert latest change.
6776 (decl_value_expr_insert): Likewise.
6778 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6780 * gimplify.c (nonlocal_vlas): Delete.
6781 (nonlocal_vla_vars): Likewise.
6782 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6784 (gimplify_body): Do not create and destroy nonlocal_vlas.
6785 * tree-nested.c: Include diagnostic.h.
6786 (use_pointer_in_frame): Tweak.
6787 (lookup_field_for_decl): Add assertion and declare the transformation.
6788 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6789 internal error when the reference is in a wrong context. Do not
6790 create a debug decl by default.
6791 (note_nonlocal_block_vlas): Delete.
6792 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6793 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6794 create a debug decl by default.
6795 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6796 call is in a wrong context.
6797 (fixup_vla_decls): New function.
6798 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6799 debug variables were created.
6800 * tree.c (decl_value_expr_lookup): Add checking assertion.
6801 (decl_value_expr_insert): Likewise.
6803 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6806 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6807 * tree-scalar-evolution.c (interpret_expr): Likewise.
6808 (expression_expensive_p): Likewise.
6809 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6810 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6811 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6812 (ssa_defined_by_minus_one_stmt_p): New.
6814 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6817 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6818 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6819 * dojump.c (do_jump): Likewise.
6820 * expr.c (expand_expr_real_2): Check operand type's sign.
6821 * fold-const.c (const_unop): Handle ABSU_EXPR.
6822 (fold_abs_const): Likewise.
6823 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6824 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6825 (strip_sign_op_1): Likesise.
6826 * match.pd: Add new pattern to generate ABSU_EXPR.
6827 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6828 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6829 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6830 * tree-inline.c (estimate_operator_cost): Likewise.
6831 * tree-pretty-print.c (dump_generic_node): Likewise.
6832 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6833 * tree.def (ABSU_EXPR): New.
6835 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6838 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6839 documented as preserved for backward compatibility only.
6840 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6842 PR rtl-optimization/86108
6843 * bb-reorder.c (create_forwarder_block): Renamed to ...
6844 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6845 jump from new landing pad to the second part.
6846 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6849 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6852 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6853 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6854 Only call store_expr for halves if the mode is the same.
6857 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6858 Fix up comment formatting.
6860 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6862 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6863 (typed_splay_tree::closure,
6864 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6865 (typed_splay_tree::typed_splay_tree,
6866 typed_splay_tree::operator =): Declared private.
6867 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6868 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6869 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6870 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6871 typed_splay_tree::splay_tree_splay,
6872 typed_splay_tree::splay_tree_foreach_helper,
6873 typed_splay_tree::splay_tree_insert,
6874 typed_splay_tree::splay_tree_remove,
6875 typed_splay_tree::splay_tree_lookup,
6876 typed_splay_tree::splay_tree_predecessor,
6877 typed_splay_tree::splay_tree_successor,
6878 typed_splay_tree::splay_tree_min,
6879 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6880 (typed_splay_tree::root, typed_splay_tree::comp,
6881 typed_splay_tree::delete_key,
6882 typed_splay_tree::delete_value): New data members.
6883 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6884 typed_splay_tree::remove.
6886 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6888 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6889 -mginv and -mno-ginv to the assembler.
6890 * config/mips/mips.opt (-mcrc): New option.
6892 * doc/invoke.text (-mcrc): Document.
6895 2018-06-15 Nick Clifton <nickc@redhat.com>
6898 * tree.c (escaped_string): New class. Converts an unescaped
6899 string into its escaped equivalent.
6900 (warn_deprecated_use): Use the new class to convert the
6901 deprecation message, if present.
6902 (test_escaped_strings): New self test.
6903 (test_c_tests): Add test_escaped_strings.
6904 * doc/extend.texi (deprecated): Add a note that the
6905 deprecation message is affected by the -fmessage-length
6906 option, and that control characters will be escaped.
6907 (#pragma GCC error): Document this pragma.
6908 (#pragma GCC warning): Likewise.
6909 * doc/invoke.texi (-fmessage-length): Document this option's
6910 effect on the #warning and #error preprocessor directives and
6911 the deprecated attribute.
6913 2018-06-15 Richard Biener <rguenther@suse.de>
6915 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6916 here, also noting vector size used.
6917 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6918 size used in MSG_OPTIMIZED_LOCATIONS dump.
6919 (pass_slp_vectorize::execute): Adjust.
6921 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6924 * config/arc/arc.c (arc_return_address_register): Fix
6927 2018-06-15 Richard Biener <rguenther@suse.de>
6930 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6931 leave useless conversion stripping to force_gimple_operand_gsi.
6932 (gimplify_build2): Likewise.
6933 (gimplify_build1): Likewise.
6935 2018-06-15 Richard Biener <rguenther@suse.de>
6938 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6939 before adjusting their block.
6941 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6943 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6944 multilibs for *-*-rtems*.
6945 * config/riscv/t-rtems: New file.
6947 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6950 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6951 unsigned_type_for returns NULL.
6954 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6955 subregs of multi-word pseudos unless the float mode has word size.
6957 2018-06-14 Richard Biener <rguenther@suse.de>
6960 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6961 duplicate and harmful type_hash_canon.
6962 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6964 2018-06-14 Richard Biener <rguenther@suse.de>
6967 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6970 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6972 * config/rtems.h (STDINT_LONG32): Define.
6974 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6975 Prachi Godbole <prachi.godbole@imgtec.com>
6977 * config/mips/mips-cpus.def: Define P6600.
6978 * config/mips/mips-tables.opt: Regenerate.
6979 * config/mips/mips.c (mips_ucbranch_type): New enum.
6980 (mips_rtx_cost_data): Add support for P6600.
6981 (mips_issue_rate): Likewise.
6982 (mips_multipass_dfa_lookahead): Likewise.
6983 (mips_avoid_hazard): Likewise.
6984 (mips_reorg_process_insns): Likewise.
6985 (mips_classify_branch_p6600): New function.
6986 * config/mips/mips.h (TUNE_P6600): New define.
6987 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6988 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6989 * config/mips/mips.md: Include p6600.md.
6990 (processor): Add p6600.
6991 * config/mips/p6600.md: New file.
6992 * doc/invoke.texi: Add p6600 to supported architectures.
6994 2018-06-13 Martin Sebor <msebor@redhat.com>
6996 PR tree-optimization/86114
6997 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6999 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
7001 2018-06-13 Richard Biener <rguenther@suse.de>
7003 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
7004 Properly set vector type of the intermediate stmt.
7005 * tree-vect-stmts.c (vectorizable_operation): The destination
7006 var always has vectype_out type.
7008 2018-06-13 Jeff Law <law@redhat.com>
7010 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
7011 integer 0 for argument to print_rtl_with_bb.
7012 (rl78_reorg): Likewise.
7014 2018-06-13 David Malcolm <dmalcolm@redhat.com>
7016 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
7017 from rtx to rtx_insn *.
7018 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
7020 (add_sched_insns_for_speculation): Likewise for local "target",
7021 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7022 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
7023 from rtx_insn ** to rtx_code_label **.
7024 (reorg_emit_nops): Likewise.
7025 (c6x_reorg): Likewise for local "call_labels".
7026 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
7028 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
7029 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
7030 the loops over LABEL_REFS.
7031 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
7033 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7034 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
7035 (split_branches): Strengthen local "olabel" from rtx to
7036 rtx_insn *, adding a safe_as_a cast.
7037 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
7039 (add_insn_after): Likewise for first two params.
7040 (add_insn_before): Likewise.
7041 (remove_insn): Likewise for param.
7042 (emit_pattern_before_noloc): Likewise for second and third params.
7043 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
7044 (emit_call_insn_before_noloc): Likewise.
7045 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
7047 (emit_barrier_before): Likewise.
7048 (emit_label_before): Strengthen "label" param from "rtx" to
7049 "rtx_code_label *". Strengthen "before" param from "rtx" to
7051 (emit_insn_after_1): Strengthen "after" param from "rtx" to
7053 (emit_pattern_after_noloc): Likewise.
7054 (emit_insn_after_noloc): Likewise.
7055 (emit_jump_insn_after_noloc): Likewise.
7056 (emit_call_insn_after_noloc): Likewise.
7057 (emit_debug_insn_after_noloc): Likewise.
7058 (emit_barrier_after): Likewise.
7059 (emit_label_after): Likewise for both params.
7060 (emit_pattern_after_setloc): Likewise for "after" param. Convert
7061 "loc" param from "int" to "location_t".
7062 (emit_insn_after_setloc): Likewise.
7063 (emit_jump_insn_after_setloc): Likewise.
7064 (emit_call_insn_after_setloc): Likewise.
7065 (emit_debug_insn_after_setloc): Likewise.
7066 (emit_pattern_before_setloc): Likewise for "before" param. Convert
7067 "loc" param from "int" to "location_t".
7068 (emit_pattern_before): Convert NULL_RTX to NULL.
7069 (emit_insn_before_setloc): Convert "loc" param from "int" to
7071 (emit_jump_insn_before_setloc): Likewise.
7072 (emit_call_insn_before_setloc): Likewise.
7073 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
7074 rtx_insn *. Convert "loc" param from "int" to "location_t".
7075 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
7076 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
7077 Convert 3rd param from "int" to "location_t".
7078 (emit_barrier_before, emit_barrier_after, next_real_insn):
7079 Strengthen param from rtx to rtx_insn *.
7080 (emit_label_before): Strengthen 1st param from "rtx" to
7081 "rtx_code_label *". Strengthen 2nd param from "rtx" to
7083 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
7084 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
7085 Strengthen 2nd param from "rtx" to "rtx_insn *".
7086 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
7087 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
7088 Likewise. Convert 3rd param from "int" to "location_t".
7089 (emit_label_after): Strengthen 1st param from "rtx" to
7091 (next_real_insn, remove_insn): Strengthen param from "rtx" to
7093 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
7094 from "rtx" to "rtx_insn *".
7096 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
7098 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
7099 bodies streamed in with -Q.
7100 * dumpfile.c (dump_files): Add lto-stream-out dump file.
7101 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7102 * gimple-streamer-out.c: Include gimple-pretty-print.h
7103 (output_bb): Dump stmts streamed.
7104 * lto-section-out.c: Include print-tree.h
7105 (lto_begin_section): Dump sections created.
7106 (lto_output_decl_index): Dump decl encoded.
7107 * lto-streamer-out.c: Include print-tree.h
7108 (create_output_block): Dump output block created.
7109 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
7110 (output_function): Dump function output.
7111 (output_constructor): Dump constructor streamed.
7112 (write_global_stream): Output indexes encoded.
7113 (produce_asm_for_decls): Dump streams encoded.
7114 * lto-streamer.c (streamer_dump_file): New global var.
7115 * lto-streamer.h (streamer_dump_file): Declare.
7116 * passes.c (ipa_write_summaries): Initialize streamer dump.
7117 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
7120 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
7123 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
7124 offsets for register save directives. Emit a second batch of save
7125 directives, if need be, when the function accesses prior frames.
7127 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7129 * config/arc/fpu.md (fmasf4): Force operand to register.
7130 (fnmasf4): Likewise.
7132 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7134 * config/arc/arc-protos.h (arc_pad_return): Remove.
7135 * config/arc/arc.c (machine_function): Remove force_short_suffix
7137 (arc_print_operand): Adjust printing of '&'.
7138 (arc_verify_short): Remove conditional printing of short suffix.
7139 (arc_final_prescan_insn): Remove reference to size_reason.
7140 (pad_return): New function.
7141 (arc_reorg): Call pad_return.
7142 (arc_pad_return): Remove.
7143 (arc_init_machine_status): Remove reference to force_short_suffix.
7144 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
7145 (attr length): When attribute iscompact is true force to 2
7146 regardless; in the case of maybe check if we want to force the
7147 instruction to have 4 bytes length.
7148 (nopv): Change it to generate 4 byte long nop as well.
7149 (blockage): New pattern.
7150 (simple_return): Remove call to arc_pad_return.
7151 (p_return_i): Likewise.
7153 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7155 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
7157 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7159 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
7162 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7164 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
7165 for ARC700 and ARCv2.
7167 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
7170 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
7171 operands[2] instead of operands[1].
7174 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7176 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
7177 case, check whether the outer register overlaps an unallocatable
7178 register, not just whether it fits the required class.
7180 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7182 * poly-int.h (can_div_trunc_p): Add new overload in which all values
7184 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
7185 (memrefs_conflict_p): Likewise.
7186 (init_alias_analysis): Likewise.
7187 * cfgexpand.c (expand_debug_expr): Likewise.
7188 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
7189 * cse.c (fold_rtx): Likewise.
7190 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
7191 * expr.c (emit_block_move_hints): Likewise.
7192 (clear_storage_hints, push_block, emit_push_insn): Likewise.
7193 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
7194 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
7195 (emit_group_store): Likewise.
7196 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
7197 to read the PRE/POST_MODIFY increment.
7198 * calls.c (store_one_arg): Use strip_offset.
7199 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
7201 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
7203 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
7204 (simplify_binary_operation_1): Extend CONST_INT handling to
7206 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
7207 than a HOST_WIDE_INT.
7208 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
7210 (adjust_mems, add_stores): Update accodingly.
7211 (vt_canonicalize_addr): Track polynomial offsets.
7212 (emit_note_insn_var_location): Likewise.
7213 (vt_add_function_parameter): Likewise.
7214 (vt_initialize): Likewise.
7216 2018-06-12 Jeff Law <law@redhat.com>
7218 * config.gcc (alpha*-*-freebsd*): Remove.
7219 * config/alpha/freebsd.h: Remove.
7221 2018-06-12 David Malcolm <dmalcolm@redhat.com>
7224 * spellcheck-tree.c (levenshtein_distance): Rename to...
7225 (get_edit_distance): ...this, and update for underlying renaming.
7226 * spellcheck-tree.h (levenshtein_distance): Rename to...
7227 (get_edit_distance): ...this.
7228 * spellcheck.c (levenshtein_distance): Rename to...
7229 (get_edit_distance): ...this. Convert from Levenshtein distance
7230 to Damerau-Levenshtein distance by supporting transpositions of
7231 adjacent characters. Rename "v1" to "v_next" and "v0" to
7233 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
7234 (selftest::test_edit_distance_unit_test_oneway): ...this, and
7235 update for underlying renaming.
7236 (selftest::levenshtein_distance_unit_test): Rename to...
7237 (selftest::test_get_edit_distance_unit): ...this, and update for
7238 underlying renaming.
7239 (selftest::test_find_closest_string): Add example from PR 69968
7240 where transposition helps
7241 (selftest::test_metric_conditions): Update for renaming.
7242 (selftest::test_metric_conditions): Likewise.
7243 (selftest::spellcheck_c_tests): Likewise.
7244 * spellcheck.h (levenshtein_distance): Rename both overloads to...
7245 (get_edit_distance): ...this.
7246 (best_match::consider): Update for renaming.
7248 2018-06-12 Martin Sebor <msebor@redhat.com>
7250 PR tree-optimization/85259
7251 * builtins.c (compute_objsize): Handle constant offsets.
7252 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
7253 true iff a warning has been issued.
7254 * gimple.h (gimple_nonartificial_location): New function.
7255 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
7256 gimple_nonartificial_location and handle -Wno-system-headers.
7257 (handle_builtin_stxncpy): Same.
7259 2018-06-12 Martin Sebor <msebor@redhat.com>
7262 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7264 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7266 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7267 BUILTIN_VEC_XST entries for pointer to double and long long.
7269 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7272 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7274 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7277 2018-06-12 Martin Liska <mliska@suse.cz>
7279 * doc/options.texi: Document IntegerRange.
7281 2018-06-12 Martin Liska <mliska@suse.cz>
7283 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7284 * opt-functions.awk: Handle Deprecated flag.
7285 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7287 (read_cmdline_option): Report warning for a deprecated option.
7288 * opts.h (struct cl_option): Add new field cl_deprecated.
7289 (CL_ERR_DEPRECATED): New.
7291 2018-06-12 Martin Liska <mliska@suse.cz>
7293 * doc/options.texi: Document Deprecated option flag.
7295 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7297 * config/arc/arc-arch.h (arc_extras): New enum.
7298 (arc_cpu_t):Add field extra.
7299 (arc_cpu_types): Consider the extras.
7300 * config/arc/arc-cpus.def: Add extras info.
7301 * config/arc/arc-opts.h (processor_type): Consider extra field.
7302 * config/arc/arc.c (arc_override_options): Handle extra field.
7304 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7306 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7307 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7309 * config/arc/arc.c (arc_init): Update pic warning.
7310 (irq_range): Update irq range parsing warnings.
7311 (arc_override_options): Update various warning messages.
7312 (arc_handle_aux_attribute): Likewise.
7314 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7316 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7318 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7320 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7321 compiler messages directives.
7323 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7325 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7326 * config/mips/mips-tables.opt: Regenerate.
7327 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7329 * doc/invoke.texi: Document -march=i6500.
7331 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7333 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7334 (i6400_gpmul): Add cpu_unit.
7335 (i6400_gpdiv): Likewise.
7336 (i6400_msa_add_d): Update reservations.
7337 (i6400_msa_int_add) Likewise.
7338 (i6400_msa_short_logic3) Likewise.
7339 (i6400_msa_short_logic2) Likewise.
7340 (i6400_msa_short_logic) Likewise.
7341 (i6400_msa_move) Likewise.
7342 (i6400_msa_cmp) Likewise.
7343 (i6400_msa_short_float2) Likewise.
7344 (i6400_msa_div_d) Likewise.
7345 (i6400_msa_long_logic1) Likewise.
7346 (i6400_msa_long_logic2) Likewise.
7347 (i6400_msa_mult) Likewise.
7348 (i6400_msa_long_float2) Likewise.
7349 (i6400_msa_long_float4) Likewise.
7350 (i6400_msa_long_float5) Likewise.
7351 (i6400_msa_long_float8) Likewise.
7352 (i6400_fpu_fadd): Include frint type.
7353 (i6400_fpu_store): New define_insn_reservation.
7354 (i6400_fpu_load): Likewise.
7355 (i6400_fpu_move): Likewise.
7356 (i6400_fpu_fcmp): Likewise.
7357 (i6400_fpu_fmadd): Likewise.
7358 (i6400_int_mult): Include imul3nc type and update reservation.
7359 (i6400_int_div): Include idiv3 type and update reservation.
7360 (i6400_int_load): Update to check type not move_type.
7361 (i6400_int_store): Likewise.
7362 (i6400_int_prefetch): Set zero latency.
7364 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7366 * gcc.c: Document new %@{...} sequence.
7367 (LINK_COMMAND_SPEC): Use it for the -L switches.
7368 (cpp_unique_options): Use it for the -I switches.
7369 (at_file_argbuf): New global variable.
7370 (in_at_file): Likewise.
7371 (alloc_args): Create at_file_argbuf.
7372 (clear_args): Truncate at_file_argbuf.
7373 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7374 (open_at_file): New function.
7375 (close_at_file): Likewise.
7376 (create_at_file): Delete.
7377 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7380 (validate_switches_from_spec): Deal with %@{...} sequence.
7381 (validate_switches): Likewise.
7382 (driver::finalize): Call clear_args.
7384 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7386 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7388 2018-06-11 Martin Sebor <msebor@redhat.com>
7390 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7391 (Wno-class-memaccess): Add @opindex.
7392 (Wno-templates, Wno-multiple-inheritance): Same.
7393 (Wno-virtual-inheritance, Wno-namespaces): Same.
7394 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7395 (Wno-format-overflow, Wno-format-truncation): Same.
7396 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7397 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7398 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7399 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7400 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7401 (Wno-misspelled-isr): Same.
7403 2018-06-11 Martin Sebor <msebor@redhat.com>
7405 * PR tree-optimization/86083
7406 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7408 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7410 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7412 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7415 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7416 on the correct operand.
7417 (*movdi_internal64): Ditto.
7419 2018-06-11 Martin Liska <mliska@suse.cz>
7421 PR tree-optimization/86089
7422 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7424 2018-06-11 Julia Koval <julia.koval@intel.com>
7426 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7427 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7428 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7430 2018-06-11 Olivier Hainque <hainque@adacore.com>
7432 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7433 for Ada with strict dwarf2.
7435 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7438 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7441 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7443 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7445 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7447 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7450 2018-06-08 Martin Liska <mliska@suse.cz>
7452 * tree-cfg.h (debug_function): Fix argument type to match
7455 2018-06-08 Martin Liska <mliska@suse.cz>
7457 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7458 Remove usage of MPX-related (and removed) fields.
7459 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7461 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7463 * cfg.c (debug): Use TDF_NONE rather than 0.
7464 * cfghooks.c (debug): Likewise.
7465 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7466 (struct dump_option_value_info): Convert to...
7467 (struct kv_pair): ...this template type.
7468 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7470 (optinfo_verbosity_options): Likewise.
7471 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7473 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7474 than int for "optgroup_flags" param.
7475 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7477 (dump_dec): Likewise.
7478 (dump_finish): Use TDF_NONE rather than 0.
7479 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7480 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7481 than 0. Update for change to option_ptr.
7482 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7483 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7484 0. Update for changes to optinfo_verbosity_options and
7486 (opt_info_switch_p): Convert optgroup_flags from int to
7488 (dump_basic_block): Use dump_flags_t rather than int
7489 for "dump_kind" param.
7490 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7491 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7492 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7493 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7494 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7495 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7496 TDF_NONE): Convert from macros to...
7497 (enum dump_flag): ...this new enum.
7498 (dump_flags_t): Update to use enum.
7499 (operator|, operator&, operator~, operator|=, operator&=):
7500 Implement for dump_flags_t.
7501 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7502 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7503 Convert from macros to...
7504 (enum optgroup_flag): ...this new enum.
7505 (optgroup_flags_t): New typedef.
7506 (operator|, operator|=): Implement for optgroup_flags_t.
7507 (struct dump_file_info): Convert field "alt_flags" to
7508 dump_flags_t. Convert field "optgroup_flags" to
7510 (dump_basic_block): Use dump_flags_t rather than int for param.
7511 (dump_generic_expr_loc): Likewise.
7512 (dump_dec): Likewise.
7513 (dump_register): Convert param "optgroup_flags" to
7515 (opt_info_enable_passes): Likewise.
7516 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7518 * gimple-pretty-print.c (debug): Likewise.
7519 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7520 (merged_store_group::apply_stores): Likewise.
7521 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7522 * gimple.c (verify_gimple_pp): Likewise.
7523 * graphite-poly.c (print_pbb_body): Likewise.
7524 * passes.c (pass_manager::register_one_dump_file): Convert
7525 local "optgroup_flags" to optgroup_flags_t.
7526 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7528 (debug_body): Likewise.
7529 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7530 to optgroup_flags_t.
7531 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7533 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7534 (convert_mult_to_fma): Likewise.
7535 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7536 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7537 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7538 "dump_kind" to dump_flags_t.
7540 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7542 * config/rs6000/rs6000.c (min, max): Delete.
7544 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7546 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7547 -mabi=spe and -mabi=no-spe.
7549 2018-06-08 Martin Liska <mliska@suse.cz>
7551 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7552 where we expect an existing summary.
7554 2018-06-08 Martin Liska <mliska@suse.cz>
7556 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7557 * ipa-inline.h (estimate_edge_growth): Likewise.
7559 2018-06-08 Martin Liska <mliska@suse.cz>
7561 * cgraph.c (function_version_hasher::hash): Use
7562 cgraph_node::get_uid ().
7563 (function_version_hasher::equal):
7564 * cgraph.h (cgraph_node::get_uid): New method.
7565 * ipa-inline.c (update_caller_keys): Use
7566 cgraph_node::get_uid ().
7567 (update_callee_keys): Likewise.
7568 * ipa-utils.c (searchc): Likewise.
7569 (ipa_reduced_postorder): Likewise.
7570 * lto-cgraph.c (input_node): Likewise.
7571 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7572 * symbol-summary.h (symtab_insertion): Likewise.
7573 (symtab_removal): Likewise.
7574 (symtab_duplication): Likewise.
7575 * tree-pretty-print.c (dump_function_header): Likewise.
7576 * tree-sra.c (convert_callers_for_node): Likewise.
7578 2018-06-08 Martin Liska <mliska@suse.cz>
7580 * cgraph.c (symbol_table::create_edge): Always assign a new
7582 (symbol_table::free_edge): Do not recycle numbers.
7583 * cgraph.h (cgraph_edge::get): New method.
7584 * symbol-summary.h (symtab_removal): Use it.
7585 (symtab_duplication): Likewise.
7586 (call_summary::hashable_uid): Remove.
7588 2018-06-08 Martin Liska <mliska@suse.cz>
7590 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7591 (initialize_growth_caches): Remove.
7592 (free_growth_caches): Likewise.
7593 (do_estimate_edge_time): Use edge_growth_cache.
7594 (do_estimate_edge_size): Likewise.
7595 (do_estimate_edge_hints): Likewise.
7596 * ipa-inline.c (reset_edge_caches): Likewise.
7597 (recursive_inlining): Likewise.
7598 (inline_small_functions): Likewise.
7599 * ipa-inline.h (initialize_growth_caches): Remove.
7600 (estimate_edge_size): Likewise.
7601 (estimate_edge_time): Likewise.
7602 (estimate_edge_hints): Likewise.
7603 (reset_edge_growth_cache): Likewise.
7604 * symbol-summary.h (call_summary::remove): New method.
7606 2018-06-08 Martin Liska <mliska@suse.cz>
7608 * ipa-cp.c (class edge_clone_summary): New summary.
7609 (grow_edge_clone_vectors): Remove.
7610 (ipcp_edge_duplication_hook): Remove.
7611 (class edge_clone_summary_t): New call_summary class.
7612 (ipcp_edge_removal_hook): Remove.
7613 (edge_clone_summary_t::duplicate): New function.
7614 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7615 (create_specialized_node): Likewise.
7616 (ipcp_driver): Initialize edge_clone_summaries and do not
7619 2018-06-08 Martin Liska <mliska@suse.cz>
7621 * symbol-summary.h (get): New function.
7622 (call_summary::m_initialize_when_cloning): New class member.
7624 2018-06-08 Martin Liska <mliska@suse.cz>
7626 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7627 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7628 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7629 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7630 (remove_cgraph_node_from_order): Use the removed_nodes set.
7631 (do_per_function_toporder): Likwise.
7632 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7633 instead of summary_uid.
7634 (symtab_removal): Likewise.
7635 (symtab_duplication): Likewise.
7637 2018-06-08 Martin Liska <mliska@suse.cz>
7639 * ipa-cp.c (ipcp_store_bits_results): Use
7640 ipcp_transformation_sum.
7641 (ipcp_store_vr_results): Likewise.
7642 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7644 (ipcp_transformation_initialize): ... this.
7645 (ipa_set_node_agg_value_chain):
7646 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7647 (write_ipcp_transformation_info): Likewise.
7648 (read_ipcp_transformation_info): Likewise.
7649 (ipcp_update_bits): Likewise.
7650 (ipcp_update_vr): Likewise.
7651 (ipcp_transform_function): Likewise.
7652 * ipa-prop.h: Rename ipcp_transformation_summary to
7653 ipcp_transformation.
7654 (class ipcp_transformation_t): New function summary.
7655 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7656 (ipa_get_agg_replacements_for_node): Likewise.
7658 2018-06-08 Martin Liska <mliska@suse.cz>
7660 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7662 (class funct_state_summary_t): New function_summary class.
7663 (has_function_state): Remove.
7664 (get_function_state): Likewise.
7665 (set_function_state): Likewise.
7666 (add_new_function): Likewise.
7667 (funct_state_summary_t::insert): New function.
7668 (duplicate_node_data): Remove.
7669 (remove_node_data): Remove.
7670 (funct_state_summary_t::duplicate): New function.
7671 (register_hooks): Create new funct_state_summaries.
7672 (pure_const_generate_summary): Use it.
7673 (pure_const_write_summary): Likewise.
7674 (pure_const_read_summary): Likewise.
7675 (propagate_pure_const): Likewise.
7676 (propagate_nothrow): Likewise.
7677 (dump_malloc_lattice): Likewise.
7678 (propagate_malloc): Likewise.
7679 (execute): Do not register hooks, just remove summary
7681 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7684 2018-06-08 Martin Liska <mliska@suse.cz>
7686 * ipa-reference.c (remove_node_data): Remove.
7687 (duplicate_node_data): Likewise.
7688 (class ipa_ref_var_info_summary_t): New class.
7689 (class ipa_ref_opt_summary_t): Likewise.
7690 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7691 (get_reference_optimization_summary): Use
7692 ipa_ref_opt_sum_summaries.
7693 (set_reference_vars_info): Remove.
7694 (set_reference_optimization_summary): Likewise.
7695 (ipa_init): Create summaries.
7696 (init_function_info): Use function summary.
7697 (ipa_ref_opt_summary_t::duplicate): New function.
7698 (ipa_ref_opt_summary_t::remove): New function.
7699 (get_read_write_all_from_node): Fix GNU coding style.
7700 (propagate): Use function summary.
7701 (write_node_summary_p): Fix GNU coding style.
7702 (stream_out_bitmap): Likewise.
7703 (ipa_reference_read_optimization_summary): Use function summary.
7704 (ipa_reference_c_finalize): Do not release hooks.
7706 2018-06-08 Martin Liska <mliska@suse.cz>
7708 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7709 (analyze_function_body): Extract multiple calls of get_create.
7710 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7711 * ipa-inline.c (recursive_inlining): Use ::get method.
7712 * ipa-inline.h (estimate_edge_growth): Likewise.
7714 2018-06-08 Martin Liska <mliska@suse.cz>
7716 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7718 (hsa_function_summary::hsa_function_summary): Use the new enum
7720 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7721 * hsa-gen.c (hsa_get_host_function): Likewise.
7722 (get_brig_function_name): Likewise.
7723 * ipa-hsa.c (process_hsa_functions): Likewise.
7724 (ipa_hsa_write_summary): Likewise.
7725 * symbol-summary.h (symtab_duplication): Use ::get function/
7726 (get): New function.
7728 2018-06-08 Martin Liska <mliska@suse.cz>
7730 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7732 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7733 (hsa_register_kernel): Likewise.
7734 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7735 * hsa-gen.c (hsa_get_host_function): Likewise.
7736 (get_brig_function_name): Likewise.
7737 (generate_hsa): Likewise.
7738 (pass_gen_hsail::execute): Likewise.
7739 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7740 (devirtualization_time_bonus): Likewise.
7741 (ipcp_propagate_stage): Likewise.
7742 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7743 (edge_set_predicate): Likewise.
7744 (evaluate_conditions_for_known_args): Likewise.
7745 (evaluate_properties_for_edge): Likewise.
7746 (ipa_fn_summary::reset): Likewise.
7747 (ipa_fn_summary_t::duplicate): Likewise.
7748 (dump_ipa_call_summary): Likewise.
7749 (ipa_dump_fn_summary): Likewise.
7750 (analyze_function_body): Likewise.
7751 (compute_fn_summary): Likewise.
7752 (estimate_edge_devirt_benefit): Likewise.
7753 (estimate_edge_size_and_time): Likewise.
7754 (estimate_calls_size_and_time): Likewise.
7755 (estimate_node_size_and_time): Likewise.
7756 (inline_update_callee_summaries): Likewise.
7757 (remap_edge_change_prob): Likewise.
7758 (remap_edge_summaries): Likewise.
7759 (ipa_merge_fn_summary_after_inlining): Likewise.
7760 (ipa_update_overall_fn_summary): Likewise.
7761 (read_ipa_call_summary): Likewise.
7762 (inline_read_section): Likewise.
7763 (write_ipa_call_summary): Likewise.
7764 (ipa_fn_summary_write): Likewise.
7765 (ipa_free_fn_summary): Likewise.
7766 * ipa-hsa.c (process_hsa_functions): Likewise.
7767 (ipa_hsa_write_summary): Likewise.
7768 (ipa_hsa_read_section): Likewise.
7769 * ipa-icf.c (sem_function::merge): Likewise.
7770 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7771 (do_estimate_edge_time): Likewise.
7772 (estimate_size_after_inlining): Likewise.
7773 (estimate_growth): Likewise.
7774 (growth_likely_positive): Likewise.
7775 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7776 (inline_call): Likewise.
7777 * ipa-inline.c (caller_growth_limits): Likewise.
7778 (can_inline_edge_p): Likewise.
7779 (can_inline_edge_by_limits_p): Likewise.
7780 (compute_uninlined_call_time): Likewise.
7781 (compute_inlined_call_time): Likewise.
7782 (want_inline_small_function_p): Likewise.
7783 (edge_badness): Likewise.
7784 (update_caller_keys): Likewise.
7785 (update_callee_keys): Likewise.
7786 (recursive_inlining): Likewise.
7787 (inline_small_functions): Likewise.
7788 (inline_to_all_callers_1): Likewise.
7789 (dump_overall_stats): Likewise.
7790 (early_inline_small_functions): Likewise.
7791 (early_inliner): Likewise.
7792 * ipa-inline.h (estimate_edge_growth): Likewise.
7793 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7794 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7795 * ipa-prop.h (IPA_NODE_REF): Likewise.
7796 (IPA_EDGE_REF): Likewise.
7797 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7798 (propagate_malloc): Likewise.
7799 * ipa-split.c (execute_split_functions): Likewise.
7800 * symbol-summary.h: Rename get to get_create.
7802 (get_create): Likewise.
7803 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7805 2018-06-08 Martin Liska <mliska@suse.cz>
7807 * symbol-summary.h (release): Move definition out of class
7809 (symtab_removal): Likewise.
7810 (symtab_duplication): Likewise.
7812 2018-06-08 Martin Liska <mliska@suse.cz>
7814 * symbol-summary.h (function_summary): Move constructor
7815 implementation out of class declaration.
7816 (release): Likewise.
7817 (symtab_insertion): Likewise.
7818 (symtab_removal): Likewise.
7819 (symtab_duplication): Likewise.
7822 2018-06-08 Martin Liska <mliska@suse.cz>
7824 * Makefile.in: Remove support for MPX (macros, related functions,
7825 fields in cgraph_node, ...).
7826 * builtin-types.def (BT_BND): Likewise.
7827 (BT_FN_BND_CONST_PTR): Likewise.
7828 (BT_FN_CONST_PTR_BND): Likewise.
7829 (BT_FN_VOID_PTR_BND): Likewise.
7830 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7831 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7832 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7833 (expand_builtin_mempcpy_with_bounds): Likewise.
7834 (expand_builtin_memset_with_bounds): Likewise.
7835 (expand_builtin_memset_args): Likewise.
7836 (std_expand_builtin_va_start): Likewise.
7837 (expand_builtin): Likewise.
7838 (expand_builtin_with_bounds): Likewise.
7839 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7840 (DEF_LIB_BUILTIN_CHKP): Likewise.
7841 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7842 (DEF_CHKP_BUILTIN): Likewise.
7843 (BUILT_IN_MEMCPY): Likewise.
7844 (BUILT_IN_MEMMOVE): Likewise.
7845 (BUILT_IN_MEMPCPY): Likewise.
7846 (BUILT_IN_MEMSET): Likewise.
7847 (BUILT_IN_STPCPY): Likewise.
7848 (BUILT_IN_STRCAT): Likewise.
7849 (BUILT_IN_STRCHR): Likewise.
7850 (BUILT_IN_STRCPY): Likewise.
7851 (BUILT_IN_STRLEN): Likewise.
7852 (BUILT_IN_MEMCPY_CHK): Likewise.
7853 (BUILT_IN_MEMMOVE_CHK): Likewise.
7854 (BUILT_IN_MEMPCPY_CHK): Likewise.
7855 (BUILT_IN_MEMSET_CHK): Likewise.
7856 (BUILT_IN_STPCPY_CHK): Likewise.
7857 (BUILT_IN_STRCAT_CHK): Likewise.
7858 (BUILT_IN_STRCPY_CHK): Likewise.
7859 * calls.c (store_bounds): Likewise.
7860 (emit_call_1): Likewise.
7861 (special_function_p): Likewise.
7862 (maybe_warn_nonstring_arg): Likewise.
7863 (initialize_argument_information): Likewise.
7864 (finalize_must_preallocate): Likewise.
7865 (compute_argument_addresses): Likewise.
7866 (expand_call): Likewise.
7867 * cfgexpand.c (expand_call_stmt): Likewise.
7868 (expand_return): Likewise.
7869 (expand_gimple_stmt_1): Likewise.
7870 (pass_expand::execute): Likewise.
7871 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7872 (cgraph_node::remove): Likewise.
7873 (cgraph_node::dump): Likewise.
7874 (cgraph_node::verify_node): Likewise.
7875 * cgraph.h (chkp_function_instrumented_p): Likewise.
7876 (symtab_node::get_alias_target): Likewise.
7877 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7878 (cgraph_local_p): Likewise.
7879 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7880 (cgraph_edge::rebuild_references): Likewise.
7881 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7882 (walk_polymorphic_call_targets): Likewise.
7883 (cgraph_node::expand_thunk): Likewise.
7884 (symbol_table::output_weakrefs): Likewise.
7885 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7886 (ix86_handle_option): Likewise.
7887 * config/i386/constraints.md: Likewise.
7888 * config/i386/i386-builtin-types.def (BND): Likewise.
7892 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7893 (BDESC_FIRST): Likewise.
7895 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7896 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7897 * config/i386/i386.c (enum reg_class): Likewise.
7898 (ix86_target_string): Likewise.
7899 (ix86_option_override_internal): Likewise.
7900 (ix86_conditional_register_usage): Likewise.
7901 (ix86_valid_target_attribute_inner_p): Likewise.
7902 (ix86_set_indirect_branch_type): Likewise.
7903 (ix86_set_current_function): Likewise.
7904 (ix86_function_arg_regno_p): Likewise.
7905 (init_cumulative_args): Likewise.
7906 (ix86_function_arg_advance): Likewise.
7907 (ix86_function_arg): Likewise.
7908 (ix86_pass_by_reference): Likewise.
7909 (ix86_function_value_regno_p): Likewise.
7910 (ix86_function_value_1): Likewise.
7911 (ix86_function_value_bounds): Likewise.
7912 (ix86_return_in_memory): Likewise.
7913 (ix86_setup_incoming_vararg_bounds): Likewise.
7914 (ix86_va_start): Likewise.
7915 (indirect_thunk_need_prefix): Likewise.
7916 (print_reg): Likewise.
7917 (ix86_print_operand): Likewise.
7918 (ix86_expand_call): Likewise.
7919 (ix86_output_function_return): Likewise.
7920 (reg_encoded_number): Likewise.
7921 (BDESC_VERIFYS): Likewise.
7922 (ix86_init_mpx_builtins): Likewise.
7923 (ix86_init_builtins): Likewise.
7924 (ix86_emit_cmove): Likewise.
7925 (ix86_emit_move_max): Likewise.
7926 (ix86_expand_builtin): Likewise.
7927 (ix86_builtin_mpx_function): Likewise.
7928 (ix86_get_arg_address_for_bt): Likewise.
7929 (ix86_load_bounds): Likewise.
7930 (ix86_store_bounds): Likewise.
7931 (ix86_load_returned_bounds): Likewise.
7932 (ix86_store_returned_bounds): Likewise.
7933 (ix86_class_likely_spilled_p): Likewise.
7934 (ix86_hard_regno_mode_ok): Likewise.
7935 (x86_order_regs_for_local_alloc): Likewise.
7936 (ix86_mitigate_rop): Likewise.
7937 (ix86_bnd_prefixed_insn_p): Likewise.
7938 (ix86_mpx_bound_mode): Likewise.
7939 (ix86_make_bounds_constant): Likewise.
7940 (ix86_initialize_bounds): Likewise.
7941 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7942 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7943 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7944 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7945 (TARGET_CHKP_BOUND_MODE): Likewise.
7946 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7947 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7948 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7949 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7950 * config/i386/i386.h (TARGET_MPX): Likewise.
7951 (TARGET_MPX_P): Likewise.
7952 (VALID_BND_REG_MODE): Likewise.
7953 (FIRST_BND_REG): Likewise.
7954 (LAST_BND_REG): Likewise.
7955 (enum reg_class): Likewise.
7956 (BND_REG_P): Likewise.
7957 (BND_REGNO_P): Likewise.
7958 (BNDmode): Likewise.
7959 (ADJUST_INSN_LENGTH): Likewise.
7960 * config/i386/i386.md: Likewise.
7961 * config/i386/i386.opt: Likewise.
7962 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7963 (defined): Likewise.
7964 (LINK_MPX): Likewise.
7965 (MPX_SPEC): Likewise.
7966 (LIBMPX_SPEC): Likewise.
7967 (LIBMPXWRAPPERS_SPEC): Likewise.
7968 (CHKP_SPEC): Likewise.
7969 * config/i386/predicates.md: Likewise.
7970 * dbxout.c (dbxout_type): Likewise.
7971 * doc/extend.texi: Likewise.
7972 * doc/invoke.texi: Likewise.
7973 * doc/md.texi: Likewise.
7974 * doc/tm.texi: Likewise.
7975 * doc/tm.texi.in: Likewise.
7976 * dwarf2out.c (is_base_type): Likewise.
7977 (gen_formal_types_die): Likewise.
7978 (gen_subprogram_die): Likewise.
7979 (gen_type_die_with_usage): Likewise.
7980 (gen_decl_die): Likewise.
7981 (dwarf2out_late_global_decl): Likewise.
7982 * expr.c (expand_assignment): Likewise.
7983 (emit_storent_insn): Likewise.
7984 (store_expr_with_bounds): Likewise.
7985 (store_expr): Likewise.
7986 (expand_expr_real_1): Likewise.
7987 * expr.h (store_expr_with_bounds): Likewise.
7988 * function.c (use_register_for_decl): Likewise.
7989 (struct bounds_parm_data): Likewise.
7990 (assign_parms_augmented_arg_list): Likewise.
7991 (assign_parm_find_entry_rtl): Likewise.
7992 (assign_parm_is_stack_parm): Likewise.
7993 (assign_parm_load_bounds): Likewise.
7994 (assign_bounds): Likewise.
7995 (assign_parms): Likewise.
7996 (expand_function_start): Likewise.
7997 * gcc.c (CHKP_SPEC): Likewise.
7998 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7999 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
8000 (wrestrict_dom_walker::check_call): Likewise.
8001 * gimple.c (gimple_build_call_from_tree): Likewise.
8002 * gimple.h (enum gf_mask): Likewise.
8003 (gimple_call_with_bounds_p): Likewise.
8004 (gimple_call_set_with_bounds): Likewise.
8005 * gimplify.c (gimplify_init_constructor): Likewise.
8006 * ipa-cp.c (initialize_node_lattices): Likewise.
8007 (propagate_constants_across_call): Likewise.
8008 (find_more_scalar_values_for_callers_subset): Likewise.
8009 * ipa-hsa.c (process_hsa_functions): Likewise.
8010 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
8011 * ipa-icf.c (sem_function::merge): Likewise.
8012 * ipa-inline.c (early_inliner): Likewise.
8013 * ipa-pure-const.c (warn_function_noreturn): Likewise.
8014 (warn_function_cold): Likewise.
8015 (propagate_pure_const): Likewise.
8016 * ipa-ref.h (enum GTY): Likewise.
8017 * ipa-split.c (find_retbnd): Likewise.
8018 (consider_split): Likewise.
8019 (split_function): Likewise.
8020 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8021 * ipa.c (walk_polymorphic_call_targets): Likewise.
8022 (symbol_table::remove_unreachable_nodes): Likewise.
8023 (process_references): Likewise.
8024 (cgraph_build_static_cdtor_1): Likewise.
8025 * lto-cgraph.c (lto_output_node): Likewise.
8026 (output_refs): Likewise.
8027 (compute_ltrans_boundary): Likewise.
8028 (input_overwrite_node): Likewise.
8029 (input_node): Likewise.
8030 (input_cgraph_1): Likewise.
8031 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
8032 * passes.c (pass_manager::execute_early_local_passes): Likewise.
8033 (class pass_chkp_instrumentation_passes): Likewise.
8034 (make_pass_chkp_instrumentation_passes): Likewise.
8035 * passes.def: Likewise.
8036 * rtl.h (struct GTY): Likewise.
8037 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
8038 * stor-layout.c (layout_type): Likewise.
8039 * symtab.c: Likewise.
8040 * target.def: Likewise.
8041 * targhooks.c (default_chkp_bound_type): Likewise.
8042 (default_chkp_bound_mode): Likewise.
8043 (default_builtin_chkp_function): Likewise.
8044 (default_chkp_function_value_bounds): Likewise.
8045 (default_chkp_make_bounds_constant): Likewise.
8046 (default_chkp_initialize_bounds): Likewise.
8047 * targhooks.h (default_chkp_bound_type): Likewise.
8048 (default_chkp_bound_mode): Likewise.
8049 (default_builtin_chkp_function): Likewise.
8050 (default_chkp_function_value_bounds): Likewise.
8051 (default_chkp_make_bounds_constant): Likewise.
8052 (default_chkp_initialize_bounds): Likewise.
8053 * toplev.c (compile_file): Likewise.
8054 (process_options): Likewise.
8055 * tree-core.h (DEF_BUILTIN): Likewise.
8056 (DEF_BUILTIN_CHKP): Likewise.
8057 * tree-inline.c (declare_return_variable): Likewise.
8058 (remap_gimple_stmt): Likewise.
8059 (copy_bb): Likewise.
8060 (initialize_inlined_parameters): Likewise.
8061 (expand_call_inline): Likewise.
8062 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
8063 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
8064 (make_pass_ipa_chkp_produce_thunks): Likewise.
8065 (make_pass_chkp): Likewise.
8066 (make_pass_chkp_opt): Likewise.
8067 (make_pass_chkp_instrumentation_passes): Likewise.
8068 * tree-pretty-print.c (dump_generic_node): Likewise.
8069 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
8070 * tree-ssa-dce.c (propagate_necessity): Likewise.
8071 (eliminate_unnecessary_stmts): Likewise.
8072 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
8073 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
8074 * tree-ssa-sccvn.h: Likewise.
8075 * tree-ssa-strlen.c (get_string_length): Likewise.
8076 (valid_builtin_call): Likewise.
8077 (adjust_last_stmt): Likewise.
8078 (handle_builtin_strchr): Likewise.
8079 (handle_builtin_strcpy): Likewise.
8080 (handle_builtin_stxncpy): Likewise.
8081 (handle_builtin_memcpy): Likewise.
8082 (handle_builtin_strcat): Likewise.
8083 (strlen_check_and_optimize_stmt): Likewise.
8084 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
8085 * tree-streamer-in.c: Likewise.
8086 * tree-streamer.c (record_common_node): Likewise.
8087 * tree.c (tree_code_size): Likewise.
8088 (wide_int_to_tree_1): Likewise.
8089 (type_contains_placeholder_1): Likewise.
8090 (build_common_tree_nodes): Likewise.
8091 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
8092 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
8093 (POINTER_BOUNDS_P): Likewise.
8094 (BOUNDED_TYPE_P): Likewise.
8095 (BOUNDED_P): Likewise.
8096 (CALL_WITH_BOUNDS_P): Likewise.
8097 (pointer_bounds_type_node): Likewise.
8098 * value-prof.c (gimple_ic): Likewise.
8099 * var-tracking.c (vt_add_function_parameters): Likewise.
8100 * varasm.c (make_decl_rtl): Likewise.
8101 (assemble_start_function): Likewise.
8102 (output_constant): Likewise.
8103 (maybe_assemble_visibility): Likewise.
8104 * varpool.c (ctor_for_folding): Likewise.
8105 * chkp-builtins.def: Remove.
8106 * ipa-chkp.c: Remove.
8107 * ipa-chkp.h: Remove.
8108 * rtl-chkp.c: Remove.
8109 * rtl-chkp.h: Remove.
8110 * tree-chkp-opt.c: Remove.
8111 * tree-chkp.c: Remove.
8112 * tree-chkp.h: Remove.
8114 2018-06-07 Carl Love <cel@us.ibm.com>
8116 * config/rs6000/vsx.md (vextract_fp_from_shorth,
8117 vextract_fp_from_shortl): Add BE support.
8119 2018-06-07 Paul Koning <ni1d@arrl.net>
8121 * compare-elim.c (try_merge_compare): Don't merge compare if
8122 address contains a side effect.
8123 (try_eliminate_compare): Likewise.
8125 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
8127 * config.gcc: Support "tremont".
8128 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
8129 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8131 * config/i386/i386.c (m_TREMONT): Define.
8132 (processor_target_table): Add "tremont".
8133 (PTA_TREMONT): Define.
8134 (ix86_lea_outperforms): Add TARGET_TREMONT.
8135 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
8136 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
8137 and M_INTEL_GOLDMONT_PLUS.
8138 (fold_builtin_cpu): Add "tremont".
8139 (ix86_add_stmt_cost): Add TARGET_TREMONT.
8140 (ix86_option_override_internal): Add "tremont".
8141 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
8142 (processor_type): Add PROCESSOR_TREMONT.
8143 * config/i386/x86-tune.def: Add m_TREMONT.
8144 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
8146 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8148 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
8149 symbol defined for msp430i* devices to be lower case.
8151 2018-06-07 Richard Biener <rguenther@suse.de>
8153 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
8154 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
8155 Properly wrap signed arithmetic if overflow wraps.
8157 2018-06-07 Jakub Jelinek <jakub@redhat.com>
8159 PR tree-optimization/69615
8160 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
8161 of a cast from a same precision integral SSA_NAME in a bb dominated
8162 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
8163 cast to utype if rhs2 has already a compatible type.
8165 2018-06-07 Richard Biener <rguenther@suse.de>
8167 PR tree-optimization/85935
8168 * graphite-scop-detection.c (find_params_in_bb): Analyze
8169 condition operands with respect to the correct loop. Assert
8170 the analysis doesn't fail.
8172 2018-06-04 Carl Love <cel@us.ibm.com>
8174 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
8175 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
8176 as it is slightly cheaper.
8177 (first_match_or_eos_index_<mode>):
8178 Calculate index using natural element order.
8179 (first_match_index_<mode>):
8180 Calculate index using natural element order.
8181 (first_match_or_eos_index_<mode>):
8182 Calculate index using natural order.
8183 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
8184 for BE and LE modes.
8185 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
8186 P9V_BUILTIN_VCLZLSBB_V16QI.
8187 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
8190 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8192 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
8193 indentation and line wrap for many prototypes. Add missing
8194 @smallexample directives around block of prototypes for vec_xl and
8197 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
8199 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
8200 track if we pass or return IEEE 128-bit floating point.
8201 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
8202 C++ mangling that is compatible with GCC 8.1.
8203 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8204 (init_cumulative_args): Note if we pass or return IEEE 128-bit
8205 floating point types.
8206 (rs6000_function_arg_advance_1): Likewise.
8207 (rs6000_mangle_type): Optionally generate mangled names that match
8208 what GCC 8.1 generated for IEEE 128-bit floating point types.
8209 (rs6000_globalize_decl_name): If we have an external function that
8210 passes or returns IEEE 128-bit types, generate a weak reference
8211 from the mangled name used in GCC 8.1 to the current mangled
8213 (rs6000_init_builtins): Make __ibm128 use the long double type if
8214 long double is IBM extended double. Make __float128 use the long
8215 double type if long double is IEEE 128-bit.
8218 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
8219 macro for __ibm128 built-in functions.
8220 (PACK_IF): Add __ibm128 pack/unpack functions.
8221 (UNPACK_IF): Likewise.
8222 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
8223 enable long double built-in functions if long double is IEEE
8224 128-bit floating point.
8225 (rs6000_invalid_builtin): Update long double built-in function
8227 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
8228 functions, adjust the built-in function to use the long double
8229 built-in function if __ibm128 and long double are the same type.
8230 * doc/extend.texi (PowerPC builtins): Update documention for
8231 __builtin_{,un}pack_longdouble. Add documentation for
8232 __builtin_{,un}pack_ibm128.
8234 2018-06-06 Jim Wilson <jimw@sifive.com>
8236 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
8237 (struct machine_function): New field interrupt_mode.
8238 (riscv_handle_type_attribute): New function. Add forward declaration.
8239 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
8240 (riscv_expand_epilogue): Check interrupt_mode field.
8241 (riscv_set_current_function): Check interrupt attribute args and
8242 set interrupt_mode field.
8243 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
8244 (riscv_sret, riscv_uret): New.
8245 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
8246 new arguments to interrupt attribute.
8248 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
8251 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
8252 Don't handle -mcpu=power8 if -mpower9-vector is also used.
8254 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8256 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
8257 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
8258 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8259 several redundant entries.
8261 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8263 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8264 type from "rtx" to "rtx_insn *".
8265 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8266 for local "call_insn", removing cast.
8267 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8269 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8271 PR tree-optimization/86066
8272 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8273 for BIT_INSERT_EXPR stores.
8275 2018-06-06 Richard Biener <rguenther@suse.de>
8277 PR tree-optimization/86062
8278 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8279 component refs ontop
8280 of to be offsetted base.
8282 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8284 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8285 to be static and remove check on interrupt attribute name.
8287 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8289 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8290 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8292 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8295 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8297 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8298 Remove second argument, change how error is called.
8299 (aarch64_layout_arg): Remove second argument from
8300 aarch64_err_no_fpadvsimd call.
8301 (aarch64_init_cumulative_args): Ditto.
8302 (aarch64_gimplify_va_arg_expr): Ditto.
8303 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8305 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8307 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8308 (*simple_return_indirect_internal<mode>): Rename from
8309 simple_return_indirect_internal. Use W mode iterator.
8310 (rstorssp): New expander.
8311 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8312 (clrssbsy): New expander.
8313 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8315 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8317 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8319 (cmse_check_pointed_object): Likewise.
8321 2018-06-05 Martin Liska <mliska@suse.cz>
8323 PR gcov-profile/47618
8324 * doc/invoke.texi: Document how -fprofile-dir format
8327 2018-06-05 Richard Biener <rguenther@suse.de>
8329 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8330 removal pretend DOM info isn't available so we do not update
8331 it and only remove edges, not dominated blocks. Actually free
8332 DOM info in case we removed something. Remove unreachable blocks.
8333 (mfb_keep_latches): Work with either DOM info or marked backedges.
8334 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8335 first. Mark backedges if DOM info isn't available.
8336 (Re-)compute DOM info after cleanup_control_flow_pre.
8338 2018-06-05 Richard Biener <rguenther@suse.de>
8340 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8341 (locus_discrim_hasher::hash): Adjust.
8342 (locus_discrim_hasher::equal): Likewise.
8343 (next_discriminator_for_locus): Work on line directly.
8344 (same_line_p): Pass in expanded locus1 as well.
8345 (assign_discriminators): Avoid redundant location expansions.
8347 2018-06-05 Richard Biener <rguenther@suse.de>
8349 PR tree-optimization/86046
8350 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8351 if required after clearing TREE_ADDRESSABLE.
8353 2018-06-05 Richard Biener <rguenther@suse.de>
8355 PR tree-optimization/86047
8356 * tree-ssa-loop.c (for_each_index): Glob handling of all
8357 decls and constants and really handle all of them.
8359 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8362 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8363 qualifier_void_pointer and qualifier_const_void_pointer.
8364 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8365 (arm_init_builtins): Handle the above.
8366 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8367 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8368 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8371 2018-06-05 Martin Liska <mliska@suse.cz>
8373 * auto-profile.c (read_autofdo_file): Do not use
8374 gcov_ctr_summary struct.
8375 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8376 * coverage.c (struct counts_entry): Likewise.
8377 (read_counts_file): Read just single summary entry.
8378 (get_coverage_counts): Use gcov_summary struct.
8379 * coverage.h (get_coverage_counts): Likewise.
8380 * gcov-dump.c (dump_working_sets): Likewise.
8381 (tag_summary): Dump just single summary.
8382 * gcov-io.c (gcov_write_summary): Write just histogram
8384 (gcov_read_summary): Read just single summary.
8385 (compute_working_sets): Use gcov_summary struct.
8386 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8387 of GCOV_COUNTERS_SUMMABLE.
8388 (GCOV_COUNTERS_SUMMABLE): Remove.
8389 (GCOV_FIRST_VALUE_COUNTER): Replace with
8390 GCOV_COUNTER_V_INTERVAL.
8391 (struct gcov_ctr_summary): Remove.
8392 (struct gcov_summary): Directly use fields of former
8394 (compute_working_sets): Use gcov_summary struct.
8395 * gcov.c (read_count_file): Do not use ctrs fields.
8396 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8398 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8400 * profile.c: Likewise.
8401 * profile.h: Likewise.
8403 2018-06-05 Martin Liska <mliska@suse.cz>
8405 PR gcov-profile/84846
8406 * gcov.c (output_lines): Print working directory only
8407 in intermediate format.
8409 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8411 * config/s390/s390-builtin-types.def: Add void function type.
8412 * config/s390/s390-builtins.def: Use the function type for the
8415 2018-06-04 Jim Wilson <jimw@sifive.com>
8417 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8419 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8420 and maybe_eh_return. Change regno to unsigned int. Use new args to
8421 handle EH_RETURN_DATA_REGNO registers properly.
8422 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8423 (riscv_expand_epilogue): Update comment. Change argument name and
8424 type. Update code to use new name and type. Pass new args to
8425 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8427 * config/riscv/riscv.md (NORMAL_RETURN): New.
8428 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8429 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8430 (eh_return): Call gen_eh_return_internal and emit barrier.
8431 (eh_return_internal): Call riscv_expand_epilogue.
8433 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8435 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8436 bit_insertion field and declare can_be_merged_into method.
8437 (merged_store_group::can_be_merged_into): New method.
8438 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8439 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8440 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8442 2018-06-04 Richard Biener <rguenther@suse.de>
8444 PR tree-optimization/85955
8445 * builtins.c (fold_builtin_sincos): Convert pointers to
8446 destination to appropriate type before dereferencing.
8448 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8450 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8452 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8454 * expr.c (expand_expr_real_1): Force the operand into memory if
8455 its TYPE_MODE is BLKmode and if there is no integer mode for
8456 the number of bits being extracted.
8458 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8462 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8463 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8465 2018-06-04 Richard Biener <rguenther@suse.de>
8467 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8468 (cleanup_tree_cfg_noloop): ... single caller. Do
8469 start_recording_case_labels later.
8471 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8473 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8474 to _IMMINTRIN_H_INCLUDED.
8475 * config/i386/pconfigintrin.h: Ditto.
8476 * config/i386/waitpkgintrin.h: Ditto.
8477 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8478 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8479 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8480 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8481 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8482 waitpkgintrin.h and cldemoteintrin.h.
8484 2018-06-04 Richard Biener <rguenther@suse.de>
8486 PR tree-optimization/86038
8487 * tracer.c (find_best_successor): Check probability for
8488 being initialized, bail out if not.
8490 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8493 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8494 of bits to ignore when comparing architectures.
8496 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8498 PR tree-optimization/69615
8499 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8500 maximum or minimum of the type, try to merge it also as if
8501 range1 is + [-, x - 1] or + [x + 1, -].
8504 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8506 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8508 PR tree-optimization/86034
8509 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8510 the unsigned bitfield type in a bit insertion sequence if it does not
8511 have a larger precision than the bitfield size.
8512 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8514 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8516 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8518 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8520 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8521 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8522 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8523 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8525 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8527 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8528 Disable -fdelete-null-pointer-checks for ELF toolchain.
8530 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8531 Kito Cheng <kito.cheng@gmail.com>
8533 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8534 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8535 (nds32*-*-*): Add glibc and uclibc conditions.
8536 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8537 (TARGET_EXCEPT_UNWIND_INFO): Define.
8538 * config/nds32/elf.h: New file.
8539 * config/nds32/linux.h: New file.
8540 * config/nds32/nds32-elf.opt: New file.
8541 * config/nds32/nds32-linux.opt: New file.
8542 * config/nds32/nds32-fp-as-gp.c
8543 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8544 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8546 (nds32_asm_file_end): Ditto.
8547 (nds32_print_operand): Ditto.
8548 (nds32_insert_attributes): Ditto.
8549 (nds32_init_libfuncs): New function.
8550 (TARGET_HAVE_TLS): Define.
8551 (TARGET_INIT_LIBFUNCS): Define.
8552 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8554 (TARGET_ELF): Apply different mcmodel setting.
8555 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8556 been migrated into elf.h and linux.h files.
8557 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8558 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8559 (mcmodel): The content has been migrated into nds32-elf.opt and
8560 nds32-linux.opt files.
8561 * config/nds32/t-elf: New file.
8562 * config/nds32/t-linux: New file.
8564 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8565 Shiva Chen <shiva0217@gmail.com>
8567 * config/nds32/constants.md (unspec_volatile_element): Add
8568 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8569 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8571 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8572 (make_pass_nds32_fp_as_gp): Declare.
8573 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8575 (nds32_asm_function_end_prologue): Remove unused asm output.
8576 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8577 (nds32_asm_file_start): Output necessary fp_as_gp information.
8578 (nds32_option_override): Adjust register usage.
8579 (nds32_expand_prologue): Consider fp_as_gp situation.
8580 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8581 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8584 (simple_return): Ditto.
8585 (omit_fp_begin): Output special directive for fp_as_gp.
8586 (omit_fp_end): Output special directive for fp_as_gp.
8587 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8588 mforbid-fp-as-gp): New options.
8590 2018-06-01 Mark Wielaard <mark@klomp.org>
8592 * dwarf2out.c (dwarf2out_finish): Remove generation of
8593 DW_AT_loclists_base.
8595 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8597 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8598 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8599 (struct merged_store_group): Add bit_insertion field.
8600 (dump_char_array): Use standard hexadecimal format.
8601 (merged_store_group::merged_store_group): Set bit_insertion to false.
8602 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8603 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8604 also print the mask in the dump file.
8605 (pass_store_merging::gate): Minor tweak.
8606 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8607 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8608 stores with INTEGER_CST stores.
8609 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8610 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8611 and use it throughout. Generate bit insertion sequences if need be.
8612 (pass_store_merging::process_store): Remove redundant condition.
8613 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8615 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8617 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8618 the 128-bit floating point types. Fix function comment.
8620 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8622 * config/aarch64/aarch64-simd.md
8623 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8625 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8628 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8630 PR tree-optimization/85989
8631 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8633 (backprop::intersect_uses): Check it when deciding whether this
8634 is a backedge reference.
8635 (backprop::process_block): Add each phi to m_visited_phis
8636 after visiting it, then clear it at the end.
8638 2018-06-01 Richard Biener <rguenther@suse.de>
8640 * tree-vectorizer.h (vect_dr_stmt): New function.
8641 (vect_get_load_cost): Adjust.
8642 (vect_get_store_cost): Likewise.
8643 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8644 Use vect_dr_stmt instead of DR_SMTT.
8645 (vect_record_base_alignments): Likewise.
8646 (vect_calculate_target_alignment): Likewise.
8647 (vect_compute_data_ref_alignment): Likewise and make static.
8648 (vect_update_misalignment_for_peel): Likewise.
8649 (vect_verify_datarefs_alignment): Likewise.
8650 (vector_alignment_reachable_p): Likewise.
8651 (vect_get_data_access_cost): Likewise. Pass down
8652 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8653 (vect_get_peeling_costs_all_drs): Likewise.
8654 (vect_peeling_hash_get_lowest_cost): Likewise.
8655 (vect_enhance_data_refs_alignment): Likewise.
8656 (vect_find_same_alignment_drs): Likewise.
8657 (vect_analyze_data_refs_alignment): Likewise.
8658 (vect_analyze_group_access_1): Likewise.
8659 (vect_analyze_group_access): Likewise.
8660 (vect_analyze_data_ref_access): Likewise.
8661 (vect_analyze_data_ref_accesses): Likewise.
8662 (vect_vfa_segment_size): Likewise.
8663 (vect_small_gap_p): Likewise.
8664 (vectorizable_with_step_bound_p): Likewise.
8665 (vect_prune_runtime_alias_test_list): Likewise.
8666 (vect_analyze_data_refs): Likewise.
8667 (vect_supportable_dr_alignment): Likewise.
8668 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8669 (vect_gen_prolog_loop_niters): Likewise.
8670 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8671 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8673 (vect_recog_mask_conversion_pattern): Likewise.
8674 (vect_try_gather_scatter_pattern): Likewise.
8675 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8676 to vect_get_store_cost.
8677 (vect_get_store_cost): Get stmt_info instead of DR.
8678 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8679 (vect_get_load_cost): Get stmt_info instead of DR.
8681 2018-06-01 Richard Biener <rguenther@suse.de>
8684 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8685 vars wrapped in MEM_REFs.
8687 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8689 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8690 Fix subreg tests so that we only return a choice between
8691 GENERAL_REGS and FP_REGS if the original classes included both.
8693 2018-06-01 Richard Biener <rguenther@suse.de>
8696 * tree-ssa-structalias.c (get_function_part_constraint):
8697 Handle NULL fi->decl.
8698 (find_func_aliases_for_call): Properly handle indirect
8699 fi from direct call.
8700 (find_func_clobbers): Likewise.
8701 (ipa_pta_execute): Likewise.
8702 (create_variable_info_for): For functions that are ifunc_resolver
8703 resolve to a varinfo that contains the result of the resolver call.
8704 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8707 2018-05-31 Michael Collison <michael.collison@arm.com>
8709 * config/aarch64/aarch64.md:
8710 (*fix_to_zero_extenddfdi2): New pattern.
8711 * gcc.target/aarch64/fix_extend1.c: New testcase.
8713 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8717 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8718 and BUILT_IN_STRNCMP_EQ.
8719 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8720 BUILT_IN_STRNCMP_EQ.
8721 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8722 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8723 (gimple_fold_builtin): Likewise.
8724 * tree-ssa-strlen.c (compute_string_length): New function.
8725 (determine_min_obsize): New function.
8726 (handle_builtin_string_cmp): New function to handle calls to
8727 string compare functions.
8728 (strlen_optimize_stmt): Add handling to builtin string compare
8730 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8731 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8732 * tree.c (build_common_builtin_nodes): Add new defines of
8733 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8735 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8738 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8739 functions with naked attribute.
8741 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8743 * config/i386/sse.md (avx_vec_concat<mode>):
8744 Substitute concat_tg_mode mode attribute with xtg_mode.
8745 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8746 (concat_tg_mode): Remove mode attribute.
8748 2018-05-31 Martin Sebor <msebor@redhat.com>
8751 * calls.c (alloc_max_size): Correct a logic error/typo.
8752 Treat excessive arguments as infinite. Warn for invalid arguments.
8753 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8755 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8758 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8759 and movx for Haswell.
8761 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8762 Cesar Philippidis <cesar@codesourcery.com>
8765 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8766 when emitting error on private/firstprivate reductions.
8767 * omp-low.c (lower_omp_target): Avoid reference-type processing
8768 on pointers for firstprivate clause.
8770 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8772 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8775 * config/aarch64/aarch64-simd.md
8776 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8777 (aarch64_ld1_x3_<mode>): Likewise
8778 (aarch64_st1x2<VALLDIF:mode>): Likewise
8779 (aarch64_st1_x2_<mode>): Likewise
8780 (aarch64_st1x3<VALLDIF:mode>): Likewise
8781 (aarch64_st1_x3_<mode>): Likewise
8782 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8783 (vld1_s8_x3): Likewise.
8784 (vld1_u16_x3): Likewise.
8785 (vld1_s16_x3): Likewise.
8786 (vld1_u32_x3): Likewise.
8787 (vld1_s32_x3): Likewise.
8788 (vld1_u64_x3): Likewise.
8789 (vld1_s64_x3): Likewise.
8790 (vld1_f16_x3): Likewise.
8791 (vld1_f32_x3): Likewise.
8792 (vld1_f64_x3): Likewise.
8793 (vld1_p8_x3): Likewise.
8794 (vld1_p16_x3): Likewise.
8795 (vld1_p64_x3): Likewise.
8796 (vld1q_u8_x3): Likewise.
8797 (vld1q_s8_x3): Likewise.
8798 (vld1q_u16_x3): Likewise.
8799 (vld1q_s16_x3): Likewise.
8800 (vld1q_u32_x3): Likewise.
8801 (vld1q_s32_x3): Likewise.
8802 (vld1q_u64_x3): Likewise.
8803 (vld1q_s64_x3): Likewise.
8804 (vld1q_f16_x3): Likewise.
8805 (vld1q_f32_x3): Likewise.
8806 (vld1q_f64_x3): Likewise.
8807 (vld1q_p8_x3): Likewise.
8808 (vld1q_p16_x3): Likewise.
8809 (vld1q_p64_x3): Likewise.
8810 (vst1_s64_x2): Likewise.
8811 (vst1_u64_x2): Likewise.
8812 (vst1_f64_x2): Likewise.
8813 (vst1_s8_x2): Likewise.
8814 (vst1_p8_x2): Likewise.
8815 (vst1_s16_x2): Likewise.
8816 (vst1_p16_x2): Likewise.
8817 (vst1_s32_x2): Likewise.
8818 (vst1_u8_x2): Likewise.
8819 (vst1_u16_x2): Likewise.
8820 (vst1_u32_x2): Likewise.
8821 (vst1_f16_x2): Likewise.
8822 (vst1_f32_x2): Likewise.
8823 (vst1_p64_x2): Likewise.
8824 (vst1q_s8_x2): Likewise.
8825 (vst1q_p8_x2): Likewise.
8826 (vst1q_s16_x2): Likewise.
8827 (vst1q_p16_x2): Likewise.
8828 (vst1q_s32_x2): Likewise.
8829 (vst1q_s64_x2): Likewise.
8830 (vst1q_u8_x2): Likewise.
8831 (vst1q_u16_x2): Likewise.
8832 (vst1q_u32_x2): Likewise.
8833 (vst1q_u64_x2): Likewise.
8834 (vst1q_f16_x2): Likewise.
8835 (vst1q_f32_x2): Likewise.
8836 (vst1q_f64_x2): Likewise.
8837 (vst1q_p64_x2): Likewise.
8838 (vst1_s64_x3): Likewise.
8839 (vst1_u64_x3): Likewise.
8840 (vst1_f64_x3): Likewise.
8841 (vst1_s8_x3): Likewise.
8842 (vst1_p8_x3): Likewise.
8843 (vst1_s16_x3): Likewise.
8844 (vst1_p16_x3): Likewise.
8845 (vst1_s32_x3): Likewise.
8846 (vst1_u8_x3): Likewise.
8847 (vst1_u16_x3): Likewise.
8848 (vst1_u32_x3): Likewise.
8849 (vst1_f16_x3): Likewise.
8850 (vst1_f32_x3): Likewise.
8851 (vst1_p64_x3): Likewise.
8852 (vst1q_s8_x3): Likewise.
8853 (vst1q_p8_x3): Likewise.
8854 (vst1q_s16_x3): Likewise.
8855 (vst1q_p16_x3): Likewise.
8856 (vst1q_s32_x3): Likewise.
8857 (vst1q_s64_x3): Likewise.
8858 (vst1q_u8_x3): Likewise.
8859 (vst1q_u16_x3): Likewise.
8860 (vst1q_u32_x3): Likewise.
8861 (vst1q_u64_x3): Likewise.
8862 (vst1q_f16_x3): Likewise.
8863 (vst1q_f32_x3): Likewise.
8864 (vst1q_f64_x3): Likewise.
8865 (vst1q_p64_x3): Likewise.
8867 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8869 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8870 user_label_prefix to name.
8872 * tree-core.h: Update comment about the format of NAME string
8873 passed to handler in attribute_spec.
8875 * config/msp430/msp430.md: Remove erroneous subreg expression from
8876 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8877 zero_extend{q,h}isi2.
8879 2018-05-30 Borislav Petkov <bp@suse.de>
8881 * doc/extend.texi: Document some architecture specific
8882 constraints and sort entries.
8884 2018-05-30 Martin Sebor <msebor@redhat.com>
8887 * builtins.c (expand_builtin_stpcpy_1): New function.
8888 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8889 only if the former succeeds.
8891 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8893 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8896 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8898 * doc/invoke.texi (-flinker-output): Document
8900 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8902 * passes.c (ipa_write_summaries): Only modify statements if body
8904 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8905 incrementally linking.
8906 (ipa_passes): Likewise.
8907 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8908 pass down resolution info.
8909 * common.opt (flag_incremental_link): Update info.
8910 * gcc.c (plugin specs): Turn flinker-output=* to
8911 -plugin-opt=-linker-output-known
8912 * toplev.c (compile_file): Also cut compilation when doing incremental
8914 * flag-types. (enum lto_partition_model): Add
8915 LTO_LINKER_OUTPUT_NOLTOREL.
8916 (invoke.texi): Add -flinker-output docs.
8917 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8918 link same way as WPA; do not stream in dead initializers.
8920 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8921 darf2out_register_external_decl): Support incremental link.
8923 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8925 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8927 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8929 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8930 it down to simple_object_copy_lto_debug_sections.
8931 (run_gcc): Determine incremental LTO link time and configure
8932 lto1 into non-wpa mode, disable renaming of debug sections.
8934 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8936 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8937 descriptions of various incorrectly documented functions.
8939 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8942 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8943 address check not strict.
8945 2018-05-30 Richard Biener <rguenther@suse.de>
8947 PR tree-optimization/85964
8948 * tracer.c (better_p): Drop initialized count check, we only
8949 call the function with initialized counts now.
8950 (find_best_successor): Do find a best edge if one
8951 has uninitialized count.
8952 (find_best_predecessor): Likewise. Do BB frequency check only
8953 if count is initialized.
8955 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8957 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8958 (aarch64_ldrstr_offset_compare): New.
8959 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8960 load/store orderings.
8961 (aarch64_gen_adjusted_ldpstp): Likewise.
8963 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8965 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8966 Check for subset of GENERAL_REGS and FP_REGS.
8967 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8970 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8972 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8973 and wi::to_poly_offset. Add the current offset and then check
8974 whether the sum fits, rather than using an unchecked addition of
8975 a checked term. Check for a shwi rather than a uhwi.
8976 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8977 (store_constructor): Use poly_int_tree_p.
8978 (expand_expr_real_1): Likewise.
8979 * function.c (assign_temp): Likewise.
8980 * fold-const.c (const_binop): Use poly_int_tree_p and
8982 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8984 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8985 to_poly_offset for MEM offsets.
8986 * ipa-icf.c (sem_variable::equals): Likewise.
8987 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8988 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8989 wi::to_poly_offset for BIT_FIELD_REF offsets.
8990 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8992 * var-tracking.c (emit_note_insn_var_location): Use
8993 tree_to_poly_uint64.
8995 2018-05-29 Jim Wilson <jimw@sifive.com>
8997 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8999 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
9002 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
9003 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
9005 (sse4_1_round<mode>2): Use nonimmediate_operand
9006 for operand 1 predicate.
9008 2018-05-29 Martin Sebor <msebor@redhat.com>
9009 Richard Biener <rguenther@suse.de>
9012 * calls.c (get_size_range): Call determine_value_range instead
9013 of get_value_range..
9014 * tree-vrp.h (determine_value_range): Declared new function.
9015 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
9017 2018-05-29 Richard Biener <rguenther@suse.de>
9019 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
9020 sure to use non-pattern stmts for get_earlier_stmt arguments.
9021 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
9022 called on pattern stmts.
9023 (get_later_stmt): Likewise.
9025 2018-05-29 Martin Liska <mliska@suse.cz>
9027 PR gcov-profile/85759
9028 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
9031 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9033 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
9035 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
9039 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9040 VEC_PACK_FLOAT_EXPR): New tree codes.
9041 * tree-pretty-print.c (op_code_prio): Handle
9042 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9043 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9044 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9045 * tree-inline.c (estimate_operator_cost): Likewise.
9046 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
9047 * fold-const.c (const_binop): Likewise.
9048 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
9049 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9050 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9051 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
9052 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9053 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9054 * expr.c (expand_expr_real_2): Likewise.
9055 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
9056 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
9057 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
9059 * optabs.c (expand_widen_pattern_expr): For
9060 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
9061 sign from result type rather than operand's type.
9062 (expand_binop_directly): For vec_packu_float_optab and
9063 vec_packs_float_optab allow result type to be different from operand's
9065 * optabs-tree.c (optab_for_tree_code): Handle
9066 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9067 VEC_PACK_FLOAT_EXPR. Formatting fixes.
9068 * tree-vect-generic.c (expand_vector_operations_1): Handle
9069 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9070 VEC_PACK_FLOAT_EXPR.
9071 * tree-vect-stmts.c (supportable_widening_operation): Handle
9073 (supportable_narrowing_operation): Handle FLOAT_EXPR.
9074 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
9075 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
9076 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
9077 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
9079 (vec_pack<floatprefix>_float_<mode>): New expander.
9080 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
9082 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
9083 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
9084 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
9085 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
9086 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
9088 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
9089 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
9090 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9091 VEC_PACK_FLOAT_EXPR): Document.
9093 2018-05-29 Richard Biener <rguenther@suse.de>
9095 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
9097 (stmt_vec_info_vec): Make pointer.
9098 (init_stmt_vec_info_vec): Remove.
9099 (free_stmt_vec_info_vec): Likewise.
9100 (set_stmt_vec_info_vec): New function.
9101 (free_stmt_vec_infos): Likewise.
9102 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
9103 (set_vinfo_for_stmt): Likewise.
9104 (get_earlier_stmt): Likewise.
9105 (get_later_stmt): Likewise.
9106 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
9107 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
9108 (vec_info::~vec_info): Free stmt_vec_infos.
9109 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
9110 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
9111 (pass_slp_vectorize::execute): Likewise.
9112 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
9113 (free_stmt_vec_info_vec): Likewise.
9114 (set_stmt_vec_info_vec): New function.
9115 (free_stmt_vec_infos): Likewise.
9116 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
9117 the global stmt_vec_info_vec.
9118 * tree-parloops.c (gather_scalar_reductions): Use
9119 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
9122 2018-05-29 Richard Biener <rguenther@suse.de>
9124 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
9126 2018-05-29 Martin Liska <mliska@suse.cz>
9127 David Malcolm <dmalcolm@redhat.com>
9129 * vec.c (test_reverse): New.
9130 (vec_c_tests): Add new test.
9131 * vec.h (vl_ptr>::reverse): New function.
9133 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
9135 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
9137 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
9140 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
9142 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
9144 2018-05-28 Richard Biener <rguenther@suse.de>
9146 PR tree-optimization/85933
9147 * tree-vect-data-refs.c (vect_record_base_alignments): Only
9148 look at stmts marked as vectorizable.
9150 2018-05-28 Richard Biener <rguenther@suse.de>
9152 PR tree-optimization/85934
9153 * tree-vect-generic.c (expand_vector_operations_1): Hoist
9154 vector boolean check before scalar optimization.
9156 2018-05-28 Jakub Jelinek <jakub@redhat.com>
9158 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
9161 2018-05-28 Mark Wielaard <mark@klomp.org>
9163 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9164 if it is an expression containing a minus sign.
9166 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
9168 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
9170 2018-05-27 Paul Koning <ni1d@arrl.net>
9172 * config/pdp11/pdp11.md (truncsihi2): Remove.
9174 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
9175 Chung-Ju Wu <jasonwucj@gmail.com>
9177 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
9179 (unaligned_store_dw): Ditto.
9180 * config/nds32/nds32-memory-manipulation.c
9181 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
9182 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
9183 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
9184 (emit_setmem_word_loop): Rename to ...
9185 (emit_setmem_doubleword_loop): ... this.
9186 (nds32_gen_dup_4_byte_to_word_value): New function.
9187 (nds32_gen_dup_8_byte_to_double_word_value): New function.
9188 (nds32_expand_setmem_loop): Refine implementation.
9189 (nds32_expand_setmem_loop_v3m): Ditto.
9190 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
9193 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9195 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
9197 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9199 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
9200 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9201 machine->attr_no_prologue_p.
9202 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
9203 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
9204 (nds32_expand_epilogue): Consider attr_naked_p.
9205 (nds32_expand_epilogue_v3pop): Likewise.
9206 (nds32_can_use_return_insn): Likewise.
9207 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
9208 attr_no_prologue_p fields.
9209 * config/nds32/nds32.opt (mret-in-naked-func): New option.
9211 2018-05-27 Jakub Jelinek <jakub@redhat.com>
9214 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
9216 * config/i386/sse.md
9217 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
9219 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
9221 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
9223 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
9225 (*<floatsuffix>floatv2div2sf2): Rename to ...
9226 (*float<floatunssuffix>v2div2sf2): ... this.
9227 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
9228 (float<floatunssuffix>v2div2sf2_mask): ... this.
9229 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
9230 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
9231 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
9233 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
9235 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9237 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9239 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9241 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9243 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
9244 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
9245 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
9246 gen_ufix_truncv8dfv8si2.
9247 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
9248 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
9249 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
9250 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
9251 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
9252 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
9253 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
9254 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
9256 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9260 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9262 2018-05-25 Jim Wilson <jimw@sifive.com>
9264 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9265 * config/riscv/riscv.c (struct machine_function): Add
9266 interrupt_handler_p and attribute_checked_p fields.
9267 (riscv_attribute_table): Add interrupt.
9268 (riscv_interrupt_type_p): New.
9269 (riscv_save_reg_p): Save extra regs for interrupt handler.
9270 (riscv_use_save_libcall): Return false for interrupt handler.
9271 (riscv_first_stack_step): Add forward declaration.
9272 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9273 for interrupt handler with large frame. Use it for saved reg list.
9274 (riscv_expand_prologue): Move flag_stack_usage_info support to
9275 eliminate duplication.
9276 (riscv_expand_epilogue): Generate mret for interrupt handler.
9277 (riscv_epilogue_uses): New.
9278 (riscv_can_use_return_insn): Return false for interrupt handler.
9279 (riscv_function_ok_for_sibcall): Likewise.
9280 (riscv_set_current_function): Add interrupt handler support.
9281 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9282 * config/riscv/riscv.md (UNSPECV_MRET): New.
9284 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9286 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9288 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9290 PR tree-optimization/85712
9291 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9292 this candidate has already been replaced in-situ by a copy.
9294 2018-05-25 Jason Merrill <jason@redhat.com>
9296 PR c++/80485 - inline function non-zero address.
9297 * symtab.c (nonzero_address): Check DECL_COMDAT.
9299 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9302 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9303 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9304 extension of SImode operation. Use const123_operand predicate.
9305 (*saddsi_1): Remove.
9306 (*saddl_se_1): Ditto.
9308 (*ssubl_se_1): Ditto.
9309 * config/alpha/predicates.md (const123_operand): New predicate.
9310 * config/alpha/constraints.md (P): Use IN_RANGE.
9312 2018-05-25 Richard Biener <rguenther@suse.de>
9314 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9316 (ref_maybe_used_by_stmt_p): Likewise.
9317 (stmt_may_clobber_ref_p): Likewise.
9318 (stmt_may_clobber_ref_p_1): Likewise.
9319 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9321 (ref_maybe_used_by_stmt_p): Likewise.
9322 (stmt_may_clobber_ref_p): Likewise.
9323 (stmt_may_clobber_ref_p_1): Likewise.
9324 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9325 the alias oracle to disambiguate DRs with stmts DR analysis
9327 (vect_analyze_data_refs): Do not give up on not analyzable
9328 DRs for BB vectorization. Remove code truncating the dataref
9331 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9334 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9335 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9336 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9338 2018-05-25 Richard Biener <rguenther@suse.de>
9340 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9341 function, combining stmt data ref gathering and fatal analysis
9343 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9344 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9345 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9346 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9347 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9349 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9351 PR tree-optimization/85720
9352 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9353 SCC if all partitions are builtins.
9354 (version_loop_by_alias_check): New parameter. Generate cancelable
9355 runtime alias check if all partitions are builtins.
9356 (distribute_loop): Update call to above function.
9358 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9360 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9361 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9362 (parm_default_def_partition_arg): Ditto.
9363 (set_parm_default_def_partition): Ditto.
9364 (get_parm_default_def_partitions): Ditto and make it static.
9365 (get_undefined_value_partitions): Ditto and make it static.
9366 (remove_ssa_form): Refactor call to init_var_map here.
9367 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9368 computation for loop region.
9369 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9370 (register_default_def): Delete.
9371 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9372 (parm_default_def_partition_arg): Ditto.
9373 (set_parm_default_def_partition): Ditto.
9374 (get_parm_default_def_partitions): Ditto and make it static.
9375 (get_undefined_value_partitions): Ditto and make it static.
9376 (coalesce_with_default, coalesce_with_default): Update comment.
9377 (create_coalesce_list_for_region): New func factored out from
9378 create_outofssa_var_map.
9379 (populate_coalesce_list_for_outofssa): New func factored out from
9380 create_outofssa_var_map and coalesce_ssa_name.
9381 (create_outofssa_var_map): Delete.
9382 (coalesce_ssa_name): Refactor to support live range computation.
9383 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9384 (get_parm_default_def_partitions): Delete.
9385 (get_undefined_value_partitions): Ditto.
9386 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9387 computation for loop region.
9388 (new_tree_live_info, loe_visit_block): Ditto.
9389 (live_worklist, set_var_live_on_entry): Ditto.
9390 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9391 * tree-ssa-live.h (struct _var_map): New fields.
9392 (init_var_map): Change decl.
9393 (region_contains_p): New.
9395 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9397 * tree-ssa-live.h (live_merge_and_clear): Delete.
9399 2018-05-25 Richard Biener <rguenther@suse.de>
9402 * tree-dump.c (dequeue_and_dump): Remove access to removed
9403 operand 2 of a SWITCH_EXPR.
9405 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9407 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9408 multiplication and division.
9409 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9410 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9411 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9412 (cond_udiv_optab, cond_umod_optab): New optabs.
9413 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9414 (IFN_COND_RDIV): New internal functions.
9415 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9416 TRUNC_MOD_EXPR and RDIV_EXPR.
9417 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9418 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9420 (SVE_INT_BINARY): Include mult.
9421 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9422 (optab, sve_int_op): Handle mult.
9423 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9425 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9426 for SVE_INT_BINARY_SD.
9428 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9430 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9431 (optab, sve_int_op): Handle div and udiv.
9432 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9433 for SVE_INT_BINARY_SD.
9434 (*<optab><mode>3): New insn for the same.
9436 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9438 * tree-vect-patterns.c: Include predict.h.
9439 (vect_recog_divmod_pattern): Restrict check for division support
9440 to when optimizing for size.
9442 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9444 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9445 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9446 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9447 (gimple_match_op::set_op): Likewise.
9448 (gimple_resimplify4): Declare.
9449 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9450 (expr::gen_transform): Likewise.
9451 (decision_tree::gen): Generate a simplification routine for 4 operands.
9452 * gimple-match-head.c (gimple_simplify): Add an overload for
9453 4 operands. In the top-level function, handle up to 4 call
9454 arguments and call gimple_resimplify4.
9455 (gimple_resimplify4): New function.
9456 (build_call_internal): Pass a fourth operand.
9457 (maybe_push_to_seq): Likewise.
9458 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9459 Fold VEC_COND_EXPRs of an operation and a default value into
9460 an IFN_COND_* function if possible.
9461 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9463 (SVE_COND_FP_BINARY): Include them.
9464 (optab, sve_fp_op): Handle them.
9465 (SVE_INT_BINARY_REV): New code iterator.
9466 (SVE_COND_FP_BINARY_REV): New int iterator.
9467 (commutative): New int attribute.
9468 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9470 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9472 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9473 (*cond_<optab><mode>): New patterns for reversed operands.
9475 2018-05-25 Richard Biener <rguenther@suse.de>
9477 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9478 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9479 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9480 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9481 * tree-vect-data-refs.c (everywhere): Adjust users.
9482 * tree-vect-loop.c (everywhere): Likewise.
9483 * tree-vect-slp.c (everywhere): Likewise.
9484 * tree-vect-stmts.c (everywhere): Likewise.
9485 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9487 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9489 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9491 (gcc_cv_as_section_exclude): ... this.
9492 Try Solaris as #exclude syntax.
9493 * configure: Regenerate.
9494 * config.in: Regenerate.
9495 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9497 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9498 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9500 * varasm.c (default_elf_asm_named_section): Don't check if
9501 HAVE_GAS_SECTION_EXCLUDE is defined.
9503 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9505 * doc/md.texi: Update the documentation of the cond_* optabs
9506 to mention the new final operand. Fix GET_MODE_NUNITS call.
9507 Describe the scalar case too.
9508 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9509 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9511 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9512 (get_conditional_internal_fn): Update comment.
9513 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9514 accumulator value as a final argument to conditional functions.
9515 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9516 a define_expand and add an "else" operand. Assert for now that
9517 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9518 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9519 (*cond_<optab><mode>): New patterns.
9520 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9521 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9522 (UNSPEC_COND_EOR): Delete.
9523 (optab): Remove associated mappings.
9524 (SVE_INT_BINARY): New code iterator.
9525 (sve_int_op): Remove int attribute and add "minus" to the code
9527 (SVE_COND_INT_OP): Delete.
9528 (SVE_COND_FP_OP): Rename to...
9529 (SVE_COND_FP_BINARY): ...this.
9531 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9533 * optabs.c (can_reuse_operands_p): New function.
9534 (maybe_legitimize_operands): Try to reuse the results for
9537 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9539 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9540 Add {q} suffix to insn mnemonic.
9542 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9544 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9545 (msp430_warn_func_return): New.
9547 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9549 * fold-const.c (tree_nonzero_bits): New function.
9550 * fold-const.h (tree_nonzero_bits): Likewise.
9551 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9552 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9554 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9558 * varasm.c (assemble_alias): Check ifunc_resolver only on
9561 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9564 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9565 when memory input operand is handled.
9567 2018-05-24 Luis Machado <luis.machado@linaro.org>
9569 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9571 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9573 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9575 * match.pd: Delay FMA folds until after vectorization.
9577 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9580 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9581 address check not strict.
9583 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9585 * gimple-match.h (gimple_match_op): New class.
9586 (mprts_hook): Replace parameters with a gimple_match_op *.
9587 (maybe_build_generic_op): Likewise.
9588 (gimple_simplified_result_is_gimple_val): Replace parameters with
9589 a const gimple_match_op *.
9590 (gimple_simplify): Replace code_helper * and tree * parameters with
9591 a gimple_match_op * parameter.
9592 (gimple_resimplify1): Replace code_helper *, tree and tree *
9593 parameters with a gimple_match_op * parameter.
9594 (gimple_resimplify2): Likewise.
9595 (gimple_resimplify3): Likewise.
9596 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9597 parameters with a gimple_match_op * parameter.
9598 * gimple-match-head.c (gimple_simplify): Change prototypes of
9599 auto-generated functions to take a gimple_match_op * instead of
9600 separate code_helper * and tree * parameters. Make the same
9601 change in the top-level overload and update calls to the
9602 gimple_resimplify routines. Update calls to the auto-generated
9603 functions and to maybe_push_res_to_seq in the publicly-facing
9604 operation-specific gimple_simplify overloads.
9605 (gimple_match_op::MAX_NUM_OPS): Define.
9606 (gimple_resimplify1): Replace rcode and ops with a single res_op
9607 parameter. Update call to gimple_simplify.
9608 (gimple_resimplify2): Likewise.
9609 (gimple_resimplify3): Likewise.
9610 (mprts_hook): Replace parameters with a gimple_match_op *.
9611 (maybe_build_generic_op): Likewise.
9612 (build_call_internal): Replace type, nargs and ops with
9613 a gimple_match_op *.
9614 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9615 with a single gimple_match_op *. Update calls to mprts_hook,
9616 build_call_internal and gimple_simplified_result_is_gimple_val.
9617 Factor out code that is common to the tree_code and combined_fn cases.
9618 * genmatch.c (expr::gen_transform): Replace tem_code and
9619 tem_ops with a gimple_match_op called tem_op. Update calls
9620 to the gimple_resimplify functions and maybe_push_res_to_seq.
9621 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9622 res_ops. Update call to the gimple_resimplify functions.
9623 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9624 (decision_tree::gen): Make the functions take a gimple_match_op *
9625 called res_op instead of separate res_code and res_ops parameters.
9626 Update call accordingly.
9627 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9628 and ops with a single res_op parameter. Update calls to
9629 maybe_build_generic_op and maybe_push_res_to_seq.
9630 (fold_stmt_1): Update calls to gimple_simplify and
9631 replace_stmt_with_simplification.
9632 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9633 and gimple_simplified_result_is_gimple_val.
9634 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9636 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9637 with a gimple_match_op *.
9638 (vn_nary_build_or_lookup): Likewise. Update call to
9639 vn_nary_build_or_lookup_1.
9640 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9641 gimple_match_op *. Update calls to the gimple_resimplify routines
9642 and to gimple_simplified_result_is_gimple_val.
9643 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9644 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9645 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9646 (visit_nary_op): Likewise.
9647 (visit_reference_op_load): Likewise.
9649 2018-05-23 Luis Machado <luis.machado@linaro.org>
9651 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9652 modifier for printing the step amount.
9654 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9657 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9660 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9662 * doc/sourcebuild.texi (Endianness): New subsubsection.
9664 2018-05-23 Luis Machado <luis.machado@linaro.org>
9666 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9667 <prefetch_dynamic_strides>: New const bool field.
9668 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9669 prefetch_dynamic_strides.
9670 (exynosm1_prefetch_tune): Likewise.
9671 (thunderxt88_prefetch_tune): Likewise.
9672 (thunderx_prefetch_tune): Likewise.
9673 (thunderx2t99_prefetch_tune): Likewise.
9674 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9676 (aarch64_override_options_internal): Update to set
9677 PARAM_PREFETCH_DYNAMIC_STRIDES.
9678 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9679 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9680 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9681 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9682 prefetch-dynamic-strides setting.
9684 2018-05-23 Luis Machado <luis.machado@linaro.org>
9686 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9687 <minimum_stride>: New const int field.
9688 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9689 minimum_stride field defaulting to -1.
9690 (exynosm1_prefetch_tune): Likewise.
9691 (thunderxt88_prefetch_tune): Likewise.
9692 (thunderx_prefetch_tune): Likewise.
9693 (thunderx2t99_prefetch_tune): Likewise.
9694 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9695 <default_opt_level>: Set to 3.
9696 (aarch64_override_options_internal): Update to set
9697 PARAM_PREFETCH_MINIMUM_STRIDE.
9698 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9699 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9700 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9701 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9702 stride is constant and is below the minimum stride threshold.
9704 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9706 * config/arm/arm-cpus.in (mode26): Delete.
9707 (armv4): Delete mode26 reference.
9708 * config/arm/arm.c (arm_configure_build_target): Delete use of
9711 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9713 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9715 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9716 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9718 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9719 (floatunsdidf2): Ditto.
9721 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9723 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9724 (fixuns_trunc<mode>si2_avx512f): Ditto.
9725 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9726 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9727 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9729 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9731 PR rtl-optimization/79985
9732 * df-scan.c (df_insn_refs_collect): Remove special case for
9733 global registers and asm statements.
9735 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9737 * extend.texi (Global Register Variables): Rewrite the bullet list.
9738 Note that the register is available for allocation. Note that access
9739 via inline asm must use constraints. Add note about async-signal
9740 handlers. Remove paragraph about automagic register selection.
9742 2018-05-23 Richard Biener <rguenther@suse.de>
9744 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9745 of fixed offset from memset VN.
9747 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9749 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9751 (alloc_cand_and_find_basis): Initialize first_interp field.
9752 (slsr_process_mul): Modify first_interp field.
9753 (slsr_process_add): Likewise.
9754 (slsr_process_cast): Modify first_interp field for each new
9756 (slsr_process_copy): Likewise.
9757 (dump_candidate): Dump first_interp field.
9758 (replace_mult_candidate): Process all interpretations, not just
9760 (replace_rhs_if_not_dup): Likewise.
9761 (replace_one_candidate): Likewise.
9763 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9765 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9767 (aarch64_needs_frame_chain): New function.
9768 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9770 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9773 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9774 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9775 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9776 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9777 as true for strict-align.
9778 (aarch64_can_inline_p): Perform checks even when callee has no
9779 attributes to check for strict alignment.
9780 * doc/extend.texi (AArch64 Function Attributes): Document
9782 * doc/invoke.texi: (AArch64 Options): Likewise.
9784 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9786 PR tree-optimization/85853
9787 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9788 the handling of the root of the node to...
9789 (vect_slp_analyze_node_operations_1): ...this new function,
9790 and run the whole thing with the child nodes' def types
9791 set according to their SLP node's def type.
9793 2018-05-23 Richard Biener <rguenther@suse.de>
9796 * tree-data-ref.c (create_runtime_alias_checks): Defer
9797 and ignore overflow warnings.
9799 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9801 PR tree-optimization/85822
9802 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9805 2018-05-23 Richard Biener <rguenther@suse.de>
9807 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9808 memset constants via native_interpret_expr.
9810 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9813 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9815 (cgraph_node::create_alias): Likewise.
9816 (cgraph_node::get_availability): Check ifunc_resolver instead
9817 of looking up ifunc attribute.
9818 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9819 * varasm.c (do_assemble_alias): Likewise.
9820 (assemble_alias): Likewise.
9821 (default_binds_local_p_3): Likewise.
9822 * cgraph.h (cgraph_node): Add ifunc_resolver.
9823 (cgraph_node::only_called_directly_or_aliased_p): Return false
9825 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9827 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9828 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9829 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9830 instead of looking up ifunc attribute.
9832 2018-05-22 Luis Machado <luis.machado@linaro.org>
9834 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9836 2018-05-22 Martin Sebor <msebor@redhat.com>
9839 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9840 only when expasion succeeds.
9841 (expand_builtin_strcmp): Same.
9842 (expand_builtin_strncmp): Same.
9844 2018-05-22 Martin Sebor <msebor@redhat.com>
9846 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9848 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9849 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9851 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9852 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9854 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9855 Add check for legitimate_address.
9856 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9857 (aarch64_swap_ldrstr_operands): New.
9858 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9861 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9862 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9864 * config/aarch64/aarch64.md: New patterns to generate stp
9866 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9867 single words and double words.
9868 (load_pair_sw, load_pair_dw): Likewise.
9869 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9871 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9873 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9874 for different mode ldpstp and add peephole for merged zero stores.
9876 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9878 (aarch64_gen_store_pair): Rename calls to match new patterns.
9879 (aarch64_gen_load_pair): Rename calls to match new patterns.
9880 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9881 (load_pair<DREG:mode><DREG2:mode>): ... This.
9882 (store_pair<mode>): Rename to...
9883 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9884 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9886 (V_INT_EQUIV): Handle SImode.
9887 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9890 2018-05-22 Martin Sebor <msebor@redhat.com>
9893 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9894 or ajust the presumed bound on an operation to avoid unnecessary
9897 2018-05-22 Martin Sebor <msebor@redhat.com>
9899 PR tree-optimization/85826
9900 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9901 assuming that a DECL necesarily has a constant size.
9903 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9906 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9908 2018-05-22 Richard Biener <rguenther@suse.de>
9910 PR tree-optimization/85834
9911 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9912 non-constant and non-zero memset arguments.
9914 2018-05-22 Martin Liska <mliska@suse.cz>
9917 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9919 2018-05-22 Richard Biener <rguenther@suse.de>
9921 PR tree-optimization/85863
9922 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9923 comparisons when vectype is specified.
9924 (vectorizable_condition): Do not specify vectype for
9925 vect_is_simple_cond when SLP vectorizing.
9927 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9930 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9931 define __ibm128 as long double.
9932 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9933 as a distinct type when IEEE 128-bit support is enabled.
9934 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9935 128-bit types to use the correct functions.
9936 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9937 convert between 128-bit floating point types that have different
9938 modes but the same representation, instead of using gen_lowpart to
9940 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9942 (IFKF_reg): New attributes to give the register constraints for
9944 (extend<mode>tf2_internal): New insns to mark an explicit
9945 conversion between 128-bit floating point types that have a
9946 different mode but share the same representation.
9948 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9950 PR tree-optimization/85814
9951 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9952 a null return from get_strinfo when unsharing the next
9953 strinfo in the chain.
9955 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9958 * varasm.c (weak_finish): Clean up weak_decls.
9960 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9962 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9963 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9964 UNSPEC_UADALP values.
9965 * config/aarch64/iterators.md (ABAL): New int iterator.
9968 (sur): Add mappings for the above.
9969 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9971 (aarch64_<sur>abal<mode>_4): Likewise.
9972 (aarch64_<sur>adalp<mode>_3): Likewise.
9973 (<sur>sadv16qi): New define_expand.
9975 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9977 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9978 (*movdf_internal): Ditto.
9979 (*rcpsf2_sse): Ditto.
9980 (*rsqrtsf2_sse): Ditto.
9981 (*sqrt<mode>2_sse): Ditto.
9983 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9985 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9987 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9988 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9989 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9990 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9992 * config/aarch64/arm_neon.h: Likewise.
9993 * config/aarch64/iterators.md (VQ_I): New.
9995 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9997 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9999 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
10002 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10004 * config/nds32/constraints.md (S): New constraint.
10005 * config/nds32/nds32.md (call_internal): Use constraint S.
10006 (call_value_internal): Likewise.
10007 (sibcall_internal): Likewise.
10008 (sibcall_value_internal): Likewise.
10010 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10011 Chung-Ju Wu <jasonwucj@gmail.com>
10013 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
10014 into consideration.
10016 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10017 Chung-Ju Wu <jasonwucj@gmail.com>
10019 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
10020 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
10021 (nds32_rtx_costs_impl): Simplify.
10022 (nds32_address_cost_impl): Simplify.
10023 (nds32_init_rtx_costs): New function.
10024 (nds32_rtx_costs_speed_prefer): Likewise.
10025 (nds32_rtx_costs_size_prefer): Likewise.
10026 (nds32_address_cost_speed_prefer): Likewise.
10027 (nds32_address_cost_speed_fwprop): Likewise.
10028 (nds32_address_cost_size_prefer): Likewise.
10029 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
10030 * config/nds32/nds32.c (nds32_option_override): Use
10031 nds32_init_rtx_costs function.
10033 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10035 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
10036 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
10037 (TARGET_PIPELINE_N8): Likewise.
10038 (TARGET_PIPELINE_N10): Likewise.
10039 (TARGET_PIPELINE_N13): Likewise.
10040 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10042 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
10044 * config/nds32/nds32-fpu.md: Update copyright year.
10046 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10048 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
10050 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10052 * config/nds32/nds32.c
10053 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
10054 * config/nds32/nds32.opt (minline-asm-r15): New option.
10056 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10058 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
10060 * config/nds32/nds32.md (abssi2): New pattern.
10062 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
10064 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10065 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
10066 Merge insn pattern from sse_cvtsi2ss<round_name> and
10067 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
10068 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
10069 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
10070 using SWI48 mode iterator.
10071 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
10072 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
10073 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
10074 pattern from sse_cvttss2si<round_saeonly_name>
10075 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
10076 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
10077 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
10078 using SWI48 mode iterator.
10079 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
10080 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
10081 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
10082 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
10083 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
10084 using SWI48 mode iterator.
10085 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
10086 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
10087 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
10088 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
10089 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
10090 SWI48 mode iterator.
10091 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
10092 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
10093 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
10094 pattern from sse_cvttsd2si<round_saeonly_name>
10095 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
10097 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10099 * config/nds32/nds32-md-auxiliary.c
10100 (nds32_valid_smw_lwm_base_p): Refine.
10101 (nds32_output_smw_single_word): Refine.
10102 (nds32_output_smw_double_word): New.
10103 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
10105 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10107 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
10108 (nds32_output_stack_pop): Refine.
10109 (nds32_expand_unaligned_load): Refine.
10110 (nds32_expand_unaligned_store): Refine.
10112 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10113 Chung-Ju Wu <jasonwucj@gmail.com>
10115 * config/nds32/constants.md: Add TP_REGNUM constant.
10116 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
10117 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
10119 * config/nds32/nds32-doubleword.md: Consider flag_pic.
10120 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
10121 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
10122 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
10123 and PIC code generation.
10124 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
10126 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
10128 * config/nds32/nds32.md: Support TLS and PIC.
10129 * config/nds32/nds32.c: Support TLS and PIC.
10130 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
10131 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
10134 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10136 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
10137 mode with E_ prefix.
10139 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10140 Chung-Ju Wu <jasonwucj@gmail.com>
10142 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
10143 * config/nds32/nds32-md-auxiliary.c
10144 (symbolic_reference_mentioned_p): New.
10145 (nds32_legitimize_ict_address): New.
10146 (nds32_expand_ict_move): New.
10147 (nds32_indirect_call_referenced_p): New.
10148 (nds32_symbol_binds_local_p): Delete.
10149 (nds32_long_call_p): Modify.
10150 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
10151 * config/nds32/nds32-protos.h
10152 (symbolic_reference_mentioned_p): Declare.
10153 (nds32_legitimize_ict_address): Declare.
10154 (nds32_expand_ict_move): Declare.
10155 (nds32_indirect_call_referenced_p): Declare.
10156 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
10157 (nds32_relax_group): Use nds32_ict_const_p as condition.
10158 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
10159 (nds32_asm_file_start): Output ict_model directive in asm code.
10160 (nds32_legitimate_address_p): Consider indirect call.
10161 (nds32_print_operand): Consider indirect call.
10162 (nds32_print_operand_address): Consider indirect call.
10163 (nds32_insert_attributes): Handle "indirect_call" attribute.
10164 (TARGET_LEGITIMATE_ADDRESS_P): Define.
10165 (TARGET_LEGITIMATE_CONSTANT_P): Define.
10166 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
10167 (TARGET_DELEGITIMIZE_ADDRESS): Define.
10168 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
10169 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
10170 (TARGET_ICT_MODEL_SMALL): Define.
10171 (TARGET_ICT_MODEL_LARGE): Define.
10172 * config/nds32/nds32.md (movsi): Consider ict model.
10173 (call, call_value): Consider ict model.
10174 (sibcall, sibcall_value): Consider ict model.
10175 * config/nds32/nds32.opt (mict-model): New option.
10176 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
10179 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
10180 Monk Chiang <sh.chiang04@gmail.com>
10181 Jim Wilson <jimw@sifive.com>
10183 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
10184 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
10185 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
10186 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
10187 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
10188 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
10189 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
10190 compute save_libcall_adjustment properly.
10191 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
10192 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
10193 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
10194 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
10195 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
10196 (ABI_SPEC): Handle mabi=ilp32e.
10197 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
10198 (RVE): Add RVE mask.
10199 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
10200 <-march>: Add rv32e as an example.
10202 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
10205 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
10206 (intra_create_variable_infos): Handle C++ constructors.
10208 2018-05-18 Martin Liska <mliska@suse.cz>
10210 * passes.def: Remove a redundant pass.
10212 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
10215 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10217 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10219 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10221 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
10223 (armv2, armv2a, armv3, armv3m): Delete architectures.
10224 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
10225 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10226 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
10228 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
10229 (*mulsidi3adddi): Likewise.
10230 (mulsidi3): Likewise.
10231 (*mulsidi3_nov6): Likewise.
10232 (umulsidi3): Likewise.
10233 (umulsidi3_nov6): Likewise.
10234 (umaddsidi4): Likewise.
10235 (*umulsidi3adddi): Likewise.
10236 (smulsi3_highpart): Likewise.
10237 (*smulsi3_highpart_nov6): Likewise.
10238 (umulsi3_highpart): Likewise.
10239 (*umulsi3_highpart_nov6): Likewise.
10240 * config/arm/arm.h (arm_arch3m): Delete.
10241 * config/arm/arm.c (arm_arch3m): Delete.
10242 (arm_option_override_internal): Update armv3-related comment.
10243 (arm_configure_build_target): Delete use of isa_bit_mode32.
10244 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
10245 (arm_rtx_costs_internal): Delete check of arm_arch3m.
10246 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
10247 (mulsa3): Likewise.
10248 (mulusa3): Likewise.
10249 * config/arm/arm-protos.h (arm_arch3m): Delete.
10250 * config/arm/arm-tables.opt: Regenerate.
10251 * config/arm/arm-tune.md: Likewise.
10252 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
10253 deleted architectures.
10255 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10257 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
10258 (armv5t, armv5te): New features.
10259 (ARMv5, ARMv5e): Delete fgroups.
10260 (ARMv5t, ARMv5te): Adjust for above changes.
10261 (ARMv6m): Likewise.
10262 (armv5, armv5e): Delete arches.
10263 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10265 (*call_reg_arm): Likewise.
10266 (*call_value_reg_armv5): Likewise.
10267 (*call_value_reg_arm): Likewise.
10268 (*call_symbol): Likewise.
10269 (*call_value_symbol): Likewise.
10270 (*sibcall_insn): Likewise.
10271 (*sibcall_value_insn): Likewise.
10272 (clzsi2): Likewise.
10273 (prefetch): Likewise.
10274 (define_split and define_peephole2 dependent on arm_arch5):
10276 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10278 (TARGET_ARM_QBIT): Likewise.
10279 (TARGET_DSP_MULTIPLY): Likewise.
10280 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10281 (arm_arch5, arm_arch5e): Delete.
10282 (arm_arch5t, arm_arch5te): Declare.
10283 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10284 (arm_arch5t): Declare.
10285 (arm_option_reconfigure_globals): Update for the above.
10286 (arm_options_perform_arch_sanity_checks): Update comment, replace
10287 use of arm_arch5 with arm_arch5t.
10288 (use_return_insn): Likewise.
10289 (arm_emit_call_insn): Likewise.
10290 (output_return_instruction): Likewise.
10291 (arm_final_prescan_insn): Likewise.
10292 (arm_coproc_builtin_available): Likewise.
10293 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10294 arm_arch5e with arm_arch5t and arm_arch5te.
10295 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10296 (arm_arch5t, arm_arch5te): Declare.
10297 * config/arm/arm-tables.opt: Regenerate.
10298 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10299 * config/arm/t-multilib: Likewise.
10300 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10301 instead of arm_arch5.
10302 (*call_reg_thumb1): Likewise.
10303 (*call_value_reg_thumb1_v5): Likewise.
10304 (*call_value_reg_thumb1): Likewise.
10305 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10307 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10309 2018-05-18 Martin Liska <mliska@suse.cz>
10311 PR gcov-profile/84846
10312 * doc/gcov.texi: Document -t option of gcov tool.
10314 2018-05-18 Martin Liska <mliska@suse.cz>
10316 PR gcov-profile/84846
10317 * gcov.c (print_usage): Add new -t option.
10318 (process_args): Handle the option.
10319 (generate_results): Use stdout as output when requested by
10322 2018-05-18 Martin Liska <mliska@suse.cz>
10324 PR gcov-profile/84846
10325 * coverage.c (coverage_init): Write PWD to .gcno file.
10326 * doc/gcov.texi: Document how working directory is printed.
10327 * gcov-dump.c (dump_gcov_file): Print PWD.
10328 * gcov.c (output_intermediate_file): Likewise.
10329 (read_graph_file): Read PWD string.
10330 (output_lines): Print PWD.
10332 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10334 PR middle-end/85817
10335 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10336 for retval and return false if all args to phi are zero.
10338 2018-05-18 Richard Biener <rguenther@suse.de>
10340 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10342 (evrp_dom_walker::before_dom_children): Call it.
10344 2018-05-18 Richard Biener <rguenther@suse.de>
10346 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10347 results when processing array refs with variable index.
10349 2018-05-18 Toon Moene <toon@moene.org>
10351 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10352 directly after that of -floop-interchange. Indicate that both
10353 options are enabled by default when specifying -O3.
10355 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10357 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10358 iterator. Delete separate integer-mode vec_set<mode> expander.
10359 (aarch64_simd_vec_setv2di): Delete.
10360 (vec_setv2di): Delete.
10361 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10362 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10363 the "w, r" alternative.
10365 2018-05-18 Martin Liska <mliska@suse.cz>
10367 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10368 * tree-pass.h (make_pass_lower_switch_O0): New function.
10369 * tree-switch-conversion.c (node_has_low_bound): Remove.
10370 (node_has_high_bound): Likewise.
10371 (node_is_bounded): Likewise.
10372 (class pass_lower_switch): Make it a template type and create
10374 (pass_lower_switch::execute): Add template argument.
10375 (make_pass_lower_switch): New function.
10376 (make_pass_lower_switch_O0): New function.
10377 (do_jump_if_equal): Remove.
10378 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10379 all the hard work to tree optimization passes.
10381 2018-05-18 Martin Liska <mliska@suse.cz>
10383 * dbgcnt.c (limit_low): Renamed from limit.
10384 (limit_high): New variable.
10385 (dbg_cnt_is_enabled): Check for upper limit.
10386 (dbg_cnt): Adjust dumping.
10387 (dbg_cnt_set_limit_by_index): Add new argument for high
10389 (dbg_cnt_set_limit_by_name): Likewise.
10390 (dbg_cnt_process_single_pair): Parse new format.
10391 (dbg_cnt_process_opt): Use strtok.
10392 (dbg_cnt_list_all_counters): Remove 'value' and add
10394 * doc/invoke.texi: Document changes.
10396 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10398 * doc/sourcebuild.texi (scalar_all_fma): Document.
10399 * tree.def (FMA_EXPR): Delete.
10400 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10401 * internal-fn.c (ternary_direct): New macro.
10402 (expand_ternary_optab_fn): Likewise.
10403 (direct_ternary_optab_supported_p): Likewise.
10404 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10405 * builtins.c (fold_builtin_fma): Delete.
10406 (fold_builtin_3): Don't call it.
10407 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10408 * expr.c (expand_expr_real_2): Likewise.
10409 * fold-const.c (operand_equal_p): Likewise.
10410 (fold_ternary_loc): Likewise.
10411 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10412 * gimple.c (DEFTREECODE): Likewise.
10413 * gimplify.c (gimplify_expr): Likewise.
10414 * optabs-tree.c (optab_for_tree_code): Likewise.
10415 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10416 * tree-eh.c (operation_could_trap_p): Likewise.
10417 (stmt_could_throw_1_p): Likewise.
10418 * tree-inline.c (estimate_operator_cost): Likewise.
10419 * tree-pretty-print.c (dump_generic_node): Likewise.
10420 (op_code_prio): Likewise.
10421 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10422 * tree-ssa-operands.c (get_expr_operands): Likewise.
10423 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10424 * fold-const-call.h (fold_fma): Delete.
10425 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10426 CFN_FNMA and CFN_FNMS.
10427 (fold_fma): Delete.
10428 * genmatch.c (combined_fn): New enum.
10429 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10430 (commutative_op): New function.
10431 (commutate): Use it. Handle more than 2 operands.
10432 (dt_operand::gen_gimple_expr): Use commutative_op.
10433 (parser::parse_expr): Allow :c to be used with non-binary
10434 operators if the commutative operand is known.
10435 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10436 CFN_FMS, CFN_FNMA and CFN_FNMS.
10437 (backprop::process_assign_use): Remove FMA_EXPR handling.
10438 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10439 (gen_hsa_fma): New function.
10440 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10441 IFN_FNMA and IFN_FNMS.
10442 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10443 * gimple-fold.h (follow_all_ssa_edges): Declare.
10444 * gimple-fold.c (follow_all_ssa_edges): New function.
10445 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10446 gimple_build interface and use follow_all_ssa_edges to fold the result.
10447 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10448 instead of checking for optabs directly.
10449 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10450 rather than FMA_EXPRs.
10451 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10452 call to IFN_FMA instead of an FMA_EXPR.
10454 2018-05-17 Jim Wilson <jimw@sifive.com>
10456 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10457 sign extended promoted subreg, and the range does not have the sign bit
10458 set, then do a sign extend.
10460 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10461 test, check for sign extended subreg and/or constant operands, and
10462 do a sign extend in that case.
10464 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10466 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10467 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10469 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10470 Change logics_shift_reg to logics_shift_imm.
10471 (thunderx2t99_fp_loadpair_basic): Delete.
10472 (thunderx2t99_fp_storepair_basic): Delete.
10473 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10474 (thunderx2t99_asimd_polynomial): Delete.
10475 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10476 and neon_fp_mul_d_scalar_q.
10477 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10478 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10479 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10480 (thunderx2t99_asimd_lut): Add missing tbl types.
10481 (thunderx2t99_asimd_ext): Delete.
10482 (thunderx2t99_asimd_load1_1_mult): Delete.
10483 (thunderx2t99_asimd_load1_2_mult): Delete.
10484 (thunderx2t99_asimd_load1_ldp): New.
10485 (thunderx2t99_asimd_load1): New.
10486 (thunderx2t99_asimd_load2): Add missing *load2* types.
10487 (thunderx2t99_asimd_load3): New.
10488 (thunderx2t99_asimd_load4): New.
10489 (thunderx2t99_asimd_store1_1_mult): Delete.
10490 (thunderx2t99_asimd_store1_2_mult): Delete.
10491 (thunderx2t99_asimd_store2_mult): Delete.
10492 (thunderx2t99_asimd_store2_onelane): Delete.
10493 (thunderx2t99_asimd_store_stp): New.
10494 (thunderx2t99_asimd_store1): New.
10495 (thunderx2t99_asimd_store2): New.
10496 (thunderx2t99_asimd_store3): New.
10497 (thunderx2t99_asimd_store4): New.
10499 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10501 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10502 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10504 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10505 Segher Boessenkool <segher@kernel.crashing.org>
10508 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10511 2018-05-17 Richard Biener <rguenther@suse.de>
10513 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10514 for pruning loop and prune defs feeding only already visited PHIs.
10516 2018-05-17 Richard Biener <rguenther@suse.de>
10518 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10520 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10521 Richard Biener <rguenther@suse.de>
10523 PR tree-optimization/85793
10524 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10525 for VMAT_ELEMENTWISE.
10527 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10529 * internal-fn.h (lookup_internal_fn): Declare
10530 * internal-fn.c (lookup_internal_fn): New function.
10531 * gimple.c (gimple_build_call_from_tree): Handle calls to
10532 internal functions.
10533 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10534 internal function names.
10535 * tree-pretty-print.c (dump_generic_node): Likewise.
10536 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10538 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10540 * gimple-fold.h (gimple_build): Make the function forms take
10541 combined_fn rather than built_in_function.
10542 (gimple_simplify): Likewise.
10543 * gimple-match-head.c (gimple_simplify): Likewise.
10544 * gimple-fold.c (gimple_build): Likewise.
10545 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10546 rather than gimple_build_call_internal.
10547 (get_initial_defs_for_reduction): Likewise.
10548 (vect_create_epilog_for_reduction): Likewise.
10549 (vectorizable_live_operation): Likewise.
10551 2018-05-17 Martin Liska <mliska@suse.cz>
10553 * gimple-ssa-sprintf.c (format_directive): Do not use
10554 space in between 'G_' and '('.
10556 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10559 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10560 even if the mask is not all ones.
10563 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10565 (ix86_gimple_fold_builtin): Likewise.
10568 * config/i386/i386.c: Include tree-vector-builder.h.
10569 (ix86_vector_shift_count): New function.
10570 (ix86_fold_builtin): Fold shift builtins by scalar count.
10571 (ix86_gimple_fold_builtin): Likewise.
10573 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10574 _mm512_setzero): New intrinsics.
10576 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10577 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10579 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10580 code generation for cases where splatting a value is not useful.
10581 * simplify-rtx.c (simplify_ternary_operation): Simplify
10582 vec_merge across a vec_duplicate and a paradoxical subreg forming
10583 a vector mode to a vec_concat.
10585 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10587 * config.gcc: Support "goldmont-plus".
10588 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10590 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10591 PROCESSOR_GOLDMONT_PLUS.
10592 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10593 (processor_target_table): Add "goldmont-plus".
10594 (PTA_GOLDMONT_PLUS): Define.
10595 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10596 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10597 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10598 (fold_builtin_cpu): Add "goldmont-plus".
10599 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10600 (ix86_option_override_internal): Add "goldmont-plus".
10601 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10602 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10603 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10604 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10606 2018-05-17 Richard Biener <rguenther@suse.de>
10608 PR tree-optimization/85757
10609 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10610 remove defs that only feed that PHI from further processing.
10612 2018-05-16 Jim Wilson <jimw@sifive.com>
10614 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10616 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10618 2018-05-16 Mark Wielaard <mark@klomp.org>
10620 * dwarf2out.c (count_index_strings): New function.
10621 (output_indirect_strings): Call count_index_strings and generate
10622 header for dwarf_version >= 5.
10624 2018-05-16 Mark Wielaard <mark@klomp.org>
10626 * dwarf2out.c (dwarf_FORM): New function.
10627 (set_indirect_string): Use dwarf_FORM.
10628 (reset_indirect_string): Likewise.
10629 (size_of_die): Likewise.
10630 (value_format): Likewise.
10631 (output_die): Likewise.
10632 (add_skeleton_AT_string): Likewise.
10633 (output_macinfo_op): Likewise.
10634 (index_string): Likewise.
10635 (output_index_string_offset): Likewise.
10636 (output_index_string): Likewise.
10637 (count_index_strings): Likewise.
10639 2018-05-16 Carl Love <cel@us.ibm.com>
10641 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10642 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10644 2018-05-16 Martin Jambor <mjambor@suse.cz>
10646 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10647 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10649 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10651 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10652 (fnma<mode>4): Likewise.
10653 (fms<mode>4): Likewise.
10654 (fnms<mode>4): Likewise.
10655 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10656 (aarch64_fnma<mode>4): Likewise.
10657 (aarch64_fms<mode>4): Likewise.
10658 (aarch64_fnms<mode>4): Likewise.
10659 (aarch64_fnmadd<mode>4): Likewise.
10661 2018-05-16 Jason Merrill <jason@redhat.com>
10663 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10665 2018-05-16 Richard Biener <rguenther@suse.de>
10667 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10668 (dump_stmt_cost): Declare.
10669 (add_stmt_cost): Dump cost we add.
10670 (add_stmt_costs): New function.
10671 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10672 No longer exported.
10673 (vect_analyze_stmt): Adjust prototype.
10674 (vectorizable_condition): Likewise.
10675 (vectorizable_live_operation): Likewise.
10676 (vectorizable_reduction): Likewise.
10677 (vectorizable_induction): Likewise.
10678 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10679 cost vector to pass to vectorizable_ and record afterwards.
10680 (vect_model_reduction_cost): Take cost vector argument and adjust.
10681 (vect_model_induction_cost): Likewise.
10682 (vectorizable_reduction): Likewise.
10683 (vectorizable_induction): Likewise.
10684 (vectorizable_live_operation): Likewise.
10685 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10686 SLP_TREE_NUMBER_OF_VEC_STMTS.
10687 (vect_analyze_slp_cost_1): Remove.
10688 (vect_analyze_slp_cost): Likewise.
10689 (vect_slp_analyze_node_operations): Take visited args and
10690 a target cost vector. Avoid processing already visited stmt sets.
10691 (vect_slp_analyze_operations): Use a local cost vector to gather
10692 costs and register those of non-discarded instances.
10693 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10694 (vect_schedule_slp_instance): Remove copying of
10695 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10697 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10698 adding cost. Record cost entry location.
10699 (vect_prologue_cost_for_slp_op): Function to compute cost of
10700 a constant or invariant generated for SLP vect in the prologue,
10701 split out from vect_analyze_slp_cost_1.
10702 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10703 (vect_model_promotion_demotion_cost): Likewise.
10704 (vect_model_store_cost): Likewise, make static.
10705 (vect_model_load_cost): Likewise.
10706 (vectorizable_bswap): Add cost vector arg and adjust.
10707 (vectorizable_call): Likewise.
10708 (vectorizable_simd_clone_call): Likewise.
10709 (vectorizable_conversion): Likewise.
10710 (vectorizable_assignment): Likewise.
10711 (vectorizable_shift): Likewise.
10712 (vectorizable_operation): Likewise.
10713 (vectorizable_store): Likewise.
10714 (vectorizable_load): Likewise.
10715 (vectorizable_condition): Likewise.
10716 (vectorizable_comparison): Likewise.
10717 (can_vectorize_live_stmts): Likewise.
10718 (vect_analyze_stmt): Likewise.
10719 (vect_transform_stmt): Adjust calls to vectorizable_*.
10720 * tree-vectorizer.c: Include gimple-pretty-print.h.
10721 (dump_stmt_cost): New function.
10723 2018-05-16 Richard Biener <rguenther@suse.de>
10725 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10726 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10727 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10728 (check_name): New callback.
10729 (dse_classify_store): Track cycles via a visited bitmap of PHI
10730 defs and simplify handling of in-loop and across loop dead stores
10731 and properly fail for loop-variant refs. Handle byte-tracking with
10732 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10735 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10737 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10738 (vect_get_mask_type_for_stmt): Likewise.
10739 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10741 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10742 to determine the statement's vector type and the vector type that
10743 should be used for calculating nunits. Deal with cases in which
10744 the type has to be deferred.
10745 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10746 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10747 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10748 (vect_determine_vf_for_stmt): New functions, split out from...
10749 (vect_determine_vectorization_factor): ...here.
10750 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10751 (vect_get_mask_type_for_stmt): New functions, split out from
10752 vect_determine_vectorization_factor.
10754 2018-05-16 Richard Biener <rguenther@suse.de>
10756 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10757 verify the [VEC_]COND_EXPR embedded comparison.
10759 2018-05-15 Martin Sebor <msebor@redhat.com>
10761 PR tree-optimization/85753
10762 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10763 RECORD_TYPE in addition to ARRAY_TYPE.
10765 2018-05-15 Martin Sebor <msebor@redhat.com>
10767 PR middle-end/85643
10768 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10770 2018-05-15 Richard Biener <rguenther@suse.de>
10772 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10773 add by_clobber_p one. Change algorithm to collect all defs
10774 representing uses we need to walk and try reducing them to
10775 a single one before failing.
10776 (dse_dom_walker::dse_optimize_stmt): Adjust.
10778 2018-05-13 Mark Wielaard <mark@klomp.org>
10780 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10781 (size_of_loc_descr): Likewise.
10782 (output_loc_operands): Likewise.
10783 (output_loc_operands_raw): Likewise.
10784 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10785 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10786 (hash_loc_operands): Likewise.
10787 (compare_loc_operands): Likewise.
10789 2018-05-14 Mark Wielaard <mark@klomp.org>
10791 * dwarf2out.c (count_index_addrs): New function.
10792 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10794 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10796 PR tree-optimization/83648
10797 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10798 return value as malloc candidate.
10800 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10803 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10804 param as true in call to suggest_attribute.
10806 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10808 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10809 -mreadonly-in-sdata.
10811 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10813 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10815 (aarch64_crypto_aesd_fused): Likewise.
10817 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10819 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10820 (movsi_aarch64): Likewise.
10821 (load_pairsi): Likewise.
10822 (load_pairdi): Likewise.
10823 (store_pairsi): Likewise.
10824 (store_pairdi): Likewise.
10825 (load_pairsf): Likewise.
10826 (load_pairdf): Likewise.
10827 (store_pairsf): Likewise.
10828 (store_pairdf): Likewise.
10829 (zero_extend): Likewise.
10830 (trunc): Swap alternatives.
10831 (fcvt_target): Add '?' to prefer w over r.
10833 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10836 * config/i386/i386.md: Disallow non-commutative arithmetics in
10837 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10838 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10839 in the peephole2 before it.
10841 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10843 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10844 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10845 (ix86_handle_option): Handle -mcldemote.
10846 * config.gcc: New header.
10847 * config/i386/cldemoteintrin.h: New file.
10848 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10849 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10851 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10852 OPTION_MASK_ISA_CLDEMOTE.
10853 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10854 (ix86_valid_target_attribute_inner_p): Ditto.
10855 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10856 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10857 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10858 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10859 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10861 * config/i386/i386.opt: Add -mcldemote.
10862 * config/i386/x86intrin.h: New header.
10863 * doc/invoke.texi: Add -mcldemote.
10865 2018-05-14 Richard Biener <rguenther@suse.de>
10867 * doc/match-and-simplify.texi: Adjust :s documentation.
10869 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10871 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10872 intended memcpy size.
10873 (REORDER_45): Likewise.
10875 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10877 * sort.cc: New file.
10878 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10879 * vec.c (qsort_chk): Use gcc_qsort.
10880 * Makefile.in (OBJS-libcommon): Add sort.o.
10881 (build/sort.o): New target. Use it...
10882 (BUILD_RTL): ... here, and...
10883 (build/gencfn-macros): ... here, and...
10884 (build/genmatch): ... here.
10886 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10887 Chung-Ju Wu <jasonwucj@gmail.com>
10889 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10890 * config/nds32/nds32-graywolf.md: New file.
10891 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10892 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10894 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10895 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10896 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10897 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10898 * config/nds32/pipelines.md: Include n15 settings.
10900 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10901 Chung-Ju Wu <jasonwucj@gmail.com>
10903 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10904 * config/nds32/nds32-n13.md: New file.
10905 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10906 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10908 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10909 * config/nds32/nds32.md (pipeline_model): Add n13.
10910 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10911 * config/nds32/pipelines.md: Include n13 settings.
10913 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10914 Chung-Ju Wu <jasonwucj@gmail.com>
10916 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10917 * config/nds32/nds32-n10.md: New file.
10918 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10919 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10921 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10922 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10923 * config/nds32/nds32.md (pipeline_model): Add n10.
10924 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10925 * config/nds32/pipelines.md: Include n10 settings.
10927 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10928 Kito Cheng <kito.cheng@gmail.com>
10929 Chung-Ju Wu <jasonwucj@gmail.com>
10931 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10932 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10933 Add enum values for DSP extension instructions.
10934 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10936 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10937 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10938 New code iterators.
10939 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10940 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10941 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10942 * config/nds32/nds32-intrinsic.md: Likewise.
10943 * config/nds32/nds32_intrinsic.h: Likewise.
10944 * config/nds32/nds32-md-auxiliary.c: Likewise.
10945 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10946 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10947 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10948 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10949 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10950 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10951 TYPE_DMAC in switch statement.
10952 * config/nds32/nds32.c: New checking and implementation for DSP
10953 extension instructions.
10954 * config/nds32/nds32.h: Likewise.
10955 * config/nds32/nds32.md: Likewise.
10956 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10957 * config/nds32/predicates.md: Implement new predicates for DSP
10960 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10962 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10963 Reformat alternatives and attributes so it is easier to identify
10964 which constraints/attributes go with which instruction.
10965 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10966 (mov<mode>_softfloat32, FMOVE64): Likewise.
10967 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10968 (mov<mode>_softfloat64, FMOVE64): Likewise.
10970 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10972 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10974 (Basic PowerPC Built-in Functions): The new name of the
10975 subsection previously known as "PowerPC Built-in Functions".
10976 (Basic PowerPC Built-in Functions Available on all Configurations):
10978 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10979 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10980 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10981 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10983 2018-05-11 Martin Jambor <mjambor@suse.cz>
10986 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10989 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10992 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10994 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10996 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10997 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10998 (ix86_handle_option): Handle -mwaitpkg.
10999 * config.gcc: New header.
11000 * config/i386/cpuid.h (bit_WAITPKG): New bit.
11001 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
11002 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
11004 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11005 OPTION_MASK_ISA_WAITPKG.
11006 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
11007 (ix86_option_override_internal): Add PTA_WAITPKG.
11008 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
11009 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
11010 IX86_BUILTIN_TPAUSE.
11011 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
11012 __builtin_ia32_umwait and __builtin_ia32_tpause.
11013 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
11014 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
11015 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
11016 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
11017 UNSPECV_TPAUSE): New.
11018 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
11019 * config/i386/i386.opt: Add -mwaitpkg.
11020 * config/i386/waitpkgintrin.h: New file.
11021 * config/i386/x86intrin.h: New header.
11022 * doc/invoke.texi: Add -mwaitpkg.
11024 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11027 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
11029 (cortex-m0): Use armv6s-m isa.
11030 (cortex-m0plus): Likewise.
11031 (cortex-m1): Likewise.
11032 (cortex-m0.small-multiply): Likewise.
11033 (cortex-m0plus.small-multiply): Likewise.
11034 (cortex-m1.small-multiply): Likewise.
11036 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
11037 Jakub Jelinek <jakub@redhat.com>
11039 PR tree-optimization/85692
11040 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
11041 source permute as well.
11043 2018-05-11 Martin Liska <mliska@suse.cz>
11046 * doc/extend.texi: Document LLVM style format for no_sanitize
11049 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11051 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
11052 mode_supports_vsx_dform_quad to mode_supports_dq_form.
11053 (mode_supports_vsx_dform_quad): Likewise.
11054 (mode_supports_vmx_dform): Move these functions to be next to the
11055 other mode_supports functions.
11056 (mode_supports_dq_form): Likewise.
11057 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
11058 mode_supports_dq_form.
11059 (reg_offset_addressing_ok_p): Likewise.
11060 (offsettable_ok_by_alignment): Likewise.
11061 (rs6000_legitimate_offset_address_p): Likewise.
11062 (legitimate_lo_sum_address_p): Likewise.
11063 (rs6000_legitimize_address): Likewise.
11064 (rs6000_legitimize_reload_address): Likewise.
11065 (rs6000_secondary_reload_inner): Likewise.
11066 (rs6000_preferred_reload_class): Likewise.
11067 (rs6000_output_move_128bit): Likewise.
11069 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11071 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
11072 Generate SImode target register for null target.
11073 <case IX86_BUILTIN_XGETBV>: Ditto.
11074 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
11075 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
11077 2018-05-10 Carl Love <cel@us.ibm.com>
11079 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11080 dcbtt and dcbtstt if operands[2] is 0.
11082 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11085 * config/i386/sse.md (usadv64qi): New expander.
11087 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
11089 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
11090 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
11091 -maltivec=be support.
11092 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
11093 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
11094 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
11095 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
11096 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
11097 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
11098 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
11099 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
11100 altivec_vsumsws): Adjust.
11101 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
11102 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
11103 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
11105 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
11106 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
11107 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
11108 (altivec_lve<VI_char>x): Delete expand.
11109 (*altivec_lve<VI_char>x_internal): Rename to...
11110 (altivec_lve<VI_char>x): ... this.
11111 (altivec_lvxl_<mode>): Delete expand.
11112 (*altivec_lvxl_<mode>_internal): Rename to ...
11113 (altivec_lvxl_<mode>): ... this.
11114 (altivec_stvxl_<mode>): Delete expand.
11115 (*altivec_stvxl_<mode>_internal): Rename to ...
11116 (altivec_stvxl_<mode>): ... this.
11117 (altivec_stve<VI_char>x): Delete expand.
11118 (*altivec_stve<VI_char>x_internal): Rename to ...
11119 (altivec_stve<VI_char>x): ... this.
11120 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
11121 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
11122 reduc_plus_scal_<mode>): Adjust.
11123 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
11125 (rs6000_cpu_cpp_builtins): Adjust.
11126 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
11127 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
11128 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
11129 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
11130 -maltivec=be support.
11131 (rs6000_split_vec_extract_var): Adjust.
11132 (rs6000_split_v4si_init): Adjust.
11133 (swap_selector_for_mode): Delete.
11134 (altivec_expand_lvx_be, altivec_expand_stvx_be,
11135 altivec_expand_stvex_be): Delete.
11136 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
11137 -maltivec=be support.
11138 (rs6000_gimple_fold_builtin): Ditto.
11139 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
11141 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
11142 (TARGET_DIRECT_MOVE_64BIT): Adjust.
11143 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
11144 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
11145 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
11146 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
11147 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
11148 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
11149 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
11150 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
11151 anonymous split): Adjust.
11152 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
11153 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
11155 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
11157 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
11158 when --with-gxx-include-dir is also specified.
11159 * configure: Regenerate.
11161 2018-05-09 Jim Wilson <jimw@sifive.com>
11164 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11165 * config/riscv/t-withmultilib: New.
11166 * config/riscv/withmultilib.h: New.
11167 * doc/install.texi: Document RISC-V --with-multilib-list support.
11169 2018-05-09 Richard Biener <rguenther@suse.de>
11171 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
11173 (vect_bb_vectorization_profitable_p): Adjust. Compute
11174 actual scalar cost using the cost vector and the add_stmt_cost
11177 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11179 PR rtl-optimization/85645
11180 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
11181 in the REG_CFA_REGISTER note for LR, don't leave it empty.
11183 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11185 PR rtl-optimization/85645
11186 * shrink-wrap.c (spread_components): Return a boolean saying if
11187 anything was changed.
11188 (try_shrink_wrapping_separate): Iterate spread_components until
11189 nothing changes anymore.
11191 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11193 PR rtl-optimization/85645
11194 * regrename.c (build_def_use): Also kill the chains that include the
11195 destination of a REG_CFA_REGISTER note.
11197 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11199 PR rtl-optimization/85645
11200 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
11201 insn that has a REG_CFA_REGISTER note.
11203 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
11205 * cfgexpand.c (expand_clobber): New function.
11206 (expand_gimple_stmt_1): Use it.
11207 * tree-vect-stmts.c (vect_clobber_variable): New function,
11209 (vectorizable_simd_clone_call): ...here.
11210 (vectorizable_store): Emit a clobber either side of an
11211 IFN_STORE_LANES sequence.
11212 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
11214 2018-05-09 Tom de Vries <tom@codesourcery.com>
11217 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
11218 (define_insn "trap_if_false"): Add exit after trap.
11220 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
11222 PR rtl-optimization/85638
11223 * bb-reorder.c: Include common/common-target.h.
11224 (create_forwarder_block): New function extracted from...
11225 (fix_up_crossing_landing_pad): ...here. Rename into...
11226 (dw2_fix_up_crossing_landing_pad): ...this.
11227 (sjlj_fix_up_crossing_landing_pad): New function.
11228 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
11229 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
11230 from both partitions and exit the loop after one iteration.
11232 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11235 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11237 (Basic PowerPC Built-in Functions): The new name of the
11238 subsection previously known as "PowerPC Built-in Functions".
11239 (Basic PowerPC Built-in Functions Available on all Configurations):
11241 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11243 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11245 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11247 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11250 2018-05-08 Jim Wilson <jimw@sifive.com>
11252 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
11253 (LD_EMUL_SUFFIX): New.
11254 (LINK_SPEC): Use it.
11256 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11258 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11260 (Basic PowerPC Built-in Functions): The new name of the
11261 subsection previously known as "PowerPC Built-in Functions".
11262 (Basic PowerPC Built-in Functions Available on all Configurations):
11264 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11266 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11268 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11270 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11273 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11276 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11277 after cmpelim optimization.
11279 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11281 * config.gcc: Support "goldmont".
11282 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11283 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11284 PROCESSOR_GOLDMONT.
11285 * config/i386/i386.c (m_GOLDMONT): Define.
11286 (processor_target_table): Add "goldmont".
11287 (PTA_GOLDMONT): Define.
11288 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11289 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11290 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11291 (fold_builtin_cpu): Add "goldmont".
11292 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11293 (ix86_option_override_internal): Add "goldmont".
11294 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11295 (processor_type): Add PROCESSOR_GOLDMONT.
11296 * config/i386/i386.md: Add CPU "glm".
11297 * config/i386/glm.md: New file.
11298 * config/i386/x86-tune.def: Add m_GOLDMONT.
11299 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11301 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11304 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11306 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11307 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11308 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11311 * config/i386/i386.c (ix86_fold_builtin): Handle
11312 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11315 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11316 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11318 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11321 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11322 (check_arch): Likewise.
11323 (check_fpu): Return the result rather than printing it.
11324 (end arch): Fix operator precedence.
11325 (end cpu): Likewise.
11326 (END): Print the result from check_fpu.
11328 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11329 Alan Hayward <alan.hayward@arm.com>
11330 David Sherwood <david.sherwood@arm.com>
11332 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11333 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11334 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11335 (*fcmuo<mode>_and): New patterns.
11337 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11339 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11340 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11341 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11342 (cmp_op, sve_imm_con): New code attributes.
11343 (SVE_COND_INT_CMP, imm_con): Delete.
11344 (cmp_op): Remove above unspecs from int attribute.
11345 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11347 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11348 comparison-specific unspecs.
11349 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11350 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11351 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11352 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11353 (*vec_fcm<cmp_op><mode>): Rename to...
11354 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11355 (*vec_fcmuo<mode>): Rename to...
11356 (*fcmuo<mode>): ...this and adjust likewise.
11357 (*pred_fcm<cmp_op><mode>): New pattern.
11358 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11359 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11361 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11363 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11364 (aarch64_emit_sve_predicated_cond): New function.
11365 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11366 (aarch64_emit_unspec_cond_or): Replace with...
11367 (aarch64_emit_sve_or_conds): ...this new function. Use
11368 aarch64_emit_sve_ptrue_op for the individual comparisons and
11369 aarch64_emit_binop to OR them together.
11370 (aarch64_emit_inverted_unspec_cond): Replace with...
11371 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11372 aarch64_emit_sve_ptrue_op for the comparison and
11373 aarch64_emit_unop to invert the result.
11374 (aarch64_expand_sve_vec_cmp_float): Update after the above
11375 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11377 2018-05-07 Nathan Sidwell <nathan@acm.org>
11379 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11380 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11381 (Backwards Compatibility): Likewise.
11383 2018-05-07 Luis Machado <luis.machado@linaro.org>
11387 2018-05-07 Luis Machado <luis.machado@linaro.org>
11389 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11390 <prefetch_dynamic_strides>: New const bool field.
11391 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11392 prefetch_dynamic_strides.
11393 (exynosm1_prefetch_tune): Likewise.
11394 (thunderxt88_prefetch_tune): Likewise.
11395 (thunderx_prefetch_tune): Likewise.
11396 (thunderx2t99_prefetch_tune): Likewise.
11397 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11399 (aarch64_override_options_internal): Update to set
11400 PARAM_PREFETCH_DYNAMIC_STRIDES.
11401 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11402 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11403 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11404 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11405 prefetch-dynamic-strides setting.
11407 2018-05-07 Luis Machado <luis.machado@linaro.org>
11409 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11410 <minimum_stride>: New const int field.
11411 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11412 minimum_stride field.
11413 (exynosm1_prefetch_tune): Likewise.
11414 (thunderxt88_prefetch_tune): Likewise.
11415 (thunderx_prefetch_tune): Likewise.
11416 (thunderx2t99_prefetch_tune): Likewise.
11417 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11418 (aarch64_override_options_internal): Update to set
11419 PARAM_PREFETCH_MINIMUM_STRIDE.
11420 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11421 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11422 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11423 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11424 stride is constant and is below the minimum stride threshold.
11426 2018-05-07 Luis Machado <luis.machado@linaro.org>
11428 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
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 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11473 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11474 the type is addressable. Don't force op into register if it has
11477 2018-05-05 Roland McGrath <mcgrathr@google.com>
11480 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11481 any section for which we don't know a specific type it should have,
11482 regardless of name. Previously this was done only for the exact
11483 names ".init_array", ".fini_array", and ".preinit_array".
11484 (default_elf_asm_named_section): Add comment about
11485 relationship with default_section_type_flags and SECTION_NOTYPE.
11486 (get_section): Don't consider it a type conflict if one side has
11487 SECTION_NOTYPE and the other doesn't, as long as neither has the
11488 SECTION_BSS et al used in the default_section_type_flags logic.
11490 2018-05-05 Tom de Vries <tom@codesourcery.com>
11493 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11494 (workaround_barsyncs): New function.
11495 (nvptx_reorg): Use workaround_barsyncs.
11496 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11497 (define_expand "nvptx_membar_cta"): New define_expand.
11498 (define_insn "*nvptx_membar_cta"): New insn.
11500 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11502 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11503 To improve optimization opportunities.
11504 * builtin-types.def: The new needed builtin types for the above.
11506 2018-05-04 Richard Biener <rguenther@suse.de>
11508 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11509 * gimple-ssa-store-merging.c
11510 (imm_store_chain_info::output_merged_store): Remove redundant create,
11511 release split_store vector contents on failure.
11512 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11513 scalar stmt vector on cache hit.
11515 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11517 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11519 * config.gcc (powerpc-xilinx-eabi*): Remove.
11520 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11522 (fusion_addis_mem_combo_load): Ditto.
11523 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11525 (rs6000_cpu_cpp_builtins): Ditto.
11526 * config/rs6000/rs6000-linux.c
11527 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11528 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11529 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11531 (rs6000_setup_reg_addr_masks): Ditto.
11532 (rs6000_init_hard_regno_mode_ok): Ditto.
11533 (rs6000_option_override_internal): Ditto.
11534 (legitimate_lo_sum_address_p): Ditto.
11535 (rs6000_legitimize_address): Ditto.
11536 (rs6000_legitimize_reload_address): Ditto.
11537 (rs6000_legitimate_address_p): Ditto.
11538 (abi_v4_pass_in_fpr): Ditto.
11539 (setup_incoming_varargs): Ditto.
11540 (rs6000_gimplify_va_arg): Ditto.
11541 (rs6000_split_multireg_move): Ditto.
11542 (rs6000_savres_strategy): Ditto.
11543 (rs6000_emit_prologue_components): Ditto.
11544 (rs6000_emit_epilogue_components): Ditto.
11545 (rs6000_emit_prologue): Ditto.
11546 (rs6000_emit_epilogue): Ditto.
11547 (rs6000_elf_file_end): Ditto.
11548 (rs6000_function_value): Ditto.
11549 (rs6000_libcall_value): Ditto.
11550 * config/rs6000/rs6000.h: Ditto.
11551 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11552 (TARGET_MINMAX): ... this. New.
11553 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11554 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11555 (*movsi_internal1_single): Delete.
11556 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11557 mfpu=, mxilinx-fpu): Delete.
11558 * config/rs6000/singlefp.h: Delete.
11559 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11560 * config/rs6000/t-rs6000: Ditto.
11561 * config/rs6000/t-xilinx: Delete.
11562 * config/rs6000/titan.md: Adjust for fp_type removal.
11563 * config/rs6000/vsx.md: Remove Xilinx FP support.
11564 (VStype_simple): Delete.
11565 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11566 * config/rs6000/xfpu.h: Delete.
11567 * config/rs6000/xfpu.md: Delete.
11568 * config/rs6000/xilinx.h: Delete.
11569 * config/rs6000/xilinx.opt: Delete.
11570 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11571 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11573 2018-05-04 Tom de Vries <tom@codesourcery.com>
11576 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11579 2018-05-04 Richard Biener <rguenther@suse.de>
11581 PR middle-end/85627
11582 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11583 (expand_complex_div_wide): Likewise.
11584 (expand_complex_operations_1): Likewise.
11585 (expand_complex_libcall): Preserve EH info of the original stmt.
11586 (tree_lower_complex): Handle removed blocks.
11587 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11588 on complex multiplication and division libcall builtins.
11590 2018-05-04 Richard Biener <rguenther@suse.de>
11592 PR middle-end/85574
11593 * fold-const.c (negate_expr_p): Restrict negation of operand
11594 zero of a division to when we know that can happen without
11596 (fold_negate_expr_1): Likewise.
11598 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11601 * real.h (real_nextafter): Declare.
11602 * real.c (real_nextafter): New function.
11603 * fold-const-call.c (fold_const_nextafter): New function.
11604 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11605 CASE_CFN_NEXTTOWARD.
11606 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11607 even when arg1_mode is different from arg0_mode.
11609 2018-05-03 Nathan Sidwell <nathan@acm.org>
11611 * doc/extend.texi (Deprecated Features): Remove
11612 -ffriend-injection.
11613 (Backwards Compatibility): Likewise.
11614 * doc/invoke.texi (C++ Language Options): Likewise.
11615 (C++ Dialect Options): Likewise.
11617 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11620 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11621 _mm512_mask_mullox_epi64): New intrinsics.
11623 2018-05-03 Tom de Vries <tom@codesourcery.com>
11626 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11627 dump files): Add offload-tree.
11629 2018-05-03 Richard Biener <rguenther@suse.de>
11631 PR tree-optimization/85615
11632 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11633 to loops not nested in BBs loop father to avoid creating multi-entry
11636 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11638 PR tree-optimization/70291
11639 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11640 arguments. Change return type to tree. Emit libcall as a new
11641 statement rather than replacing existing one when inplace_p is true.
11642 (expand_complex_multiplication_components): New function.
11643 (expand_complex_multiplication): Expand floating-point complex
11644 multiplication using the above.
11645 (expand_complex_division): Rename inner_type parameter to type.
11646 Update expand_complex_libcall call-site.
11647 (expand_complex_operations_1): Update expand_complex_multiplication
11648 and expand_complex_division call-sites.
11650 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11653 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11654 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11655 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11656 the highest significant bit of the shift count mask is clear. In
11657 check whether and[sq]i3 is needed verify that all significant bits
11658 of the shift count other than the highest are set.
11660 2018-05-02 Tom de Vries <tom@codesourcery.com>
11663 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11664 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11665 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11666 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11667 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11668 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11669 __builtin_goacc_parlevel_size.
11671 2018-05-02 Richard Biener <rguenther@suse.de>
11673 PR tree-optimization/85597
11674 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11675 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11678 2018-05-02 Tom de Vries <tom@codesourcery.com>
11681 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11682 dump files): Add ltrans-tree.
11684 2018-05-02 Tom de Vries <tom@codesourcery.com>
11687 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11688 dump files): Add wpa-ipa.
11690 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11692 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11693 powerpc*-*-linux*paired* target.
11694 * config/rs6000/750cl.h: Delete.
11695 * config/rs6000/paired.h: Delete.
11696 * config/rs6000/paired.md: Delete.
11697 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11699 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11700 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11701 comment. Remove paired float support.
11702 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11703 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11705 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11706 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11708 * config/rs6000/rs6000.c: Remove paired float support.
11709 (paired_expand_vector_init, paired_expand_vector_move,
11710 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11711 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11712 paired_expand_builtin, paired_expand_predicate_builtin,
11713 paired_init_builtins): Delete.
11714 * config/rs6000/rs6000.h: Remove paired float support.
11715 * config/rs6000/rs6000.md: Remove paired float support.
11716 (move_from_CR_ov_bit): Delete.
11717 * config/rs6000/rs6000.opt (mpaired): Delete.
11718 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11719 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11721 2018-05-02 Richard Biener <rguenther@suse.de>
11723 PR middle-end/85567
11724 * gimplify.c (gimplify_save_expr): When in SSA form allow
11725 SAVE_EXPRs to compute to SSA vars.
11727 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11730 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11731 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11732 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11733 clobber operands[2], instead use a new pseudo. Formatting fixes.
11735 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11737 PR tree-optimization/85586
11738 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11739 exit early for statements in the same group if the accesses are
11742 2018-05-02 Tom de Vries <tom@codesourcery.com>
11745 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11748 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11750 PR tree-optimization/85143
11751 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11753 2018-05-01 Tom de Vries <tom@codesourcery.com>
11756 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11757 not found" error message.
11759 2018-05-01 Tom de Vries <tom@codesourcery.com>
11762 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11763 * vec.c (test_ordered_remove_if): New function.
11764 (vec_c_tests): Call test_ordered_remove_if.
11765 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11766 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11767 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11768 VEC_ORDERED_REMOVE_IF.
11770 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11772 PR tree-optimization/82665
11773 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11774 pointer subtraction where arguments come from a memchr call.
11776 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11778 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11779 --push-state --as-needed and --pop-state instead of --as-needed and
11780 --no-as-needed if ld supports it.
11781 * configure: Regenerated.
11784 * doc/install.texi2html: Replace _002d with - and _002a with * in
11785 generated html files using sed.
11787 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11790 * gcc-rich-location.c (blank_line_before_p): New function.
11791 (use_new_line): New function.
11792 (gcc_rich_location::add_fixit_insert_formatted): New function.
11793 * gcc-rich-location.h
11794 (gcc_rich_location::add_fixit_insert_formatted): New function.
11796 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11798 * selftest.c (assert_streq): Rename "expected" and "actual" to
11799 "val1" and "val2". Extend NULL-handling to cover both inputs
11800 symmetrically, while still requiring both to be non-NULL for a pass.
11801 * selftest.h (assert_streq): Rename "expected" and "actual" to
11803 (ASSERT_EQ): Likewise.
11804 (ASSERT_EQ_AT): Likewise.
11805 (ASSERT_KNOWN_EQ): Likewise.
11806 (ASSERT_KNOWN_EQ_AT): Likewise.
11807 (ASSERT_NE): Likewise.
11808 (ASSERT_MAYBE_NE): Likewise.
11809 (ASSERT_MAYBE_NE_AT): Likewise.
11810 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11811 the assertion to pass.
11812 (ASSERT_STREQ_AT): Likewise.
11814 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11816 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11817 interaction with -pie.
11819 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11821 * selftest.h: Fix alphabetization of per-source-file selftest
11824 2018-04-30 Jason Merrill <jason@redhat.com>
11826 PR c++/61982 - dead stores to destroyed objects.
11827 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11830 2018-04-30 Jason Merrill <jason@redhat.com>
11832 * tree.c (build_clobber): New.
11833 * tree.h: Declare it.
11834 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11836 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11838 * diagnostic-show-locus.c (layout::layout): Update for
11839 location_get_source_line returning a char_span.
11840 (struct char_span): Move to input.h.
11841 (struct correction): Update for fields in char_span becoming
11843 (struct source_line): Update for location_get_source_line
11844 returning a char_span.
11845 (layout::print_line): Likewise.
11846 * edit-context.c (edited_file::print_content): Likewise.
11847 (edited_file::print_diff_hunk): Likewise.
11848 (edited_file::print_run_of_changed_lines): Likewise.
11849 (edited_file::get_num_lines): Likewise.
11850 (edited_line::edited_line): Likewise.
11851 * final.c (asm_show_source): Likewise.
11852 * input.c (location_get_source_line): Convert return type
11853 from const char * to char_span, losing the final "line_len"
11855 (dump_location_info): Update for the above.
11856 (get_substring_ranges_for_loc): Likewise. Use a char_span
11857 when handling the literal within the line.
11858 (test_reading_source_line): Update for location_get_source_line
11859 returning a char_span.
11860 * input.h (class char_span): Move here from
11861 diagnostic-show-locus.c, converting from a struct to a class.
11862 Make data members private.
11863 (char_span::operator bool): New.
11864 (char_span::length): New.
11865 (char_span::get_buffer): New.
11866 (char_span::operator[]): New.
11867 (char_span::subspan): Make const.
11868 (char_span::xstrdup): New.
11869 (location_get_source_line): Convert return type from const char *
11870 to char_span, losing the final "line_size" param.
11872 2018-04-30 Jan Hubicka <jh@suse.cz>
11874 * lto-wrapper.c (ltrans_priorities): New static var.
11875 (cmp_priority): New.
11876 (run_gcc): Read priorities and if doing parallel build order
11877 the Makefile by them.
11879 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11881 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11883 2018-04-30 Richard Biener <rguenther@suse.de>
11885 * tree-cfg.c (verify_address): Remove base argument, add
11886 flag whether to check TREE_ADDRESSABLE and do that.
11887 (verify_expr): Remove.
11888 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11889 (verify_gimple_assign_single): Likewise.
11890 (verify_gimple_switch): Likewise.
11891 (verify_expr_location_1): Dereference tp once. Add (disabled)
11892 piece from verify_expr.
11893 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11895 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11897 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11899 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11901 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11902 (small_data_pattern): Likewise.
11903 (arc_rewrite_small_data): Likewise.
11904 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11905 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11906 (get_symbol_alignment): New function.
11907 (legitimate_small_data_address_p): Likewise.
11908 (legitimate_scaled_address): Update, call
11909 legitimate_small_data_address_p.
11910 (output_sdata): New static variable.
11911 (arc_print_operand): Update how we handle small data operands.
11912 (arc_print_operand_address): Likewise.
11913 (arc_legitimate_address_p): Update, use
11914 legitimate_small_data_address_p.
11915 (arc_rewrite_small_data_p): Remove.
11916 (arc_rewrite_small_data_1): Likewise.
11917 (arc_rewrite_small_data): Likewise.
11918 (small_data_pattern): Likewise.
11919 (compact_sda_memory_operand): Update to use
11920 legitimate_small_data_address_p and get_symbol_alignment.
11921 (prepare_move_operands): Don't rewite sdata pattern.
11922 (prepare_extend_operands): Remove.
11923 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11925 (zero_extendqisi2): Likewise.
11926 (zero_extendhisi2): Likewise.
11927 (extendqihi2): Likewise.
11928 (extendqisi2): Likewise.
11929 (extendhisi2): Likewise.
11930 (addsi3): Likewise.
11931 (subsi3): Likewise.
11932 (andsi3): Likewise.
11933 * config/arc/constraints.md (Usd): Change it to memory constraint.
11935 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11937 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11938 as source of std instructions.
11939 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11940 allow 6-bit constants as source for store instructions.
11941 (movdi_insn): Update instruction pattern to allow 6-bit constants
11942 as source for store instructions.
11944 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11946 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11948 2018-04-30 Nathan Sidwell <nathan@acm.org>
11949 Sandra Loosemore <sandra@codesourcery.com>
11951 * dumpfile.c (dump_open): Allow '-' for stdout.
11952 * doc/invoke.texi (Developer Options): Document dump filename
11953 determination early. Document stdin/stdout selection.
11955 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11957 Microblaze Target: PIC data text relative
11959 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11960 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11962 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11963 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11964 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11965 New addressing mode for data-text relative position indepenedent code.
11966 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11967 'ADDRESS_SYMBOLIC_TXT_REL'.
11968 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11969 (microblaze_legitimate_pic_operand): Exclude function calls from
11970 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11971 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11973 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11974 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11975 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11976 for 'address + offset'.
11977 (microblaze_expand_prologue): Add new function prologue call for
11979 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11980 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11981 table in case of TARGET_PIC_DATA_TEXT_REL.
11982 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11983 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11984 Add new macros 'UNSPEC_TEXT',
11985 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11986 + exclude function calls from 'UNSPEC_PLT' in case of data text
11988 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11989 new target hook for generating address diff vector tables in case of
11991 * doc/tm.texi : Regenerate.
11992 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11993 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11994 of addr diff vector generation.
11995 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11996 target hook definition.
11997 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11998 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11999 * doc/invoke.texi (Add new pic option): Add new microblaze pic
12000 option for data text relative.
12002 2018-04-30 Richard Biener <rguenther@suse.de>
12004 * tree-chrec.h (evolution_function_is_constant_p): Remove
12006 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
12008 2018-04-30 Richard Biener <rguenther@suse.de>
12011 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12013 2018-04-30 Richard Biener <rguenther@suse.de>
12015 PR tree-optimization/28364
12016 PR tree-optimization/85275
12017 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
12018 copying first exit test.
12020 2018-04-28 Mark Wielaard <mark@klomp.org>
12022 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
12023 dwarf_version >= 5.
12024 (dwarf_AT): Handle DW_AT_addr_base.
12025 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
12027 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
12030 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
12031 (*ashl<dwi>3_doubleword_mask_1): Ditto.
12032 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
12033 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
12035 2018-04-28 Richard Biener <rguenther@suse.de>
12037 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
12038 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
12039 to reflect use. Only add interesting stmts.
12041 2018-04-27 Martin Jambor <mjambor@suse.cz>
12044 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
12045 the jump function allows for passing through aggregate values.
12047 2018-04-27 David Malcolm <dmalcolm@redhat.com>
12049 * input.h (in_system_header_at): Convert from macro to inline
12051 (from_macro_expansion_at): Likewise.
12052 (from_macro_definition_at): Likewise.
12054 2018-04-27 Jeff Law <law@redhat.com>
12056 * config.gcc: Mark tile* targets as deprecated/obsolete.
12058 2018-04-27 Richard Biener <rguenther@suse.de>
12060 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
12063 2018-04-27 Richard Biener <rguenther@suse.de>
12065 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
12067 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
12069 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
12070 with Yd constraint. Set "preferred_for_speed" attribute from
12071 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
12072 with Yd constraint.
12073 (*movdi_internal): Ditto.
12074 (movti_interunit splitters): Remove
12075 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
12076 (movdi_interunit splitters): Ditto.
12077 * config/i386/constraints.md (Ye): Remove.
12078 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
12080 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12083 * config/aarch64/constraints.md (Usg): Limit to 31.
12084 (Usj): Limit to 63.
12086 2018-04-27 Jakub Jelinek <jakub@redhat.com>
12088 PR tree-optimization/85529
12089 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
12090 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
12091 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
12092 zero extension or masking of the MSB bit.
12093 (optimize_range_tests): Add FIRST_BB argument, pass it through
12094 to optimize_range_tests_var_bound.
12095 (maybe_optimize_range_tests, reassociate_bb): Adjust
12096 optimize_range_tests callers.
12098 2018-04-26 Richard Biener <rguenther@suse.de>
12099 Jakub Jelinek <jakub@redhat.com>
12101 * cgraph.h (symbol_table): Just declare debug method here.
12102 * symtab.c (symbol_table::debug): Define.
12104 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
12106 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
12108 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
12110 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
12111 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
12112 (*movdi_internal): Substitute Yi and Yj constraint with x
12113 and Ym and Yn constraint with y constraint. Update "isa"
12114 attribute and set "preferred_for_speed" attribute from
12115 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
12116 (*movsi_internal): Ditto.
12117 (*movdf_internal): Ditto.
12118 (*movsf_internal): Ditto.
12119 (*zero_extendsidi2): Ditto.
12120 * config/i386/sse.md (vec_set<mode>_0): Ditto.
12121 (sse2_loadld): Ditto.
12122 (*vec_extract<ssevecmodelower>_0): Ditto.
12123 (*vec_extractv4si_0_zext_sse4): Ditto.
12124 (vec_concatv2di): Ditto.
12125 (*vec_dup<mode>): Ditto.
12126 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
12127 * config/i386/constraints.md (Yi): Remove.
12132 2018-04-26 Nathan Sidwell <nathan@acm.org>
12134 * dumpfile.c (dump_open): New.
12135 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
12136 (dump_finish): Detect stdio/stderr by value not name.
12138 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
12140 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
12142 2018-04-26 Tom de Vries <tom@codesourcery.com>
12145 * config/nvptx/nvptx.c (verify_neutering_jumps)
12146 (verify_neutering_labels): New function
12147 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12149 2018-04-26 Tom de Vries <tom@codesourcery.com>
12152 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12153 (nvptx_single): Use needs_neutering_p to skip over insns that do not
12156 2018-04-26 Richard Biener <rguenther@suse.de>
12157 Tom de Vries <tom@codesourcery.com>
12160 * lto-streamer-out.c (output_function): Fixup loops if required to match
12161 discovery done in the reader.
12163 2018-04-26 Richard Biener <rguenther@suse.de>
12165 PR tree-optimization/85116
12166 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
12167 have a loop exit from the single latch predecessor. Remove
12168 case of header with just condition.
12169 (ch_base::copy_headers): Exclude infinite loops from any
12171 (pass_ch::execute): Record exits.
12173 2018-04-26 Richard Biener <rguenther@suse.de>
12175 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
12176 prologue cost vector and pass it to vect_get_load_cost.
12177 (vect_get_peeling_costs_all_drs): Likewise.
12178 (vect_peeling_hash_get_lowest_cost): Likewise.
12179 (vect_enhance_data_refs_alignment): Likewise.
12181 2018-04-26 Richard Biener <rguenther@suse.de>
12183 PR middle-end/85450
12184 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
12185 checking of integer<->pointer conversions.
12186 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
12187 sign-/zero-extending pointer types.
12188 (expand_omp_for_static_chunk): Likewise.
12190 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
12191 Jean Lee <xiaoyur347@gmail.com>
12193 * config/mips/mips.c (mips_asan_shadow_offset): New function.
12194 (TARGET_ASAN_SHADOW_OFFSET): Define.
12195 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
12196 true for -fsanitize=address.
12198 2018-04-25 Mark Wielaard <mark@klomp.org>
12200 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
12203 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12205 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
12206 than "alu", remove explicit "memory" and "imm_disp" attributes.
12207 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
12209 PR middle-end/85414
12210 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
12211 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
12212 gen_lowpart_no_emit.
12214 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
12217 * config/i386/i386.c (ix86_expand_builtin): Change memory
12218 operand to XI, extend p0 to Pmode.
12219 * config/i386/i386.md: Change unspec volatile and operand
12220 1 mode to XI, change operand 0 mode to P.
12222 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12224 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
12225 GET_MODE_MASK before any checking.
12226 (nds32_can_use_bset_p): Likewise.
12227 (nds32_can_use_btgl_p): Likewise.
12229 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12231 * config/nds32/nds32-doubleword.md: New define_split pattern for
12232 illegal register number.
12234 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12236 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
12238 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12240 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
12242 2018-04-25 Richard Biener <rguenther@suse.de>
12244 * lto-streamer.h (LTO_major_version): Bump to 8.
12246 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12248 * BASE-VER: Set to 9.0.0.
12250 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
12252 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
12253 in __abskf2 and __powikf2.
12255 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12258 * config/aarch64/constraints.md (Usg, Usj): New constraints.
12259 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
12260 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12261 Use the above on operand 2. Reindent.
12262 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12264 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12267 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12269 * config/i386/i386.opt (mcet): Removed.
12270 * doc/install.texi: Remove -mcet documentation.
12271 * doc/invoke.texi: Likewise.
12273 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12276 * doc/install.texi: Remove -mcet from bootstrap-cet.
12278 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12281 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12282 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12286 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12287 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12288 containing a CONST_VECTOR.
12290 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12292 * doc/install.texi: Update newlib dependency for nvptx.
12294 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12297 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12298 instead of INTVAL when shifting x left.
12300 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12302 PR tree-optimization/85478
12303 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12304 vect_grouped_store_supported for single element vectors.
12306 2018-04-24 Richard Biener <rguenther@suse.de>
12309 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12310 load cost increase to the case of non-constant step.
12312 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12315 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12316 destination if any_malformed_asm.
12318 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12320 PR middle-end/85496
12321 * expr.c (store_field): In the bitfield case, if the value comes from
12322 a function call and is returned in registers by means of a PARALLEL,
12323 do not change the mode of the temporary unless BLKmode and VOIDmode.
12325 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12327 PR rtl-optimization/85423
12328 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12329 dependencies to debug insns when the previous insn is non-debug.
12331 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12333 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12334 enums into a single definition.
12335 (fls): Fix predicates and printing.
12338 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12340 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12341 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12342 and short u6 immediate.
12343 (check_if_valid_sleep_operand): Remove.
12344 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12346 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12348 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12349 flag_always_save_lp condition.
12350 * config/nds32/nds32.opt (malways-save-lp): New option.
12352 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12354 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12355 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12356 * config/nds32/nds32.h
12357 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12358 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12360 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12362 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12363 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12365 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12366 Chung-Ju Wu <jasonwucj@gmail.com>
12368 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12370 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12371 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12373 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12375 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12377 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12379 * config/nds32/nds32-protos.h (nds32_data_alignment,
12380 nds32_local_alignment): Declare.
12381 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12382 nds32_local_alignment): New functions.
12383 (TARGET_CONSTANT_ALIGNMENT): Define.
12384 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12386 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12388 * config/nds32/nds32.c
12389 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12390 (TARGET_MODES_TIEABLE_P): Likewise.
12392 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12394 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12395 level Ofast and Og.
12397 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12398 Chung-Ju Wu <jasonwucj@gmail.com>
12400 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12401 for unaligned access.
12402 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12404 * config/nds32/nds32-intrinsic.md: Likewise.
12405 * config/nds32/nds32_intrinsic.h: Likewise.
12406 * config/nds32/nds32.h (nds32_builtins): Likewise.
12407 * config/nds32/nds32.opt (munaligned-access): New option.
12408 * config/nds32/nds32.c (nds32_asm_file_start): Display
12409 flag_unaligned_access status.
12411 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12413 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12414 -mno-relax is present.
12415 * config/riscv/linux.h (LINK_SPEC): Ditto.
12417 2018-04-20 Martin Sebor <msebor@redhat.com>
12420 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12422 (gimple_fold_builtin_stxcpy_chk): Same.
12423 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12425 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12428 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12429 __powikf2 when long double is IEEE 128-bit.
12431 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12433 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12434 step to make sure stack always aligned.
12436 2018-04-20 Carl Love <cel@us.ibm.com>
12439 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12440 size check for arg0.
12442 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12443 Tom de Vries <tom@codesourcery.com>
12446 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12447 Emit insns for calls too.
12448 (nvptx_find_par): Always look for worker-level predecessor insn.
12449 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12451 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12452 (nvptx_process_pars): Propagate frames for calls.
12454 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12457 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12459 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12460 (ix86_handle_option): Don't handle OPT_mibt.
12461 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12463 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12465 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12466 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12467 (ix86_target_macros): Define __CET__ with flag_cf_protection
12468 for -fcf-protection.
12469 * config/i386/i386.c (isa2_opts): Remove -mibt.
12470 * config/i386/i386.h (TARGET_IBT): Removed.
12471 (TARGET_IBT_P): Likewise.
12472 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12473 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12474 * config/i386/i386.opt (mcet): Update help message.
12475 (mshstk): Likewise.
12477 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12478 -mcet as an alias for -mshstk.
12480 2018-04-20 Richard Biener <rguenther@suse.de>
12482 PR middle-end/85475
12483 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12484 complexity by forcing a single use of the multiply operand.
12486 2018-04-20 Martin Jambor <mjambor@suse.cz>
12489 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12490 recursion dependency to only apply to non-clones.
12492 2018-04-20 Martin Jambor <mjambor@suse.cz>
12495 * ipa-cp.c (create_specialized_node): Check that clones of
12496 self-recursive edges exist during IPA-CP.
12498 2018-04-19 Toon Moene <toon@moene.org>
12500 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12503 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12505 PR tree-optimization/85467
12506 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12507 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12508 VECTOR_CST element to type.
12510 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12513 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12514 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12515 (builtin_longjmp): Likewise.
12516 (save_stack_nonlocal): New pattern.
12517 (restore_stack_nonlocal): Likewise.
12519 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12522 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12523 Replace ASM_OUTPUT_LABEL with fprintf.
12525 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12528 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12529 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12530 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12531 define __IBT__ and __SHSTK__ for -fcf-protection.
12532 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12534 (ix86_trampoline_init): Likewise.
12535 (x86_output_mi_thunk): Likewise.
12536 (ix86_notrack_prefixed_insn_p): Likewise.
12537 (ix86_option_override_internal): Don't disallow -fcf-protection.
12538 * config/i386/i386.md (rdssp<mode>): Also enable for
12540 (incssp<mode>): Likewise.
12541 (nop_endbr): Likewise.
12542 * config/i386/i386.opt (mcet): Change help message to built-in
12545 (mshstk): Likewise.
12546 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12547 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12548 enable CET built-in functions.
12550 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12552 * common/config/i386/i386-common.c
12553 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12554 OPTION_MASK_ISA_MOVDIRI_UNSET,
12555 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12556 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12557 * config.gcc (movdirintrin.h): New header.
12558 * config/i386/cpuid.h (bit_MOVDIRI,
12559 bit_MOVDIR64B): New bits.
12560 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12562 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12563 (VOID, PVOID, PCVOID)): New function types.
12564 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12565 __builtin_ia32_directstoreu_u64,
12566 __builtin_ia32_movdir64b): New builtins.
12567 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12568 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12570 (ix86_valid_target_attribute_inner_p): Ditto.
12571 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12572 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12573 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12574 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12575 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12576 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12577 (movdiri<mode>, movdir64b_<mode>): New.
12578 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12579 * config/i386/immintrin.h: Include movdirintrin.h.
12580 * config/i386/movdirintrin.h: New file.
12581 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12583 2018-04-19 Richard Biener <rguenther@suse.de>
12585 PR middle-end/85455
12586 * cfg.c (clear_bb_flags): When loop state says we have
12587 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12589 2018-04-19 Richard Biener <rguenther@suse.de>
12591 PR tree-optimization/84737
12592 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12593 copying restrict info.
12594 (vect_setup_realignment): Use it.
12595 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12596 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12597 the first DR to all generated stores.
12598 (vectorizable_load): Likewise for loads.
12600 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12602 PR tree-optimization/85446
12603 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12604 the integral and pointer types to have the same precision.
12606 * doc/install.texi: Document --disable-cet being the default and
12609 2018-04-18 Martin Liska <mliska@suse.cz>
12611 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12614 2018-04-18 Martin Liska <mliska@suse.cz>
12617 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12620 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12621 arguments if they are comparable.
12623 2018-04-18 Martin Liska <mliska@suse.cz>
12626 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12629 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12632 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12635 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12636 ENDBR after calling __morestack.
12638 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12641 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12642 by using gcc_base_ver to generate a gcc_driver_version, and use
12643 it when generating GCC_DRIVER_NAME.
12644 * configure: Regenerate.
12646 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12649 * config.gcc: Obsolete powerpc*-*-*spe*.
12651 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12654 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12655 (stabstr_D): Change type of unum from unsigned int to
12656 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12659 2018-04-17 Jim Wilson <jimw@sifive.com>
12662 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12663 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12664 Set arg_pointer_offset after using pretend_args_size.
12666 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12668 PR rtl-optimization/85431
12669 * dse.c (record_store): Ignore zero width stores.
12672 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12673 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12674 __builtin_stack_restore rather than after it.
12675 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12676 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12677 argument instead of virtual_dynamic_stack_rtx.
12679 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12681 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12683 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12684 Add note to error message to explain internal mapping of overloaded
12685 built-in function name to non-overloaded built-in function name.
12686 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12689 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12692 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12693 where the inputs overlap with the output.
12695 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12697 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12698 (=v, v) alternative and explicit "memory" attribute.
12699 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12700 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12702 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12703 "sselog1" type instead of "sselog".
12704 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12705 "sselog". Remove explicit "memory" attribute.
12706 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12707 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12709 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12710 "sselog1" type instead of "sselog". Remove explicit "memory"
12712 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12713 use "sselog1" type instead of "sselog". Remove explicit "memory"
12715 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12716 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12718 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12719 "sselog1" type instead of "sselog". Remove explicit "memory"
12721 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12722 use "sselog1" type instead of "sselog". Remove explicit "memory"
12726 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12728 PR middle-end/85414
12729 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12732 2018-04-17 Martin Jambor <mjambor@suse.cz>
12735 * ipa-cp.c (create_specialized_node): Call
12736 expand_all_artificial_thunks if necessary.
12738 2018-04-17 Martin Liska <mliska@suse.cz>
12741 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12742 in message, remote space in between '_G' and '('.
12744 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12747 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12748 avx512f_vmcmp<mode>3<round_saeonly_name>,
12749 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12750 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12751 avx512f_rndscale<mode><round_saeonly_name>,
12752 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12753 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12754 Use %<iptr>2 instead of %2 for -masm=intel.
12755 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12756 avx512f_vcvttss2usi<round_saeonly_name>,
12757 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12759 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12760 avx512f_vcvttsd2usi<round_saeonly_name>,
12761 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12762 Use %q1 instead of %1 for -masm=intel.
12763 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12764 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12765 of %3 for -masm=intel.
12766 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12768 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12770 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12772 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12773 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12775 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12777 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12778 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12779 %g1 and one with %0 and %1.
12780 (avx512er_vmrcp28<mode><round_saeonly_name>,
12781 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12782 %1 for -masm=intel.
12783 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12784 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12785 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12786 of %0 and %{%4%} for -masm=intel.
12787 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12788 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12789 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12790 order of %0 and %{%5%}%{z%} for -masm=intel.
12792 2018-04-17 Jan Hubicka <jh@suse.cz>
12795 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12797 2018-04-17 Martin Liska <mliska@suse.cz>
12800 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12801 for target_clone error message. Make default implementation
12802 clone to be a local declaration.
12803 (separate_attrs): Add new argument and check for an empty
12805 (expand_target_clones): Handle it.
12806 (ipa_target_clone): Make redirection just for target_clones
12809 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12810 Tom de Vries <tom@codesourcery.com>
12812 PR middle-end/84955
12813 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12814 tiled basic blocks without omp continue statements.
12816 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12819 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12820 vec_extract expression as having side effects to make sure it gets
12823 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12826 * config/i386/i386.c (get_builtin_code_for_version): Check
12829 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12832 * config.gcc: Support "skylake".
12833 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12835 * config/i386/i386.c (m_SKYLAKE): Define.
12836 (processor_target_table): Add "skylake".
12837 (ix86_option_override_internal): Add "skylake".
12838 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12839 PROCESSOR_CANNONLAKE.
12840 (get_builtin_code_for_version): Fix priority for
12841 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12842 PROCESSOR_SKYLAKE-AVX512.
12843 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12844 (processor_type): Add PROCESSOR_SKYLAKE.
12846 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12847 Jason Merrill <jason@redhat.com>
12850 * convert.c (convert_to_integer_1): Use direct recursion for
12851 enumeral types and types with a precision less than the number
12852 of bits in their mode.
12854 2018-04-16 Julia Koval <julia.koval@intel.com>
12857 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12858 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12860 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12863 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12864 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12865 and -mno-direct-move.
12867 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12870 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12871 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12872 Ensure negative shifts result in {0}.
12874 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12876 PR rtl-optimization/79916
12877 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12878 regs (if any) to define how to gnerate SD moves when LRA is in
12881 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12883 PR rtl-optimization/85393
12884 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12885 * except.c (expand_dw2_landing_pad_for_region): Make static.
12886 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12887 a label and unconditional jump to old_bb, rather than
12888 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12891 PR rtl-optimization/85376
12892 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12893 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12894 instead of a specific value.
12896 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12897 Bin Cheng <bin.cheng@arm.com>
12899 PR tree-optimization/82965
12900 PR tree-optimization/83991
12901 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12902 by_profile_only parameter.
12903 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12904 information if the loop was predicted to iterate too many times.
12905 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12907 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12910 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12913 2018-04-13 Martin Liska <mliska@suse.cz>
12914 Jakub Jelinek <jakub@redhat.com>
12916 PR middle-end/81657
12917 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12918 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12919 * builtins.c (expand_builtin_memory_copy_args): Use
12920 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12921 handle dest_addr == pc_rtx.
12923 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12926 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12927 asked to not generate direct moves.
12928 (fix_trunc<mode>si2_stfiwx): Similar.
12929 (fix_trunc<mode>si2_internal): Similar.
12931 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12934 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12935 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12936 lookup if dest in some wider mode is known to be const0_rtx and
12937 if so, record permanent equivalence for it to be ZERO_EXTEND of
12938 the narrower mode destination.
12940 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12942 * lto-streamer-out.c (output_function): Revert 259346.
12943 * omp-expand.c (expand_oacc_for): Likewise.
12945 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12947 PR rtl-optimization/85354
12948 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12949 * sel-sched.c (sel_global_init): ... here.
12951 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12954 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12955 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12956 mode for PE-COFF targets.
12957 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12958 (i386_pe_asm_lto_end): Likewise.
12959 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12960 (TARGET_ASM_LTO_END): Likewise.
12961 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12962 (i386_pe_asm_lto_start): New function.
12963 (i386_pe_asm_lto_end): Likewise.
12965 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12966 Richard Biener <rguenther@suse.de>
12968 PR middle-end/84955
12969 * lto-streamer-out.c (output_function): Fix CFG loop state before
12971 * omp-expand.c (expand_oacc_for): Handle calls to internal
12972 functions like regular functions.
12974 2018-04-12 Richard Biener <rguenther@suse.de>
12977 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12978 for the early LTO debug to properly generate references to it
12979 during DIE emission. Do not re-use that for the skeleton for
12981 (dwarf2out_early_finish): Likewise.
12983 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12986 * config/i386/sse.md
12987 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12988 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12989 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12990 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12991 and output is a reg, avoid creating invalid lowpart subreg, but
12992 instead split into a 512-bit move. Don't split if not AVX512VL,
12993 input is xmm16+ reg and output is a mem.
12994 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12995 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12996 xmm16+ reg and output is a mem.
12998 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13000 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
13001 also for flag_dwarf2_cfi_asm.
13003 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13005 PR rtl-optimization/85342
13006 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
13007 a bool scalar var inside of the loop instead. Don't try to update
13008 recog_data.operand after failed apply_change_group.
13010 2018-04-12 Tom de Vries <tom@codesourcery.com>
13013 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
13014 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13015 array with flexible array member as array without given dimension.
13016 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
13017 argument for undefined param to true.
13019 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
13022 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13023 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
13024 from PowerPC section.
13025 * config/rs6000/sysv4.opt (mcall-): Improve help text.
13026 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
13027 help text that is too long.
13028 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
13029 help text that is too long.
13030 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
13031 help text that is too long.
13033 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
13035 * config/alpha/alpha.md (stack_probe_internal): Rename
13036 from "probe_stack". Update all callers.
13038 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13040 PR rtl-optimization/84566
13041 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
13042 sched_macro_fuse_insns.
13044 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13047 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
13048 (compute_block_dependences): ... from here.
13050 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13052 PR tree-optimization/85331
13053 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
13054 from int to HOST_WIDE_INT.
13056 2018-04-11 Martin Jambor <mjambor@suse.cz>
13059 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
13060 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
13061 not the same as the source val.
13062 (cgraph_edge_brings_value_p): New parameter.
13063 (gather_edges_for_value): Pass destination value to
13064 cgraph_edge_brings_value_p.
13065 (perhaps_add_new_callers): Likewise.
13066 (get_info_about_necessary_edges): Likewise and exclude values brought
13067 only by self-recursive edges.
13068 (create_specialized_node): Redirect only clones of self-calling edges.
13069 (+self_recursive_pass_through_p): New function.
13070 (find_more_scalar_values_for_callers_subset): Use it.
13071 (find_aggregate_values_for_callers_subset): Likewise.
13072 (known_aggs_to_agg_replacement_list): Removed.
13073 (decide_whether_version_node): Re-calculate known constants for all
13074 remaining context clones.
13076 2018-04-11 Richard Biener <rguenther@suse.de>
13079 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
13080 from early DWARF output.
13081 (dwarf2out_early_finish): Output line info unconditionally into
13082 early DWARF and add reference to it.
13084 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13087 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
13088 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
13089 other than V2DFmode using iptr mode attribute.
13090 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
13092 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13094 PR rtl-optimization/84659
13095 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
13097 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13100 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
13102 (output_loc_list): Pass address of a dummy size variable even in the
13103 locview handling loop.
13104 (index_location_lists): Add comment on why skip_loc_list_entry can't
13107 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
13110 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
13113 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
13116 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13117 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
13118 and -mstring-compare-inline-limit.
13120 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13123 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
13124 for stack clash protection in a register whenever we need it to be in
13127 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13129 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
13130 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
13132 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13135 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
13137 (mlong-double-): Ditto.
13138 * config/rs6000/sysv4.opt (msdata=): Ditto.
13139 (mtls-size=): Ditto.
13141 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13143 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13144 erroneous entries for
13145 "vector int vec_ldl (int, long int *)", and
13146 "vector unsigned int vec_ldl (int, unsigned long int *)".
13147 Add comments and entries for
13148 "vector bool char vec_ldl (int, bool char *)",
13149 "vector bool short vec_ldl (int, bool short *)",
13150 "vector bool int vec_ldl (int, bool int *)",
13151 "vector bool long long vec_ldl (int, bool long long *)",
13152 "vector pixel vec_ldl (int, pixel *)",
13153 "vector long long vec_ldl (int, long long *)",
13154 "vector unsigned long long vec_ldl (int, unsigned long long *)".
13155 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
13156 type tree bool_long_long_type_node and correct definition of
13157 bool_V2DI_type_node to make reference to this new type tree.
13158 (rs6000_mangle_type): Replace erroneous reference to
13159 bool_long_type_node with bool_long_long_type_node.
13160 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
13161 comments to emphasize sign distinctions for char and int types and
13162 replace RS6000_BTI_bool_long constant with
13163 RS6000_BTI_bool_long_long constant. Also add comment to restrict
13164 use of RS6000_BTI_pixel.
13165 (bool_long_type_node): Remove this macro definition.
13166 (bool_long_long_type_node): New macro definition
13168 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13170 PR rtl-optimization/85300
13171 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
13172 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
13173 simplify_unary_operation fails.
13175 2018-04-10 Martin Liska <mliska@suse.cz>
13177 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
13178 cgraph_edge and ipa_ref.
13180 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13184 * config/i386/sse.md
13185 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
13186 computation of the VEC_MERGE selector from mask.
13187 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
13188 Fix decoding of the VEC_MERGE selector into mask.
13190 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13192 PR tree-optimization/85286
13193 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13195 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13197 * final.c (final_1): Set insn_last_address as well as
13198 insn_current_address.
13200 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13203 * explow.c (emit_stack_probe): Call validize_mem on memory location
13204 before passing it to gen_probe_stack. Create address operand and
13205 legitimize it for the probe_stack_address case.
13207 2018-04-09 Jan Hubicka <jh@suse.cz>
13210 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
13211 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
13212 * tree.c (free_lang_data_in_type): Fix handling of binfos;
13214 (free_lang_data): Rebuild type inheritance graph.
13216 2018-04-09 Martin Sebor <msebor@redhat.com>
13218 * invoke.texi (-finline-small-functions): Mention other optimization
13220 (-findirect-inlining, -fpartial-inlining): Same.
13221 (-finline-functions-called-once): Same.
13222 (-freorder-blocks-and-partition): Same.
13224 2018-04-09 Jan Hubicka <jh@suse.cz>
13227 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
13228 jumps; choose last target that matches the criteria (i.e.
13229 no partition changes for non-crossing jumps).
13230 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
13231 support for redirecting crossing jumps to non-crossing.
13233 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
13235 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
13236 also for naked functions.
13238 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
13240 * config/arc/arc.md (add_shift): New pattern.
13241 (add_shift2): Likewise.
13242 (sub_shift): Likewise.
13243 (sub_shift_cmp0_noout): Likewise.
13244 (compare_si_ashiftsi): Likewise.
13245 (xbfu_cmp0_noout): New combine pattern.
13246 (xbfu_cmp0"): Likewise.
13247 (movsi_set_cc_insn): Place the predicable variant first.
13248 (commutative_binary_cmp0_noout): Remove clobber.
13249 (commutative_binary_cmp0): New pattern.
13250 (noncommutative_binary_cmp0): Likewise.
13251 (noncommutative_binary_cmp0_noout): Likewise.
13252 (noncommutative_binary_comparison_result_used): Removed.
13253 (rsub_cmp0): New pattern.
13254 (rsub_cmp0_noout): Likewise.
13255 (extzvsi): Changed, keep only meaningful variants.
13256 (SQH, SEZ): New iterators.
13257 (SQH_postfix): New mode attribute.
13258 (SEZ_prefix): New code attribute.
13259 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
13260 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13261 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13262 of numerical value.
13263 (noncommutative_operator): Check the availability of barrel
13266 2018-04-09 Richard Biener <rguenther@suse.de>
13268 PR tree-optimization/85284
13269 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13270 Only use the niter constraining form of simple_iv when the exit
13271 is always executed.
13273 2018-04-09 Tom de Vries <tom@codesourcery.com>
13276 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13277 (define_expand "*memory_barrier"): New define_expand.
13278 (define_insn "memory_barrier"): New insn.
13280 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13282 PR rtl-optimization/80463
13283 PR rtl-optimization/83972
13284 PR rtl-optimization/83480
13286 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13287 correct producer for the insn.
13288 (tidy_control_flow): Fixup seqnos in case of debug insns.
13290 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13292 PR rtl-optimization/83913
13294 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13295 different sched-times when merging exprs.
13297 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13299 PR rtl-optimization/83962
13301 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13302 tidy_fallthru_edge and tidy_control_flow.
13304 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13306 PR rtl-optimization/83530
13308 * sel-sched.c (force_next_insn): New global variable.
13309 (remove_insn_for_debug): When force_next_insn is true, also leave only
13310 next insn in the ready list.
13311 (sel_sched_region): When the region wasn't scheduled, make another pass
13312 over it with force_next_insn set to 1.
13314 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13316 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13318 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13319 for interrupt control.
13320 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13321 functions for interrupt control.
13322 * config/nds32/nds32-intrinsic.md: Likewise.
13323 * config/nds32/nds32_intrinsic.h: Likewise.
13324 * config/nds32/nds32.h (nds32_builtins): Likewise.
13326 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13328 * config/nds32/nds32.c (nds32_init_machine_status,
13329 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13330 strict_aligned_p field.
13331 (nds32_expand_to_rtl_hook): New function.
13332 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13333 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13335 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13336 Chung-Ju Wu <jasonwucj@gmail.com>
13338 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13339 * config/nds32/nds32-n7.md: New file.
13340 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13341 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13343 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13344 * config/nds32/nds32.md (pipeline_model): Add n7.
13345 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13346 * config/nds32/pipelines.md: Include n7 settings.
13348 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13349 Chung-Ju Wu <jasonwucj@gmail.com>
13351 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13352 * config/nds32/nds32-e8.md: New file.
13353 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13354 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13356 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13357 * config/nds32/nds32.md (pipeline_model): Add e8.
13358 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13359 * config/nds32/pipelines.md: Include e8 settings.
13361 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13362 Chung-Ju Wu <jasonwucj@gmail.com>
13364 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13365 * config/nds32/nds32-n8.md: New file.
13366 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13367 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13369 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13370 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13371 * config/nds32/nds32.md (pipeline_model): Add n8.
13372 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13373 * config/nds32/pipelines.md: Include n8 settings.
13375 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13376 Chung-Ju Wu <jasonwucj@gmail.com>
13378 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13379 * config/nds32/nds32-n9-2r1w.md: New file.
13380 * config/nds32/nds32-n9-3r2w.md: New file.
13381 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13382 nds32_register_ports): New or modify for cpu n9.
13383 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13385 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13386 * config/nds32/nds32-utils.c: New file.
13387 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13388 TARGET_MUL_SLOW): Define.
13389 * config/nds32/nds32.md (pipeline_model): New attribute.
13390 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13391 New options that support cpu n9.
13392 * config/nds32/pipelines.md: Include n9 settings.
13393 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13395 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13397 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13398 information if necessary.
13399 (output_cond_branch_compare_zero): Likewise.
13400 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13401 (nds32_target_alignment): Refine for alignment.
13402 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13403 (FUNCTION_BOUNDARY): Modify.
13404 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13406 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13408 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13410 * config/nds32/constants.md (unspec_volatile_element): Add values for
13411 TLB operation and data prefetch.
13412 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13413 functions for TLB operation and data prefetch.
13414 * config/nds32/nds32-intrinsic.md: Likewise.
13415 * config/nds32/nds32_intrinsic.h: Likewise.
13416 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13417 (nds32_print_operand): Likewise.
13418 * config/nds32/nds32.h (nds32_builtins): Likewise.
13420 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13421 Andrew Pinski <pinsika@gcc.gnu.org>
13423 PR middle-end/82976
13424 * match.pd: Use constant_boolean_node of correct type instead of
13425 boolean_true_node or boolean_false_node for simplifying
13426 pointer comparisons to zero.
13428 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13430 PR tree-optimization/80021
13431 * tree.c (verify_type_variant): Make error call in verify_variant_match
13432 translatable and remove final full stop.
13434 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13436 * config/nds32/constants.md (unspec_volatile_element): Add
13437 UNSPEC_VOLATILE_EH_RETURN.
13438 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13439 nds32_output_stack_pop): Support dwarf exception handling process.
13440 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13441 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13442 exception handling process.
13443 (nds32_compute_stack_frame): Likewise.
13444 (nds32_return_addr_rtx): Likewise.
13445 (nds32_initial_elimination_offset): Likewise.
13446 (nds32_expand_prologue): Likewise.
13447 (nds32_expand_epilogue): Likewise.
13448 (nds32_dynamic_chain_address): New function.
13449 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13450 exception handling.
13451 (DYNAMIC_CHAIN_ADDRESS): Define.
13452 (EH_RETURN_DATA_REGNO): Define.
13453 (EH_RETURN_STACKADJ_RTX): Define.
13454 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13455 patterns for dwarf exception handling.
13457 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13459 * config/nds32/nds32.h: Clean up obsolete macros.
13461 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13463 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13464 Add enum values for particular instructions.
13465 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13466 particular intrinsic functions.
13467 * config/nds32/nds32-intrinsic.md: Likewise.
13468 * config/nds32/nds32_intrinsic.h: Likewise.
13469 * config/nds32/nds32.h (nds32_builtins): Likewise.
13470 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13471 (btst, ave): New patterns for particular instructions.
13473 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13475 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13476 Add enum values for atomic load/store and memory sync.
13477 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13479 * config/nds32/nds32-intrinsic.md: Likewise.
13480 * config/nds32/nds32_intrinsic.h: Likewise.
13481 * config/nds32/nds32.h (nds32_builtins): Likewise.
13483 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13485 PR tree-optimization/85257
13486 * fold-const.c (native_encode_vector): If not all elts could fit
13487 and off is -1, return 0 rather than offset.
13488 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13489 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13490 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13491 adjust buffer in native_interpret_expr call.
13493 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13495 * config/nds32/constants.md (unspec_volatile_element): Add cache
13496 control enum values.
13497 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13498 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13499 * config/nds32/nds32.c (nds32_cctl_names): New.
13500 (nds32_print_operand): Handle cache control register names.
13501 * config/nds32/nds32.h (nds32_builtins): New enum values.
13502 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13504 * config/nds32/nds32.md (type): Add mmu.
13505 * config/nds32/pipelines.md (simple_insn): Add mmu.
13507 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13509 * config/nds32/nds32.md (type): Remove call.
13510 * config/nds32/pipelines.md (simple_insn): Likewise.
13512 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13514 * config/nds32/constants.md (unspec_volatile_element): Add
13515 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13516 UNSPEC_VOLATILE_FMFCFG.
13517 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13518 description for fmfcfg and fmfcsr.
13519 (bdesc_1arg): Add fmtcsr.
13520 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13521 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13522 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13523 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13524 unspec_fmfcfg): New patterns.
13525 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13526 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13527 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13528 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13529 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13530 __nds32__fmfcfg): Define.
13532 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13534 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13535 intrinsic register names.
13536 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13537 intrinsic register enum values and macros.
13539 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13541 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13542 for load/store addressing form.
13543 (nds32_print_operand_address): Likewise.
13545 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13548 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13549 based on LABEL_REF. Remove useless assertion.
13550 (pic_address_needs_scratch): Fix formatting.
13551 (sparc_legitimize_pic_address): Minor tweaks.
13552 (sparc_delegitimize_address): Adjust assertion accordingly.
13553 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13554 into symbolic_operand.
13555 (movsi_high_pic_label_ref): Likewise.
13556 (movsi_lo_sum_pic_label_ref): Likewise.
13557 (movdi_pic_label_ref): Likewise.
13558 (movdi_high_pic_label_ref): Likewise.
13559 (movdi_lo_sum_pic_label_ref): Likewise.
13561 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13563 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13564 custom LIB_SPEC setup.
13566 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13567 Kito Cheng <kito.cheng@gmail.com>
13569 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13570 * config/riscv/freebsd.h: New.
13572 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13574 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13575 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13578 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13579 Kito Cheng <kito.cheng@gmail.com>
13581 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13582 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13583 * config/nds32/nds32-protos.h (nds32_output_call,
13584 nds32_output_return): Declare.
13585 * config/nds32/nds32.md: Refine all the call and return patterns.
13587 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13590 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13591 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13593 PR rtl-optimization/84872
13594 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13595 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13596 EDGE_CROSSING edge.
13598 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13600 * expr.c (copy_blkmode_to_reg): Revert 254862.
13601 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13603 2018-04-06 Richard Biener <rguenther@suse.de>
13605 PR middle-end/85244
13606 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13607 after seeing a component reference with an adjacent field. Treat
13608 refs to arrays at struct end of external decls similar to
13609 refs to unconstrained commons.
13611 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13614 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13615 look through SAVE_EXPRs with non-side-effects argument. Adjust
13617 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13620 2018-04-06 Richard Biener <rguenther@suse.de>
13622 PR middle-end/85180
13623 * alias.c (find_base_term): New wrapper around find_base_term
13624 unwinding CSELIB_VAL_PTR changes.
13625 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13628 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13630 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13632 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13633 constant definitions.
13634 ("nop"): lr 0,0 -> nopr r0
13635 ("nop_lr0", "nop_lr1"): New insn definitions.
13637 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13639 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13640 NDS32_V3PUSH_AVAILABLE_P macro.
13642 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13643 Chung-Ju Wu <jasonwucj@gmail.com>
13645 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13646 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13647 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13648 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13649 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13650 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13651 * config/nds32/constraints.md: New constraints and checking for hard
13652 float configuration.
13653 * config/nds32/iterators.md: New mode iterator and attribute for hard
13654 float configuration.
13655 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13657 * config/nds32/nds32-fpu.md: New file.
13658 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13659 deal with hard float code generation.
13660 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13662 (abi_type, float_reg_number): New enum type.
13663 * config/nds32/nds32-predicates.c: New predicates for hard float.
13664 * config/nds32/nds32-protos.h: Declare functions for hard float.
13665 * config/nds32/nds32.c: Implementation for hard float configuration.
13666 * config/nds32/nds32.h: Definitions for hard float configuration.
13667 * config/nds32/nds32.md: Include hard float machine description and
13668 modify patterns for hard float configuration.
13669 * config/nds32/nds32.opt: New options for hard float configuration.
13670 * config/nds32/predicates.md: New predicates for hard float
13673 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13675 * common/config/nds32/nds32-common.c
13676 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13678 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13680 PR middle-end/85195
13681 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13682 CONSTRUCTOR_ELT (ctor, ...)->value.
13684 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13687 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13689 2018-04-05 Tom de Vries <tom@codesourcery.com>
13692 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13695 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13696 Kito Cheng <kito.cheng@gmail.com>
13698 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13699 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13700 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13701 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13703 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13704 Kito Cheng <kito.cheng@gmail.com>
13706 * config/nds32/constraints.md (Ufe): New memory constraint.
13707 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13708 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13709 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13711 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13712 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13714 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13716 * config/nds32/nds32.md: Use optimize_size in the condition for
13717 alu-shift instructions.
13719 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13721 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13723 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13725 * config/nds32/nds32.md (negsi2): Refine pattern.
13727 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13728 Chung-Ju Wu <jasonwucj@gmail.com>
13730 * config/nds32/iterators.md (shift_rotate): New code iterator.
13731 (shift): New code attribute.
13732 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13733 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13734 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13735 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13736 bit-wise operations.
13737 (andsi3, *andsi3): Ditto.
13738 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13739 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13740 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13741 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13742 nds32_ior_operand, nds32_xor_operand): New predicates.
13744 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13746 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13747 (addsi3, subsi3): ... this.
13749 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13751 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13753 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13755 * config/nds32/nds32.md: Adjust indention.
13757 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13759 * config/nds32/nds32.md (feature): New attribute.
13761 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13763 * config/nds32/nds32.md (subtype): New attribute.
13765 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13768 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13769 expansion to perform a bitwise AND of the argument followed by a
13770 boolean negation of the result.
13772 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13774 PR rtl-optimization/84878
13775 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13776 the basic block. Assert the use reference is not artificial and that
13777 it has an associated insn.
13779 2018-04-04 Michael Matz <matz@suse.de>
13781 * builtins.c (compute_objsize): Pass correct operand
13782 to array_at_struct_end_p.
13784 2018-04-04 Richard Biener <rguenther@suse.de>
13787 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13788 from contexts for DINFO_LEVEL_TERSE and below.
13790 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13792 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13793 resiter_operand condition.
13794 * config/nds32/nds32.md (*move<mode>): Ditto.
13796 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13797 Monk Chiang <sh.chiang04@gmail.com>
13799 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13801 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13803 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13805 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13806 Kito Cheng <kito.cheng@gmail.com>
13808 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13809 nds32_cond_code_str, output_cond_branch,
13810 output_cond_branch_compare_zero, nds32_expand_cbranch,
13811 nds32_expand_cstore, nds32_expand_movcc,
13812 nds32_output_cbranchsi4_equality_zero,
13813 nds32_output_cbranchsi4_equality_reg,
13814 nds32_output_cbranchsi4_equality_reg_or_const_int,
13815 nds32_output_cbranchsi4_greater_less_zero: New functions.
13816 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13817 nds32_expand_cstore, nds32_expand_movcc,
13818 nds32_output_cbranchsi4_equality_zero,
13819 nds32_output_cbranchsi4_equality_reg,
13820 nds32_output_cbranchsi4_equality_reg_or_const_int,
13821 nds32_output_cbranchsi4_greater_less_zero): Declare.
13822 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13823 nds32_rimm11s_operand): New predicates.
13824 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13825 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13828 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13830 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13831 * config/nds32/nds32.md: Ditto.
13832 * config/nds32/pipelines.md: Ditto.
13834 2018-04-04 Richard Biener <rguenther@suse.de>
13836 PR tree-optimization/85168
13837 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13838 propagating abnormals.
13840 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13842 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13844 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13845 Kito Cheng <kito.cheng@gmail.com>
13847 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13848 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13849 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13850 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13851 * config/nds32/nds32.md (sibcall_internal): New.
13852 (sibcall_register): Remove.
13853 (sibcall_immediate): Remove.
13854 (sibcall_value_internal): New.
13855 (sibcall_value_register): Remove.
13856 (sibcall_value_immediate): Remove.
13857 * config/nds32/predicates.md (nds32_general_register_operand): New.
13858 (nds32_call_address_operand): New.
13860 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13862 PR rtl-optimization/85167
13863 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13864 bb_defs if *split_p, instead preinitialize it to NULL.
13866 PR tree-optimization/85156
13867 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13868 evaluating the argument multiple times.
13870 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13872 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13874 (_mm_cvtpd_ps): Likewise.
13875 (_mm_cvttpd_epi32): Likewise.
13876 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13877 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13878 vector, pixel, and bool following altivec.h include.
13880 2018-04-03 Martin Sebor <msebor@redhat.com>
13882 * doc/extend.texi (Common Function Attributes): Clarify.
13883 (const attribute): Likewise.
13884 (pure attribute): Likewise.
13886 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13889 * config/i386/i386.c (ix86_expand_vector_set): Use
13890 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13892 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13894 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13895 instructions when changing rounding bits to preserve precision bits
13896 in the x87 control word.
13898 2018-04-03 Martin Liska <mliska@suse.cz>
13900 PR tree-optimization/82491
13901 * rtl.h (strip_offset_and_add): Replace += suboffset with
13902 poly_uint64 () + suboffset.
13904 2018-03-29 Martin Liska <mliska@suse.cz>
13905 Martin Jambor <mjambor@suse.cz>
13908 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13909 param_type is not an integral or pointer type.
13911 2018-04-03 Richard Biener <rguenther@suse.de>
13913 * sese.h (recompute_all_dominators): Remove.
13915 2018-04-02 Martin Sebor <msebor@redhat.com>
13917 * doc/invoke.texi (-Wrestrict): Fix typos.
13919 2018-04-02 Jim Wilson <jimw@sifive.com>
13921 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13922 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13923 (<optab>di3, <optab>si3_extend): Likewise.
13924 (<optab>si3_mask, <optab>si3_mask_1): New.
13925 (<optab>di3_mask, <optab>di3_mask_1): New.
13926 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13927 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13928 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13930 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13932 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13935 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13937 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13938 (nds32_canonicalize_comparison): New function.
13940 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13941 Kito Cheng <kito.cheng@gmail.com>
13942 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13944 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13945 * config/nds32/constants.md (unspec_volatile_element): Add
13946 UNSPEC_VOLATILE_RELAX_GROUP.
13947 * config/nds32/nds32-relax-opt.c: New file.
13948 * config/nds32/nds32-predicates.c
13949 (nds32_symbol_load_store_p): New function.
13950 * config/nds32/nds32-protos.h
13951 (nds32_symbol_load_store_p): Declare function.
13952 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13953 * config/nds32/nds32.c
13954 (nds32_register_pass): New function to register pass.
13955 (nds32_register_passes): New function to register passes.
13956 * config/nds32/nds32.md (relax_group): New pattern.
13957 * config/nds32/nds32.opt (mrelax-hint): New option.
13958 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13960 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13962 * config/nds32/t-nds32: Modify files dependency.
13964 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13966 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13967 (PROFILE_HOOK): Define its implementation.
13969 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13971 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13972 type and 32-bit size.
13974 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13976 PR middle-end/85090
13977 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13978 (V_128_256): New mode iterator.
13979 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13980 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13981 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13983 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13984 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13986 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13989 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13990 NaN inputs correctly.
13992 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13995 * config/rs6000/vsx.md (??r): New mode attribute.
13996 (*vsx_mov<mode>_64bit): Use it.
13997 (*vsx_mov<mode>_32bit): Likewise.
13999 2018-03-30 Martin Sebor <msebor@redhat.com>
14001 PR tree-optimization/84818
14002 * builtins.c (check_access): Use warning_n.
14004 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14007 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
14009 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
14012 2018-03-30 Julia Koval <julia.koval@intel.com>
14015 * x86-tune.def (movx, partial_reg_dependency): Enable for
14018 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
14020 PR inline-asm/84985
14021 * lra-constraints.c (process_alt_operands): Move setting
14022 this_alternative_matches below.
14024 2018-03-29 Martin Liska <mliska@suse.cz>
14027 * doc/invoke.texi: Document how LTO works with debug info.
14028 Describe auto-load support of binutils. Mention 'x86-64'
14029 as valid option value of -march option.
14031 2018-03-29 Jakub Jelinek <jakub@redhat.com>
14033 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
14036 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
14037 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
14038 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
14041 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14044 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
14045 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
14046 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
14047 for RS6000_BTM_POWERPC64.
14048 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
14049 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
14050 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
14053 (DIVDEU): Likewise.
14055 2018-03-28 Carl Love <cel@us.ibm.com>
14058 2017-09-27 Carl Love <cel@us.ibm.com>
14060 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14061 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14062 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14065 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14067 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14068 instead of __vector bool.
14069 (_mm_max_pu8): Likewise.
14070 (_mm_min_pi16): Likewise.
14072 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14075 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
14076 (DIVWEUO): Likewise.
14077 (DIVDEO): Likewise.
14078 (DIVDEUO): Likewise.
14079 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
14080 DIVWEUO and DIVDEUO.
14081 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
14082 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
14083 (div_extend): Likewise.
14084 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
14086 (__builtin_divweuo): Likewise.
14087 (__builtin_divdeo): Likewise.
14088 (__builtin_divdeuo): Likewise.
14090 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14093 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
14094 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
14096 PR tree-optimization/82004
14097 * gimple-match-head.c (optimize_pow_to_exp): New function.
14098 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
14099 Don't fold to exp if optimize_pow_to_exp is false.
14101 2018-03-28 Martin Liska <mliska@suse.cz>
14104 * calls.c (initialize_argument_information): Fix trailing space.
14105 * common.opt: Fix typo and provide better explanation for
14106 -fsanitize-coverage option.
14107 * config/i386/i386.opt: Fix typo.
14109 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14110 Martin Liska <mliska@suse.cz>
14113 * gimplify.c (asan_poison_variable): Don't do the check for
14114 gimplify_omp_ctxp here.
14115 (gimplify_decl_expr): Do it here.
14116 (gimplify_target_expr): Likewise.
14118 2018-03-28 Martin Liska <mliska@suse.cz>
14121 * config/i386/i386.c (ix86_function_arg_advance): Do not call
14122 chkp_type_bounds_count if MPX is not enabled.
14124 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
14126 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
14128 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
14131 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
14132 function to create the function decl for complex long double
14133 multiply and divide for -mabi=ieeelongdouble.
14134 (init_float128_ieee): Call it.
14136 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
14139 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
14140 -fcf-protection=branch -mibt.
14141 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
14143 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14146 * config/arm/arm.c (arm_valid_symbolic_address): Handle
14147 arm_word_relocations.
14149 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
14152 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
14153 extern array declarations.
14155 2018-03-27 Richard Biener <rguenther@suse.de>
14157 PR middle-end/84067
14158 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
14159 explicit single_use checks.
14161 2018-03-27 Richard Biener <rguenther@suse.de>
14163 PR tree-optimization/85082
14164 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
14167 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14169 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
14170 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
14171 Turn on fasynchronous-unwind-tables and funwind-tables.
14173 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
14176 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
14177 (*bmi_blsr_<mode>_ccz): Ditto.
14179 2018-03-26 Tom de Vries <tom@codesourcery.com>
14181 PR tree-optimization/85063
14182 * omp-general.c (offloading_function_p): New function. Factor out
14184 * omp-offload.c (pass_omp_target_link::gate): ... here.
14185 * omp-general.h (offloading_function_p): Declare.
14186 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
14187 with attribute omp declare target for offloading functions.
14189 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
14191 PR tree-optimization/84005
14192 * tree-data-ref.h (get_base_for_alignment): Declare.
14193 * tree-data-ref.c (get_base_for_alignment_1): New function.
14194 (get_base_for_alignment): Likewise.
14195 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
14196 get_base_for_alignment to find a suitable base object, instead
14197 of always using drb->base_address.
14199 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14201 PR inline-asm/85022
14202 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
14203 known size by default.
14205 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
14207 PR inline-asm/85030
14208 * lra-constraints.c (process_alt_operands): Don't match BLKmode
14209 and non BLKmode operands.
14211 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14214 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
14215 Clean up attributes.
14217 2018-03-23 Richard Biener <rguenther@suse.de>
14220 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
14221 we are going to emit early debug for LTO.
14223 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14225 PR inline-asm/85034
14226 * function.c (match_asm_constraints_1): Don't optimize if input
14227 doesn't satisfy general_operand predicate for output's mode.
14229 PR inline-asm/85022
14230 * alias.c (write_dependence_p): Don't require for x_canonicalized
14231 non-VOIDmode if x has VOIDmode.
14234 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
14235 just don't try to optimize it rather than assert it never happens.
14237 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14239 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
14240 macro expansions for definition of ST_INTERNAL_<mode> and
14241 LD_INTERNAL_<mode> builtins.
14242 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
14244 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
14246 (altivec_expand_st_builtin): Likewise.
14247 (altivec_expand_builtin): Remove calls to deleted functions.
14248 (rs6000_address_for_altivec): Delete this function.
14249 * config/rs6000/vector.md: Remove expands for
14250 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
14252 2018-03-22 Sudakshina Das <sudi.das@arm.com>
14255 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
14256 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
14257 re-computing once computed.
14258 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
14259 (arm_init_machine_status): Initialize
14260 machine->static_chain_stack_bytes.
14262 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14265 * doc/extend.texi: Add four new prototypes for vec_ld.
14266 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14267 definitions for more logical presentation.
14268 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14269 entries for V1TI variants of __builtin_altivec_ld builtin.
14270 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14271 handling of V1TI variant of LVX icode pattern.
14272 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14273 (rs6000_gimple_fold_builtin): Likewise.
14274 (altivec_init_builtins): Add code to define
14275 __builtin_altivec_lvx_v1ti function.
14277 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14279 PR inline-asm/84941
14280 * function.c (match_asm_constraints_1): Don't do the optimization
14281 if input isn't a REG, SUBREG, MEM or constant.
14283 2018-03-22 Tom de Vries <tom@codesourcery.com>
14285 PR tree-optimization/84956
14286 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14287 bb_has_abnormal_pred.
14289 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14292 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14293 DECL_INITIAL (decl) to decl at the end.
14294 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14295 adjust the comment.
14297 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14299 * doc/extend.texi (__builtin_tgmath): Document when complex
14300 integer types are treated as _Complex _Float64.
14302 2018-03-21 Tom de Vries <tom@codesourcery.com>
14304 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14306 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14308 PR tree-optimization/84960
14309 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14310 if it is ENTRY block, move them into single succ of ENTRY in that case.
14312 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14314 PR tree-optimization/84811
14315 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14316 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14317 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14318 (known_subrange_p): Update accordingly. Cast each value involved
14319 in the size comparison, rather than casting the result of the
14322 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14324 PR tree-optimization/84982
14325 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14326 by flipping the least significant bit rather than all bits from
14327 bitpos to bitpos + bitsize - 1.
14329 2018-03-21 Nathan Sidwell <nathan@acm.org>
14331 * doc/extend.texi (Deprecated Features): Remove mention of
14332 long-deleted deprecations.
14334 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14337 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14338 * configure: Regenerate.
14340 2018-03-21 Tom de Vries <tom@codesourcery.com>
14342 PR tree-optimization/83126
14343 * tree-parloops.c (num_phis): New function.
14344 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14346 2018-03-21 Nathan Sidwell <nathan@acm.org>
14348 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14349 mention anon-struct/union members and trailing attributes.
14351 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14353 PR tree-optimization/84969
14354 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14355 builtin memset partitions if they set different rhs values.
14357 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14359 PR rtl-optimization/84989
14360 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14361 VEC_DUPLICATE with scalar result mode.
14363 2018-03-21 Martin Liska <mliska@suse.cz>
14366 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14367 not intended return statement.
14369 2018-03-21 Martin Liska <mliska@suse.cz>
14372 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14373 (chkp_find_bound_slots_1): Limit number of iterations.
14375 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14378 * Minor grammar fixes for x86 options.
14380 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14383 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14384 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14386 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14389 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14390 (altivec_lvx_<mode>_1op): Likewise.
14391 (altivec_stvx_<mode>_2op): Likewise.
14392 (altivec_stvx_<mode>_1op): Likewise.
14393 (altivec_lvx_<VM2:mode>): New define_expand.
14394 (altivec_stvx_<VM2:mode>): Likewise.
14395 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14396 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14397 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14398 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14399 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14400 (rs6000_gen_lvx): Likewise.
14401 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14402 (altivec_expand_stv_builtin): Likewise.
14403 (altivec_expand_builtin): Likewise.
14404 * config/rs6000/vector.md: Likewise.
14406 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14409 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14412 2018-03-20 Richard Biener <rguenther@suse.de>
14415 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14416 sign-conversions as zero, fall back to standard scalar_stmt
14419 2018-03-20 Martin Liska <mliska@suse.cz>
14422 * predict.c (rebuild_frequencies): Handle case when we have
14423 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14425 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14428 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14429 flag_section_anchors.
14430 * varasm.c (use_blocks_for_decl_p): Remove hack for
14431 dw2_force_const_mem.
14434 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14436 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14437 be created, use lowpart_subreg of operands[0] rather than operands[0]
14439 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14440 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14441 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14442 and n constraint instead of aarch64_shift_imm_di and Usd.
14443 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14444 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14446 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14449 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14450 to favor GPR over NEON registers.
14451 (<shift>di3_neon): Likewise.
14453 2018-03-20 Tom de Vries <tom@codesourcery.com>
14456 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14457 (nvptx_process_pars): Emit bar.sync asap and alap.
14459 2018-03-20 Tom de Vries <tom@codesourcery.com>
14462 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14463 seen_label if seen_label is already set.
14465 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14468 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14469 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14470 Use 1U instead of 1. Formatting fixes.
14473 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14474 instead of TREE_TYPE (s1) for the return value.
14476 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14478 PR tree-optimization/84946
14479 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14480 bitsize + bitsize in poly_uint64 rather than poly_int64.
14483 * dwarf2asm.c: Include fold-const.c.
14484 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14485 of decl rather than decl itself.
14487 PR rtl-optimization/84643
14488 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14490 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14493 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14494 calling assemble_variable.
14496 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14499 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14500 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14502 2018-03-19 Jim Wilson <jimw@sifive.com>
14505 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14506 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14507 (riscv_first_stack_step): Likewise.
14508 (riscv_option_override): Use STACK_BOUNDARY instead of
14509 MIN_STACK_BOUNDARY.
14510 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14511 MIN_STACK_BOUNDARY.
14512 (BIGGEST_ALIGNMENT): Set to 128.
14513 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14514 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14517 2018-03-19 Richard Biener <rguenther@suse.de>
14519 PR tree-optimization/84933
14520 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14521 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14523 2018-03-19 Richard Biener <rguenther@suse.de>
14525 PR tree-optimization/84859
14526 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14527 (cond_if_else_store_replacement): Perform sinking operation on
14528 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14529 Generalize what a BB with a single eligible store is.
14531 2018-03-19 Richard Biener <rguenther@suse.de>
14533 PR tree-optimization/84929
14534 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14535 chrec_is_positive against non-chrec arg.
14537 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14540 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14542 2018-03-18 Martin Liska <mliska@suse.cz>
14544 PR rtl-optimization/84635
14545 * regrename.c (build_def_use): Use matches_mode only when
14548 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14550 PR tree-optimization/84913
14551 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14552 vectorize chains of COND_EXPRs.
14554 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14556 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14558 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14560 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14562 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14564 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14566 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14567 Kito Cheng <kito.cheng@gmail.com>
14569 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14570 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14571 (nds32_adjust_reg_alloc_order): New function.
14572 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14574 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14576 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14577 nds32_print_operand, nds32_print_operand_address): Use
14578 HOST_WIDE_INT_PRINT_DEC instead.
14580 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14582 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14584 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14587 * config/i386/i386.c (initial_ix86_tune_features,
14588 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14589 unsigned long long.
14590 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14591 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14592 rather than 1u << ix86_tune. Formatting fix.
14593 (ix86_option_override_internal): Change ix86_arch_mask from
14594 unsigned int to unsigned HOST_WIDE_INT, initialize to
14595 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14596 (ix86_function_specific_restore): Likewise.
14598 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14601 * postreload.c (reload_combine_recognize_pattern): Perform
14602 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14603 truncate_int_for_mode the result for the destination's mode.
14606 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14607 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14609 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14612 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14615 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14616 regno_allocno_class_array and sorted_pseudos.
14617 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14618 insns where regno is used.
14620 2018-03-16 Martin Liska <mliska@suse.cz>
14623 * multiple_target.c (create_dispatcher_calls): Redirect
14624 reference in the symbol table.
14626 2018-03-16 Martin Liska <mliska@suse.cz>
14629 * multiple_target.c (create_dispatcher_calls): Redirect also
14632 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14636 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14637 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14638 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14640 2018-03-16 Julia Koval <julia.koval@intel.com>
14642 * doc/invoke.texi (Skylake Server): Add CLWB.
14643 Cannonlake): Remove CLWB.
14645 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14647 PR tree-optimization/84841
14648 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14650 (FLOAT_ONE_CONST_TYPE): Define.
14651 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14652 (sort_by_operand_rank): Put entries with higher constant_type last
14653 rather than first to match comments.
14655 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14657 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14660 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14663 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14664 check and use address of target always.
14666 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14669 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14670 (indirect_thunk_bnd_needed): Likewise.
14671 (indirect_thunks_used): Likewise.
14672 (indirect_thunks_bnd_used): Likewise.
14673 (indirect_return_needed): New.
14674 (indirect_return_bnd_needed): Likewise.
14675 (output_indirect_thunk_function): Add a bool argument for
14677 (output_indirect_thunk_function): Don't generate alias for
14678 function return thunk.
14679 (ix86_code_end): Call output_indirect_thunk_function to generate
14680 function return thunks.
14681 (ix86_output_function_return): Set indirect_return_bnd_needed
14682 and indirect_return_needed instead of indirect_thunk_bnd_needed
14683 and indirect_thunk_needed.
14685 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14687 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14688 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14689 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14691 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14692 Paul Hua <paul.hua.gm@gmail.com>
14695 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14697 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14699 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14700 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14701 resp. SFmode cases.
14703 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14706 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14707 instead of GET_MODE_SIZE when comparing Units.
14709 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14712 * varasm.c (hash_section): Return an unchangeble hash value
14713 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14714 Return !aarch64_can_use_per_function_literal_pools_p ().
14716 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14719 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14720 rather than address of cmode as last argument to prepare_cmp_insn.
14722 2018-03-15 Julia Koval <julia.koval@intel.com>
14724 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14725 F_AVX512VNNI, F_AVX512BITALG): New.
14727 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14730 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14731 insn for floating-point loads and stores.
14733 2018-03-14 Carl Love <cel@us.ibm.com>
14735 * config/rs6000/rs6000-c.c: Add macro definitions for
14736 ALTIVEC_BUILTIN_VEC_PERMXOR.
14737 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14738 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14739 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14740 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14742 * config/doc/extend.texi: Add prototypes for vec_permxor.
14744 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14747 * diagnostic-show-locus.c (class layout_point): Convert m_line
14748 from int to linenum_type.
14749 (line_span::comparator): Use linenum "compare" function when
14750 comparing line numbers.
14751 (test_line_span): New function.
14752 (layout_range::contains_point): Convert param "row" from int to
14754 (layout_range::intersects_line_p): Likewise.
14755 (layout::will_show_line_p): Likewise.
14756 (layout::print_source_line): Likewise.
14757 (layout::should_print_annotation_line_p): Likewise.
14758 (layout::print_annotation_line): Likewise.
14759 (layout::print_leading_fixits): Likewise.
14760 (layout::annotation_line_showed_range_p): Likewise.
14761 (struct line_corrections): Likewise for field m_row.
14762 (line_corrections::line_corrections): Likewise for param "row".
14763 (layout::print_trailing_fixits): Likewise.
14764 (layout::get_state_at_point): Likewise.
14765 (layout::get_x_bound_for_row): Likewise.
14766 (layout::print_line): Likewise.
14767 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14768 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14769 * input.c (selftest::test_linenum_comparisons): New function.
14770 (selftest::input_c_tests): Call it.
14771 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14772 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14773 * selftest.h (ASSERT_GT): New macro.
14774 (ASSERT_GT_AT): New macro.
14775 (ASSERT_LT): New macro.
14776 (ASSERT_LT_AT): New macro.
14778 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14780 PR rtl-optimization/84780
14781 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14783 2018-03-14 Martin Liska <mliska@suse.cz>
14785 * tree.c (record_node_allocation_statistics): Use
14786 get_stats_node_kind.
14787 (get_stats_node_kind): New function extracted from
14788 record_node_allocation_statistics.
14789 (free_node): Use get_stats_node_kind.
14791 2018-03-14 Richard Biener <rguenther@suse.de>
14793 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14794 that the value-set of ANTIC_IN doesn't grow.
14797 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14799 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14800 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14802 2018-03-14 Julia Koval <julia.koval@intel.com>
14804 * config.gcc (icelake-client, icelake-server): New.
14806 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14807 (initial_ix86_arch_features): Ditto.
14808 (PTA_SKYLAKE): Add SGX.
14809 (PTA_ICELAKE): Remove.
14810 (PTA_ICELAKE_CLIENT): New.
14811 (PTA_ICELAKE_SERVER): New.
14812 (ix86_option_override_internal): Split up icelake on icelake client and
14814 (get_builtin_code_for_version): Ditto.
14815 (fold_builtin_cpu): Ditto.
14816 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14817 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14818 * config/i386/i386.h (processor_type): Ditto.
14819 * doc/invoke.texi: Ditto.
14821 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14824 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14825 INTEGER_CST offset, add it together with bitpos / 8 and
14826 sign extend based on POINTER_SIZE.
14830 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14833 * config/i386/constraints.md (Yc): New register constraint.
14834 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14835 Use Yc constraint for alternative 2 of operand 0. Remove
14836 preferred_for_speed attribute.
14838 2018-03-14 Richard Biener <rguenther@suse.de>
14840 PR tree-optimization/84830
14841 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14842 with the old one to avoid oscillations.
14844 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14847 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14849 (assign_by_spills): Return a flag of reload assignment failure.
14850 Do not process the reload assignment failures. Do not spill other
14851 reload pseudos if they has the same reg class. Update n if
14853 (lra_assign): Add a return arg. Set up from the result of
14854 assign_by_spills call.
14855 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14856 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14857 usage_insns if it is not NULL.
14858 (spill_hard_reg_in_range): New function.
14859 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14860 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14861 function prototypes.
14862 (lra_assign): Change prototype.
14863 * lra.c (lra): Add code to deal with fails by splitting hard reg
14866 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14868 * config/riscv/riscv.opt (mrelax): New option.
14869 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14870 "norelax" when riscv_mrelax is disabled.
14871 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14873 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14876 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14877 reassociation for int modes.
14879 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14881 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14882 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14884 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14885 * config/aarch64/aarch64-sve.md
14886 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14887 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14888 (*extend<mode><Vwide>2): Rename to...
14889 (aarch64_sve_extend<mode><Vwide>2): ...this.
14890 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14891 renaming the old pattern to...
14892 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14894 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14895 define_expand, renaming the old pattern to...
14896 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14897 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14898 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14899 account when deciding which SVE instruction the optab should use.
14900 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14902 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14904 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14905 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14906 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14907 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14908 (tlsdesc_small_<mode>): Turn a define_expand and use
14909 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14910 (tlsdesc_small_advsimd_<mode>): ...this.
14911 (tlsdesc_small_sve_<mode>): New pattern.
14913 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14915 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14916 (UNSPEC_UMUL_HIGHPART): New constants.
14917 (MUL_HIGHPART): New int iteraor.
14918 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14919 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14921 (*<su>mul<mode>3_highpart): New define_insn.
14923 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14926 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14929 2018-03-13 Martin Liska <mliska@suse.cz>
14932 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14934 (sem_item_optimizer::~sem_item_optimizer): Release it.
14935 (sem_item_optimizer::merge_classes): Register variable aliases.
14936 (sem_item_optimizer::fixup_pt_set): New function.
14937 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14938 * ipa-icf.h: Declare new variables and functions.
14940 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14942 PR middle-end/84834
14943 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14944 integer_pow2p@2 and test integer_pow2p in condition.
14945 (A < 0 ? C : 0): Similarly for @1.
14947 PR middle-end/84831
14948 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14949 characters starting at p contain '\0' character, don't look beyond
14953 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14954 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14957 * reg-stack.c (change_stack): Change update_end var from int to
14958 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14959 also call set_block_for_insn on the newly added insns and rescan.
14962 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14963 on the last operand.
14966 * tree.c (stabilize_reference_1): Return save_expr (e) for
14967 STATEMENT_LIST even if it doesn't have side-effects.
14969 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14971 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14973 2018-03-12 Renlin Li <renlin.li@arm.com>
14975 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14976 aarch64_output_scalar_simd_mov_immediate.
14978 2018-03-12 Martin Sebor <msebor@redhat.com>
14980 PR tree-optimization/83456
14981 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14982 for perfectly overlapping calls to memcpy.
14983 (gimple_fold_builtin_memory_chk): Same.
14984 (gimple_fold_builtin_strcpy): Handle no-warning.
14985 (gimple_fold_builtin_stxcpy_chk): Same.
14986 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14988 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14990 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14991 parameter. Use it for SFmode.
14992 (rs6000_function_arg_advance_1): Adjust.
14993 (rs6000_function_arg): Adjust.
14994 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14996 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14998 PR rtl-optimization/84169
14999 PR rtl-optimization/84780
15000 * combine.c (can_combine_p): Check for a 2-insn combination whether
15001 the destination register is used between the two insns, too.
15003 2018-03-12 Richard Biener <rguenther@suse.de>
15005 PR tree-optimization/84803
15006 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
15007 for refs DR analysis didn't process.
15009 2018-03-12 Richard Biener <rguenther@suse.de>
15011 PR tree-optimization/84777
15012 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
15013 force-vectorize loops ignore whether we are optimizing for size.
15015 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
15017 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
15018 (TARGET_MD_ASM_ADJUST): Define.
15020 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
15021 Kito Cheng <kito.cheng@gmail.com>
15022 Chung-Ju Wu <jasonwucj@gmail.com>
15024 * config/nds32/nds32.c (nds32_compute_stack_frame,
15025 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
15026 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
15027 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
15028 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
15029 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
15030 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
15031 * config/nds32/nds32.md (prologue, epilogue): Use macro
15032 NDS32_V3PUSH_AVAILABLE_P to do checking.
15034 2018-03-11 Jakub Jelinek <jakub@redhat.com>
15037 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
15038 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
15039 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
15040 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
15041 addition of most attributes on !orig_type_die or the attribute not
15042 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
15044 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15045 Chung-Ju Wu <jasonwucj@gmail.com>
15047 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
15048 __NDS32_VH__ macro.
15049 * config/nds32/nds32.opt (mvh): New option.
15051 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15052 Chung-Ju Wu <jasonwucj@gmail.com>
15054 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
15056 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
15057 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
15060 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15061 Chung-Ju Wu <jasonwucj@gmail.com>
15063 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
15065 * config/nds32/nds32-multiple.md (strlensi): New pattern.
15066 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
15068 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
15069 Kito Cheng <kito.cheng@gmail.com>
15070 Chung-Ju Wu <jasonwucj@gmail.com>
15072 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
15073 UNSPEC_FFMISM and UNSPEC_FLMISM.
15074 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
15075 for ffb, ffmism and flmism.
15076 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
15077 (unspec_ffmism): Ditto.
15078 (unspec_flmism): Ditto.
15079 (nds32_expand_builtin_impl): Check if string extension is available.
15080 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
15081 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
15083 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
15086 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15089 * lra-assigns.c (assign_by_spills): Return a flag of reload
15090 assignment failure. Do not process the reload assignment
15091 failures. Do not spill other reload pseudos if they has the same
15093 (lra_assign): Add a return arg. Set up from the result of
15094 assign_by_spills call.
15095 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15096 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15097 usage_insns if it is not NULL.
15098 (spill_hard_reg_in_range): New function.
15099 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15100 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15101 function prototypes.
15102 (lra_assign): Change prototype.
15103 * lra.c (lra): Add code to deal with fails by splitting hard reg
15106 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
15109 * config/i386/i386.opt: Replace Enforcment with Enforcement.
15111 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
15114 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
15115 (dw_val_node): Add val_symbolic_view.
15116 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
15117 (symview_upper_bound): New.
15118 (new_line_info_table): Initialize symviews_since_reset.
15119 (dwarf2out_source_line): Count symviews_since_reset and set
15120 symview_upper_bound.
15121 (dw_val_equal_p): Handle symview.
15122 (add_AT_symview): New.
15123 (print_dw_val): Handle symview.
15124 (attr_checksum, attr_checksum_ordered): Likewise.
15125 (same_dw_val_p, size_of_die): Likewise.
15126 (value_format, output_die): Likewise.
15127 (add_high_low_attributes): Use add_AT_symview for entry_view.
15128 (dwarf2out_finish): Reset symview_upper_bound, clear
15131 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
15134 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
15135 Add strict argument and use it.
15136 (rs6000_split_multireg_move): Update for new strict argument.
15137 (mem_operand_gpr): Disallow all non-offsettable addresses.
15138 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
15140 2018-03-09 Jakub Jelinek <jakub@redhat.com>
15143 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
15144 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
15145 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
15148 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
15149 decl, use remap_type if we want to use the type.
15151 2018-03-09 Martin Sebor <msebor@redhat.com>
15153 PR tree-optimization/84526
15154 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15156 (builtin_access::generic_overlap): Be prepared to handle non-array
15159 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
15161 PR rtl-optimization/84682
15162 * lra-constraints.c (process_address_1): Check is_address flag
15163 for address constraints.
15164 (process_alt_operands): Likewise.
15165 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
15166 preprocess_constraints.
15167 * recog.h (preprocess_constraints): Add oploc parameter.
15169 * recog.c (preprocess_constraints): Test address_operand for
15170 CT_ADDRESS constraints.
15172 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15175 * lra-assigns.c (assign_by_spills): Return a flag of reload
15176 assignment failure. Do not process the reload assignment
15177 failures. Do not spill other reload pseudos if they has the same
15179 (lra_assign): Add a return arg. Set up from the result of
15180 assign_by_spills call.
15181 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15182 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15183 usage_insns if it is not NULL.
15184 (spill_hard_reg_in_range): New function.
15185 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15186 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15187 function prototypes.
15188 (lra_assign): Change prototype.
15189 * lra.c (lra): Add code to deal with fails by splitting hard reg
15192 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15195 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
15196 Accept complain bool parameter. Only emit errors if it is true.
15197 (arm_parse_cpu_option_name): Likewise.
15198 (arm_target_thumb_only): Adjust callers of the above.
15199 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
15200 prototype to take a default true bool parameter.
15201 (arm_parse_arch_option_name): Likewise.
15203 2018-03-09 David Malcolm <dmalcolm@redhat.com>
15204 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
15208 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
15209 * configure: Regenerate.
15210 * configure.ac ("linker --version-script option"): New.
15211 ("linker soname option"): New.
15213 2018-03-09 Richard Biener <rguenther@suse.de>
15215 PR tree-optimization/84775
15216 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
15217 immediate uses of predicate stmts and mark them modified.
15220 PR tree-optimization/84178
15221 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15223 (version_loop_for_if_conversion): Delay update_ssa call.
15224 (tree_if_conversion): Delay update_ssa until after predicate
15227 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
15230 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
15231 when the function accesses prior frames.
15233 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15236 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
15237 gen_llsym, otherwise call maybe_gen_llsym.
15239 PR inline-asm/84742
15240 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15241 has ',' character inside of it.
15243 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15246 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15247 as clobbering CC_REGNUM.
15249 2018-03-08 Richard Biener <rguenther@suse.de>
15251 PR middle-end/84552
15252 * tree-scalar-evolution.c: Include tree-into-ssa.h.
15253 (follow_copies_to_constant): Do not follow SSA names registered
15256 2018-03-08 Richard Biener <rguenther@suse.de>
15258 PR tree-optimization/84178
15259 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15261 (version_loop_for_if_conversion): Delay update_ssa call.
15262 (tree_if_conversion): Delay update_ssa until after predicate
15265 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15267 PR tree-optimization/84178
15268 * tree-if-conv.c (release_bb_predicate): Remove the
15269 the assertion that the stmts have NULL use_ops.
15270 Discard the statements, asserting that they haven't
15271 yet been added to a BB.
15273 2018-03-08 Richard Biener <rguenther@suse.de>
15275 PR tree-optimization/84746
15276 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15277 (phi_translate): Pass in destination ANTIC_OUT set.
15278 (phi_translate_1): Likewise. For a simplified result lookup
15279 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15280 (phi_translate_set): Adjust.
15281 (do_pre_regular_insertion): Likewise.
15282 (do_pre_partial_partial_insertion): Likewise.
15284 2018-03-08 Martin Liska <mliska@suse.cz>
15286 PR gcov-profile/84735
15287 * doc/gcov.texi: Document usage of profile files.
15288 * gcov-io.h: Document changes in the format.
15290 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15294 * dwarf2out.c (struct dw_line_info_table): Update comments for
15296 (FORCE_RESET_NEXT_VIEW): New.
15297 (FORCE_RESETTING_VIEW_P): New.
15298 (RESETTING_VIEW_P): Check for -1 too.
15299 (ZERO_VIEW_P): Likewise.
15300 (new_line_info_table): Force-reset next view.
15301 (dwarf2out_begin_function): Likewise.
15302 (dwarf2out_source_line): Simplify zero_view_p initialization.
15303 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15304 view directly. Omit view when omitting .loc at line 0.
15306 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15308 PR tree-optimization/84740
15309 * tree-switch-conversion.c (process_switch): Call build_constructors
15310 only if info.phi_count is non-zero.
15312 PR tree-optimization/84739
15313 * tree-tailcall.c (find_tail_calls): Check call arguments against
15314 DECL_ARGUMENTS (current_function_decl) rather than
15315 DECL_ARGUMENTS (func) when checking for tail recursion.
15317 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15319 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15320 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15321 Volker Reichelt's entry and add entries for people that perform
15322 GCC fuzzy testing and report numerous bugs.
15324 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15327 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15328 readonly data in sdata, if that is disabled.
15329 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15330 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15331 -mreadonly-in-sdata option.
15333 2018-03-07 Martin Sebor <msebor@redhat.com>
15335 PR tree-optimization/84468
15336 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15337 basic block when looking for nul assignment.
15339 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15342 * except.h (output_function_exception_table): Adjust prototype.
15343 * except.c (output_function_exception_table): Remove FNNAME parameter
15344 and add SECTION parameter. Ouput one part of the table at a time.
15345 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15346 the first part of the exception table and emit unwind directives.
15347 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15348 (i386_pe_seh_cold_init): Likewise.
15349 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15350 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15351 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15352 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15353 * config/i386/winnt.c: Include except.h.
15354 (struct seh_frame_state): Add reg_offset, after_prologue and
15355 in_cold_section fields.
15356 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15357 (i386_pe_seh_cold_init): New function.
15358 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15359 to seh->in_cold_section.
15360 (seh_emit_push): Record the offset of the push.
15361 (seh_emit_save): Record the offet of the save.
15362 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15363 Test seh->after_prologue to disregard the epilogue.
15364 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15365 (i386_pe_end_cold_function): New function.
15367 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15370 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15371 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15374 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15376 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15377 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15379 PR middle-end/84723
15380 * multiple_target.c: Include tree-inline.h and intl.h.
15381 (expand_target_clones): Diagnose and fail if node->definition and
15382 !tree_versionable_function_p (node->decl).
15384 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15386 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15388 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15389 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15390 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15392 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15395 * combine.c (try_combine): Use reg_or_subregno instead of handling
15396 just paradoxical SUBREGs and REGs.
15398 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15400 * config/arc/arc.c (arc_finalize_pic): Remove function.
15401 (arc_must_save_register): We use single base PIC register, remove
15402 checks to save/restore the PIC register.
15403 (arc_expand_prologue): Likewise.
15404 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15406 (arc_verify_short): Likewise.
15407 (arc_attr_type): Likewise.
15408 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15409 (walk_stores): Likewise.
15410 (arc_address_cost): Make it static.
15411 (arc_verify_short): Likewise.
15412 (branch_dest): Likewise.
15413 (arc_attr_type): Likewise.
15414 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15415 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15416 (arc_final_prescan_insn): Remove inserting the nops due to
15417 hardware hazards. It is done in reorg step.
15418 (insn_length_variant_t): Remove.
15419 (insn_length_parameters_t): Likewise.
15420 (arc_insn_length_parameters): Likewise.
15421 (arc_get_insn_variants): Likewise.
15422 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15424 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15426 PR inline-asm/84683
15427 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15430 PR tree-optimization/84687
15431 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15433 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15435 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15437 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15438 Rename to ppc_speculation_barrier.
15439 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15440 __builtin_ppc_speculation_barrier.
15442 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15445 * combine.c (combine_simplify_rtx): Don't try to simplify if
15446 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15449 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15451 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15452 to 32 bytes when compiling for POWER9.
15454 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15457 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15458 regparm >= 3 with no arg reg available also for calls with
15459 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15462 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15464 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15466 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15469 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15471 2018-03-05 Richard Biener <rguenther@suse.de>
15473 PR tree-optimization/84486
15474 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15475 When inserting a __builtin_assume_aligned call set the LHS
15476 SSA name alignment info accordingly.
15478 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15480 PR tree-optimization/84114
15481 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15482 Avoid reassociation of FLOAT_MODE addition.
15484 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15486 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15487 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15488 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15489 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15490 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15491 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15494 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15495 __builtin_ia32_wbinvd): New builtins.
15496 (SPECIAL_ARGS2): New.
15497 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15498 (SPECIAL_ARGS2): New.
15499 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15500 (ix86_valid_target_attribute_inner_p): Ditto.
15501 (ix86_init_mmx_sse_builtins): Add special_args2.
15502 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15503 TARGET_WBNOINVD_P): New.
15504 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15505 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15506 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15507 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15508 * config/i386/pconfigintrin.h: New file.
15509 * config/i386/wbnoinvdintrin.h: Ditto.
15510 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15512 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15514 2018-03-05 Richard Biener <rguenther@suse.de>
15516 PR tree-optimization/84670
15517 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15519 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15520 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15521 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15522 doesn't grow if all successors (recursively) were visited at least
15525 2018-03-05 Richard Biener <rguenther@suse.de>
15527 PR tree-optimization/84650
15528 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15529 if executed in the loop pipeline.
15531 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15533 * doc/configfiles.texi (Configuration Files): Move info about
15534 conditionalizing $target-protos.h to...
15535 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15536 differs from $target-protos.h.
15538 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15539 Chung-Ju Wu <jasonwucj@gmail.com>
15541 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15542 * config/nds32/nds32-multiple.md (setmemsi): Define.
15543 * config/nds32/nds32-memory-manipulation.c
15544 (nds32_gen_dup_4_byte_to_word_value): New.
15545 (emit_setmem_word_loop): New.
15546 (emit_setmem_byte_loop): New.
15547 (nds32_expand_setmem_loop): New.
15548 (nds32_expand_setmem_loop_v3m): New.
15549 (nds32_expand_setmem_unroll): New.
15550 (nds32_expand_setmem): New.
15552 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15553 Chung-Ju Wu <jasonwucj@gmail.com>
15555 * config/nds32/nds32-memory-manipulation.c
15556 (nds32_emit_load_store): New.
15557 (nds32_emit_post_inc_load_store): New.
15558 (nds32_emit_mem_move): New.
15559 (nds32_emit_mem_move_block): New.
15560 (nds32_expand_movmemsi_loop_unknown_size): New.
15561 (nds32_expand_movmemsi_loop_known_size): New.
15562 (nds32_expand_movmemsi_loop): New.
15563 (nds32_expand_movmemsi_unroll): New.
15564 (nds32_expand_movmemqi): Rename ...
15565 (nds32_expand_movmemsi): ... to this.
15566 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15567 (movmemsi): ... to this.
15568 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15569 (nds32_expand_movmemsi): ... to this.
15571 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15572 Monk Chiang <sh.chiang04@gmail.com>
15573 Chung-Ju Wu <jasonwucj@gmail.com>
15575 * config/nds32/nds32-protos.h
15576 (nds32_expand_load_multiple): New arguments.
15577 (nds32_expand_store_multiple): Ditto.
15578 (nds32_valid_multiple_load_store): Rename ...
15579 (nds32_valid_multiple_load_store_p): ... to this.
15580 * config/nds32/nds32-memory-manipulation.c
15581 (nds32_expand_load_multiple): Refine implementation.
15582 (nds32_expand_store_multiple): Ditto.
15583 * config/nds32/nds32-multiple.md
15584 (load_multiple): Update nds32_expand_load_multiple interface.
15585 (store_multiple): Update nds32_expand_store_multiple interface.
15586 * config/nds32/nds32-predicates.c
15587 (nds32_valid_multiple_load_store): Rename ...
15588 (nds32_valid_multiple_load_store_p): ... to this and refine
15590 * config/nds32/predicates.md
15591 (nds32_load_multiple_and_update_address_operation): New predicate.
15592 (nds32_store_multiple_and_update_address_operation): New predicate.
15594 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15595 Chung-Ju Wu <jasonwucj@gmail.com>
15597 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15598 (combo): New attribute.
15599 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15601 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15603 * config/nds32/nds32.opt: Change -mcmodel= default value.
15605 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15606 Monk Chiang <sh.chiang04@gmail.com>
15607 Chung-Ju Wu <jasonwucj@gmail.com>
15609 * config/nds32/constants.md (unspec_element): New enum.
15610 * config/nds32/constraints.md (Umw): New constraint.
15611 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15612 * config/nds32/nds32-intrinsic.md: Likewise.
15613 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15614 (nds32_valid_smw_lwm_base_p): New.
15615 (nds32_output_smw_single_word): New.
15616 (nds32_output_lmw_single_word): New.
15617 (nds32_expand_unaligned_load): New.
15618 (nds32_expand_unaligned_store): New.
15619 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15620 (nds32_output_smw_single_word): Declare.
15621 (nds32_output_lmw_single_word): Declare.
15622 (nds32_expand_unaligned_load): Declare.
15623 (nds32_expand_unaligned_store): Declare.
15624 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15625 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15626 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15627 NDS32_BUILTIN_UASTORE_DW.
15628 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15631 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15632 Kito Cheng <kito.cheng@gmail.com>
15633 Chung-Ju Wu <jasonwucj@gmail.com>
15635 * config/nds32/nds32-intrinsic.c
15636 (nds32_expand_builtin_null_ftype_reg): Delete.
15637 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15638 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15639 (nds32_read_argument): New.
15640 (nds32_legitimize_target): Ditto.
15641 (nds32_legitimize_argument): Ditto.
15642 (nds32_check_constant_argument): Ditto.
15643 (nds32_expand_unop_builtin): Ditto.
15644 (nds32_expand_unopimm_builtin): Ditto.
15645 (nds32_expand_binop_builtin): Ditto.
15646 (nds32_builtin_decl_impl): Ditto.
15647 (builtin_description): Ditto.
15648 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15649 (nds32_init_builtins_impl): Ditto.
15650 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15651 (nds32_builtin_decl): New.
15652 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15653 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15655 2018-03-02 Jeff Law <law@redhat.com>
15657 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15658 (redundant_insn, fill_simple_delay_slots): Likewise.
15659 (fill_slots_from_thread): Likewise.
15660 * resource.c (mark_referenced_resources): Likewise.
15661 (mark_set_resources, find_dead_or_set_registers): Likewise.
15663 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15665 * substring-locations.h (format_warning_va): Formatting fix for
15666 ATTRIBUTE_GCC_DIAG.
15667 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15669 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15670 * substring-locations.c: Include intl.h.
15671 (format_warning_va): Turned into small wrapper around
15672 format_warning_n_va, renamed to ...
15673 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15674 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15676 (format_warning_at_substring_n): New function.
15677 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15678 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15679 format_warning_at_substring with just a shorter name instead of
15680 const function pointer.
15681 (fmtwarn_n): New function.
15682 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15683 appropriate, get rid of all the fmtstr temporaries, move conditionals
15684 with G_() wrapped string literals directly into fmtwarn arguments,
15685 cast dir.len to (int), formatting fixes.
15687 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15689 * doc/invoke.texi: Remove "Cilk Plus" references.
15691 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15692 Richard Biener <rguenther@suse.de>
15695 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15696 for error or warning attributes if CALL_FROM_THUNK_P is set.
15699 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15702 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15703 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15706 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15707 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15709 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15710 instead of -1U in last predictors element's probability member.
15712 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15715 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15716 arguments if they are comparable.
15718 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15720 PR tree-optimization/84634
15721 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15722 masks and masked_loop_p with a single loop_masks, making sure it's
15723 null for bb vectorization.
15725 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15727 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15728 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15729 loop->force_vectorize to check whether there is no alias.
15731 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15734 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15736 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15738 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15739 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15740 instead of a loop around prev_real_insn.
15741 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15744 PR inline-asm/84625
15745 * config/i386/i386.c (ix86_print_operand): Use conditional
15746 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15749 2018-03-02 Richard Biener <rguenther@suse.de>
15751 PR tree-optimization/84427
15752 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15753 (bitmap_set_subtract_values): Rewrite to handle multiple
15756 (prune_clobbered_mems): Likewise.
15757 (phi_translate): Take edge instead of pred/phiblock.
15758 (phi_translate_1): Likewise.
15759 (phi_translate_set): Likewise. Insert all translated
15760 exprs for a value into the set, keeping possibly multiple
15761 expressions per value.
15762 (compute_antic_aux): Adjust for phi_translate changes.
15763 When intersecting union the expressions and prune those
15764 not in the final value set, keeping possibly multiple
15765 expressions per value. Do not use value-insertion
15766 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15767 all expressions. Add verification that the value-sets
15768 only shrink during iteration.
15769 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15770 (do_pre_regular_insertion): Likewise.
15771 (do_pre_partial_partial_insertion): Likewise.
15773 2018-03-02 Richard Biener <rguenther@suse.de>
15776 * config/darwin.c (saved_debug_info_level): New static global.
15777 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15778 (darwin_asm_lto_end): Restore debug info generation settings.
15780 2018-03-01 Martin Liska <mliska@suse.cz>
15783 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15784 volatile arguments.
15786 2018-03-01 Richard Biener <rguenther@suse.de>
15789 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15790 type annotation with LTO when debug was disabled at compile-time.
15792 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15794 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15796 (mips_final_postscan_insn): Likewise.
15798 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15800 PR rtl-optimization/84528
15801 * alias.c (init_alias_target): Add commentary.
15802 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15803 a unique base value if the frame pointer is not eliminated
15804 to the stack pointer.
15806 2018-03-01 Tom de Vries <tom@codesourcery.com>
15808 PR rtl-optimization/83327
15809 * lra-int.h (hard_regs_spilled_into): Declare.
15810 * lra.c (hard_regs_spilled_into): Define.
15811 (init_reg_info): Init hard_regs_spilled_into.
15812 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15813 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15814 (process_bb_lives): Handle hard_regs_spilled_into.
15815 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15816 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15818 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15820 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15821 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15822 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15823 * config/rs6000/aix72.h: New file.
15825 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15827 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15828 instead of warning_at with conditional singular and plural messages
15832 * stor-layout.c (update_alignment_for_field): For
15833 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15834 && !DECL_PACKED (field), do the alignment update, just use
15835 only desired_align instead of MAX (type_align, desired_align)
15837 (place_field): Don't do known_align < desired_align handling
15838 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15839 is non-NULL, instead do it after rli->prev_field handling and
15840 only if not within a bitfield word. For DECL_PACKED (field)
15841 use type_align of BITS_PER_UNIT.
15843 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15845 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15846 superfluous parentheses and trailing spaces.
15848 2018-02-28 Richard Biener <rguenther@suse.de>
15850 PR tree-optimization/84584
15851 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15852 SCoPs with fake exit edge.
15854 2018-02-28 Martin Liska <mliska@suse.cz>
15857 * timevar.c (timer::print): Fix format to properly print 100%
15860 2018-02-28 Richard Biener <rguenther@suse.de>
15862 PR middle-end/84607
15863 * genmatch.c (capture_info::walk_match): Do not mark
15864 captured expressions without operands as expr_p given
15865 they act more like predicates and should be subject to
15866 "lost tail" side-effect preserving.
15868 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15870 PR rtl-optimization/81611
15871 * auto-inc-dec.c (attempt_change): Move dead note from
15872 mem_insn if it's the next use of regno
15873 (find_address): Take address use of reg holding
15874 non-incremented value. Add parm to limit search to the named
15876 (merge_in_block): Attempt to use a mem insn that is the next
15877 use of the original regno.
15879 2018-02-27 Martin Sebor <msebor@redhat.com>
15882 * doc/invoke.texi (-Wmissing-attributes): New option.
15883 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15885 2018-02-27 Martin Sebor <msebor@redhat.com>
15887 PR translation/84207
15888 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15889 n argument to unsigned HOST_WIDE_INT.
15890 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15891 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15892 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15893 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15895 2018-02-27 Richard Biener <rguenther@suse.de>
15897 PR tree-optimization/84512
15898 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15899 Do not use the estimate returned from record_stmt_cost for
15900 the scalar iteration cost but sum properly using add_stmt_cost.
15902 2018-02-27 Richard Biener <rguenther@suse.de>
15904 PR tree-optimization/84466
15905 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15906 Adjust last change to less strictly validate use operands.
15908 2018-02-27 Martin Liska <mliska@suse.cz>
15910 PR gcov-profile/84548
15911 * gcov.c (process_file): Allow partial overlap and consider it
15912 also as group functions.
15913 (output_lines): Properly calculate range of lines for a group.
15915 2018-02-27 Martin Liska <mliska@suse.cz>
15917 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15918 'ggc' suffixes. Change first column width.
15919 (timer::print): Fix formatting of the column.
15921 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15923 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15924 preserve inline entry blocks for the sake of debug inline
15925 entry point markers alone.
15926 (remove_unused_locals): Suggest in comments a better place to
15927 force the preservation of inline entry blocks that are
15928 otherwise unused, but do not preserve them.
15930 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15932 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15934 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15937 * config/i386/constraints.md (Bs): Replace
15938 ix86_indirect_branch_register with
15939 TARGET_INDIRECT_BRANCH_REGISTER.
15941 * config/i386/i386.md (indirect_jump): Likewise.
15942 (tablejump): Likewise.
15943 (*sibcall_memory): Likewise.
15944 (*sibcall_value_memory): Likewise.
15945 Peepholes of indirect call and jump via memory: Likewise.
15946 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15947 (*sibcall_value_GOT_32): Likewise.
15948 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15949 (GOT_memory_operand): Likewise.
15950 (call_insn_operand): Likewise.
15951 (sibcall_insn_operand): Likewise.
15952 (GOT32_symbol_operand): Likewise.
15953 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15955 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15957 PR rtl-optimization/83496
15958 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15959 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15960 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15961 redundant insn, if any.
15962 (relax_delay_slots): Likewise.
15963 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15965 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15967 PR tree-optimization/83965
15968 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15969 that grouped statements are part of a reduction chain. Return
15970 true if the statement is not marked as a reduction itself but
15971 is part of a group.
15972 (vect_recog_dot_prod_pattern): Don't check whether the statement
15973 is part of a group here.
15974 (vect_recog_sad_pattern): Likewise.
15975 (vect_recog_widen_sum_pattern): Likewise.
15977 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15980 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15982 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15985 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15987 (ix86_output_indirect_function_return): New prototype.
15988 (ix86_split_simple_return_pop_internal): Likewise.
15989 * config/i386/i386.c (indirect_return_via_cx): New.
15990 (indirect_return_via_cx_bnd): Likewise.
15991 (indirect_thunk_name): Handle return va CX_REG.
15992 (output_indirect_thunk_function): Create alias for
15993 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15994 (ix86_output_indirect_jmp): Remove the bool argument.
15995 (ix86_output_indirect_function_return): New function.
15996 (ix86_split_simple_return_pop_internal): Likewise.
15997 * config/i386/i386.md (*indirect_jump): Don't pass false
15998 to ix86_output_indirect_jmp.
15999 (*tablejump_1): Likewise.
16000 (simple_return_pop_internal): Change it to define_insn_and_split.
16001 Call ix86_split_simple_return_pop_internal to split it for
16002 -mfunction-return=.
16003 (simple_return_indirect_internal): Call
16004 ix86_output_indirect_function_return instead of
16005 ix86_output_indirect_jmp.
16007 2018-02-26 Jakub Jelinek <jakub@redhat.com>
16010 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
16011 memset and value initialization afterwards.
16013 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
16015 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
16017 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16020 * common/config/aarch64/aarch64-common.c
16021 (aarch_option_optimization_table[]): Switch
16022 off fomit-frame-pointer
16024 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16025 Chung-Ju Wu <jasonwucj@gmail.com>
16027 * config/nds32/nds32-multiple.md (load_multiple): Disallow
16029 (store_multiple): Ditto.
16031 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16033 * config.gcc: Add --with-cpu support for nds32 target.
16034 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
16035 * config/nds32/nds32.opt: Add -mcpu= option.
16037 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
16039 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
16040 isel=yes): Warn for these deprecated options.
16042 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
16044 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
16045 ISA_2_5_MASKS_EMBEDDED.
16047 2018-02-23 Jakub Jelinek <jakub@redhat.com>
16049 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
16050 p->max as pointers rather than using iterative_hash_expr.
16052 2018-02-23 Carl Love <cel@us.ibm.com>
16054 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
16055 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
16057 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
16058 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
16059 P8V_BUILTIN_VEC_VUNSIGNED2.
16061 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
16064 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
16065 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
16067 * lra-constraints.c (curr_insn_transform): Set up
16068 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
16070 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
16071 * lra-eliminations.c (spill_pseudos): Ditto.
16072 (process_insn_for_elimination): Ditto.
16073 * lra-lives.c (reg_early_clobber_p): Use the new macros.
16074 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
16075 LRA_NON_CLOBBERED_ALT.
16077 2018-02-22 Martin Sebor <msebor@redhat.com>
16079 PR tree-optimization/84480
16080 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
16081 to maybe_diag_stxncpy_trunc. Call it.
16082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
16083 from gimple_fold_builtin_strcpy. Print inlining stack.
16084 (handle_builtin_stxncpy): Print inlining stack.
16085 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
16087 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
16090 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
16091 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
16092 and -fcheck-pointer-bounds are used together.
16093 (indirect_thunk_prefix): New enum.
16094 (indirect_thunk_need_prefix): New function.
16095 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
16096 "_nt" instead of "_bnd" for NOTRACK prefix.
16097 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
16098 (output_indirect_thunk_function): Likewise.
16100 (ix86_code_end): Update output_indirect_thunk_function calls.
16101 (ix86_output_indirect_branch_via_reg): Replace
16102 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
16103 (ix86_output_indirect_branch_via_push): Likewise.
16104 (ix86_output_function_return): Likewise.
16105 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
16106 incompatible with -fcf-protection=branch and
16107 -fcheck-pointer-bounds.
16109 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16112 * config/aarch64/aarch64.c (aarch64_print_address_internal):
16113 Change gcc_assert call to output_operand_lossage.
16115 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16117 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
16119 2018-02-22 DJ Delorie <dj@redhat.com>
16120 Sebastian Perta <sebastian.perta@renesas.com>
16121 Oleg Endo <olegendo@gcc.gnu.org>
16123 * config/rx/rx.c (rx_rtx_costs): New function.
16124 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
16126 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
16128 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
16130 2018-02-22 Martin Liska <mliska@suse.cz>
16133 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
16134 Add "native" as a possible value.
16136 2018-02-22 Martin Liska <mliska@suse.cz>
16139 * config/i386/i386.c (ix86_option_override_internal):
16140 Add "native" as a possible value for -march and -mtune.
16142 2018-02-22 Jakub Jelinek <jakub@redhat.com>
16145 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
16146 to all type variants.
16148 PR tree-optimization/84503
16149 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
16150 width as info->bitpos + info->bitsize - start.
16151 (merged_store_group::merge_overlapping): Simplify width computation.
16152 (check_no_overlap): New function.
16153 (imm_store_chain_info::try_coalesce_bswap): Compute expected
16154 start + width and last_order of the group, fail if check_no_overlap
16156 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
16157 to group if check_no_overlap fails.
16159 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16161 * config/rs6000/altivec.md: Delete contraint arguments to
16162 define_expand, define_split, and define_peephole2, and in
16163 define_insn_and_split if always unused.
16164 * config/rs6000/darwin.md: Ditto.
16165 * config/rs6000/dfp.md: Ditto.
16166 * config/rs6000/rs6000.md: Ditto.
16167 * config/rs6000/sync.md: Ditto.
16168 * config/rs6000/vector.md: Ditto.
16169 * config/rs6000/vsx.md: Ditto.
16171 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16173 * config/rs6000/altivec.md: Write output control strings as braced
16174 blocks instead of double-quoted strings.
16175 * config/rs6000/darwin.md: Ditto.
16176 * config/rs6000/rs6000.md: Ditto.
16177 * config/rs6000/vector.md: Ditto.
16178 * config/rs6000/vsx.md: Ditto.
16180 2018-02-21 Jason Merrill <jason@redhat.com>
16182 PR c++/84314 - ICE with templates and fastcall attribute.
16183 * attribs.c (build_type_attribute_qual_variant): Remove assert.
16185 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16187 * ipa-cp.c (determine_versionability): Fix comment typos.
16189 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16192 * ipa-cp.c (determine_versionability): Do not version functions caling
16195 2018-02-21 Martin Liska <mliska@suse.cz>
16198 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
16199 Add "native" as a possible value.
16200 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
16201 the macro when native cpu detection is available.
16203 2018-02-21 Martin Liska <mliska@suse.cz>
16206 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
16207 Add "native" as a possible value.
16208 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
16209 when native cpu detection is available.
16211 2018-02-21 Jakub Jelinek <jakub@redhat.com>
16212 Martin Sebor <msebor@redhat.com>
16214 PR tree-optimization/84478
16215 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
16217 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
16218 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
16219 support which is conservatively correct, for 2 only stay conservative
16220 for maxlen. Formatting and comment capitalization fixes. Add STRICT
16221 argument to the 2 argument get_range_strlen, adjust 6 arg
16222 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
16224 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
16225 (gimple_fold_builtin_strlen): Pass true as last argument to
16228 2018-02-20 Martin Sebor <msebor@redhat.com>
16230 PR middle-end/84095
16231 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
16232 (builtin_memref::set_base_and_offset): Same. Handle inner references.
16233 (builtin_memref::builtin_memref): Factor out parts into
16234 set_base_and_offset and call it.
16236 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16238 PR middle-end/84406
16239 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
16240 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
16241 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
16242 search at the associated MODE_INT.
16244 2018-02-20 Jeff Law <law@redhat.com>
16246 PR middle-end/82123
16247 PR tree-optimization/81592
16248 PR middle-end/79257
16249 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
16250 for range data rather than using global data.
16251 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
16252 range data rather than using global data.
16253 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
16254 pass it to children as needed.
16255 (struct directive::fmtresult): Similarly.
16256 (struct directive::set_width): Similarly.
16257 (struct directive::set_precision): Similarly.
16258 (format_integer, format_directive, parse_directive): Similarly.
16259 (format_none): Accept unnamed vr_values parameter.
16260 (format_percent, format_floating, format_character): Similarly.
16261 (format_string, format_plain): Similarly.
16262 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16263 the EVRP range analyzer for range data rather than using global data.
16264 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16265 gimple-ssa-evrp-analyze.h
16266 (class sprintf_dom_walker): Add after_dom_children member function.
16267 Add evrp_range_analyzer member.
16268 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16269 range analyzer as needed.
16270 (sprintf_dom_walker::after_dom_children): New member function.
16271 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16273 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16274 (evrp_range_analyzer::pop_to_marker): Likewise.
16276 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16278 PR tree-optimization/84419
16279 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16280 with the required type if its current type is compatible but
16283 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16285 PR middle-end/82004
16286 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16287 after vectorization.
16289 2018-02-20 Martin Liska <mliska@suse.cz>
16292 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16293 possible values if we don't have a hint.
16295 2018-02-20 Martin Liska <mliska@suse.cz>
16299 * final.c (shorten_branches): Build align_tab array with one
16301 * opts.c (finish_options): Add alignment option limit check.
16302 (MAX_CODE_ALIGN): Likewise.
16303 (MAX_CODE_ALIGN_VALUE): Likewise.
16304 * doc/invoke.texi: Document maximum allowed option value for
16305 all -falign-* options.
16307 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16310 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16311 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16312 * var-tracking.c (emit_note_insn_var_location): Remove all references
16313 to NOTE_INSN_CALL_ARG_LOCATION.
16314 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16315 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16316 Use copy_rtx_if_shared.
16317 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16318 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16319 (dwarf2out_var_location): Remove handling of
16320 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16322 * final.c (final_scan_insn): Remove all references to
16323 NOTE_INSN_CALL_ARG_LOCATION.
16324 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16325 before dumping final insns.
16326 * except.c (emit_note_eh_region_end): Remove all references to
16327 NOTE_INSN_CALL_ARG_LOCATION.
16328 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16329 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16330 * config/arc/arc.c (hwloop_optimize): Likewise.
16331 * config/arm/arm.c (create_fix_barrier): Likewise.
16332 * config/s390/s390.c (s390_chunkify_start): Likewise.
16333 * config/sh/sh.c (find_barrier): Likewise.
16334 * config/i386/i386.c (rest_of_insert_endbranch,
16335 ix86_seh_fixup_eh_fallthru): Likewise.
16336 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16337 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16338 * config/frv/frv.c (frv_function_prologue): Likewise.
16339 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16341 (note_outside_basic_block_p): Remove all references to
16342 NOTE_INSN_CALL_ARG_LOCATION.
16343 * gengtype.c (adjust_field_rtx_def): Likewise.
16344 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16346 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16347 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16350 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16353 PR tree-optimization/84452
16354 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16355 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16358 2018-02-19 Martin Liska <mliska@suse.cz>
16361 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16363 2018-02-19 Martin Liska <mliska@suse.cz>
16364 Richard Sandiford <richard.sandiford@linaro.org>
16366 PR tree-optimization/82491
16367 * gimple-fold.c (get_base_constructor): Make earlier bail out
16370 2018-02-19 Carl Love <cel@us.ibm.com>
16372 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16373 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16375 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16376 P8V_BUILTIN_VEC_NEG.
16378 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16380 * config/rl78/rl78.md (movdf): New define expand.
16382 2018-02-19 Martin Liska <mliska@suse.cz>
16385 * doc/invoke.texi: Fix typo.
16386 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16388 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16390 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16391 handle rs6000_single_float and rs6000_double_float specially for
16394 2018-02-16 Jeff Law <law@redhat.com>
16396 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16397 (pop_regs): Use it.
16399 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16402 * ipa-inline.c (inline_small_functions): Fix a typo.
16404 2018-02-16 Nathan Sidwell <nathan@acm.org>
16406 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16408 2018-02-16 Carl Love <cel@us.ibm.com>
16410 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16411 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16412 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16413 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16414 expansion to P8V_BUILTIN_VEC_FLOAT2.
16416 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16418 PR rtl-optimization/70023
16419 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16420 src_regno into account.
16422 2018-02-16 Carl Love <cel@us.ibm.com>
16424 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16425 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16426 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16427 * config/rs6000/rs6000.c: Remove case statements for
16428 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16429 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16430 and P9V_BUILTIN_VEC_VINSERT4B.
16431 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16432 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16433 * config/rs6000/vsx.md:
16434 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16437 2018-02-16 Carl Love <cel@us.ibm.com>
16439 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16441 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16443 * config/rs6000/rs6000-c.c: Add the definitions for
16444 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16445 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16446 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16447 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16448 definition for insert4b and define insn *insert3b_internal.
16449 * doc/extend.texi: Add documentation for vec_extract4b.
16451 2018-02-16 Nathan Sidwell <nathan@acm.org>
16453 * doc/extend.texi (Backwards Compatibility): Mention friend
16454 injection. Note for-scope is deprecated.
16455 * doc/invoke.texi (-ffriend-injection): Deprecate.
16457 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16459 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16460 that moved to I2, also allow destinations that are a paradoxical
16461 subreg (instead of a normal reg).
16463 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16466 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16469 2018-02-16 Richard Biener <rguenther@suse.de>
16471 PR tree-optimization/84037
16472 PR tree-optimization/84016
16474 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16475 Adjust vec_construct for the fact we need additional higher latency
16476 128bit inserts for AVX256 and AVX512 vector builds.
16477 (ix86_add_stmt_cost): Scale vector construction cost for
16480 2018-02-16 Richard Biener <rguenther@suse.de>
16482 PR tree-optimization/84417
16483 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16484 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16485 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16487 2018-02-16 Martin Liska <mliska@suse.cz>
16490 * internal-fn.def (ASAN_CHECK): Set proper flags.
16491 (ASAN_MARK): Likewise.
16493 2018-02-16 Julia Koval <julia.koval@intel.com>
16495 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16496 from PTA_CANNONLAKE.
16498 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16501 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16502 Use ++iter rather than iter++ for std::list iterators.
16503 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16504 defer deleting them until all nodes in the forest are processed. Do
16505 free even leaf nodes. Change to_process into auto_vec.
16508 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16509 * vec.h (vec_default_construct): Use memset instead of placement new
16510 if BROKEN_VALUE_INITIALIZATION is defined.
16511 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16512 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16515 PR rtl-optimization/83723
16516 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16517 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16518 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16520 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16522 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16524 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16526 PR rtl-optimization/81443
16527 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16528 from inner REGs to paradoxical SUBREGs.
16530 2018-02-16 Richard Biener <rguenther@suse.de>
16532 PR tree-optimization/84399
16533 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16534 For operands we can analyze at their definition make sure we can
16535 analyze them at each use as well.
16537 2018-02-16 Richard Biener <rguenther@suse.de>
16539 PR tree-optimization/84190
16540 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16541 volatile accesses if the decl isn't volatile.
16543 2018-02-15 Jason Merrill <jason@redhat.com>
16545 PR c++/84314 - ICE with templates and fastcall attribute.
16546 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16547 TYPE_CANONICAL on an existing type.
16549 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16551 PR tree-optimization/84383
16552 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16553 dstoff nor call operand_equal_p if dstbase is NULL.
16555 PR tree-optimization/84334
16556 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16557 also a CONSTANT_CLASS_P, punt.
16559 2018-02-14 Jim Wilson <jimw@sifive.com>
16561 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16562 first SMALL_OPERAND check. New local min_second_step. Move assert
16563 to where locals are set. Add TARGET_RVC support.
16564 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16566 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16568 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16570 2018-02-14 Martin Sebor <msebor@redhat.com>
16572 PR tree-optimization/83698
16573 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16574 arrays constrain the offset range to their bounds.
16575 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16576 (builtin_access::overlap): Avoid setting the size of overlap if it's
16578 (maybe_diag_overlap): Also consider arrays when deciding what values
16579 of offsets to include in diagnostics.
16581 2018-02-14 Martin Sebor <msebor@redhat.com>
16584 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16585 that correspond to the kind of a declaration.
16587 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16590 * config/pa/pa.md: Load address of PIC label using the linkage table
16591 if the label is nonlocal.
16593 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16595 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16596 warning message if user requests -maltivec=be.
16597 * doc/invoke.texi: Document deprecation of -maltivec=be.
16599 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16602 * config/rs6000/rs6000-c.c: Update definitions for
16603 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16604 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16606 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16609 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16610 add _get_ssp intrinsics. Remove argument from
16611 __builtin_ia32_rdssp[d|q].
16612 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16613 * config/i386/i386-builtin.def: Remove argument from
16614 __builtin_ia32_rdssp[d|q].
16615 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16616 ix86_expand_special_args_builtin for _rdssp[d|q].
16617 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16618 Clear register before usage.
16619 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16620 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16622 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16624 PR tree-optimization/84357
16625 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16626 operand 1 of an ARRAY_REF too.
16628 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16631 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16632 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16634 (set_of_reg): New struct.
16635 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16636 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16637 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16639 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16640 Split into bitclr, bitset, bitinvert patterns if appropriate.
16641 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16642 use rx_fuse_in_memory_bitop.
16643 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16644 to named insn, correct maximum insn length.
16646 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16649 * machmode.def: Define a complex mode for PARTIAL_INT.
16650 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16652 * doc/rtl.texi: Document CSPImode.
16653 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16655 (msp430_hard_regno_nregs_with_padding): Likewise.
16657 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16660 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16662 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16664 PR rtl-optimization/84169
16665 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16666 we generated a parallel as new i3 and we split that to new i2 and i3
16667 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16668 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16669 those to i2, not i1. Partially rewrite this scan code.
16671 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16674 * stor-layout.c (place_field): For variable length fields, adjust
16675 offset_align afterwards not just based on the field's alignment,
16676 but also on the size.
16678 PR middle-end/84309
16679 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16680 of exps and logs in the use_exp2 case.
16682 2018-02-13 Jeff Law <law@redhat.com>
16684 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16685 entry for "vector".
16687 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16690 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16694 * common.opt (gas-loc-support, gas-locview-support): New.
16695 (ginline-points, ginternal-reset-location-views): New.
16696 * doc/invoke.texi: Document them. Use @itemx where intended.
16697 (gvariable-location-views): Adjust.
16698 * target.def (reset_location_view): New.
16699 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16700 (TARGET_RESET_LOCATION_VIEW): New.
16701 * doc/tm.texi: Rebuilt.
16702 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16703 (dwarf2out_default_as_locview_support): New.
16704 (output_asm_line_debug_info): Use option variables.
16705 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16706 (output_loc_list): Likewise.
16707 (add_high_low_attributes): Check option variables.
16708 Don't output entry view attribute in strict mode.
16709 (gen_inlined_subroutine_die): Check option variables.
16710 (dwarf2out_inline_entry): Likewise.
16711 (init_sections_and_labels): Likewise.
16712 (dwarf2out_early_finish): Likewise.
16713 (maybe_reset_location_view): New, from...
16714 (dwarf2out_var_location): ... here. Call it.
16715 * debug.h (dwarf2out_default_as_loc_support): Declare.
16716 (dwarf2out_default_as_locview_support): Declare.
16717 * hooks.c (hook_int_rtx_insn_0): New.
16718 * hooks.h (hook_int_rtx_insn_0): Declare.
16719 * toplev.c (process_options): Take -gas-loc-support and
16720 -gas-locview-support from dwarf2out. Enable
16721 -gvariable-location-views by default only with locview
16722 assembler support. Enable -ginternal-reset-location-views by
16723 default only if the target defines the corresponding hook.
16724 Enable -ginline-points by default if location views are
16725 enabled; force it disabled if statement frontiers are
16727 * tree-inline.c (expand_call_inline): Check option variables.
16728 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16730 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16732 PR tree-optimization/84321
16733 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16734 handling. Also check whether the anti-range contains any values
16735 that satisfy the mask; switch to a VR_RANGE if not.
16737 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16740 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16742 2018-02-13 Martin Jambor <mjambor@suse.cz>
16745 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16746 of call statements, also set location of a load to a temporary.
16748 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16750 * config/rl78/rl78.c (add_vector_labels): New function.
16751 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16752 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16753 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16754 which checks that no arguments are passed.
16755 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16756 * doc/extend.texi: Documentation for the new attribute.
16758 2018-02-13 Andreas Schwab <schwab@suse.de>
16760 * config/riscv/linux.h (CPP_SPEC): Define.
16762 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16765 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16766 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16767 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16768 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16769 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16770 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16771 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16772 temporarily for AES and PCLMUL builtins.
16774 PR tree-optimization/84339
16775 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16776 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16779 PR middle-end/84309
16780 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16781 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16782 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16784 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16786 * omp-simd-clone.h: New file.
16787 * omp-simd-clone.c: Include omp-simd-clone.h.
16788 (expand_simd_clones): No longer static.
16789 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16790 cgraph.h and omp-simd-clone.h.
16791 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16792 (vect_recog_widen_shift_pattern): Formatting fix.
16793 (vect_pattern_recog_1): Don't check optab for calls.
16796 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16797 operands[2] into a REG before using gen_lowpart on it.
16799 2018-02-12 Jeff Law <law@redhat.com>
16802 * config/sh/sh.c (find_barrier): Consider a sibling call
16805 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16806 successfully back substituting a reg.
16808 2018-02-12 Richard Biener <rguenther@suse.de>
16810 PR tree-optimization/84037
16811 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16812 parameter, move visited init to caller.
16813 (vect_slp_analyze_operations): Separate cost from validity
16814 check, initialize visited once for all instances.
16815 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16817 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16819 (vect_model_promotion_demotion_cost): Likewise.
16820 (vectorizable_bswap): Guard cost modeling with !slp_node
16821 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16823 (vectorizable_call): Likewise.
16824 (vectorizable_conversion): Likewise.
16825 (vectorizable_assignment): Likewise.
16826 (vectorizable_shift): Likewise.
16827 (vectorizable_operation): Likewise.
16828 (vectorizable_store): Likewise.
16829 (vectorizable_load): Likewise.
16830 (vectorizable_condition): Likewise.
16831 (vectorizable_comparison): Likewise.
16833 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16836 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16837 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16838 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16840 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16842 PR middle-end/83665
16843 * params.def (inline-min-speedup): Increase from 8 to 15.
16844 (max-inline-insns-auto): Decrease from 40 to 30.
16845 * ipa-split.c (consider_split): Add some buffer for function to
16846 be considered inlining candidate.
16847 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16850 2018-02-12 Richard Biener <rguenther@suse.de>
16852 PR tree-optimization/84037
16853 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16854 matched stmts if we cannot swap the non-matched ones.
16856 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16858 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16859 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16860 _mm_maskz_scalef_round_ss): New intrinsics.
16861 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16862 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16863 __builtin_ia32_scalefss_round): Remove.
16864 (__builtin_ia32_scalefsd_mask_round,
16865 __builtin_ia32_scalefss_mask_round): New intrinsics.
16866 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16867 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16868 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16869 "<round_constraint>")): Changed to ...
16870 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16871 "<round_scalar_constraint>")): ... this.
16872 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16873 %0, %1, %2<round_op3>}"): Changed to ...
16874 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16875 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16876 %2<round_scalar_mask_op3>}"): ... this.
16877 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16879 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16881 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16882 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16883 (_mm_maskz_sqrt_round_ss): New intrinsics.
16884 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16885 (__builtin_ia32_sqrtsd_mask_round)
16886 (__builtin_ia32_sqrtss_mask_round): New builtins.
16887 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16888 (__builtin_ia32_sqrtss_round): Remove.
16889 (__builtin_ia32_sqrtsd_mask_round)
16890 (__builtin_ia32_sqrtss_mask_round): New builtins.
16891 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16892 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16893 ((match_operand:VF_128 1 "vector_operand"
16894 "xBm,<round_constraint>")): Changed to ...
16895 ((match_operand:VF_128 1 "vector_operand"
16896 "xBm,<round_scalar_constraint>")): ... this.
16897 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16898 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16899 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16900 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16901 %<iptr>1<round_scalar_mask_op3>}): ... this.
16902 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16903 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16905 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16908 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16909 Cast vec_cmpeq result to correct type.
16910 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16911 Cast vec_cmpgt result to correct type.
16913 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16915 * final.c (final_scan_insn_1): Renamed from...
16916 (final_scan_insn): ... this. New wrapper, to recover
16917 seen from the outermost call in recursive ones.
16918 * config/sparc/sparc.c (output_return): Drop seen from call.
16919 (output_sibcall): Likewise.
16920 * config/visium/visium.c (output_branch): Likewise.
16922 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16924 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16927 2018-02-10 Alan Modra <amodra@gmail.com>
16930 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16931 Specify LR as an input.
16933 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16936 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16937 remove_member_access_dummy_vars): New functions.
16938 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16939 lower_omp_1, execute_lower_omp): Use them.
16941 PR rtl-optimization/84308
16942 * shrink-wrap.c (spread_components): Release todo vector.
16944 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16946 PR rtl-optimization/57193
16947 * ira-color.c (struct allocno_color_data): Add member
16948 conflict_allocno_hard_prefs.
16949 (update_conflict_allocno_hard_prefs): New.
16950 (bucket_allocno_compare_func): Add a preference based on
16951 conflict_allocno_hard_prefs.
16952 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16953 (color_allocnos): Remove a dead code. Initiate
16954 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16956 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16959 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16960 constraint from =wa to wa. Avoid a subreg on the output operand,
16961 instead use a pseudo and subreg it in a move.
16962 (p9_xxbrd_<mode>): Changed to ...
16963 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16964 (p9_xxbrd_v2df): New expander.
16965 (p9_xxbrw_<mode>): Changed to ...
16966 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16967 (p9_xxbrw_v4sf): New expander.
16969 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16971 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16973 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16976 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16977 multiply in 32-bit mode.
16978 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16979 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16982 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16984 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16985 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16986 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16987 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16989 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16992 * dwarf2out.c (is_trivial_indirect_ref): New function.
16993 (dwarf2out_late_global_decl): Do not generate a location
16994 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16995 and that are not defined in the current unit.
16997 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16999 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
17000 instead of a libcall for UNORDERED.
17002 2018-02-09 Tamar Christina <tamar.christina@arm.com>
17005 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
17006 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
17008 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17011 * config/s390/s390.c (s390_set_current_function): Invoke
17012 s390_indirect_branch_settings also if fndecl didn't change.
17014 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17016 * config/rs6000/rs6000.md (blockage): Set length to zero.
17018 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17020 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
17022 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17025 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
17026 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
17030 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
17031 PARALLEL incoming that failed vt_get_decl_and_offset check.
17033 PR middle-end/84237
17034 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
17035 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
17037 (get_variable_section): For decls in named .bss* sections pass true as
17038 second argument to bss_initializer_p.
17040 2018-02-09 Marek Polacek <polacek@redhat.com>
17041 Jakub Jelinek <jakub@redhat.com>
17044 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
17045 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
17046 Sync some changes from cxx_fold_indirect_ref.
17048 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17050 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
17052 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
17053 (BLOCK_INLINE_ENTRY_LABEL): New.
17054 (dwarf2out_var_location): Disregard inline entry markers.
17055 (inline_entry_data): New struct.
17056 (inline_entry_data_hasher): New hashtable type.
17057 (inline_entry_data_hasher::hash): New.
17058 (inline_entry_data_hasher::equal): New.
17059 (inline_entry_data_table): New variable.
17060 (add_high_low_attributes): Add DW_AT_entry_pc and
17061 DW_AT_GNU_entry_view attributes if a pending entry is found
17062 in inline_entry_data_table. Add old entry_pc attribute only
17063 if debug nonbinding markers are disabled.
17064 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
17065 markers are enabled.
17066 (block_within_block_p, dwarf2out_inline_entry): New.
17067 (dwarf2out_finish): Check that no entries remained in
17068 inline_entry_data_table.
17069 * final.c (reemit_insn_block_notes): Handle inline entry notes.
17070 (final_scan_insn, notice_source_line): Likewise.
17071 (rest_of_clean_state): Skip inline entry markers.
17072 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
17074 * gimple.c (gimple_build_debug_inline_entry): New.
17075 * gimple.h (enum gimple_debug_subcode): Add
17076 GIMPLE_DEBUG_INLINE_ENTRY.
17077 (gimple_build_debug_inline_entry): Declare.
17078 (gimple_debug_inline_entry_p): New.
17079 (gimple_debug_nonbind_marker_p): Adjust.
17080 * insn-notes.def (INLINE_ENTRY): New.
17081 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
17082 inline entry marker notes.
17083 (print_insn): Likewise.
17084 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
17085 (INSN_DEBUG_MARKER_KIND): Likewise.
17086 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
17087 * tree-inline.c (expand_call_inline): Build and insert
17088 debug_inline_entry stmt.
17089 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
17090 inline entry blocks early, if nonbind markers are enabled.
17091 (dump_scope_block): Dump fragment info.
17092 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
17093 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
17094 (gimple_build_debug_inline_entry): New.
17095 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
17096 Enable/disable inline entry points too.
17097 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
17098 (DEBUG_INSN): Describe inline entry markers.
17100 * common.opt (gvariable-location-views): New.
17101 (gvariable-location-views=incompat5): New.
17102 * config.in: Rebuilt.
17103 * configure: Rebuilt.
17104 * configure.ac: Test assembler for view support.
17105 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
17106 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
17107 * dwarf2out.c (var_loc_view): New typedef.
17108 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
17109 (dwarf2out_locviews_in_attribute): New.
17110 (dwarf2out_locviews_in_loclist): New.
17111 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
17112 (enum dw_line_info_opcode): Add LI_adv_address.
17113 (struct dw_line_info_table): Add view.
17114 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
17115 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
17116 (zero_view_p): New variable.
17117 (ZERO_VIEW_P): New macro.
17118 (output_asm_line_debug_info): New.
17119 (struct var_loc_node): Add view.
17120 (add_AT_view_list, AT_loc_list): New.
17121 (add_var_loc_to_decl): Add view param. Test it against last.
17122 (new_loc_list): Add view params. Record them.
17123 (AT_loc_list_ptr): Handle loc and view lists.
17124 (view_list_to_loc_list_val_node): New.
17125 (print_dw_val): Handle dw_val_class_view_list.
17126 (size_of_die): Likewise.
17127 (value_format): Likewise.
17128 (loc_list_has_views): New.
17129 (gen_llsym): Set vl_symbol too.
17130 (maybe_gen_llsym, skip_loc_list_entry): New.
17131 (dwarf2out_maybe_output_loclist_view_pair): New.
17132 (output_loc_list): Output view list or entries too.
17133 (output_view_list_offset): New.
17134 (output_die): Handle dw_val_class_view_list.
17135 (output_dwarf_version): New.
17136 (output_compilation_unit_header): Use it.
17137 (output_skeleton_debug_sections): Likewise.
17138 (output_rnglists, output_line_info): Likewise.
17139 (output_pubnames, output_aranges): Update version comments.
17140 (output_one_line_info_table): Output view numbers in asm comments.
17141 (dw_loc_list): Determine current endview, pass it to new_loc_list.
17142 Call maybe_gen_llsym.
17143 (loc_list_from_tree_1): Adjust.
17144 (add_AT_location_description): Create view list attribute if
17145 needed, check it's absent otherwise.
17146 (convert_cfa_to_fb_loc_list): Adjust.
17147 (maybe_emit_file): Call output_asm_line_debug_info for test.
17148 (dwarf2out_var_location): Reset views as needed. Precompute
17149 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
17150 attribute. Set view.
17151 (new_line_info_table): Reset next view.
17152 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
17153 (dwarf2out_source_line): Likewise. Output view resets and labels to
17154 the assembler, or select appropriate line info opcodes.
17155 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
17156 (optimize_string_length): Catch it. Adjust.
17157 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
17158 dw_val_class_view_list, and remove it if no longer needed.
17159 (hash_loc_list): Hash view numbers.
17160 (loc_list_hasher::equal): Compare them.
17161 (optimize_location_lists): Check whether a view list symbol is
17162 needed, and whether the locview attribute is present, and
17163 whether they match. Remove the locview attribute if no longer
17165 (index_location_lists): Call skip_loc_list_entry for test.
17166 (dwarf2out_finish): Call output_asm_line_debug_info for test.
17167 Use output_dwarf_version.
17168 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
17169 (struct dw_val_node): Add val_view_list.
17170 * final.c (SEEN_NEXT_VIEW): New.
17171 (set_next_view_needed): New.
17172 (clear_next_view_needed): New.
17173 (maybe_output_next_view): New.
17174 (final_start_function): Rename to...
17175 (final_start_function_1): ... this. Take pointer to FIRST,
17176 add SEEN parameter. Emit param bindings in the initial view.
17177 (final_start_function): Reintroduce SEEN-less interface.
17178 (final): Rename to...
17179 (final_1): ... this. Take SEEN parameter. Output final pending
17180 next view at the end.
17181 (final): Reintroduce seen-less interface.
17182 (final_scan_insn): Output pending next view before switching
17183 sections or ending a block. Mark the next view as needed when
17184 outputting variable locations. Notify debug backend of section
17185 changes, and of location view changes.
17186 (rest_of_handle_final): Adjust.
17187 * toplev.c (process_options): Autodetect value for debug variable
17188 location views option. Warn on incompat5 without -gdwarf-5.
17189 * doc/invoke.texi (gvariable-location-views): New.
17190 (gvariable-location-views=incompat5): New.
17191 (gno-variable-location-views): New.
17193 2018-02-08 David Malcolm <dmalcolm@redhat.com>
17195 PR tree-optimization/84136
17196 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
17197 that the result of find_edge is non-NULL.
17199 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17202 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
17203 storing integer register in SImode. Fix cost of 256 and 512
17204 byte aligned SSE register store.
17206 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17208 * config/i386/i386.c (ix86_multiplication_cost): Fix
17209 multiplication cost for TARGET_AVX512DQ.
17211 2018-02-08 Marek Polacek <polacek@redhat.com>
17213 PR tree-optimization/84238
17214 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
17217 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17219 PR tree-optimization/84265
17220 * tree-vect-stmts.c (vectorizable_store): Don't treat
17221 VMAT_CONTIGUOUS accesses as grouped.
17222 (vectorizable_load): Likewise.
17224 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17226 PR tree-optimization/81635
17227 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
17228 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
17229 (test_round_for_mask): New functions.
17230 (wide_int_cc_tests): Call test_round_for_mask.
17231 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
17232 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
17233 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
17234 range returned by get_range_info.
17236 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17239 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
17240 * symtab.c: Include builtins.h
17241 (symtab_node::output_to_lto_symbol_table_p): Move here
17242 from lto-streamer-out.c:output_symbol_p.
17243 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
17244 (output_symbol_p): Move all logic to symtab.c
17245 (produce_symtab): Update.
17247 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17249 * config/s390/s390-opts.h (enum indirect_branch): Define.
17250 * config/s390/s390-protos.h (s390_return_addr_from_memory)
17251 (s390_indirect_branch_via_thunk)
17252 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
17253 (enum s390_indirect_branch_type): Define.
17254 * config/s390/s390.c (struct s390_frame_layout, struct
17255 machine_function): Remove.
17256 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
17257 (indirect_branch_table_label_no, indirect_branch_table_name):
17259 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
17260 (enum s390_indirect_branch_option): Define.
17261 (s390_return_addr_from_memory): New function.
17262 (s390_handle_string_attribute): New function.
17263 (s390_attribute_table): Add new attribute handler.
17264 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17265 (s390_indirect_branch_via_thunk): New function.
17266 (s390_indirect_branch_via_inline_thunk): New function.
17267 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17268 sibling call optimization for non z10 compiles.
17269 (s390_emit_call): Force indirect branch target to be a single
17270 register. Add r1 clobber for non-z10 compiles.
17271 (s390_emit_epilogue): Emit return jump via return_use expander.
17272 (s390_reorg): Handle JUMP_INSNs as execute targets.
17273 (s390_option_override_internal): Perform validity checks for the
17274 new command line options.
17275 (s390_indirect_branch_attrvalue): New function.
17276 (s390_indirect_branch_settings): New function.
17277 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17278 (s390_output_indirect_thunk_function): New function.
17279 (s390_code_end): Implement target hook.
17280 (s390_case_values_threshold): Implement target hook.
17281 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17283 * config/s390/s390.h (struct s390_frame_layout)
17284 (struct machine_function): Move here from s390.c.
17285 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17286 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17287 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17288 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17289 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17290 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17291 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17292 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17293 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17294 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17295 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17296 (mnemonic attribute): Add values which aren't recognized
17298 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17299 pattern for branch conversion. Fix mnemonic attribute.
17300 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17301 indirect branch via thunk if requested.
17302 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17303 ("*indirect_jump"): Disable for branch conversion using out of
17305 ("indirect_jump_via_thunk<mode>_z10")
17306 ("indirect_jump_via_thunk<mode>")
17307 ("indirect_jump_via_inlinethunk<mode>_z10")
17308 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17309 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17310 ("casesi_jump_via_inlinethunk<mode>_z10")
17311 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17312 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17313 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17314 ("*indirect2_jump"): Disable for branch conversion.
17315 ("casesi_jump"): Turn into expander and expand patterns for branch
17317 ("return_use"): New expander.
17318 ("*return"): Emit return via thunk and rename it to ...
17319 ("*return<mode>"): ... this one.
17320 * config/s390/s390.opt: Add new options and and enum for the
17323 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17325 * lra-constraints.c (match_reload): Unconditionally use
17326 gen_lowpart_SUBREG, rather than selecting between that
17327 and equivalent gen_rtx_SUBREG code.
17329 2018-02-08 Richard Biener <rguenther@suse.de>
17331 PR tree-optimization/84233
17332 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17333 changed flag instead of boguously re-using phi_inserted.
17335 2018-02-08 Martin Jambor <mjambor@suse.cz>
17337 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17338 static local variables.
17340 2018-02-08 Richard Biener <rguenther@suse.de>
17342 PR tree-optimization/84278
17343 * tree-vect-stmts.c (vectorizable_store): When looking for
17344 smaller vector types to perform grouped strided loads/stores
17345 make sure the mode is supported by the target.
17346 (vectorizable_load): Likewise.
17348 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17350 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17351 Increase LDP/STP opportunities by adding adjacent callee-saves.
17353 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17355 PR rtl-optimization/84068
17356 PR rtl-optimization/83459
17357 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17359 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17361 PR tree-optimization/84224
17362 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17363 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17364 non-zero arguments.
17366 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17369 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17370 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17371 count, remove one USE.
17373 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17375 * doc/install.texi (Configuration): Document the
17376 --with-long-double-format={ibm,ieee} PowerPC configuration
17380 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17381 Convert from define_expand to be define_insn_and_split. Rework
17382 float/double/_Float128 conversions to QI/HI/SImode to work with
17383 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17384 conversions to QI/HImode types did a store and then a load to
17385 truncate the value. For conversions to VSX registers, don't split
17386 the insn, instead emit the code directly. Use the code iterator
17387 any_fix to combine signed and unsigned conversions.
17388 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17389 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17390 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17391 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17392 (fix_<mode>di2_hw): Likewise.
17393 (fixuns_<mode>di2_hw): Likewise.
17394 (fix_<mode>si2_hw): Likewise.
17395 (fixuns_<mode>si2_hw): Likewise.
17396 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17397 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17398 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17399 fix<uns>_trunc<SFDF:mode>si2_p8.
17400 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17401 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17402 (fix<uns>_<mode>_mem): Likewise.
17403 (fctiw<u>z_<mode>_mem): Likewise.
17404 (fix<uns>_<mode>_mem): Likewise.
17405 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17406 the register allocator from doing a direct move to the GPRs to do
17407 a store, and instead use the ISA 3.0 store byte/half-word from
17408 vector register instruction. For IEEE 128-bit floating point,
17409 also optimize stores of 32-bit ints.
17410 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17412 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17414 * genextract.c (push_pathstr_operand): New function to support
17416 (walk_rtx): Call push_pathstr_operand.
17417 (print_path): Support [a-zA-Z].
17419 2018-02-07 Richard Biener <rguenther@suse.de>
17421 PR tree-optimization/84037
17422 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17423 (cse_and_gimplify_to_preheader): Declare.
17424 (vect_get_place_in_interleaving_chain): Likewise.
17425 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17427 (_loop_vec_info::~_loop_vec_info): Delete it.
17428 (cse_and_gimplify_to_preheader): New function.
17429 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17430 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17431 (vectorizable_load): Likewise. For grouped stores always base
17432 the IV on the first element.
17433 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17434 condition before gimplifying.
17436 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17438 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17439 *DIV_EXPR and *MOD_EXPR.
17441 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17444 * config/i386/i386.c (ix86_option_override_internal): Mask out
17445 the CF_SET bit when checking -fcf-protection.
17447 2018-02-07 Tom de Vries <tom@codesourcery.com>
17450 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17453 2018-02-07 Richard Biener <rguenther@suse.de>
17455 PR tree-optimization/84204
17456 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17459 PR tree-optimization/84205
17460 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17461 special-case isl_ast_op_zdiv_r.
17463 PR tree-optimization/84223
17464 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17465 Only add conditions from within the region.
17466 (gather_bbs::after_dom_children): Adjust.
17468 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17471 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17472 * config/avr/avr.md: Only post-reload split REG-REG moves if
17473 either register is GENERAL_REG_P.
17475 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17477 PR tree-optimization/84235
17478 * tree-ssa-scopedtables.c
17479 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17480 if the subtraction is performed in floating point type where NaNs are
17481 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17482 build 1. Formatting fix.
17484 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17487 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17488 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17489 and skip it regardless of bb boundaries. Use CALL_P macro,
17490 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17491 unnecessarily, formatting fix.
17493 2018-02-06 Michael Collison <michael.collison@arm.com>
17495 * config/arm/thumb2.md:
17496 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17497 (*thumb_mov_notscc): Ditto.
17499 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17502 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17503 unsigned_fix, not "s".
17505 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17507 * configure.ac (gcc_fn_eh_frame_ro): New function.
17508 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17509 correct .eh_frame permissions.
17510 * configure: Regenerate.
17512 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17514 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17515 irrelevant options.
17517 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17519 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17520 Display warning message for -mno-speculate-indirect-jumps.
17522 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17524 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17526 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17528 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17530 PR tree-optimization/84225
17531 * tree-eh.c (find_trapping_overflow): Only call
17532 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17534 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17537 * config/i386/i386.c: Reimplement the check of possible options
17538 -mibt/-mshstk conbination. Change error messages.
17539 * doc/invoke.texi: Fix a typo: remove extra '='.
17541 2018-02-06 Marek Polacek <polacek@redhat.com>
17543 PR tree-optimization/84228
17544 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17546 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17549 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17550 emitted arch directives.
17551 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17552 __ARM_FEATURE_COPROC before changing architectures.
17554 2018-02-06 Richard Biener <rguenther@suse.de>
17556 * config/i386/i386.c (print_reg): Fix typo.
17557 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17559 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17561 * configure: Regenerate.
17563 2018-02-05 Martin Sebor <msebor@redhat.com>
17565 PR tree-optimization/83369
17566 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17569 2018-02-05 Martin Liska <mliska@suse.cz>
17571 * doc/invoke.texi: Cherry-pick upstream r323995.
17573 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17575 * ira.c (ira_init_register_move_cost): Adjust comment.
17577 2018-02-05 Martin Liska <mliska@suse.cz>
17579 PR gcov-profile/84137
17580 * doc/gcov.texi: Fix typo in documentation.
17582 2018-02-05 Martin Liska <mliska@suse.cz>
17584 PR gcov-profile/83879
17585 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17586 using dlopen functionality.
17588 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17590 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17591 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17592 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17593 _mm_maskz_range_round_ss): New intrinsics.
17594 (__builtin_ia32_rangesd128_round)
17595 (__builtin_ia32_rangess128_round): Remove.
17596 (__builtin_ia32_rangesd128_mask_round,
17597 __builtin_ia32_rangess128_mask_round): New builtins.
17598 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17599 __builtin_ia32_rangess128_round): Remove.
17600 (__builtin_ia32_rangesd128_mask_round,
17601 __builtin_ia32_rangess128_mask_round): New builtins.
17602 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17603 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17604 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17605 "<round_saeonly_constraint>")): Changed to ...
17606 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17607 "<round_saeonly_scalar_constraint>")): ... this.
17608 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17609 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17610 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17611 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17612 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17614 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17616 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17618 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17619 Remove all values except native, 8540 and 8548.
17621 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17623 * config/i386/i386.c (ix86_output_function_return): Pass
17624 INVALID_REGNUM, instead of -1, as invalid register number to
17625 indirect_thunk_name and output_indirect_thunk.
17627 2018-02-02 Julia Koval <julia.koval@intel.com>
17629 * config.gcc: Add -march=icelake.
17630 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17631 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17632 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17633 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17634 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17635 (processor_target_table): Add icelake.
17636 (ix86_option_override_internal): Handle new PTAs.
17637 (get_builtin_code_for_version): Handle icelake.
17638 (M_INTEL_COREI7_ICELAKE): New.
17639 (fold_builtin_cpu): Handle icelake.
17640 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17641 * doc/invoke.texi: Add -march=icelake.
17643 2018-02-02 Julia Koval <julia.koval@intel.com>
17645 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17646 to wide_int_bitmask.
17647 * wide-int-bitmask.h: New.
17649 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17652 * config/i386/i386.md: Replace Pmode with word_mode in
17653 builtin_setjmp_setup and builtin_longjmp to support x32.
17655 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17659 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17661 (rs6000_supported_cpu_names): New static variable.
17662 (linux_cpu_translation_table): Likewise.
17663 (elf_platform) <cpu>: Define new static variable and use it.
17664 Translate kernel AT_PLATFORM name to canonical name if needed.
17665 Error if platform name is unknown.
17667 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17670 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17672 2018-02-01 Jeff Law <law@redhat.com>
17675 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17676 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17677 the scratch if RELEASE_VIA_POP is false.
17678 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17679 If we have to save a temporary register, decrement SIZE appropriately.
17680 Pass new arguments to release_scratch_register_on_entry.
17681 (ix86_adjust_stack_and_probe): Likewise.
17682 (ix86_emit_probe_stack_range): Pass new arguments to
17683 release_scratch_register_on_entry.
17685 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17687 PR rtl-optimization/84157
17688 * combine.c (change_zero_ext): Use REG_P predicate in
17689 front of HARD_REGISTER_P predicate.
17691 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17693 * config/avr/avr.c (avr_option_override): Move disabling of
17694 -fdelete-null-pointer-checks to...
17695 * common/config/avr/avr-common.c (avr_option_optimization_table):
17698 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17700 PR tree-optimization/81635
17701 * tree-data-ref.c (split_constant_offset_1): For types that
17702 wrap on overflow, try to use range info to prove that wrapping
17705 2018-02-01 Renlin Li <renlin.li@arm.com>
17708 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17709 TAILCALL_ADDR_REGS.
17710 (aarch64_register_move_cost): Likewise.
17711 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17712 TAILCALL_ADDR_REGS.
17713 (REG_CLASS_NAMES): Likewise.
17714 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17715 TAILCALL_ADDR_REGS. Remove IP registers.
17716 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17718 2018-02-01 Richard Biener <rguenther@suse.de>
17720 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17721 for specifying RPO order and allow NULL for that.
17722 * domwalk.c (dom_walker::dom_walker): Likewise.
17723 (dom_walker::walk): Handle NULL RPO order.
17724 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17726 (rewrite_update_dom_walker): Likewise.
17727 (mark_def_dom_walker): Likewise.
17729 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17731 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17732 (aarch64_maybe_expand_sve_subreg_move): Declare.
17733 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17734 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17736 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17737 that are semantically a reverse operation.
17738 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17739 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17740 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17742 (aarch64_can_change_mode_class): For big-endian, forbid changes
17743 between two SVE modes if they have different element sizes.
17745 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17747 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17748 the TImode handling for big-endian targets.
17750 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17752 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17753 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17755 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17756 Remove BSWAP handing for big-endian targets and use the form of
17757 LD1RQ appropariate for the mode.
17759 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17761 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17762 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17763 duplicated element.
17765 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17768 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17769 check for operands that need to go through aarch64_sve_reload_be.
17771 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17773 PR tree-optimization/81661
17774 PR tree-optimization/84117
17775 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17776 * tree-eh.c: Include gimplify.h.
17777 (find_trapping_overflow, replace_trapping_overflow,
17778 rewrite_to_non_trapping_overflow): New functions.
17779 * tree-vect-loop.c: Include tree-eh.h.
17780 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17781 * tree-data-ref.c: Include tree-eh.h.
17782 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17784 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17786 PR rtl-optimization/84123
17787 * combine.c (change_zero_ext): Check if hard register satisfies
17788 can_change_dest_mode before calling gen_lowpart_SUBREG.
17790 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17793 * ira.c (ira_init_register_move_cost): Remove assert.
17795 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17797 PR rtl-optimization/84071
17798 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17799 * doc/tm.texi: Regenerate.
17801 2018-01-31 Richard Biener <rguenther@suse.de>
17803 PR tree-optimization/84132
17804 * tree-data-ref.c (analyze_miv_subscript): Properly
17805 check whether evolution_function_is_affine_multivariate_p
17806 before calling gcd_of_steps_may_divide_p.
17808 2018-01-31 Julia Koval <julia.koval@intel.com>
17811 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17812 * config/i386/i386.md (rdpid_rex64) New.
17813 (rdpid): Make 32bit only.
17815 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17818 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17819 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17821 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17824 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17826 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17828 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17830 PR rtl-optimization/84071
17831 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17832 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17834 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17836 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17837 (arc_attribute_table): Add 'aux' attribute.
17838 (arc_in_small_data_p): Consider aux like variables.
17839 (arc_is_aux_reg_p): New function.
17840 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17841 (arc_get_aux_arg): New function.
17842 (prepare_move_operands): Handle aux-register access.
17843 (arc_handle_aux_attribute): New function.
17844 * doc/extend.texi (ARC Variable attributes): Add subsection.
17846 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17848 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17849 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17850 (arc_attribute_table): Add 'uncached' attribute.
17851 (arc_print_operand): Print '.di' flag for uncached memory
17853 (arc_in_small_data_p): Do not consider for small data the uncached
17855 (arc_is_uncached_mem_p): New function.
17856 * config/arc/predicates.md (compact_store_memory_operand): Check
17857 for uncached memory accesses.
17858 (nonvol_nonimm_operand): Likewise.
17859 * doc/extend.texi (ARC Type Attribute): New subsection.
17861 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17864 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17865 falign-loops=): Add Optimization flag.
17867 2018-01-30 Jeff Law <law@redhat.com>
17870 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17871 INT_REGISTERS_SAVED. Check it prior to calling
17872 get_scratch_register_on_entry.
17873 (ix86_adjust_stack_and_probe): Similarly.
17874 (ix86_emit_probe_stack_range): Similarly.
17875 (ix86_expand_prologue): Corresponding changes.
17877 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17880 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17881 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17883 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17886 * lra-constraints.c (curr_insn_transform): Process AND in the
17889 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17891 PR rtl-optimization/83986
17892 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17893 dependence against last_pending_memory_flush in addition to
17894 pending_jump_insns.
17896 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17898 PR tree-optimization/81611
17899 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17902 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17905 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17908 2018-01-30 Richard Biener <rguenther@suse.de>
17909 Jakub Jelinek <jakub@redhat.com>
17911 PR tree-optimization/84111
17912 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17913 inner loops added during recursion, as they don't have up-to-date
17916 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17919 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17920 (can_inline_edge_by_limits_p): ... here.
17921 (can_early_inline_edge_p, check_callers,
17922 update_caller_keys, update_callee_keys, recursive_inlining,
17923 add_new_edges_to_heap, speculation_useful_p,
17924 inline_small_functions,
17925 inline_small_functions, flatten_function,
17926 inline_to_all_callers_1): Update.
17928 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17930 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17933 2018-01-30 Richard Biener <rguenther@suse.de>
17935 PR tree-optimization/83008
17936 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17937 invariant and constant vector uses in stmts when they need
17938 more than one stmt.
17940 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17943 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17944 * configure: Regenerate.
17946 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17948 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17950 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17951 Use gen_rtx_REG rather than gen_lowpart.
17953 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17955 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17956 rather than 0 when creating partial subregs.
17958 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17960 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17963 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17966 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17967 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17968 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17969 flags. This restores the settings used before the 2017-07-24.
17970 Turning off pre increment/decrement/modify allows IVOPTS to
17971 optimize DF/SF loops where the index is an int.
17973 2018-01-29 Richard Biener <rguenther@suse.de>
17974 Kelvin Nilsen <kelvin@gcc.gnu.org>
17977 * tree-vect-stmts.c (vectorizable_call): Don't call
17978 targetm.vectorize_builtin_md_vectorized_function if callee is
17981 2018-01-22 Carl Love <cel@us.ibm.com>
17983 * doc/extend.tex: Fix typo in second arg in
17984 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17986 2018-01-29 Richard Biener <rguenther@suse.de>
17988 PR tree-optimization/84086
17989 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17990 (flush_ssaname_freelist): When SSA names were released reset
17991 the SCEV hash table.
17993 2018-01-29 Richard Biener <rguenther@suse.de>
17995 PR tree-optimization/84057
17996 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17997 removed paths when removing edges.
17999 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
18001 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
18002 -mfunction-return=@var{choice}.
18004 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18006 PR diagnostic/84034
18007 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
18008 Handle CR like TAB.
18009 (layout::print_source_line): Likewise.
18010 (test_get_line_width_without_trailing_whitespace): Add test cases.
18012 2018-01-27 Jakub Jelinek <jakub@redhat.com>
18014 PR middle-end/84040
18015 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
18018 2018-01-26 Jim Wilson <jimw@sifive.com>
18020 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
18022 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
18025 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18027 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
18028 and CMP + SUB-immediate -> SUBS.
18030 2018-01-26 Martin Sebor <msebor@redhat.com>
18032 PR tree-optimization/83896
18033 * tree-ssa-strlen.c (get_string_len): Rename...
18034 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
18035 Avoid assuming length is constant.
18036 (handle_char_store): Use HOST_WIDE_INT for string length.
18038 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
18041 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
18042 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
18044 2018-01-26 Richard Biener <rguenther@suse.de>
18046 PR rtl-optimization/84003
18047 * dse.c (record_store): Only record redundant stores when
18048 the earlier store aliases at least all accesses the later one does.
18050 2018-01-26 Jakub Jelinek <jakub@redhat.com>
18052 PR rtl-optimization/83985
18053 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
18054 REG_CFA_RESTORE insns.
18055 (delete_unmarked_insns): Don't ignore separate shrink wrapping
18056 REG_CFA_RESTORE insns here.
18059 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
18060 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
18062 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18064 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
18065 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
18066 (arc_init): Likewise.
18067 (arc_override_options): Likewise.
18068 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
18070 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
18072 * config/arc/arc.h (TARGET_DBNZ): Define.
18073 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
18074 properly set the tune attribute.
18075 (dbnz): Use TARGET_DBNZ guard.
18076 * config/arc/arc.opt (mtune): Add core3 option.
18078 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18080 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
18081 recognize new pic like addresses.
18082 (arc_delegitimize_address): Clean up.
18084 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18086 * config/arc/arc-arches.def: Option mrf16 valid for all
18088 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
18089 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
18090 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
18091 * config/arc/arc-tables.opt: Regenerate.
18092 * config/arc/arc.c (arc_conditional_register_usage): Handle
18093 reduced register file case.
18094 (arc_file_start): Set must have build attributes.
18095 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
18096 mrf16 option value.
18097 * config/arc/arc.opt (mrf16): Add new option.
18098 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
18099 * config/arc/genmultilib.awk: Handle new mrf16 option.
18100 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
18101 * config/arc/t-multilib: Regenerate.
18102 * doc/invoke.texi (ARC Options): Document mrf16 option.
18104 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18106 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
18107 * config/arc/arc.c (arc_handle_secure_attribute): New function.
18108 (arc_attribute_table): Add 'secure_call' attribute.
18109 (arc_print_operand): Print secure call operand.
18110 (arc_function_ok_for_sibcall): Don't optimize tail calls when
18112 (arc_is_secure_call_p): New function. * config/arc/arc.md
18113 (call_i): Add support for sjli instruction.
18114 (call_value_i): Likewise.
18115 * config/arc/constraints.md (Csc): New constraint.
18117 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18118 John Eric Martin <John.Martin@emmicro-us.com>
18120 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
18121 * config/arc/arc.c (_arc_jli_section): New struct.
18122 (arc_jli_section): New type.
18123 (rc_jli_sections): New static variable.
18124 (arc_handle_jli_attribute): New function.
18125 (arc_attribute_table): Add jli_always and jli_fixed attribute.
18126 (arc_file_end): New function.
18127 (TARGET_ASM_FILE_END): Define.
18128 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
18129 (arc_add_jli_section): New function.
18130 (jli_call_scan): Likewise.
18131 (arc_reorg): Call jli_call_scan.
18132 (arc_output_addsi): Remove 'S' from printing asm operand.
18133 (arc_is_jli_call_p): New function.
18134 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
18136 (movhi_insn): Likewise.
18137 (movsi_insn): Likewise.
18138 (movsi_set_cc_insn): Likewise.
18139 (loadqi_update): Likewise.
18140 (load_zeroextendqisi_update): Likewise.
18141 (load_signextendqisi_update): Likewise.
18142 (loadhi_update): Likewise.
18143 (load_zeroextendhisi_update): Likewise.
18144 (load_signextendhisi_update): Likewise.
18145 (loadsi_update): Likewise.
18146 (loadsf_update): Likewise.
18147 (movsicc_insn): Likewise.
18148 (bset_insn): Likewise.
18149 (bxor_insn): Likewise.
18150 (bclr_insn): Likewise.
18151 (bmsk_insn): Likewise.
18152 (bicsi3_insn): Likewise.
18153 (cmpsi_cc_c_insn): Likewise.
18154 (movsi_ne): Likewise.
18155 (movsi_cond_exec): Likewise.
18156 (clrsbsi2): Likewise.
18157 (norm_f): Likewise.
18167 (call_i): Remove 'S' asm letter, add jli instruction.
18168 (call_value_i): Likewise.
18169 * config/arc/arc.op (mjli-always): New option.
18170 * config/arc/constraints.md (Cji): New constraint.
18171 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
18173 (subsf3_fpx): Likewise.
18174 (mulsf3_fpx): Likewise.
18175 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
18177 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
18178 function attrbutes.
18179 * doc/invoke.texi (ARC): Document mjli-always option.
18181 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
18183 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
18184 avoid addition with 0 and use incw and decw where possible.
18186 2018-01-26 Richard Biener <rguenther@suse.de>
18188 PR tree-optimization/81082
18189 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
18190 association if it requires casting to unsigned.
18191 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
18192 from fold_plusminus_mult_expr to catch important cases late when
18193 range info is available.
18195 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18197 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
18198 * configure.ac (hidden_linkonce): New test.
18199 * configure: Regenerate.
18200 * config.in: Regenerate.
18202 2018-01-26 Julia Koval <julia.koval@intel.com>
18204 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
18205 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
18206 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
18207 _mm_mask_bitshuffle_epi64_mask): Fix type.
18208 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
18209 USI_FTYPE_V4DI_V4DI_USI): Remove.
18210 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
18211 __builtin_ia32_vpshufbitqmb256_mask,
18212 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
18213 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
18214 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
18216 2018-01-26 Alan Modra <amodra@gmail.com>
18219 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
18220 UNSPEC_VBPERMQ. Sort other unspecs.
18222 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
18224 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
18226 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
18228 PR middle-end/83055
18229 * predict.c (drop_profile): Do not push/pop cfun; update also
18231 (handle_missing_profiles): Fix logic looking for zero profiles.
18233 2018-01-25 Jakub Jelinek <jakub@redhat.com>
18235 PR middle-end/83977
18236 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
18237 on functions with #pragma omp declare simd or functions with simd
18239 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
18240 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
18241 Remove trailing \n from warning_at calls.
18243 2018-01-25 Tom de Vries <tom@codesourcery.com>
18246 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18247 for neutered workers.
18249 2018-01-24 Joseph Myers <joseph@codesourcery.com>
18252 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
18253 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
18255 2018-01-24 Jeff Law <law@redhat.com>
18258 * i386.c (get_probe_interval): Move to earlier point.
18259 (ix86_compute_frame_layout): If -fstack-clash-protection and
18260 the frame is larger than the probe interval, then use pushes
18261 to save registers rather than reg->mem moves.
18262 (ix86_expand_prologue): Remove conditional for int_registers_saved
18265 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18268 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18269 min/max for never referenced object.
18271 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18273 PR middle-end/83977
18274 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18276 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18277 attributes from DECL_ATTRIBUTES (decl) without affecting
18278 DECL_ATTRIBUTES (current_function_decl).
18279 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18280 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18282 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18284 PR tree-optimization/83979
18285 * fold-const.c (fold_comparison): Use constant_boolean_node
18286 instead of boolean_{true,false}_node.
18288 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18290 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18293 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18295 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18296 Simplify the clause that sets the length attribute.
18297 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18298 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18299 clause that sets the length attribute.
18300 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18302 2018-01-24 Tom de Vries <tom@codesourcery.com>
18305 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18306 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18307 Add strict parameter.
18308 (prevent_branch_around_nothing): Insert dummy insn between branch to
18309 label and label with no ptx insn inbetween.
18310 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18312 2018-01-24 Tom de Vries <tom@codesourcery.com>
18315 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18316 for neutered threads in warp.
18317 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18319 2018-01-24 Richard Biener <rguenther@suse.de>
18321 PR tree-optimization/83176
18322 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18325 2018-01-24 Richard Biener <rguenther@suse.de>
18327 PR tree-optimization/82819
18328 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18329 code generating pluses that are no-ops in the target precision.
18331 2018-01-24 Richard Biener <rguenther@suse.de>
18333 PR middle-end/84000
18334 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18336 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18338 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18339 to merge probabilities.
18340 * predict.c (probably_never_executed): Also mark as cold functions
18341 with global 0 profile and guessed local profile.
18342 * profile-count.c (profile_probability::combine_with_count): New
18344 * profile-count.h (profile_probability::operator*,
18345 profile_probability::operator*=, profile_probability::operator/,
18346 profile_probability::operator/=): Reduce precision to adjusted
18347 and set value to guessed on contradictory divisions.
18348 (profile_probability::combine_with_freq): Remove.
18349 (profile_probability::combine_wiht_count): Declare.
18350 (profile_count::force_nonzero):: Set to adjusted.
18351 (profile_count::probability_in):: Set quality to adjusted.
18352 * tree-ssa-tail-merge.c (replace_block_by): Use
18353 combine_with_count.
18355 2018-01-23 Andrew Waterman <andrew@sifive.com>
18356 Jim Wilson <jimw@sifive.com>
18358 * config/riscv/riscv.c (riscv_stack_boundary): New.
18359 (riscv_option_override): Set riscv_stack_boundary. Handle
18360 riscv_preferred_stack_boundary_arg.
18361 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18362 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18363 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18364 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18365 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18366 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18368 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18371 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18372 of struct ix86_frame.
18373 (ix86_expand_epilogue): Likewise. Add a local variable for
18374 the reg_save_offset field in struct ix86_frame.
18376 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18378 PR tree-optimization/82604
18379 * tree-loop-distribution.c (enum partition_kind): New enum item
18380 PKIND_PARTIAL_MEMSET.
18381 (partition_builtin_p): Support above new enum item.
18382 (generate_code_for_partition): Ditto.
18383 (compute_access_range): Differentiate cases that equality can be
18384 proven at all loops, the innermost loops or no loops.
18385 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18386 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18387 (finalize_partitions, distribute_loop): Don't fuse partition of
18388 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18389 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18390 parloop is enabled.
18392 2018-01-23 Martin Liska <mliska@suse.cz>
18394 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18395 order to ignore the predictor.
18396 (PRED_POLYMORPHIC_CALL): Likewise.
18397 (PRED_RECURSIVE_CALL): Likewise.
18399 2018-01-23 Martin Liska <mliska@suse.cz>
18401 * tree-profile.c (tree_profiling): Print function header to
18402 aware reader which function we are working on.
18403 * value-prof.c (gimple_find_values_to_profile): Do not print
18404 not interesting value histograms.
18406 2018-01-23 Martin Liska <mliska@suse.cz>
18408 * profile-count.h (enum profile_quality): Add
18409 profile_uninitialized as the first value. Do not number values
18410 as they are zero based.
18411 (profile_count::verify): Update sanity check.
18412 (profile_probability::verify): Likewise.
18414 2018-01-23 Nathan Sidwell <nathan@acm.org>
18416 * doc/invoke.texi (ffor-scope): Deprecate.
18418 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18420 PR tree-optimization/83510
18421 * domwalk.c (set_all_edges_as_executable): New function.
18422 (dom_walker::dom_walker): Convert bool param
18423 "skip_unreachable_blocks" to enum reachability. Move setup of
18424 edge flags to set_all_edges_as_executable and only do it when
18425 reachability is REACHABLE_BLOCKS.
18426 * domwalk.h (enum dom_walker::reachability): New enum.
18427 (dom_walker::dom_walker): Convert bool param
18428 "skip_unreachable_blocks" to enum reachability.
18429 (set_all_edges_as_executable): New decl.
18430 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18431 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18433 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18434 but converting true to REACHABLE_BLOCKS.
18435 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18437 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18438 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18439 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18441 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18442 if check_all_array_refs will be called.
18444 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18446 * tree.c (selftest::test_location_wrappers): Add more test
18449 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18451 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18452 (selftest::test_bit_in_range): Likewise.
18454 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18457 * doc/sourcebuild.texi (vect_float): Say that the selector
18458 only describes the situation when -funsafe-math-optimizations is on.
18459 (vect_float_strict): Document.
18461 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18463 PR tree-optimization/83965
18464 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18465 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18466 instead of checking only for a reduction.
18467 (vect_recog_widen_sum_pattern): Likewise.
18469 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18471 * predict.c (probably_never_executed): Only use precise profile info.
18472 (compute_function_frequency): Skip after inlining hack since we now
18473 have quality checking.
18475 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18477 * profile-count.h (profile_probability::very_unlikely,
18478 profile_probability::unlikely, profile_probability::even): Set
18479 precision to guessed.
18481 2018-01-23 Richard Biener <rguenther@suse.de>
18483 PR tree-optimization/83963
18484 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18485 Properly terminate dominator walk when crossing the exit edge not
18486 when visiting its source block.
18488 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18491 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18492 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18494 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18496 PR tree-optimization/83957
18497 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18498 semicolon after for body surrounded by braces.
18500 PR tree-optimization/83081
18501 * profile-count.h (profile_probability::split): New method.
18502 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18503 Use profile_probability::split.
18504 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18505 when splitting a single conditional jump into 2.
18507 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18509 PR tree-optimization/69452
18510 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18513 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18515 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18516 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18517 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18519 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18521 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18523 * config/rl78/rl78.md (movdi): New define_expand.
18524 * config/rl78/rl78.c (rl78_split_movdi): New function.
18526 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18529 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18531 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18532 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18533 128-bit to produce an UNSPEC move to get the double word with the
18534 signbit and then a shift directly to do signbit.
18535 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18536 implementation with a new version that just does either a direct
18537 move or a regular move. Move memory interface to separate insns.
18538 Move insns so they are next to the expander.
18539 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18540 with signbit move. Split big and little endian case.
18541 (signbit<mode>2_dm_mem_le): Likewise.
18542 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18543 (signbit<mode>2_dm2): Likewise.
18545 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18547 * config/rl78/rl78.md (anddi3): New define_expand.
18549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18551 * config/rl78/rl78.md (umindi3): New define_expand.
18553 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18555 * config/rl78/rl78.md (smindi3): New define_expand.
18557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18559 * config/rl78/rl78.md (smaxdi3): New define_expand.
18561 2018-01-22 Carl Love <cel@us.ibm.com>
18563 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18564 LVX_V1TI): Add macro expansion.
18565 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18566 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18567 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18568 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18569 Change check to determine if the instruction is a byte reversing
18570 entry. Fix typo in comment.
18571 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18572 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18573 Add def_builtin calls for new builtins.
18574 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18575 Add define_insn expansion.
18577 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18579 * config/rl78/rl78.md (umaxdi3): New define_expand.
18581 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18583 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18584 for non-QImode registers.
18586 2018-01-22 Richard Biener <rguenther@suse.de>
18588 PR tree-optimization/83963
18589 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18590 including the loop exit block.
18591 (scop_detection::merge_sese): Likewise.
18592 (scop_detection::add_scop): Do it here instead.
18594 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18596 * doc/sourcebuild.texi (arm_softfloat): Document.
18598 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18601 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18602 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18603 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18605 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18606 David Edelsohn <dje.gcc@gmail.com>
18609 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18610 Change "crset eq" to "crset 2".
18611 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18612 (*call_indirect_aix<mode>_nospec): Likewise.
18613 (*call_value_indirect_aix<mode>_nospec): Likewise.
18614 (*call_indirect_elfv2<mode>_nospec): Likewise.
18615 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18616 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18617 change assembly output from . to $.
18618 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18619 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18620 (*tablejump<mode>_internal1_nospec): Likewise.
18622 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18625 * config/sh/sh_optimize_sett_clrt.cc:
18626 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18628 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18630 PR tree-optimization/83940
18631 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18632 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18633 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18634 use it to pass back the definition type.
18635 (vect_check_store_rhs): Likewise rhs_dt_out.
18636 (vect_build_gather_load_calls): Add a mask_dt argument and use
18637 it instead of a call to vect_is_simple_use.
18638 (vectorizable_store): Update calls to vect_check_load_store_mask
18639 and vect_check_store_rhs. Use the dt returned by the latter instead
18640 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18641 instead of calls to vect_is_simple_use. Pass the scalar rather
18642 than the vector operand to vect_is_simple_use when handling
18643 second and subsequent copies of an rhs value.
18644 (vectorizable_load): Update calls to vect_check_load_store_mask
18645 and vect_build_gather_load_calls. Use the cached mask_dt and
18646 gs_info.offset_dt instead of calls to vect_is_simple_use.
18648 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18650 PR middle-end/83945
18651 * tree-emutls.c: Include gimplify.h.
18652 (lower_emutls_2): New function.
18653 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18654 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18655 it before further processing.
18658 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18659 UINTVAL (trueop1) instead of INTVAL (op1).
18661 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18665 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18666 INCOMING_FRAME_SP_OFFSET if not defined.
18667 (scan_trace): Add ENTRY argument. If true and
18668 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18669 emit a note to adjust the CFA offset.
18670 (create_cfi_notes): Adjust scan_trace callers.
18671 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18672 INCOMING_FRAME_SP_OFFSET in the CIE.
18673 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18674 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18676 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18677 * doc/tm.texi: Regenerated.
18679 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18681 PR rtl-optimization/83147
18682 * lra-constraints.c (remove_inheritance_pseudos): Use
18683 lra_substitute_pseudo_within_insn.
18685 2018-01-19 Tom de Vries <tom@codesourcery.com>
18686 Cesar Philippidis <cesar@codesourcery.com>
18689 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18691 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18694 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18695 spaces for function labels.
18697 2018-01-19 Martin Liska <mliska@suse.cz>
18699 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18700 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18701 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18702 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18703 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18704 (PRED_CONST_RETURN): Change from 69 to 65.
18705 (PRED_NULL_RETURN): Change from 91 to 71.
18706 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18707 (PRED_LOOP_GUARD): Change from 66 to 73.
18709 2018-01-19 Martin Liska <mliska@suse.cz>
18711 * predict.c (predict_insn_def): Add new assert.
18712 (struct branch_predictor): Change type to signed integer.
18713 (test_prediction_value_range): Amend test to cover
18714 PROB_UNINITIALIZED.
18715 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18716 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18717 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18718 (PRED_LOOP_IV_COMPARE): Likewise.
18719 * predict.h (PROB_UNINITIALIZED): Define new constant.
18721 2018-01-19 Martin Liska <mliska@suse.cz>
18723 * predict.c (dump_prediction): Add new format for
18724 analyze_brprob.py script which is enabled with -details
18726 * profile-count.h (precise_p): New function.
18728 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18730 PR tree-optimization/83922
18731 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18732 there are no statements that need masking.
18733 (vect_active_double_reduction_p): New function.
18734 (vect_analyze_loop_operations): Use it when handling phis that
18735 are not in the loop header.
18737 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18739 PR tree-optimization/83914
18740 * tree-vect-loop.c (vectorizable_induction): Don't convert
18741 init_expr or apply the peeling adjustment for inductions
18742 that are nested within the vectorized loop.
18744 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18746 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18749 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18753 * function.h (gimplify_parameters): Add gimple_seq * argument.
18754 * function.c: Include gimple.h and options.h.
18755 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18756 for the added local temporaries if needed.
18757 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18758 if there are any parameter cleanups, wrap whole body into a
18759 try/finally with the cleanups.
18761 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18764 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18765 Use GET_MODE_CLASS for scalar floating point.
18767 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18771 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18772 Fix call of call_cgraph_insertion_hooks.
18774 2018-01-18 Martin Sebor <msebor@redhat.com>
18776 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18778 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18781 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18784 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18787 * common.opt: (-ffile-prefix-map): New option.
18788 * opts.c (common_handle_option): Defer it.
18789 * opts-global.c (handle_common_deferred_options): Handle it.
18790 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18791 * file-prefix-map.h: New file.
18792 (remap_debug_filename, add_debug_prefix_map): ...here.
18793 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18794 * final.c (debug_prefix_map, add_debug_prefix_map
18795 remap_debug_filename): Move to...
18796 * file-prefix-map.c: New file.
18797 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18798 generalize, get rid of alloca(), use strrchr() instead of strchr().
18799 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18800 Implement in terms of add_prefix_map().
18801 (remap_macro_filename, remap_debug_filename): Implement in term of
18803 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18804 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18805 * dbxout.c: Include file-prefix-map.h.
18806 * varasm.c: Likewise.
18807 * vmsdbgout.c: Likewise.
18808 * xcoffout.c: Likewise.
18809 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18810 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18811 * doc/invoke.texi (-ffile-prefix-map): Document.
18812 (-fdebug-prefix-map): Update description.
18814 2018-01-18 Martin Liska <mliska@suse.cz>
18816 * config/i386/i386.c (indirect_thunk_name): Document that also
18818 (output_indirect_thunk): Document why both instructions
18819 (pause and lfence) are generated.
18821 2018-01-18 Richard Biener <rguenther@suse.de>
18823 PR tree-optimization/83887
18824 * graphite-scop-detection.c
18825 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18826 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18827 (scop_detection::merge_sese): Re-implement with a flood-fill
18828 algorithm that properly finds a SESE region if it exists.
18830 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18833 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18834 pointer_diff optimizations use view_convert instead of convert.
18836 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18838 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18839 Generate different code for -mno-speculate-indirect-jumps.
18840 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18841 (*call_indirect_aix<mode>): Disable for
18842 -mno-speculate-indirect-jumps.
18843 (*call_indirect_aix<mode>_nospec): New define_insn.
18844 (*call_value_indirect_aix<mode>): Disable for
18845 -mno-speculate-indirect-jumps.
18846 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18847 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18848 -mno-speculate-indirect-jumps.
18849 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18851 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18853 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18854 long double type, set the flags for noting the default long double
18855 type, even if we don't pass or return a long double type.
18857 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18860 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18863 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18865 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18866 support for merge[hl].
18867 (fold_mergehl_helper): New helper function.
18868 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18869 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18870 (altivec_vmrglw_direct): Add xxmrglw insn.
18872 2018-01-17 Andrew Waterman <andrew@sifive.com>
18874 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18875 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18877 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18880 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18881 call the lto_location_cache before reading the
18882 DECL_SOURCE_LOCATION of the types.
18884 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18885 Richard Sandiford <richard.sandiford@linaro.org>
18887 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18888 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18889 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18890 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18891 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18893 * config/aarch64/constraints.md (aarch64_movti_operand):
18895 * config/aarch64/predicates.md (Uti): Add new constraint.
18897 2018-01-17 Carl Love <cel@us.ibm.com>
18899 * config/rs6000/vsx.md (define_expand xl_len_r,
18900 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18901 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18903 (define_expand, define_insn): Move the shift left from the
18904 define_insn to the define_expand for lxvl and stxvl instructions.
18905 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18906 and XL_LEN_R definitions to PURE.
18908 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18910 * config/i386/i386.c (indirect_thunk_name): Declare regno
18911 as unsigned int. Compare regno with INVALID_REGNUM.
18912 (output_indirect_thunk): Ditto.
18913 (output_indirect_thunk_function): Ditto.
18914 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18915 in the call to output_indirect_thunk_function.
18917 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18919 PR middle-end/83884
18920 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18921 rather than the size of inner_type to determine the stack slot size
18922 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18924 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18927 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18930 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18932 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18933 endian Linux systems to optionally enable multilibs for selecting
18934 the long double type if the user configured an explicit type.
18935 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18936 have no long double multilibs if not defined.
18937 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18938 warn if the user used -mabi={ieee,ibm}longdouble and we built
18939 multilibs for long double.
18940 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18941 appropriate multilib option.
18942 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18944 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18945 for building long double multilibs.
18946 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18948 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18950 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18953 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18955 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18958 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18961 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18964 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18966 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18967 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18969 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18971 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18972 different rtl trees depending on TARGET_64BIT.
18973 (rs6000_gen_lvx): Likewise.
18975 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18977 * config/visium/visium.md (nop): Tweak comment.
18978 (hazard_nop): Likewise.
18980 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18982 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18983 -mspeculate-indirect-jumps.
18984 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18985 for -mno-speculate-indirect-jumps.
18986 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18987 (*call_value_indirect_elfv2<mode>): Disable for
18988 -mno-speculate-indirect-jumps.
18989 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18990 (indirect_jump): Emit different RTL for
18991 -mno-speculate-indirect-jumps.
18992 (*indirect_jump<mode>): Disable for
18993 -mno-speculate-indirect-jumps.
18994 (*indirect_jump<mode>_nospec): New define_insn.
18995 (tablejump): Emit different RTL for
18996 -mno-speculate-indirect-jumps.
18997 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18998 (tablejumpsi_nospec): New define_expand.
18999 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
19000 (tablejumpdi_nospec): New define_expand.
19001 (*tablejump<mode>_internal1): Disable for
19002 -mno-speculate-indirect-jumps.
19003 (*tablejump<mode>_internal1_nospec): New define_insn.
19004 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
19007 2018-01-16 Artyom Skrobov tyomitch@gmail.com
19009 * caller-save.c (insert_save): Drop unnecessary parameter. All
19012 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19013 Richard Biener <rguenth@suse.de>
19016 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
19017 return early, inline manually is_gimple_sizepos. Make sure if we
19018 call gimplify_expr we don't end up with a gimple constant.
19019 * tree.c (variably_modified_type_p): Don't return true for
19020 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
19021 * gimplify.h (is_gimple_sizepos): Remove.
19023 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19025 PR tree-optimization/83857
19026 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
19027 vectorizable_live_operation for pure SLP statements.
19028 (vectorizable_live_operation): Handle PHIs.
19030 2018-01-16 Richard Biener <rguenther@suse.de>
19032 PR tree-optimization/83867
19033 * tree-vect-stmts.c (vect_transform_stmt): Precompute
19034 nested_in_vect_loop_p since the scalar stmt may get invalidated.
19036 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19039 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
19040 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
19041 If off is not INTEGER_CST, issue a may not be aligned warning
19042 rather than isn't aligned. Use isn%'t rather than isn't.
19043 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
19045 <case MULT_EXPR>: Improve the case when bottom and one of the
19046 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
19047 operand, in that case check if the other operand is multiple of
19048 bottom divided by the INTEGER_CST operand.
19050 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19053 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
19054 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
19055 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
19056 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
19057 * config/pa/pa.c (pa_function_arg_advance): Likewise.
19058 (pa_function_arg, pa_arg_partial_bytes): Likewise.
19059 (pa_function_arg_size): New function.
19061 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19063 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
19064 in a separate statement.
19066 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19068 PR tree-optimization/83847
19069 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
19070 group gathers and scatters.
19072 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19074 PR rtl-optimization/86620
19075 * params.def (max-sched-ready-insns): Bump minimum value to 1.
19077 PR rtl-optimization/83213
19078 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
19079 to last if both are JUMP_INSNs.
19081 PR tree-optimization/83843
19082 * gimple-ssa-store-merging.c
19083 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
19084 store_immediate_info for bswap/nop orig_stores.
19086 2018-01-15 Andrew Waterman <andrew@sifive.com>
19088 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
19090 <UDIV>: Increase cost if !TARGET_DIV.
19092 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
19094 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
19095 (define_attr "cr_logical_3op"): New.
19096 (cceq_ior_compare): Adjust.
19097 (cceq_ior_compare_complement): Adjust.
19098 (*cceq_rev_compare): Adjust.
19099 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
19100 (is_cracked_insn): Adjust.
19101 (insn_must_be_first_in_group): Adjust.
19102 * config/rs6000/40x.md: Adjust.
19103 * config/rs6000/440.md: Adjust.
19104 * config/rs6000/476.md: Adjust.
19105 * config/rs6000/601.md: Adjust.
19106 * config/rs6000/603.md: Adjust.
19107 * config/rs6000/6xx.md: Adjust.
19108 * config/rs6000/7450.md: Adjust.
19109 * config/rs6000/7xx.md: Adjust.
19110 * config/rs6000/8540.md: Adjust.
19111 * config/rs6000/cell.md: Adjust.
19112 * config/rs6000/e300c2c3.md: Adjust.
19113 * config/rs6000/e500mc.md: Adjust.
19114 * config/rs6000/e500mc64.md: Adjust.
19115 * config/rs6000/e5500.md: Adjust.
19116 * config/rs6000/e6500.md: Adjust.
19117 * config/rs6000/mpc.md: Adjust.
19118 * config/rs6000/power4.md: Adjust.
19119 * config/rs6000/power5.md: Adjust.
19120 * config/rs6000/power6.md: Adjust.
19121 * config/rs6000/power7.md: Adjust.
19122 * config/rs6000/power8.md: Adjust.
19123 * config/rs6000/power9.md: Adjust.
19124 * config/rs6000/rs64.md: Adjust.
19125 * config/rs6000/titan.md: Adjust.
19127 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19129 * config/i386/predicates.md (indirect_branch_operand): Rewrite
19130 ix86_indirect_branch_register logic.
19132 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19134 * config/i386/constraints.md (Bs): Update
19135 ix86_indirect_branch_register check. Don't check
19136 ix86_indirect_branch_register with GOT_memory_operand.
19138 * config/i386/predicates.md (GOT_memory_operand): Don't check
19139 ix86_indirect_branch_register here.
19140 (GOT32_symbol_operand): Likewise.
19142 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19144 * config/i386/predicates.md (constant_call_address_operand):
19145 Rewrite ix86_indirect_branch_register logic.
19146 (sibcall_insn_operand): Likewise.
19148 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19150 * config/i386/constraints.md (Bs): Replace
19151 ix86_indirect_branch_thunk_register with
19152 ix86_indirect_branch_register.
19154 * config/i386/i386.md (indirect_jump): Likewise.
19155 (tablejump): Likewise.
19156 (*sibcall_memory): Likewise.
19157 (*sibcall_value_memory): Likewise.
19158 Peepholes of indirect call and jump via memory: Likewise.
19159 * config/i386/i386.opt: Likewise.
19160 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19161 (GOT_memory_operand): Likewise.
19162 (call_insn_operand): Likewise.
19163 (sibcall_insn_operand): Likewise.
19164 (GOT32_symbol_operand): Likewise.
19166 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19168 PR middle-end/83837
19169 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
19170 type rather than type addr's type points to.
19171 (expand_omp_atomic_mutex): Likewise.
19172 (expand_omp_atomic): Likewise.
19174 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19177 * config/i386/i386.c (output_indirect_thunk_function): Use
19178 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
19179 for __x86_return_thunk.
19181 2018-01-15 Richard Biener <rguenther@suse.de>
19183 PR middle-end/83850
19184 * expmed.c (extract_bit_field_1): Fix typo.
19186 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19189 * config/arm/iterators.md (VF): New mode iterator.
19190 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
19191 Remove integer-related logic from pattern.
19192 (neon_vabd<mode>_3): Likewise.
19194 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19196 PR middle-end/82694
19197 * common.opt (fstrict-overflow): No longer an alias.
19198 (fwrapv-pointer): New option.
19199 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
19200 also for pointer types based on flag_wrapv_pointer.
19201 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
19202 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
19203 opts->x_flag_wrapv got set.
19204 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
19205 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
19206 POINTER_TYPE_OVERFLOW_UNDEFINED.
19207 * match.pd: Likewise in address comparison pattern.
19208 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
19210 2018-01-15 Richard Biener <rguenther@suse.de>
19213 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
19214 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
19215 Reset type names to their identifier if their TYPE_DECL doesn't
19216 have linkage (and thus is used for ODR and devirt).
19217 (save_debug_info_for_decl): Remove.
19218 (save_debug_info_for_type): Likewise.
19219 (add_tree_to_fld_list): Adjust.
19220 * tree-pretty-print.c (dump_generic_node): Make dumping of
19221 type names more robust.
19223 2018-01-15 Richard Biener <rguenther@suse.de>
19225 * BASE-VER: Bump to 8.0.1.
19227 2018-01-14 Martin Sebor <msebor@redhat.com>
19230 * builtins.c (check_access): Avoid warning when the no-warning bit
19233 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
19235 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
19236 * ira-color (allocno_hard_regs_compare): Likewise.
19238 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
19241 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
19242 Use .pushsection/.popsection.
19244 2018-01-14 Martin Sebor <msebor@redhat.com>
19247 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
19249 2018-01-14 Jakub Jelinek <jakub@redhat.com>
19251 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
19252 entry from extra_headers.
19253 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
19254 extra_headers, make the list bitwise identical to the i?86-*-* one.
19256 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19258 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
19259 -mcmodel=large with -mindirect-branch=thunk,
19260 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19261 -mfunction-return=thunk-extern.
19262 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19263 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19264 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19266 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19268 * config/i386/i386.c (print_reg): Print the name of the full
19269 integer register without '%'.
19270 (ix86_print_operand): Handle 'V'.
19271 * doc/extend.texi: Document 'V' modifier.
19273 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19275 * config/i386/constraints.md (Bs): Disallow memory operand for
19276 -mindirect-branch-register.
19278 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19279 (GOT_memory_operand): Likewise.
19280 (call_insn_operand): Likewise.
19281 (sibcall_insn_operand): Likewise.
19282 (GOT32_symbol_operand): Likewise.
19283 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19284 for -mindirect-branch-register.
19285 (tablejump): Likewise.
19286 (*sibcall_memory): Likewise.
19287 (*sibcall_value_memory): Likewise.
19288 Disallow peepholes of indirect call and jump via memory for
19289 -mindirect-branch-register.
19290 (*call_pop): Replace m with Bw.
19291 (*call_value_pop): Likewise.
19292 (*sibcall_pop_memory): Replace m with Bs.
19293 * config/i386/i386.opt (mindirect-branch-register): New option.
19294 * doc/invoke.texi: Document -mindirect-branch-register option.
19296 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19298 * config/i386/i386-protos.h (ix86_output_function_return): New.
19299 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19300 set function_return_type.
19301 (indirect_thunk_name): Add ret_p to indicate thunk for function
19303 (output_indirect_thunk_function): Pass false to
19304 indirect_thunk_name.
19305 (ix86_output_indirect_branch_via_reg): Likewise.
19306 (ix86_output_indirect_branch_via_push): Likewise.
19307 (output_indirect_thunk_function): Create alias for function
19308 return thunk if regno < 0.
19309 (ix86_output_function_return): New function.
19310 (ix86_handle_fndecl_attribute): Handle function_return.
19311 (ix86_attribute_table): Add function_return.
19312 * config/i386/i386.h (machine_function): Add
19313 function_return_type.
19314 * config/i386/i386.md (simple_return_internal): Use
19315 ix86_output_function_return.
19316 (simple_return_internal_long): Likewise.
19317 * config/i386/i386.opt (mfunction-return=): New option.
19318 (indirect_branch): Mention -mfunction-return=.
19319 * doc/extend.texi: Document function_return function attribute.
19320 * doc/invoke.texi: Document -mfunction-return= option.
19322 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19324 * config/i386/i386-opts.h (indirect_branch): New.
19325 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19326 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19327 with local indirect jump when converting indirect call and jump.
19328 (ix86_set_indirect_branch_type): New.
19329 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19330 (indirectlabelno): New.
19331 (indirect_thunk_needed): Likewise.
19332 (indirect_thunk_bnd_needed): Likewise.
19333 (indirect_thunks_used): Likewise.
19334 (indirect_thunks_bnd_used): Likewise.
19335 (INDIRECT_LABEL): Likewise.
19336 (indirect_thunk_name): Likewise.
19337 (output_indirect_thunk): Likewise.
19338 (output_indirect_thunk_function): Likewise.
19339 (ix86_output_indirect_branch_via_reg): Likewise.
19340 (ix86_output_indirect_branch_via_push): Likewise.
19341 (ix86_output_indirect_branch): Likewise.
19342 (ix86_output_indirect_jmp): Likewise.
19343 (ix86_code_end): Call output_indirect_thunk_function if needed.
19344 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19346 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19347 (ix86_attribute_table): Add indirect_branch.
19348 * config/i386/i386.h (machine_function): Add indirect_branch_type
19349 and has_local_indirect_jump.
19350 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19352 (tablejump): Likewise.
19353 (*indirect_jump): Use ix86_output_indirect_jmp.
19354 (*tablejump_1): Likewise.
19355 (simple_return_indirect_internal): Likewise.
19356 * config/i386/i386.opt (mindirect-branch=): New option.
19357 (indirect_branch): New.
19360 (thunk-inline): Likewise.
19361 (thunk-extern): Likewise.
19362 * doc/extend.texi: Document indirect_branch function attribute.
19363 * doc/invoke.texi: Document -mindirect-branch= option.
19365 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19368 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19370 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19372 * ipa-inline.c (want_inline_small_function_p): Return false if
19373 inlining has already failed with CIF_FINAL_ERROR.
19374 (update_caller_keys): Call want_inline_small_function_p before
19376 (update_callee_keys): Likewise.
19378 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19380 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19382 (rs6000_quadword_masked_address_p): Likewise.
19383 (quad_aligned_load_p): Likewise.
19384 (quad_aligned_store_p): Likewise.
19385 (const_load_sequence_p): Add comment to describe the outer-most loop.
19386 (mimic_memory_attributes_and_flags): New function.
19387 (rs6000_gen_stvx): Likewise.
19388 (replace_swapped_aligned_store): Likewise.
19389 (rs6000_gen_lvx): Likewise.
19390 (replace_swapped_aligned_load): Likewise.
19391 (replace_swapped_load_constant): Capitalize argument name in
19392 comment describing this function.
19393 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19394 and stores that access quad-word aligned addresses and replace
19395 with stvx or lvx instructions when appropriate.
19396 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19397 New function prototype.
19398 (rs6000_quadword_masked_address_p): Likewise.
19399 (rs6000_gen_lvx): Likewise.
19400 (rs6000_gen_stvx): Likewise.
19401 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19402 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19403 when memory address is aligned.
19404 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19405 this split to select lvx instruction when memory address is aligned.
19406 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19407 instruction when memory address is aligned.
19408 (*vsx_le_perm_load_v16qi): Likewise.
19409 (four unnamed splitters): Modify to select the stvx instruction
19410 when memory is aligned.
19412 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19414 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19415 which appears in the queue multiple times.
19417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19418 Alan Hayward <alan.hayward@arm.com>
19419 David Sherwood <david.sherwood@arm.com>
19421 * tree-vectorizer.h (vec_lower_bound): New structure.
19422 (_loop_vec_info): Add check_nonzero and lower_bounds.
19423 (LOOP_VINFO_CHECK_NONZERO): New macro.
19424 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19425 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19426 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19427 fields. Make seg_len the distance travelled, not including the
19429 (dr_direction_indicator): Declare.
19430 (dr_zero_step_indicator): Likewise.
19431 (dr_known_forward_stride_p): Likewise.
19432 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19434 (runtime_alias_check_p): Allow runtime alias checks with
19436 (operator ==): Compare access_size and align.
19437 (prune_runtime_alias_test_list): Rework for new distinction between
19438 the access_size and seg_len.
19439 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19441 (get_segment_min_max): New function.
19442 (create_intersect_range_checks): Use it.
19443 (dr_step_indicator): New function.
19444 (dr_direction_indicator): Likewise.
19445 (dr_zero_step_indicator): Likewise.
19446 (dr_known_forward_stride_p): Likewise.
19447 * tree-loop-distribution.c (data_ref_segment_size): Return
19448 DR_STEP * (niters - 1).
19449 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19451 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19452 (vect_preserves_scalar_order_p): New function, split out from...
19453 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19454 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19455 (vect_vfa_access_size): New function.
19456 (vect_vfa_align): Likewise.
19457 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19458 (dump_lower_bound): New function.
19459 (vect_check_lower_bound): Likewise.
19460 (vect_small_gap_p): Likewise.
19461 (vectorizable_with_step_bound_p): Likewise.
19462 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19463 depencies if the vectorization factor is 1. Convert the checks
19464 for nonzero steps into checks on the bounds of DR_STEP. Try using
19465 a bunds check for variable steps if the minimum required step is
19466 relatively small. Update calls to the dr_with_seg_len
19467 constructor and to vect_compile_time_alias.
19468 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19470 (vect_loop_versioning): Call it.
19471 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19473 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19476 Alan Hayward <alan.hayward@arm.com>
19477 David Sherwood <david.sherwood@arm.com>
19479 * doc/sourcebuild.texi (vect_scatter_store): Document.
19480 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19482 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19484 * genopinit.c (main): Add supports_vec_scatter_store and
19485 supports_vec_scatter_store_cached to target_optabs.
19486 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19487 IFN_MASK_SCATTER_STORE.
19488 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19490 * internal-fn.h (internal_store_fn_p): Declare.
19491 (internal_fn_stored_value_index): Likewise.
19492 * internal-fn.c (scatter_store_direct): New macro.
19493 (expand_scatter_store_optab_fn): New function.
19494 (direct_scatter_store_optab_supported_p): New macro.
19495 (internal_store_fn_p): New function.
19496 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19497 IFN_MASK_SCATTER_STORE.
19498 (internal_fn_mask_index): Likewise.
19499 (internal_fn_stored_value_index): New function.
19500 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19501 for scatter stores.
19502 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19503 * optabs-query.c (supports_vec_scatter_store_p): New function.
19504 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19505 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19506 true for scatter stores.
19507 (vect_gather_scatter_fn_p): Handle scatter stores too.
19508 (vect_check_gather_scatter): Consider using scatter stores if
19509 supports_vec_scatter_store_p.
19510 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19511 scatter stores too.
19512 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19513 internal_fn_stored_value_index.
19514 (check_load_store_masking): Handle scatter stores too.
19515 (vect_get_store_rhs): Make public.
19516 (vectorizable_call): Use internal_store_fn_p.
19517 (vectorizable_store): Handle scatter store internal functions.
19518 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19519 when deciding whether the end of the group has been reached.
19520 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19521 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19522 (mask_scatter_store<mode>): New insns.
19524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19525 Alan Hayward <alan.hayward@arm.com>
19526 David Sherwood <david.sherwood@arm.com>
19528 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19529 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19530 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19532 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19533 Use vect_truncate_gather_scatter_offset if we can't treat the
19534 operation as a normal gather load or scatter store.
19535 (get_group_load_store_type): Take the gather_scatter_info
19536 as argument. Try using a gather load or scatter store for
19537 single-element groups.
19538 (get_load_store_type): Update calls to get_group_load_store_type
19539 and vect_use_strided_gather_scatters_p.
19541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19542 Alan Hayward <alan.hayward@arm.com>
19543 David Sherwood <david.sherwood@arm.com>
19545 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19546 optional tree argument.
19547 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19549 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19550 but continue to use the current value as a fallback.
19551 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19552 to compare the updates.
19553 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19554 (get_load_store_type): Use it when handling a strided access.
19555 (vect_get_strided_load_store_ops): New function.
19556 (vect_get_data_ptr_increment): Likewise.
19557 (vectorizable_load): Handle strided gather loads. Always pass
19558 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19561 Alan Hayward <alan.hayward@arm.com>
19562 David Sherwood <david.sherwood@arm.com>
19564 * doc/md.texi (gather_load@var{m}): Document.
19565 (mask_gather_load@var{m}): Likewise.
19566 * genopinit.c (main): Add supports_vec_gather_load and
19567 supports_vec_gather_load_cached to target_optabs.
19568 * optabs-tree.c (init_tree_optimization_optabs): Use
19569 ggc_cleared_alloc to allocate target_optabs.
19570 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19571 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19573 * internal-fn.h (internal_load_fn_p): Declare.
19574 (internal_gather_scatter_fn_p): Likewise.
19575 (internal_fn_mask_index): Likewise.
19576 (internal_gather_scatter_fn_supported_p): Likewise.
19577 * internal-fn.c (gather_load_direct): New macro.
19578 (expand_gather_load_optab_fn): New function.
19579 (direct_gather_load_optab_supported_p): New macro.
19580 (direct_internal_fn_optab): New function.
19581 (internal_load_fn_p): Likewise.
19582 (internal_gather_scatter_fn_p): Likewise.
19583 (internal_fn_mask_index): Likewise.
19584 (internal_gather_scatter_fn_supported_p): Likewise.
19585 * optabs-query.c (supports_at_least_one_mode_p): New function.
19586 (supports_vec_gather_load_p): Likewise.
19587 * optabs-query.h (supports_vec_gather_load_p): Declare.
19588 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19589 and memory_type field.
19590 (NUM_PATTERNS): Bump to 15.
19591 * tree-vect-data-refs.c: Include internal-fn.h.
19592 (vect_gather_scatter_fn_p): New function.
19593 (vect_describe_gather_scatter_call): Likewise.
19594 (vect_check_gather_scatter): Try using internal functions for
19595 gather loads. Recognize existing calls to a gather load function.
19596 (vect_analyze_data_refs): Consider using gather loads if
19597 supports_vec_gather_load_p.
19598 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19599 (vect_get_gather_scatter_offset_type): Likewise.
19600 (vect_convert_mask_for_vectype): Likewise.
19601 (vect_add_conversion_to_patterm): Likewise.
19602 (vect_try_gather_scatter_pattern): Likewise.
19603 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19604 (vect_vect_recog_func_ptrs): Add it.
19605 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19606 internal_fn_mask_index and internal_gather_scatter_fn_p.
19607 (check_load_store_masking): Take the gather_scatter_info as an
19608 argument and handle gather loads.
19609 (vect_get_gather_scatter_ops): New function.
19610 (vectorizable_call): Check internal_load_fn_p.
19611 (vectorizable_load): Likewise. Handle gather load internal
19613 (vectorizable_store): Update call to check_load_store_masking.
19614 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19615 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19616 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19617 (aarch64_gather_scale_operand_d): New predicates.
19618 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19619 (mask_gather_load<mode>): New insns.
19621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19622 Alan Hayward <alan.hayward@arm.com>
19623 David Sherwood <david.sherwood@arm.com>
19625 * optabs.def (fold_left_plus_optab): New optab.
19626 * doc/md.texi (fold_left_plus_@var{m}): Document.
19627 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19628 * internal-fn.c (fold_left_direct): Define.
19629 (expand_fold_left_optab_fn): Likewise.
19630 (direct_fold_left_optab_supported_p): Likewise.
19631 * fold-const-call.c (fold_const_fold_left): New function.
19632 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19633 * tree-parloops.c (valid_reduction_p): New function.
19634 (gather_scalar_reductions): Use it.
19635 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19636 (vect_finish_replace_stmt): Declare.
19637 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19638 (needs_fold_left_reduction_p): New function, split out from...
19639 (vect_is_simple_reduction): ...here. Accept reductions that
19640 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19641 (vect_force_simple_reduction): Also store the reduction type in
19642 the assignment's STMT_VINFO_REDUC_TYPE.
19643 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19644 (merge_with_identity): New function.
19645 (vect_expand_fold_left): Likewise.
19646 (vectorize_fold_left_reduction): Likewise.
19647 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19648 scalar phi in place for it. Check for target support and reject
19649 cases that would reassociate the operation. Defer the transform
19650 phase to vectorize_fold_left_reduction.
19651 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19652 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19653 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19657 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19658 call to ifc_temp_var.
19660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19661 Alan Hayward <alan.hayward@arm.com>
19662 David Sherwood <david.sherwood@arm.com>
19664 * target.def (legitimize_address_displacement): Take the original
19665 offset as a poly_int.
19666 * targhooks.h (default_legitimize_address_displacement): Update
19668 * targhooks.c (default_legitimize_address_displacement): Likewise.
19669 * doc/tm.texi: Regenerate.
19670 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19671 as an argument, moving assert of ad->disp == ad->disp_term to...
19672 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19673 Try calling targetm.legitimize_address_displacement before expanding
19674 the address rather than afterwards, and adjust for the new interface.
19675 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19676 Match the new hook interface. Handle SVE addresses.
19677 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19678 new hook interface.
19680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19682 * Makefile.in (OBJS): Add early-remat.o.
19683 * target.def (select_early_remat_modes): New hook.
19684 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19685 * doc/tm.texi: Regenerate.
19686 * targhooks.h (default_select_early_remat_modes): Declare.
19687 * targhooks.c (default_select_early_remat_modes): New function.
19688 * timevar.def (TV_EARLY_REMAT): New timevar.
19689 * passes.def (pass_early_remat): New pass.
19690 * tree-pass.h (make_pass_early_remat): Declare.
19691 * early-remat.c: New file.
19692 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19694 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19697 Alan Hayward <alan.hayward@arm.com>
19698 David Sherwood <david.sherwood@arm.com>
19700 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19701 vfm1 with a bound_epilog parameter.
19702 (vect_do_peeling): Update calls accordingly, and move the prologue
19703 call earlier in the function. Treat the base bound_epilog as 0 for
19704 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19705 this base when peeling for gaps.
19706 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19707 with fully-masked loops.
19708 (vect_estimate_min_profitable_iters): Handle the single peeled
19709 iteration in that case.
19711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19712 Alan Hayward <alan.hayward@arm.com>
19713 David Sherwood <david.sherwood@arm.com>
19715 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19716 single-element interleaving even if the size is not a power of 2.
19717 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19718 accesses for single-element interleaving if the group size is
19721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19722 Alan Hayward <alan.hayward@arm.com>
19723 David Sherwood <david.sherwood@arm.com>
19725 * doc/md.texi (fold_extract_last_@var{m}): Document.
19726 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19727 * optabs.def (fold_extract_last_optab): New optab.
19728 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19729 * internal-fn.c (fold_extract_direct): New macro.
19730 (expand_fold_extract_optab_fn): Likewise.
19731 (direct_fold_extract_optab_supported_p): Likewise.
19732 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19733 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19734 EXTRACT_LAST_REDUCTION.
19735 (get_initial_def_for_reduction): Do not create an initial vector
19736 for EXTRACT_LAST_REDUCTION reductions.
19737 (vectorizable_reduction): Leave the scalar phi in place for
19738 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19739 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19740 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19741 transform phase to vectorizable_condition.
19742 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19744 (vect_finish_stmt_generation): ...here.
19745 (vect_finish_replace_stmt): New function.
19746 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19747 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19749 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19752 Alan Hayward <alan.hayward@arm.com>
19753 David Sherwood <david.sherwood@arm.com>
19755 * doc/md.texi (extract_last_@var{m}): Document.
19756 * optabs.def (extract_last_optab): New optab.
19757 * internal-fn.def (EXTRACT_LAST): New internal function.
19758 * internal-fn.c (cond_unary_direct): New macro.
19759 (expand_cond_unary_optab_fn): Likewise.
19760 (direct_cond_unary_optab_supported_p): Likewise.
19761 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19762 loops using EXTRACT_LAST.
19763 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19764 (extract_last_<mode>): ...this optab.
19765 (vec_extract<mode><Vel>): Update accordingly.
19767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19768 Alan Hayward <alan.hayward@arm.com>
19769 David Sherwood <david.sherwood@arm.com>
19771 * target.def (empty_mask_is_expensive): New hook.
19772 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19773 * doc/tm.texi: Regenerate.
19774 * targhooks.h (default_empty_mask_is_expensive): Declare.
19775 * targhooks.c (default_empty_mask_is_expensive): New function.
19776 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19777 if the target says that empty masks are expensive.
19778 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19780 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19783 Alan Hayward <alan.hayward@arm.com>
19784 David Sherwood <david.sherwood@arm.com>
19786 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19787 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19788 (vect_use_loop_mask_for_alignment_p): New function.
19789 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19790 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19791 niters_skip argument. Make sure that the first niters_skip elements
19792 of the first iteration are inactive.
19793 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19794 Update call to vect_set_loop_masks_directly.
19795 (get_misalign_in_elems): New function, split out from...
19796 (vect_gen_prolog_loop_niters): ...here.
19797 (vect_update_init_of_dr): Take a code argument that specifies whether
19798 the adjustment should be added or subtracted.
19799 (vect_update_init_of_drs): Likewise.
19800 (vect_prepare_for_masked_peels): New function.
19801 (vect_do_peeling): Skip prologue peeling if we're using a mask
19802 instead. Update call to vect_update_inits_of_drs.
19803 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19805 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19806 alignment. Do not include the number of peeled iterations in
19807 the minimum threshold in that case.
19808 (vectorizable_induction): Adjust the start value down by
19809 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19810 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19811 Take the number of skipped iterations into account when calculating
19813 * tree-vect-stmts.c (vect_gen_while_not): New function.
19815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19816 Alan Hayward <alan.hayward@arm.com>
19817 David Sherwood <david.sherwood@arm.com>
19819 * doc/sourcebuild.texi (vect_fully_masked): Document.
19820 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19821 default value to 0.
19822 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19824 (vect_analyze_loop_2): ...here. Don't check the vectorization
19825 factor against the number of loop iterations if the loop is
19828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19829 Alan Hayward <alan.hayward@arm.com>
19830 David Sherwood <david.sherwood@arm.com>
19832 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19833 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19834 (dump_groups): Update accordingly.
19835 (iv_use::mem_type): New member variable.
19836 (address_p): New function.
19837 (record_use): Add a mem_type argument and initialize the new
19839 (record_group_use): Add a mem_type argument. Use address_p.
19840 Remove obsolete null checks of base_object. Update call to record_use.
19841 (find_interesting_uses_op): Update call to record_group_use.
19842 (find_interesting_uses_cond): Likewise.
19843 (find_interesting_uses_address): Likewise.
19844 (get_mem_type_for_internal_fn): New function.
19845 (find_address_like_use): Likewise.
19846 (find_interesting_uses_stmt): Try find_address_like_use before
19847 calling find_interesting_uses_op.
19848 (addr_offset_valid_p): Use the iv mem_type field as the type
19849 of the addressed memory.
19850 (add_autoinc_candidates): Likewise.
19851 (get_address_cost): Likewise.
19852 (split_small_address_groups_p): Use address_p.
19853 (split_address_groups): Likewise.
19854 (add_iv_candidate_for_use): Likewise.
19855 (autoinc_possible_for_pair): Likewise.
19856 (rewrite_groups): Likewise.
19857 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19858 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19859 (get_alias_ptr_type_for_ptr_address): New function.
19860 (rewrite_use_address): Rewrite address uses in calls that were
19861 identified by find_address_like_use.
19863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19864 Alan Hayward <alan.hayward@arm.com>
19865 David Sherwood <david.sherwood@arm.com>
19867 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19869 * gimple-expr.h (is_gimple_addressable: Likewise.
19870 * gimple-expr.c (is_gimple_address): Likewise.
19871 * internal-fn.c (expand_call_mem_ref): New function.
19872 (expand_mask_load_optab_fn): Use it.
19873 (expand_mask_store_optab_fn): Likewise.
19875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19876 Alan Hayward <alan.hayward@arm.com>
19877 David Sherwood <david.sherwood@arm.com>
19879 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19880 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19881 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19882 (cond_umax@var{mode}): Document.
19883 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19884 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19885 (cond_umin_optab, cond_umax_optab): New optabs.
19886 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19887 (COND_IOR, COND_XOR): New internal functions.
19888 * internal-fn.h (get_conditional_internal_fn): Declare.
19889 * internal-fn.c (cond_binary_direct): New macro.
19890 (expand_cond_binary_optab_fn): Likewise.
19891 (direct_cond_binary_optab_supported_p): Likewise.
19892 (get_conditional_internal_fn): New function.
19893 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19894 Cope with reduction statements that are vectorized as calls rather
19896 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19897 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19898 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19899 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19900 (UNSPEC_COND_EOR): New unspecs.
19901 (optab): Add mappings for them.
19902 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19903 (sve_int_op, sve_fp_op): New int attributes.
19905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19906 Alan Hayward <alan.hayward@arm.com>
19907 David Sherwood <david.sherwood@arm.com>
19909 * optabs.def (while_ult_optab): New optab.
19910 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19911 * internal-fn.def (WHILE_ULT): New internal function.
19912 * internal-fn.h (direct_internal_fn_supported_p): New override
19913 that takes two types as argument.
19914 * internal-fn.c (while_direct): New macro.
19915 (expand_while_optab_fn): New function.
19916 (convert_optab_supported_p): Likewise.
19917 (direct_while_optab_supported_p): New macro.
19918 * wide-int.h (wi::udiv_ceil): New function.
19919 * tree-vectorizer.h (rgroup_masks): New structure.
19920 (vec_loop_masks): New typedef.
19921 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19922 and fully_masked_p.
19923 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19924 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19925 (vect_max_vf): New function.
19926 (slpeel_make_loop_iterate_ntimes): Delete.
19927 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19928 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19929 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19930 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19931 internal-fn.h, stor-layout.h and optabs-query.h.
19932 (vect_set_loop_mask): New function.
19933 (add_preheader_seq): Likewise.
19934 (add_header_seq): Likewise.
19935 (interleave_supported_p): Likewise.
19936 (vect_maybe_permute_loop_masks): Likewise.
19937 (vect_set_loop_masks_directly): Likewise.
19938 (vect_set_loop_condition_masked): Likewise.
19939 (vect_set_loop_condition_unmasked): New function, split out from
19940 slpeel_make_loop_iterate_ntimes.
19941 (slpeel_make_loop_iterate_ntimes): Rename to..
19942 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19943 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19944 (vect_do_peeling): Update call accordingly.
19945 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19947 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19948 mask_compare_type, can_fully_mask_p and fully_masked_p.
19949 (release_vec_loop_masks): New function.
19950 (_loop_vec_info): Use it to free the loop masks.
19951 (can_produce_all_loop_masks_p): New function.
19952 (vect_get_max_nscalars_per_iter): Likewise.
19953 (vect_verify_full_masking): Likewise.
19954 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19955 retries, and free the mask rgroups before retrying. Check loop-wide
19956 reasons for disallowing fully-masked loops. Make the final decision
19957 about whether use a fully-masked loop or not.
19958 (vect_estimate_min_profitable_iters): Do not assume that peeling
19959 for the number of iterations will be needed for fully-masked loops.
19960 (vectorizable_reduction): Disable fully-masked loops.
19961 (vectorizable_live_operation): Likewise.
19962 (vect_halve_mask_nunits): New function.
19963 (vect_double_mask_nunits): Likewise.
19964 (vect_record_loop_mask): Likewise.
19965 (vect_get_loop_mask): Likewise.
19966 (vect_transform_loop): Handle the case in which the final loop
19967 iteration might handle a partial vector. Call vect_set_loop_condition
19968 instead of slpeel_make_loop_iterate_ntimes.
19969 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19970 (check_load_store_masking): New function.
19971 (prepare_load_store_mask): Likewise.
19972 (vectorizable_store): Handle fully-masked loops.
19973 (vectorizable_load): Likewise.
19974 (supportable_widening_operation): Use vect_halve_mask_nunits for
19976 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19977 (vect_gen_while): New function.
19978 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19979 (aarch64_uqdec<mode>): New insn.
19981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19982 Alan Hayward <alan.hayward@arm.com>
19983 David Sherwood <david.sherwood@arm.com>
19985 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19986 (reduc_xor_scal_optab): New optabs.
19987 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19988 (reduc_xor_scal_@var{m}): Document.
19989 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19990 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19991 internal functions.
19992 * fold-const-call.c (fold_const_call): Handle them.
19993 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19994 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19995 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19996 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19997 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19998 (UNSPEC_XORV): New unspecs.
19999 (optab): Add entries for them.
20000 (BITWISEV): New int iterator.
20001 (bit_reduc_op): New int attributes.
20003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20004 Alan Hayward <alan.hayward@arm.com>
20005 David Sherwood <david.sherwood@arm.com>
20007 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
20008 * internal-fn.def (VEC_SHL_INSERT): New internal function.
20009 * optabs.def (vec_shl_insert_optab): New optab.
20010 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
20011 (duplicate_and_interleave): Likewise.
20012 * tree-vect-loop.c: Include internal-fn.h.
20013 (neutral_op_for_slp_reduction): New function, split out from
20014 get_initial_defs_for_reduction.
20015 (get_initial_def_for_reduction): Handle option 2 for variable-length
20016 vectors by loading the neutral value into a vector and then shifting
20017 the initial value into element 0.
20018 (get_initial_defs_for_reduction): Replace the code argument with
20019 the neutral value calculated by neutral_op_for_slp_reduction.
20020 Use gimple_build_vector for constant-length vectors.
20021 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
20022 but the first group_size elements have a neutral value.
20023 Use duplicate_and_interleave otherwise.
20024 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
20025 Update call to get_initial_defs_for_reduction. Handle SLP
20026 reductions for variable-length vectors by creating one vector
20027 result for each scalar result, with the elements associated
20028 with other scalar results stubbed out with the neutral value.
20029 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
20030 Require IFN_VEC_SHL_INSERT for double reductions on
20031 variable-length vectors, or SLP reductions that have
20032 a neutral value. Require can_duplicate_and_interleave_p
20033 support for variable-length unchained SLP reductions if there
20034 is no neutral value, such as for MIN/MAX reductions. Also require
20035 the number of vector elements to be a multiple of the number of
20036 SLP statements when doing variable-length unchained SLP reductions.
20037 Update call to vect_create_epilog_for_reduction.
20038 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
20039 and remove initial values.
20040 (duplicate_and_interleave): Make public.
20041 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
20042 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
20044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20045 Alan Hayward <alan.hayward@arm.com>
20046 David Sherwood <david.sherwood@arm.com>
20048 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
20049 (can_duplicate_and_interleave_p): New function.
20050 (vect_get_and_check_slp_defs): Take the vector of statements
20051 rather than just the current one. Remove excess parentheses.
20052 Restriction rejectinon of vect_constant_def and vect_external_def
20053 for variable-length vectors to boolean types, or types for which
20054 can_duplicate_and_interleave_p is false.
20055 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
20056 (duplicate_and_interleave): New function.
20057 (vect_get_constant_vectors): Use gimple_build_vector for
20058 constant-length vectors and suitable variable-length constant
20059 vectors. Use duplicate_and_interleave for other variable-length
20060 vectors. Don't defer the update when inserting new statements.
20062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20063 Alan Hayward <alan.hayward@arm.com>
20064 David Sherwood <david.sherwood@arm.com>
20066 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
20067 min_profitable_iters doesn't go negative.
20069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20070 Alan Hayward <alan.hayward@arm.com>
20071 David Sherwood <david.sherwood@arm.com>
20073 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
20074 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
20075 * optabs.def (vec_mask_load_lanes_optab): New optab.
20076 (vec_mask_store_lanes_optab): Likewise.
20077 * internal-fn.def (MASK_LOAD_LANES): New internal function.
20078 (MASK_STORE_LANES): Likewise.
20079 * internal-fn.c (mask_load_lanes_direct): New macro.
20080 (mask_store_lanes_direct): Likewise.
20081 (expand_mask_load_optab_fn): Handle masked operations.
20082 (expand_mask_load_lanes_optab_fn): New macro.
20083 (expand_mask_store_optab_fn): Handle masked operations.
20084 (expand_mask_store_lanes_optab_fn): New macro.
20085 (direct_mask_load_lanes_optab_supported_p): Likewise.
20086 (direct_mask_store_lanes_optab_supported_p): Likewise.
20087 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
20089 (vect_load_lanes_supported): Likewise.
20090 * tree-vect-data-refs.c (strip_conversion): New function.
20091 (can_group_stmts_p): Likewise.
20092 (vect_analyze_data_ref_accesses): Use it instead of checking
20093 for a pair of assignments.
20094 (vect_store_lanes_supported): Take a masked_p parameter.
20095 (vect_load_lanes_supported): Likewise.
20096 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
20097 vect_store_lanes_supported and vect_load_lanes_supported.
20098 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
20099 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
20100 parameter. Don't allow gaps for masked accesses.
20101 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
20102 and vect_load_lanes_supported.
20103 (get_load_store_type): Take a masked_p parameter and update
20104 call to get_group_load_store_type.
20105 (vectorizable_store): Update call to get_load_store_type.
20106 Handle IFN_MASK_STORE_LANES.
20107 (vectorizable_load): Update call to get_load_store_type.
20108 Handle IFN_MASK_LOAD_LANES.
20110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20111 Alan Hayward <alan.hayward@arm.com>
20112 David Sherwood <david.sherwood@arm.com>
20114 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
20116 * config/aarch64/aarch64-protos.h
20117 (aarch64_sve_struct_memory_operand_p): Declare.
20118 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
20119 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
20120 (VPRED, vpred): Handle SVE structure modes.
20121 * config/aarch64/constraints.md (Utx): New constraint.
20122 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
20123 (aarch64_sve_struct_nonimmediate_operand): New predicates.
20124 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
20125 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
20126 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
20127 structure modes. Split into pieces after RA.
20128 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
20129 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
20131 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
20132 SVE structure modes.
20133 (aarch64_classify_address): Likewise.
20134 (sizetochar): Move earlier in file.
20135 (aarch64_print_operand): Handle SVE register lists.
20136 (aarch64_array_mode): New function.
20137 (aarch64_sve_struct_memory_operand_p): Likewise.
20138 (TARGET_ARRAY_MODE): Redefine.
20140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20141 Alan Hayward <alan.hayward@arm.com>
20142 David Sherwood <david.sherwood@arm.com>
20144 * target.def (array_mode): New target hook.
20145 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
20146 * doc/tm.texi: Regenerate.
20147 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
20148 * hooks.c (hook_optmode_mode_uhwi_none): New function.
20149 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
20150 targetm.array_mode.
20151 * stor-layout.c (mode_for_array): Likewise. Support polynomial
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 * fold-const.c (fold_binary_loc): Check the argument types
20159 rather than the result type when testing for a vector operation.
20161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20163 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
20164 * doc/tm.texi: Regenerate.
20166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20167 Alan Hayward <alan.hayward@arm.com>
20168 David Sherwood <david.sherwood@arm.com>
20170 * doc/invoke.texi (-msve-vector-bits=): Document new option.
20171 (sve): Document new AArch64 extension.
20172 * doc/md.texi (w): Extend the description of the AArch64
20173 constraint to include SVE vectors.
20174 (Upl, Upa): Document new AArch64 predicate constraints.
20175 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
20177 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
20178 (msve-vector-bits=): New option.
20179 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
20180 SVE when these are disabled.
20181 (sve): New extension.
20182 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
20183 modes. Adjust their number of units based on aarch64_sve_vg.
20184 (MAX_BITSIZE_MODE_ANY_MODE): Define.
20185 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
20186 aarch64_addr_query_type.
20187 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
20188 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
20189 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
20190 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
20191 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
20192 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
20193 (aarch64_simd_imm_zero_p): Delete.
20194 (aarch64_check_zero_based_sve_index_immediate): Declare.
20195 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20196 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20197 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20198 (aarch64_sve_float_mul_immediate_p): Likewise.
20199 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20200 rather than an rtx.
20201 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
20202 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
20203 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
20204 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
20205 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
20206 (aarch64_regmode_natural_size): Likewise.
20207 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
20208 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
20210 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
20211 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
20212 for VG and the SVE predicate registers.
20213 (V_ALIASES): Add a "z"-prefixed alias.
20214 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
20215 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
20216 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
20217 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
20218 (REG_CLASS_NAMES): Add entries for them.
20219 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
20220 and the predicate registers.
20221 (aarch64_sve_vg): Declare.
20222 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
20223 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
20224 (REGMODE_NATURAL_SIZE): Define.
20225 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
20227 * config/aarch64/aarch64.c: Include cfgrtl.h.
20228 (simd_immediate_info): Add a constructor for series vectors,
20229 and an associated step field.
20230 (aarch64_sve_vg): New variable.
20231 (aarch64_dbx_register_number): Handle VG and the predicate registers.
20232 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
20233 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
20234 (VEC_ANY_DATA, VEC_STRUCT): New constants.
20235 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
20236 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
20237 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
20238 (aarch64_get_mask_mode): New functions.
20239 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
20240 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
20241 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
20242 predicate modes and predicate registers. Explicitly restrict
20243 GPRs to modes of 16 bytes or smaller. Only allow FP registers
20244 to store a vector mode if it is recognized by
20245 aarch64_classify_vector_mode.
20246 (aarch64_regmode_natural_size): New function.
20247 (aarch64_hard_regno_caller_save_mode): Return the original mode
20249 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
20250 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
20251 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
20252 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
20254 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
20255 does not overlap dest if the function is frame-related. Handle
20257 (aarch64_split_add_offset): New function.
20258 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
20259 them aarch64_add_offset.
20260 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20261 and update call to aarch64_sub_sp.
20262 (aarch64_add_cfa_expression): New function.
20263 (aarch64_expand_prologue): Pass extra temporary registers to the
20264 functions above. Handle the case in which we need to emit new
20265 DW_CFA_expressions for registers that were originally saved
20266 relative to the stack pointer, but now have to be expressed
20267 relative to the frame pointer.
20268 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20270 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20271 IP0 and IP1 values for SVE frames.
20272 (aarch64_expand_vec_series): New function.
20273 (aarch64_expand_sve_widened_duplicate): Likewise.
20274 (aarch64_expand_sve_const_vector): Likewise.
20275 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20276 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20277 into the register, rather than emitting a SET directly.
20278 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20279 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20280 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20281 (offset_9bit_signed_scaled_p): New functions.
20282 (aarch64_replicate_bitmask_imm): New function.
20283 (aarch64_bitmask_imm): Use it.
20284 (aarch64_cannot_force_const_mem): Reject expressions involving
20285 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20286 (aarch64_classify_index): Handle SVE indices, by requiring
20287 a plain register index with a scale that matches the element size.
20288 (aarch64_classify_address): Handle SVE addresses. Assert that
20289 the mode of the address is VOIDmode or an integer mode.
20290 Update call to aarch64_classify_symbol.
20291 (aarch64_classify_symbolic_expression): Update call to
20292 aarch64_classify_symbol.
20293 (aarch64_const_vec_all_in_range_p): New function.
20294 (aarch64_print_vector_float_operand): Likewise.
20295 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20296 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20297 and the FP immediates 1.0 and 0.5.
20298 (aarch64_print_address_internal): Handle SVE addresses.
20299 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20300 (aarch64_regno_regclass): Handle predicate registers.
20301 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20303 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20304 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20305 (aarch64_convert_sve_vector_bits): New function.
20306 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20307 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20308 rather than an rtx.
20309 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20310 Handle SVE vector and predicate modes. Accept VL-based constants
20311 that need only one temporary register, and VL offsets that require
20312 no temporary registers.
20313 (aarch64_conditional_register_usage): Mark the predicate registers
20314 as fixed if SVE isn't available.
20315 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20316 Return true for SVE vector and predicate modes.
20317 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20318 rather than an unsigned int. Handle SVE modes.
20319 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20321 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20323 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20324 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20325 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20326 (aarch64_sve_float_mul_immediate_p): New functions.
20327 (aarch64_sve_valid_immediate): New function.
20328 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20329 Explicitly reject structure modes. Check for INDEX constants.
20330 Handle PTRUE and PFALSE constants.
20331 (aarch64_check_zero_based_sve_index_immediate): New function.
20332 (aarch64_simd_imm_zero_p): Delete.
20333 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20334 vector modes. Accept constants in the range of CNT[BHWD].
20335 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20336 ask for an Advanced SIMD mode.
20337 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20338 (aarch64_simd_vector_alignment): Handle SVE predicates.
20339 (aarch64_vectorize_preferred_vector_alignment): New function.
20340 (aarch64_simd_vector_alignment_reachable): Use it instead of
20342 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20343 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20345 (MAX_VECT_LEN): Delete.
20346 (expand_vec_perm_d): Add a vec_flags field.
20347 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20348 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20349 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20351 (aarch64_evpc_rev): Rename to...
20352 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20353 (aarch64_evpc_rev_global): New function.
20354 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20355 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20357 (aarch64_evpc_sve_tbl): New function.
20358 (aarch64_expand_vec_perm_const_1): Update after rename of
20359 aarch64_evpc_rev. Handle SVE permutes too, trying
20360 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20361 than aarch64_evpc_tbl.
20362 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20363 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20364 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20365 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20366 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20367 (aarch64_expand_sve_vcond): New functions.
20368 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20369 of aarch64_vector_mode_p.
20370 (aarch64_dwarf_poly_indeterminate_value): New function.
20371 (aarch64_compute_pressure_classes): Likewise.
20372 (aarch64_can_change_mode_class): Likewise.
20373 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20374 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20375 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20376 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20377 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20378 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20379 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20380 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20382 (Dn, Dl, Dr): Accept const as well as const_vector.
20383 (Dz): Likewise. Compare against CONST0_RTX.
20384 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20385 of "vector" where appropriate.
20386 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20387 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20388 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20389 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20390 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20391 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20392 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20393 (v_int_equiv): Extend to SVE modes.
20394 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20396 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20397 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20398 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20399 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20400 (SVE_COND_FP_CMP): New int iterators.
20401 (perm_hilo): Handle the new unpack unspecs.
20402 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20404 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20405 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20406 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20407 (aarch64_equality_operator, aarch64_constant_vector_operand)
20408 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20409 (aarch64_sve_nonimmediate_operand): Likewise.
20410 (aarch64_sve_general_operand): Likewise.
20411 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20412 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20413 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20414 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20415 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20416 (aarch64_sve_float_arith_immediate): Likewise.
20417 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20418 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20419 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20420 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20421 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20422 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20423 (aarch64_sve_float_arith_operand): Likewise.
20424 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20425 (aarch64_sve_float_mul_operand): Likewise.
20426 (aarch64_sve_vec_perm_operand): Likewise.
20427 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20428 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20429 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20430 as well as const_vector.
20431 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20432 in file. Use CONST0_RTX and CONSTM1_RTX.
20433 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20434 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20435 Use aarch64_simd_imm_zero.
20436 * config/aarch64/aarch64-sve.md: New file.
20437 * config/aarch64/aarch64.md: Include it.
20438 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20439 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20440 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20441 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20442 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20443 (sve): New attribute.
20444 (enabled): Disable instructions with the sve attribute unless
20446 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20447 aarch64_expand_mov_immediate.
20448 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20449 CNT[BHSD] immediates.
20450 (movti): Split CONST_POLY_INT moves into two halves.
20451 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20452 Split additions that need a temporary here if the destination
20453 is the stack pointer.
20454 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20455 (*add<mode>3_poly_1): New instruction.
20456 (set_clobber_cc): New expander.
20458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20460 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20461 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20462 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20463 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20464 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20465 Change innermode from fixed_mode_size to machine_mode.
20466 (simplify_subreg): Update call accordingly. Handle a constant-sized
20467 subreg of a variable-length CONST_VECTOR.
20469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20470 Alan Hayward <alan.hayward@arm.com>
20471 David Sherwood <david.sherwood@arm.com>
20473 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20474 (add_offset_to_base): New function, split out from...
20475 (create_mem_ref): ...here. When handling a scale other than 1,
20476 check first whether the address is valid without the offset.
20477 Add it into the base if so, leaving the index and scale as-is.
20479 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20482 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20483 fold_for_warn before checking if arg2 is INTEGER_CST.
20485 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20487 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20488 (store_multiple_operation): Delete.
20489 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20490 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20491 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20492 guarded by TARGET_STRING.
20493 (rs6000_output_load_multiple): Delete.
20494 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20495 OPTION_MASK_STRING / TARGET_STRING handling.
20496 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20497 (const rs6000_opt_masks) <"string">: Change mask to 0.
20498 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20499 (MASK_STRING): Delete.
20500 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20502 (load_multiple): Delete.
20509 (store_multiple): Delete.
20516 (movmemsi_8reg): Delete.
20517 (corresponding unnamed define_insn): Delete.
20518 (movmemsi_6reg): Delete.
20519 (corresponding unnamed define_insn): Delete.
20520 (movmemsi_4reg): Delete.
20521 (corresponding unnamed define_insn): Delete.
20522 (movmemsi_2reg): Delete.
20523 (corresponding unnamed define_insn): Delete.
20524 (movmemsi_1reg): Delete.
20525 (corresponding unnamed define_insn): Delete.
20526 * config/rs6000/rs6000.opt (mno-string): New.
20527 (mstring): Replace by deprecation warning stub.
20528 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20530 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20532 * regrename.c (regrename_do_replace): If replacing the same
20533 reg multiple times, try to reuse last created gen_raw_REG.
20536 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20537 main to workaround a bug in GDB.
20539 2018-01-12 Tom de Vries <tom@codesourcery.com>
20542 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20544 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20546 PR rtl-optimization/80481
20547 * ira-color.c (get_cap_member): New function.
20548 (allocnos_conflict_by_live_ranges_p): Use it.
20549 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20550 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20552 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20555 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20556 (*saddl_se_1): Ditto.
20557 (*ssubsi_1): Ditto.
20558 (*ssubl_se_1): Ditto.
20560 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20562 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20563 rather than wi::to_widest for DR_INITs.
20564 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20565 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20566 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20568 (vect_analyze_group_access_1): Note that here.
20570 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20572 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20573 polynomial type sizes.
20575 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20577 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20578 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20579 (gimple_add_tmp_var): Likewise.
20581 2018-01-12 Martin Liska <mliska@suse.cz>
20583 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20584 (gimple_alloc_sizes): Likewise.
20585 (dump_gimple_statistics): Use PRIu64 in printf format.
20586 * gimple.h: Change uint64_t to int.
20588 2018-01-12 Martin Liska <mliska@suse.cz>
20590 * tree-core.h: Use uint64_t instead of int.
20591 * tree.c (tree_node_counts): Likewise.
20592 (tree_node_sizes): Likewise.
20593 (dump_tree_statistics): Use PRIu64 in printf format.
20595 2018-01-12 Martin Liska <mliska@suse.cz>
20597 * Makefile.in: As qsort_chk is implemented in vec.c, add
20598 vec.o to linkage of gencfn-macros.
20599 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20600 passing the info to record_node_allocation_statistics.
20601 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20603 * ggc-common.c (struct ggc_usage): Add operator== and use
20604 it in operator< and compare function.
20605 * mem-stats.h (struct mem_usage): Likewise.
20606 * vec.c (struct vec_usage): Remove operator< and compare
20607 function. Can be simply inherited.
20609 2018-01-12 Martin Jambor <mjambor@suse.cz>
20612 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20613 * tree-ssa-math-opts.c: Include domwalk.h.
20614 (convert_mult_to_fma_1): New function.
20615 (fma_transformation_info): New type.
20616 (fma_deferring_state): Likewise.
20617 (cancel_fma_deferring): New function.
20618 (result_of_phi): Likewise.
20619 (last_fma_candidate_feeds_initial_phi): Likewise.
20620 (convert_mult_to_fma): Added deferring logic, split actual
20621 transformation to convert_mult_to_fma_1.
20622 (math_opts_dom_walker): New type.
20623 (math_opts_dom_walker::after_dom_children): New method, body moved
20624 here from pass_optimize_widening_mul::execute, added deferring logic
20626 (pass_optimize_widening_mul::execute): Moved most of code to
20627 math_opts_dom_walker::after_dom_children.
20628 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20629 * config/i386/i386.c (ix86_option_override_internal): Added
20630 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20632 2018-01-12 Richard Biener <rguenther@suse.de>
20635 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20636 inline instance vars.
20638 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20641 * config/rx/rx.c (rx_is_restricted_memory_address):
20642 Handle SUBREG case.
20644 2018-01-12 Richard Biener <rguenther@suse.de>
20646 PR tree-optimization/80846
20647 * target.def (split_reduction): New target hook.
20648 * targhooks.c (default_split_reduction): New function.
20649 * targhooks.h (default_split_reduction): Declare.
20650 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20651 target requests first reduce vectors by combining low and high
20653 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20654 (get_vectype_for_scalar_type_and_size): Export.
20655 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20656 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20657 * doc/tm.texi: Regenerate.
20658 * config/i386/i386.c (ix86_split_reduction): Implement
20659 TARGET_VECTORIZE_SPLIT_REDUCTION.
20661 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20664 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20665 in PIC mode except for TARGET_VXWORKS_RTP.
20666 * config/sparc/sparc.c: Include cfgrtl.h.
20667 (TARGET_INIT_PIC_REG): Define.
20668 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20669 (sparc_pic_register_p): New predicate.
20670 (sparc_legitimate_address_p): Use it.
20671 (sparc_legitimize_pic_address): Likewise.
20672 (sparc_delegitimize_address): Likewise.
20673 (sparc_mode_dependent_address_p): Likewise.
20674 (gen_load_pcrel_sym): Remove 4th parameter.
20675 (load_got_register): Adjust call to above. Remove obsolete stuff.
20676 (sparc_expand_prologue): Do not call load_got_register here.
20677 (sparc_flat_expand_prologue): Likewise.
20678 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20679 (sparc_use_pseudo_pic_reg): New function.
20680 (sparc_init_pic_reg): Likewise.
20681 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20682 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20684 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20686 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20687 Add item for branch_cost.
20689 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20691 PR rtl-optimization/83565
20692 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20693 not extend the result to a larger mode for rotate operations.
20694 (num_sign_bit_copies1): Likewise.
20696 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20699 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20701 Use values-Xc.o for -pedantic.
20702 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20704 2018-01-12 Martin Liska <mliska@suse.cz>
20707 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20709 (possible_polymorphic_call_targets): Use it.
20710 (ipa_devirt): Likewise.
20712 2018-01-12 Martin Liska <mliska@suse.cz>
20714 * profile-count.h (enum profile_quality): Use 0 as invalid
20715 enum value of profile_quality.
20717 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20719 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20720 -mext-string options.
20722 2018-01-12 Richard Biener <rguenther@suse.de>
20724 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20725 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20726 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20728 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20730 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20732 * configure.ac (--with-long-double-format): Add support for the
20733 configuration option to change the default long double format on
20735 * config.gcc (powerpc*-linux*-*): Likewise.
20736 * configure: Regenerate.
20737 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20738 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20739 used without modification.
20741 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20743 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20744 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20745 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20746 MISC_BUILTIN_SPEC_BARRIER.
20747 (rs6000_init_builtins): Likewise.
20748 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20750 (speculation_barrier): New define_insn.
20751 * doc/extend.texi: Document __builtin_speculation_barrier.
20753 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20756 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20757 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20758 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20760 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20761 integral modes instead of "ss" and "sd".
20762 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20763 vectors with 32-bit and 64-bit elements.
20764 (vecdupssescalarmodesuffix): New mode attribute.
20765 (vec_dup<mode>): Use it.
20767 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20770 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20771 frame if argument is passed on stack.
20773 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20776 * ree.c (combine_reaching_defs): Optimize also
20777 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20778 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20780 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20782 PR middle-end/83189
20783 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20785 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20787 PR middle-end/83718
20788 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20789 after they are computed.
20791 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20793 PR tree-optimization/83695
20794 * gimple-loop-linterchange.cc
20795 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20796 reset cached scev information after interchange.
20797 (pass_linterchange::execute): Remove call to scev_reset_htab.
20799 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20801 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20802 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20803 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20804 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20805 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20806 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20807 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20808 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20809 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20810 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20811 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20812 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20813 (V_lane_reg): Likewise.
20814 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20816 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20817 (vfmal_lane_low<mode>_intrinsic,
20818 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20819 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20820 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20821 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20822 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20823 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20825 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20827 * config/arm/arm-cpus.in (fp16fml): New feature.
20828 (ALL_SIMD): Add fp16fml.
20829 (armv8.2-a): Add fp16fml as an option.
20830 (armv8.3-a): Likewise.
20831 (armv8.4-a): Add fp16fml as part of fp16.
20832 * config/arm/arm.h (TARGET_FP16FML): Define.
20833 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20835 * config/arm/arm-modes.def (V2HF): Define.
20836 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20837 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20838 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20839 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20840 vfmsl_low, vfmsl_high): New set of builtins.
20841 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20842 (vfml_op): New code attribute.
20843 (VFMLHALVES): New int iterator.
20844 (VFML, VFMLSEL): New mode attributes.
20845 (V_reg): Define mapping for V2HF.
20846 (V_hi, V_lo): New mode attributes.
20847 (VF_constraint): Likewise.
20848 (vfml_half, vfml_half_selector): New int attributes.
20849 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20851 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20852 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20854 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20855 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20856 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20857 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20859 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20860 Document new effective target and option set.
20862 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20864 * config/arm/arm-cpus.in (armv8_4): New feature.
20865 (ARMv8_4a): New fgroup.
20866 (armv8.4-a): New arch.
20867 * config/arm/arm-tables.opt: Regenerate.
20868 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20869 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20870 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20871 Add matching rules for -march=armv8.4-a and extensions.
20872 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20874 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20877 * config/rx/rx.md (BW): New mode attribute.
20878 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20880 2018-01-11 Richard Biener <rguenther@suse.de>
20882 PR tree-optimization/83435
20883 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20884 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20885 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20887 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20888 Alan Hayward <alan.hayward@arm.com>
20889 David Sherwood <david.sherwood@arm.com>
20891 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20893 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20894 (aarch64_print_address_internal): Use it to check for a zero offset.
20896 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20897 Alan Hayward <alan.hayward@arm.com>
20898 David Sherwood <david.sherwood@arm.com>
20900 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20901 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20902 Return a poly_int64 rather than a HOST_WIDE_INT.
20903 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20904 rather than a HOST_WIDE_INT.
20905 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20906 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20907 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20908 final_offset from HOST_WIDE_INT to poly_int64.
20909 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20910 to_constant when getting the number of units in an Advanced SIMD
20912 (aarch64_builtin_vectorized_function): Check for a constant number
20914 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20916 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20917 attribute instead of GET_MODE_NUNITS.
20918 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20919 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20920 GET_MODE_SIZE for fixed-size registers.
20921 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20922 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20923 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20924 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20925 (aarch64_print_operand, aarch64_print_address_internal)
20926 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20927 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20928 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20929 Handle polynomial GET_MODE_SIZE.
20930 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20931 wider than SImode without modification.
20932 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20933 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20934 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20935 passing and returning SVE modes.
20936 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20937 rather than GEN_INT.
20938 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20939 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20940 (aarch64_allocate_and_probe_stack_space): Likewise.
20941 (aarch64_layout_frame): Cope with polynomial offsets.
20942 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20943 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20944 polynomial offsets.
20945 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20946 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20947 poly_int64 rather than a HOST_WIDE_INT.
20948 (aarch64_get_separate_components, aarch64_process_components)
20949 (aarch64_expand_prologue, aarch64_expand_epilogue)
20950 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20951 (aarch64_anchor_offset): New function, split out from...
20952 (aarch64_legitimize_address): ...here.
20953 (aarch64_builtin_vectorization_cost): Handle polynomial
20954 TYPE_VECTOR_SUBPARTS.
20955 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20957 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20958 number of elements from the PARALLEL rather than the mode.
20959 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20960 rather than GET_MODE_BITSIZE.
20961 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20962 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20963 (aarch64_expand_vec_perm_const_1): Handle polynomial
20964 d->perm.length () and d->perm elements.
20965 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20966 Apply to_constant to d->perm elements.
20967 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20968 polynomial CONST_VECTOR_NUNITS.
20969 (aarch64_move_pointer): Take amount as a poly_int64 rather
20971 (aarch64_progress_pointer): Avoid temporary variable.
20972 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20973 the mode attribute instead of GET_MODE.
20975 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20976 Alan Hayward <alan.hayward@arm.com>
20977 David Sherwood <david.sherwood@arm.com>
20979 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20980 x exists before using it.
20981 (aarch64_add_constant_internal): Rename to...
20982 (aarch64_add_offset_1): ...this. Replace regnum with separate
20983 src and dest rtxes. Handle the case in which they're different,
20984 including when the offset is zero. Replace scratchreg with an rtx.
20985 Use 2 additions if there is no spare register into which we can
20986 move a 16-bit constant.
20987 (aarch64_add_constant): Delete.
20988 (aarch64_add_offset): Replace reg with separate src and dest
20989 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20990 Use aarch64_add_offset_1.
20991 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20992 an rtx rather than an int. Take the delta as a poly_int64
20993 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20994 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20995 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20996 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20997 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20998 and aarch64_add_sp.
20999 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
21000 aarch64_add_constant.
21002 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21004 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
21005 Use scalar_float_mode.
21007 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21009 * config/aarch64/aarch64-simd.md
21010 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
21011 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
21012 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
21013 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
21014 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
21015 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
21016 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
21017 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
21018 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
21019 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
21021 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21024 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
21025 targ_options->x_arm_arch_string is non NULL.
21027 2018-01-11 Tamar Christina <tamar.christina@arm.com>
21029 * config/aarch64/aarch64.h
21030 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
21032 2018-01-11 Sudakshina Das <sudi.das@arm.com>
21035 * expmed.c (emit_store_flag_force): Swap if const op0
21036 and change VOIDmode to mode of op0.
21038 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21040 PR rtl-optimization/83761
21041 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
21042 than bytes to mode_for_size.
21044 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21046 PR middle-end/83189
21047 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
21048 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
21051 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21053 PR middle-end/83575
21054 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
21055 when in layout mode.
21056 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
21057 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
21060 2018-01-10 Michael Collison <michael.collison@arm.com>
21062 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
21063 * config/aarch64/aarch64-option-extension.def: Add
21064 AARCH64_OPT_EXTENSION of 'fp16fml'.
21065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21066 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
21067 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
21068 * config/aarch64/constraints.md (Ui7): New constraint.
21069 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
21070 (VFMLA_SEL_W): Ditto.
21073 (VFMLA16_LOW): New int iterator.
21074 (VFMLA16_HIGH): Ditto.
21075 (UNSPEC_FMLAL): New unspec.
21076 (UNSPEC_FMLSL): Ditto.
21077 (UNSPEC_FMLAL2): Ditto.
21078 (UNSPEC_FMLSL2): Ditto.
21079 (f16mac): New code attribute.
21080 * config/aarch64/aarch64-simd-builtins.def
21081 (aarch64_fmlal_lowv2sf): Ditto.
21082 (aarch64_fmlsl_lowv2sf): Ditto.
21083 (aarch64_fmlalq_lowv4sf): Ditto.
21084 (aarch64_fmlslq_lowv4sf): Ditto.
21085 (aarch64_fmlal_highv2sf): Ditto.
21086 (aarch64_fmlsl_highv2sf): Ditto.
21087 (aarch64_fmlalq_highv4sf): Ditto.
21088 (aarch64_fmlslq_highv4sf): Ditto.
21089 (aarch64_fmlal_lane_lowv2sf): Ditto.
21090 (aarch64_fmlsl_lane_lowv2sf): Ditto.
21091 (aarch64_fmlal_laneq_lowv2sf): Ditto.
21092 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
21093 (aarch64_fmlalq_lane_lowv4sf): Ditto.
21094 (aarch64_fmlsl_lane_lowv4sf): Ditto.
21095 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
21096 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
21097 (aarch64_fmlal_lane_highv2sf): Ditto.
21098 (aarch64_fmlsl_lane_highv2sf): Ditto.
21099 (aarch64_fmlal_laneq_highv2sf): Ditto.
21100 (aarch64_fmlsl_laneq_highv2sf): Ditto.
21101 (aarch64_fmlalq_lane_highv4sf): Ditto.
21102 (aarch64_fmlsl_lane_highv4sf): Ditto.
21103 (aarch64_fmlalq_laneq_highv4sf): Ditto.
21104 (aarch64_fmlsl_laneq_highv4sf): Ditto.
21105 * config/aarch64/aarch64-simd.md:
21106 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
21107 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21108 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
21109 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21110 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
21111 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
21112 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
21113 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
21114 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
21115 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
21116 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
21117 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
21118 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
21119 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
21120 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
21121 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
21122 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
21123 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
21124 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
21125 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
21126 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
21127 (vfmlsl_low_u32): Ditto.
21128 (vfmlalq_low_u32): Ditto.
21129 (vfmlslq_low_u32): Ditto.
21130 (vfmlal_high_u32): Ditto.
21131 (vfmlsl_high_u32): Ditto.
21132 (vfmlalq_high_u32): Ditto.
21133 (vfmlslq_high_u32): Ditto.
21134 (vfmlal_lane_low_u32): Ditto.
21135 (vfmlsl_lane_low_u32): Ditto.
21136 (vfmlal_laneq_low_u32): Ditto.
21137 (vfmlsl_laneq_low_u32): Ditto.
21138 (vfmlalq_lane_low_u32): Ditto.
21139 (vfmlslq_lane_low_u32): Ditto.
21140 (vfmlalq_laneq_low_u32): Ditto.
21141 (vfmlslq_laneq_low_u32): Ditto.
21142 (vfmlal_lane_high_u32): Ditto.
21143 (vfmlsl_lane_high_u32): Ditto.
21144 (vfmlal_laneq_high_u32): Ditto.
21145 (vfmlsl_laneq_high_u32): Ditto.
21146 (vfmlalq_lane_high_u32): Ditto.
21147 (vfmlslq_lane_high_u32): Ditto.
21148 (vfmlalq_laneq_high_u32): Ditto.
21149 (vfmlslq_laneq_high_u32): Ditto.
21150 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
21151 (AARCH64_FL_FOR_ARCH8_4): New.
21152 (AARCH64_ISA_F16FML): New ISA flag.
21153 (TARGET_F16FML): New feature flag for fp16fml.
21154 (doc/invoke.texi): Document new fp16fml option.
21156 2018-01-10 Michael Collison <michael.collison@arm.com>
21158 * config/aarch64/aarch64-builtins.c:
21159 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
21160 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21161 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
21162 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
21163 (AARCH64_ISA_SHA3): New ISA flag.
21164 (TARGET_SHA3): New feature flag for sha3.
21165 * config/aarch64/iterators.md (sha512_op): New int attribute.
21166 (CRYPTO_SHA512): New int iterator.
21167 (UNSPEC_SHA512H): New unspec.
21168 (UNSPEC_SHA512H2): Ditto.
21169 (UNSPEC_SHA512SU0): Ditto.
21170 (UNSPEC_SHA512SU1): Ditto.
21171 * config/aarch64/aarch64-simd-builtins.def
21172 (aarch64_crypto_sha512hqv2di): New builtin.
21173 (aarch64_crypto_sha512h2qv2di): Ditto.
21174 (aarch64_crypto_sha512su0qv2di): Ditto.
21175 (aarch64_crypto_sha512su1qv2di): Ditto.
21176 (aarch64_eor3qv8hi): Ditto.
21177 (aarch64_rax1qv2di): Ditto.
21178 (aarch64_xarqv2di): Ditto.
21179 (aarch64_bcaxqv8hi): Ditto.
21180 * config/aarch64/aarch64-simd.md:
21181 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
21182 (aarch64_crypto_sha512su0qv2di): Ditto.
21183 (aarch64_crypto_sha512su1qv2di): Ditto.
21184 (aarch64_eor3qv8hi): Ditto.
21185 (aarch64_rax1qv2di): Ditto.
21186 (aarch64_xarqv2di): Ditto.
21187 (aarch64_bcaxqv8hi): Ditto.
21188 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
21189 (vsha512h2q_u64): Ditto.
21190 (vsha512su0q_u64): Ditto.
21191 (vsha512su1q_u64): Ditto.
21192 (veor3q_u16): Ditto.
21193 (vrax1q_u64): Ditto.
21194 (vxarq_u64): Ditto.
21195 (vbcaxq_u16): Ditto.
21196 * config/arm/types.md (crypto_sha512): New type attribute.
21197 (crypto_sha3): Ditto.
21198 (doc/invoke.texi): Document new sha3 option.
21200 2018-01-10 Michael Collison <michael.collison@arm.com>
21202 * config/aarch64/aarch64-builtins.c:
21203 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
21204 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21205 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
21206 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
21207 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
21208 (AARCH64_ISA_SM4): New ISA flag.
21209 (TARGET_SM4): New feature flag for sm4.
21210 * config/aarch64/aarch64-simd-builtins.def
21211 (aarch64_sm3ss1qv4si): Ditto.
21212 (aarch64_sm3tt1aq4si): Ditto.
21213 (aarch64_sm3tt1bq4si): Ditto.
21214 (aarch64_sm3tt2aq4si): Ditto.
21215 (aarch64_sm3tt2bq4si): Ditto.
21216 (aarch64_sm3partw1qv4si): Ditto.
21217 (aarch64_sm3partw2qv4si): Ditto.
21218 (aarch64_sm4eqv4si): Ditto.
21219 (aarch64_sm4ekeyqv4si): Ditto.
21220 * config/aarch64/aarch64-simd.md:
21221 (aarch64_sm3ss1qv4si): Ditto.
21222 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
21223 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
21224 (aarch64_sm4eqv4si): Ditto.
21225 (aarch64_sm4ekeyqv4si): Ditto.
21226 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
21227 (sm3part_op): Ditto.
21228 (CRYPTO_SM3TT): Ditto.
21229 (CRYPTO_SM3PART): Ditto.
21230 (UNSPEC_SM3SS1): New unspec.
21231 (UNSPEC_SM3TT1A): Ditto.
21232 (UNSPEC_SM3TT1B): Ditto.
21233 (UNSPEC_SM3TT2A): Ditto.
21234 (UNSPEC_SM3TT2B): Ditto.
21235 (UNSPEC_SM3PARTW1): Ditto.
21236 (UNSPEC_SM3PARTW2): Ditto.
21237 (UNSPEC_SM4E): Ditto.
21238 (UNSPEC_SM4EKEY): Ditto.
21239 * config/aarch64/constraints.md (Ui2): New constraint.
21240 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
21241 * config/arm/types.md (crypto_sm3): New type attribute.
21242 (crypto_sm4): Ditto.
21243 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
21244 (vsm3tt1aq_u32): Ditto.
21245 (vsm3tt1bq_u32): Ditto.
21246 (vsm3tt2aq_u32): Ditto.
21247 (vsm3tt2bq_u32): Ditto.
21248 (vsm3partw1q_u32): Ditto.
21249 (vsm3partw2q_u32): Ditto.
21250 (vsm4eq_u32): Ditto.
21251 (vsm4ekeyq_u32): Ditto.
21252 (doc/invoke.texi): Document new sm4 option.
21254 2018-01-10 Michael Collison <michael.collison@arm.com>
21256 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
21257 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
21258 (AARCH64_FL_FOR_ARCH8_4): New.
21259 (AARCH64_FL_V8_4): New flag.
21260 (doc/invoke.texi): Document new armv8.4-a option.
21262 2018-01-10 Michael Collison <michael.collison@arm.com>
21264 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21265 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21266 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21267 * config/aarch64/aarch64-option-extension.def: Add
21268 AARCH64_OPT_EXTENSION of 'sha2'.
21269 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21270 (crypto): Disable sha2 and aes if crypto disabled.
21271 (crypto): Enable aes and sha2 if enabled.
21272 (simd): Disable sha2 and aes if simd disabled.
21273 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21275 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21276 (TARGET_SHA2): New feature flag for sha2.
21277 (TARGET_AES): New feature flag for aes.
21278 * config/aarch64/aarch64-simd.md:
21279 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21280 conditional on TARGET_AES.
21281 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21282 (aarch64_crypto_sha1hsi): Make pattern conditional
21284 (aarch64_crypto_sha1hv4si): Ditto.
21285 (aarch64_be_crypto_sha1hv4si): Ditto.
21286 (aarch64_crypto_sha1su1v4si): Ditto.
21287 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21288 (aarch64_crypto_sha1su0v4si): Ditto.
21289 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21290 (aarch64_crypto_sha256su0v4si): Ditto.
21291 (aarch64_crypto_sha256su1v4si): Ditto.
21292 (doc/invoke.texi): Document new aes and sha2 options.
21294 2018-01-10 Martin Sebor <msebor@redhat.com>
21296 PR tree-optimization/83781
21297 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21300 2018-01-11 Martin Sebor <msebor@gmail.com>
21301 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21303 PR tree-optimization/83501
21304 PR tree-optimization/81703
21306 * tree-ssa-strlen.c (get_string_cst): Rename...
21307 (get_string_len): ...to this. Handle global constants.
21308 (handle_char_store): Adjust.
21310 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21311 Jim Wilson <jimw@sifive.com>
21313 * config/riscv/riscv-protos.h (riscv_output_return): New.
21314 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21315 (riscv_attribute_table, riscv_output_return),
21316 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21317 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21318 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21320 (riscv_expand_prologue): Add early return for naked function.
21321 (riscv_expand_epilogue): Likewise.
21322 (riscv_function_ok_for_sibcall): Return false for naked function.
21323 (riscv_set_current_function): New.
21324 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21325 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21326 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21327 * doc/extend.texi (RISC-V Function Attributes): New.
21329 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21331 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21332 check for 128-bit long double before checking TCmode.
21333 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21334 128-bit long doubles before checking TFmode or TCmode.
21335 (FLOAT128_IBM_P): Likewise.
21337 2018-01-10 Martin Sebor <msebor@redhat.com>
21339 PR tree-optimization/83671
21340 * builtins.c (c_strlen): Unconditionally return zero for the empty
21342 Use -Warray-bounds for warnings.
21343 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21344 for non-constant array indices with COMPONENT_REF, arrays of
21345 arrays, and pointers to arrays.
21346 (gimple_fold_builtin_strlen): Determine and set length range for
21347 non-constant character arrays.
21349 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21351 PR middle-end/81897
21352 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21355 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21357 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21359 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21362 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21363 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21364 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21365 indexed_or_indirect_operand predicate.
21366 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21367 (*vsx_le_perm_load_v8hi): Likewise.
21368 (*vsx_le_perm_load_v16qi): Likewise.
21369 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21370 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21371 (*vsx_le_perm_store_v8hi): Likewise.
21372 (*vsx_le_perm_store_v16qi): Likewise.
21373 (eight unnamed splitters): Likewise.
21375 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21377 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21378 * config/rs6000/emmintrin.h: Likewise.
21379 * config/rs6000/mmintrin.h: Likewise.
21380 * config/rs6000/xmmintrin.h: Likewise.
21382 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21385 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21387 * tree.c (tree_nop_conversion): Return true for location wrapper
21389 (maybe_wrap_with_location): New function.
21390 (selftest::check_strip_nops): New function.
21391 (selftest::test_location_wrappers): New function.
21392 (selftest::tree_c_tests): Call it.
21393 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21394 (maybe_wrap_with_location): New decl.
21395 (EXPR_LOCATION_WRAPPER_P): New macro.
21396 (location_wrapper_p): New inline function.
21397 (tree_strip_any_location_wrapper): New inline function.
21399 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21402 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21403 stack_realign_offset for the largest alignment of stack slot
21405 (ix86_find_max_used_stack_alignment): New function.
21406 (ix86_finalize_stack_frame_flags): Use it. Set
21407 max_used_stack_alignment if we don't realign stack.
21408 * config/i386/i386.h (machine_function): Add
21409 max_used_stack_alignment.
21411 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21413 * config/arm/arm.opt (-mbranch-cost): New option.
21414 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21417 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21420 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21421 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21423 2018-01-10 Richard Biener <rguenther@suse.de>
21426 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21427 early out so it also covers the case where we have a non-NULL
21430 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21432 PR tree-optimization/83753
21433 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21434 for non-strided grouped accesses if the number of elements is 1.
21436 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21439 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21440 * i386.h (TARGET_USE_GATHER): Define.
21441 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21443 2018-01-10 Martin Liska <mliska@suse.cz>
21446 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21447 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21449 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21450 CLEANUP_NO_PARTITIONING is not set.
21452 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21454 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21455 for vectors, as a partial revert of r254296.
21456 * rtl.h (const_vec_p): Delete.
21457 (const_vec_duplicate_p): Don't test for vector CONSTs.
21458 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21459 * expmed.c (make_tree): Likewise.
21462 * common.md (E, F): Use CONSTANT_P instead of checking for
21464 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21465 checking for CONST_VECTOR.
21467 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21469 PR middle-end/83575
21470 * predict.c (force_edge_cold): Handle in more sane way edges
21471 with no prediction.
21473 2018-01-09 Carl Love <cel@us.ibm.com>
21475 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21477 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21478 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21479 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21480 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21481 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21482 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21483 * config/rs6000/rs6000-protos.h: Add extern defition for
21484 rs6000_generate_float2_double_code.
21485 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21487 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21488 (float2_v2df): Add define_expand.
21490 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21493 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21494 op_mode in the force_to_mode call.
21496 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21498 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21499 instead of checking each element individually.
21500 (aarch64_evpc_uzp): Likewise.
21501 (aarch64_evpc_zip): Likewise.
21502 (aarch64_evpc_ext): Likewise.
21503 (aarch64_evpc_rev): Likewise.
21504 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21505 instead of checking each element individually. Return true without
21507 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21508 whether all selected elements come from the same input, instead of
21509 checking each element individually. Remove calls to gen_rtx_REG,
21510 start_sequence and end_sequence and instead assert that no rtl is
21513 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21515 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21516 order of HIGH and CONST checks.
21518 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21520 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21521 if the destination isn't an SSA_NAME.
21523 2018-01-09 Richard Biener <rguenther@suse.de>
21525 PR tree-optimization/83668
21526 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21528 (canonicalize_loop_form): ... here, renamed from ...
21529 (canonicalize_loop_closed_ssa_form): ... this and amended to
21530 swap successor edges for loop exit blocks to make us use
21531 the RPO order we need for initial schedule generation.
21533 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21535 PR tree-optimization/64811
21536 * match.pd: When optimizing comparisons with Inf, avoid
21537 introducing or losing exceptions from comparisons with NaN.
21539 2018-01-09 Martin Liska <mliska@suse.cz>
21542 * asan.c (shadow_mem_size): Add gcc_assert.
21544 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21546 Don't save registers in main().
21549 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21550 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21551 * config/avr/avr.c (avr_set_current_function): Don't error if
21552 naked, OS_task or OS_main are specified at the same time.
21553 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21555 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21557 * common/config/avr/avr-common.c (avr_option_optimization_table):
21558 Switch on -mmain-is-OS_task for optimizing compilations.
21560 2018-01-09 Richard Biener <rguenther@suse.de>
21562 PR tree-optimization/83572
21563 * graphite.c: Include cfganal.h.
21564 (graphite_transform_loops): Connect infinite loops to exit
21565 and remove fake edges at the end.
21567 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21569 * ipa-inline.c (edge_badness): Revert accidental checkin.
21571 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21574 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21575 symbols; not inline clones.
21577 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21580 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21581 hard registers. Formatting fixes.
21583 PR preprocessor/83722
21584 * gcc.c (try_generate_repro): Pass
21585 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21586 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21589 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21590 Kito Cheng <kito.cheng@gmail.com>
21592 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21593 (riscv_leaf_function_p): Delete.
21594 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21596 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21598 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21600 (do_ifelse): New function.
21601 (do_isel): New function.
21602 (do_sub3): New function.
21603 (do_add3): New function.
21604 (do_load_mask_compare): New function.
21605 (do_overlap_load_compare): New function.
21606 (expand_compare_loop): New function.
21607 (expand_block_compare): Call expand_compare_loop() when appropriate.
21608 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21609 option description.
21610 (-mblock-compare-inline-loop-limit): New option.
21612 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21615 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21616 Reverse order of second and third operands in first alternative.
21617 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21618 of first and second elements in UNSPEC_VPERMR vector.
21619 (altivec_expand_vec_perm_le): Likewise.
21621 2018-01-08 Jeff Law <law@redhat.com>
21623 PR rtl-optimizatin/81308
21624 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21625 (process_switch): If group_case_labels makes a change, then set
21627 (pass_convert_switch::execute): If a switch is converted, then
21628 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21630 PR rtl-optimization/81308
21631 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21634 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21636 PR target/83663 - Revert r255946
21637 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21638 generation for cases where splatting a value is not useful.
21639 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21640 across a vec_duplicate and a paradoxical subreg forming a vector
21641 mode to a vec_concat.
21643 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21645 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21646 -march=armv8.3-a variants.
21647 * config/arm/t-multilib: Likewise.
21648 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21650 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21652 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21654 (cceq_ior_compare_complement): Give it a name so I can use it, and
21655 change boolean_or_operator predicate to boolean_operator so it can
21656 be used to generate a crand.
21657 (eqne): New code iterator.
21658 (bd/bd_neg): New code_attrs.
21659 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21660 a single define_insn.
21661 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21662 decrement (bdnzt/bdnzf/bdzt/bdzf).
21663 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21664 with the new names of the branch decrement patterns, and added the
21665 names of the branch decrement conditional patterns.
21667 2018-01-08 Richard Biener <rguenther@suse.de>
21669 PR tree-optimization/83563
21670 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21673 2018-01-08 Richard Biener <rguenther@suse.de>
21675 PR middle-end/83713
21676 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21678 2018-01-08 Richard Biener <rguenther@suse.de>
21680 PR tree-optimization/83685
21681 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21682 references to abnormals.
21684 2018-01-08 Richard Biener <rguenther@suse.de>
21687 * dwarf2out.c (output_indirect_strings): Handle empty
21688 skeleton_debug_str_hash.
21689 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21691 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21693 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21694 (emit_store_direct): Likewise.
21695 (arc_trampoline_adjust_address): Likewise.
21696 (arc_asm_trampoline_template): New function.
21697 (arc_initialize_trampoline): Use asm_trampoline_template.
21698 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21699 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21700 * config/arc/arc.md (flush_icache): Delete pattern.
21702 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21704 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21705 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21708 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21711 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21712 by not USED_FOR_TARGET.
21713 (make_pass_resolve_sw_modes): Likewise.
21715 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21717 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21720 2018-01-08 Richard Biener <rguenther@suse.de>
21722 PR middle-end/83580
21723 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21725 2018-01-08 Richard Biener <rguenther@suse.de>
21727 PR middle-end/83517
21728 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21730 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21732 PR middle-end/81897
21733 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21734 basic blocks with a small number of successors.
21735 (convert_control_dep_chain_into_preds): Improve handling of
21737 (dump_predicates): Split apart into...
21738 (dump_pred_chain): ...here...
21739 (dump_pred_info): ...and here.
21740 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21741 (can_chain_union_be_invalidated_p): Improve check for invalidation
21743 (uninit_uses_cannot_happen): Avoid unnecessary if
21744 convert_control_dep_chain_into_preds yielded nothing.
21746 2018-01-06 Martin Sebor <msebor@redhat.com>
21748 PR tree-optimization/83640
21749 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21750 subtracting negative offset from size.
21751 (builtin_access::overlap): Adjust offset bounds of the access to fall
21752 within the size of the object if possible.
21754 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21756 PR rtl-optimization/83699
21757 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21758 extract_bit_field_as_subreg to cases in which the extracted
21759 value is also a vector.
21761 * lra-constraints.c (process_alt_operands): Test for the equivalence
21762 substitutions when detecting a possible reload cycle.
21764 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21767 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21768 by default if flag_selective_schedling{,2}. Formatting fixes.
21770 PR rtl-optimization/83682
21771 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21772 if it has non-VECTOR_MODE element mode.
21773 (vec_duplicate_p): Likewise.
21775 PR middle-end/83694
21776 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21777 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21779 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21782 * config/i386/i386-builtin.def
21783 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21784 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21785 Require also OPTION_MASK_ISA_AVX512F in addition to
21786 OPTION_MASK_ISA_GFNI.
21787 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21788 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21789 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21790 to OPTION_MASK_ISA_GFNI.
21791 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21792 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21793 OPTION_MASK_ISA_AVX512BW.
21794 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21795 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21796 addition to OPTION_MASK_ISA_GFNI.
21797 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21798 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21799 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21800 to OPTION_MASK_ISA_GFNI.
21801 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21802 a requirement for all ISAs rather than any of them with a few
21804 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21806 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21807 bitmasks to be enabled with 3 exceptions, instead of requiring any
21808 enabled ISA with lots of exceptions.
21809 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21810 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21811 Change avx512bw in isa attribute to avx512f.
21812 * config/i386/sgxintrin.h: Add license boilerplate.
21813 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21814 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21815 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21816 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21818 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21819 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21820 temporarily sse2 rather than sse if not enabled already.
21823 * config/i386/sse.md (VI248_VLBW): Rename to ...
21824 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21825 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21826 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21827 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21828 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21829 mode iterator instead of VI248_VLBW.
21831 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21833 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21834 (record_modified): Skip clobbers; add debug output.
21835 (param_change_prob): Use sreal frequencies.
21837 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21839 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21840 punt for user-aligned variables.
21842 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21844 * tree-chrec.c (chrec_contains_symbols): Return true for
21847 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21850 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21851 of (x|y) == x for BICS pattern.
21853 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21855 PR tree-optimization/83605
21856 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21857 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21860 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21862 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21863 * config/epiphany/rtems.h: New file.
21865 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21866 Uros Bizjak <ubizjak@gmail.com>
21869 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21870 QIreg_operand instead of register_operand predicate.
21871 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21872 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21873 comments instead of -fmitigate[-_]rop.
21875 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21878 * cgraphunit.c (symbol_table::compile): Switch to text_section
21879 before calling assembly_start debug hook.
21880 * run-rtl-passes.c (run_rtl_passes): Likewise.
21883 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21885 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21886 range_int_cst_p rather than !symbolic_range_p before calling
21887 extract_range_from_multiplicative_op_1.
21889 2018-01-04 Jeff Law <law@redhat.com>
21891 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21892 redundant test in assertion.
21894 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21896 * doc/rtl.texi: Document machine_mode wrapper classes.
21898 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21900 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21901 using tree_to_uhwi.
21903 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21905 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21906 the VEC_PERM_EXPR fold to fail.
21908 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21911 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21912 to switched_sections.
21914 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21917 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21918 test for d.testing.
21920 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21923 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21924 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21926 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21929 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21930 is BLKmode and bitpos not zero or mode change is needed.
21932 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21935 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21938 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21941 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21942 instead of MULT rtx. Update all corresponding splitters.
21943 (*saddl_se): Ditto.
21944 (*ssub<modesuffix>): Ditto.
21945 (*ssubl_se): Ditto.
21946 (*cmp_sadd_di): Update split patterns.
21947 (*cmp_sadd_si): Ditto.
21948 (*cmp_sadd_sidi): Ditto.
21949 (*cmp_ssub_di): Ditto.
21950 (*cmp_ssub_si): Ditto.
21951 (*cmp_ssub_sidi): Ditto.
21952 * config/alpha/predicates.md (const23_operand): New predicate.
21953 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21954 Look for ASHIFT, not MULT inner operand.
21955 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21957 2018-01-04 Martin Liska <mliska@suse.cz>
21959 PR gcov-profile/83669
21960 * gcov.c (output_intermediate_file): Add version to intermediate
21962 * doc/gcov.texi: Document new field 'version' in intermediate
21963 file format. Fix location of '-k' option of gcov command.
21965 2018-01-04 Martin Liska <mliska@suse.cz>
21968 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21970 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21972 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21974 2018-01-03 Martin Sebor <msebor@redhat.com>
21976 PR tree-optimization/83655
21977 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21978 checking calls with invalid arguments.
21980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21982 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21983 (vectorizable_mask_load_store): Delete.
21984 (vectorizable_call): Return false for masked loads and stores.
21985 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21986 instead of gimple_assign_rhs1.
21987 (vectorizable_load): Handle IFN_MASK_LOAD.
21988 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21992 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21994 (vectorizable_mask_load_store): ...here.
21995 (vectorizable_load): ...and here.
21997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21999 * tree-vect-stmts.c (vect_build_all_ones_mask)
22000 (vect_build_zero_merge_argument): New functions, split out from...
22001 (vectorizable_load): ...here.
22003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22005 * tree-vect-stmts.c (vect_check_store_rhs): New function,
22007 (vectorizable_mask_load_store): ...here.
22008 (vectorizable_store): ...and here.
22010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22012 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
22014 (vectorizable_mask_load_store): ...here.
22016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22018 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
22019 (vect_model_store_cost): Take a vec_load_store_type instead of a
22021 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
22022 (vect_model_store_cost): Take a vec_load_store_type instead of a
22024 (vectorizable_mask_load_store): Update accordingly.
22025 (vectorizable_store): Likewise.
22026 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
22028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22030 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
22031 IFN_MASK_LOAD calls here rather than...
22032 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
22034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22035 Alan Hayward <alan.hayward@arm.com>
22036 David Sherwood <david.sherwood@arm.com>
22038 * expmed.c (extract_bit_field_1): For vector extracts,
22039 fall back to extract_bit_field_as_subreg if vec_extract
22042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22043 Alan Hayward <alan.hayward@arm.com>
22044 David Sherwood <david.sherwood@arm.com>
22046 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
22047 they are variable or constant sized.
22048 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
22049 slots for constant-sized data.
22051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22052 Alan Hayward <alan.hayward@arm.com>
22053 David Sherwood <david.sherwood@arm.com>
22055 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
22056 handling COND_EXPRs with boolean comparisons, try to find a better
22057 basis for the mask type than the boolean itself.
22059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22061 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
22062 is calculated and how it can be overridden.
22063 * genmodes.c (max_bitsize_mode_any_mode): New variable.
22064 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
22066 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
22069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22070 Alan Hayward <alan.hayward@arm.com>
22071 David Sherwood <david.sherwood@arm.com>
22073 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
22074 Remove the mode argument.
22075 (aarch64_simd_valid_immediate): Remove the mode and inverse
22077 * config/aarch64/iterators.md (bitsize): New iterator.
22078 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
22079 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
22080 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
22081 aarch64_simd_valid_immediate.
22082 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
22083 (aarch64_reg_or_bic_imm): Likewise.
22084 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
22085 with an insn_type enum and msl with a modifier_type enum.
22086 Replace element_width with a scalar_mode. Change the shift
22087 to unsigned int. Add constructors for scalar_float_mode and
22088 scalar_int_mode elements.
22089 (aarch64_vect_float_const_representable_p): Delete.
22090 (aarch64_can_const_movi_rtx_p)
22091 (aarch64_simd_scalar_immediate_valid_for_move)
22092 (aarch64_simd_make_constant): Update call to
22093 aarch64_simd_valid_immediate.
22094 (aarch64_advsimd_valid_immediate_hs): New function.
22095 (aarch64_advsimd_valid_immediate): Likewise.
22096 (aarch64_simd_valid_immediate): Remove mode and inverse
22097 arguments. Rewrite to use the above. Use const_vec_duplicate_p
22098 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
22099 and aarch64_float_const_representable_p on the result.
22100 (aarch64_output_simd_mov_immediate): Remove mode argument.
22101 Update call to aarch64_simd_valid_immediate and use of
22102 simd_immediate_info.
22103 (aarch64_output_scalar_simd_mov_immediate): Update call
22106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22107 Alan Hayward <alan.hayward@arm.com>
22108 David Sherwood <david.sherwood@arm.com>
22110 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
22111 (mode_nunits): Likewise CONST_MODE_NUNITS.
22112 * machmode.def (ADJUST_NUNITS): Document.
22113 * genmodes.c (mode_data::need_nunits_adj): New field.
22114 (blank_mode): Update accordingly.
22115 (adj_nunits): New variable.
22116 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
22118 (emit_mode_size_inline): Set need_bytesize_adj for all modes
22119 listed in adj_nunits.
22120 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
22121 listed in adj_nunits. Don't emit case statements for such modes.
22122 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
22123 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
22124 nothing if adj_nunits is nonnull.
22125 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
22126 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
22127 (emit_mode_fbit): Update use of print_maybe_const_decl.
22128 (emit_move_size): Likewise. Treat the array as non-const
22130 (emit_mode_adjustments): Handle adj_nunits.
22132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22134 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
22135 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
22136 (VECTOR_MODES): Use it.
22137 (make_vector_modes): Take the prefix as an argument.
22139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22140 Alan Hayward <alan.hayward@arm.com>
22141 David Sherwood <david.sherwood@arm.com>
22143 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
22144 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
22145 for MODE_VECTOR_BOOL.
22146 * machmode.def (VECTOR_BOOL_MODE): Document.
22147 * genmodes.c (VECTOR_BOOL_MODE): New macro.
22148 (make_vector_bool_mode): New function.
22149 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
22151 * lto-streamer-in.c (lto_input_mode_table): Likewise.
22152 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
22154 * stor-layout.c (int_mode_for_mode): Likewise.
22155 * tree.c (build_vector_type_for_mode): Likewise.
22156 * varasm.c (output_constant_pool_2): Likewise.
22157 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
22158 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
22159 for MODE_VECTOR_BOOL.
22160 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
22161 of mode class checks.
22162 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
22163 instead of a list of mode class checks.
22164 (expand_vector_scalar_condition): Likewise.
22165 (type_for_widest_vector_mode): Handle BImode as an inner mode.
22167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22168 Alan Hayward <alan.hayward@arm.com>
22169 David Sherwood <david.sherwood@arm.com>
22171 * machmode.h (mode_size): Change from unsigned short to
22173 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
22174 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22175 or if measurement_type is not polynomial.
22176 (fixed_size_mode::includes_p): Check for constant-sized modes.
22177 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
22178 return a poly_uint16 rather than an unsigned short.
22179 (emit_mode_size): Change the type of mode_size from unsigned short
22180 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
22181 (emit_mode_adjustments): Cope with polynomial vector sizes.
22182 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22184 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22186 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
22187 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
22188 * caller-save.c (setup_save_areas): Likewise.
22189 (replace_reg_with_saved_mem): Likewise.
22190 * calls.c (emit_library_call_value_1): Likewise.
22191 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
22192 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
22193 (gen_lowpart_for_combine): Likewise.
22194 * convert.c (convert_to_integer_1): Likewise.
22195 * cse.c (equiv_constant, cse_insn): Likewise.
22196 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
22197 (cselib_subst_to_values): Likewise.
22198 * dce.c (word_dce_process_block): Likewise.
22199 * df-problems.c (df_word_lr_mark_ref): Likewise.
22200 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
22201 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
22202 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
22203 (rtl_for_decl_location): Likewise.
22204 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
22205 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
22206 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
22207 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
22208 (expand_expr_real_1): Likewise.
22209 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
22210 (pad_below): Likewise.
22211 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22212 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
22213 * ira.c (get_subreg_tracking_sizes): Likewise.
22214 * ira-build.c (ira_create_allocno_objects): Likewise.
22215 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
22216 (ira_sort_regnos_for_alter_reg): Likewise.
22217 * ira-costs.c (record_operand_costs): Likewise.
22218 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
22219 (resolve_simple_move): Likewise.
22220 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
22221 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
22222 (lra_constraints): Likewise.
22223 (CONST_POOL_OK_P): Reject variable-sized modes.
22224 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
22225 (add_pseudo_to_slot, lra_spill): Likewise.
22226 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22227 * optabs-query.c (get_best_extraction_insn): Likewise.
22228 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22229 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
22230 (expand_mult_highpart, valid_multiword_target_p): Likewise.
22231 * recog.c (offsettable_address_addr_space_p): Likewise.
22232 * regcprop.c (maybe_mode_change): Likewise.
22233 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
22234 * regrename.c (build_def_use): Likewise.
22235 * regstat.c (dump_reg_info): Likewise.
22236 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
22237 (find_reloads, find_reloads_subreg_address): Likewise.
22238 * reload1.c (eliminate_regs_1): Likewise.
22239 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
22240 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
22241 (simplify_binary_operation_1, simplify_subreg): Likewise.
22242 * targhooks.c (default_function_arg_padding): Likewise.
22243 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
22244 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
22245 (verify_gimple_assign_ternary): Likewise.
22246 * tree-inline.c (estimate_move_cost): Likewise.
22247 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22248 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
22249 (get_address_cost_ainc): Likewise.
22250 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
22251 (vect_supportable_dr_alignment): Likewise.
22252 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22253 (vectorizable_reduction): Likewise.
22254 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
22255 (vectorizable_operation, vectorizable_load): Likewise.
22256 * tree.c (build_same_sized_truth_vector_type): Likewise.
22257 * valtrack.c (cleanup_auto_inc_dec): Likewise.
22258 * var-tracking.c (emit_note_insn_var_location): Likewise.
22259 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
22260 (ADDR_VEC_ALIGN): Likewise.
22262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22263 Alan Hayward <alan.hayward@arm.com>
22264 David Sherwood <david.sherwood@arm.com>
22266 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22268 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22269 or if measurement_type is polynomial.
22270 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22271 * combine.c (make_extraction): Likewise.
22272 * dse.c (find_shift_sequence): Likewise.
22273 * dwarf2out.c (mem_loc_descriptor): Likewise.
22274 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22275 (extract_bit_field, extract_low_bits): Likewise.
22276 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22277 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22278 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22279 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22280 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22281 * reload.c (find_reloads): Likewise.
22282 * reload1.c (alter_reg): Likewise.
22283 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22284 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22285 * tree-if-conv.c (predicate_mem_writes): Likewise.
22286 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22287 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22288 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22289 * valtrack.c (dead_debug_insert_temp): Likewise.
22290 * varasm.c (mergeable_constant_section): Likewise.
22291 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22294 Alan Hayward <alan.hayward@arm.com>
22295 David Sherwood <david.sherwood@arm.com>
22297 * expr.c (expand_assignment): Cope with polynomial mode sizes
22298 when assigning to a CONCAT.
22300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22301 Alan Hayward <alan.hayward@arm.com>
22302 David Sherwood <david.sherwood@arm.com>
22304 * machmode.h (mode_precision): Change from unsigned short to
22306 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22308 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22309 or if measurement_type is not polynomial.
22310 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22311 in which the mode is already known to be a scalar_int_mode.
22312 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22313 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22315 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22316 for GET_MODE_PRECISION.
22317 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22318 for GET_MODE_PRECISION.
22319 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22321 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22322 (expand_field_assignment, make_extraction): Likewise.
22323 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22324 (get_last_value): Likewise.
22325 * convert.c (convert_to_integer_1): Likewise.
22326 * cse.c (cse_insn): Likewise.
22327 * expr.c (expand_expr_real_1): Likewise.
22328 * lra-constraints.c (simplify_operand_subreg): Likewise.
22329 * optabs-query.c (can_atomic_load_p): Likewise.
22330 * optabs.c (expand_atomic_load): Likewise.
22331 (expand_atomic_store): Likewise.
22332 * ree.c (combine_reaching_defs): Likewise.
22333 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22334 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22335 * tree.h (type_has_mode_precision_p): Likewise.
22336 * ubsan.c (instrument_si_overflow): Likewise.
22338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22339 Alan Hayward <alan.hayward@arm.com>
22340 David Sherwood <david.sherwood@arm.com>
22342 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22343 polynomial numbers of units.
22344 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22345 (valid_vector_subparts_p): New function.
22346 (build_vector_type): Remove temporary shim and take the number
22347 of units as a poly_uint64 rather than an int.
22348 (build_opaque_vector_type): Take the number of units as a
22349 poly_uint64 rather than an int.
22350 * tree.c (build_vector_from_ctor): Handle polynomial
22351 TYPE_VECTOR_SUBPARTS.
22352 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22353 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22354 (build_vector_from_val): If the number of units is variable,
22355 use build_vec_duplicate_cst for constant operands and
22356 VEC_DUPLICATE_EXPR otherwise.
22357 (make_vector_type): Remove temporary is_constant ().
22358 (build_vector_type, build_opaque_vector_type): Take the number of
22359 units as a poly_uint64 rather than an int.
22360 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22362 * cfgexpand.c (expand_debug_expr): Likewise.
22363 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22364 (store_constructor, expand_expr_real_1): Likewise.
22365 (const_scalar_mask_from_tree): Likewise.
22366 * fold-const-call.c (fold_const_reduction): Likewise.
22367 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22368 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22369 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22370 (fold_relational_const): Likewise.
22371 (native_interpret_vector): Likewise. Change the size from an
22372 int to an unsigned int.
22373 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22374 TYPE_VECTOR_SUBPARTS.
22375 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22376 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22377 duplicating a non-constant operand into a variable-length vector.
22378 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22379 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22380 * ipa-icf.c (sem_variable::equals): Likewise.
22381 * match.pd: Likewise.
22382 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22383 * print-tree.c (print_node): Likewise.
22384 * stor-layout.c (layout_type): Likewise.
22385 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22386 * tree-cfg.c (verify_gimple_comparison): Likewise.
22387 (verify_gimple_assign_binary): Likewise.
22388 (verify_gimple_assign_ternary): Likewise.
22389 (verify_gimple_assign_single): Likewise.
22390 * tree-pretty-print.c (dump_generic_node): Likewise.
22391 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22392 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22393 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22394 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22395 (vect_shift_permute_load_chain): Likewise.
22396 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22397 (expand_vector_condition, optimize_vector_constructor): Likewise.
22398 (lower_vec_perm, get_compute_type): Likewise.
22399 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22400 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22401 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22402 (vect_recog_mask_conversion_pattern): Likewise.
22403 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22404 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22405 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22406 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22407 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22408 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22409 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22410 (supportable_widening_operation): Likewise.
22411 (supportable_narrowing_operation): Likewise.
22412 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22414 * varasm.c (output_constant): Likewise.
22416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22417 Alan Hayward <alan.hayward@arm.com>
22418 David Sherwood <david.sherwood@arm.com>
22420 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22421 so that both the length == 3 and length != 3 cases set up their
22422 own permute vectors. Add comments explaining why we know the
22423 number of elements is constant.
22424 (vect_permute_load_chain): Likewise.
22426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22427 Alan Hayward <alan.hayward@arm.com>
22428 David Sherwood <david.sherwood@arm.com>
22430 * machmode.h (mode_nunits): Change from unsigned char to
22432 (ONLY_FIXED_SIZE_MODES): New macro.
22433 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22434 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22435 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22437 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22438 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22439 or if measurement_type is not polynomial.
22440 * genmodes.c (ZERO_COEFFS): New macro.
22441 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22443 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22444 Use ZERO_COEFFS when emitting initializers.
22445 * data-streamer.h (bp_pack_poly_value): New function.
22446 (bp_unpack_poly_value): Likewise.
22447 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22448 for GET_MODE_NUNITS.
22449 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22450 for GET_MODE_NUNITS.
22451 * tree.c (make_vector_type): Remove temporary shim and make
22452 the real function take the number of units as a poly_uint64
22453 rather than an int.
22454 (build_vector_type_for_mode): Handle polynomial nunits.
22455 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22456 * emit-rtl.c (const_vec_series_p_1): Likewise.
22457 (gen_rtx_CONST_VECTOR): Likewise.
22458 * fold-const.c (test_vec_duplicate_folding): Likewise.
22459 * genrecog.c (validate_pattern): Likewise.
22460 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22461 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22462 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22463 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22464 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22465 * rtlanal.c (subreg_get_info): Likewise.
22466 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22467 (vect_grouped_load_supported): Likewise.
22468 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22469 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22470 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22471 (simplify_const_unary_operation, simplify_binary_operation_1)
22472 (simplify_const_binary_operation, simplify_ternary_operation)
22473 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22474 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22475 instead of CONST_VECTOR_NUNITS.
22476 * varasm.c (output_constant_pool_2): Likewise.
22477 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22478 explicit-encoded elements in the XVEC for variable-length vectors.
22480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22482 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22485 Alan Hayward <alan.hayward@arm.com>
22486 David Sherwood <david.sherwood@arm.com>
22488 * coretypes.h (fixed_size_mode): Declare.
22489 (fixed_size_mode_pod): New typedef.
22490 * builtins.h (target_builtins::x_apply_args_mode)
22491 (target_builtins::x_apply_result_mode): Change type to
22492 fixed_size_mode_pod.
22493 * builtins.c (apply_args_size, apply_result_size, result_vector)
22494 (expand_builtin_apply_args_1, expand_builtin_apply)
22495 (expand_builtin_return): Update accordingly.
22497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22499 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22500 * cselib.c (cselib_hash_rtx): Likewise.
22501 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22502 CONST_VECTOR encoding.
22504 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22505 Jeff Law <law@redhat.com>
22508 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22509 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22510 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22511 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22514 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22515 explicitly probe *sp in a noreturn function if there were any callee
22516 register saves or frame pointer is needed.
22518 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22521 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22522 BLKmode for ternary, binary or unary expressions.
22525 * var-tracking.c (delete_vta_debug_insn): New inline function.
22526 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22527 insns from get_insns () to NULL instead of each bb separately.
22528 Use delete_vta_debug_insn. No longer static.
22529 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22530 delete_vta_debug_insns callers.
22531 * rtl.h (delete_vta_debug_insns): Declare.
22532 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22533 instead of variable_tracking_main.
22535 2018-01-03 Martin Sebor <msebor@redhat.com>
22537 PR tree-optimization/83603
22538 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22539 arguments past the endof the argument list in functions declared
22540 without a prototype.
22541 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22542 Avoid checking when arguments are null.
22544 2018-01-03 Martin Sebor <msebor@redhat.com>
22547 * doc/extend.texi (attribute const): Fix a typo.
22548 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22549 issuing -Wsuggest-attribute for void functions.
22551 2018-01-03 Martin Sebor <msebor@redhat.com>
22553 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22554 offset_int::from instead of wide_int::to_shwi.
22555 (maybe_diag_overlap): Remove assertion.
22556 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22557 * gimple-ssa-sprintf.c (format_directive): Same.
22558 (parse_directive): Same.
22559 (sprintf_dom_walker::compute_format_length): Same.
22560 (try_substitute_return_value): Same.
22562 2018-01-03 Jeff Law <law@redhat.com>
22564 PR middle-end/83654
22565 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22566 non-constant residual for zero at runtime and avoid probing in
22567 that case. Reorganize code for trailing problem to mirror handling
22570 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22572 PR tree-optimization/83501
22573 * tree-ssa-strlen.c (get_string_cst): New.
22574 (handle_char_store): Call get_string_cst.
22576 2018-01-03 Martin Liska <mliska@suse.cz>
22578 PR tree-optimization/83593
22579 * tree-ssa-strlen.c: Include tree-cfg.h.
22580 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22581 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22582 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22584 (strlen_dom_walker::before_dom_children): Call
22585 gimple_purge_dead_eh_edges. Dump tranformation with details
22587 (strlen_dom_walker::before_dom_children): Update call by adding
22588 new argument cleanup_eh.
22589 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22591 2018-01-03 Martin Liska <mliska@suse.cz>
22594 * cif-code.def (VARIADIC_THUNK): New enum value.
22595 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22598 2018-01-03 Jan Beulich <jbeulich@suse.com>
22600 * sse.md (mov<mode>_internal): Tighten condition for when to use
22601 vmovdqu<ssescalarsize> for TI and OI modes.
22603 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22605 Update copyright years.
22607 2018-01-03 Martin Liska <mliska@suse.cz>
22610 * ipa-visibility.c (function_and_variable_visibility): Skip
22611 functions with noipa attribure.
22613 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22615 * gcc.c (process_command): Update copyright notice dates.
22616 * gcov-dump.c (print_version): Ditto.
22617 * gcov.c (print_version): Ditto.
22618 * gcov-tool.c (print_version): Ditto.
22619 * gengtype.c (create_file): Ditto.
22620 * doc/cpp.texi: Bump @copying's copyright year.
22621 * doc/cppinternals.texi: Ditto.
22622 * doc/gcc.texi: Ditto.
22623 * doc/gccint.texi: Ditto.
22624 * doc/gcov.texi: Ditto.
22625 * doc/install.texi: Ditto.
22626 * doc/invoke.texi: Ditto.
22628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22630 * vector-builder.h (vector_builder::m_full_nelts): Change from
22631 unsigned int to poly_uint64.
22632 (vector_builder::full_nelts): Update prototype accordingly.
22633 (vector_builder::new_vector): Likewise.
22634 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22635 (vector_builder::operator ==): Likewise.
22636 (vector_builder::finalize): Likewise.
22637 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22638 Take the number of elements as a poly_uint64 rather than an
22640 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22641 from unsigned int to poly_uint64.
22642 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22643 (vec_perm_indices::new_vector): Likewise.
22644 (vec_perm_indices::length): Likewise.
22645 (vec_perm_indices::nelts_per_input): Likewise.
22646 (vec_perm_indices::input_nelts): Likewise.
22647 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22648 number of elements per input as a poly_uint64 rather than an
22649 unsigned int. Use the original encoding for variable-length
22650 vectors, rather than clamping each individual element.
22651 For the second and subsequent elements in each pattern,
22652 clamp the step and base before clamping their sum.
22653 (vec_perm_indices::series_p): Handle polynomial element counts.
22654 (vec_perm_indices::all_in_range_p): Likewise.
22655 (vec_perm_indices_to_tree): Likewise.
22656 (vec_perm_indices_to_rtx): Likewise.
22657 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22658 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22659 (tree_vector_builder::new_binary_operation): Handle polynomial
22660 element counts. Return false if we need to know the number
22661 of elements at compile time.
22662 * fold-const.c (fold_vec_perm): Punt if the number of elements
22663 isn't known at compile time.
22665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22667 * vec-perm-indices.h (vec_perm_builder): Change element type
22668 from HOST_WIDE_INT to poly_int64.
22669 (vec_perm_indices::element_type): Update accordingly.
22670 (vec_perm_indices::clamp): Handle polynomial element_types.
22671 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22672 (vec_perm_indices::all_in_range_p): Likewise.
22673 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22675 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22676 polynomial vec_perm_indices element types.
22677 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22678 * fold-const.c (fold_vec_perm): Likewise.
22679 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22680 * tree-vect-generic.c (lower_vec_perm): Likewise.
22681 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22682 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22683 element type to HOST_WIDE_INT.
22685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22686 Alan Hayward <alan.hayward@arm.com>
22687 David Sherwood <david.sherwood@arm.com>
22689 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22690 rather than an int. Use plus_constant.
22691 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22692 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22695 Alan Hayward <alan.hayward@arm.com>
22696 David Sherwood <david.sherwood@arm.com>
22698 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22699 a HOST_WIDE_INT to a poly_int64.
22701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22702 Alan Hayward <alan.hayward@arm.com>
22703 David Sherwood <david.sherwood@arm.com>
22705 * calls.c (load_register_parameters): Cope with polynomial
22706 mode sizes. Require a constant size for BLKmode parameters
22707 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22708 forces a parameter to be padded at the lsb end in order to
22709 fill a complete number of words, require the parameter size
22710 to be ordered wrt UNITS_PER_WORD.
22712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22713 Alan Hayward <alan.hayward@arm.com>
22714 David Sherwood <david.sherwood@arm.com>
22716 * reload1.c (spill_stack_slot_width): Change element type
22717 from unsigned int to poly_uint64_pod.
22718 (alter_reg): Treat mode sizes as polynomial.
22720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22721 Alan Hayward <alan.hayward@arm.com>
22722 David Sherwood <david.sherwood@arm.com>
22724 * reload.c (complex_word_subreg_p): New function.
22725 (reload_inner_reg_of_subreg, push_reload): Use it.
22727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22728 Alan Hayward <alan.hayward@arm.com>
22729 David Sherwood <david.sherwood@arm.com>
22731 * lra-constraints.c (process_alt_operands): Reject matched
22732 operands whose sizes aren't ordered.
22733 (match_reload): Refer to this check here.
22735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22736 Alan Hayward <alan.hayward@arm.com>
22737 David Sherwood <david.sherwood@arm.com>
22739 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22740 that the mode size is in the set {1, 2, 4, 8, 16}.
22742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22743 Alan Hayward <alan.hayward@arm.com>
22744 David Sherwood <david.sherwood@arm.com>
22746 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22747 Use plus_constant instead of gen_rtx_PLUS.
22749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22750 Alan Hayward <alan.hayward@arm.com>
22751 David Sherwood <david.sherwood@arm.com>
22753 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22754 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22755 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22756 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22757 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22758 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22759 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22760 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22761 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22762 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22764 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22765 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22766 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22767 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22768 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22769 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22770 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22771 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22772 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22773 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22775 * expr.c (emit_move_resolve_push): Treat the input and result
22776 of PUSH_ROUNDING as a poly_int64.
22777 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22778 (emit_push_insn): Likewise.
22779 * lra-eliminations.c (mark_not_eliminable): Likewise.
22780 * recog.c (push_operand): Likewise.
22781 * reload1.c (elimination_effects): Likewise.
22782 * rtlanal.c (nonzero_bits1): Likewise.
22783 * calls.c (store_one_arg): Likewise. Require the padding to be
22784 known at compile time.
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 * expr.c (emit_single_push_insn_1): 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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22798 rather than an int.
22800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22801 Alan Hayward <alan.hayward@arm.com>
22802 David Sherwood <david.sherwood@arm.com>
22804 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22805 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22806 via stack temporaries. Treat the mode size as polynomial too.
22808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22809 Alan Hayward <alan.hayward@arm.com>
22810 David Sherwood <david.sherwood@arm.com>
22812 * expr.c (expand_expr_real_2): When handling conversions involving
22813 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22814 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22815 as a poly_uint64 too.
22817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22818 Alan Hayward <alan.hayward@arm.com>
22819 David Sherwood <david.sherwood@arm.com>
22821 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22824 Alan Hayward <alan.hayward@arm.com>
22825 David Sherwood <david.sherwood@arm.com>
22827 * combine.c (can_change_dest_mode): Handle polynomial
22828 REGMODE_NATURAL_SIZE.
22829 * expmed.c (store_bit_field_1): Likewise.
22830 * expr.c (store_constructor): Likewise.
22831 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22832 and polynomial REGMODE_NATURAL_SIZE.
22833 (gen_lowpart_common): Likewise.
22834 * reginfo.c (record_subregs_of_mode): Likewise.
22835 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22842 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
22850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22851 Alan Hayward <alan.hayward@arm.com>
22852 David Sherwood <david.sherwood@arm.com>
22854 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22855 in a POINTER_PLUS_EXPR.
22857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22858 Alan Hayward <alan.hayward@arm.com>
22859 David Sherwood <david.sherwood@arm.com>
22861 * omp-simd-clone.c (simd_clone_subparts): New function.
22862 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22863 (ipa_simd_modify_function_body): Likewise.
22865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22866 Alan Hayward <alan.hayward@arm.com>
22867 David Sherwood <david.sherwood@arm.com>
22869 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22870 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22871 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22872 (expand_vector_condition, vector_element): Likewise.
22873 (subparts_gt): New function.
22874 (get_compute_type): Use subparts_gt.
22875 (count_type_subparts): Delete.
22876 (expand_vector_operations_1): Use subparts_gt instead of
22877 count_type_subparts.
22879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22880 Alan Hayward <alan.hayward@arm.com>
22881 David Sherwood <david.sherwood@arm.com>
22883 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22884 (vect_compile_time_alias): ...this new function. Do the calculation
22885 on poly_ints rather than trees.
22886 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22894 (vect_schedule_slp_instance): Likewise.
22896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22897 Alan Hayward <alan.hayward@arm.com>
22898 David Sherwood <david.sherwood@arm.com>
22900 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22901 constant and extern definitions for variable-length vectors.
22902 (vect_get_constant_vectors): Note that the number of units
22903 is known to be constant.
22905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22906 Alan Hayward <alan.hayward@arm.com>
22907 David Sherwood <david.sherwood@arm.com>
22909 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22910 of units as polynomial. Choose between WIDE and NARROW based
22913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22914 Alan Hayward <alan.hayward@arm.com>
22915 David Sherwood <david.sherwood@arm.com>
22917 * tree-vect-stmts.c (simd_clone_subparts): New function.
22918 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22921 Alan Hayward <alan.hayward@arm.com>
22922 David Sherwood <david.sherwood@arm.com>
22924 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22925 vectors as polynomial. Use build_index_vector for
22926 IFN_GOMP_SIMD_LANE.
22928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22929 Alan Hayward <alan.hayward@arm.com>
22930 David Sherwood <david.sherwood@arm.com>
22932 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22933 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22934 for variable-length vectors.
22935 (vectorizable_mask_load_store): Treat the number of units as
22936 polynomial, asserting that it is constant if the condition has
22937 already been enforced.
22938 (vectorizable_store, vectorizable_load): 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-loop.c (vectorizable_live_operation): Treat the number
22945 of units as polynomial. Punt if we can't tell at compile time
22946 which vector contains the final result.
22948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22949 Alan Hayward <alan.hayward@arm.com>
22950 David Sherwood <david.sherwood@arm.com>
22952 * tree-vect-loop.c (vectorizable_induction): Treat the number
22953 of units as polynomial. Punt on SLP inductions. Use an integer
22954 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22955 cast of such a series for variable-length floating-point
22958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22959 Alan Hayward <alan.hayward@arm.com>
22960 David Sherwood <david.sherwood@arm.com>
22962 * tree.h (build_index_vector): Declare.
22963 * tree.c (build_index_vector): New function.
22964 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22965 of units as polynomial, forcibly converting it to a constant if
22966 vectorizable_reduction has already enforced the condition.
22967 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22968 to create a {1,2,3,...} vector.
22969 (vectorizable_reduction): Treat the number of units as polynomial.
22970 Choose vectype_in based on the largest scalar element size rather
22971 than the smallest number of units. Enforce the restrictions
22974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22975 Alan Hayward <alan.hayward@arm.com>
22976 David Sherwood <david.sherwood@arm.com>
22978 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22979 number of units as polynomial.
22981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22982 Alan Hayward <alan.hayward@arm.com>
22983 David Sherwood <david.sherwood@arm.com>
22985 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22986 * target.def (autovectorize_vector_sizes): Return the vector sizes
22987 by pointer, using vector_sizes rather than a bitmask.
22988 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22989 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22990 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22992 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22993 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22994 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22995 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22996 * omp-general.c (omp_max_vf): Likewise.
22997 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22998 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22999 * tree-vect-loop.c (vect_analyze_loop): Likewise.
23000 * tree-vect-slp.c (vect_slp_bb): Likewise.
23001 * doc/tm.texi: Regenerate.
23002 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23004 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
23005 the vector size as a poly_uint64 rather than an unsigned int.
23006 (current_vector_size): Change from an unsigned int to a poly_uint64.
23007 (get_vectype_for_scalar_type): Update accordingly.
23008 * tree.h (build_truth_vector_type): Take the size and number of
23009 units as a poly_uint64 rather than an unsigned int.
23010 (build_vector_type): Add a temporary overload that takes
23011 the number of units as a poly_uint64 rather than an unsigned int.
23012 * tree.c (make_vector_type): Likewise.
23013 (build_truth_vector_type): Take the number of units as a poly_uint64
23014 rather than an unsigned int.
23016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23017 Alan Hayward <alan.hayward@arm.com>
23018 David Sherwood <david.sherwood@arm.com>
23020 * target.def (get_mask_mode): Take the number of units and length
23021 as poly_uint64s rather than unsigned ints.
23022 * targhooks.h (default_get_mask_mode): Update accordingly.
23023 * targhooks.c (default_get_mask_mode): Likewise.
23024 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
23025 * doc/tm.texi: Regenerate.
23027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23028 Alan Hayward <alan.hayward@arm.com>
23029 David Sherwood <david.sherwood@arm.com>
23031 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
23032 * omp-general.c (omp_max_vf): Likewise.
23033 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
23034 (expand_omp_simd): Handle polynomial safelen.
23035 * omp-low.c (omplow_simd_context): Add a default constructor.
23036 (omplow_simd_context::max_vf): Change from int to poly_uint64.
23037 (lower_rec_simd_input_clauses): Update accordingly.
23038 (lower_rec_input_clauses): Likewise.
23040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23041 Alan Hayward <alan.hayward@arm.com>
23042 David Sherwood <david.sherwood@arm.com>
23044 * tree-vectorizer.h (vect_nunits_for_cost): New function.
23045 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
23046 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
23047 (vect_analyze_slp_cost): Likewise.
23048 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
23049 (vect_model_load_cost): Likewise.
23051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23052 Alan Hayward <alan.hayward@arm.com>
23053 David Sherwood <david.sherwood@arm.com>
23055 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
23056 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
23057 from an unsigned int * to a poly_uint64_pod *.
23058 (calculate_unrolling_factor): New function.
23059 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
23061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23062 Alan Hayward <alan.hayward@arm.com>
23063 David Sherwood <david.sherwood@arm.com>
23065 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
23066 from an unsigned int to a poly_uint64.
23067 (_loop_vec_info::slp_unrolling_factor): Likewise.
23068 (_loop_vec_info::vectorization_factor): Change from an int
23070 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
23071 (vect_get_num_vectors): New function.
23072 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
23073 (vect_get_num_copies): Use vect_get_num_vectors.
23074 (vect_analyze_data_ref_dependences): Change max_vf from an int *
23075 to an unsigned int *.
23076 (vect_analyze_data_refs): Change min_vf from an int * to a
23078 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23079 than an unsigned HOST_WIDE_INT.
23080 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
23081 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23082 to an unsigned int *.
23083 (vect_analyze_data_ref_dependences): Likewise.
23084 (vect_compute_data_ref_alignment): Handle polynomial vf.
23085 (vect_enhance_data_refs_alignment): Likewise.
23086 (vect_prune_runtime_alias_test_list): Likewise.
23087 (vect_shift_permute_load_chain): Likewise.
23088 (vect_supportable_dr_alignment): Likewise.
23089 (dependence_distance_ge_vf): Take the vectorization factor as a
23090 poly_uint64 rather than an unsigned HOST_WIDE_INT.
23091 (vect_analyze_data_refs): Change min_vf from an int * to a
23093 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
23094 vfm1 as a poly_uint64 rather than an int. Make the same change
23095 for the returned bound_scalar.
23096 (vect_gen_vector_loop_niters): Handle polynomial vf.
23097 (vect_do_peeling): Likewise. Update call to
23098 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
23099 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
23101 * tree-vect-loop.c (vect_determine_vectorization_factor)
23102 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
23103 (vect_get_known_peeling_cost): Likewise.
23104 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
23105 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
23106 (vect_transform_loop): Likewise. Use the lowest possible VF when
23107 updating the upper bounds of the loop.
23108 (vect_min_worthwhile_factor): Make static. Return an unsigned int
23109 rather than an int.
23110 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
23111 polynomial unroll factors.
23112 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
23113 (vect_make_slp_decision): Likewise.
23114 (vect_supported_load_permutation_p): Likewise, and polynomial
23116 (vect_analyze_slp_cost): Handle polynomial vf.
23117 (vect_slp_analyze_node_operations): Likewise.
23118 (vect_slp_analyze_bb_1): Likewise.
23119 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23120 than an unsigned HOST_WIDE_INT.
23121 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
23122 (vectorizable_load): Handle polynomial vf.
23123 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
23125 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
23127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23128 Alan Hayward <alan.hayward@arm.com>
23129 David Sherwood <david.sherwood@arm.com>
23131 * match.pd: Handle bit operations involving three constants
23132 and try to fold one pair.
23134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23136 * tree-vect-loop-manip.c: Include gimple-fold.h.
23137 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
23138 niters_maybe_zero parameters. Handle other cases besides a step of 1.
23139 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
23140 Add a path that uses a step of VF instead of 1, but disable it
23142 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
23143 and niters_no_overflow parameters. Update calls to
23144 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
23145 Create a new SSA name if the latter choses to use a ste other
23146 than zero, and return it via niters_vector_mult_vf_var.
23147 * tree-vect-loop.c (vect_transform_loop): Update calls to
23148 vect_do_peeling, vect_gen_vector_loop_niters and
23149 slpeel_make_loop_iterate_ntimes.
23150 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
23151 (vect_gen_vector_loop_niters): Update declarations after above changes.
23153 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
23155 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
23156 128-bit round to integer instructions.
23157 (ceil<mode>2): Likewise.
23158 (btrunc<mode>2): Likewise.
23159 (round<mode>2): Likewise.
23161 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23163 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
23164 unaligned VSX load/store on P8/P9.
23165 (expand_block_clear): Allow the use of unaligned VSX
23166 load/store on P8/P9.
23168 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23170 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
23172 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
23173 swap associated with both a load and a store.
23175 2018-01-02 Andrew Waterman <andrew@sifive.com>
23177 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
23178 * config/riscv/riscv.md (clear_cache): Use it.
23180 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
23182 * web.c: Remove out-of-date comment.
23184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23186 * expr.c (fixup_args_size_notes): Check that any existing
23187 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
23188 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
23189 (emit_single_push_insn): ...here.
23191 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23193 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
23194 (const_vector_encoded_nelts): New function.
23195 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
23196 (const_vector_int_elt, const_vector_elt): Declare.
23197 * emit-rtl.c (const_vector_int_elt_1): New function.
23198 (const_vector_elt): Likewise.
23199 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
23200 of CONST_VECTOR_ELT.
23202 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23204 * expr.c: Include rtx-vector-builder.h.
23205 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
23206 directly on the tree encoding.
23207 (const_vector_from_tree): Likewise.
23208 * optabs.c: Include rtx-vector-builder.h.
23209 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
23210 sequence of "u" values.
23211 * vec-perm-indices.c: Include rtx-vector-builder.h.
23212 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
23213 directly on the vec_perm_indices encoding.
23215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23217 * doc/rtl.texi (const_vector): Describe new encoding scheme.
23218 * Makefile.in (OBJS): Add rtx-vector-builder.o.
23219 * rtx-vector-builder.h: New file.
23220 * rtx-vector-builder.c: Likewise.
23221 * rtl.h (rtx_def::u2): Add a const_vector field.
23222 (CONST_VECTOR_NPATTERNS): New macro.
23223 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
23224 (CONST_VECTOR_DUPLICATE_P): Likewise.
23225 (CONST_VECTOR_STEPPED_P): Likewise.
23226 (CONST_VECTOR_ENCODED_ELT): Likewise.
23227 (const_vec_duplicate_p): Check for a duplicated vector encoding.
23228 (unwrap_const_vec_duplicate): Likewise.
23229 (const_vec_series_p): Check for a non-duplicated vector encoding.
23230 Say that the function only returns true for integer vectors.
23231 * emit-rtl.c: Include rtx-vector-builder.h.
23232 (gen_const_vec_duplicate_1): Delete.
23233 (gen_const_vector): Call gen_const_vec_duplicate instead of
23234 gen_const_vec_duplicate_1.
23235 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
23236 (gen_const_vec_duplicate): Use rtx_vector_builder.
23237 (gen_const_vec_series): Likewise.
23238 (gen_rtx_CONST_VECTOR): Likewise.
23239 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
23240 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23241 Build a new vector rather than modifying a CONST_VECTOR in-place.
23242 (handle_special_swappables): Update call accordingly.
23243 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
23244 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23245 Build a new vector rather than modifying a CONST_VECTOR in-place.
23246 (handle_special_swappables): Update call accordingly.
23248 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23250 * simplify-rtx.c (simplify_const_binary_operation): Use
23251 CONST_VECTOR_ELT instead of XVECEXP.
23253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23255 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
23256 the selector elements to be different from the data elements
23257 if the selector is a VECTOR_CST.
23258 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
23259 ssizetype for the selector.
23261 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23263 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23264 before testing each element individually.
23265 * tree-vect-generic.c (lower_vec_perm): Likewise.
23267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23269 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23270 * selftest-run-tests.c (selftest::run_tests): Call it.
23271 * vector-builder.h (vector_builder::operator ==): New function.
23272 (vector_builder::operator !=): Likewise.
23273 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23274 (vec_perm_indices::all_from_input_p): New function.
23275 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23276 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23277 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23278 instead of reading the VECTOR_CST directly. Detect whether both
23279 vector inputs are the same before constructing the vec_perm_indices,
23280 and update the number of inputs argument accordingly. Use the
23281 utility functions added above. Only construct sel2 if we need to.
23283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23285 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23286 the broadcast of the low byte.
23287 (expand_mult_highpart): Use an explicit encoding for the permutes.
23288 * optabs-query.c (can_mult_highpart_p): Likewise.
23289 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23290 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23291 (vectorizable_bswap): Likewise.
23292 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23293 explicit encoding for the power-of-2 permutes.
23294 (vect_permute_store_chain): Likewise.
23295 (vect_grouped_load_supported): Likewise.
23296 (vect_permute_load_chain): Likewise.
23298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23300 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23301 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23302 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23303 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23304 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23305 (vect_gen_perm_mask_any): Likewise.
23307 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23309 * int-vector-builder.h: New file.
23310 * vec-perm-indices.h: Include int-vector-builder.h.
23311 (vec_perm_indices): Redefine as an int_vector_builder.
23312 (auto_vec_perm_indices): Delete.
23313 (vec_perm_builder): Redefine as a stand-alone class.
23314 (vec_perm_indices::vec_perm_indices): New function.
23315 (vec_perm_indices::clamp): Likewise.
23316 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23317 (vec_perm_indices::new_vector): New function.
23318 (vec_perm_indices::new_expanded_vector): Update for new
23319 vec_perm_indices class.
23320 (vec_perm_indices::rotate_inputs): New function.
23321 (vec_perm_indices::all_in_range_p): Operate directly on the
23322 encoded form, without computing elided elements.
23323 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23324 encoding. Update for new vec_perm_indices class.
23325 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23326 the given vec_perm_builder.
23327 (expand_vec_perm_var): Update vec_perm_builder constructor.
23328 (expand_mult_highpart): Use vec_perm_builder instead of
23329 auto_vec_perm_indices.
23330 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23331 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23332 or double series encoding as appropriate.
23333 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23334 vec_perm_indices instead of auto_vec_perm_indices.
23335 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23336 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23337 (vect_permute_store_chain): Likewise.
23338 (vect_grouped_load_supported): Likewise.
23339 (vect_permute_load_chain): Likewise.
23340 (vect_shift_permute_load_chain): Likewise.
23341 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23342 (vect_transform_slp_perm_load): Likewise.
23343 (vect_schedule_slp_instance): Likewise.
23344 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23345 (vectorizable_mask_load_store): Likewise.
23346 (vectorizable_bswap): Likewise.
23347 (vectorizable_store): Likewise.
23348 (vectorizable_load): Likewise.
23349 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23350 vec_perm_indices instead of auto_vec_perm_indices. Use
23351 tree_to_vec_perm_builder to read the vector from a tree.
23352 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23353 vec_perm_builder instead of a vec_perm_indices.
23354 (have_whole_vector_shift): Use vec_perm_builder and
23355 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23356 truncation to calc_vec_perm_mask_for_shift.
23357 (vect_create_epilog_for_reduction): Likewise.
23358 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23359 from auto_vec_perm_indices to vec_perm_indices.
23360 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23361 instead of changing individual elements.
23362 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23363 the vector in d.perm.
23364 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23365 from auto_vec_perm_indices to vec_perm_indices.
23366 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23367 instead of changing individual elements.
23368 (arm_vectorize_vec_perm_const): Use new_vector to install
23369 the vector in d.perm.
23370 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23371 Update vec_perm_builder constructor.
23372 (rs6000_expand_interleave): Likewise.
23373 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23374 (rs6000_expand_interleave): Likewise.
23376 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23378 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23379 to qimode could truncate the indices.
23380 * optabs.c (expand_vec_perm_var): Likewise.
23382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23384 * Makefile.in (OBJS): Add vec-perm-indices.o.
23385 * vec-perm-indices.h: New file.
23386 * vec-perm-indices.c: Likewise.
23387 * target.h (vec_perm_indices): Replace with a forward class
23389 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23390 * optabs.h: Include vec-perm-indices.h.
23391 (expand_vec_perm): Delete.
23392 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23393 (expand_vec_perm_const): Declare.
23394 * target.def (vec_perm_const_ok): Replace with...
23395 (vec_perm_const): ...this new hook.
23396 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23397 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23398 * doc/tm.texi: Regenerate.
23399 * optabs.def (vec_perm_const): Delete.
23400 * doc/md.texi (vec_perm_const): Likewise.
23401 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23402 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23403 expand_vec_perm for constant permutation vectors. Assert that
23404 the mode of variable permutation vectors is the integer equivalent
23405 of the mode that is being permuted.
23406 * optabs-query.h (selector_fits_mode_p): Declare.
23407 * optabs-query.c: Include vec-perm-indices.h.
23408 (selector_fits_mode_p): New function.
23409 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23410 is defined, instead of checking whether the vec_perm_const_optab
23411 exists. Use targetm.vectorize.vec_perm_const instead of
23412 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23413 fit in the vector mode before using a variable permute.
23414 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23415 vec_perm_indices instead of an rtx.
23416 (expand_vec_perm): Replace with...
23417 (expand_vec_perm_const): ...this new function. Take the selector
23418 as a vec_perm_indices rather than an rtx. Also take the mode of
23419 the selector. Update call to shift_amt_for_vec_perm_mask.
23420 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23421 Use vec_perm_indices::new_expanded_vector to expand the original
23422 selector into bytes. Check whether the indices fit in the vector
23423 mode before using a variable permute.
23424 (expand_vec_perm_var): Make global.
23425 (expand_mult_highpart): Use expand_vec_perm_const.
23426 * fold-const.c: Includes vec-perm-indices.h.
23427 * tree-ssa-forwprop.c: Likewise.
23428 * tree-vect-data-refs.c: Likewise.
23429 * tree-vect-generic.c: Likewise.
23430 * tree-vect-loop.c: Likewise.
23431 * tree-vect-slp.c: Likewise.
23432 * tree-vect-stmts.c: Likewise.
23433 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23435 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23436 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23437 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23438 (aarch64_vectorize_vec_perm_const): ...this new function.
23439 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23440 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23441 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23442 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23443 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23444 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23445 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23447 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23448 check for NEON modes.
23449 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23450 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23451 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23452 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23454 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23455 the old VEC_PERM_CONST conditions.
23456 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23457 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23458 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23459 (ia64_vectorize_vec_perm_const_ok): Merge into...
23460 (ia64_vectorize_vec_perm_const): ...this new function.
23461 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23462 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23463 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23464 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23465 * config/mips/mips.c (mips_expand_vec_perm_const)
23466 (mips_vectorize_vec_perm_const_ok): Merge into...
23467 (mips_vectorize_vec_perm_const): ...this new function.
23468 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23469 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23470 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23471 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23472 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23473 (rs6000_expand_vec_perm_const): Delete.
23474 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23476 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23477 (altivec_expand_vec_perm_const_le): Take each operand individually.
23478 Operate on constant selectors rather than rtxes.
23479 (altivec_expand_vec_perm_const): Likewise. Update call to
23480 altivec_expand_vec_perm_const_le.
23481 (rs6000_expand_vec_perm_const): Delete.
23482 (rs6000_vectorize_vec_perm_const_ok): Delete.
23483 (rs6000_vectorize_vec_perm_const): New function.
23484 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23485 an element count and rtx array.
23486 (rs6000_expand_extract_even): Update call accordingly.
23487 (rs6000_expand_interleave): Likewise.
23488 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23489 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23490 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23491 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23492 (rs6000_expand_vec_perm_const): Delete.
23493 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23494 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23495 (altivec_expand_vec_perm_const_le): Take each operand individually.
23496 Operate on constant selectors rather than rtxes.
23497 (altivec_expand_vec_perm_const): Likewise. Update call to
23498 altivec_expand_vec_perm_const_le.
23499 (rs6000_expand_vec_perm_const): Delete.
23500 (rs6000_vectorize_vec_perm_const_ok): Delete.
23501 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23502 reference to the SPE evmerge intructions.
23503 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23504 an element count and rtx array.
23505 (rs6000_expand_extract_even): Update call accordingly.
23506 (rs6000_expand_interleave): Likewise.
23507 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23508 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23510 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23514 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23515 vector mode and that that mode matches the mode of the data
23517 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23518 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23519 directly using expand_vec_perm_1 when forcing selectors into
23521 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23525 * optabs-query.h (can_vec_perm_p): Delete.
23526 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23527 * optabs-query.c (can_vec_perm_p): Split into...
23528 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23529 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23530 particular selector is valid.
23531 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23532 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23533 (vect_grouped_load_supported): Likewise.
23534 (vect_shift_permute_load_chain): Likewise.
23535 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23536 (vect_transform_slp_perm_load): Likewise.
23537 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23538 (vectorizable_bswap): Likewise.
23539 (vect_gen_perm_mask_checked): Likewise.
23540 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23541 implementations of variable permutation vectors into account
23542 when deciding which selector to use.
23543 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23544 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23545 with a false third argument.
23546 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23547 to test whether the constant selector is valid and can_vec_perm_var_p
23548 to test whether a variable selector is valid.
23550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23552 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23553 * optabs-query.c (can_vec_perm_p): Likewise.
23554 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23555 instead of vec_perm_indices.
23556 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23557 (vect_gen_perm_mask_checked): Likewise,
23558 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23559 (vect_gen_perm_mask_checked): Likewise,
23561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23563 * optabs-query.h (qimode_for_vec_perm): Declare.
23564 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23565 (qimode_for_vec_perm): ...this new function.
23566 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23568 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23570 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23571 does not have a conditional at the top.
23573 2018-01-02 Richard Biener <rguenther@suse.de>
23575 * ipa-inline.c (big_speedup_p): Fix expression.
23577 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23580 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23583 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23587 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23588 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23589 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23590 cond_taken_branch_cost 3->4.
23592 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23594 PR tree-optimization/83581
23595 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23596 TODO_cleanup_cfg if any changes have been made.
23598 PR middle-end/83608
23599 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23600 convert_modes if target mode has the right side, but different mode
23603 PR middle-end/83609
23604 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23605 last argument when extracting from CONCAT. If either from_real or
23606 from_imag is NULL, use expansion through memory. If result is not
23607 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23608 the parts directly to inner mode, if even that fails, use expansion
23611 PR middle-end/83623
23612 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23613 check for bswap in mode rather than HImode and use that in expand_unop
23616 Copyright (C) 2018 Free Software Foundation, Inc.
23618 Copying and distribution of this file, with or without modification,
23619 are permitted in any medium without royalty provided the copyright
23620 notice and this notice are preserved.