1 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5 register_operand when calling ix86_set_reg_reg_cost.
6 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
7 Set *total to 0 for operands that satisfy x86_64_immediate_operand
8 predicate and to 1 otherwise.
10 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
12 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
13 emit a diagnostic that it is not supported and reset the option.
14 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
15 supported and consume the option. (ASM_FINAL_SPEC): New.
17 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
19 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
22 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
24 C-SKY port: Documentation
26 * doc/extend.texi (C-SKY Function Attributes): New section.
27 * doc/invoke.texi (Option Summary): Add C-SKY options.
28 (C-SKY Options): New section.
29 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
31 2018-08-17 Jojo <jijie_rong@c-sky.com>
32 Huibin Wang <huibin_wang@c-sky.com>
33 Sandra Loosemore <sandra@codesourcery.com>
34 Chung-Lin Tang <cltang@codesourcery.com>
36 C-SKY port: Backend implementation
39 * common/config/csky/*: New.
41 2018-08-17 Jojo <jijie_rong@c-sky.com>
42 Huibin Wang <huibin_wang@c-sky.com>
43 Sandra Loosemore <sandra@codesourcery.com>
44 Chung-Lin Tang <cltang@codesourcery.com>
45 Andrew Jenner <andrew@codesourcery.com>
49 * config.gcc (csky-*-*): New.
50 * configure.ac: Add csky to targets for dwarf2 debug_line support.
51 * configure: Regenerated.
53 2018-08-17 David Malcolm <dmalcolm@redhat.com>
55 * dump-context.h: Include "dumpfile.h".
56 (dump_context::dump_printf_va): Convert final param from va_list
57 to va_list *. Convert from ATTRIBUTE_PRINTF to
58 ATTRIBUTE_GCC_DUMP_PRINTF.
59 (dump_context::dump_printf_loc_va): Likewise.
60 * dumpfile.c: Include "stringpool.h".
61 (make_item_for_dump_printf_va): Delete.
62 (make_item_for_dump_printf): Delete.
63 (class dump_pretty_printer): New class.
64 (dump_pretty_printer::dump_pretty_printer): New ctor.
65 (dump_pretty_printer::emit_items): New member function.
66 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
68 (dump_pretty_printer::emit_item): New member function.
69 (dump_pretty_printer::stash_item): New member function.
70 (dump_pretty_printer::format_decoder_cb): New member function.
71 (dump_pretty_printer::decode_format): New member function.
72 (dump_context::dump_printf_va): Reimplement in terms of
74 (dump_context::dump_printf_loc_va): Convert final param from va_list
76 (dump_context::begin_scope): Reimplement call to
77 make_item_for_dump_printf.
78 (dump_printf): Update for change to dump_printf_va.
79 (dump_printf_loc): Likewise.
80 (selftest::test_capture_of_dump_calls): Convert "stmt" from
81 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
83 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
84 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
85 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
86 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
87 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
88 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
89 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
90 within a dump_printf_loc call to "%wu".
91 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
92 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
93 missing space after "=".
94 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
95 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
96 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
97 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
98 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
99 duplicate "vectorized" from message.
101 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
103 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
104 polyNxK_t element's TYPE_STRING_FLAG.
106 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
108 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
109 (they were unnamed before). Fix comments.
111 2018-08-17 Nathan Sidwell <nathan@acm.org>
113 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
115 2018-08-17 Richard Biener <rguenther@suse.de>
117 PR tree-optimization/86841
118 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
120 2018-08-17 Martin Liska <mliska@suse.cz>
122 * common.opt: Remove Warn, Init and Report for options with
123 Ignore/Deprecated flag. Warning is done automatically for
125 * config/i386/i386.opt: Likewise.
126 * config/ia64/ia64.opt: Likewise.
127 * config/rs6000/rs6000.opt: Likewise.
128 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
129 Remove usage of flag_check_pointer_bounds.
130 * lto-wrapper.c (merge_and_complain): Do not handle
131 OPT_fcheck_pointer_bounds.
132 (append_compiler_options): Likewise.
133 * opt-functions.awk: Do not handle Deprecated.
134 * optc-gen.awk: Check that Var, Report and Init are not
135 used for an option with Ignore/Deprecated flag.
136 * opts-common.c (decode_cmdline_option): Do not report
138 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
140 * opts.h (struct cl_option): Remove cl_deprecated flag.
141 (CL_ERR_DEPRECATED): Remove error enum value.
143 2018-08-17 Richard Biener <rguenther@suse.de>
146 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
147 across a va-arg-pack using call adjust its return value accordingly.
149 2018-08-16 Martin Sebor <msebor@redhat.com>
151 PR tree-optimization/86853
152 * gimple-ssa-sprintf.c (struct format_result): Rename member.
153 (struct fmtresult): Add member and initialize it in ctors.
154 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
155 (format_string): Handle %S the same as %ls. Set MAYFAIL.
156 (format_directive): Set POSUNDER4K when MAYFAIL is set.
157 (parse_directive): Handle %C same as %c.
158 (sprintf_dom_walker::compute_format_length): Adjust.
159 (is_call_safe): Adjust.
161 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
163 * builtins.c (c_strlen): Add new parameter eltsize. Use it
164 for determining how to count the elements.
165 * builtins.h (c_strlen): Adjust prototype.
166 * expr.c (string_constant): Add new parameter mem_size.
167 Set *mem_size appropriately.
168 * expr.h (string_constant): Adjust protoype.
169 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
170 * gimple-fold.h (get_range_strlen): Adjust prototype.
171 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
172 (format_string): Call get_string_length with eltsize.
174 2018-08-16 David Malcolm <dmalcolm@redhat.com>
176 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
177 to emit the span, rather than setting it as the prefix.
179 2018-08-16 David Malcolm <dmalcolm@redhat.com>
181 * diagnostic-show-locus.c (layout::start_annotation_line): Add
182 "margin_char" parameter, defaulting to space. Use it in place
183 of pp_space for the initial part of the margin.
184 (layout::print_leading_fixits): Use '+' when filling the margin
185 of line-insertion fix-it hints.
187 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
189 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
192 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
194 * config/rs6000/altivec.md: Don't set length attribute to the default
196 * config/rs6000/darwin.md: Ditto.
197 * config/rs6000/dfp.md: Ditto.
198 * config/rs6000/htm.md: Ditto.
199 * config/rs6000/rs6000.md: Ditto.
200 * config/rs6000/sync.md: Ditto.
201 * config/rs6000/vsx.md: Ditto.
203 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
205 * config/rs6000/altivec.md: Don't set length attribute to the default
206 value, for branch instructions.
207 * config/rs6000/darwin.md: Ditto.
208 * config/rs6000/rs6000.md: Ditto.
210 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
212 * config/rs6000/rs6000.md (length): Always define as const_int 4.
213 (unnamed conditional branch define_insn): Set length to 4 or 8
215 (<bd>_<mode>): Similar, for alternative 0.
216 (<bd>tf_<mode>): Ditto.
218 2018-08-16 Tamar Christina <tamar.christina@arm.com>
220 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
222 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
224 * doc/rtl.texi: Replace old RTX class names with new names.
227 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
229 * expmed.h (canonicalize_comparison): New declaration.
230 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
231 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
232 * optabs.c (prepare_cmp_insn): Likewise.
233 * rtl.h (unsigned_condition_p): New function which checks if a
234 comparison operator is unsigned.
236 2018-08-16 Nathan Sidwell <nathan@acm.org>
238 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
239 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
241 2018-08-16 Tamar Christina <tamar.christina@arm.com>
244 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
245 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
246 (mov<mov>): ..this and enable unconditionally.
248 2018-08-16 Tamar Christina <tamar.christina@arm.com>
250 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
252 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
254 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
255 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
256 "Common" with "Target".
258 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
260 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
261 * doc/invoke.texi (mmitigate-rop): Remove.
262 * config/i386/i386.c: Do not include "regrename.h".
263 (ix86_rop_should_change_byte_p, reg_encoded_number)
264 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
266 (ix86_reorg): Remove call to ix86_mitigate_rop.
267 * config/i386/i386.md (attr "modrm_class"): Remove.
268 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
269 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
270 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
272 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
274 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
275 allow folding of mergeh() and mergel() for the float and double types.
276 (fold_mergehl_helper): Rework to handle building a permute tree
279 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
281 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
284 2018-08-15 David Malcolm <dmalcolm@redhat.com>
286 * common.opt (fdiagnostics-show-labels): New option.
287 * diagnostic-show-locus.c (class layout_range): Add field
289 (class layout): Add field "m_show_labels_p".
290 (layout_range::layout_range): Add param "label" and use it to
292 (make_range): Pass in NULL for new "label" param of layout_range's
294 (layout::layout): Initialize m_show_labels_p.
295 (layout::maybe_add_location_range): Pass in loc_range->m_label
296 when constructing layout_range instances.
297 (struct line_label): New struct.
298 (layout::print_any_labels): New member function.
299 (layout::print_line): Call it if label-printing is enabled.
300 (selftest::test_one_liner_labels): New test.
301 (selftest::test_diagnostic_show_locus_one_liner): Call it.
302 * diagnostic.c (diagnostic_initialize): Initialize
303 context->show_labels_p.
304 * diagnostic.h (struct diagnostic_context): Add field
306 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
307 -fno-diagnostics-show-labels.
308 * dwarf2out.c (gen_producer_string): Add
309 OPT_fdiagnostics_show_labels to the ignored options.
310 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
312 (gcc_rich_location::maybe_add_expr): Likewise.
313 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
314 label" param, defaulting to NULL.
315 (gcc_rich_location::add_expr): Add "label" param.
316 (gcc_rich_location::maybe_add_expr): Likewise.
317 (class text_range_label): New class.
318 (class range_label_for_type_mismatch): New class.
319 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
320 of format_warning_va.
321 (fmtwarn_n): Likewise for new params of format_warning_n_va.
322 * lto-wrapper.c (merge_and_complain): Add
323 OPT_fdiagnostics_show_labels to the "pick one setting" options.
324 (append_compiler_options): Likewise to the dropped options.
325 (append_diag_options): Likewise to the passed-on options.
326 * opts.c (common_handle_option): Handle the new option.
327 * selftest-diagnostic.c
328 (test_diagnostic_context::test_diagnostic_context): Enable
330 * substring-locations.c: Include "gcc-rich-location.h".
331 (format_warning_n_va): Add "fmt_label" and "param_label" params
332 and use them as appropriate.
333 (format_warning_va): Add "fmt_label" and "param_label" params,
334 passing them on to format_warning_n_va.
335 (format_warning_at_substring): Likewise.
336 (format_warning_at_substring_n): Likewise.
337 * substring-locations.h (format_warning_va): Add "fmt_label" and
338 "param_label" params.
339 (format_warning_n_va): Likewise.
340 (format_warning_at_substring): Likewise.
341 (format_warning_at_substring_n): Likewise.
342 * toplev.c (general_init): Initialize global_dc->show_labels_p.
344 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
347 * builtins.c (expand_builtin_memcmp): Do not expand the call
348 when overflow is detected.
350 2018-08-15 Martin Sebor <msebor@redhat.com>
352 PR tree-optimization/71625
353 * config/aarch64/aarch64-builtins.c
354 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
356 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
358 * config/s390/s390.c (s390_reorg): Remove loop.
360 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
363 (darwin_function_switched_text_sections): Delete.
364 * gcc/config/darwin.h
365 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
367 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
370 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
371 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
372 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
374 2018-08-15 Martin Liska <mliska@suse.cz>
376 PR tree-optimization/86925
377 * predict.c (expr_expected_value_1): When taking
378 later predictor, assign also probability.
379 Use fold_build2_initializer_loc in order to fold
380 the expression in -frounding-math.
382 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
384 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
386 (expand_vec_perm_1): Try the new method.
388 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
391 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
392 Check whether lra_live_max_point is 0 before dividing.
394 2018-08-14 Martin Sebor <msebor@redhat.com>
396 PR tree-optimization/86650
397 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
398 (vrp_prop::check_mem_ref): Same.
400 2018-08-13 Liu Hao <lh_mouse@126.com>
402 * pretty-print.c (eat_esc_sequence): Swap the foreground and
403 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
404 and clear it thereafter, as it only works for DBCS.
406 2018-08-13 Liu Hao <lh_mouse@126.com>
408 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
409 handle returned by _get_osf_handle().
411 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
413 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
414 for folding vec_perm.
416 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
418 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
419 Add support for gimple-folding of vec_pack() and vec_unpack()
422 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
424 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
425 vec_xst variants to the list.
426 (rs6000_gimple_fold_builtin): Add support for folding unaligned
427 vector loads and stores.
429 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
431 * config.gcc (rs6000-ibm-aix4.x): Delete.
432 (rs6000-ibm-aix5.1): Delete.
433 (rs6000-ibm-aix5.2): Delete.
434 (rs6000-ibm-aix5.3): Delete.
435 * config/rs6000/aix43.h: Delete.
436 * config/rs6000/aix51.h: Delete.
437 * config/rs6000/aix52.h: Delete.
438 * config/rs6000/t-aix43: Delete.
440 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
442 * config/s390/s390.c (s390_decompose_constant_pool_ref):
444 (s390_decompose_address): Factor out constant pool ref
447 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
449 * config/nds32/nds32-predicates.c
450 (nds32_can_use_bclr_p): Change return type as bool.
451 (nds32_can_use_bset_p): Ditto.
452 (nds32_can_use_btgl_p): Ditto.
453 (nds32_can_use_bitci_p): Ditto.
454 * config/nds32/nds32-protos.h
455 (nds32_can_use_bclr_p): Change declaration.
456 (nds32_can_use_bset_p): Ditto.
457 (nds32_can_use_btgl_p): Ditto.
458 (nds32_can_use_bitci_p): Ditto.
460 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
462 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
463 Support -msched-prolog-epilog option.
464 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
466 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
468 * common/config/nds32/nds32-common.c
469 (nds32_option_optimization_table): Enalbe -malways-align.
471 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
473 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
475 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
476 OPT_misr_secure_ case.
477 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
478 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
479 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
481 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
482 (nds32_isr_info): New field security_level.
483 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
484 * config/nds32/nds32.md (return_internal): Consider critical attribute.
485 * config/nds32/nds32.opt (misr-secure): New option.
486 * config/nds32/nds32_init.inc: New file.
487 * config/nds32/nds32_isr.h: New file.
489 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
491 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
492 Update comment for atomic instructions.
493 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
494 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
496 (atomic_loaddi): Revise fence expansion to only emit fence prior to
497 load for __ATOMIC_SEQ_CST model.
498 (atomic_loaddi_1): Remove float register target.
499 (atomic_storedi): Handle CONST_INT values.
500 (atomic_storedi_1): Remove float register source. Add special case
502 (memory_barrier): New expander and insn.
504 2018-08-11 Jakub Jelinek <jakub@redhat.com>
506 PR tree-optimization/86835
507 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
508 new_stmt after def_gsi, make sure to insert new_square_stmt after
509 that stmt, not 2 stmts before it.
511 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
514 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
517 2018-08-10 Martin Liska <mliska@suse.cz>
520 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
522 * builtins.c (expand_builtin_expect_with_probability):
524 (expand_builtin_expect_with_probability): New function.
525 (build_builtin_expect_predicate): Add new argumnet probability
526 for BUILT_IN_EXPECT_WITH_PROBABILITY.
527 (fold_builtin_expect):
530 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
531 * builtins.h (fold_builtin_expect): Set new argument.
532 * doc/extend.texi: Document __builtin_expect_with_probability.
533 * doc/invoke.texi: Likewise.
534 * gimple-fold.c (gimple_fold_call): Pass new argument.
535 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
536 also BUILT_IN_EXPECT_WITH_PROBABILITY.
537 * predict.c (get_predictor_value): New function.
538 (expr_expected_value): Add new argument probability. Assume
539 that predictor and probability are always non-null.
540 (expr_expected_value_1): Likewise. For __builtin_expect and
541 __builtin_expect_with_probability set probability. Handle
542 combination in binary expressions.
543 (tree_predict_by_opcode): Simplify code by simply calling
545 (pass_strip_predict_hints::execute): Add handling of
546 BUILT_IN_EXPECT_WITH_PROBABILITY.
547 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
549 * tree.h (DECL_BUILT_IN_P): New function.
551 2018-08-10 Martin Liska <mliska@suse.cz>
553 PR tree-optimization/85799
554 * passes.def: Add argument for pass_strip_predict_hints.
555 * predict.c (class pass_strip_predict_hints): Add new argument
557 (strip_predictor_early): New function.
558 (pass_strip_predict_hints::execute): Call the function to
560 (strip_predict_hints): New function.
561 * predict.def: Fix comment.
563 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
565 * Makefile.in: Clarify which tm.texi to copy over to assert the
566 right to grant a GFDL license for all.
568 2018-08-09 Jeff Law <law@redhat.com>
570 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
573 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
575 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
578 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
580 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
581 reductions for variable-length vectors.
583 2018-08-09 David Malcolm <dmalcolm@redhat.com>
586 * common.opt (fdiagnostics-show-line-numbers): New option.
587 * diagnostic-show-locus.c (class layout): Add fields
588 "m_show_line_numbers_p" and "m_linenum_width";
589 (num_digits): New function.
590 (test_num_digits): New function.
591 (layout::layout): Initialize new fields. Update m_x_offset
592 logic to handle any left margin.
593 (layout::print_source_line): Print line number when requested.
594 (layout::start_annotation_line): New member function.
595 (layout::print_annotation_line): Call it.
596 (layout::print_leading_fixits): Likewise.
597 (layout::print_trailing_fixits): Likewise. Update calls to
598 move_to_column for new parameter.
599 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
600 it to potentially call start_annotation_line.
601 (layout::show_ruler): Call start_annotation_line.
602 (selftest::test_line_numbers_multiline_range): New selftest.
603 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
604 and selftest::test_line_numbers_multiline_range.
605 * diagnostic.c (diagnostic_initialize): Initialize
607 * diagnostic.h (struct diagnostic_context): Add field
608 "show_line_numbers_p".
609 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
610 -fno-diagnostics-show-line-numbers.
611 * dwarf2out.c (gen_producer_string): Add
612 OPT_fdiagnostics_show_line_numbers to the ignored options.
613 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
614 one setting" options.
615 (append_compiler_options): Likewise to the dropped options.
616 (append_diag_options): Likewise to the passed-on options.
617 * opts.c (common_handle_option): Handle the new option.
618 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
620 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
622 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
623 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
624 third argument of type "const signed char" to descriptions of
625 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
626 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
627 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
628 __builtin_bcdsub_ov functions.
630 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
632 PR tree-optimization/86858
633 * tree-vect-loop.c (vect_is_simple_reduction): Restore
634 flow_bb_inside_loop_p calls.
636 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
638 PR tree-optimization/86871
639 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
640 instead of gimple_assign_lhs.
642 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
645 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
646 register constraint to operand 0.
647 (add<mode>3_carryinC): Likewise.
648 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
650 2018-08-09 Martin Liska <mliska@suse.cz>
653 * common.opt: Remove extra line.
655 2018-08-09 Martin Liska <mliska@suse.cz>
657 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
658 at the end of a line, make first letter capital and end up
659 a sentence with a dot.
660 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
661 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
662 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
663 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
664 (PARAM_MAX_ISL_OPERATIONS): Likewise.
665 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
666 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
667 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
668 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
669 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
670 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
671 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
672 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
673 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
674 (PARAM_TREE_REASSOC_WIDTH): Likewise.
675 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
676 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
677 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
679 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
682 * config/s390/s390.c (s390_option_override_internal): Reduce the
683 stack-clash-protection-probe-interval param if it would be too big
686 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
689 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
690 * config/m68k/m68k.c (handle_move_double): Don't call
691 m68k_final_prescan_insn.
692 (m68k_adjust_decorated_operand): Renamed from
693 m68k_final_prescan_insn, remove first and third operand and
695 (print_operand): Call it.
696 (print_operand_address): Call it.
698 2018-08-08 Nathan Sidwell <nathan@acm.org>
700 * diagnostic.c (diagnostic_report_current_module): Use
701 linemap_included_from & linemap_included_from_linemap.
703 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
705 * config/aarch64/aarch64-cores.def: Add phecda core.
706 * config/aarch64/aarch64-tune.md: Regenerate.
707 * doc/invoke.texi: Add phecda core.
709 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
712 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
714 * config/s390/s390.md ("movti"): Add more alternatives for
715 constant to GPR copies.
717 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
719 * config/s390/s390.c: Fix whitespace damage throughout the file.
720 * config/s390/s390.h: Likewise.
721 * config/s390/tpf.h: Likewise.
723 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
725 * config/s390/s390.c (s390_loadrelative_operand_p):
726 Remove TARGET_CPU_ZARCH usages.
727 (s390_rtx_costs): Likewise.
728 (s390_legitimate_constant_p): Likewise.
729 (s390_cannot_force_const_mem): Likewise.
730 (legitimate_reload_constant_p): Likewise.
731 (s390_preferred_reload_class): Likewise.
732 (legitimize_pic_address): Likewise.
733 (legitimize_tls_address): Likewise.
734 (s390_split_branches): Removed.
735 (s390_add_execute): Removed.
736 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
737 (s390_mainpool_start): Likewise.
738 (s390_mainpool_finish): Likewise.
739 (s390_mainpool_cancel): Removed.
740 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
741 (s390_chunkify_cancel): Likewise.
742 (s390_return_addr_rtx): Likewise.
743 (s390_register_info): Remove split_branches_pending_p uages.
744 (s390_optimize_register_info): Likewise.
745 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
746 split_branches_pending_p usages.
747 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
748 (s390_load_got): Likewise.
749 (s390_expand_split_stack_prologue): Likewise.
750 (output_asm_nops): Likewise.
751 (s390_function_profiler): Likewise.
752 (s390_emit_call): Likewise.
753 (s390_conditional_register_usage): Likewise.
754 (s390_optimize_prologue): Likewise.
755 (s390_reorg): Remove TARGET_CPU_ZARCH and
756 split_branches_pending_p usages.
757 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
759 (s390_output_indirect_thunk_function): Likewise.
760 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
761 (TARGET_CPU_ZARCH_P): Removed.
762 (struct machine_function): Remove split_branches_pending_p.
763 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
765 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
767 * common/config/s390/s390-common.c (processor_flags_table):
769 * config.gcc: Remove with_arch/with_tune support.
770 * config/s390/2064.md: Remove cpu attribute comparisons.
771 * config/s390/driver-native.c (s390_host_detect_local_cpu):
773 * config/s390/linux.h (ASM_SPEC):
774 Remove -march support.
775 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
776 Use a table to get an arch level.
777 * config/s390/s390-opts.h (enum processor_type):
780 (processor_table): Remove entries, add arch_level values.
781 (s390_issue_rate): Remove cases.
782 (s390_option_override): Adjust
783 s390_option_override_internal() call.
784 (s390_option_override_internal): Remove deprecation warning.
785 (s390_valid_target_attribute_tree): Adjust
786 s390_option_override_internal() call.
787 * config/s390/s390.h (struct s390_processor):
788 Share with s390-c.c, add arch_level field.
789 * config/s390/s390.md:
790 Remove occurrences in cpu attribute.
791 * config/s390/s390.opt: Remove -march/-mtune support.
792 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
793 * doc/invoke.texi: Remove deprecation warning.
795 2018-08-08 Luis Machado <luis.machado@linaro.org>
797 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
799 (qdf24xx_tunings): Set vector cost structure to
802 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
803 <register_sextend>: Set to 3.
805 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
808 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
809 * config/aarch64/aarch64-simd.md
810 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
811 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
812 (aarch64_frecpx<mode>): ...this new pattern.
813 * config/aarch64/aarch64-simd-builtins.def: Remove comment
814 about aarch64_frecp<FRECP:frecp_suffix><mode>.
816 2018-08-07 Martin Liska <mliska@suse.cz>
819 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
820 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
821 * predict.def (PRED_MALLOC_NONNULL): New predictor.
822 * doc/extend.texi: Document that malloc attribute adds
825 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
828 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
829 Define to speculation_safe_value_not_needed.
831 2018-08-06 Jeff Law <law@redhat.com>
833 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
834 the vr_values instance to cprop_into_stmt.
835 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
836 (cprop_operand): Also query EVRP to determine if OP is a constant.
838 2018-08-06 Nathan Sidwell <nathan@acm.org>
840 * diagnostic.c (diagnostic_report_current_module): Reroll
841 included-at loop. Translate text.
843 2018-08-06 David Malcolm <dmalcolm@redhat.com>
845 * function-tests.c (selftest::test_expansion_to_rtl): Call
846 free_after_compilation.
848 2018-08-06 Alan Hayward <alan.hayward@arm.com>
850 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
852 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
854 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
855 loops with memory block operations from getting unrolled.
857 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
860 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
861 Define to speculation_safe_value_not_needed.
863 2018-08-06 Jeff Law <law@redhat.com>
865 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
868 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
871 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
872 with all enabled __intN types.
874 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
876 2018-08-06 Alan Hayward <alan.hayward@arm.com>
878 * alias.c (record_set): Check for clobber high.
879 * cfgexpand.c (expand_gimple_stmt): Likewise.
880 * combine-stack-adj.c (single_set_for_csa): Likewise.
881 * combine.c (find_single_use_1): Likewise.
882 (set_nonzero_bits_and_sign_copies): Likewise.
883 (get_combine_src_dest): Likewise.
884 (is_parallel_of_n_reg_sets): Likewise.
885 (try_combine): Likewise.
886 (record_dead_and_set_regs_1): Likewise.
887 (reg_dead_at_p_1): Likewise.
888 (reg_dead_at_p): Likewise.
889 * dce.c (deletable_insn_p): Likewise.
890 (mark_nonreg_stores_1): Likewise.
891 (mark_nonreg_stores_2): Likewise.
892 * df-scan.c (df_find_hard_reg_defs): Likewise.
893 (df_uses_record): Likewise.
894 (df_get_call_refs): Likewise.
895 * dwarf2out.c (mem_loc_descriptor): Likewise.
896 * haifa-sched.c (haifa_classify_rtx): Likewise.
897 * ira-build.c (create_insn_allocnos): Likewise.
898 * ira-costs.c (scan_one_insn): Likewise.
899 * ira.c (equiv_init_movable_p): Likewise.
900 (rtx_moveable_p): Likewise.
901 (interesting_dest_for_shprep): Likewise.
902 * jump.c (mark_jump_label_1): Likewise.
903 * postreload-gcse.c (record_opr_changes): Likewise.
904 * postreload.c (reload_cse_simplify): Likewise.
905 (struct reg_use): Add source expr.
906 (reload_combine): Check for clobber high.
907 (reload_combine_note_use): Likewise.
908 (reload_cse_move2add): Likewise.
909 (move2add_note_store): Likewise.
910 * print-rtl.c (print_pattern): Likewise.
911 * recog.c (decode_asm_operands): Likewise.
912 (store_data_bypass_p): Likewise.
913 (if_test_bypass_p): Likewise.
914 * regcprop.c (kill_clobbered_value): Likewise.
915 (kill_set_value): Likewise.
916 * reginfo.c (reg_scan_mark_refs): Likewise.
917 * reload1.c (maybe_fix_stack_asms): Likewise.
918 (eliminate_regs_1): Likewise.
919 (elimination_effects): Likewise.
920 (mark_not_eliminable): Likewise.
921 (scan_paradoxical_subregs): Likewise.
922 (forget_old_reloads_1): Likewise.
923 * reorg.c (find_end_label): Likewise.
924 (try_merge_delay_insns): Likewise.
925 (redundant_insn): Likewise.
926 (own_thread_p): Likewise.
927 (fill_simple_delay_slots): Likewise.
928 (fill_slots_from_thread): Likewise.
929 (dbr_schedule): Likewise.
930 * resource.c (update_live_status): Likewise.
931 (mark_referenced_resources): Likewise.
932 (mark_set_resources): Likewise.
933 * rtl.c (copy_rtx): Likewise.
934 * rtlanal.c (reg_referenced_p): Likewise.
935 (single_set_2): Likewise.
936 (noop_move_p): Likewise.
937 (note_stores): Likewise.
938 * sched-deps.c (sched_analyze_reg): Likewise.
939 (sched_analyze_insn): Likewise.
941 2018-08-06 Alan Hayward <alan.hayward@arm.com>
943 * cse.c (invalidate_reg): New function extracted from...
944 (invalidate): ...here.
945 (canonicalize_insn): Check for clobber high.
946 (invalidate_from_clobbers): invalidate clobber highs.
947 (invalidate_from_sets_and_clobbers): Likewise.
948 (count_reg_usage): Check for clobber high.
949 (insn_live_p): Likewise.
950 * cselib.c (cselib_expand_value_rtx_1):Likewise.
951 (cselib_invalidate_regno): Check for clobber in setter.
952 (cselib_invalidate_rtx): Pass through setter.
953 (cselib_invalidate_rtx_note_stores):
954 (cselib_process_insn): Check for clobber high.
955 * cselib.h (cselib_invalidate_rtx): Add operand.
957 2018-08-06 Alan Hayward <alan.hayward@arm.com>
959 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
960 (mark_not_eliminable): Likewise.
961 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
962 * lra-lives.c (process_bb_lives): Check for clobber high.
963 * lra.c (new_insn_reg): Remember clobber highs.
964 (collect_non_operand_hard_regs): Check for clobber high.
965 (lra_set_insn_recog_data): Likewise.
966 (add_regs_to_insn_regno_info): Likewise.
967 (lra_update_insn_regno_info): Likewise.
969 2018-08-06 Alan Hayward <alan.hayward@arm.com>
971 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
972 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
974 2018-08-06 Alan Hayward <alan.hayward@arm.com>
976 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
977 (copy_insn_1): Likewise.
978 (gen_hard_reg_clobber_high): New gen function.
979 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
980 * genemit.c (gen_exp): Likewise.
981 (gen_emit_seq): Pass through info.
982 (gen_insn): Check for CLOBBER_HIGH.
983 (gen_expand): Pass through info.
984 (gen_split): Likewise.
985 (output_add_clobbers): Likewise.
986 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
987 (remove_clobbers): Likewise.
988 * rtl.h (gen_hard_reg_clobber_high): New declaration.
990 2018-08-06 Alan Hayward <alan.hayward@arm.com>
992 * doc/rtl.texi (clobber_high): Add.
993 (parallel): Add in clobber high
994 * rtl.c (rtl_check_failed_code3): Add function.
995 * rtl.def (CLOBBER_HIGH): Add expression.
996 * rtl.h (RTL_CHECKC3): Add macro.
997 (rtl_check_failed_code3): Add declaration.
1000 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
1003 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
1004 cfun->machine->max_used_stack_alignment if needed.
1006 2018-08-04 Martin Sebor <msebor@redhat.com>
1008 PR tree-optimization/86571
1009 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
1012 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
1015 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1018 2018-08-03 Jeff Law <law@redhat.com>
1021 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1022 Define to speculation_safe_value_not_needed.
1024 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1026 * doc/gcov.texi (-x): Remove duplicate "to".
1027 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1028 (-Wif-not-aligned): Remove duplicate "is".
1029 (-flto): Remove duplicate "the".
1030 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1031 duplicate "v5.00.b".
1032 (MSP430 Options): Remove duplicate "and" from the description
1033 of "-mgprel-sec=regexp".
1034 (x86 Options): Remove duplicate copies of "vmldLog102" and
1035 vmlsLog104 from description of "-mveclibabi=type".
1037 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
1039 * internal-fn.h (first_commutative_argument): Declare.
1040 * internal-fn.c (first_commutative_argument): New function.
1041 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
1042 restrictions for pattern statements. Use first_commutative_argument
1043 to look for commutative operands in calls to internal functions.
1045 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
1047 * Makefile.in (wide-int-range.o): New.
1048 * tree-vrp.c: Move all the wide_int_* functions to...
1049 * wide-int-range.cc: ...here.
1050 * tree-vrp.h: Move all the wide_int_* prototypes to...
1051 * wide-int-range.h: ...here.
1053 2018-08-03 Tom de Vries <tdevries@suse.de>
1055 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
1057 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
1058 * except.c (output_function_exception_table): Do early exit if
1059 targetm_common.except_unwind_info (&global_options) == UI_NONE.
1061 2018-08-03 Martin Liska <mliska@suse.cz>
1063 * predict.c (dump_prediction): Change to 2 digits
1066 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
1068 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
1070 (falkor_am_1_gtov_gtov): ... a new insn reservation.
1072 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
1074 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
1075 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
1076 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
1077 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
1078 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
1080 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1082 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1083 pp_take_prefix when saving the existing prefix.
1084 * diagnostic.c (diagnostic_append_note): Likewise.
1085 * langhooks.c (lhd_print_error_function): Likewise.
1086 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1087 param's type. Free the existing prefix.
1088 (pp_take_prefix): New function.
1089 (pretty_printer::pretty_printer): Drop the prefix parameter.
1090 Rename the length parameter to match the comment.
1091 (pretty_printer::~pretty_printer): Free the prefix.
1092 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1094 (struct pretty_printer): Drop the "const" from "prefix" field's
1095 type and clarify memory management.
1096 (pp_set_prefix): Drop the "const" from the 2nd param.
1097 (pp_take_prefix): New decl.
1099 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1101 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
1102 for word_mode_ok here instead of passing as argument.
1103 (expand_block_compare): Change select_block_compare_mode() call.
1104 (expand_strncmp_gpr_sequence): New function.
1105 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
1107 2018-08-02 Jeff Law <law@redhat.com>
1110 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1111 Define to speculation_safe_value_not_needed.
1114 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1115 Define to speculation_safe_value_not_needed.
1117 2018-08-02 Tom de Vries <tdevries@suse.de>
1120 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1121 function. Return UI_TARGET unconditionally.
1122 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1123 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1125 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1127 * genemit.c (print_overload_test): Fix typo.
1129 2018-08-02 Richard Biener <rguenther@suse.de>
1131 PR tree-optimization/86816
1132 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1133 which checks for value availability before querying it.
1134 (gvn_uses_equal): Use it.
1135 (same_succ_hash): Likewise.
1136 (gimple_equal_p): Likewise.
1138 2018-08-02 Nick Clifton <nickc@redhat.com>
1141 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1142 Define to speculation_safe_value_not_needed.
1145 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1146 Define to speculation_safe_value_not_needed.
1149 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1150 Define to speculation_safe_value_not_needed.
1153 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1154 Define to speculation_safe_value_not_needed.
1157 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1158 Define to speculation_safe_value_not_needed.
1161 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1162 Define to speculation_safe_value_not_needed.
1165 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1166 Define to speculation_safe_value_not_needed.
1169 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1170 Define to speculation_safe_value_not_needed.
1173 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1174 speculation_safe_value_not_needed.
1177 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1178 to speculation_safe_value_not_needed.
1180 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1182 * doc/md.texi: Expand the documentation of instruction names
1183 to mention port-local uses. Document '@' in pattern names.
1184 * read-md.h (overloaded_instance, overloaded_name): New structs.
1186 (md_reader::handle_overloaded_name): New member function.
1187 (md_reader::get_overloads): Likewise.
1188 (md_reader::m_first_overload): New member variable.
1189 (md_reader::m_next_overload_ptr): Likewise.
1190 (md_reader::m_overloads_htab): Likewise.
1191 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1192 m_next_overload_ptr and m_overloads_htab.
1193 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1194 (get_mode_token, get_code_token, get_int_token): New functions.
1195 (map_attr_string): Add an optional argument that passes back
1196 the associated iterator.
1197 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1198 (md_reader::handle_overloaded_name, add_overload_instance): New
1200 (apply_iterators): Handle '@' names. Report an error if '@'
1201 is used without iterators.
1202 (initialize_iterators): Initialize the new iterator_group fields.
1203 * genopinit.c (handle_overloaded_code_for)
1204 (handle_overloaded_gen): New functions.
1205 (main): Use them to print declarations of maybe_code_for_* and
1206 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1207 * genemit.c (print_overload_arguments, print_overload_test)
1208 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1209 (main): Use it to print definitions of maybe_code_for_* and
1210 maybe_gen_* functions.
1211 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1212 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1213 instead of explicit mode checks.
1214 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1215 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1216 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1217 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1218 (aarch64_expand_compare_and_swap): Likewise
1219 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1220 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1221 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1222 (aarch64_constant_pool_reload_icode): Delete.
1223 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1224 instead of aarch64_constant_pool_reload_icode. Use
1225 code_for_aarch64_reload_mov instead of explicit mode checks.
1226 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1227 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1228 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1229 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1230 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1231 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1232 (aarch64_atomic_load_op_code): Delete.
1233 (aarch64_emit_atomic_load_op): Likewise.
1234 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1235 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1236 instead of aarch64_emit_atomic_load_op.
1237 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1238 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1239 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1240 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1241 character before the pattern name.
1242 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1243 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1244 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1245 (aarch64_frecps<mode>): Likewise.
1246 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1247 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1248 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1249 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1250 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1252 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1254 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1255 Allow HFmode constants if TARGET_FP_F16INST.
1257 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1260 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1261 No longer check last store for clobber of address register.
1263 2018-08-02 Martin Liska <mliska@suse.cz>
1265 PR gcov-profile/86817
1266 * gcov.c (process_all_functions): New function.
1268 (process_file): Move functions processing to
1269 process_all_functions.
1271 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1273 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1274 "const" to the "gimple *" and "rtx_insn *" parameters.
1275 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1277 (dump_location_t::dump_location_t): Likewise.
1279 2018-08-01 Martin Sebor <msebor@redhat.com>
1281 PR tree-optimization/86650
1282 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1283 rather than a "gcall *". Directly pass the data of interest
1284 to percent_K_format, rather than building a temporary CALL_EXPR
1286 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1287 (gimple_fold_builtin_strncat): Adjust.
1288 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1289 gcall* argument with gimple*.
1290 * gimple-ssa-warn-restrict.c (check_call): Same.
1291 (wrestrict_dom_walker::before_dom_children): Same.
1292 (builtin_access::builtin_access): Same.
1293 (check_bounds_or_overlap): Same
1294 (maybe_diag_overlap): Same.
1295 (maybe_diag_offset_bounds): Same.
1296 * tree-diagnostic.c (default_tree_printer): Move usage of
1297 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1299 * tree-pretty-print.c (percent_K_format): Add argument.
1300 * tree-pretty-print.h: Add argument.
1301 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1302 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1303 (maybe_diag_stxncpy_trunc): Same.
1304 (handle_builtin_stxncpy): Same.
1305 (handle_builtin_strcat): Same.
1307 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1309 * match.pd: Optimise pointer range checks.
1311 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1313 PR tree-optimization/86758
1314 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1315 to remove pattern statements.
1317 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1319 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1320 result of dfs_enumerate_from when constructing stmt_vec_infos,
1321 instead of additionally calling get_loop_body.
1323 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1325 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1327 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1328 When creating an iv, assert that the step is not known to be zero.
1329 (vect_setup_realignment): Update call accordingly.
1330 * tree-vect-stmts.c (vectorizable_store): Likewise.
1331 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1333 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1335 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1336 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1337 (vectorizable_reduction): Likewise.
1338 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1339 (vect_detect_hybrid_slp_stmts): Likewise.
1340 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1342 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1344 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1345 (wide_int_set_zero_nonzero_bits): ...this.
1346 (zero_nonzero_bits_from_vr): Rename to...
1347 (vrp_set_zero_nonzero_bits): ...this.
1348 (extract_range_from_multiplicative_op_1): Abstract wide int
1350 (wide_int_range_multiplicative_op): ...here.
1351 (extract_range_from_binary_expr_1): Extract wide int binary
1352 operations into their own functions.
1353 (wide_int_range_lshift): New.
1354 (wide_int_range_can_optimize_bit_op): New.
1355 (wide_int_range_shift_undefined_p): New.
1356 (wide_int_range_bit_xor): New.
1357 (wide_int_range_bit_ior): New.
1358 (wide_int_range_bit_and): New.
1359 (wide_int_range_trunc_mod): New.
1360 (extract_range_into_wide_ints): New.
1361 (vrp_shift_undefined_p): New.
1362 (extract_range_from_multiplicative_op): New.
1363 (vrp_can_optimize_bit_op): New.
1364 * tree-vrp.h (value_range::dump): New.
1365 (wide_int_range_multiplicative_op): New.
1366 (wide_int_range_lshift):New.
1367 (wide_int_range_shift_undefined_p): New.
1368 (wide_int_range_bit_xor): New.
1369 (wide_int_range_bit_ior): New.
1370 (wide_int_range_bit_and): New.
1371 (wide_int_range_trunc_mod): New.
1372 (zero_nonzero_bits_from_bounds): Rename to...
1373 (wide_int_set_zero_nonzero_bits): ...this.
1374 (zero_nonzero_bits_from_vr): Rename to...
1375 (vrp_set_zero_nonzero_bits): ...this.
1376 (range_easy_mask_min_max): Rename to...
1377 (wide_int_range_can_optimize_bit_op): this.
1378 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1379 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1381 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1383 * tree-vectorizer.h (vect_orig_stmt): New function.
1384 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1385 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1386 (vect_create_epilog_for_reduction): Likewise.
1387 (vectorizable_live_operation): Likewise.
1388 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1389 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1390 * tree-vect-stmts.c (vectorizable_call): Likewise.
1391 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1393 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1395 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1397 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1398 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1399 (vect_transform_loop): Likewise.
1400 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1402 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1404 * tree-vectorizer.h (vect_schedule_slp): Return void.
1405 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1406 (vect_schedule_slp): Likewise.
1408 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1410 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1412 (vect_transform_loop): Update calls accordingly. Schedule SLP
1413 instances before the main loop, if any exist.
1415 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1417 PR tree-optimization/86749
1418 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1419 If the lhs is used in a COND_EXPR, check that it is being used
1420 as the "then" or "else" value.
1422 2018-08-01 Tom de Vries <tdevries@suse.de>
1425 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1426 speculation_safe_value_not_needed.
1428 2018-08-01 Richard Biener <rguenther@suse.de>
1430 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1433 2018-08-01 Martin Liska <mliska@suse.cz>
1435 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1436 format how successful transformation is dumped.
1437 (gimple_mod_pow2_value_transform): Likewise.
1438 (gimple_mod_subtract_transform): Likewise.
1439 (gimple_stringops_transform): Likewise.
1441 2018-08-01 Martin Liska <mliska@suse.cz>
1444 * value-prof.c (interesting_stringop_to_profile_p):
1445 Simplify the code and add BUILT_IN_MEMMOVE.
1446 (gimple_stringops_transform): Likewise.
1448 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1450 * config/aarch64/aarch64-simd.md
1451 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1452 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1453 use GPI iterator instead of SI mode.
1455 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1457 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1458 rs6000_speculation_barrier.
1459 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1460 new barrier pattern name.
1462 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1464 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1465 (speculation_barrier): New insn.
1467 2018-08-01 Richard Biener <rguenther@suse.de>
1470 * graphite.h: Include isl/id.h and isl/space.h to allow build
1473 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1476 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1477 mode for COFF targets.
1478 * defaults.h (TARGET_COFF): Define.
1479 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1480 TARGET_COFF): Define.
1481 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1482 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1483 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1485 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1487 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1488 * gimple-streamer-out.c (output_bb): Save it.
1489 * lto-streamer-in.c (input_struct_function_base): Restore
1490 instance discriminator if available. Create map on demand.
1491 * lto-streamer-out.c (output_struct_function_base): Save it if
1493 * final.c (decl_to_instance_map): Document LTO strategy.
1495 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1496 Olivier Hainque <hainque@adacore.com>
1498 * debug.h (decl_to_instance_map_t): New type.
1499 (decl_to_instance_map): Declare.
1500 (maybe_create_decl_to_instance_map): New inline function.
1501 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1502 to track basic block discriminators.
1503 (final_start_function_1): Initialize them.
1504 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1506 (decl_to_instance_map): New variable.
1507 (map_decl_to_instance, maybe_set_discriminator): New functions.
1508 (notice_source_line): Set discriminator.
1510 2018-07-31 Ian Lance Taylor <iant@golang.org>
1512 * targhooks.c (default_have_speculation_safe_value): Add
1515 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1517 * dump-context.h: Include "pretty-print.h".
1518 (dump_context::refresh_dumps_are_enabled): New decl.
1519 (dump_context::emit_item): New decl.
1520 (class dump_context): Add fields "m_test_pp" and
1522 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1523 (temp_dump_context::get_dumped_text): New decl.
1524 (class temp_dump_context): Add field "m_pp".
1525 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1526 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1528 (set_dump_file): Update for above change.
1529 (set_alt_dump_file): Likewise.
1530 (dump_loc): New overload, taking a pretty_printer *.
1531 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1532 to any test pretty-printer.
1533 (make_item_for_dump_gimple_stmt): New function, adapted from
1534 optinfo::add_gimple_stmt.
1535 (dump_context::dump_gimple_stmt): Call it, and use the result,
1536 eliminating the direct usage of dump_file and alt_dump_file in
1537 favor of indirectly using them via emit_item.
1538 (make_item_for_dump_gimple_expr): New function, adapted from
1539 optinfo::add_gimple_expr.
1540 (dump_context::dump_gimple_expr): Call it, and use the result,
1541 eliminating the direct usage of dump_file and alt_dump_file in
1542 favor of indirectly using them via emit_item.
1543 (make_item_for_dump_generic_expr): New function, adapted from
1545 (dump_context::dump_generic_expr): Call it, and use the result,
1546 eliminating the direct usage of dump_file and alt_dump_file in
1547 favor of indirectly using them via emit_item.
1548 (make_item_for_dump_printf_va): New function, adapted from
1549 optinfo::add_printf_va.
1550 (make_item_for_dump_printf): New function.
1551 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1552 and use the result, eliminating the direct usage of dump_file and
1553 alt_dump_file in favor of indirectly using them via emit_item.
1554 (make_item_for_dump_dec): New function.
1555 (dump_context::dump_dec): Call it, and use the result,
1556 eliminating the direct usage of dump_file and alt_dump_file in
1557 favor of indirectly using them via emit_item.
1558 (make_item_for_dump_symtab_node): New function, adapted from
1559 optinfo::add_symtab_node.
1560 (dump_context::dump_symtab_node): Call it, and use the result,
1561 eliminating the direct usage of dump_file and alt_dump_file in
1562 favor of indirectly using them via emit_item.
1563 (dump_context::begin_scope): Reimplement, avoiding direct usage
1564 of dump_file and alt_dump_file in favor of indirectly using them
1566 (dump_context::emit_item): New member function.
1567 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1568 Set up test pretty-printer on the underlying context. Call
1569 refresh_dumps_are_enabled.
1570 (temp_dump_context::~temp_dump_context): Call
1571 refresh_dumps_are_enabled.
1572 (temp_dump_context::get_dumped_text): New member function.
1573 (selftest::verify_dumped_text): New function.
1574 (ASSERT_DUMPED_TEXT_EQ): New macro.
1575 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1576 and then without optinfo enabled. Add uses of
1577 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1578 * dumpfile.h: Update comment for the dump_* API.
1579 * optinfo-emit-json.cc
1580 (selftest::test_building_json_from_dump_calls): Update for new
1581 param for temp_dump_context ctor.
1582 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1583 and "m_owned" field.
1584 (optinfo_item::~optinfo_item): Likewise.
1585 (optinfo::add_item): New member function.
1586 (optinfo::emit): Update comment.
1587 (optinfo::add_string): Delete.
1588 (optinfo::add_printf): Delete.
1589 (optinfo::add_printf_va): Delete.
1590 (optinfo::add_gimple_stmt): Delete.
1591 (optinfo::add_gimple_expr): Delete.
1592 (optinfo::add_tree): Delete.
1593 (optinfo::add_symtab_node): Delete.
1594 (optinfo::add_dec): Delete.
1595 * optinfo.h (class dump_context): New forward decl.
1596 (optinfo::add_item): New decl.
1597 (optinfo::add_string): Delete.
1598 (optinfo::add_printf): Delete.
1599 (optinfo::add_printf_va): Delete.
1600 (optinfo::add_gimple_stmt): Delete.
1601 (optinfo::add_gimple_expr): Delete.
1602 (optinfo::add_tree): Delete.
1603 (optinfo::add_symtab_node): Delete.
1604 (optinfo::add_dec): Delete.
1605 (optinfo::add_poly_int): Delete.
1606 (optinfo_item::optinfo_item): Remove "owned" param.
1607 (class optinfo_item): Remove field "m_owned".
1609 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1612 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1613 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1614 requested variable alignment.
1615 (expand_one_ssa_partition): Likewise.
1616 (expand_one_var): Likewise.
1618 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1620 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1621 to speculation_safe_value_not_needed.
1623 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1625 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1626 * targhooks.c (speculation_safe_value_not_needed): New function.
1627 * target.def (have_speculation_safe_value): Update documentation.
1628 * doc/tm.texi: Regenerated.
1630 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1632 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1633 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1635 (despeculate_copy<ALLI:mode>_insn): New insn.
1636 (despeculate_copyti_insn): New insn.
1637 (despeculate_simple<ALLI:mode>): New insn
1638 (despeculate_simpleti): New insn.
1639 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1641 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1642 aarch64_speculation_safe_value.
1643 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1645 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1647 * config/aarch64/aarch64-speculation.cc: New file.
1648 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1649 before pass_reorder_blocks.
1650 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1652 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1653 X14 and X15 when tracking speculation.
1654 * config/aarch64/aarch64.md (register name constants): Add
1655 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1656 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1657 (speculation_barrier): New insn attribute.
1658 (cmp<mode>): Allow SP in comparisons.
1659 (speculation_tracker): New insn.
1660 (speculation_barrier): Add speculation_barrier attribute.
1661 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1662 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1663 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1665 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1667 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1668 aarch64_track_speculation is true.
1669 (tb<optab><mode>1): Likewise.
1670 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1671 generate CB[N]Z when tracking speculation.
1672 (aarch64_split_compare_and_swap): Likewise.
1673 (aarch64_split_atomic_op): Likewise.
1675 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1677 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1679 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1681 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1682 (speculation_barrier): New insn.
1684 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1686 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1687 * config/arm/arm.md (speculation_barrier): New expand.
1688 (speculation_barrier_insn): New pattern.
1690 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1692 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1693 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1694 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1695 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1697 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1698 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1699 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1700 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1701 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1702 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1703 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1704 * builtins.c (expand_speculation_safe_value): New function.
1705 (expand_builtin): Call it.
1706 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1707 * doc/extend.texi: Document __builtin_speculation_safe_value.
1708 * doc/md.texi: Document "speculation_barrier" pattern.
1709 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1710 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1711 * doc/tm.texi: Regenerated.
1712 * target.def (have_speculation_safe_value, speculation_safe_value): New
1714 * targhooks.c (default_have_speculation_safe_value): New function.
1715 (default_speculation_safe_value): New function.
1716 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1717 (default_speculation_safe_value): Add prototype.
1719 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1721 * dump-context.h (dump_context::dump_loc): New decl.
1722 * dumpfile.c (dump_context::dump_loc): New member function.
1723 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1724 and dump_gimple_stmt.
1725 (dump_context::dump_gimple_expr_loc): Likewise, using
1727 (dump_context::dump_generic_expr_loc): Likewise, using
1729 (dump_context::dump_printf_loc_va): Likewise, using
1731 (dump_context::begin_scope): Explicitly using the global function
1732 "dump_loc", rather than the member function.
1734 2018-07-31 Martin Sebor <msebor@redhat.com>
1736 PR tree-optimization/86741
1737 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1739 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1741 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1742 depend on whether prefetch instructions will be emitted or not.
1743 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1744 will be emitted or not.
1745 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1746 (TARGET_SETMEM_PFD): New macros.
1748 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1750 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1751 (NULL_STMT_VEC_INFO): Delete.
1752 (stmt_vec_info::operator*): Likewise.
1753 (stmt_vec_info::operator gimple *): Likewise.
1754 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1755 of NULL_STMT_VEC_INFO.
1756 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1757 (vect_reassociating_reduction_p): Likewise.
1758 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1759 (vectorizable_store): Likewise.
1760 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1761 (vec_info::free_stmt_vec_infos): Likewise.
1763 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1765 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1766 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1770 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1771 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1772 (vec_info::free_stmt_vec_info): New private member functions.
1773 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1774 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1775 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1776 set_stmt_vec_info_vec and free_stmt_vec_infos.
1777 * tree-vect-loop.c (_loop_vec_info): Remove call to
1778 set_stmt_vec_info_vec.
1779 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1780 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1781 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1782 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1783 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1784 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1785 with the clearing in the xcalloc.
1786 (stmt_vec_info_vec): Delete.
1787 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1788 (vectorize_loops): Likewise.
1789 (vec_info::~vec_info): Remove argument from call to
1790 free_stmt_vec_infos.
1791 (vec_info::add_stmt): Remove vinfo argument from call to
1794 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1796 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1797 rather than a gimple stmt.
1798 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1799 information for pattern statements when passed the original
1800 statement; instead wait to be passed the pattern statement itself.
1801 Don't call set_vinfo_for_stmt here.
1802 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1803 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1804 stmt_vec_infos here.
1805 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1806 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1807 stmt_vec_infos entry.
1809 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1811 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1812 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1813 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1814 * tree-vect-stmts.c (vectorizable_call): Likewise.
1815 (vectorizable_simd_clone_call): Likewise.
1817 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1819 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1820 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1821 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1822 * tree-vect-loop.c (vect_transform_loop): Likewise.
1823 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1824 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1826 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1828 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1829 (vect_dr_stmt): Delete.
1830 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1831 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1833 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1834 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1835 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1836 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1837 (vect_analyze_data_refs): Likewise.
1838 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1840 (vect_find_same_alignment_drs): Likewise.
1841 (vect_slp_analyze_node_dependences): Update calls accordingly.
1842 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1843 instead of DR_VECT_AUX.
1844 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1845 of a vector data references. Use vec_info::lookup_dr instead of
1847 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1848 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1849 instead of DR_VECT_AUX.
1851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1853 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1855 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1857 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1858 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1859 (vect_gen_prolog_loop_niters): Likewise.
1861 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1863 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1864 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1865 (vect_known_alignment_in_bytes, vect_dr_behavior)
1866 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1867 instead of data_references. Update calls to other routines for
1868 which the same change has been made.
1869 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1870 dr_vec_infos instead of stmt_vec_infos.
1871 (vect_analyze_data_ref_dependence): Update call accordingly.
1872 (vect_slp_analyze_data_ref_dependence)
1873 (vect_record_base_alignments): Use DR_VECT_AUX.
1874 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1875 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1876 (vector_alignment_reachable_p, vect_get_data_access_cost)
1877 (vect_peeling_supportable, vect_analyze_group_access_1)
1878 (vect_analyze_group_access, vect_analyze_data_ref_access)
1879 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1880 (vect_compile_time_alias, vect_small_gap_p)
1881 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1882 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1883 instead of data_references. Update calls to other routines for
1884 which the same change has been made.
1885 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1886 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1887 (vect_slp_analyze_and_verify_node_alignment)
1888 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1889 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1890 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1892 (_vect_peel_info::dr): Replace with...
1893 (_vect_peel_info::dr_info): ...this new field.
1894 (vect_peeling_hash_get_most_frequent)
1895 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1896 (vect_peeling_hash_get_lowest_cost):
1897 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1898 routines for which the same change has been made.
1899 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1901 * tree-vect-loop-manip.c (get_misalign_in_elems)
1902 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1904 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1905 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1906 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1907 (get_group_load_store_type, get_negative_load_store_type)
1908 (vect_get_data_ptr_increment, vectorizable_store)
1909 (vectorizable_load): Likewise.
1910 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1911 Update calls to other routines for which the same change has been made.
1913 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1915 * tree-vectorizer.h (vec_info::move_dr): New member function.
1916 (dataref_aux): Rename to...
1917 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1918 (_stmt_vec_info::dr_aux): Update accordingly.
1919 (_stmt_vec_info::data_ref_info): Delete.
1920 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1921 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1922 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1923 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1925 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1926 (STMT_VINFO_DR_INFO): New macro.
1927 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1928 (set_dr_misalignment): Update after rename of dataref_aux.
1929 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1930 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1931 initialization of STMT_VINFO_DATA_REF.
1932 * tree-vectorizer.c (vec_info::move_dr): New function.
1933 * tree-vect-patterns.c (vect_recog_bool_pattern)
1934 (vect_recog_mask_conversion_pattern)
1935 (vect_recog_gather_scatter_pattern): Use it.
1936 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1937 the "dr" and "stmt" fields of dr_vec_info instead of
1938 STMT_VINFO_DATA_REF.
1940 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1942 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1943 (is_pattern_stmt_p): Use it.
1944 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1945 on pattern statements.
1947 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1949 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1950 original stmt as a stmt_vec_info rather than a gimple stmt.
1951 (vect_pattern_recog_1): Take the statement directly as a
1952 stmt_vec_info, rather than via a gimple_stmt_iterator.
1953 Update call to vect_mark_pattern_stmts.
1954 (vect_pattern_recog): Update calls accordingly.
1956 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1958 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1959 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1960 a vect_def_type for the first argument.
1961 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1962 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1963 operand if it isn't defined by a vectorized statement.
1964 (vect_build_gather_load_calls): Remove the mask_dt argument and
1965 update calls to vect_get_vec_def_for_stmt_copy.
1966 (vectorizable_bswap): Likewise the dt argument.
1967 (vectorizable_call): Update calls to vectorizable_bswap and
1968 vect_get_vec_def_for_stmt_copy.
1969 (vectorizable_simd_clone_call, vectorizable_assignment)
1970 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1971 (vectorizable_comparison): Update calls to
1972 vect_get_vec_def_for_stmt_copy.
1973 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1975 (vect_get_loop_based_defs): Remove dt argument and update call
1976 to vect_get_vec_def_for_stmt_copy.
1977 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1978 and vect_get_vec_def_for_stmt_copy.
1979 (vectorizable_load): Update calls to vect_build_gather_load_calls
1980 and vect_get_vec_def_for_stmt_copy.
1981 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1982 (vectorizable_reduction, vectorizable_live_operation): Update calls
1983 to vect_get_vec_def_for_stmt_copy.
1985 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1987 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1988 and gimple stmt arguments with a stmt_vec_info.
1989 (vect_record_base_alignments): Update calls accordingly.
1990 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1991 and gimple stmt arguments with a stmt_vec_info.
1992 (vect_build_slp_tree_1): Remove vinfo argument and update call
1993 to vect_record_max_nunits.
1994 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1995 and vect_record_max_nunits.
1997 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1999 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
2000 file and take a stmt_vec_info instead of a gimple stmt.
2001 (supportable_widening_operation, vect_finish_replace_stmt)
2002 (vect_finish_stmt_generation, vect_get_store_rhs)
2003 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
2004 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
2005 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
2006 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
2007 (vect_create_data_ref_ptr, bump_vector_ptr)
2008 (vect_permute_store_chain, vect_setup_realignment)
2009 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
2010 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
2011 (vectorizable_reduction, vectorizable_induction)
2012 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
2013 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
2014 than gimple stmts as arguments.
2015 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
2016 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
2017 (can_group_stmts_p, vect_check_gather_scatter)
2018 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2019 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
2020 (vect_permute_load_chain, vect_shift_permute_load_chain)
2021 (vect_transform_grouped_load)
2022 (vect_record_grouped_load_vectors): Likewise.
2023 * tree-vect-loop.c (vect_fixup_reduc_chain)
2024 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
2025 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
2026 (vectorizable_reduction, vectorizable_induction)
2027 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
2028 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
2029 (vect_get_load_store_mask): Likewise.
2030 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
2031 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
2032 * tree-vect-stmts.c (vect_mark_relevant)
2033 (is_simple_and_all_uses_invariant)
2034 (exist_non_indexing_operands_for_use_p, process_use)
2035 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
2036 (vect_get_vec_def_for_operand, vect_get_vec_defs)
2037 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
2038 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
2039 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
2040 (get_negative_load_store_type, get_load_store_type)
2041 (vect_check_load_store_mask, vect_check_store_rhs)
2042 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
2043 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2044 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2045 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2046 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
2047 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
2048 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
2049 (supportable_widening_operation): Likewise.
2051 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2053 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
2054 a stmt_vec_info instead of a gcall.
2055 (vect_check_gather_scatter): Update call accordingly.
2056 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
2058 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
2059 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
2060 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
2061 instead of a gimple stmt.
2062 (vect_transform_loop): Update calls accordingly.
2063 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
2064 stmt_vec_infos instead of gimple stmts.
2065 (vect_analyze_slp_instance): Update use accordingly.
2066 * tree-vect-stmts.c (read_vector_array, write_vector_array)
2067 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
2068 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
2069 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
2070 (vect_gen_widened_results_half, vect_get_loop_based_defs)
2071 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
2072 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
2073 down to subroutines.
2075 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2077 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
2078 of the worklist from a vector of gimple stmts to a vector of
2080 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2081 (vect_mark_stmts_to_be_vectorized): Likewise
2083 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2085 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
2086 statement before passing it to vect_analyze_stmt.
2087 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
2088 the chain of phi vector definitions. Track the exit phi via its
2090 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
2091 STMT_VINFO_REDUC_DEF.
2092 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
2093 stmt_vec_infos to handle the statement chains.
2094 (vect_get_slp_defs): Record the first statement in the node
2095 using a stmt_vec_info.
2096 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
2097 statements here and pass their stmt_vec_info down to subroutines.
2098 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
2099 down to vect_finish_stmt_generation.
2100 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
2101 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
2102 stmt_vec_infos to subroutines.
2103 (vect_remove_stores): Use stmt_vec_infos to handle the statement
2106 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2108 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
2109 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
2110 (vect_permute_store_chain, vect_setup_realignment)
2111 (vect_permute_load_chain, vect_shift_permute_load_chain)
2112 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
2113 stmts internally, and when passing values to other vectorizer routines.
2114 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2115 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2116 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2117 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2118 (vectorizable_reduction, vectorizable_induction)
2119 (vectorizable_live_operation, vect_transform_loop_stmt)
2120 (vect_transform_loop): Likewise.
2121 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2122 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2123 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2124 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2125 (vect_slp_analyze_node_operations_1): Likewise.
2126 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2127 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2128 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2129 (vect_finish_stmt_generation_1, get_group_load_store_type)
2130 (get_load_store_type, vect_build_gather_load_calls)
2131 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2132 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2133 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2134 (vectorizable_store, vectorizable_load, vectorizable_condition)
2135 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2136 (supportable_widening_operation): Likewise.
2137 (vect_get_vector_types_for_stmt): Likewise.
2138 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2140 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2142 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2143 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2144 (vect_permute_store_chain, vect_permute_load_chain)
2145 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2146 repeated stmt_vec_info lookups.
2147 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2148 (vect_update_ivs_after_vectorizer): Likewise.
2149 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2150 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2151 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2152 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2153 (vect_bb_slp_scalar_cost): Likewise.
2154 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2156 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2158 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2159 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2160 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2161 to get gassigns and gcalls, rather than operating on generc gimple
2163 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2164 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2165 (vectorizable_load, vect_analyze_stmt): Likewise.
2166 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2168 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2170 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2171 return stmt_vec_infos rather than gimple stmts. Do not accept
2173 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2175 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2176 Update use of get_later_stmt.
2177 (vect_get_constant_vectors): Update call accordingly.
2178 (vect_schedule_slp_instance): Likewise
2179 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2180 (vect_slp_analyze_instance_dependence): Likewise.
2181 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2183 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2185 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2186 (stmt_info_for_cost::stmt_info): ...this new field.
2187 (add_stmt_costs): Update accordingly.
2188 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2189 (vect_get_known_peeling_cost): Likewise.
2190 (vect_estimate_min_profitable_iters): Likewise.
2191 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2193 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2195 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2196 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2197 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2199 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2201 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2203 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2204 a gimple stmt to a stmt_vec_info.
2205 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2207 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2209 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2210 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2211 (_loop_vec_info::reduction_chains): Likewise.
2212 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2214 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2216 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2218 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2219 a gimple stmt to a stmt_vec_info.
2220 (_stmt_vec_info::next_element): Likewise.
2221 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2222 (vect_slp_analyze_and_verify_node_alignment)
2223 (vect_analyze_group_access_1, vect_analyze_group_access)
2224 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2225 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2226 (vect_supportable_dr_alignment): Update accordingly.
2227 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2228 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2229 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2230 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2231 * tree-vect-slp.c (vect_build_slp_tree_1)
2232 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2233 (vect_split_slp_store_group, vect_analyze_slp_instance)
2234 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2235 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2236 (get_group_load_store_type, get_load_store_type)
2237 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2238 (vect_transform_stmt, vect_remove_stores): Likewise.
2240 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2242 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2244 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2245 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2246 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2247 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2248 (vector_alignment_reachable_p, vect_get_data_access_cost)
2249 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2250 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2251 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2252 (vect_analyze_group_access_1, vect_analyze_group_access)
2253 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2254 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2255 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2256 result of vect_dr_stmt and use the stmt_vec_info instead of
2257 the associated gimple stmt.
2258 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2259 (vect_gen_prolog_loop_niters): Likewise.
2260 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2262 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2264 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2265 to a vec<stmt_vec_info>.
2266 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2267 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2269 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2270 to a vec<stmt_vec_info>.
2271 (bst_traits::value_type, bst_traits::value_type): Likewise.
2272 (bst_traits::hash): Update accordingly.
2273 (vect_get_and_check_slp_defs): Change the stmts parameter from
2274 a vec<gimple *> to a vec<stmt_vec_info>.
2275 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2276 (vect_build_slp_tree): Likewise.
2277 (vect_build_slp_tree_2): Likewise. Update uses of
2278 SLP_TREE_SCALAR_STMTS.
2279 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2280 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2281 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2282 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2283 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2284 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2285 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2286 (vect_get_constant_vectors, vect_get_slp_defs)
2287 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2288 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2289 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2290 instead of gimple stmts.
2291 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2292 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2293 (vect_slp_analyze_instance_dependence): Update uses of
2294 SLP_TREE_SCALAR_STMTS.
2295 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2296 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2297 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2298 (get_initial_defs_for_reduction): Likewise.
2299 (vect_create_epilog_for_reduction): Likewise.
2300 (vectorize_fold_left_reduction): Likewise.
2301 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2302 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2303 (can_vectorize_live_stmts): Likewise.
2305 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2307 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2308 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2309 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2311 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2313 (gather_scalar_reductions): Update after above interface changes.
2314 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2315 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2317 (vect_force_simple_reduction): Likewise.
2318 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2319 LOOP_VINFO_REDUCTIONS.
2320 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2322 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2324 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2325 a gimple stmt to a stmt_vec_info.
2326 * tree-vect-loop.c (vect_active_double_reduction_p)
2327 (vect_force_simple_reduction, vectorizable_reduction): Update
2330 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2332 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2333 vec<gimple *> to a vec<stmt_vec_info>.
2334 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2335 the reduction_phis argument from a vec<gimple *> to a
2337 (vectorizable_reduction): Likewise the phis local variable that
2338 is passed to vect_create_epilog_for_reduction. Update for new type
2339 of SLP_TREE_VEC_STMTS.
2340 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2341 (vectorizable_live_operation): Likewise.
2342 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2343 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2345 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2347 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2348 a gimple stmt to a stmt_vec_info.
2349 (vectorizable_condition, vectorizable_live_operation)
2350 (vectorizable_reduction, vectorizable_induction): Pass back the
2351 vectorized statement as a stmt_vec_info.
2352 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2353 use of STMT_VINFO_VEC_STMT.
2354 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2355 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2356 as stmt_vec_infos rather than gimple stmts.
2357 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2359 (vectorizable_live_operation): Likewise.
2360 (vectorizable_reduction, vectorizable_induction): Likewise,
2361 updating use of STMT_VINFO_VEC_STMT.
2362 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2363 of STMT_VINFO_VEC_STMT.
2364 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2365 (vectorizable_simd_clone_call, vectorizable_conversion)
2366 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2367 (vectorizable_store, vectorizable_load, vectorizable_condition)
2368 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2369 from a gimple stmt to a stmt_vec_info.
2370 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2371 pointer to a stmt_vec_info to the vectorizable_* routines.
2373 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2375 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2376 a gimple stmt to a stmt_vec_info.
2377 (is_pattern_stmt_p): Update accordingly.
2378 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2379 (vect_record_grouped_load_vectors): Likewise.
2380 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2381 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2382 (vect_model_reduction_cost): Likewise.
2383 (vect_create_epilog_for_reduction): Likewise.
2384 (vectorizable_reduction, vectorizable_induction): Likewise.
2385 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2386 Return the stmt_vec_info for the pattern statement.
2387 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2388 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2389 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2390 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2391 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2392 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2393 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2394 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2396 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2398 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2399 (vect_finish_stmt_generation): Likewise.
2400 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2401 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2402 (vect_build_gather_load_calls): Use the return value of the above
2403 functions instead of a separate call to vinfo_for_stmt. Use narrow
2404 scopes for the input gimple stmt and wider scopes for the associated
2405 stmt_vec_info. Use vec_info::lookup_def when setting these
2406 stmt_vec_infos from an SSA_NAME definition.
2407 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2408 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2409 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2410 (vectorizable_store, vectorizable_load, vectorizable_condition)
2411 (vectorizable_comparison): Likewise.
2412 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2413 (vectorizable_reduction): Likewise.
2415 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2417 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2418 stmt_vec_info * parameter before the optional gimple **.
2419 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2420 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2421 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2422 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2423 (vectorizable_live_operation): Likewise.
2424 * tree-vect-patterns.c (type_conversion_p): Likewise.
2425 (vect_look_through_possible_promotion): Likewise.
2426 (vect_recog_rotate_pattern): Likewise.
2427 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2429 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2431 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2432 a typedef to a wrapper class.
2433 (NULL_STMT_VEC_INFO): New macro.
2434 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2435 (stmt_vec_info::operator*): New function.
2436 (stmt_vec_info::operator gimple *): Likewise.
2437 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2438 (add_stmt_costs): Likewise.
2439 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2440 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2441 (vect_get_known_peeling_cost): Likewise.
2442 (vect_estimate_min_profitable_iters): Likewise.
2443 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2444 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2445 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2446 (vectorizable_store, free_stmt_vec_infos): Likewise.
2447 (new_stmt_vec_info): Change return type of xcalloc to
2450 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2452 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2453 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2454 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2455 a single_imm_use-based sequence.
2456 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2458 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2460 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2461 * tree-vectorizer.c (vec_info::lookup_def): New function.
2462 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2463 (vect_widened_op_tree): Likewise.
2464 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2465 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2466 (vectorizable_reduction): Likewise.
2467 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2469 (vect_is_slp_reduction): Update calls accordingly. Use
2470 vec_info::lookup_def.
2471 (vect_is_simple_reduction): Likewise
2472 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2474 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2476 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2477 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2478 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2480 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2481 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2482 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2483 (vect_is_slp_reduction, vectorizable_induction)
2484 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2485 * tree-vect-patterns.c (vect_init_pattern_stmt):
2486 (vect_determine_min_output_precision_1, vect_determine_precisions)
2487 (vect_pattern_recog): Likewise.
2488 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2489 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2490 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2491 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2492 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2493 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2494 info field from a loop to a loop_vec_info.
2496 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2498 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2499 (vec_info::add_stmt): Declare.
2500 * tree-vectorizer.c (vec_info::add_stmt): New function.
2501 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2502 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2503 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2504 (vectorizable_induction): Likewise.
2505 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2506 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2507 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2508 (vectorizable_load): Likewise.
2509 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2510 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2511 (vect_recog_gather_scatter_pattern): Likewise.
2512 (append_pattern_def_seq): Likewise. Remove a check that is
2513 performed by add_stmt itself.
2515 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2517 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2518 which make_ssa_name was called with new_stmt before new_stmt
2521 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2523 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2525 (vect_is_slp_reduction): ...here...
2526 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2527 that are already known to be false.
2529 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2531 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2532 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2533 STMT_VINFO_NUM_SLP_USES when it's true.
2534 (vect_free_slp_instance): Add a final_p parameter and pass it to
2536 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2537 (vect_analyze_slp_instance): Likewise.
2538 (vect_slp_analyze_operations): Likewise.
2539 (vect_slp_analyze_bb_1): Likewise.
2540 * tree-vectorizer.c (vec_info): Likewise.
2541 * tree-vect-loop.c (vect_transform_loop): Likewise.
2543 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2545 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2546 function is not called for second and subsequent members of
2549 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2551 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2552 cases for nested loops from here to ...
2553 (vect_create_epilog_for_reduction): ...here. Only call
2554 vect_is_simple_use for inner-loop reductions.
2556 2018-07-31 Martin Liska <mliska@suse.cz>
2558 PR gcov-profile/85338
2559 PR gcov-profile/85350
2560 PR gcov-profile/85372
2561 * profile.c (struct location_triplet): New.
2562 (struct location_triplet_hash): Likewise.
2563 (output_location): Do not output a BB that
2564 is already recorded for a line.
2565 (branch_prob): Use streamed_locations.
2567 2018-07-31 Martin Liska <mliska@suse.cz>
2569 PR gcov-profile/85370
2570 * coverage.c (coverage_begin_function): Do not mark target
2571 clones as artificial functions.
2573 2018-07-31 Martin Liska <mliska@suse.cz>
2575 PR gcov-profile/83813
2576 PR gcov-profile/84758
2577 PR gcov-profile/85217
2578 PR gcov-profile/85332
2579 * profile.c (branch_prob): Do not record GOTO expressions
2580 for GIMPLE statements which locations are already streamed.
2582 2018-07-31 Olivier Hainque <hainque@adacore.com>
2584 * gcc.c (handle_spec_function): Accept a soft_matched_part
2585 argument, as do_spec_1. Pass it down to ...
2586 (eval_spec_function): Accept a soft_matched_part argument,
2587 and pass it down to ...
2588 (do_spec_2): Accept a soft_matched_part argument, and pass
2589 it down to do_spec_1.
2590 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2591 (handle_braces): Update call to handle_spec_function.
2592 (driver::set_up_specs): Update calls to do_spec_2.
2593 (compare_debug_dump_opt_spec_function): Likewise.
2594 (compare_debug_self_opt_spec_function): Likewise.
2596 2018-07-31 Olivier Hainque <hainque@adacore.com>
2598 * common.opt (nolibc): New option.
2599 * doc/invoke.texi (Link Options): Document it.
2600 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2601 * config/alpha/linux.h: Likewise.
2602 * config/arc/elf.h: Likewise.
2603 * config/arm/uclinux-elf.h: Likewise.
2604 * config/arm/unknown-elf.h: Likewise.
2605 * config/avr/avrlibc.h: Likewise.
2606 * config/bfin/bfin.h: Likewise.
2607 * config/bfin/linux.h: Likewise.
2608 * config/bfin/uclinux.h: Likewise.
2609 * config/darwin.h: Likewise.
2610 * config/darwin10.h: Likewise.
2611 * config/darwin12.h: Likewise.
2612 * config/gnu-user.h: Likewise.
2613 * config/lm32/uclinux-elf.h: Likewise.
2614 * config/pa/pa-hpux11.h: Likewise.
2615 * config/pa/pa64-hpux.h: Likewise.
2616 * config/sparc/sparc.h: Likewise.
2618 2018-07-31 Olivier Hainque <hainque@adacore.com>
2620 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2621 undefined variables.
2623 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2626 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2629 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2631 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2632 terminated string literal.
2634 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2636 PR rtl-optimization/85160
2637 * combine.c (is_just_move): New function.
2638 (try_combine): Allow combining two instructions into two if neither of
2639 the original instructions was a move.
2641 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2644 * doc/extend.texi (Global Register Variables): Discourage use of type
2646 (Local Register Variables): Likewise.
2648 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2650 PR tree-optimization/86506
2651 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2653 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2656 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2657 hard_regno, make sure no insn between `from` and `to` clobbers it.
2659 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2660 Tom de Vries <tdevries@suse.de>
2662 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2663 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2664 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2665 PTX_DEFAULT_RUNTIME_DIM.
2666 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2668 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2670 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2671 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2672 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2674 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2676 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2678 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2679 predicate to reflect toc fusion has been deleted.
2680 (toc_fusion_mem_raw): Delete, no longer used.
2681 (toc_fusion_mem_wrapped): Likewise.
2682 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2684 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2685 Delete, no longer used.
2686 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2687 meant to be used for toc fusion.
2688 (rs6000_debug_print_mode): Delete toc fusion debugging.
2689 (rs6000_debug_reg_global): Likewise.
2690 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2691 fusion and secondary reload support that were never used.
2692 (rs6000_option_override_internal): Delete TOC fusion, that was only
2693 partially defined, and it did not work unless you also used the
2695 (rs6000_legitimate_address_p): Delete TOC fusion support.
2696 (rs6000_opt_masks): Likewise.
2697 (fusion_wrap_memory_address): Delete function, no longer used.
2698 (fusion_split_address); Delete TOC fusion support.
2699 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2700 longer used with toc fusion being deleted.
2701 (TARGET_TOC_FUSION_FP): Likewise.
2702 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2704 (toc fusion spliter): Delete TOC fusion support.
2705 (toc_fusionload_<mode>): Likewise.
2706 (toc_fusionload_di): Likewise.
2707 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2708 longer needs to be named. Rename predicate to delete TOC fusion.
2709 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2710 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2711 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2712 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2713 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2715 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2717 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2718 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2719 __int128 in built-in function prototypes.
2720 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2721 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2723 2018-07-27 Martin Sebor <msebor@redhat.com>
2725 PR tree-optimization/86696
2726 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2727 types, including enums.
2728 (handle_char_store): Be prepared for the above function to fail.
2730 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2732 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2733 when optimization level is lower than 2 or optimize for size.
2735 2018-07-26 Martin Sebor <msebor@redhat.com>
2737 PR tree-optimization/86043
2738 PR tree-optimization/86042
2739 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2740 (get_string_cst_length): Rename...
2741 (get_min_string_length): ...to this. Add argument.
2742 (handle_char_store): Extend to handle multi-character stores by
2744 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2745 * tree.h (initializer_zerop): Add argument.
2747 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2750 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2751 declare target to variables if they have always,{to,from,tofrom} map
2754 2018-07-26 Martin Liska <mliska@suse.cz>
2757 * lto-wrapper.c: Add linker_output as prefix
2758 for ltrans_output_file.
2760 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2762 PR rtl-optimization/85805
2763 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2764 value for hard registers if that was written in the same mode.
2766 2018-07-26 Martin Liska <mliska@suse.cz>
2768 PR gcov-profile/86536
2769 * gcov.c (format_gcov): Use printf format %.*f directly
2770 and do not handle special values.
2772 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2774 * common/config/arc/arc-common.c (arc_option_optimization_table):
2775 Update default optimizations for size.
2777 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2779 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2780 * config/arc/constraints.md (CfZ): New constraint.
2781 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2782 (subsf3_fpu): Likewise.
2783 (cmpsf_fpu): Likewise.
2784 (cmpsf_fpu_uneq): Likewise.
2786 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2788 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2790 (arc_is_uncached_mem_p): uncached applies to both the variable and
2793 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2795 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2798 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2800 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2801 field "m_scopes" from vec to auto_vec.
2803 2018-07-25 Martin Liska <mliska@suse.cz>
2805 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2808 2018-07-25 Richard Biener <rguenther@suse.de>
2811 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2812 special wrt context_die late.
2813 (gen_subprogram_die): Re-use DIEs in local scope.
2815 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2817 PR tree-optimization/86644
2818 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2820 2018-07-25 Martin Liska <mliska@suse.cz>
2823 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2824 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2826 2018-07-25 Martin Liska <mliska@suse.cz>
2829 * params.def: Explain ASan abbreviation and provide
2830 a documentation link.
2832 2018-07-24 Martin Sebor <msebor@redhat.com>
2834 PR tree-optimization/86622
2835 PR tree-optimization/86532
2836 * builtins.h (string_length): Declare.
2837 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2838 (check_access): Be prepared for non-constant length ranges.
2839 (string_length): Make extern.
2840 * expr.c (string_constant): Only handle the minor non-constant
2841 array index. Use string_constant to compute the length of
2842 a generic string constant.
2844 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2846 PR tree-optimization/86618
2847 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2848 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2850 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2852 PR tree-optimization/86636
2853 * json.cc (json::object::set): Fix comment. Add assertions.
2854 (json::array::append): Move here from json.h. Add comment and an
2856 (json::string::string): Likewise.
2857 * json.h (json::array::append): Move to json.cc.
2858 (json::string::string): Likewise.
2859 * optinfo-emit-json.cc
2860 (optrecord_json_writer::impl_location_to_json): Assert that we
2861 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2862 wrapper around it. Expand the location once, rather than three
2864 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2865 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2867 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2868 places. Fix some overlong lines.
2870 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2872 * config/aarch64/aarch64-simd.md
2873 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2874 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2875 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2876 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2877 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2878 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2879 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2880 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2881 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2883 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2886 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2887 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2888 and abs_d == d, do the power of two handling if profitable.
2890 2018-07-24 Richard Biener <rguenther@suse.de>
2892 * match.pd: Add BIT_FIELD_REF canonicalizations.
2894 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2897 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2899 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2901 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2902 terminated STRING_CST object.
2904 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2906 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2908 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2910 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2911 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2912 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2913 the elements into a register.
2914 (rs6000_split_v4si_init_di_reg): Delete.
2915 (rs6000_split_v4si_init): Delete.
2916 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2917 (vsx_init_v4si): Rewrite as a define_expand.
2919 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2921 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2922 zero_extend argument from memory): New.
2924 2018-07-22 Martin Sebor <msebor@redhat.com>
2927 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2928 diagnosing calls with unknown arguments unless -Walloca-larger-than
2929 is restricted to less than PTRDIFF_MAX bytes.
2931 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2933 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2935 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2937 * pretty-print.c (text_info::set_location): Remove redundant
2938 "line_table" parameter from call to rich_location::set_range.
2940 2018-07-20 Martin Sebor <msebor@redhat.com>
2943 * builtins.c (expand_builtin_alloca): Adjust.
2944 * calls.c (alloc_max_size): Simplify.
2945 * cgraphunit.c (cgraph_node::expand): Adjust.
2946 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2948 (frame_larger_than_size): Same.
2949 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2950 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2952 * doc/invoke.texi (GCC Command Options): Document option arguments.
2953 Explain byte-size arguments and suffixes.
2954 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2955 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2956 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2957 * doc/options.texi (UInteger): Expand.
2958 (Host_Wide_Int, ByteSize): Document new properties.
2959 * final.c (final_start_function_1): Include sizes in an error message.
2960 * function.c (frame_offset_overflow): Same.
2961 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2962 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2963 Diagnose unbounded alloca calls only for limits of less than
2965 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2966 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2968 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2969 only for limits of less than PTRDIFF_MAX.
2970 * langhooks-def.h (lhd_handle_option): Change function argument
2972 * langhooks.c (lhd_handle_option): Same.
2973 * langhooks.h (handle_option): Same.
2974 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2976 (var_type, var_type_struct): Same.
2977 (var_set): Handle ByteSize flag.
2978 * optc-gen.awk: Add comments to output to ease debugging. Make
2979 use of HOST_WIDE_INT where appropriate.
2980 * opts-gen-save.awk: Use %lx to format unsigned long.
2981 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2982 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2983 arguments. Parse bytes-size suffixes.
2984 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2985 (enum_value_to_arg): Same.
2986 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2987 (handle_option): Adjust.
2988 (generate_option): Change function argument to HOST_WIDE_INT.
2989 (cmdline_handle_error): Adjust.
2990 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2991 (set_option): Change function argument to HOST_WIDE_INT.
2992 (option_enabled): Handle cl_host_wide_int.
2993 (get_option_state): Handle CLVC_SIZE.
2994 (control_warning_option): Same.
2995 * opts.c (common_handle_option): Change function argument to
2996 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2997 OPT_Wvla_larger_than_.
2998 * opts.h (enum cl_var_type): Add an enumerator.
2999 * stor-layout.c (layout_decl): Print a more meaningful warning.
3000 * toplev.c (output_stack_usage): Adjust.
3002 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
3004 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
3005 call to inline_expand_builtin_string_cmp.
3006 (expand_builtin_strcmp): Likewise.
3007 (expand_builtin_strncmp): Likewise.
3008 (inline_string_cmp): Delete the last parameter, change char_type_node
3009 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
3011 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
3012 the inlining expansion on target where the type of the call has same or
3013 narrower precision than unsigned char.
3015 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3017 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
3018 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
3019 * common.opt (fsave-optimization-record): New option.
3020 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
3021 * doc/invoke.texi (-fsave-optimization-record): New option.
3022 * dumpfile.c: Include "optinfo-emit-json.h".
3023 (struct kv_pair): Move to coretypes.h.
3024 (optgroup_options): Make non-static.
3025 (dump_context::end_scope): Call
3026 optimization_records_maybe_pop_dump_scope.
3027 * dumpfile.h (optgroup_options): New decl.
3028 * json.cc: New file.
3030 * optinfo-emit-json.cc: New file.
3031 * optinfo-emit-json.h: New file.
3032 * optinfo.cc: Include "optinfo-emit-json.h".
3033 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
3034 (optinfo_enabled_p): Check optimization_records_enabled_p.
3035 (optinfo_wants_inlining_info_p): Likewise.
3036 * optinfo.h: Update comment.
3037 * profile-count.c (profile_quality_as_string): New function.
3038 * profile-count.h (profile_quality_as_string): New decl.
3039 (profile_count::quality): New accessor.
3040 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
3041 and optinfo_emit_json_cc_tests.
3042 * selftest.h (selftest::json_cc_tests): New decl.
3043 (selftest::optinfo_emit_json_cc_tests): New decl.
3044 * toplev.c: Include "optinfo-emit-json.h".
3045 (compile_file): Call optimization_records_finish.
3046 (do_compile): Call optimization_records_start.
3047 * tree-ssa-live.c: Include optinfo.h.
3048 (remove_unused_scope_block_p): Retain inlining information if
3049 optinfo_wants_inlining_info_p returns true.
3051 2018-07-20 Richard Biener <rguenther@suse.de>
3054 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
3055 to cover -flto-partition=none.
3057 2018-07-20 Martin Liska <mliska@suse.cz>
3059 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
3060 (get_decl_source_range): Remove unused function.
3062 2018-07-20 Richard Biener <rguenther@suse.de>
3064 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
3065 (struct vn_phi_s): Likewise.
3066 (struct vn_reference_s): Likewise.
3067 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
3068 for searching the slot of an entry known to be in the hash itself.
3069 (vn_phi_hasher::equal): Likewise.
3070 (vn_reference_hasher::equal): Likewise.
3071 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
3073 (optimistic_info, current_info): Remove, keeping only valid_info.
3074 (vn_reference_lookup_1): Remove fallback lookup.
3075 (vn_reference_lookup_2): Likewise.
3076 (vn_nary_op_lookup_1): Likewise.
3077 (vn_phi_lookup): Likewise.
3078 (vn_nary_build_or_lookup_1): Make sure to not chain the built
3080 (vn_reference_insert): Adjust, chain the inserted hash element
3081 at last_inserted_ref.
3082 (vn_reference_insert_pieces): Likewise.
3083 (visit_reference_op_call): Likewise.
3084 (vn_nary_op_insert_into): Chain the inserted hash element at
3086 (vn_nary_op_insert_pieces): Adjust.
3087 (vn_nary_op_insert): Likewise.
3088 (vn_nary_op_insert_stmt): Likewise.
3089 (vn_phi_insert): Adjust, chain the inserted hash element at
3091 (process_scc): Remove clearing and copying the optimistic
3092 table. Instead remove elements inserted during an optimistic
3093 iteration from the single table we maintain.
3094 (init_scc_vn): Adjust.
3095 (free_scc_vn): Likewise.
3096 (sccvn_dom_walker::record_cond): Likewise.
3097 (sccvn_dom_walker::after_dom_children): Likewise.
3099 2018-07-19 Martin Sebor <msebor@redhat.com>
3101 PR tree-optimization/84047
3102 PR tree-optimization/83776
3103 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
3104 (check_array_bounds): Call it.
3106 2018-07-19 Martin Sebor <msebor@redhat.com>
3108 * align.h (align_flags): Use member initialization.
3110 2018-07-19 David Malcolm <dmalcolm@redhat.com>
3112 * Makefile.in (OBJS): Add optinfo.o.
3113 * coretypes.h (class symtab_node): New forward decl.
3114 (struct cgraph_node): New forward decl.
3115 (class varpool_node): New forward decl.
3116 * dump-context.h: New file.
3117 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3119 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3120 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3121 (set_alt_dump_file): Likewise.
3122 (dump_context::~dump_context): New dtor.
3123 (dump_gimple_stmt): Move implementation to...
3124 (dump_context::dump_gimple_stmt): ...this new member function.
3125 Add the stmt to any pending optinfo, creating one if need be.
3126 (dump_gimple_stmt_loc): Move implementation to...
3127 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3128 Start a new optinfo and add the stmt to it.
3129 (dump_gimple_expr): Move implementation to...
3130 (dump_context::dump_gimple_expr): ...this new member function.
3131 Add the stmt to any pending optinfo, creating one if need be.
3132 (dump_gimple_expr_loc): Move implementation to...
3133 (dump_context::dump_gimple_expr_loc): ...this new member function.
3134 Start a new optinfo and add the stmt to it.
3135 (dump_generic_expr): Move implementation to...
3136 (dump_context::dump_generic_expr): ...this new member function.
3137 Add the tree to any pending optinfo, creating one if need be.
3138 (dump_generic_expr_loc): Move implementation to...
3139 (dump_context::dump_generic_expr_loc): ...this new member
3140 function. Add the tree to any pending optinfo, creating one if
3142 (dump_printf): Move implementation to...
3143 (dump_context::dump_printf_va): ...this new member function. Add
3144 the text to any pending optinfo, creating one if need be.
3145 (dump_printf_loc): Move implementation to...
3146 (dump_context::dump_printf_loc_va): ...this new member function.
3147 Start a new optinfo and add the stmt to it.
3148 (dump_dec): Move implementation to...
3149 (dump_context::dump_dec): ...this new member function. Add the
3150 value to any pending optinfo, creating one if need be.
3151 (dump_context::dump_symtab_node): New member function.
3152 (dump_context::get_scope_depth): New member function.
3153 (dump_context::begin_scope): New member function.
3154 (dump_context::end_scope): New member function.
3155 (dump_context::ensure_pending_optinfo): New member function.
3156 (dump_context::begin_next_optinfo): New member function.
3157 (dump_context::end_any_optinfo): New member function.
3158 (dump_context::s_current): New global.
3159 (dump_context::s_default): New global.
3160 (dump_scope_depth): Delete global.
3161 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3162 (dump_symtab_node): New function.
3163 (get_dump_scope_depth): Reimplement in terms of dump_context.
3164 (dump_begin_scope): Likewise.
3165 (dump_end_scope): Likewise.
3166 (selftest::temp_dump_context::temp_dump_context): New ctor.
3167 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3168 (selftest::verify_item): New function.
3169 (ASSERT_IS_TEXT): New macro.
3170 (ASSERT_IS_TREE): New macro.
3171 (ASSERT_IS_GIMPLE): New macro.
3172 (selftest::test_capture_of_dump_calls): New test.
3173 (selftest::dumpfile_c_tests): Call it.
3174 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3175 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3176 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3177 descriptive comment.
3178 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3179 (dump_node, dump_bb): Move these unrelated decls.
3180 (class dump_manager): Add leading comment.
3181 * optinfo.cc: New file.
3182 * optinfo.h: New file.
3184 2018-07-19 Michael Collison <michael.collison@arm.com>
3185 Richard Henderson <rth@redhat.com>
3187 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3188 (subti): Handle op1 zero.
3189 (subvti4, usub4ti4): New.
3190 (*sub<GPI>3_compare1_imm): New.
3191 (sub<GPI>3_carryinCV): New.
3192 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3193 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3195 2018-07-19 Michael Collison <michael.collison@arm.com>
3196 Richard Henderson <rth@redhat.com>
3198 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3199 (addti3): Create simpler code if low part is already known to be 0.
3200 (addvti4, uaddvti4): New.
3201 (*add<GPI>3_compareC_cconly_imm): New.
3202 (*add<GPI>3_compareC_cconly): New.
3203 (*add<GPI>3_compareC_imm): New.
3204 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3205 handle constants within this pattern..
3206 (*add<GPI>3_compareV_cconly_imm): New.
3207 (*add<GPI>3_compareV_cconly): New.
3208 (*add<GPI>3_compareV_imm): New.
3209 (add<GPI>3_compareV): New.
3210 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3211 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3212 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3213 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3214 with 'comparison' operator.
3215 (*add<GPI>3_compareV_cconly_imm): Ditto.
3216 (*add<GPI>3_compareV_cconly): Ditto.
3217 (*add<GPI>3_compareV_imm): Ditto.
3218 (add<GPI>3_compareV): Ditto.
3219 (add<mode>3_carryinC): Ditto.
3220 (*add<mode>3_carryinC_zero): Ditto.
3221 (*add<mode>3_carryinC): Ditto.
3222 (add<mode>3_carryinV): Ditto.
3223 (*add<mode>3_carryinV_zero): Ditto.
3224 (*add<mode>3_carryinV): Ditto.
3226 2018-07-19 Michael Collison <michael.collison@arm.com>
3227 Richard Henderson <rth@redhat.com>
3229 * config/aarch64/aarch64-modes.def (CC_V): New.
3230 * config/aarch64/aarch64-protos.h
3231 (aarch64_addti_scratch_regs): Declare
3232 (aarch64_subvti_scratch_regs): Declare.
3233 (aarch64_expand_subvti): Declare.
3234 (aarch64_gen_unlikely_cbranch): Declare
3235 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3236 for signed overflow using CC_Vmode.
3237 (aarch64_get_condition_code_1): Handle CC_Vmode.
3238 (aarch64_gen_unlikely_cbranch): New function.
3239 (aarch64_addti_scratch_regs): New function.
3240 (aarch64_subvti_scratch_regs): New function.
3241 (aarch64_expand_subvti): New function.
3243 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3245 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3247 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3248 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3251 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3254 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3255 address check not strict.
3257 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3259 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3261 (store_pair_lanes<mode>): Likewise.
3262 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3263 enum value 'ADDR_QUERY_LDP_STP_N'.
3264 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3265 (aarch64_print_address_internal): Add declaration.
3266 (aarch64_print_ldpstp_address): Remove.
3267 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3268 (aarch64_print_operand): Change printing of 'y'.
3269 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3270 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3271 'true' rather than '1'.
3272 * config/aarch64/constraints.md (Uml): Likewise.
3273 (Uml): Rename to Umn.
3276 2018-07-19 Richard Biener <rguenther@suse.de>
3278 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3280 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3281 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3282 (vn_reference_hasher): Likewise.
3283 (struct vn_tables_s): Remove obstack and alloc-pool members.
3284 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3285 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3286 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3287 (vn_reference_insert_pieces): Likewise.
3288 (alloc_vn_nary_op_noinit): Adjust.
3289 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3290 (vn_phi_eq): Adjust.
3291 (shared_lookup_phiargs): Remove.
3292 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3293 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3294 (visit_reference_op_call): Likewise.
3295 (copy_nary, copy_phi, copy_reference): Remove.
3296 (process_scc): Rewind the obstack when iterating. Do not
3297 copy the elements to valid_info but just move them from one
3298 hashtable to the other.
3299 (allocate_vn_table): Adjust.
3300 (free_vn_table): Likewise.
3301 (init_scc_vn): Likewise.
3302 (free_scc_vn): Likewise.
3304 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3307 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3308 indirect_return as function type attribute.
3309 (ix86_attribute_table): Change indirect_return to function
3311 * doc/extend.texi: Update indirect_return attribute.
3313 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3315 * wide-int.h (widest2_int): New.
3316 * gimple-fold.c (arith_overflowed_p): Use it.
3317 * tree.h (widest2_int_cst): New.
3318 * tree-vrp.c (wide_int_binop_overflow): Rename from
3319 vrp_int_const_binop.
3320 Rewrite to work on trees.
3321 (extract_range_from_multiplicative_op_1): Abstract code to...
3322 (wide_int_range_min_max): ...here.
3323 (wide_int_range_cross_product): ...and here.
3324 (extract_range_from_binary_expr_1): Abstract overflow code to...
3325 (wide_int_range_mult_wrapping): ...here.
3326 * tree-vrp.h (wide_int_range_cross_product): New.
3327 (wide_int_range_mult_wrapping): New.
3329 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3330 Julia Koval <julia.koval@intel.com>
3332 * config/i386/x86-tune-costs.h (skylake_memcpy,
3333 skylake_memset): Replace rep_prefix with unrolling for size 512.
3335 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3338 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3339 comparision with EQ_EXPR in last stmt.
3341 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3343 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3344 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3345 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3346 previously known as "PowerPC AltiVec Built-in Functions". Move
3347 some material to new subsubsections "PowerPC AltiVec Built-in
3348 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3350 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3351 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3352 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3353 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3355 2018-07-18 Richard Biener <rguenther@suse.de>
3357 PR tree-optimization/86557
3358 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3361 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3363 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3365 2018-07-17 Jeff Law <law@redhat.com>
3367 * config/arm/arm.c (get_label_padding): Update for recent
3368 changes to label_to_alignment.
3370 PR tree-optimization/86010
3371 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3373 * config/mips/mips.c (vr4130_align_insns): Update for recent
3374 changes to label_to_alignment.
3376 * config/frv/frv.c (frv_label_align): Update for recent changes
3377 to label_to_alignment.
3379 * config/nios2/nios2.c (nios2_label_align): Update for recent
3380 changes which dropped ALIGN_LABELS_LOG.
3382 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3384 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3385 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3387 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3389 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3390 deprecated align_labels_log.
3392 2018-07-17 Richard Biener <rguenther@suse.de>
3395 * dwarf2out.c (init_sections_and_labels): Always generate
3396 a debug_line_str_section for early LTO debug.
3397 (dwarf2out_finish): Reset debug_line_str_hash output early.
3398 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3400 (dwarf2out_early_finish): Output debug_line_str.
3402 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3404 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3405 index register on z196 or later.
3407 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3409 * config/s390/s390.c (s390_default_align): Set default function
3411 (s390_override_options_after_change): Call s390_default align.
3412 (s390_option_override_internal): Call s390_default_align.
3413 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3415 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3418 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3419 to _looptemp_ clauses, other than the first two.
3421 2018-07-17 Martin Liska <mliska@suse.cz>
3423 * opts.c: Do not enable OPT_falign_* for -Os.
3425 2018-07-17 Martin Liska <mliska@suse.cz>
3427 * align.h (MAX_CODE_ALIGN): New.
3428 (MAX_CODE_ALIGN_VALUE): New.
3429 * common/config/i386/i386-common.c (ix86_handle_option):
3430 (MAX_CODE_ALIGN): Moved to align.h.
3431 * final.c (MAX_CODE_ALIGN): Likewise.
3432 * opts.c (parse_and_check_align_values):
3433 (MAX_CODE_ALIGN): Likewise.
3434 (MAX_CODE_ALIGN_VALUE): Likewise.
3436 2018-07-17 Martin Liska <mliska@suse.cz>
3438 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3439 in order to fulfil coding style.
3440 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3441 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3442 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3443 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3444 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3445 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3446 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3447 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3448 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3450 2018-07-17 Martin Liska <mliska@suse.cz>
3452 * align.h: New file.
3453 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3455 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3456 align_flags of label_to_alignment.
3457 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3459 * config/m68k/m68k.c: Do not use removed align_labels_value and
3461 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3462 (LOOP_ALIGN): Likewise.
3463 (LABEL_ALIGN): Likewise.
3464 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3465 Remove not used macro.
3466 (rs6000_loop_align): Change return type to align_flags.
3467 (rs6000_loop_align_max_skip): Remove.
3468 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3469 Change return type to align_flags.
3470 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3471 Remove not used macro.
3472 (rs6000_loop_align): Change return type to align_flags.
3473 (rs6000_loop_align_max_skip): Remove.
3474 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3475 * config/rx/rx-protos.h (rx_align_for_label): Make it
3477 * config/rx/rx.c (rx_align_for_label): Change return type
3479 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3481 into align_flags class.
3482 (LABEL_ALIGN): Likewise.
3483 (LOOP_ALIGN): Likewise.
3484 * config/s390/s390.c (s390_label_align): Use align_flags
3486 (s390_asm_output_function_label): Likewise.
3487 * config/sh/sh.c (sh_override_options_after_change):
3488 Use align_flags class directly without macros.
3489 (find_barrier): Likewise.
3490 (barrier_align): Likewise.
3491 (sh_loop_align): Likewise.
3492 * config/spu/spu.c (spu_option_override):
3493 Use align_flags_tuple::get_value instead of removed macros.
3494 (spu_sched_init): Likewise.
3495 * config/spu/spu.h (GTY): Likewise.
3496 * config/visium/visium.c (visium_option_override):
3497 Set "8" as default secondary alignment.
3498 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3499 in order to guarantee secondary alignment of 8.
3500 * coretypes.h: Include align.h header file.
3501 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3502 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3503 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3504 * doc/tm.texi.in: Likewise.
3505 * final.c (struct label_alignment): Remove not used structure.
3506 (LABEL_ALIGN): Change type to align_flags.
3507 (LOOP_ALIGN): Likewise.
3508 (JUMP_ALIGN): Likewise.
3509 (default_loop_align_max_skip): Remove.
3510 (default_label_align_max_skip): Likewise.
3511 (default_jump_align_max_skip): Likewise.
3512 (default_label_align_after_barrier_max_skip):
3513 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3514 (LABEL_TO_MAX_SKIP): Remove.
3515 (label_to_alignment): Return align_flags type instead of integer.
3516 (label_to_max_skip): Remove.
3517 (align_fuzz): Use align_flags type.
3518 (compute_alignments): Use align_flags type and use align_flags::max
3519 to combine multiple alignments.
3520 (grow_label_align): Grow vec instead of C array.
3521 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3522 (shorten_branches): Use align_flags type and use align_flags::max
3523 to combine multiple alignments.
3524 (final_scan_insn_1): Remove usage of secondary alignment that comes
3525 from label alignment, but instead use proper secondary alignment
3526 which is computed in grow_label_align.
3527 * flags.h (struct align_flags_tuple): Move to align.h.
3528 (struct align_flags): Likewise.
3529 (state_align_loops): Rename to align_loops.
3530 (state_align_jumps): Rename to align_jumps.
3531 (state_align_labels): Rename to align_labels.
3532 (state_align_functions): Rename to align_functions.
3533 (align_loops_log): Remove.
3534 (align_jumps_log): Remove.
3535 (align_labels_log): Remove.
3536 (align_functions_log): Remove.
3537 (align_loops_max_skip): Remove.
3538 (align_jumps_max_skip): Remove.
3539 (align_labels_max_skip): Remove.
3540 (align_functions_max_skip): Remove.
3541 (align_loops_value): Remove.
3542 (align_jumps_value): Remove.
3543 (align_labels_value): Remove.
3544 (align_functions_value): Remove.
3545 * output.h (label_to_alignment): Change return type to align_flags.
3546 (label_to_max_skip): Remove.
3547 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3548 jump_align_max_skip macros.
3549 * targhooks.h (default_loop_align_max_skip): Remove.
3550 (default_label_align_max_skip): Likewise.
3551 (default_jump_align_max_skip): Likewise.
3552 (default_label_align_after_barrier_max_skip): Remove.
3553 * toplev.c (read_log_maxskip): Use ::normalize function.
3554 (parse_N_M): Remove not used argument and also call ::normalize.
3555 (parse_alignment_opts): Do not pass unused arguments.
3556 * varasm.c (assemble_start_function): Use directly align_functions
3557 instead of removed macros.
3558 * system.h: Do not poison removed macros.
3560 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3563 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3564 and cond temporaries don't have reference type if iterator has
3565 pointer type. For init use &for_pre_body instead of pre_p if
3566 for_pre_body is non-empty.
3568 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3570 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3571 double-double modes to SFmode directly directly.
3572 (trunc<mode>sf2_fprs): Delete.
3574 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3576 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3577 for conversions between IFmode and the decimal floating point modes.
3578 (init_float128_ieee): Use the correct names for conversions between
3579 KFmode and the decimal floating point modes.
3581 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3583 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3584 for the conversions between TDmode and IFmode.
3585 (init_float128_ieee): Use more correct names for the conversions
3586 between TDmode and KFmode.
3588 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3590 PR tree-optimization/86526
3591 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3592 (inline_expand_builtin_string_cmp): Likewise.
3593 (inline_string_cmp): Likewise. Use c_readstr instead of
3594 builtin_memcpy_read_str. Add unit_mode temporary.
3596 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3599 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3600 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3602 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3604 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3605 Alphabetize prototypes of built-in functions, separating out
3606 built-in functions that are listed in this section but should be
3607 described elsewhere.
3609 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3612 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3613 conditional move when trapping comparison was split to a
3614 non-trapping one (and vice versa).
3616 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3618 * config/s390/s390.c (s390_function_profiler): Generate nops
3619 instead of profiler call sequences.
3620 * config/s390/s390.opt: Add the new option.
3622 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3624 * config/s390/s390.c (s390_function_profiler): Generate
3625 __mcount_loc section.
3626 * config/s390/s390.opt: Add the new option.
3628 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3630 * common.opt: Add the new warning.
3631 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3632 %r0,__fentry__" when -mfentry is specified.
3633 (s390_option_override_internal): Disallow -mfentry for 31-bit
3635 * config/s390/s390.opt: Add the new option.
3637 2018-07-16 Richard Biener <rguenther@suse.de>
3640 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3641 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3643 2018-07-16 Martin Liska <mliska@suse.cz>
3646 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3649 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3651 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3653 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3655 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3657 * config/arc/arc-cpus.def (hs4x): New cpu.
3659 * config/arc/arc-tables.opt: Regenerate.
3660 * config/arc/arc.c (arc_sched_issue_rate): New function.
3661 (TARGET_SCHED_ISSUE_RATE): Define.
3662 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3663 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3665 (attr tune): Add ARCHS4x tune values.
3666 (attr tune_dspmpy): Define.
3667 (*tst): Correct instruction type.
3668 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3669 * config/arc/arcHS4x.md: New file.
3670 * config/arc/fpu.md: Update instruction type attributes.
3671 * config/arc/t-multilib: Regenerate.
3673 2018-07-16 Tom de Vries <tdevries@suse.de>
3676 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3678 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3680 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3683 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3685 PR tree-optimization/86514
3686 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3687 conversion to a boolean type from a type with greater precision.
3689 2018-07-16 Tom de Vries <tdevries@suse.de>
3691 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3692 dump_flags request TDF_SLIM.
3694 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3696 * fold-const.c (int_const_binop_1): Abstract...
3697 (wide_int_binop): ...wide int code here.
3698 (poly_int_binop): ...poly int code here.
3699 Abstract the rest of int_const_binop_1 into int_const_binop.
3700 * fold-const.h (wide_int_binop): New.
3701 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3702 Remove useless PLUS/MINUS_EXPR case.
3703 (zero_nonzero_bits_from_vr): Move wide int code...
3704 (zero_nonzero_bits_from_bounds): ...here.
3705 (extract_range_from_binary_expr_1): Move mask optimization code...
3706 (range_easy_mask_min_max): ...here.
3707 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3708 (range_easy_mask_min_max): New.
3710 2018-07-15 Jeff Law <law@redhat.com>
3713 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3716 2018-07-14 Jim Wilson <jimw@sifive.com>
3718 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3720 2018-07-14 Paul Koning <ni1d@arrl.net>
3722 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3724 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3726 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3729 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3730 Steve Munroe <munroesj52@gmail.com>
3732 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3733 (_mm_andnot_si128): Likewise.
3734 (_mm_or_si128): Likewise.
3735 (_mm_xor_si128): Likewise.
3737 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3740 * builtins.c (expand_builtin_memcmp): Inline the calls first
3741 when result_eq is false.
3742 (expand_builtin_strcmp): Inline the calls first.
3743 (expand_builtin_strncmp): Likewise.
3744 (inline_string_cmp): New routine. Expand a string compare
3745 call by using a sequence of char comparison.
3746 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3747 a call to str(n)cmp/memcmp.
3748 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3750 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3752 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3754 * config/arm/driver-arm.c: Include arm-native.h.
3755 (host_detect_local_cpu): Use auto-generated data tables.
3756 (vendors, arm_cpu_table): Delete. Move part information to ...
3757 * config/arm/arm-cpus.in: ... here.
3758 * config/arm/parsecpu.awk (gen_native): New function.
3759 (vendor, part): New CPU fields.
3760 (END): Add support for building the native CPU detection tables.
3761 * config/arm/t-arm (arm-native.h): Add build rule.
3762 (driver-arm.o): Add dependency on arm-native.h.
3764 2018-07-13 Richard Biener <rguenther@suse.de>
3767 * match.pd (addr1 - addr2): Allow either of the operand to
3770 2018-07-13 Tom de Vries <tdevries@suse.de>
3772 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3775 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3777 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3778 arrays instead of numbered variables.
3780 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3782 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3783 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3785 2018-07-13 Richard Biener <rguenther@suse.de>
3788 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3789 instead of get_context_die.
3791 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3792 Richard Biener <rguenther@suse.de>
3795 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3796 that the loop latch destination where phi is defined.
3798 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3800 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3801 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3802 (riscv_set_current_function): Extract getting interrupt type to new
3804 (riscv_get_interrupt_type): New function.
3805 (riscv_merge_decl_attributes): New function, checking interrupt type is
3807 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3809 2018-07-12 Paul Koning <ni1d@arrl.net>
3811 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3814 2018-07-12 Paul Koning <ni1d@arrl.net>
3816 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3817 zero reference, add doloop_end instead.
3818 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3819 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3820 detail for doloop_end.
3822 2018-07-12 Martin Sebor <msebor@redhat.com>
3825 * attribs.c (decl_attributes): Reject conflicting attributes before
3826 calling attribute handlers.
3828 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3830 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3832 (gcc::dump_manager::get_dump_file_name): likewise.
3833 (dump_begin): Likewise.
3834 * dumpfile.h (dump_begin): Update prototype.
3835 (gcc::dump_manager::get_dump_file_name,
3836 gcc::dump_manager::get_dump_file_name): Update prototype.
3838 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3840 * internal-fn.h (vectorizable_internal_fn_p): New function.
3841 * tree-vect-slp.c (compatible_calls_p): Likewise.
3842 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3843 to internal functions.
3844 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3846 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3848 * fold-const.h (inverse_conditions_p): Declare.
3849 * fold-const.c (inverse_conditions_p): New function.
3850 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3851 that test the inverse condition of a conditional internal function.
3852 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3853 * internal-fn.c (internal_fn_mask_index): Handle conditional
3855 (vectorized_internal_fn_supported_p): New function.
3856 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3857 (any_pred_load_store): Replace with...
3858 (need_to_predicate): ...this new variable.
3859 (redundant_ssa_names): New variable.
3860 (ifcvt_can_use_mask_load_store): Move initial checks to...
3861 (ifcvt_can_predicate): ...this new function. Handle tree codes
3862 for which a conditional internal function exists.
3863 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3864 instead of ifcvt_can_use_mask_load_store. Update after variable
3866 (predicate_load_or_store): New function, split out from
3867 predicate_mem_writes.
3868 (check_redundant_cond_expr): New function.
3869 (value_available_p): Likewise.
3870 (predicate_rhs_code): Likewise.
3871 (predicate_mem_writes): Rename to...
3872 (predicate_statements): ...this. Use predicate_load_or_store
3873 and predicate_rhs_code.
3874 (combine_blocks, tree_if_conversion): Update after above name changes.
3875 (ifcvt_local_dce): Handle redundant_ssa_names.
3876 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3877 general conditional functions.
3878 * tree-vect-stmts.c (vectorizable_call): Likewise.
3880 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3881 Alan Hayward <alan.hayward@arm.com>
3882 David Sherwood <david.sherwood@arm.com>
3884 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3885 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3886 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3887 plus and minus and convert them into IFN_COND_FMA-based sequences.
3888 (convert_mult_to_fma): Handle conditional plus and minus.
3890 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3892 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3893 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3894 (cond_fnms_optab): New optabs.
3895 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3897 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3898 * internal-fn.h (get_conditional_internal_fn): Declare.
3899 (get_unconditional_internal_fn): Likewise.
3900 * internal-fn.c (cond_ternary_direct): New macro.
3901 (expand_cond_ternary_optab_fn): Likewise.
3902 (direct_cond_ternary_optab_supported_p): Likewise.
3903 (FOR_EACH_COND_FN_PAIR): Likewise.
3904 (get_conditional_internal_fn): New function.
3905 (get_unconditional_internal_fn): Likewise.
3906 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3907 (gimple_match_op::gimple_match_op): Add a new overload for 5
3909 (gimple_match_op::set_op): Likewise.
3910 (gimple_resimplify5): Declare.
3911 * genmatch.c (decision_tree::gen): Generate simplifications for
3913 * gimple-match-head.c (gimple_simplify): Define an overload for
3914 5 operands. Handle calls with 5 arguments in the top-level overload.
3915 (convert_conditional_op): Handle conversions from unconditional
3916 internal functions to conditional ones.
3917 (gimple_resimplify5): New function.
3918 (build_call_internal): Pass a fifth operand.
3919 (maybe_push_res_to_seq): Likewise.
3920 (try_conditional_simplification): Try converting conditional
3921 internal functions to unconditional internal functions.
3922 Handle 3-operand unconditional forms.
3923 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3924 Define ternary equivalents of the current rules for binary conditional
3926 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3928 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3929 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3930 (optab): Handle them.
3931 (SVE_COND_FP_TERNARY): New int iterator.
3932 (sve_fmla_op, sve_fmad_op): New int attributes.
3933 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3934 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3935 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3937 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3939 * target.def (preferred_else_value): New target hook.
3940 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3941 * doc/tm.texi: Regenerate.
3942 * targhooks.h (default_preferred_else_value): Declare.
3943 * targhooks.c (default_preferred_else_value): New function.
3944 * internal-fn.h (conditional_internal_fn_code): Declare.
3945 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3946 (get_conditional_internal_fn): Use it.
3947 (conditional_internal_fn_code): New function.
3948 * gimple-match.h (gimple_match_cond): New struct.
3949 (gimple_match_op): Add a cond member function.
3950 (gimple_match_op::gimple_match_op): Update all forms to take a
3952 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3953 for the suboperation, but don't specify a particular else_value.
3954 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3955 (visit_nary_op, visit_reference_op_load): Pass
3956 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3957 * gimple-match-head.c: Include tree-eh.h
3958 (convert_conditional_op): New function.
3959 (maybe_resimplify_conditional_op): Likewise.
3960 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3961 (gimple_resimplify2): Likewise.
3962 (gimple_resimplify3): Likewise.
3963 (gimple_resimplify4): Likewise.
3964 (maybe_push_res_to_seq): Return null for conditional operations.
3965 (try_conditional_simplification): New function.
3966 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3968 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3970 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3972 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3974 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3976 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3978 (hash_tree): Do not hash DECL_FCONTEXT
3979 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3980 Do not stream DECL_FCONTEXT.
3981 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3982 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3984 2018-07-12 Richard Biener <rguenther@suse.de>
3987 * dwarf2out.c (gen_block_die): Only output blocks when they have
3988 at least one !DECL_IGNORED_P variable.
3990 2018-07-12 Richard Biener <rguenther@suse.de>
3993 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3994 Remove -mieee-fp handling.
3996 2018-07-12 Richard Biener <rguenther@suse.de>
3998 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3999 left-over from last patch.
4001 2018-07-12 Jakub Jelinek <jakub@redhat.com>
4003 PR tree-optimization/86492
4004 * gimple-ssa-store-merging.c
4005 (imm_store_chain_info::coalesce_immediate_stores): Call
4006 check_no_overlap even for the merge_overlapping case. Formatting fix.
4008 2018-07-12 Richard Biener <rguenther@suse.de>
4011 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
4012 move possibly trapping operations into the conditional.
4014 2018-07-12 Richard Biener <rguenther@suse.de>
4016 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4017 (vn_lookup_simplify_result): Remove recursion limit applied
4019 (vn_nary_build_or_lookup_1): Adjust.
4020 (try_to_simplify): Likewise.
4021 * gimple-match-head.c (gimple_resimplify1): Instead apply one
4023 (gimple_resimplify2): Likewise.
4024 (gimple_resimplify3): Likewise.
4025 (gimple_resimplify4): Likewise.
4027 2018-07-11 Jakub Jelinek <jakub@redhat.com>
4029 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4030 Use __mmask64 type instead of __mmask8 for __M argument.
4031 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
4032 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
4034 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
4035 __mmask16 for __M argument.
4036 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
4037 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
4038 to __mmask16 instead of __mmask8.
4039 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
4040 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
4041 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
4042 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
4043 instead of __mmask16 for __U argument.
4044 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
4045 __mmask16 instead of __mmask8 for __U argument.
4046 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
4048 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
4050 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
4052 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
4054 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
4056 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
4058 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
4060 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
4062 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
4064 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
4066 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
4067 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
4068 return type as well as __M argument type and all casts from __mmask8
4070 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
4071 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
4072 return type as well as __M argument type and all casts from __mmask8
4074 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
4075 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
4076 return type as well as __M argument type and all casts from __mmask8
4078 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
4079 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
4080 return type as well as __M argument type and all casts from __mmask8
4082 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
4083 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
4086 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
4088 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
4089 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
4092 2018-07-11 Paul Koning <ni1d@arrl.net>
4094 * doc/md.texi (define_subst): Document how multiple occurrences of
4095 the same argument in the replacement pattern are handled.
4097 2018-07-11 Paul Koning <ni1d@arrl.net>
4099 * doc/extend.texi (Common Variable Attributes): Move "mode" into
4101 (Common Type Attributes): Add "mode" attribute.
4103 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
4105 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
4106 stream DECL_ORIGINAL_TYPE.
4107 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
4108 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
4109 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
4110 Do not walk original type.
4111 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
4113 (write_ts_decl_non_common_tree_pointers): Do not stream
4115 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4116 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4118 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4120 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4121 thread twice from the same starting edge.
4123 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4125 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4126 * gimple.c (gimple_call_nonnull_result_p): ...here...
4127 (gimple_call_nonnull_arg): ...and here.
4128 * gimple.h (gimple_call_nonnull_result_p): New.
4129 (gimple_call_nonnull_arg): New.
4131 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4133 * config/arm/arm-cpus.in: Move information from fpu field of each
4134 cpu definition to the isa field.
4135 * config/arm/parsecpu.awk (fpu): Delete match rule.
4136 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4138 2018-07-11 Richard Biener <rguenther@suse.de>
4141 * dwarf2out.c (init_sections_and_labels): Use
4142 output_asm_line_debug_info consistently.
4143 (dwarf2out_early_finish): Likewise.
4144 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4147 2018-07-11 Richard Biener <rguenther@suse.de>
4149 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4150 Rework father_bb setting in a way to avoid propagating constants
4151 multiple times on a loop body.
4153 2018-07-10 Mark Wielaard <mark@klomp.org>
4156 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4158 2018-07-10 Richard Biener <rguenther@suse.de>
4160 * hash-map.h (hash_map::iterator::operator*): Return
4161 references to key and value.
4163 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4166 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4167 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4168 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4169 (gimplify_omp_for): For composite loops, move outer
4170 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4171 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4172 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4173 TREE_LIST for both the original class iterator and the "last" helper
4174 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4175 loop, remember has_decl_expr from outer composite loops for the
4176 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4178 2018-07-09 Martin Sebor <msebor@redhat.com>
4182 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4183 accessing implicitly initialized array elements.
4184 * expr.c (string_constant): Handle string initializers of
4185 character arrays within aggregates.
4186 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4187 Store element offset. As a special case, handle zero size.
4188 (fold_nonarray_ctor_reference): Same.
4189 (fold_ctor_reference): Add argument. Store subobject offset.
4190 * gimple-fold.h (fold_ctor_reference): Add argument.
4192 2018-07-09 Paul Koning <ni1d@arrl.net>
4194 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4195 (pdp11_insn_cost): New function.
4196 (pdp11_md_asm_adjust): New function.
4197 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4198 (pdp11_rtx_costs): Update to match machine better.
4199 (output_addr_const_pdp11): Correct format mismatch warnings.
4200 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4201 * config/pdp11/pdp11.md: General change to add base_cost and/or
4202 length attributes for use by new pdp11_insn_cost function.
4203 (MIN_BRANCH): Correct definition.
4205 (doloop_end): Use standard pattern name for looping pattern.
4206 (doloop_end_nocc): New.
4207 (movsf): Add another constraint alternative.
4208 (zero_extendqihi2): Add constraint alternatives for not in place
4210 (zero_extendhisi2): Remove.
4211 (shift patterns): Add CC handling variants.
4215 (define_peephole2): New peephole to recognize mov that sets CC for
4218 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4220 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4221 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4224 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4227 * real.c (real_nextafter): Return true if result is denormal.
4229 2018-07-09 Martin Liska <mliska@suse.cz>
4231 * common.opt: Add back wrongly removed attribute.
4233 2018-07-09 Richard Biener <rguenther@suse.de>
4236 * dwarf2out.c (gen_block_die): For an early generated DIE
4237 always output high/low PC attributes.
4239 2018-07-09 Tom de Vries <tdevries@suse.de>
4241 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4242 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4243 onto VAR_DECL with abstract origin.
4245 2018-07-07 Jim Wilson <jimw@sifive.com>
4247 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4249 2018-07-07 Tom de Vries <tdevries@suse.de>
4251 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4253 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4255 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4257 (combine_bound): Use wide-int overflow calculation instead of
4259 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4261 * fold-const.c (int_const_binop_2): Same.
4262 (extract_muldiv_1): Same.
4263 (fold_div_compare): Same.
4264 (fold_abs_const): Same.
4266 * poly-int.h (add): Same.
4270 * predict.c (predict_iv_comparison): Same.
4271 * profile-count.c (slow_safe_scale_64bit): Same.
4272 * simplify-rtx.c (simplify_const_binary_operation): Same.
4273 * tree-chrec.c (tree_fold_binomial): Same.
4274 * tree-data-ref.c (split_constant_offset_1): Same.
4275 * tree-if-conv.c (idx_within_array_bound): Same.
4276 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4277 * tree-ssa-phiopt.c (minmax_replacement): Same.
4278 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4279 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4280 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4281 * wide-int.cc (wi::add_large): Same.
4282 (wi::mul_internal): Same.
4283 (wi::sub_large): Same.
4284 (wi::divmod_internal): Same.
4285 * wide-int.h: Change overflow type to overflow_type for neg, add,
4286 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4287 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4288 mul_internal, divmod_internal.
4289 (overflow_type): New enum.
4290 (accumulate_overflow): New.
4292 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4294 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4295 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4297 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4299 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4300 argument is checked for zero before entering loop, avoid checking again.
4302 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4304 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4305 * gimplify.h (generic_expr_could_trap_p): Likwise.
4306 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4308 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4310 PR tree-optimization/86401
4311 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4312 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4313 (fold_bit_and_mask): ... here. New helper function for match.pd.
4314 * fold-const.h (fold_bit_and_mask): Declare.
4315 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4317 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4320 * target.def (translate_mode_attribute): New hook.
4321 * targhooks.h (default_translate_mode_attribute): Declare.
4322 * targhooks.c (default_translate_mode_attribute): New function.
4323 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4324 * doc/tm.texi: Regenerate.
4325 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4326 (rs6000_translate_mode_attribute): New function.
4328 2018-07-06 Paul Koning <ni1d@arrl.net>
4330 * doc/md.texi (define_split): Document DONE and FAIL.
4331 (define_peephole2): Ditto.
4333 2018-07-05 Jeff Law <law@redhat.com>
4335 PR tree-optimization/86010
4336 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4337 both the head and tail of mem* and str* calls.
4339 2018-07-05 Jim Wilson <jimw@sifive.com>
4341 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4343 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4345 * config/aarch64/aarch64-simd.md: correct flags text for
4346 MIN_EXPR replacement.
4348 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4350 * configure: Regenerated.
4352 2018-07-05 Carl Love <cel@us.ibm.com>
4354 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4355 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4356 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4357 VSX_BUILTIN_DOUBLEL_V4SF.
4359 2018-07-05 Martin Sebor <msebor@redhat.com>
4362 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4363 than its domain to compute its the upper bound of a char array.
4365 2018-07-05 Nathan Sidwell <nathan@acm.org>
4367 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4368 * doc/cpp.texi: Update comment.
4369 * doc/tm.texi: Rebuilt.
4370 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4371 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4372 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4373 headers affected by extern "C".
4374 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4375 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4376 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4377 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4378 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4379 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4380 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4381 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4382 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4383 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4384 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4385 config/sparc/sp64-elf.h, config/spu/spu.h,
4386 config/stormy16/stormy16.h, config/v850/v850.h,
4387 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4388 define NO_IMPLICIT_EXTERN_C.
4389 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4391 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4394 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4395 instead of GET_MODE_SIZE when comparing Units.
4397 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4400 * rtlanal.c (set_noop_p): Constrain on mode change,
4401 include hard-reg-set.h
4403 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4405 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4407 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4410 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4413 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4415 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4417 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4420 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4422 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4424 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4427 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4429 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4431 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4432 neg_*_op* variables.
4434 2018-07-04 Martin Liska <mliska@suse.cz>
4436 * tree-switch-conversion.c: Define
4437 max_ratio_for_speed and max_ratio_for_size constants.
4439 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4440 Martin Liska <mliska@suse.cz>
4445 * common.opt: Rename align options with 'str_' prefix.
4446 * common/config/i386/i386-common.c (set_malign_value): New
4448 (ix86_handle_option): Use it to set -falign-* options/
4449 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4450 type from int to string.
4451 * config/aarch64/aarch64.c: Update default values from int
4453 * config/alpha/alpha.c (alpha_override_options_after_change):
4455 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4456 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4457 max skip conditionally.
4458 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4459 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4460 max skip conditionally.
4461 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4462 max skip conditionally.
4463 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4464 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4465 max skip conditionally.
4466 * config/i386/i386.c (struct ptt): Change type from int to
4468 (ix86_default_align): Set default values.
4469 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4470 max skip conditionally.
4471 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4472 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4473 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4474 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4475 max skip conditionally.
4476 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4477 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4478 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4479 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4480 max skip conditionally.
4481 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4482 * config/ia64/ia64.c (ia64_option_override): Set default values
4483 for alignment options.
4484 * config/m68k/m68k.c: Handle new str_align_* options.
4485 * config/mips/mips.c (mips_set_compression_mode): Change
4487 (mips_option_override): Set default values for options.
4488 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4490 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4492 * config/rx/rx.c (rx_option_override): Likewise.
4493 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4494 (LABEL_ALIGN): Use align_labels_log.
4495 (LOOP_ALIGN): Use align_loops_align.
4496 * config/s390/s390.c (s390_asm_output_function_label): Use new
4498 * config/sh/sh.c (sh_override_options_after_change):
4499 Change type of constants.
4500 * config/spu/spu.c (spu_sched_init): Likewise.
4501 * config/sparc/sparc.c (sparc_option_override): Set default
4503 * config/visium/visium.c (visium_option_override): Likewise.
4504 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4505 emit p2align format with last argument if it's not needed.
4506 * doc/invoke.texi: Document extended format of -falign-*.
4507 * final.c: Use align_labels alignment.
4508 * flags.h (struct target_flag_state): Change type to use
4510 (struct align_flags_tuple): New.
4511 (struct align_flags): Likewise.
4512 (align_loops_log): Redefine macro to use new types.
4513 (align_loops_max_skip): Redefine macro to use new types.
4514 (align_jumps_log): Redefine macro to use new types.
4515 (align_jumps_max_skip): Redefine macro to use new types.
4516 (align_labels_log): Redefine macro to use new types.
4517 (align_labels_max_skip): Redefine macro to use new types.
4518 (align_functions_log): Redefine macro to use new types.
4519 (align_loops): Redefine macro to use new types.
4520 (align_jumps): Redefine macro to use new types.
4521 (align_labels): Redefine macro to use new types.
4522 (align_functions): Redefine macro to use new types.
4523 (align_functions_max_skip): Redefine macro to use new types.
4524 (align_loops_value): New macro.
4525 (align_jumps_value): New macro.
4526 (align_labels_value): New macro.
4527 (align_functions_value): New macro.
4528 * function.c (invoke_set_current_function_hook): Propagate
4529 alignment values from flags to global variables default in
4531 * ipa-icf.c (sem_function::equals_wpa): Use
4532 cl_optimization_option_eq instead of memcmp.
4533 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4535 (cl_optimization_stream_in): Likewise.
4536 * optc-save-gen.awk: Support strings in cl_optimization.
4537 * opth-gen.awk: Likewise.
4538 * opts.c (finish_options): Remove error checking of invalid
4540 (MAX_CODE_ALIGN): Remove.
4541 (MAX_CODE_ALIGN_VALUE): Likewise.
4542 (parse_and_check_align_values): New function.
4543 (check_alignment_argument): Likewise.
4544 (common_handle_option): Use check_alignment_argument.
4545 * opts.h (parse_and_check_align_values): Declare.
4546 * toplev.c (init_alignments): Remove.
4547 (read_log_maxskip): New.
4548 (parse_N_M): Likewise.
4549 (parse_alignment_opts): Likewise.
4550 (backend_init_target): Remove usage of init_alignments.
4551 * toplev.h (parse_alignment_opts): Declare.
4552 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4554 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4555 * tree.c (cl_option_hasher::equal): New.
4556 * varasm.c: Use new global macros.
4558 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4560 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4561 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4562 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4563 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4564 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4565 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4566 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4567 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4568 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4569 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4570 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4571 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4572 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4574 2018-07-04 Martin Liska <mliska@suse.cz>
4575 Jonathan Wakely <jwakely@redhat.com>
4577 * coverage.c: Use correct type.
4578 * doc/invoke.texi: Language correction.
4580 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4583 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4584 ENDBRANCH for non-tail call which may return via indirect branch.
4585 * doc/extend.texi: Document indirect_return attribute.
4587 2018-07-03 Martin Sebor <msebor@redhat.com>
4589 PR tree-optimization/86274
4590 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4592 (format_floating): Correct handling of infinities and NaNs.
4594 2018-07-03 Martin Sebor <msebor@redhat.com>
4596 * print-tree.c (print_real_cst): New function.
4597 (print_node_brief): Call it.
4598 (print_node): Ditto.
4600 2018-07-03 Jeff Law <law@redhat.com>
4602 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4603 into a single pattern.
4605 * config/h8300/h8300.md (ors code_iterator): New.
4606 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4607 a single pattern and single splitter.
4608 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4609 (iorqi3_1, xorqi3_1): Likewise.
4610 (iorqi3, xorqi3 expanders): Similarly.
4612 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4613 (movmd_internal) into a single pattern using the P mode iterator.
4614 (movmd splitters): Similarly.
4615 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4616 (movsd splitters): Similarly.
4618 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4619 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4620 right table based on the size of the operand.
4621 * config/h8300/h8300.md (length_table): Corresponding changes. All
4622 references to "addb", "addw" and "addl" changed to "add".
4623 (btst patterns): Merge two variants into a single pattern.
4624 (tstqi, tsthi): Likewise.
4625 (addhi3_incdec, addsi3_incdec): Likewise.
4626 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4627 (mulhi3, mulsi3): Likewise.
4628 (udivhi3, udivsi3): Likewise.
4629 (divhi3, divsi3): Likewise.
4630 (andorqi3, andorhi3, andorsi3): Likewise.
4632 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4635 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4636 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4638 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4640 PR tree-optimization/85694
4641 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4642 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4644 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4645 (<u>avg<mode>3_ceil): New patterns.
4647 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4649 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4650 scan-tree-dump directive.
4651 * gcc.dg/vect/slp-perm-2.c: Likewise.
4652 * gcc.dg/vect/slp-perm-3.c: Likewise.
4653 * gcc.dg/vect/slp-perm-5.c: Likewise.
4654 * gcc.dg/vect/slp-perm-6.c: Likewise.
4655 * gcc.dg/vect/slp-perm-7.c: Likewise.
4656 * gcc.dg/vect/slp-perm-8.c: Likewise.
4658 2018-07-03 Marek Polacek <polacek@redhat.com>
4661 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4663 2018-07-03 Richard Biener <rguenther@suse.de>
4666 * tree-ssa-structalias.c (find_func_clobbers): Properly
4667 handle indirect calls.
4669 2018-07-03 Jeff Law <law@redhat.com>
4671 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4672 (shifts): New code iterator.
4673 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4674 expander. Fix HImode handling on H8/SX.
4675 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4676 (subqi3, subhi3, subsi3 expanders): Likewise.
4677 (andqi3, andhi3, andsi3 expanders): Likewise.
4678 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4679 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4680 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4681 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4682 (zero_extendqihi2, zero_extendqisi2): Likewise.
4683 (extendqihi2, extendqisi2): Likewise.
4684 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4685 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4686 (rotlqi3_1, rotlhi3_1): Likewise.
4687 (logicalhi3_sn, logicalsi3_sn): Likewise.
4688 (logicalhi3, logicalsi3): Likewise.
4690 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4692 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4693 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4694 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4695 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4696 type to append_pattern_def_seq instead of creating a stmt_vec_info
4698 (build_mask_conversion): Likewise. Remove vinfo argument.
4699 (vect_add_conversion_to_patterm): Likewise, renaming to...
4700 (vect_add_conversion_to_pattern): ...this.
4701 (vect_recog_mask_conversion_pattern): Update call to
4702 build_mask_conversion. Pass the vector type to
4703 append_pattern_def_seq here too.
4704 (vect_recog_gather_scatter_pattern): Update call to
4705 vect_add_conversion_to_pattern.
4707 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4709 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4710 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4711 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4712 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4713 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4714 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4715 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4716 append_pattern_def_seq instead of new_pattern_def_seq.
4717 (vect_recog_divmod_pattern): Do both of the above.
4718 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4721 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4723 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4724 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4725 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4726 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4727 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4728 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4729 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4730 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4731 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4732 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4733 parameter with a single stmt_vec_info.
4734 (vect_recog_func_ptr): Likewise.
4735 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4736 (vect_try_gather_scatter_pattern): ...this.
4737 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4738 the stmt_vec_info of the statement to be matched. Don't clear
4739 STMT_VINFO_RELATED_STMT.
4740 (vect_pattern_recog): Update call accordingly.
4742 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4744 PR tree-optimization/85694
4745 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4746 (uavgM3_ceil): Document new optabs.
4747 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4748 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4750 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4751 (savg_ceil_optab): New optabs.
4752 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4753 (vect_vect_recog_func_ptrs): Add it.
4754 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4755 constant directly from the associated lhs.
4757 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4759 * tree-vect-patterns.c (vect_split_statement): New function.
4760 (vect_convert_input): Use it to try to split an existing cast.
4762 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4764 * poly-int.h (print_hex): New function.
4765 * dumpfile.h (dump_dec, dump_hex): Declare.
4766 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4767 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4768 min_input_precision, operation_precision and operation_sign.
4769 * tree-vect-patterns.c (vect_get_range_info): New function.
4770 (vect_same_loop_or_bb_p, vect_single_imm_use)
4771 (vect_operation_fits_smaller_type): Delete.
4772 (vect_look_through_possible_promotion): Add an optional
4773 single_use_p parameter.
4774 (vect_recog_over_widening_pattern): Rewrite to use new
4775 stmt_vec_info infomration. Handle one operation at a time.
4776 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4777 (vect_truncatable_operation_p, vect_set_operation_type)
4778 (vect_set_min_input_precision): New functions.
4779 (vect_determine_min_output_precision_1): Likewise.
4780 (vect_determine_min_output_precision): Likewise.
4781 (vect_determine_precisions_from_range): Likewise.
4782 (vect_determine_precisions_from_users): Likewise.
4783 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4784 (vect_vect_recog_func_ptrs): Put over_widening first.
4786 (vect_pattern_recog): Call vect_determine_precisions.
4788 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4790 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4791 statements that have been replaced by further pattern statements.
4792 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4794 2018-07-03 Richard Biener <rguenther@suse.de>
4796 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4797 always set *dt. Dump vectype in vectype overload.
4798 * dumpfile.h (dump_gimple_expr): New function.
4799 (dump_gimple_expr_loc): Likewise.
4800 * dumpfile.c (dump_gimple_expr): New function.
4801 (dump_gimple_expr_loc): Likewise.
4803 2018-07-02 Jeff Law <law@redhat.com>
4805 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4806 the H8/300, H8/300H and H8/S variants into a single pattern.
4807 (movhi_h8300, movqi_h8300hs): Similarly.
4808 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4809 (QHI mode iterator): New.
4811 * config/h8300/h8300.md: Remove trailing whitespace.
4813 2018-07-02 Jim Wilson <jimw@sifive.com>
4815 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4816 instead of emit_insn for interrupt returns.
4817 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4818 (riscv_sret, riscv_uret): Likewise.
4820 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4822 * pretty-print.c (selftest::test_pp_format): Move save and restore
4823 of quotes to class auto_fix_quotes, and add an instance.
4824 * selftest.c: Include "intl.h".
4825 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4826 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4827 * selftest.h (selftest::auto_fix_quotes): New class.
4829 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4831 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4832 (aarch64_sve_prepare_conditional_op): Remove.
4833 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4834 Allow aarch64_simd_reg_or_zero as select operand; remove
4835 the aarch64_sve_prepare_conditional_op call.
4836 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4837 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4838 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4839 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4840 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4841 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4842 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4843 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4844 and a splitters to match all of the *_any patterns.
4845 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4847 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4848 (SVE_COND_FP_BINARY_REV): Remove.
4849 (sve_int_op_rev, sve_fp_op_rev): New.
4850 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4851 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4852 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4853 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4854 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4855 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4856 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4857 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4858 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4860 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4861 Remove match_dup 1 from the inner unspec.
4862 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4864 * config/aarch64/aarch64.md (movprfx): New attr.
4865 (length): Default movprfx to 8.
4866 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4867 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4868 (*<su>mul<SVE_I>3_highpart): Likewise.
4869 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4870 (*v<ASHIFT><SVE_I>3): Likewise.
4871 (*<su><MAXMIN><SVE_I>3): Likewise.
4872 (*<su><MAXMIN><SVE_F>3): Likewise.
4873 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4874 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4875 (*div<SVE_F>4): Likewise.
4877 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4879 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4882 2018-07-02 Richard Biener <rguenther@suse.de>
4884 PR tree-optimization/86363
4885 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4886 memset argument refers to a non-variable address.
4888 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4890 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4891 {PLUS,MINUS}_EXPR code to...
4892 (adjust_symbolic_bound): ...here,
4893 (combine_bound): ...here,
4894 (set_value_range_with_overflow): ...and here.
4896 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4898 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4900 (extract_range_from_abs_expr): ...here.
4902 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4904 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4905 -fno-omit-frame-pointer when not optimizing.
4907 2018-07-02 Martin Liska <mliska@suse.cz>
4910 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4911 (propagate_nothrow): Likewise.
4913 2018-07-02 Martin Liska <mliska@suse.cz>
4916 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4918 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4920 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4921 function in r262149, changing "loc" param from source_location to
4922 const dump_location_t &.
4923 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4924 declaration, as above.
4926 2018-07-01 Paul Koning <ni1d@arrl.net>
4928 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4929 -munit-asm, -mgnu-asm, -mdec-asm.
4930 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4931 (pdp11_output_labelref): New.
4932 (pdp11_output_def): New.
4933 (pdp11_output_addr_vec_elt): New.
4934 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4935 %# and %@ format codes.
4936 (pdp11_option_override): New.
4937 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4938 (pdp11_output_ident): New.
4939 (pdp11_asm_named_section): New.
4940 (pdp11_asm_init_sections): New.
4941 (pdp11_file_start): New.
4942 (pdp11_file_end): New.
4943 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4944 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4945 %o, like %c but octal.
4946 (pdp11_option_override): New.
4947 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4949 (DATA_SECTION_ASM_OP): Ditto.
4950 (READONLY_DATA_SECTION_ASM_OP): New.
4951 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4952 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4953 (ASM_OUTPUT_LABELREF): Ditto.
4954 (ASM_OUTPUT_DEF): Ditto.
4955 (ASM_OUTPUT_EXTERNAL): New.
4956 (ASM_OUTPUT_SOURCE_FILENAME): New.
4957 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4958 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4959 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4960 %# and %@ format codes.
4961 * config/pdp11/pdp11.opt (mgnu-asm): New.
4962 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4963 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4964 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4966 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4968 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4969 dereferencing path[] beyond its length.
4971 (debug_all_paths): New.
4972 (rewire_first_differing_edge): New.
4973 (adjust_paths_after_duplication): New.
4974 (duplicate_thread_path): Call adjust_paths_after_duplication.
4976 (thread_through_all_blocks): Add new argument to
4977 duplicate_thread_path.
4979 2018-06-30 Jim Wilson <jimw@sifive.com>
4981 * config/riscv/predicates.md (p2m1_shift_operand): New.
4982 (high_mask_shift_operand): New.
4983 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4984 pattern using p2m1_shift_operand.
4985 (lshsi3_zero_extend_3+2): New combiner pattern using
4986 high_mask_shift_operand.
4988 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4990 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4992 (vect_recog_rotate_pattern): ...here.
4993 (vect_convert_input): Try to insert casts of invariants in the
4995 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4996 preheader to be empty.
4998 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5000 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
5001 vector type. If given, install it in the new statement's
5003 (vect_element_precision): New function.
5004 (vect_unpromoted_value): New struct.
5005 (vect_unpromoted_value::vect_unpromoted_value): New function.
5006 (vect_unpromoted_value::set_op): Likewise.
5007 (vect_look_through_possible_promotion): Likewise.
5008 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
5009 (vect_widened_op_tree, vect_convert_input): Likewise.
5010 (vect_convert_inputs, vect_convert_output): Likewise.
5011 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
5012 to handle the optional cast of the multiplication result and
5013 vect_widened_op_tree to detect the widened multiplication itself.
5014 Do not require the input and output of promotion casts to have
5015 the same sign, but base the signedness of the operation on the
5016 input rather than the result. If the pattern includes two
5017 promotions, check that those promotions have the same sign.
5018 Do not restrict the MULT_EXPR handling to a double-width result;
5019 handle quadruple-width results and wider. Use vect_convert_inputs
5020 to convert the inputs to the common type.
5021 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
5022 to handle the optional cast of the ABS result. Also allow a sign
5023 change or a sign extension between the ABS and MINUS.
5024 Use vect_widened_op_tree to detect the widened subtraction and use
5025 vect_convert_inputs to convert the inputs to the common type.
5026 (vect_handle_widen_op_by_const): Delete.
5027 (vect_recog_widen_op_pattern): New function.
5028 (vect_recog_widen_mult_pattern): Use it.
5029 (vect_recog_widen_shift_pattern): Likewise.
5030 (vect_recog_widen_sum_pattern): Use
5031 vect_look_through_possible_promotion to handle the promoted
5034 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5036 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
5037 the containing gimple_seq *.
5038 * gimple-iterator.h (gsi_for_stmt): Declare it.
5039 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
5040 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
5041 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5042 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5043 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
5045 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
5047 (vect_mark_pattern_stmts): ...here. Handle cases in which the
5048 statement being replaced is part of an existing pattern
5049 definition sequence, inserting the new pattern statements before
5051 (vect_pattern_recog_1): Don't return a bool. If the statement
5052 is already part of a pattern, instead apply pattern matching
5053 to the pattern definition statements. Don't clear the
5054 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
5055 (vect_pattern_recog): Don't break after the first match;
5056 continue processing the pattern definition statements instead.
5057 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
5059 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5061 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5062 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5063 (vect_recog_widen_sum_pattern): Use it.
5065 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5067 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5068 phi is not a pattern statement and has not been replaced by
5069 a pattern statement.
5070 * tree-vect-patterns.c (type_conversion_p): Don't check
5071 STMT_VINFO_IN_PATTERN_P.
5072 (vect_recog_vector_vector_shift_pattern): Likewise.
5073 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
5074 the pattern statement rather than the original statement; check
5075 directly for a WIDEN_MULT_EXPR here.
5076 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
5077 vect_is_simple_use to return the pattern statement rather
5078 than the original statement; use is_pattern_stmt_p to check
5079 for such a pattern statement.
5080 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
5081 to return the pattern statement rather than the original statement;
5082 don't do the same transformation here.
5083 (vect_is_simple_use): If the defining statement has been replaced
5084 by a pattern statement, return the pattern statement instead.
5085 Remove the corresponding (local) transformation from the vectype
5088 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5090 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
5091 end and default to null.
5092 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5093 (vectorizable_reduction): Update calls accordingly, dropping the
5094 gimple ** argument if the passed-back statement isn't needed.
5095 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
5096 (vect_recog_rotate_pattern): Likewise.
5097 (vect_recog_mask_conversion_pattern): Likewise.
5098 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5099 (vect_mask_constant_operand_p): Likewise.
5100 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
5101 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
5102 (get_group_load_store_type, get_load_store_type): Likewise.
5103 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
5104 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
5105 (vectorizable_conversion, vectorizable_assignment): Likewise.
5106 (vectorizable_shift, vectorizable_operation): Likewise.
5107 (vectorizable_store, vect_is_simple_cond): Likewise.
5108 (vectorizable_condition, vectorizable_comparison): Likewise.
5109 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
5110 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
5111 and move it to the end. Cope with null def_stmt_outs.
5113 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5115 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5117 2018-06-29 Jeff Law <law@redhat.com>
5119 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5120 displacements for TARGET_V850E2V3 and newer.
5121 (TARGET_LRA_P): Remove. Defaults to LRA now.
5122 * config/v850/v850.md (sign23byte_load): Remove.
5123 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5124 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5126 2018-06-29 Martin Liska <mliska@suse.cz>
5129 * coverage.c (coverage_init): Mangle full path name.
5130 * doc/invoke.texi: Document the change.
5131 * gcov-io.c (mangle_path): New.
5132 * gcov-io.h (mangle_path): Likewise.
5133 * gcov.c (mangle_name): Use mangle_path for path mangling.
5135 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5137 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5138 if starting source register is not even.
5140 2018-06-29 Martin Liska <mliska@suse.cz>
5142 PR tree-optimization/86263
5143 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5144 Make edge redirection.
5146 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5148 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5149 (dump_scope_depth): New variable.
5150 (get_dump_scope_depth): New function.
5151 (dump_begin_scope): New function.
5152 (dump_end_scope): New function.
5153 * dumpfile.h (get_dump_scope_depth): New declaration.
5154 (dump_begin_scope): New declaration.
5155 (dump_end_scope): New declaration.
5156 (class auto_dump_scope): New class.
5157 (AUTO_DUMP_SCOPE): New macro.
5158 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5161 2018-06-29 Richard Biener <rguenther@suse.de>
5163 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5164 compute_all_dependences succeeds.
5165 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5166 exceed --param loop-max-datarefs-for-datadeps.
5168 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5170 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5172 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5175 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5176 alternative 0 in preferred_for_speed attribute.
5178 2018-06-28 Paul Koning <ni1d@arrl.net>
5180 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5181 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5182 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5183 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5184 for shift insn patterns.
5186 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5188 * cgraph.c (cgraph_node::get_body): Replace assignments to
5189 "dump_file" with calls to set_dump_file.
5190 * dumpfile.c (alt_dump_file): Make static, and group with...
5191 (alt_flags): ...this definition.
5192 (dumps_are_enabled): New variable.
5193 (refresh_dumps_are_enabled): New function.
5194 (set_dump_file): New function.
5195 (set_alt_dump_file): New function.
5196 (gcc::dump_manager::dump_start): Replace assignments to
5197 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5199 (gcc::dump_manager::dump_finish): Likewise.
5200 * dumpfile.h (alt_dump_file): Delete decl.
5201 (dumps_are_enabled): New variable decl.
5202 (set_dump_file): New function decl.
5203 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5205 * tree-nested.c (lower_nested_functions): Replace assignments to
5206 "dump_file" with calls to set_dump_file.
5208 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5210 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5211 goto_locus of each outgoing edge of each basic block.
5213 2018-06-28 Richard Biener <rguenther@suse.de>
5215 * dwarf2out.c (decl_scope_table): Remove.
5216 (push_decl_scope): Likewise.
5217 (pop_decl_scope): Likewise.
5218 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5219 (gen_struct_or_union_type_die): Likewise.
5220 (gen_tagged_type_die): Likewise.
5221 (dwarf2out_init): Do not initialize decl_scope_table.
5222 (dwarf2out_c_finalize): Do not free it.
5224 2018-06-28 Richard Biener <rguenther@suse.de>
5226 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5227 deciding whether to not re-use a DIE.
5229 2018-06-28 Richard Biener <rguenther@suse.de>
5231 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5232 DW_AT_abstract_origin attribute.
5234 2018-06-28 Martin Liska <mliska@suse.cz>
5236 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5237 Use newly introduced constants.
5238 * tree-switch-conversion.h (struct jump_table_cluster):
5239 Define max_ratio_for_size and max_ratio_for_speed.
5241 2018-06-28 Martin Liska <mliska@suse.cz>
5243 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5244 Add new checking assert to catch invalid state.
5245 (jump_table_cluster::can_be_handled): Handle single case
5247 (jump_table_cluster::is_beneficial): Bail out for such case.
5248 (bit_test_cluster::find_bit_tests):
5249 Add new checking assert to catch invalid state.
5250 (bit_test_cluster::can_be_handled): Handle single case
5252 (bit_test_cluster::is_beneficial): Bail out for such case.
5253 (switch_decision_tree::analyze_switch_statement):
5256 2018-06-28 Martin Liska <mliska@suse.cz>
5258 * common.opt: Introduce -completion option.
5259 * gcc.c (driver_handle_option): Handle it.
5260 (driver::main): Print completions if completion
5262 * opt-suggestions.c (option_proposer::get_completions):
5264 (option_proposer::suggest_completion): Likewise.
5265 (option_proposer::find_param_completions): Likewise.
5266 (verify_autocompletions): Likewise.
5267 (test_completion_valid_options): Likewise.
5268 (test_completion_valid_params): Likewise.
5269 (in_completion_p): Likewise.
5270 (empty_completion_p): Likewise.
5271 (test_completion_partial_match): Likewise.
5272 (test_completion_garbage): Likewise.
5273 (opt_proposer_c_tests): Likewise.
5274 * opt-suggestions.h: Declare new functions.
5275 * opts.c (common_handle_option): Handle OPT__completion_.
5276 * selftest-run-tests.c (selftest::run_tests): Add
5277 opt_proposer_c_tests.
5278 * selftest.c (assert_str_startswith): New.
5279 * selftest.h (assert_str_startswith): Likewise.
5280 (opt_proposer_c_tests): New.
5281 (ASSERT_STR_STARTSWITH): Likewise.
5283 2018-06-28 Martin Liska <mliska@suse.cz>
5285 * Makefile.in: Add opt-suggestions.o.
5286 * gcc-main.c: Include opt-suggestions.h.
5287 * gcc.c (driver::driver): Likewise.
5288 (driver::~driver): Remove m_option_suggestions.
5289 (driver::build_option_suggestions): Moved to option_proposer.
5290 (driver::suggest_option): Likewise.
5291 (driver::handle_unrecognized_options): Use option_proposer.
5292 * gcc.h (class driver): Add new memver m_option_proposer.
5293 * opt-suggestions.c: New file.
5294 * opt-suggestions.h: New file.
5296 2018-06-28 Martin Liska <mliska@suse.cz>
5298 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5299 (auto_string_vec::~auto_string_vec): Likewise.
5301 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5303 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5304 prevent_decl_creation_for_types fields up and add reset_location field.
5305 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5306 statement if id->reset_location is true.
5307 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5308 id->reset_location is true.
5309 (copy_phis_for_bb): Force input_location on the arguments if
5310 id->reset_location is true.
5311 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5312 is set on the function to be inlined.
5314 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5316 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5318 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5320 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5321 registers for Pmode.
5322 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5323 hard registers for the clobbered pseudo.
5325 2018-06-27 Paul Koning <ni1d@arrl.net>
5327 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5328 mutually exclusive options.
5329 * config/pdp11/constraints.md (h): New constraint.
5330 (O): Update definition to match shift code generation.
5331 (D): New constraint.
5332 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5334 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5336 (output_jump): Change arguments.
5337 (pdp11_fixed_cc_regs): New function.
5338 (pdp11_cc_mode): Ditto.
5339 (pdp11_expand_shift): Ditto.
5340 (pdp11_assemble_shift): Ditto.
5341 (pdp11_small_shift): Ditto.
5342 (pdp11_branch_cost): Remove.
5343 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5345 (pdp11_register_move_cost): Update for CC registers.
5346 (pdp11_rtx_costs): Add case for LSHIFTRT.
5347 (pdp11_output_jump): Add CCNZ mode conditional branches.
5348 (notice_update_cc_on_set): Remove.
5349 (pdp11_cc_mode): New function.
5350 (simple_memory_operand): Correct pre/post decrement case.
5351 (no_side_effect_operand): New function.
5352 (pdp11_regno_reg_class): Add CC_REGS class.
5353 (pdp11_fixed_cc_regs): New function.
5354 (pdp11_small_shift): New function.
5355 (pdp11_expand_shift): New function to expand shift insns.
5356 (pdp11_assemble_shift): New function to output shifts.
5357 (pdp11_branch_cost): Remove.
5358 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5359 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5360 (WCHAR_TYPE): Ditto.
5361 (PTRDIFF_TYPE): Ditto.
5362 (ADJUST_INSN_LENGTH): New macro.
5363 (FIXED_REGISTERS): Add CC registers.
5364 (CALL_USED_REGISTERS): Ditto.
5366 (REG_CLASS_NAMES): Ditto.
5367 (REG_CLASS_CONTENTS): Ditto.
5368 (SELECT_CC_MODE): Use new function.
5369 (TARGET_FLAGS_REGNUM): New macro.
5370 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5371 (cc0_reg_rtx): Remove.
5372 (CC_STATUS_MDEP): Remove.
5373 (CC_STATUS_MDEFP_INIT): Remove.
5374 (CC_IN_FPU): Remove.
5375 (NOTICE_UPDATE_CC): Remove.
5376 (REGISTER_NAMES): Add CC registers.
5377 (BRANCH_COST): Change to constant 1.
5378 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5380 * config/pdp11/pdp11.opt (mbcopy): Remove.
5381 (mbcopy-builtin): Remove.
5382 (mbranch-cheap): Remove.
5383 (mbranch-expensive): Remove.
5384 * config/pdp11/predicates.md (expand_shift_operand): Update to
5385 match shift code generation.
5386 (ccnz_operator): New predicate.
5387 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5388 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5389 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5391 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5392 description of O constraint.
5394 2018-06-27 Jeff Law <law@redhat.com>
5395 Austin Law <austinklaw@gmail.com>
5397 * config/v850/v850.md (addsi3_set_flags): New pattern.
5398 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5399 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5400 (zero_extendhisi2_v850_set_flags): Likewise.
5401 (zero_extendqisi2_v850_set_flags): Likewise.
5402 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5403 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5404 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5406 * config/v850/v850-protos.h (notice_update_cc): Remove.
5407 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5408 (v850_print_operand): Handle 'D' and "d".
5409 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5410 Add handling of arithmetic/logical operations compared against zero.
5411 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5412 Do not look at v850_compare_op, instead get mode from last argument.
5413 (v850_gen_compare): Remove
5414 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5415 after reload for prologue insns.
5416 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5418 (construct_save_jarl): Likewise.
5419 (TARGET_FLAGS_REGNUM): Define.
5420 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5421 (NOTICE_UPDATE_CC): Remove.
5422 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5423 than cc0. Conditionalize on reload_completed.
5424 (cmpsi_insn, setfcc_insn): Likewise.
5425 (tst1 splitter): Turn into define_and_split which sets the flags
5427 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5428 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5429 (cstoresf4, cstoredf4): Clobber the flags.
5430 (cmpsi, cmpsf, cmpdf): Remove expanders.
5431 (setf_insn): Remove pattern.
5432 (addsi3): Turn into define_and_split which clobbers the flags after
5433 reload and a suitable pattern (addsi3_clobber_flags) for use after
5435 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5436 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5437 (ashrsi3, ashrsi3_v850e2): Likewise.
5438 (bins): Clobber the flags.
5439 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5440 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5441 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5442 (call_value_internal_short, call_value_internal_long): Likewise.
5443 (callt_save_interrupt, callt_return_interrupt): Likewise.
5444 (save_interrupt, return_interrupt): Likewise.
5445 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5446 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5447 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5448 (All FP comparisons): Only allow after reload has completed.
5450 (divh, divhu): Tweak output template.
5451 (branch_z_normal, branch_z_invert): Remove
5452 (branch_nz_normal, branch_nz_invert): Likewise.
5453 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5455 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5456 * config/v850/v850.c (notice_update_cc): Remove.
5457 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5458 (CC_NO_CARRY): Likewise.
5459 (NOTICE_UPDATE_CC): Define to nothing.
5460 * config/v850/v850.md: Remove block comment on cc0 handling
5461 Remove "cc" attribute from all patterns. Remove cc_status handling
5462 from all patterns. Minor formatting fixes.
5464 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5466 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5467 (cortex-a76.cortex-a55): Likewise.
5468 * config/aarch64/aarch64-tune.md: Regenerate.
5469 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5470 cortex-a76.cortex-a55.
5472 2018-06-27 Jeff Law <law@redhat.com>
5474 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5475 (MULTILIB_DIRNAMES): Similarly.
5477 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5479 * gimple.h (gimple_return_retbnd): Delete.
5480 (gimple_return_set_retbnd): Likewise.
5481 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5482 gimple_return_set_retbnd.
5483 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5484 gimple_return_retbnd and adjust.
5485 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5486 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5487 Explicitly return NULL in a couple more cases. Move assertion
5488 on debug statements and remove unreachable code.
5489 (reset_debug_binding): Do not test id->retbnd.
5490 (expand_call_inline): Do not set it.
5492 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5494 * configure.ac: Add --disable-gcov option.
5495 * configure: Regenerate.
5496 * Makefile.in: Honour @enable_gcov@.
5497 * doc/install.texi: Document --disable-gcov.
5499 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5501 * config/arm/arm-cpus.in (cortex-a76): New entry.
5502 (cortex-a76.cortex-a55): Likewise.
5503 * config/arm/arm-tables.opt: Regenerate.
5504 * config/arm/arm-tune.md: Likewise.
5505 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5506 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5507 cortex-a76.cortex-a55.
5509 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5512 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5514 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5516 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5518 (EPILOGUE_USES): Likewise.
5520 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5522 * tree-inline.c (remap_location): New function extracted from...
5523 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5524 (copy_phis_for_bb): ...here. Call remap_location.
5525 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5527 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5529 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5530 unaligned vsx for 16B memset.
5532 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5535 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5536 ieee128_float_type_node to long_double_type_node unless
5537 TARGET_LONG_DOUBLE_128 is set.
5539 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5541 * cfgloop.c (get_loop_location): Convert return type from
5542 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5543 by implicit construction from rtx_insn *, and using
5544 dump_user_location_t::from_function_decl for the fallback case.
5545 * cfgloop.h (get_loop_location): Convert return type from
5546 location_t to dump_user_location_t.
5547 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5548 dump_printf_loc to pass in a dump_location_t rather than a
5549 location_t, via the gimple stmt.
5550 * coverage.c (get_coverage_counts): Update calls to
5551 dump_printf_loc to pass in dump_location_t rather than a
5553 * doc/optinfo.texi (Dump types): Convert example of
5554 dump_printf_loc from taking "locus" to taking "insn". Update
5555 description of the "_loc" calls to cover dump_location_t.
5556 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5558 (dump_user_location_t::dump_user_location_t): New constructors,
5559 from gimple *stmt and rtx_insn *.
5560 (dump_user_location_t::from_function_decl): New function.
5561 (dump_loc): Make static.
5562 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5563 const dump_location_t &.
5564 (dump_generic_expr_loc): Delete.
5565 (dump_printf_loc): Convert param "loc" from location_t to
5566 const dump_location_t &.
5567 (selftest::test_impl_location): New function.
5568 (selftest::dumpfile_c_tests): New function.
5569 * dumpfile.h: Include "profile-count.h".
5570 (class dump_user_location_t): New class.
5571 (struct dump_impl_location_t): New struct.
5572 (class dump_location_t): New class.
5573 (dump_printf_loc): Convert 2nd param from source_location to
5574 const dump_location_t &.
5575 (dump_generic_expr_loc): Delete.
5576 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5577 const dump_location_t &.
5578 * gimple-fold.c (fold_gimple_assign): Update call to
5579 dump_printf_loc to pass in a dump_location_t rather than a
5580 location_t, via the gimple stmt.
5581 (gimple_fold_call): Likewise.
5582 * gimple-loop-interchange.cc
5583 (loop_cand::analyze_iloop_reduction_var): Update for change to
5584 check_reduction_path.
5585 (tree_loop_interchange::interchange): Update for change to
5587 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5588 change in return-type of find_loop_location.
5589 (graphite_regenerate_ast_isl): Likewise.
5590 * graphite-optimize-isl.c (optimize_isl): Likewise.
5591 * graphite.c (graphite_transform_loops): Likewise.
5592 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5593 pass in a dump_location_t rather than a location_t, via the
5595 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5596 * ipa.c (walk_polymorphic_call_targets): Likewise.
5597 * loop-unroll.c (report_unroll): Convert "locus" param from
5598 location_t to dump_location_t.
5599 (decide_unrolling): Update for change to get_loop_location's
5601 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5602 location_t to dump_user_location_t.
5603 (grid_find_single_omp_among_assignments_1): Updates calls to
5604 dump_printf_loc to pass in a dump_location_t rather than a
5605 location_t, via the gimple stmt.
5606 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5607 location_t to dump_location_t. Updates calls to dump_printf_loc
5608 to pass in a dump_location_t rather than a location_t, via the
5610 (grid_inner_loop_gridifiable_p): Likewise.
5611 (grid_dist_follows_simple_pattern): Likewise.
5612 (grid_gfor_follows_tiling_pattern): Likewise.
5613 (grid_target_follows_gridifiable_pattern): Likewise.
5614 (grid_attempt_target_gridification): Convert initialization
5615 of local "grid" from memset to zero-initialization; FIXME: does
5616 this require C++11? Update call to dump_printf_loc to pass in a
5617 optinfo_location rather than a location_t, via the gimple stmt.
5618 * profile.c (read_profile_edge_counts): Updates call to
5619 dump_printf_loc to pass in a dump_location_t rather than a
5621 (compute_branch_probabilities): Likewise.
5622 * selftest-run-tests.c (selftest::run_tests): Call
5624 * selftest.h (dumpfile_c_tests): New decl.
5625 * tree-loop-distribution.c (pass_loop_distribution::execute):
5626 Update for change in return type of find_loop_location.
5627 * tree-parloops.c (parallelize_loops): Likewise.
5628 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5629 "locus" from location_t to dump_user_location_t.
5630 (canonicalize_loop_induction_variables): Likewise.
5631 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5632 for change in return type of find_loop_location.
5633 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5634 to dump_printf_loc to pass in a dump_location_t rather than a
5635 location_t, via the stmt.
5636 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5638 * tree-vect-loop-manip.c (find_loop_location): Convert return
5639 type from source_location to dump_user_location_t.
5640 (vect_do_peeling): Update for above change.
5641 (vect_loop_versioning): Update for change in type of
5643 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5644 from location_t to dump_user_location_t.
5645 (vect_estimate_min_profitable_iters): Update for change in type
5647 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5648 location_t to dump_location_t.
5649 (vect_slp_bb): Update for change in type of vect_location.
5650 * tree-vectorizer.c (vect_location): Convert from source_location
5651 to dump_user_location_t.
5652 (try_vectorize_loop_1): Update for change in vect_location's type.
5653 (vectorize_loops): Likewise.
5654 (increase_alignment): Likewise.
5655 * tree-vectorizer.h (vect_location): Convert from source_location
5656 to dump_user_location_t.
5657 (find_loop_location): Convert return type from source_location to
5658 dump_user_location_t.
5659 (check_reduction_path): Convert 1st param from location_t to
5660 dump_user_location_t.
5661 * value-prof.c (check_counter): Update call to dump_printf_loc to
5662 pass in a dump_user_location_t rather than a location_t; update
5663 call to error_at for change in type of "locus".
5664 (check_ic_target): Update call to dump_printf_loc to
5665 pass in a dump_user_location_t rather than a location_t, via the
5668 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5670 * config/s390/s390.h (enum processor_flags): Do not use
5671 default tune parameter when -march was specified.
5673 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5676 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5677 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5680 2018-06-26 Richard Biener <rguenther@suse.de>
5682 PR tree-optimization/86287
5684 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5685 (vect_analyze_loop): Initialize n_stmts.
5687 2018-06-26 Richard Biener <rguenther@suse.de>
5690 * fold-const.c (fold_convertible_p): Pointer extension
5693 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5696 * dwarf2out.c (loc_list_has_views): Adjust comments.
5697 (dw_loc_list): Split single cross-partition range with
5700 2018-06-25 Jeff Law <law@redhat.com>
5702 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5703 on -mbig-switch by default.
5705 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5707 (const_float_0_operand): Remove unused predicate.
5708 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5709 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5710 (recipsf2): New expander. Original pattern now called
5712 (recipdf2, recipdf2_insn): Similarly.
5713 (rsqrtsf2, rsqrtsf2_insn): Similarly
5714 (rsqrtdf2, rsqrtdf2_insn): Similarly
5716 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5718 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5719 Simplify logic for FreeBSD (twice).
5721 2018-06-25 Martin Sebor <msebor@redhat.com>
5723 PR tree-optimization/86204
5724 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5725 a strnlen result if it's less than the length of the string.
5727 2018-06-25 Martin Sebor <msebor@redhat.com>
5729 PR tree-optimization/85700
5730 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5731 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5732 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5734 2018-06-25 Martin Sebor <msebor@redhat.com>
5736 * doc/extend.texi (Zero-length arrays): Update and clarify.
5738 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5740 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5741 added IEEE/IBM long double multilib support on PowerPC little
5742 endian Linux systems.
5743 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5744 (MULTILIB_DEFAULTS): Likewise.
5745 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5747 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5748 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5749 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5751 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5754 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5755 (REORDER_45): Likewise.
5757 2018-06-25 Jeff Law <law@redhat.com>
5759 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5760 dividend to 32 bits. Adjust length.
5761 (udivmodhi4): Cleanup output template. Fix length.
5763 2018-06-25 Carl Love <cel@us.ibm.com>
5765 * config/rs6000/vsx.md: Change word selector to prefered location.
5767 2018-06-25 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/86304
5770 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5771 epilogue-if-converted loops as well.
5773 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5775 * lto-section-out.c (lto_begin_section): Do not print section
5776 name for noaddr and unnumbered dumps.
5778 2018-06-25 Richard Biener <rguenther@suse.de>
5780 * tree-vectorizer.h (struct vec_info_shared): New structure
5781 with parts split out from struct vec_info and loop_nest from
5782 struct _loop_vec_info.
5783 (struct vec_info): Adjust accordingly.
5784 (struct _loop_vec_info): Likewise.
5785 (LOOP_VINFO_LOOP_NEST): Adjust.
5786 (LOOP_VINFO_DATAREFS): Likewise.
5787 (LOOP_VINFO_DDRS): Likewise.
5788 (struct _bb_vec_info): Likewise.
5789 (BB_VINFO_DATAREFS): Likewise.
5790 (BB_VINFO_DDRS): Likewise.
5791 (struct _stmt_vec_info): Add dr_aux member.
5792 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5793 (DR_MISALIGNMENT_UNINITIALIZED): New.
5794 (set_dr_misalignment): Adjust.
5795 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5796 (vect_analyze_loop): Adjust prototype.
5797 (vect_analyze_loop_form): Likewise.
5798 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5799 Compute dependences lazily.
5800 (vect_record_base_alignments): Use shared datarefs/ddrs.
5801 (vect_verify_datarefs_alignment): Likewise.
5802 (vect_analyze_data_refs_alignment): Likewise.
5803 (vect_analyze_data_ref_accesses): Likewise.
5804 (vect_analyze_data_refs): Likewise.
5805 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5806 constructor parameter for shared part.
5807 (vect_analyze_loop_form): Pass in shared part and adjust.
5808 (vect_analyze_loop_2): Pass in storage for the number of
5809 stmts. Move loop nest finding to the caller. Compute
5811 (vect_analyze_loop): Pass in shared part.
5812 (vect_transform_loop): Verify shared datarefs are unchanged.
5813 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5814 constructor parameter for shared part.
5815 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5816 (vect_slp_bb): Verify shared datarefs are unchanged before
5818 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5820 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5821 DR_MISALIGNMENT_UNINITIALIZED.
5822 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5823 parameter for shared part.
5824 (vec_info::~vec_info): Adjust.
5825 (vec_info_shared::vec_info_shared): New.
5826 (vec_info_shared::~vec_info_shared): Likewise.
5827 (vec_info_shared::save_datarefs): Likewise.
5828 (vec_info_shared::check_datarefs): Likewise.
5829 (try_vectorize_loop_1): Construct shared part live for analyses
5830 of a single loop for multiple vector sizes.
5831 * tree-parloops.c (gather_scalar_reductions): Adjust.
5833 2018-06-25 Richard Biener <rguenther@suse.de>
5835 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5836 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5837 (vect_analyze_data_refs): Remove similar code from here and
5838 simplify accordingly.
5840 2018-06-25 Richard Biener <rguenther@suse.de>
5842 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5843 for reverse storage order accesses rather than asserting
5844 they cannot happen here.
5846 2018-06-25 Tom de Vries <tdevries@suse.de>
5849 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5850 Use data16 instead of .byte for insn prefix.
5852 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5855 * parser.c (make_char_string_pack): Pass this literal chars
5856 through cpp_interpret_string.
5857 (cp_parser_userdef_numeric_literal): Check the result of
5858 make_char_string_pack.
5860 2018-06-24 Maya Rashish <coypu@sdf.org>
5862 * ginclude/stddef.h: Simplify conditions around avoiding
5863 re-definition of __size_t.
5865 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5867 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5868 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5870 2018-06-22 Maya Rashish <coypu@sdf.org>
5872 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5873 made to OpenBSD and NetBSD through the years.
5875 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5877 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5878 behavior of vec_pack (vector double, vector double) to match
5879 behavior of vec_float2 (vector double, vector double).
5881 2018-06-22 Olivier Hainque <hainque@adacore.com>
5883 * gimplify.c (gimplify_function_tree): Prevent creation
5884 of a trampoline for the address of the current function
5885 passed to entry/exit instrumentation hooks.
5887 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5890 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5893 2018-06-22 Martin Liska <mliska@suse.cz>
5895 PR tree-optimization/86263
5896 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5897 Bail out if is_enabled is false.
5898 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5900 (jump_table_cluster::is_enabled): New function.
5902 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5904 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5905 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5906 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5907 (lto_input_ts_binfo_tree_pointers): Likewise.
5908 * tree-streamer-out.c (streamer_write_tree_bitfields,
5909 write_ts_binfo_tree_pointers): Likewise.
5910 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5912 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5914 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5916 2018-06-22 Martin Liska <mliska@suse.cz>
5918 * symbol-summary.h (get): Make it pure and inline move
5919 functionality from ::get function.
5920 (get): Remove and inline into ::get and ::get_create.
5921 (get_create): Move code from ::get function.
5923 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5926 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5927 -x assembler-with-cpp.
5929 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5931 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5932 _FILE_OFFSET_BITS=64 for C++.
5934 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5936 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5937 conversion insn that shows up when pr85657-3.c is compiled using
5938 IEEE 128-bit long double.
5939 (neg<mode>2_internal): Use the correct mode to check whether the
5940 mode is IBM extended.
5941 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5942 multiply and divide external functions from being created more
5945 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5947 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5949 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5950 the edge can be forwarded.
5951 (cfg_layout_merge_blocks): Likewise.
5953 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5955 * except.c (finish_eh_generation): Commit edge insertions only after
5956 the EH edges have been redirected from post-landing to landing pads.
5958 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5960 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5961 create_tmp_var_for to create the FRAME decl.
5962 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5964 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5966 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5967 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5968 debug statement when resetting its value.
5969 (expand_call_inline): Copy the locus of the call onto the assignment
5970 of the return value, if any. Use local variable in more cases.
5972 2018-06-21 Martin Liska <mliska@suse.cz>
5974 * ipa-pure-const.c (propagate_nothrow): Use
5975 funct_state_summaries->get.
5976 (dump_malloc_lattice): Likewise.
5977 (propagate_malloc): Likewise.
5979 2018-06-21 Richard Biener <rguenther@suse.de>
5981 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5982 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5983 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5985 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5986 BLOCK_ABSTRACT_ORIGIN unconditionally.
5988 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5990 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5992 * ipa-reference.c (ipa_reference_c_finalize): Delete
5993 ipa_ref_opt_sum_summaries and set it to NULL.
5995 2018-06-21 Tom de Vries <tdevries@suse.de>
5997 PR tree-optimization/85859
5998 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5999 test with comment from bb_no_side_effects_p.
6001 2018-06-21 Richard Biener <rguenther@suse.de>
6003 PR tree-optimization/86232
6004 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
6005 max for constant niter.
6007 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6009 * config/aarch64/aarch64-simd.md
6010 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
6012 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6014 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
6015 Make opernads of the unspec commutative.
6017 2018-06-21 Richard Biener <rguenther@suse.de>
6019 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
6020 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
6021 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6022 (vect_analyze_data_ref_dependence): Re-order checks to deal with
6024 (vect_record_base_alignments): Do not record base alignment
6025 for gathers or scatters.
6026 (vect_compute_data_ref_alignment): Drop return value that is always
6027 true. Bail out early for gathers or scatters.
6028 (vect_enhance_data_refs_alignment): Bail out early for gathers
6030 (vect_find_same_alignment_drs): Likewise.
6031 (vect_analyze_data_refs_alignment): Remove dead code.
6032 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6033 (vect_analyze_data_refs): For possible gathers or scatters do
6034 not create an alternate DR, just check their possible validity
6035 and mark them. Adjust DECL_NONALIASED handling to not rely
6037 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
6038 update inits of gathers or scatters.
6039 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
6040 Also copy gather/scatter flag to pattern vinfo.
6042 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
6044 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6045 behavior of vec_packsu (vector unsigned long long, vector unsigned
6046 long long) to match behavior of vec_packs with same signature.
6048 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
6049 Thomas Schwinge <thomas@codesourcery.com>
6050 Cesar Philippidis <cesar@codesourcery.com>
6052 * gimplify.c (gimplify_scan_omp_clauses): Add support for
6053 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6054 (gimplify_adjust_omp_clauses): Likewise.
6055 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
6056 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
6057 (gimplify_omp_target_update): Update handling of acc update and
6059 * omp-low.c (install_var_field): Remove unused parameter
6060 base_pointers_restrict.
6061 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
6062 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
6064 (omp_target_base_pointers_restrict_p): Delete.
6065 (scan_omp_target): Update call to scan_sharing_clauses.
6066 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
6068 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
6069 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6070 (convert_local_omp_clauses): Likewise.
6071 * tree-pretty-print.c (dump_omp_clause): Likewise.
6072 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
6074 (omp_clause_code_name): Likewise.
6076 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6079 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
6082 PR tree-optimization/86231
6083 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
6084 anti-range don't overwrite *vr0min before using it to compute *vr0max.
6086 2018-06-20 Tom de Vries <tdevries@suse.de>
6088 PR tree-optimization/86097
6089 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
6090 iv type if signedness of iv type is not the same as that of *nit.
6092 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6094 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
6095 EDGE_EH edges, verify they are all EDGE_EH.
6097 2018-06-20 Maya Rashish <coypu@sdf.org>
6099 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
6101 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6103 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
6104 * config/aarch64/aarch64.c (xgene1_tunings): Add
6105 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
6106 (aarch64_mode_valid_for_sched_fusion_p):
6107 Allow 16-byte modes.
6108 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
6109 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
6111 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
6113 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6114 * config/aarch64/iterators.md (VQ2): New mode iterator.
6116 2018-06-20 Martin Liska <mliska@suse.cz>
6118 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6119 Change default ratio from 10 to 8.
6121 2018-06-20 Martin Liska <mliska@suse.cz>
6123 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6125 (bit_test_cluster::find_bit_tests): Likewise.
6126 (switch_decision_tree::analyze_switch_statement): Find clusters.
6127 * tree-switch-conversion.h (struct jump_table_cluster): Document
6130 2018-06-20 Martin Liska <mliska@suse.cz>
6132 * tree-switch-conversion.c (switch_conversion::collect):
6133 Record m_uniq property.
6134 (switch_conversion::expand): Bail out for special conditions.
6135 (group_cluster::~group_cluster): New.
6136 (group_cluster::group_cluster): Likewise.
6137 (group_cluster::dump): Likewise.
6138 (jump_table_cluster::emit): New.
6139 (switch_decision_tree::fix_phi_operands_for_edges): New.
6140 (struct case_node): Remove struct.
6141 (jump_table_cluster::can_be_handled): New.
6142 (case_values_threshold): Moved to header.
6143 (reset_out_edges_aux): Likewise.
6144 (jump_table_cluster::is_beneficial): New.
6145 (bit_test_cluster::can_be_handled): Likewise.
6146 (add_case_node): Remove.
6147 (bit_test_cluster::is_beneficial): New.
6148 (case_bit_test::cmp): New.
6149 (bit_test_cluster::emit): New.
6150 (expand_switch_as_decision_tree_p): Remove.
6151 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6152 (fix_phi_operands_for_edge): Likewise.
6153 (switch_decision_tree::analyze_switch_statement): New.
6154 (compute_cases_per_edge): Move ...
6155 (switch_decision_tree::compute_cases_per_edge): ... here.
6156 (try_switch_expansion): Likewise.
6157 (switch_decision_tree::try_switch_expansion): Likewise.
6158 (record_phi_operand_mapping): Likewise.
6159 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6160 (emit_case_decision_tree): Likewise.
6161 (switch_decision_tree::emit): Likewise.
6162 (balance_case_nodes): Likewise.
6163 (switch_decision_tree::balance_case_nodes): Likewise.
6164 (dump_case_nodes): Likewise.
6165 (switch_decision_tree::dump_case_nodes): Likewise.
6166 (emit_jump): Likewise.
6167 (switch_decision_tree::emit_jump): Likewise.
6168 (emit_cmp_and_jump_insns): Likewise.
6169 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6170 (emit_case_nodes): Likewise.
6171 (switch_decision_tree::emit_case_nodes): Likewise.
6172 (conditional_probability): Remove.
6173 * tree-switch-conversion.h (enum cluster_type): New.
6175 (struct cluster): Likewise.
6176 (cluster::cluster): Likewise.
6177 (struct simple_cluster): Likewise.
6178 (simple_cluster::simple_cluster): Likewise.
6179 (struct group_cluster): Likewise.
6180 (struct jump_table_cluster): Likewise.
6181 (struct bit_test_cluster): Likewise.
6182 (struct min_cluster_item): Likewise.
6183 (struct case_tree_node): Likewise.
6184 (case_tree_node::case_tree_node): Likewise.
6185 (jump_table_cluster::case_values_threshold): Likewise.
6186 (struct case_bit_test): Likewise.
6187 (struct switch_decision_tree): Likewise.
6188 (struct switch_conversion): Likewise.
6189 (switch_decision_tree::reset_out_edges_aux): Likewise.
6191 2018-06-20 Martin Liska <mliska@suse.cz>
6193 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6194 (hoist_edge_and_branch_if_true): Likewise.
6195 (expand_switch_using_bit_tests_p): Likewise.
6196 (struct case_bit_test): Likewise.
6197 (case_bit_test_cmp): Likewise.
6198 (emit_case_bit_tests): Likewise.
6199 (switch_conversion::switch_conversion): New class.
6200 (struct switch_conv_info): Remove old struct.
6201 (collect_switch_conv_info): More to ...
6202 (switch_conversion::collect): ... this.
6203 (check_range): Likewise.
6204 (switch_conversion::check_range): Likewise.
6205 (check_all_empty_except_final): Likewise.
6206 (switch_conversion::check_all_empty_except_final): Likewise.
6207 (check_final_bb): Likewise.
6208 (switch_conversion::check_final_bb): Likewise.
6209 (create_temp_arrays): Likewise.
6210 (switch_conversion::create_temp_arrays): Likewise.
6211 (free_temp_arrays): Likewise.
6212 (gather_default_values): Likewise.
6213 (switch_conversion::gather_default_values): Likewise.
6214 (build_constructors): Likewise.
6215 (switch_conversion::build_constructors): Likewise.
6216 (constructor_contains_same_values_p): Likewise.
6217 (switch_conversion::contains_same_values_p): Likewise.
6218 (array_value_type): Likewise.
6219 (switch_conversion::array_value_type): Likewise.
6220 (build_one_array): Likewise.
6221 (switch_conversion::build_one_array): Likewise.
6222 (build_arrays): Likewise.
6223 (switch_conversion::build_arrays): Likewise.
6224 (gen_def_assigns): Likewise.
6225 (switch_conversion::gen_def_assigns): Likewise.
6226 (prune_bbs): Likewise.
6227 (switch_conversion::prune_bbs): Likewise.
6228 (fix_phi_nodes): Likewise.
6229 (switch_conversion::fix_phi_nodes): Likewise.
6230 (gen_inbound_check): Likewise.
6231 (switch_conversion::gen_inbound_check): Likewise.
6232 (process_switch): Use the newly created class.
6233 (switch_conversion::expand): New.
6234 (switch_conversion::~switch_conversion): New.
6235 * tree-switch-conversion.h: New file.
6237 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6239 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6240 tree-vect-patterns.c.
6241 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6242 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6243 (vect_recog_sad_pattern): Likewise.
6244 (vect_recog_widen_sum_pattern): Likewise.
6245 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6246 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6247 (vect_recog_rotate_pattern): Likewise.
6248 (vect_recog_mult_pattern): Likewise.
6249 (vect_recog_vector_vector_shift_pattern): Likewise.
6250 (vect_recog_divmod_pattern): Likewise.
6251 (vect_recog_mixed_size_cond_pattern): Likewise.
6252 (vect_recog_bool_pattern): Likewise.
6253 (vect_recog_mask_conversion_pattern): Likewise.
6254 (vect_try_gather_scatter_pattern): Likewise.
6255 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6256 (vect_recog_over_widening_pattern): Likewise.
6257 (vect_recog_gather_scatter_pattern): Likewise.
6258 (vect_recog_func_ptr): Move from tree-vectorizer.h
6259 (vect_vect_recog_func_ptrs): Move further down the file.
6260 (vect_recog_func): Likewise. Remove the third argument.
6261 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6262 (vect_pattern_recog_1): Expect the pattern function to do any
6263 necessary target tests. Also expect it to provide a vector type.
6264 Remove the type_in handling.
6266 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6268 * tree-vect-patterns.c (vect_pattern_detected): New function.
6269 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6270 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6271 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6272 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6273 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6274 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6275 (vect_recog_mask_conversion_pattern)
6276 (vect_try_gather_scatter_pattern): Likewise.
6278 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6280 * tree-vect-patterns.c (vect_get_internal_def): New function.
6281 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6282 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6283 (search_type_for_mask_1): Use it.
6285 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6287 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6288 redundant WIDEN_SUM_EXPR handling.
6289 (vect_recog_sad_pattern): Likewise.
6291 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6293 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6294 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6295 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6296 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6297 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6299 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6301 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6302 use the stmt_vec_info of the original bb statement for the
6303 new zero assignment, even if the call is part of a pattern.
6305 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6307 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6308 that the sequence is attached to the original statement rather
6309 than the pattern statement.
6310 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6311 PATTERN_DEF_SEQ from the original statement rather than
6312 the main pattern statement.
6313 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6314 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6315 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6317 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6319 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6320 definition statements before the early exit for statements that aren't
6322 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6324 (vect_transform_loop): ...here. Process pattern definition
6325 statements without first checking whether the main pattern
6326 statement is live or relevant.
6328 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6330 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6331 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6333 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6335 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6336 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6337 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6338 (expand_block_compare): Change select_block_compare_mode call.
6339 (expand_strncmp_align_check): Use new functions, fix comment.
6340 (emit_final_str_compare_gpr): New function.
6341 (expand_strn_compare): Refactor and clean up code.
6342 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6344 2018-06-19 Tony Reix <tony.reix@atos.com>
6345 Damien Bergamini <damien.bergamini@atos.com>
6346 David Edelsohn <dje.gcc@gmail.com>
6348 * collect2.c (static_obj): New variable.
6349 (static_libs): New variable.
6350 (is_in_list): Uncomment declaration.
6351 (main): Track AIX libraries linked statically.
6352 (is_in_list): Uncomment definition.
6353 (scan_prog_file): Don't add AIX shared libraries initializer
6354 to constructor list if linking statically.
6356 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6358 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6360 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6362 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6364 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6367 2018-06-19 Martin Liska <mliska@suse.cz>
6369 * config/i386/i386.c (ix86_can_inline_p): Do not use
6370 ipa_fn_summaries::get_create.
6371 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6373 (devirtualization_time_bonus): Likewise.
6374 (ipcp_propagate_stage): Likewise.
6375 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6376 (edge_set_predicate): Likewise.
6377 (evaluate_conditions_for_known_args): Likewise.
6378 (evaluate_properties_for_edge): Likewise.
6379 (ipa_call_summary::reset): Tranform to ...
6380 (ipa_call_summary::~ipa_call_summary): ... this.
6381 (ipa_fn_summary::reset): Transform to ...
6382 (ipa_fn_summary::~ipa_fn_summary): ... this.
6383 (ipa_fn_summary_t::remove): Rename to ...
6384 (ipa_fn_summary_t::remove_callees): ... this.
6385 (ipa_fn_summary_t::duplicate): Use placement new
6386 instead of memory copy.
6387 (ipa_call_summary_t::duplicate): Likewise.
6388 (ipa_call_summary_t::remove): Remove.
6389 (dump_ipa_call_summary): Change get_create to get.
6390 (ipa_dump_fn_summary): Dump only when summary exists.
6391 (analyze_function_body): Use symbol_summary::get instead
6393 (compute_fn_summary): Likewise.
6394 (estimate_edge_devirt_benefit): Likewise.
6395 (estimate_edge_size_and_time): Likewise.
6396 (inline_update_callee_summaries): Likewise.
6397 (remap_edge_change_prob): Likewise.
6398 (remap_edge_summaries): Likewise.
6399 (ipa_merge_fn_summary_after_inlining): Likewise.
6400 (write_ipa_call_summary): Likewise.
6401 (ipa_fn_summary_write): Likewise.
6402 (ipa_free_fn_summary): Likewise.
6403 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6404 (struct ipa_call_summary): Likewise.
6405 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6407 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6408 (estimate_size_after_inlining): Likewise.
6409 (estimate_growth): Likewise.
6410 (growth_likely_positive): Likewise.
6411 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6412 (inline_call): Likewise.
6413 * ipa-inline.c (caller_growth_limits): Likewise.
6414 (can_inline_edge_p): Likewise.
6415 (can_inline_edge_by_limits_p): Likewise.
6416 (compute_uninlined_call_time): Likewise.
6417 (compute_inlined_call_time): Likewise.
6418 (want_inline_small_function_p): Likewise.
6419 (edge_badness): Likewise.
6420 (update_caller_keys): Likewise.
6421 (update_callee_keys): Likewise.
6422 (inline_small_functions): Likewise.
6423 (inline_to_all_callers_1): Likewise.
6424 (dump_overall_stats): Likewise.
6425 (early_inline_small_functions): Likewise.
6426 (early_inliner): Likewise.
6427 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6428 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6429 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6430 * ipa-split.c (execute_split_functions): Likewise.
6431 * symbol-summary.h: Likewise.
6432 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6434 2018-06-19 Richard Biener <rguenther@suse.de>
6436 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6437 (vectorize_loops): ... here. Fix dbgcnt handling.
6438 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6440 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6443 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6444 ieee128 argument takes up only one (vector) register, not two (floating
6447 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6449 * gimplify.c (gimplify_init_constructor): Really never clear for an
6450 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6452 2018-06-19 Richard Biener <rguenther@suse.de>
6454 PR tree-optimization/86179
6455 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6456 after failed recognition.
6458 2018-06-18 Martin Sebor <msebor@redhat.com>
6461 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6462 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6463 Handle integer subtraction.
6464 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6465 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6467 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6469 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6470 param from rtx to rtx_insn *.
6471 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6473 (frv_ifcvt_modify_insn): Likwise.
6474 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6475 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6476 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6477 as_a <rtx_insn *> cast to local "unprotected_region" once
6478 it's been established that it's not NULL or pc_rtx.
6479 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6480 param "sethi" from rtx to rtx_insn *.
6481 (nds32_group_float_insns): Likewise for param "insn".
6482 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6484 (vax_output_int_subtract): Likewise.
6485 * config/vax/vax.c (vax_output_int_add): Likewise for param
6487 (vax_output_int_subtract): Likewise.
6488 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6489 (emit_pattern_after): Likewise for param "after".
6490 (emit_insn_after): Likewise.
6491 (emit_jump_insn_after): Likewise.
6492 (emit_call_insn_after): Likewise.
6493 (emit_debug_insn_after): Likewise.
6494 (emit_pattern_before): Likewise for param "before".
6495 (emit_insn_before): Likewise.
6496 (emit_jump_insn_before): Likewise.
6497 * final.c (get_insn_template): Likewise for param "insn", removing
6499 * output.h (get_insn_template): Likewise for 2nd param.
6500 * rtl.h (emit_insn_before): Likewise.
6501 (emit_jump_insn_before): Likewise.
6502 (emit_debug_insn_before_noloc): Likewise.
6503 (emit_insn_after): Likewise.
6504 (emit_jump_insn_after): Likewise.
6505 (emit_call_insn_after): Likewise.
6506 (emit_debug_insn_after): Likewise.
6507 (set_insn_deleted): Likewise for param.
6509 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6512 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6513 floating point modes, so that IFmode is numerically greater than
6514 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6515 to declare the ordering. This prevents IFmode from being
6516 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6517 machine. Include rs6000-modes.h to share the fractional values
6518 between genmodes* and the rest of the compiler.
6522 * config/rs6000/rs6000-modes.h: New file.
6523 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6524 meaning of rs6000_long_double_size so that 126..128 selects an
6525 appropriate 128-bit floating point type.
6526 (rs6000_option_override_internal): Likewise.
6527 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6528 (TARGET_LONG_DOUBLE_128): Change the meaning of
6529 rs6000_long_double_size so that 126..128 selects an appropriate
6530 128-bit floating point type.
6531 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6532 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6533 source and destination to match the standard usage.
6534 (truncifkf2): Likewise.
6535 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6536 ISA 2.07 to use an explicit clobber, instead of passing in a
6538 (copysign<mode>3_soft): Likewise.
6540 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6542 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6543 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6544 (vect_slp_analyze_instance_dependence): Likewise.
6545 (vect_enhance_data_refs_alignment): Likewise.
6546 (vect_analyze_data_refs_alignment): Likewise.
6547 (vect_slp_analyze_and_verify_instance_alignment
6548 (vect_analyze_data_ref_accesses): Likewise.
6549 (vect_prune_runtime_alias_test_list): Likewise.
6550 (vect_analyze_data_refs): Likewise.
6551 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6552 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6553 (vect_analyze_scalar_cycles_1): Likewise.
6554 (vect_get_loop_niters): Likewise.
6555 (vect_analyze_loop_form_1): Likewise.
6556 (vect_update_vf_for_slp): Likewise.
6557 (vect_analyze_loop_operations): Likewise.
6558 (vect_analyze_loop): Likewise.
6559 (vectorizable_induction): Likewise.
6560 (vect_transform_loop): Likewise.
6561 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6562 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6563 (vect_make_slp_decision): Likewise.
6564 (vect_detect_hybrid_slp): Likewise.
6565 (vect_slp_analyze_operations): Likewise.
6566 (vect_slp_bb): Likewise.
6567 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6568 (vectorizable_bswap): Likewise.
6569 (vectorizable_call): Likewise.
6570 (vectorizable_simd_clone_call): Likewise.
6571 (vectorizable_conversion): Likewise.
6572 (vectorizable_assignment): Likewise.
6573 (vectorizable_shift): Likewise.
6574 (vectorizable_operation): Likewise.
6575 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6577 2018-06-18 Martin Sebor <msebor@redhat.com>
6579 PR tree-optimization/81384
6580 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6581 * builtins.c (expand_builtin_strnlen): New function.
6582 (expand_builtin): Call it.
6583 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6584 * builtins.def (BUILT_IN_STRNLEN): New.
6585 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6586 Warn for bounds in excess of maximum object size.
6587 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6588 single-value ranges. Handle strnlen.
6589 (handle_builtin_strlen): Handle strnlen.
6590 (strlen_check_and_optimize_stmt): Same.
6591 * doc/extend.texi (Other Builtins): Document strnlen.
6593 2018-06-18 Maya Rashish <coypu@sdf.org>
6595 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6596 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6597 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6599 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6601 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6603 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6605 * tree.c (escaped_string::escape): Replace cast to char * by
6606 const_cast<char *> (unescaped).
6608 2018-06-18 Nick Clifton <nickc@redhat.com>
6611 * tree.c (escaped_string): New class. Converts an unescaped
6612 string into its escaped equivalent.
6613 (warn_deprecated_use): Use the new class to convert the
6614 deprecation message, if present.
6615 (test_escaped_strings): New self test.
6616 (test_c_tests): Add test_escaped_strings.
6617 * doc/extend.texi (deprecated): Add a note that the
6618 deprecation message is affected by the -fmessage-length
6619 option, and that control characters will be escaped.
6620 (#pragma GCC error): Document this pragma.
6621 (#pragma GCC warning): Likewise.
6622 * doc/invoke.texi (-fmessage-length): Document this option's
6623 effect on the #warning and #error preprocessor directives and
6624 the deprecated attribute.
6626 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6628 * tree.c (decl_value_expr_lookup): Revert latest change.
6629 (decl_value_expr_insert): Likewise.
6631 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6633 * gimplify.c (nonlocal_vlas): Delete.
6634 (nonlocal_vla_vars): Likewise.
6635 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6637 (gimplify_body): Do not create and destroy nonlocal_vlas.
6638 * tree-nested.c: Include diagnostic.h.
6639 (use_pointer_in_frame): Tweak.
6640 (lookup_field_for_decl): Add assertion and declare the transformation.
6641 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6642 internal error when the reference is in a wrong context. Do not
6643 create a debug decl by default.
6644 (note_nonlocal_block_vlas): Delete.
6645 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6646 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6647 create a debug decl by default.
6648 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6649 call is in a wrong context.
6650 (fixup_vla_decls): New function.
6651 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6652 debug variables were created.
6653 * tree.c (decl_value_expr_lookup): Add checking assertion.
6654 (decl_value_expr_insert): Likewise.
6656 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6659 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6660 * tree-scalar-evolution.c (interpret_expr): Likewise.
6661 (expression_expensive_p): Likewise.
6662 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6663 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6664 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6665 (ssa_defined_by_minus_one_stmt_p): New.
6667 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6670 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6671 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6672 * dojump.c (do_jump): Likewise.
6673 * expr.c (expand_expr_real_2): Check operand type's sign.
6674 * fold-const.c (const_unop): Handle ABSU_EXPR.
6675 (fold_abs_const): Likewise.
6676 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6677 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6678 (strip_sign_op_1): Likesise.
6679 * match.pd: Add new pattern to generate ABSU_EXPR.
6680 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6681 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6682 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6683 * tree-inline.c (estimate_operator_cost): Likewise.
6684 * tree-pretty-print.c (dump_generic_node): Likewise.
6685 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6686 * tree.def (ABSU_EXPR): New.
6688 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6691 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6692 documented as preserved for backward compatibility only.
6693 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6695 PR rtl-optimization/86108
6696 * bb-reorder.c (create_forwarder_block): Renamed to ...
6697 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6698 jump from new landing pad to the second part.
6699 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6702 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6705 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6706 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6707 Only call store_expr for halves if the mode is the same.
6710 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6711 Fix up comment formatting.
6713 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6715 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6716 (typed_splay_tree::closure,
6717 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6718 (typed_splay_tree::typed_splay_tree,
6719 typed_splay_tree::operator =): Declared private.
6720 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6721 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6722 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6723 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6724 typed_splay_tree::splay_tree_splay,
6725 typed_splay_tree::splay_tree_foreach_helper,
6726 typed_splay_tree::splay_tree_insert,
6727 typed_splay_tree::splay_tree_remove,
6728 typed_splay_tree::splay_tree_lookup,
6729 typed_splay_tree::splay_tree_predecessor,
6730 typed_splay_tree::splay_tree_successor,
6731 typed_splay_tree::splay_tree_min,
6732 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6733 (typed_splay_tree::root, typed_splay_tree::comp,
6734 typed_splay_tree::delete_key,
6735 typed_splay_tree::delete_value): New data members.
6736 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6737 typed_splay_tree::remove.
6739 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6741 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6742 -mginv and -mno-ginv to the assembler.
6743 * config/mips/mips.opt (-mcrc): New option.
6745 * doc/invoke.text (-mcrc): Document.
6748 2018-06-15 Nick Clifton <nickc@redhat.com>
6751 * tree.c (escaped_string): New class. Converts an unescaped
6752 string into its escaped equivalent.
6753 (warn_deprecated_use): Use the new class to convert the
6754 deprecation message, if present.
6755 (test_escaped_strings): New self test.
6756 (test_c_tests): Add test_escaped_strings.
6757 * doc/extend.texi (deprecated): Add a note that the
6758 deprecation message is affected by the -fmessage-length
6759 option, and that control characters will be escaped.
6760 (#pragma GCC error): Document this pragma.
6761 (#pragma GCC warning): Likewise.
6762 * doc/invoke.texi (-fmessage-length): Document this option's
6763 effect on the #warning and #error preprocessor directives and
6764 the deprecated attribute.
6766 2018-06-15 Richard Biener <rguenther@suse.de>
6768 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6769 here, also noting vector size used.
6770 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6771 size used in MSG_OPTIMIZED_LOCATIONS dump.
6772 (pass_slp_vectorize::execute): Adjust.
6774 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6777 * config/arc/arc.c (arc_return_address_register): Fix
6780 2018-06-15 Richard Biener <rguenther@suse.de>
6783 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6784 leave useless conversion stripping to force_gimple_operand_gsi.
6785 (gimplify_build2): Likewise.
6786 (gimplify_build1): Likewise.
6788 2018-06-15 Richard Biener <rguenther@suse.de>
6791 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6792 before adjusting their block.
6794 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6796 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6797 multilibs for *-*-rtems*.
6798 * config/riscv/t-rtems: New file.
6800 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6803 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6804 unsigned_type_for returns NULL.
6807 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6808 subregs of multi-word pseudos unless the float mode has word size.
6810 2018-06-14 Richard Biener <rguenther@suse.de>
6813 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6814 duplicate and harmful type_hash_canon.
6815 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6817 2018-06-14 Richard Biener <rguenther@suse.de>
6820 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6823 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6825 * config/rtems.h (STDINT_LONG32): Define.
6827 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6828 Prachi Godbole <prachi.godbole@imgtec.com>
6830 * config/mips/mips-cpus.def: Define P6600.
6831 * config/mips/mips-tables.opt: Regenerate.
6832 * config/mips/mips.c (mips_ucbranch_type): New enum.
6833 (mips_rtx_cost_data): Add support for P6600.
6834 (mips_issue_rate): Likewise.
6835 (mips_multipass_dfa_lookahead): Likewise.
6836 (mips_avoid_hazard): Likewise.
6837 (mips_reorg_process_insns): Likewise.
6838 (mips_classify_branch_p6600): New function.
6839 * config/mips/mips.h (TUNE_P6600): New define.
6840 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6841 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6842 * config/mips/mips.md: Include p6600.md.
6843 (processor): Add p6600.
6844 * config/mips/p6600.md: New file.
6845 * doc/invoke.texi: Add p6600 to supported architectures.
6847 2018-06-13 Martin Sebor <msebor@redhat.com>
6849 PR tree-optimization/86114
6850 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6852 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6854 2018-06-13 Richard Biener <rguenther@suse.de>
6856 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6857 Properly set vector type of the intermediate stmt.
6858 * tree-vect-stmts.c (vectorizable_operation): The destination
6859 var always has vectype_out type.
6861 2018-06-13 Jeff Law <law@redhat.com>
6863 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6864 integer 0 for argument to print_rtl_with_bb.
6865 (rl78_reorg): Likewise.
6867 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6869 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6870 from rtx to rtx_insn *.
6871 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6873 (add_sched_insns_for_speculation): Likewise for local "target",
6874 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6875 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6876 from rtx_insn ** to rtx_code_label **.
6877 (reorg_emit_nops): Likewise.
6878 (c6x_reorg): Likewise for local "call_labels".
6879 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6881 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6882 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6883 the loops over LABEL_REFS.
6884 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6886 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6887 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6888 (split_branches): Strengthen local "olabel" from rtx to
6889 rtx_insn *, adding a safe_as_a cast.
6890 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6892 (add_insn_after): Likewise for first two params.
6893 (add_insn_before): Likewise.
6894 (remove_insn): Likewise for param.
6895 (emit_pattern_before_noloc): Likewise for second and third params.
6896 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6897 (emit_call_insn_before_noloc): Likewise.
6898 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6900 (emit_barrier_before): Likewise.
6901 (emit_label_before): Strengthen "label" param from "rtx" to
6902 "rtx_code_label *". Strengthen "before" param from "rtx" to
6904 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6906 (emit_pattern_after_noloc): Likewise.
6907 (emit_insn_after_noloc): Likewise.
6908 (emit_jump_insn_after_noloc): Likewise.
6909 (emit_call_insn_after_noloc): Likewise.
6910 (emit_debug_insn_after_noloc): Likewise.
6911 (emit_barrier_after): Likewise.
6912 (emit_label_after): Likewise for both params.
6913 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6914 "loc" param from "int" to "location_t".
6915 (emit_insn_after_setloc): Likewise.
6916 (emit_jump_insn_after_setloc): Likewise.
6917 (emit_call_insn_after_setloc): Likewise.
6918 (emit_debug_insn_after_setloc): Likewise.
6919 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6920 "loc" param from "int" to "location_t".
6921 (emit_pattern_before): Convert NULL_RTX to NULL.
6922 (emit_insn_before_setloc): Convert "loc" param from "int" to
6924 (emit_jump_insn_before_setloc): Likewise.
6925 (emit_call_insn_before_setloc): Likewise.
6926 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6927 rtx_insn *. Convert "loc" param from "int" to "location_t".
6928 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6929 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6930 Convert 3rd param from "int" to "location_t".
6931 (emit_barrier_before, emit_barrier_after, next_real_insn):
6932 Strengthen param from rtx to rtx_insn *.
6933 (emit_label_before): Strengthen 1st param from "rtx" to
6934 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6936 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6937 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6938 Strengthen 2nd param from "rtx" to "rtx_insn *".
6939 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6940 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6941 Likewise. Convert 3rd param from "int" to "location_t".
6942 (emit_label_after): Strengthen 1st param from "rtx" to
6944 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6946 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6947 from "rtx" to "rtx_insn *".
6949 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6951 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6952 bodies streamed in with -Q.
6953 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6954 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6955 * gimple-streamer-out.c: Include gimple-pretty-print.h
6956 (output_bb): Dump stmts streamed.
6957 * lto-section-out.c: Include print-tree.h
6958 (lto_begin_section): Dump sections created.
6959 (lto_output_decl_index): Dump decl encoded.
6960 * lto-streamer-out.c: Include print-tree.h
6961 (create_output_block): Dump output block created.
6962 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6963 (output_function): Dump function output.
6964 (output_constructor): Dump constructor streamed.
6965 (write_global_stream): Output indexes encoded.
6966 (produce_asm_for_decls): Dump streams encoded.
6967 * lto-streamer.c (streamer_dump_file): New global var.
6968 * lto-streamer.h (streamer_dump_file): Declare.
6969 * passes.c (ipa_write_summaries): Initialize streamer dump.
6970 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6973 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6976 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6977 offsets for register save directives. Emit a second batch of save
6978 directives, if need be, when the function accesses prior frames.
6980 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6982 * config/arc/fpu.md (fmasf4): Force operand to register.
6983 (fnmasf4): Likewise.
6985 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6987 * config/arc/arc-protos.h (arc_pad_return): Remove.
6988 * config/arc/arc.c (machine_function): Remove force_short_suffix
6990 (arc_print_operand): Adjust printing of '&'.
6991 (arc_verify_short): Remove conditional printing of short suffix.
6992 (arc_final_prescan_insn): Remove reference to size_reason.
6993 (pad_return): New function.
6994 (arc_reorg): Call pad_return.
6995 (arc_pad_return): Remove.
6996 (arc_init_machine_status): Remove reference to force_short_suffix.
6997 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6998 (attr length): When attribute iscompact is true force to 2
6999 regardless; in the case of maybe check if we want to force the
7000 instruction to have 4 bytes length.
7001 (nopv): Change it to generate 4 byte long nop as well.
7002 (blockage): New pattern.
7003 (simple_return): Remove call to arc_pad_return.
7004 (p_return_i): Likewise.
7006 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7008 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
7010 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7012 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
7015 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7017 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
7018 for ARC700 and ARCv2.
7020 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
7023 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
7024 operands[2] instead of operands[1].
7027 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7029 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
7030 case, check whether the outer register overlaps an unallocatable
7031 register, not just whether it fits the required class.
7033 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7035 * poly-int.h (can_div_trunc_p): Add new overload in which all values
7037 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
7038 (memrefs_conflict_p): Likewise.
7039 (init_alias_analysis): Likewise.
7040 * cfgexpand.c (expand_debug_expr): Likewise.
7041 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
7042 * cse.c (fold_rtx): Likewise.
7043 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
7044 * expr.c (emit_block_move_hints): Likewise.
7045 (clear_storage_hints, push_block, emit_push_insn): Likewise.
7046 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
7047 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
7048 (emit_group_store): Likewise.
7049 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
7050 to read the PRE/POST_MODIFY increment.
7051 * calls.c (store_one_arg): Use strip_offset.
7052 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
7054 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
7056 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
7057 (simplify_binary_operation_1): Extend CONST_INT handling to
7059 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
7060 than a HOST_WIDE_INT.
7061 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
7063 (adjust_mems, add_stores): Update accodingly.
7064 (vt_canonicalize_addr): Track polynomial offsets.
7065 (emit_note_insn_var_location): Likewise.
7066 (vt_add_function_parameter): Likewise.
7067 (vt_initialize): Likewise.
7069 2018-06-12 Jeff Law <law@redhat.com>
7071 * config.gcc (alpha*-*-freebsd*): Remove.
7072 * config/alpha/freebsd.h: Remove.
7074 2018-06-12 David Malcolm <dmalcolm@redhat.com>
7077 * spellcheck-tree.c (levenshtein_distance): Rename to...
7078 (get_edit_distance): ...this, and update for underlying renaming.
7079 * spellcheck-tree.h (levenshtein_distance): Rename to...
7080 (get_edit_distance): ...this.
7081 * spellcheck.c (levenshtein_distance): Rename to...
7082 (get_edit_distance): ...this. Convert from Levenshtein distance
7083 to Damerau-Levenshtein distance by supporting transpositions of
7084 adjacent characters. Rename "v1" to "v_next" and "v0" to
7086 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
7087 (selftest::test_edit_distance_unit_test_oneway): ...this, and
7088 update for underlying renaming.
7089 (selftest::levenshtein_distance_unit_test): Rename to...
7090 (selftest::test_get_edit_distance_unit): ...this, and update for
7091 underlying renaming.
7092 (selftest::test_find_closest_string): Add example from PR 69968
7093 where transposition helps
7094 (selftest::test_metric_conditions): Update for renaming.
7095 (selftest::test_metric_conditions): Likewise.
7096 (selftest::spellcheck_c_tests): Likewise.
7097 * spellcheck.h (levenshtein_distance): Rename both overloads to...
7098 (get_edit_distance): ...this.
7099 (best_match::consider): Update for renaming.
7101 2018-06-12 Martin Sebor <msebor@redhat.com>
7103 PR tree-optimization/85259
7104 * builtins.c (compute_objsize): Handle constant offsets.
7105 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
7106 true iff a warning has been issued.
7107 * gimple.h (gimple_nonartificial_location): New function.
7108 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
7109 gimple_nonartificial_location and handle -Wno-system-headers.
7110 (handle_builtin_stxncpy): Same.
7112 2018-06-12 Martin Sebor <msebor@redhat.com>
7115 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7117 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7119 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7120 BUILTIN_VEC_XST entries for pointer to double and long long.
7122 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7125 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7127 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7130 2018-06-12 Martin Liska <mliska@suse.cz>
7132 * doc/options.texi: Document IntegerRange.
7134 2018-06-12 Martin Liska <mliska@suse.cz>
7136 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7137 * opt-functions.awk: Handle Deprecated flag.
7138 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7140 (read_cmdline_option): Report warning for a deprecated option.
7141 * opts.h (struct cl_option): Add new field cl_deprecated.
7142 (CL_ERR_DEPRECATED): New.
7144 2018-06-12 Martin Liska <mliska@suse.cz>
7146 * doc/options.texi: Document Deprecated option flag.
7148 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7150 * config/arc/arc-arch.h (arc_extras): New enum.
7151 (arc_cpu_t):Add field extra.
7152 (arc_cpu_types): Consider the extras.
7153 * config/arc/arc-cpus.def: Add extras info.
7154 * config/arc/arc-opts.h (processor_type): Consider extra field.
7155 * config/arc/arc.c (arc_override_options): Handle extra field.
7157 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7159 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7160 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7162 * config/arc/arc.c (arc_init): Update pic warning.
7163 (irq_range): Update irq range parsing warnings.
7164 (arc_override_options): Update various warning messages.
7165 (arc_handle_aux_attribute): Likewise.
7167 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7169 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7171 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7173 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7174 compiler messages directives.
7176 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7178 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7179 * config/mips/mips-tables.opt: Regenerate.
7180 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7182 * doc/invoke.texi: Document -march=i6500.
7184 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7186 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7187 (i6400_gpmul): Add cpu_unit.
7188 (i6400_gpdiv): Likewise.
7189 (i6400_msa_add_d): Update reservations.
7190 (i6400_msa_int_add) Likewise.
7191 (i6400_msa_short_logic3) Likewise.
7192 (i6400_msa_short_logic2) Likewise.
7193 (i6400_msa_short_logic) Likewise.
7194 (i6400_msa_move) Likewise.
7195 (i6400_msa_cmp) Likewise.
7196 (i6400_msa_short_float2) Likewise.
7197 (i6400_msa_div_d) Likewise.
7198 (i6400_msa_long_logic1) Likewise.
7199 (i6400_msa_long_logic2) Likewise.
7200 (i6400_msa_mult) Likewise.
7201 (i6400_msa_long_float2) Likewise.
7202 (i6400_msa_long_float4) Likewise.
7203 (i6400_msa_long_float5) Likewise.
7204 (i6400_msa_long_float8) Likewise.
7205 (i6400_fpu_fadd): Include frint type.
7206 (i6400_fpu_store): New define_insn_reservation.
7207 (i6400_fpu_load): Likewise.
7208 (i6400_fpu_move): Likewise.
7209 (i6400_fpu_fcmp): Likewise.
7210 (i6400_fpu_fmadd): Likewise.
7211 (i6400_int_mult): Include imul3nc type and update reservation.
7212 (i6400_int_div): Include idiv3 type and update reservation.
7213 (i6400_int_load): Update to check type not move_type.
7214 (i6400_int_store): Likewise.
7215 (i6400_int_prefetch): Set zero latency.
7217 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7219 * gcc.c: Document new %@{...} sequence.
7220 (LINK_COMMAND_SPEC): Use it for the -L switches.
7221 (cpp_unique_options): Use it for the -I switches.
7222 (at_file_argbuf): New global variable.
7223 (in_at_file): Likewise.
7224 (alloc_args): Create at_file_argbuf.
7225 (clear_args): Truncate at_file_argbuf.
7226 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7227 (open_at_file): New function.
7228 (close_at_file): Likewise.
7229 (create_at_file): Delete.
7230 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7233 (validate_switches_from_spec): Deal with %@{...} sequence.
7234 (validate_switches): Likewise.
7235 (driver::finalize): Call clear_args.
7237 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7239 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7241 2018-06-11 Martin Sebor <msebor@redhat.com>
7243 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7244 (Wno-class-memaccess): Add @opindex.
7245 (Wno-templates, Wno-multiple-inheritance): Same.
7246 (Wno-virtual-inheritance, Wno-namespaces): Same.
7247 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7248 (Wno-format-overflow, Wno-format-truncation): Same.
7249 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7250 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7251 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7252 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7253 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7254 (Wno-misspelled-isr): Same.
7256 2018-06-11 Martin Sebor <msebor@redhat.com>
7258 * PR tree-optimization/86083
7259 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7261 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7263 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7265 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7268 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7269 on the correct operand.
7270 (*movdi_internal64): Ditto.
7272 2018-06-11 Martin Liska <mliska@suse.cz>
7274 PR tree-optimization/86089
7275 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7277 2018-06-11 Julia Koval <julia.koval@intel.com>
7279 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7280 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7281 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7283 2018-06-11 Olivier Hainque <hainque@adacore.com>
7285 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7286 for Ada with strict dwarf2.
7288 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7291 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7294 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7296 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7298 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7300 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7303 2018-06-08 Martin Liska <mliska@suse.cz>
7305 * tree-cfg.h (debug_function): Fix argument type to match
7308 2018-06-08 Martin Liska <mliska@suse.cz>
7310 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7311 Remove usage of MPX-related (and removed) fields.
7312 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7314 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7316 * cfg.c (debug): Use TDF_NONE rather than 0.
7317 * cfghooks.c (debug): Likewise.
7318 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7319 (struct dump_option_value_info): Convert to...
7320 (struct kv_pair): ...this template type.
7321 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7323 (optinfo_verbosity_options): Likewise.
7324 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7326 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7327 than int for "optgroup_flags" param.
7328 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7330 (dump_dec): Likewise.
7331 (dump_finish): Use TDF_NONE rather than 0.
7332 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7333 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7334 than 0. Update for change to option_ptr.
7335 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7336 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7337 0. Update for changes to optinfo_verbosity_options and
7339 (opt_info_switch_p): Convert optgroup_flags from int to
7341 (dump_basic_block): Use dump_flags_t rather than int
7342 for "dump_kind" param.
7343 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7344 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7345 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7346 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7347 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7348 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7349 TDF_NONE): Convert from macros to...
7350 (enum dump_flag): ...this new enum.
7351 (dump_flags_t): Update to use enum.
7352 (operator|, operator&, operator~, operator|=, operator&=):
7353 Implement for dump_flags_t.
7354 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7355 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7356 Convert from macros to...
7357 (enum optgroup_flag): ...this new enum.
7358 (optgroup_flags_t): New typedef.
7359 (operator|, operator|=): Implement for optgroup_flags_t.
7360 (struct dump_file_info): Convert field "alt_flags" to
7361 dump_flags_t. Convert field "optgroup_flags" to
7363 (dump_basic_block): Use dump_flags_t rather than int for param.
7364 (dump_generic_expr_loc): Likewise.
7365 (dump_dec): Likewise.
7366 (dump_register): Convert param "optgroup_flags" to
7368 (opt_info_enable_passes): Likewise.
7369 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7371 * gimple-pretty-print.c (debug): Likewise.
7372 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7373 (merged_store_group::apply_stores): Likewise.
7374 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7375 * gimple.c (verify_gimple_pp): Likewise.
7376 * graphite-poly.c (print_pbb_body): Likewise.
7377 * passes.c (pass_manager::register_one_dump_file): Convert
7378 local "optgroup_flags" to optgroup_flags_t.
7379 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7381 (debug_body): Likewise.
7382 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7383 to optgroup_flags_t.
7384 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7386 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7387 (convert_mult_to_fma): Likewise.
7388 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7389 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7390 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7391 "dump_kind" to dump_flags_t.
7393 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7395 * config/rs6000/rs6000.c (min, max): Delete.
7397 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7399 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7400 -mabi=spe and -mabi=no-spe.
7402 2018-06-08 Martin Liska <mliska@suse.cz>
7404 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7405 where we expect an existing summary.
7407 2018-06-08 Martin Liska <mliska@suse.cz>
7409 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7410 * ipa-inline.h (estimate_edge_growth): Likewise.
7412 2018-06-08 Martin Liska <mliska@suse.cz>
7414 * cgraph.c (function_version_hasher::hash): Use
7415 cgraph_node::get_uid ().
7416 (function_version_hasher::equal):
7417 * cgraph.h (cgraph_node::get_uid): New method.
7418 * ipa-inline.c (update_caller_keys): Use
7419 cgraph_node::get_uid ().
7420 (update_callee_keys): Likewise.
7421 * ipa-utils.c (searchc): Likewise.
7422 (ipa_reduced_postorder): Likewise.
7423 * lto-cgraph.c (input_node): Likewise.
7424 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7425 * symbol-summary.h (symtab_insertion): Likewise.
7426 (symtab_removal): Likewise.
7427 (symtab_duplication): Likewise.
7428 * tree-pretty-print.c (dump_function_header): Likewise.
7429 * tree-sra.c (convert_callers_for_node): Likewise.
7431 2018-06-08 Martin Liska <mliska@suse.cz>
7433 * cgraph.c (symbol_table::create_edge): Always assign a new
7435 (symbol_table::free_edge): Do not recycle numbers.
7436 * cgraph.h (cgraph_edge::get): New method.
7437 * symbol-summary.h (symtab_removal): Use it.
7438 (symtab_duplication): Likewise.
7439 (call_summary::hashable_uid): Remove.
7441 2018-06-08 Martin Liska <mliska@suse.cz>
7443 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7444 (initialize_growth_caches): Remove.
7445 (free_growth_caches): Likewise.
7446 (do_estimate_edge_time): Use edge_growth_cache.
7447 (do_estimate_edge_size): Likewise.
7448 (do_estimate_edge_hints): Likewise.
7449 * ipa-inline.c (reset_edge_caches): Likewise.
7450 (recursive_inlining): Likewise.
7451 (inline_small_functions): Likewise.
7452 * ipa-inline.h (initialize_growth_caches): Remove.
7453 (estimate_edge_size): Likewise.
7454 (estimate_edge_time): Likewise.
7455 (estimate_edge_hints): Likewise.
7456 (reset_edge_growth_cache): Likewise.
7457 * symbol-summary.h (call_summary::remove): New method.
7459 2018-06-08 Martin Liska <mliska@suse.cz>
7461 * ipa-cp.c (class edge_clone_summary): New summary.
7462 (grow_edge_clone_vectors): Remove.
7463 (ipcp_edge_duplication_hook): Remove.
7464 (class edge_clone_summary_t): New call_summary class.
7465 (ipcp_edge_removal_hook): Remove.
7466 (edge_clone_summary_t::duplicate): New function.
7467 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7468 (create_specialized_node): Likewise.
7469 (ipcp_driver): Initialize edge_clone_summaries and do not
7472 2018-06-08 Martin Liska <mliska@suse.cz>
7474 * symbol-summary.h (get): New function.
7475 (call_summary::m_initialize_when_cloning): New class member.
7477 2018-06-08 Martin Liska <mliska@suse.cz>
7479 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7480 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7481 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7482 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7483 (remove_cgraph_node_from_order): Use the removed_nodes set.
7484 (do_per_function_toporder): Likwise.
7485 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7486 instead of summary_uid.
7487 (symtab_removal): Likewise.
7488 (symtab_duplication): Likewise.
7490 2018-06-08 Martin Liska <mliska@suse.cz>
7492 * ipa-cp.c (ipcp_store_bits_results): Use
7493 ipcp_transformation_sum.
7494 (ipcp_store_vr_results): Likewise.
7495 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7497 (ipcp_transformation_initialize): ... this.
7498 (ipa_set_node_agg_value_chain):
7499 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7500 (write_ipcp_transformation_info): Likewise.
7501 (read_ipcp_transformation_info): Likewise.
7502 (ipcp_update_bits): Likewise.
7503 (ipcp_update_vr): Likewise.
7504 (ipcp_transform_function): Likewise.
7505 * ipa-prop.h: Rename ipcp_transformation_summary to
7506 ipcp_transformation.
7507 (class ipcp_transformation_t): New function summary.
7508 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7509 (ipa_get_agg_replacements_for_node): Likewise.
7511 2018-06-08 Martin Liska <mliska@suse.cz>
7513 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7515 (class funct_state_summary_t): New function_summary class.
7516 (has_function_state): Remove.
7517 (get_function_state): Likewise.
7518 (set_function_state): Likewise.
7519 (add_new_function): Likewise.
7520 (funct_state_summary_t::insert): New function.
7521 (duplicate_node_data): Remove.
7522 (remove_node_data): Remove.
7523 (funct_state_summary_t::duplicate): New function.
7524 (register_hooks): Create new funct_state_summaries.
7525 (pure_const_generate_summary): Use it.
7526 (pure_const_write_summary): Likewise.
7527 (pure_const_read_summary): Likewise.
7528 (propagate_pure_const): Likewise.
7529 (propagate_nothrow): Likewise.
7530 (dump_malloc_lattice): Likewise.
7531 (propagate_malloc): Likewise.
7532 (execute): Do not register hooks, just remove summary
7534 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7537 2018-06-08 Martin Liska <mliska@suse.cz>
7539 * ipa-reference.c (remove_node_data): Remove.
7540 (duplicate_node_data): Likewise.
7541 (class ipa_ref_var_info_summary_t): New class.
7542 (class ipa_ref_opt_summary_t): Likewise.
7543 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7544 (get_reference_optimization_summary): Use
7545 ipa_ref_opt_sum_summaries.
7546 (set_reference_vars_info): Remove.
7547 (set_reference_optimization_summary): Likewise.
7548 (ipa_init): Create summaries.
7549 (init_function_info): Use function summary.
7550 (ipa_ref_opt_summary_t::duplicate): New function.
7551 (ipa_ref_opt_summary_t::remove): New function.
7552 (get_read_write_all_from_node): Fix GNU coding style.
7553 (propagate): Use function summary.
7554 (write_node_summary_p): Fix GNU coding style.
7555 (stream_out_bitmap): Likewise.
7556 (ipa_reference_read_optimization_summary): Use function summary.
7557 (ipa_reference_c_finalize): Do not release hooks.
7559 2018-06-08 Martin Liska <mliska@suse.cz>
7561 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7562 (analyze_function_body): Extract multiple calls of get_create.
7563 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7564 * ipa-inline.c (recursive_inlining): Use ::get method.
7565 * ipa-inline.h (estimate_edge_growth): Likewise.
7567 2018-06-08 Martin Liska <mliska@suse.cz>
7569 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7571 (hsa_function_summary::hsa_function_summary): Use the new enum
7573 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7574 * hsa-gen.c (hsa_get_host_function): Likewise.
7575 (get_brig_function_name): Likewise.
7576 * ipa-hsa.c (process_hsa_functions): Likewise.
7577 (ipa_hsa_write_summary): Likewise.
7578 * symbol-summary.h (symtab_duplication): Use ::get function/
7579 (get): New function.
7581 2018-06-08 Martin Liska <mliska@suse.cz>
7583 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7585 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7586 (hsa_register_kernel): Likewise.
7587 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7588 * hsa-gen.c (hsa_get_host_function): Likewise.
7589 (get_brig_function_name): Likewise.
7590 (generate_hsa): Likewise.
7591 (pass_gen_hsail::execute): Likewise.
7592 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7593 (devirtualization_time_bonus): Likewise.
7594 (ipcp_propagate_stage): Likewise.
7595 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7596 (edge_set_predicate): Likewise.
7597 (evaluate_conditions_for_known_args): Likewise.
7598 (evaluate_properties_for_edge): Likewise.
7599 (ipa_fn_summary::reset): Likewise.
7600 (ipa_fn_summary_t::duplicate): Likewise.
7601 (dump_ipa_call_summary): Likewise.
7602 (ipa_dump_fn_summary): Likewise.
7603 (analyze_function_body): Likewise.
7604 (compute_fn_summary): Likewise.
7605 (estimate_edge_devirt_benefit): Likewise.
7606 (estimate_edge_size_and_time): Likewise.
7607 (estimate_calls_size_and_time): Likewise.
7608 (estimate_node_size_and_time): Likewise.
7609 (inline_update_callee_summaries): Likewise.
7610 (remap_edge_change_prob): Likewise.
7611 (remap_edge_summaries): Likewise.
7612 (ipa_merge_fn_summary_after_inlining): Likewise.
7613 (ipa_update_overall_fn_summary): Likewise.
7614 (read_ipa_call_summary): Likewise.
7615 (inline_read_section): Likewise.
7616 (write_ipa_call_summary): Likewise.
7617 (ipa_fn_summary_write): Likewise.
7618 (ipa_free_fn_summary): Likewise.
7619 * ipa-hsa.c (process_hsa_functions): Likewise.
7620 (ipa_hsa_write_summary): Likewise.
7621 (ipa_hsa_read_section): Likewise.
7622 * ipa-icf.c (sem_function::merge): Likewise.
7623 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7624 (do_estimate_edge_time): Likewise.
7625 (estimate_size_after_inlining): Likewise.
7626 (estimate_growth): Likewise.
7627 (growth_likely_positive): Likewise.
7628 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7629 (inline_call): Likewise.
7630 * ipa-inline.c (caller_growth_limits): Likewise.
7631 (can_inline_edge_p): Likewise.
7632 (can_inline_edge_by_limits_p): Likewise.
7633 (compute_uninlined_call_time): Likewise.
7634 (compute_inlined_call_time): Likewise.
7635 (want_inline_small_function_p): Likewise.
7636 (edge_badness): Likewise.
7637 (update_caller_keys): Likewise.
7638 (update_callee_keys): Likewise.
7639 (recursive_inlining): Likewise.
7640 (inline_small_functions): Likewise.
7641 (inline_to_all_callers_1): Likewise.
7642 (dump_overall_stats): Likewise.
7643 (early_inline_small_functions): Likewise.
7644 (early_inliner): Likewise.
7645 * ipa-inline.h (estimate_edge_growth): Likewise.
7646 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7647 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7648 * ipa-prop.h (IPA_NODE_REF): Likewise.
7649 (IPA_EDGE_REF): Likewise.
7650 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7651 (propagate_malloc): Likewise.
7652 * ipa-split.c (execute_split_functions): Likewise.
7653 * symbol-summary.h: Rename get to get_create.
7655 (get_create): Likewise.
7656 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7658 2018-06-08 Martin Liska <mliska@suse.cz>
7660 * symbol-summary.h (release): Move definition out of class
7662 (symtab_removal): Likewise.
7663 (symtab_duplication): Likewise.
7665 2018-06-08 Martin Liska <mliska@suse.cz>
7667 * symbol-summary.h (function_summary): Move constructor
7668 implementation out of class declaration.
7669 (release): Likewise.
7670 (symtab_insertion): Likewise.
7671 (symtab_removal): Likewise.
7672 (symtab_duplication): Likewise.
7675 2018-06-08 Martin Liska <mliska@suse.cz>
7677 * Makefile.in: Remove support for MPX (macros, related functions,
7678 fields in cgraph_node, ...).
7679 * builtin-types.def (BT_BND): Likewise.
7680 (BT_FN_BND_CONST_PTR): Likewise.
7681 (BT_FN_CONST_PTR_BND): Likewise.
7682 (BT_FN_VOID_PTR_BND): Likewise.
7683 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7684 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7685 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7686 (expand_builtin_mempcpy_with_bounds): Likewise.
7687 (expand_builtin_memset_with_bounds): Likewise.
7688 (expand_builtin_memset_args): Likewise.
7689 (std_expand_builtin_va_start): Likewise.
7690 (expand_builtin): Likewise.
7691 (expand_builtin_with_bounds): Likewise.
7692 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7693 (DEF_LIB_BUILTIN_CHKP): Likewise.
7694 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7695 (DEF_CHKP_BUILTIN): Likewise.
7696 (BUILT_IN_MEMCPY): Likewise.
7697 (BUILT_IN_MEMMOVE): Likewise.
7698 (BUILT_IN_MEMPCPY): Likewise.
7699 (BUILT_IN_MEMSET): Likewise.
7700 (BUILT_IN_STPCPY): Likewise.
7701 (BUILT_IN_STRCAT): Likewise.
7702 (BUILT_IN_STRCHR): Likewise.
7703 (BUILT_IN_STRCPY): Likewise.
7704 (BUILT_IN_STRLEN): Likewise.
7705 (BUILT_IN_MEMCPY_CHK): Likewise.
7706 (BUILT_IN_MEMMOVE_CHK): Likewise.
7707 (BUILT_IN_MEMPCPY_CHK): Likewise.
7708 (BUILT_IN_MEMSET_CHK): Likewise.
7709 (BUILT_IN_STPCPY_CHK): Likewise.
7710 (BUILT_IN_STRCAT_CHK): Likewise.
7711 (BUILT_IN_STRCPY_CHK): Likewise.
7712 * calls.c (store_bounds): Likewise.
7713 (emit_call_1): Likewise.
7714 (special_function_p): Likewise.
7715 (maybe_warn_nonstring_arg): Likewise.
7716 (initialize_argument_information): Likewise.
7717 (finalize_must_preallocate): Likewise.
7718 (compute_argument_addresses): Likewise.
7719 (expand_call): Likewise.
7720 * cfgexpand.c (expand_call_stmt): Likewise.
7721 (expand_return): Likewise.
7722 (expand_gimple_stmt_1): Likewise.
7723 (pass_expand::execute): Likewise.
7724 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7725 (cgraph_node::remove): Likewise.
7726 (cgraph_node::dump): Likewise.
7727 (cgraph_node::verify_node): Likewise.
7728 * cgraph.h (chkp_function_instrumented_p): Likewise.
7729 (symtab_node::get_alias_target): Likewise.
7730 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7731 (cgraph_local_p): Likewise.
7732 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7733 (cgraph_edge::rebuild_references): Likewise.
7734 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7735 (walk_polymorphic_call_targets): Likewise.
7736 (cgraph_node::expand_thunk): Likewise.
7737 (symbol_table::output_weakrefs): Likewise.
7738 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7739 (ix86_handle_option): Likewise.
7740 * config/i386/constraints.md: Likewise.
7741 * config/i386/i386-builtin-types.def (BND): Likewise.
7745 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7746 (BDESC_FIRST): Likewise.
7748 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7749 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7750 * config/i386/i386.c (enum reg_class): Likewise.
7751 (ix86_target_string): Likewise.
7752 (ix86_option_override_internal): Likewise.
7753 (ix86_conditional_register_usage): Likewise.
7754 (ix86_valid_target_attribute_inner_p): Likewise.
7755 (ix86_set_indirect_branch_type): Likewise.
7756 (ix86_set_current_function): Likewise.
7757 (ix86_function_arg_regno_p): Likewise.
7758 (init_cumulative_args): Likewise.
7759 (ix86_function_arg_advance): Likewise.
7760 (ix86_function_arg): Likewise.
7761 (ix86_pass_by_reference): Likewise.
7762 (ix86_function_value_regno_p): Likewise.
7763 (ix86_function_value_1): Likewise.
7764 (ix86_function_value_bounds): Likewise.
7765 (ix86_return_in_memory): Likewise.
7766 (ix86_setup_incoming_vararg_bounds): Likewise.
7767 (ix86_va_start): Likewise.
7768 (indirect_thunk_need_prefix): Likewise.
7769 (print_reg): Likewise.
7770 (ix86_print_operand): Likewise.
7771 (ix86_expand_call): Likewise.
7772 (ix86_output_function_return): Likewise.
7773 (reg_encoded_number): Likewise.
7774 (BDESC_VERIFYS): Likewise.
7775 (ix86_init_mpx_builtins): Likewise.
7776 (ix86_init_builtins): Likewise.
7777 (ix86_emit_cmove): Likewise.
7778 (ix86_emit_move_max): Likewise.
7779 (ix86_expand_builtin): Likewise.
7780 (ix86_builtin_mpx_function): Likewise.
7781 (ix86_get_arg_address_for_bt): Likewise.
7782 (ix86_load_bounds): Likewise.
7783 (ix86_store_bounds): Likewise.
7784 (ix86_load_returned_bounds): Likewise.
7785 (ix86_store_returned_bounds): Likewise.
7786 (ix86_class_likely_spilled_p): Likewise.
7787 (ix86_hard_regno_mode_ok): Likewise.
7788 (x86_order_regs_for_local_alloc): Likewise.
7789 (ix86_mitigate_rop): Likewise.
7790 (ix86_bnd_prefixed_insn_p): Likewise.
7791 (ix86_mpx_bound_mode): Likewise.
7792 (ix86_make_bounds_constant): Likewise.
7793 (ix86_initialize_bounds): Likewise.
7794 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7795 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7796 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7797 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7798 (TARGET_CHKP_BOUND_MODE): Likewise.
7799 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7800 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7801 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7802 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7803 * config/i386/i386.h (TARGET_MPX): Likewise.
7804 (TARGET_MPX_P): Likewise.
7805 (VALID_BND_REG_MODE): Likewise.
7806 (FIRST_BND_REG): Likewise.
7807 (LAST_BND_REG): Likewise.
7808 (enum reg_class): Likewise.
7809 (BND_REG_P): Likewise.
7810 (BND_REGNO_P): Likewise.
7811 (BNDmode): Likewise.
7812 (ADJUST_INSN_LENGTH): Likewise.
7813 * config/i386/i386.md: Likewise.
7814 * config/i386/i386.opt: Likewise.
7815 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7816 (defined): Likewise.
7817 (LINK_MPX): Likewise.
7818 (MPX_SPEC): Likewise.
7819 (LIBMPX_SPEC): Likewise.
7820 (LIBMPXWRAPPERS_SPEC): Likewise.
7821 (CHKP_SPEC): Likewise.
7822 * config/i386/predicates.md: Likewise.
7823 * dbxout.c (dbxout_type): Likewise.
7824 * doc/extend.texi: Likewise.
7825 * doc/invoke.texi: Likewise.
7826 * doc/md.texi: Likewise.
7827 * doc/tm.texi: Likewise.
7828 * doc/tm.texi.in: Likewise.
7829 * dwarf2out.c (is_base_type): Likewise.
7830 (gen_formal_types_die): Likewise.
7831 (gen_subprogram_die): Likewise.
7832 (gen_type_die_with_usage): Likewise.
7833 (gen_decl_die): Likewise.
7834 (dwarf2out_late_global_decl): Likewise.
7835 * expr.c (expand_assignment): Likewise.
7836 (emit_storent_insn): Likewise.
7837 (store_expr_with_bounds): Likewise.
7838 (store_expr): Likewise.
7839 (expand_expr_real_1): Likewise.
7840 * expr.h (store_expr_with_bounds): Likewise.
7841 * function.c (use_register_for_decl): Likewise.
7842 (struct bounds_parm_data): Likewise.
7843 (assign_parms_augmented_arg_list): Likewise.
7844 (assign_parm_find_entry_rtl): Likewise.
7845 (assign_parm_is_stack_parm): Likewise.
7846 (assign_parm_load_bounds): Likewise.
7847 (assign_bounds): Likewise.
7848 (assign_parms): Likewise.
7849 (expand_function_start): Likewise.
7850 * gcc.c (CHKP_SPEC): Likewise.
7851 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7852 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7853 (wrestrict_dom_walker::check_call): Likewise.
7854 * gimple.c (gimple_build_call_from_tree): Likewise.
7855 * gimple.h (enum gf_mask): Likewise.
7856 (gimple_call_with_bounds_p): Likewise.
7857 (gimple_call_set_with_bounds): Likewise.
7858 * gimplify.c (gimplify_init_constructor): Likewise.
7859 * ipa-cp.c (initialize_node_lattices): Likewise.
7860 (propagate_constants_across_call): Likewise.
7861 (find_more_scalar_values_for_callers_subset): Likewise.
7862 * ipa-hsa.c (process_hsa_functions): Likewise.
7863 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7864 * ipa-icf.c (sem_function::merge): Likewise.
7865 * ipa-inline.c (early_inliner): Likewise.
7866 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7867 (warn_function_cold): Likewise.
7868 (propagate_pure_const): Likewise.
7869 * ipa-ref.h (enum GTY): Likewise.
7870 * ipa-split.c (find_retbnd): Likewise.
7871 (consider_split): Likewise.
7872 (split_function): Likewise.
7873 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7874 * ipa.c (walk_polymorphic_call_targets): Likewise.
7875 (symbol_table::remove_unreachable_nodes): Likewise.
7876 (process_references): Likewise.
7877 (cgraph_build_static_cdtor_1): Likewise.
7878 * lto-cgraph.c (lto_output_node): Likewise.
7879 (output_refs): Likewise.
7880 (compute_ltrans_boundary): Likewise.
7881 (input_overwrite_node): Likewise.
7882 (input_node): Likewise.
7883 (input_cgraph_1): Likewise.
7884 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7885 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7886 (class pass_chkp_instrumentation_passes): Likewise.
7887 (make_pass_chkp_instrumentation_passes): Likewise.
7888 * passes.def: Likewise.
7889 * rtl.h (struct GTY): Likewise.
7890 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7891 * stor-layout.c (layout_type): Likewise.
7892 * symtab.c: Likewise.
7893 * target.def: Likewise.
7894 * targhooks.c (default_chkp_bound_type): Likewise.
7895 (default_chkp_bound_mode): Likewise.
7896 (default_builtin_chkp_function): Likewise.
7897 (default_chkp_function_value_bounds): Likewise.
7898 (default_chkp_make_bounds_constant): Likewise.
7899 (default_chkp_initialize_bounds): Likewise.
7900 * targhooks.h (default_chkp_bound_type): Likewise.
7901 (default_chkp_bound_mode): Likewise.
7902 (default_builtin_chkp_function): Likewise.
7903 (default_chkp_function_value_bounds): Likewise.
7904 (default_chkp_make_bounds_constant): Likewise.
7905 (default_chkp_initialize_bounds): Likewise.
7906 * toplev.c (compile_file): Likewise.
7907 (process_options): Likewise.
7908 * tree-core.h (DEF_BUILTIN): Likewise.
7909 (DEF_BUILTIN_CHKP): Likewise.
7910 * tree-inline.c (declare_return_variable): Likewise.
7911 (remap_gimple_stmt): Likewise.
7912 (copy_bb): Likewise.
7913 (initialize_inlined_parameters): Likewise.
7914 (expand_call_inline): Likewise.
7915 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7916 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7917 (make_pass_ipa_chkp_produce_thunks): Likewise.
7918 (make_pass_chkp): Likewise.
7919 (make_pass_chkp_opt): Likewise.
7920 (make_pass_chkp_instrumentation_passes): Likewise.
7921 * tree-pretty-print.c (dump_generic_node): Likewise.
7922 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7923 * tree-ssa-dce.c (propagate_necessity): Likewise.
7924 (eliminate_unnecessary_stmts): Likewise.
7925 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7926 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7927 * tree-ssa-sccvn.h: Likewise.
7928 * tree-ssa-strlen.c (get_string_length): Likewise.
7929 (valid_builtin_call): Likewise.
7930 (adjust_last_stmt): Likewise.
7931 (handle_builtin_strchr): Likewise.
7932 (handle_builtin_strcpy): Likewise.
7933 (handle_builtin_stxncpy): Likewise.
7934 (handle_builtin_memcpy): Likewise.
7935 (handle_builtin_strcat): Likewise.
7936 (strlen_check_and_optimize_stmt): Likewise.
7937 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7938 * tree-streamer-in.c: Likewise.
7939 * tree-streamer.c (record_common_node): Likewise.
7940 * tree.c (tree_code_size): Likewise.
7941 (wide_int_to_tree_1): Likewise.
7942 (type_contains_placeholder_1): Likewise.
7943 (build_common_tree_nodes): Likewise.
7944 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7945 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7946 (POINTER_BOUNDS_P): Likewise.
7947 (BOUNDED_TYPE_P): Likewise.
7948 (BOUNDED_P): Likewise.
7949 (CALL_WITH_BOUNDS_P): Likewise.
7950 (pointer_bounds_type_node): Likewise.
7951 * value-prof.c (gimple_ic): Likewise.
7952 * var-tracking.c (vt_add_function_parameters): Likewise.
7953 * varasm.c (make_decl_rtl): Likewise.
7954 (assemble_start_function): Likewise.
7955 (output_constant): Likewise.
7956 (maybe_assemble_visibility): Likewise.
7957 * varpool.c (ctor_for_folding): Likewise.
7958 * chkp-builtins.def: Remove.
7959 * ipa-chkp.c: Remove.
7960 * ipa-chkp.h: Remove.
7961 * rtl-chkp.c: Remove.
7962 * rtl-chkp.h: Remove.
7963 * tree-chkp-opt.c: Remove.
7964 * tree-chkp.c: Remove.
7965 * tree-chkp.h: Remove.
7967 2018-06-07 Carl Love <cel@us.ibm.com>
7969 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7970 vextract_fp_from_shortl): Add BE support.
7972 2018-06-07 Paul Koning <ni1d@arrl.net>
7974 * compare-elim.c (try_merge_compare): Don't merge compare if
7975 address contains a side effect.
7976 (try_eliminate_compare): Likewise.
7978 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7980 * config.gcc: Support "tremont".
7981 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7982 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7984 * config/i386/i386.c (m_TREMONT): Define.
7985 (processor_target_table): Add "tremont".
7986 (PTA_TREMONT): Define.
7987 (ix86_lea_outperforms): Add TARGET_TREMONT.
7988 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7989 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7990 and M_INTEL_GOLDMONT_PLUS.
7991 (fold_builtin_cpu): Add "tremont".
7992 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7993 (ix86_option_override_internal): Add "tremont".
7994 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7995 (processor_type): Add PROCESSOR_TREMONT.
7996 * config/i386/x86-tune.def: Add m_TREMONT.
7997 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7999 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8001 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
8002 symbol defined for msp430i* devices to be lower case.
8004 2018-06-07 Richard Biener <rguenther@suse.de>
8006 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
8007 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
8008 Properly wrap signed arithmetic if overflow wraps.
8010 2018-06-07 Jakub Jelinek <jakub@redhat.com>
8012 PR tree-optimization/69615
8013 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
8014 of a cast from a same precision integral SSA_NAME in a bb dominated
8015 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
8016 cast to utype if rhs2 has already a compatible type.
8018 2018-06-07 Richard Biener <rguenther@suse.de>
8020 PR tree-optimization/85935
8021 * graphite-scop-detection.c (find_params_in_bb): Analyze
8022 condition operands with respect to the correct loop. Assert
8023 the analysis doesn't fail.
8025 2018-06-04 Carl Love <cel@us.ibm.com>
8027 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
8028 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
8029 as it is slightly cheaper.
8030 (first_match_or_eos_index_<mode>):
8031 Calculate index using natural element order.
8032 (first_match_index_<mode>):
8033 Calculate index using natural element order.
8034 (first_match_or_eos_index_<mode>):
8035 Calculate index using natural order.
8036 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
8037 for BE and LE modes.
8038 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
8039 P9V_BUILTIN_VCLZLSBB_V16QI.
8040 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
8043 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8045 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
8046 indentation and line wrap for many prototypes. Add missing
8047 @smallexample directives around block of prototypes for vec_xl and
8050 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
8052 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
8053 track if we pass or return IEEE 128-bit floating point.
8054 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
8055 C++ mangling that is compatible with GCC 8.1.
8056 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8057 (init_cumulative_args): Note if we pass or return IEEE 128-bit
8058 floating point types.
8059 (rs6000_function_arg_advance_1): Likewise.
8060 (rs6000_mangle_type): Optionally generate mangled names that match
8061 what GCC 8.1 generated for IEEE 128-bit floating point types.
8062 (rs6000_globalize_decl_name): If we have an external function that
8063 passes or returns IEEE 128-bit types, generate a weak reference
8064 from the mangled name used in GCC 8.1 to the current mangled
8066 (rs6000_init_builtins): Make __ibm128 use the long double type if
8067 long double is IBM extended double. Make __float128 use the long
8068 double type if long double is IEEE 128-bit.
8071 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
8072 macro for __ibm128 built-in functions.
8073 (PACK_IF): Add __ibm128 pack/unpack functions.
8074 (UNPACK_IF): Likewise.
8075 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
8076 enable long double built-in functions if long double is IEEE
8077 128-bit floating point.
8078 (rs6000_invalid_builtin): Update long double built-in function
8080 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
8081 functions, adjust the built-in function to use the long double
8082 built-in function if __ibm128 and long double are the same type.
8083 * doc/extend.texi (PowerPC builtins): Update documention for
8084 __builtin_{,un}pack_longdouble. Add documentation for
8085 __builtin_{,un}pack_ibm128.
8087 2018-06-06 Jim Wilson <jimw@sifive.com>
8089 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
8090 (struct machine_function): New field interrupt_mode.
8091 (riscv_handle_type_attribute): New function. Add forward declaration.
8092 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
8093 (riscv_expand_epilogue): Check interrupt_mode field.
8094 (riscv_set_current_function): Check interrupt attribute args and
8095 set interrupt_mode field.
8096 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
8097 (riscv_sret, riscv_uret): New.
8098 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
8099 new arguments to interrupt attribute.
8101 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
8104 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
8105 Don't handle -mcpu=power8 if -mpower9-vector is also used.
8107 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8109 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
8110 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
8111 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8112 several redundant entries.
8114 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8116 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8117 type from "rtx" to "rtx_insn *".
8118 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8119 for local "call_insn", removing cast.
8120 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8122 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8124 PR tree-optimization/86066
8125 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8126 for BIT_INSERT_EXPR stores.
8128 2018-06-06 Richard Biener <rguenther@suse.de>
8130 PR tree-optimization/86062
8131 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8132 component refs ontop
8133 of to be offsetted base.
8135 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8137 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8138 to be static and remove check on interrupt attribute name.
8140 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8142 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8143 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8145 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8148 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8150 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8151 Remove second argument, change how error is called.
8152 (aarch64_layout_arg): Remove second argument from
8153 aarch64_err_no_fpadvsimd call.
8154 (aarch64_init_cumulative_args): Ditto.
8155 (aarch64_gimplify_va_arg_expr): Ditto.
8156 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8158 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8160 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8161 (*simple_return_indirect_internal<mode>): Rename from
8162 simple_return_indirect_internal. Use W mode iterator.
8163 (rstorssp): New expander.
8164 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8165 (clrssbsy): New expander.
8166 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8168 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8170 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8172 (cmse_check_pointed_object): Likewise.
8174 2018-06-05 Martin Liska <mliska@suse.cz>
8176 PR gcov-profile/47618
8177 * doc/invoke.texi: Document how -fprofile-dir format
8180 2018-06-05 Richard Biener <rguenther@suse.de>
8182 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8183 removal pretend DOM info isn't available so we do not update
8184 it and only remove edges, not dominated blocks. Actually free
8185 DOM info in case we removed something. Remove unreachable blocks.
8186 (mfb_keep_latches): Work with either DOM info or marked backedges.
8187 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8188 first. Mark backedges if DOM info isn't available.
8189 (Re-)compute DOM info after cleanup_control_flow_pre.
8191 2018-06-05 Richard Biener <rguenther@suse.de>
8193 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8194 (locus_discrim_hasher::hash): Adjust.
8195 (locus_discrim_hasher::equal): Likewise.
8196 (next_discriminator_for_locus): Work on line directly.
8197 (same_line_p): Pass in expanded locus1 as well.
8198 (assign_discriminators): Avoid redundant location expansions.
8200 2018-06-05 Richard Biener <rguenther@suse.de>
8202 PR tree-optimization/86046
8203 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8204 if required after clearing TREE_ADDRESSABLE.
8206 2018-06-05 Richard Biener <rguenther@suse.de>
8208 PR tree-optimization/86047
8209 * tree-ssa-loop.c (for_each_index): Glob handling of all
8210 decls and constants and really handle all of them.
8212 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8215 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8216 qualifier_void_pointer and qualifier_const_void_pointer.
8217 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8218 (arm_init_builtins): Handle the above.
8219 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8220 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8221 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8224 2018-06-05 Martin Liska <mliska@suse.cz>
8226 * auto-profile.c (read_autofdo_file): Do not use
8227 gcov_ctr_summary struct.
8228 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8229 * coverage.c (struct counts_entry): Likewise.
8230 (read_counts_file): Read just single summary entry.
8231 (get_coverage_counts): Use gcov_summary struct.
8232 * coverage.h (get_coverage_counts): Likewise.
8233 * gcov-dump.c (dump_working_sets): Likewise.
8234 (tag_summary): Dump just single summary.
8235 * gcov-io.c (gcov_write_summary): Write just histogram
8237 (gcov_read_summary): Read just single summary.
8238 (compute_working_sets): Use gcov_summary struct.
8239 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8240 of GCOV_COUNTERS_SUMMABLE.
8241 (GCOV_COUNTERS_SUMMABLE): Remove.
8242 (GCOV_FIRST_VALUE_COUNTER): Replace with
8243 GCOV_COUNTER_V_INTERVAL.
8244 (struct gcov_ctr_summary): Remove.
8245 (struct gcov_summary): Directly use fields of former
8247 (compute_working_sets): Use gcov_summary struct.
8248 * gcov.c (read_count_file): Do not use ctrs fields.
8249 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8251 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8253 * profile.c: Likewise.
8254 * profile.h: Likewise.
8256 2018-06-05 Martin Liska <mliska@suse.cz>
8258 PR gcov-profile/84846
8259 * gcov.c (output_lines): Print working directory only
8260 in intermediate format.
8262 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8264 * config/s390/s390-builtin-types.def: Add void function type.
8265 * config/s390/s390-builtins.def: Use the function type for the
8268 2018-06-04 Jim Wilson <jimw@sifive.com>
8270 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8272 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8273 and maybe_eh_return. Change regno to unsigned int. Use new args to
8274 handle EH_RETURN_DATA_REGNO registers properly.
8275 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8276 (riscv_expand_epilogue): Update comment. Change argument name and
8277 type. Update code to use new name and type. Pass new args to
8278 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8280 * config/riscv/riscv.md (NORMAL_RETURN): New.
8281 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8282 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8283 (eh_return): Call gen_eh_return_internal and emit barrier.
8284 (eh_return_internal): Call riscv_expand_epilogue.
8286 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8288 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8289 bit_insertion field and declare can_be_merged_into method.
8290 (merged_store_group::can_be_merged_into): New method.
8291 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8292 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8293 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8295 2018-06-04 Richard Biener <rguenther@suse.de>
8297 PR tree-optimization/85955
8298 * builtins.c (fold_builtin_sincos): Convert pointers to
8299 destination to appropriate type before dereferencing.
8301 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8303 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8305 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8307 * expr.c (expand_expr_real_1): Force the operand into memory if
8308 its TYPE_MODE is BLKmode and if there is no integer mode for
8309 the number of bits being extracted.
8311 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8315 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8316 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8318 2018-06-04 Richard Biener <rguenther@suse.de>
8320 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8321 (cleanup_tree_cfg_noloop): ... single caller. Do
8322 start_recording_case_labels later.
8324 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8326 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8327 to _IMMINTRIN_H_INCLUDED.
8328 * config/i386/pconfigintrin.h: Ditto.
8329 * config/i386/waitpkgintrin.h: Ditto.
8330 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8331 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8332 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8333 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8334 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8335 waitpkgintrin.h and cldemoteintrin.h.
8337 2018-06-04 Richard Biener <rguenther@suse.de>
8339 PR tree-optimization/86038
8340 * tracer.c (find_best_successor): Check probability for
8341 being initialized, bail out if not.
8343 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8346 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8347 of bits to ignore when comparing architectures.
8349 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8351 PR tree-optimization/69615
8352 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8353 maximum or minimum of the type, try to merge it also as if
8354 range1 is + [-, x - 1] or + [x + 1, -].
8357 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8359 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8361 PR tree-optimization/86034
8362 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8363 the unsigned bitfield type in a bit insertion sequence if it does not
8364 have a larger precision than the bitfield size.
8365 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8367 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8369 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8371 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8373 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8374 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8375 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8376 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8378 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8380 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8381 Disable -fdelete-null-pointer-checks for ELF toolchain.
8383 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8384 Kito Cheng <kito.cheng@gmail.com>
8386 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8387 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8388 (nds32*-*-*): Add glibc and uclibc conditions.
8389 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8390 (TARGET_EXCEPT_UNWIND_INFO): Define.
8391 * config/nds32/elf.h: New file.
8392 * config/nds32/linux.h: New file.
8393 * config/nds32/nds32-elf.opt: New file.
8394 * config/nds32/nds32-linux.opt: New file.
8395 * config/nds32/nds32-fp-as-gp.c
8396 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8397 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8399 (nds32_asm_file_end): Ditto.
8400 (nds32_print_operand): Ditto.
8401 (nds32_insert_attributes): Ditto.
8402 (nds32_init_libfuncs): New function.
8403 (TARGET_HAVE_TLS): Define.
8404 (TARGET_INIT_LIBFUNCS): Define.
8405 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8407 (TARGET_ELF): Apply different mcmodel setting.
8408 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8409 been migrated into elf.h and linux.h files.
8410 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8411 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8412 (mcmodel): The content has been migrated into nds32-elf.opt and
8413 nds32-linux.opt files.
8414 * config/nds32/t-elf: New file.
8415 * config/nds32/t-linux: New file.
8417 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8418 Shiva Chen <shiva0217@gmail.com>
8420 * config/nds32/constants.md (unspec_volatile_element): Add
8421 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8422 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8424 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8425 (make_pass_nds32_fp_as_gp): Declare.
8426 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8428 (nds32_asm_function_end_prologue): Remove unused asm output.
8429 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8430 (nds32_asm_file_start): Output necessary fp_as_gp information.
8431 (nds32_option_override): Adjust register usage.
8432 (nds32_expand_prologue): Consider fp_as_gp situation.
8433 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8434 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8437 (simple_return): Ditto.
8438 (omit_fp_begin): Output special directive for fp_as_gp.
8439 (omit_fp_end): Output special directive for fp_as_gp.
8440 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8441 mforbid-fp-as-gp): New options.
8443 2018-06-01 Mark Wielaard <mark@klomp.org>
8445 * dwarf2out.c (dwarf2out_finish): Remove generation of
8446 DW_AT_loclists_base.
8448 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8450 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8451 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8452 (struct merged_store_group): Add bit_insertion field.
8453 (dump_char_array): Use standard hexadecimal format.
8454 (merged_store_group::merged_store_group): Set bit_insertion to false.
8455 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8456 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8457 also print the mask in the dump file.
8458 (pass_store_merging::gate): Minor tweak.
8459 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8460 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8461 stores with INTEGER_CST stores.
8462 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8463 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8464 and use it throughout. Generate bit insertion sequences if need be.
8465 (pass_store_merging::process_store): Remove redundant condition.
8466 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8468 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8470 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8471 the 128-bit floating point types. Fix function comment.
8473 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8475 * config/aarch64/aarch64-simd.md
8476 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8478 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8481 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8483 PR tree-optimization/85989
8484 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8486 (backprop::intersect_uses): Check it when deciding whether this
8487 is a backedge reference.
8488 (backprop::process_block): Add each phi to m_visited_phis
8489 after visiting it, then clear it at the end.
8491 2018-06-01 Richard Biener <rguenther@suse.de>
8493 * tree-vectorizer.h (vect_dr_stmt): New function.
8494 (vect_get_load_cost): Adjust.
8495 (vect_get_store_cost): Likewise.
8496 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8497 Use vect_dr_stmt instead of DR_SMTT.
8498 (vect_record_base_alignments): Likewise.
8499 (vect_calculate_target_alignment): Likewise.
8500 (vect_compute_data_ref_alignment): Likewise and make static.
8501 (vect_update_misalignment_for_peel): Likewise.
8502 (vect_verify_datarefs_alignment): Likewise.
8503 (vector_alignment_reachable_p): Likewise.
8504 (vect_get_data_access_cost): Likewise. Pass down
8505 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8506 (vect_get_peeling_costs_all_drs): Likewise.
8507 (vect_peeling_hash_get_lowest_cost): Likewise.
8508 (vect_enhance_data_refs_alignment): Likewise.
8509 (vect_find_same_alignment_drs): Likewise.
8510 (vect_analyze_data_refs_alignment): Likewise.
8511 (vect_analyze_group_access_1): Likewise.
8512 (vect_analyze_group_access): Likewise.
8513 (vect_analyze_data_ref_access): Likewise.
8514 (vect_analyze_data_ref_accesses): Likewise.
8515 (vect_vfa_segment_size): Likewise.
8516 (vect_small_gap_p): Likewise.
8517 (vectorizable_with_step_bound_p): Likewise.
8518 (vect_prune_runtime_alias_test_list): Likewise.
8519 (vect_analyze_data_refs): Likewise.
8520 (vect_supportable_dr_alignment): Likewise.
8521 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8522 (vect_gen_prolog_loop_niters): Likewise.
8523 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8524 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8526 (vect_recog_mask_conversion_pattern): Likewise.
8527 (vect_try_gather_scatter_pattern): Likewise.
8528 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8529 to vect_get_store_cost.
8530 (vect_get_store_cost): Get stmt_info instead of DR.
8531 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8532 (vect_get_load_cost): Get stmt_info instead of DR.
8534 2018-06-01 Richard Biener <rguenther@suse.de>
8537 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8538 vars wrapped in MEM_REFs.
8540 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8542 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8543 Fix subreg tests so that we only return a choice between
8544 GENERAL_REGS and FP_REGS if the original classes included both.
8546 2018-06-01 Richard Biener <rguenther@suse.de>
8549 * tree-ssa-structalias.c (get_function_part_constraint):
8550 Handle NULL fi->decl.
8551 (find_func_aliases_for_call): Properly handle indirect
8552 fi from direct call.
8553 (find_func_clobbers): Likewise.
8554 (ipa_pta_execute): Likewise.
8555 (create_variable_info_for): For functions that are ifunc_resolver
8556 resolve to a varinfo that contains the result of the resolver call.
8557 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8560 2018-05-31 Michael Collison <michael.collison@arm.com>
8562 * config/aarch64/aarch64.md:
8563 (*fix_to_zero_extenddfdi2): New pattern.
8564 * gcc.target/aarch64/fix_extend1.c: New testcase.
8566 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8570 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8571 and BUILT_IN_STRNCMP_EQ.
8572 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8573 BUILT_IN_STRNCMP_EQ.
8574 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8575 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8576 (gimple_fold_builtin): Likewise.
8577 * tree-ssa-strlen.c (compute_string_length): New function.
8578 (determine_min_obsize): New function.
8579 (handle_builtin_string_cmp): New function to handle calls to
8580 string compare functions.
8581 (strlen_optimize_stmt): Add handling to builtin string compare
8583 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8584 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8585 * tree.c (build_common_builtin_nodes): Add new defines of
8586 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8588 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8591 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8592 functions with naked attribute.
8594 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8596 * config/i386/sse.md (avx_vec_concat<mode>):
8597 Substitute concat_tg_mode mode attribute with xtg_mode.
8598 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8599 (concat_tg_mode): Remove mode attribute.
8601 2018-05-31 Martin Sebor <msebor@redhat.com>
8604 * calls.c (alloc_max_size): Correct a logic error/typo.
8605 Treat excessive arguments as infinite. Warn for invalid arguments.
8606 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8608 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8611 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8612 and movx for Haswell.
8614 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8615 Cesar Philippidis <cesar@codesourcery.com>
8618 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8619 when emitting error on private/firstprivate reductions.
8620 * omp-low.c (lower_omp_target): Avoid reference-type processing
8621 on pointers for firstprivate clause.
8623 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8625 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8628 * config/aarch64/aarch64-simd.md
8629 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8630 (aarch64_ld1_x3_<mode>): Likewise
8631 (aarch64_st1x2<VALLDIF:mode>): Likewise
8632 (aarch64_st1_x2_<mode>): Likewise
8633 (aarch64_st1x3<VALLDIF:mode>): Likewise
8634 (aarch64_st1_x3_<mode>): Likewise
8635 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8636 (vld1_s8_x3): Likewise.
8637 (vld1_u16_x3): Likewise.
8638 (vld1_s16_x3): Likewise.
8639 (vld1_u32_x3): Likewise.
8640 (vld1_s32_x3): Likewise.
8641 (vld1_u64_x3): Likewise.
8642 (vld1_s64_x3): Likewise.
8643 (vld1_f16_x3): Likewise.
8644 (vld1_f32_x3): Likewise.
8645 (vld1_f64_x3): Likewise.
8646 (vld1_p8_x3): Likewise.
8647 (vld1_p16_x3): Likewise.
8648 (vld1_p64_x3): Likewise.
8649 (vld1q_u8_x3): Likewise.
8650 (vld1q_s8_x3): Likewise.
8651 (vld1q_u16_x3): Likewise.
8652 (vld1q_s16_x3): Likewise.
8653 (vld1q_u32_x3): Likewise.
8654 (vld1q_s32_x3): Likewise.
8655 (vld1q_u64_x3): Likewise.
8656 (vld1q_s64_x3): Likewise.
8657 (vld1q_f16_x3): Likewise.
8658 (vld1q_f32_x3): Likewise.
8659 (vld1q_f64_x3): Likewise.
8660 (vld1q_p8_x3): Likewise.
8661 (vld1q_p16_x3): Likewise.
8662 (vld1q_p64_x3): Likewise.
8663 (vst1_s64_x2): Likewise.
8664 (vst1_u64_x2): Likewise.
8665 (vst1_f64_x2): Likewise.
8666 (vst1_s8_x2): Likewise.
8667 (vst1_p8_x2): Likewise.
8668 (vst1_s16_x2): Likewise.
8669 (vst1_p16_x2): Likewise.
8670 (vst1_s32_x2): Likewise.
8671 (vst1_u8_x2): Likewise.
8672 (vst1_u16_x2): Likewise.
8673 (vst1_u32_x2): Likewise.
8674 (vst1_f16_x2): Likewise.
8675 (vst1_f32_x2): Likewise.
8676 (vst1_p64_x2): Likewise.
8677 (vst1q_s8_x2): Likewise.
8678 (vst1q_p8_x2): Likewise.
8679 (vst1q_s16_x2): Likewise.
8680 (vst1q_p16_x2): Likewise.
8681 (vst1q_s32_x2): Likewise.
8682 (vst1q_s64_x2): Likewise.
8683 (vst1q_u8_x2): Likewise.
8684 (vst1q_u16_x2): Likewise.
8685 (vst1q_u32_x2): Likewise.
8686 (vst1q_u64_x2): Likewise.
8687 (vst1q_f16_x2): Likewise.
8688 (vst1q_f32_x2): Likewise.
8689 (vst1q_f64_x2): Likewise.
8690 (vst1q_p64_x2): Likewise.
8691 (vst1_s64_x3): Likewise.
8692 (vst1_u64_x3): Likewise.
8693 (vst1_f64_x3): Likewise.
8694 (vst1_s8_x3): Likewise.
8695 (vst1_p8_x3): Likewise.
8696 (vst1_s16_x3): Likewise.
8697 (vst1_p16_x3): Likewise.
8698 (vst1_s32_x3): Likewise.
8699 (vst1_u8_x3): Likewise.
8700 (vst1_u16_x3): Likewise.
8701 (vst1_u32_x3): Likewise.
8702 (vst1_f16_x3): Likewise.
8703 (vst1_f32_x3): Likewise.
8704 (vst1_p64_x3): Likewise.
8705 (vst1q_s8_x3): Likewise.
8706 (vst1q_p8_x3): Likewise.
8707 (vst1q_s16_x3): Likewise.
8708 (vst1q_p16_x3): Likewise.
8709 (vst1q_s32_x3): Likewise.
8710 (vst1q_s64_x3): Likewise.
8711 (vst1q_u8_x3): Likewise.
8712 (vst1q_u16_x3): Likewise.
8713 (vst1q_u32_x3): Likewise.
8714 (vst1q_u64_x3): Likewise.
8715 (vst1q_f16_x3): Likewise.
8716 (vst1q_f32_x3): Likewise.
8717 (vst1q_f64_x3): Likewise.
8718 (vst1q_p64_x3): Likewise.
8720 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8722 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8723 user_label_prefix to name.
8725 * tree-core.h: Update comment about the format of NAME string
8726 passed to handler in attribute_spec.
8728 * config/msp430/msp430.md: Remove erroneous subreg expression from
8729 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8730 zero_extend{q,h}isi2.
8732 2018-05-30 Borislav Petkov <bp@suse.de>
8734 * doc/extend.texi: Document some architecture specific
8735 constraints and sort entries.
8737 2018-05-30 Martin Sebor <msebor@redhat.com>
8740 * builtins.c (expand_builtin_stpcpy_1): New function.
8741 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8742 only if the former succeeds.
8744 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8746 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8749 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8751 * doc/invoke.texi (-flinker-output): Document
8753 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8755 * passes.c (ipa_write_summaries): Only modify statements if body
8757 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8758 incrementally linking.
8759 (ipa_passes): Likewise.
8760 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8761 pass down resolution info.
8762 * common.opt (flag_incremental_link): Update info.
8763 * gcc.c (plugin specs): Turn flinker-output=* to
8764 -plugin-opt=-linker-output-known
8765 * toplev.c (compile_file): Also cut compilation when doing incremental
8767 * flag-types. (enum lto_partition_model): Add
8768 LTO_LINKER_OUTPUT_NOLTOREL.
8769 (invoke.texi): Add -flinker-output docs.
8770 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8771 link same way as WPA; do not stream in dead initializers.
8773 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8774 darf2out_register_external_decl): Support incremental link.
8776 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8778 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8780 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8782 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8783 it down to simple_object_copy_lto_debug_sections.
8784 (run_gcc): Determine incremental LTO link time and configure
8785 lto1 into non-wpa mode, disable renaming of debug sections.
8787 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8789 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8790 descriptions of various incorrectly documented functions.
8792 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8795 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8796 address check not strict.
8798 2018-05-30 Richard Biener <rguenther@suse.de>
8800 PR tree-optimization/85964
8801 * tracer.c (better_p): Drop initialized count check, we only
8802 call the function with initialized counts now.
8803 (find_best_successor): Do find a best edge if one
8804 has uninitialized count.
8805 (find_best_predecessor): Likewise. Do BB frequency check only
8806 if count is initialized.
8808 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8810 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8811 (aarch64_ldrstr_offset_compare): New.
8812 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8813 load/store orderings.
8814 (aarch64_gen_adjusted_ldpstp): Likewise.
8816 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8818 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8819 Check for subset of GENERAL_REGS and FP_REGS.
8820 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8823 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8825 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8826 and wi::to_poly_offset. Add the current offset and then check
8827 whether the sum fits, rather than using an unchecked addition of
8828 a checked term. Check for a shwi rather than a uhwi.
8829 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8830 (store_constructor): Use poly_int_tree_p.
8831 (expand_expr_real_1): Likewise.
8832 * function.c (assign_temp): Likewise.
8833 * fold-const.c (const_binop): Use poly_int_tree_p and
8835 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8837 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8838 to_poly_offset for MEM offsets.
8839 * ipa-icf.c (sem_variable::equals): Likewise.
8840 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8841 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8842 wi::to_poly_offset for BIT_FIELD_REF offsets.
8843 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8845 * var-tracking.c (emit_note_insn_var_location): Use
8846 tree_to_poly_uint64.
8848 2018-05-29 Jim Wilson <jimw@sifive.com>
8850 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8852 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8855 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8856 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8858 (sse4_1_round<mode>2): Use nonimmediate_operand
8859 for operand 1 predicate.
8861 2018-05-29 Martin Sebor <msebor@redhat.com>
8862 Richard Biener <rguenther@suse.de>
8865 * calls.c (get_size_range): Call determine_value_range instead
8866 of get_value_range..
8867 * tree-vrp.h (determine_value_range): Declared new function.
8868 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8870 2018-05-29 Richard Biener <rguenther@suse.de>
8872 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8873 sure to use non-pattern stmts for get_earlier_stmt arguments.
8874 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8875 called on pattern stmts.
8876 (get_later_stmt): Likewise.
8878 2018-05-29 Martin Liska <mliska@suse.cz>
8880 PR gcov-profile/85759
8881 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8884 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8886 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8888 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8892 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8893 VEC_PACK_FLOAT_EXPR): New tree codes.
8894 * tree-pretty-print.c (op_code_prio): Handle
8895 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8896 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8897 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8898 * tree-inline.c (estimate_operator_cost): Likewise.
8899 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8900 * fold-const.c (const_binop): Likewise.
8901 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8902 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8903 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8904 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8905 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8906 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8907 * expr.c (expand_expr_real_2): Likewise.
8908 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8909 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8910 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8912 * optabs.c (expand_widen_pattern_expr): For
8913 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8914 sign from result type rather than operand's type.
8915 (expand_binop_directly): For vec_packu_float_optab and
8916 vec_packs_float_optab allow result type to be different from operand's
8918 * optabs-tree.c (optab_for_tree_code): Handle
8919 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8920 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8921 * tree-vect-generic.c (expand_vector_operations_1): Handle
8922 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8923 VEC_PACK_FLOAT_EXPR.
8924 * tree-vect-stmts.c (supportable_widening_operation): Handle
8926 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8927 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8928 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8929 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8930 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8932 (vec_pack<floatprefix>_float_<mode>): New expander.
8933 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8935 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8936 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8937 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8938 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8939 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8941 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8942 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8943 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8944 VEC_PACK_FLOAT_EXPR): Document.
8946 2018-05-29 Richard Biener <rguenther@suse.de>
8948 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8950 (stmt_vec_info_vec): Make pointer.
8951 (init_stmt_vec_info_vec): Remove.
8952 (free_stmt_vec_info_vec): Likewise.
8953 (set_stmt_vec_info_vec): New function.
8954 (free_stmt_vec_infos): Likewise.
8955 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8956 (set_vinfo_for_stmt): Likewise.
8957 (get_earlier_stmt): Likewise.
8958 (get_later_stmt): Likewise.
8959 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8960 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8961 (vec_info::~vec_info): Free stmt_vec_infos.
8962 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8963 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8964 (pass_slp_vectorize::execute): Likewise.
8965 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8966 (free_stmt_vec_info_vec): Likewise.
8967 (set_stmt_vec_info_vec): New function.
8968 (free_stmt_vec_infos): Likewise.
8969 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8970 the global stmt_vec_info_vec.
8971 * tree-parloops.c (gather_scalar_reductions): Use
8972 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8975 2018-05-29 Richard Biener <rguenther@suse.de>
8977 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8979 2018-05-29 Martin Liska <mliska@suse.cz>
8980 David Malcolm <dmalcolm@redhat.com>
8982 * vec.c (test_reverse): New.
8983 (vec_c_tests): Add new test.
8984 * vec.h (vl_ptr>::reverse): New function.
8986 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8988 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8990 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8993 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8995 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8997 2018-05-28 Richard Biener <rguenther@suse.de>
8999 PR tree-optimization/85933
9000 * tree-vect-data-refs.c (vect_record_base_alignments): Only
9001 look at stmts marked as vectorizable.
9003 2018-05-28 Richard Biener <rguenther@suse.de>
9005 PR tree-optimization/85934
9006 * tree-vect-generic.c (expand_vector_operations_1): Hoist
9007 vector boolean check before scalar optimization.
9009 2018-05-28 Jakub Jelinek <jakub@redhat.com>
9011 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
9014 2018-05-28 Mark Wielaard <mark@klomp.org>
9016 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9017 if it is an expression containing a minus sign.
9019 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
9021 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
9023 2018-05-27 Paul Koning <ni1d@arrl.net>
9025 * config/pdp11/pdp11.md (truncsihi2): Remove.
9027 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
9028 Chung-Ju Wu <jasonwucj@gmail.com>
9030 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
9032 (unaligned_store_dw): Ditto.
9033 * config/nds32/nds32-memory-manipulation.c
9034 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
9035 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
9036 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
9037 (emit_setmem_word_loop): Rename to ...
9038 (emit_setmem_doubleword_loop): ... this.
9039 (nds32_gen_dup_4_byte_to_word_value): New function.
9040 (nds32_gen_dup_8_byte_to_double_word_value): New function.
9041 (nds32_expand_setmem_loop): Refine implementation.
9042 (nds32_expand_setmem_loop_v3m): Ditto.
9043 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
9046 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9048 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
9050 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9052 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
9053 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9054 machine->attr_no_prologue_p.
9055 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
9056 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
9057 (nds32_expand_epilogue): Consider attr_naked_p.
9058 (nds32_expand_epilogue_v3pop): Likewise.
9059 (nds32_can_use_return_insn): Likewise.
9060 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
9061 attr_no_prologue_p fields.
9062 * config/nds32/nds32.opt (mret-in-naked-func): New option.
9064 2018-05-27 Jakub Jelinek <jakub@redhat.com>
9067 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
9069 * config/i386/sse.md
9070 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
9072 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
9074 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
9076 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
9078 (*<floatsuffix>floatv2div2sf2): Rename to ...
9079 (*float<floatunssuffix>v2div2sf2): ... this.
9080 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
9081 (float<floatunssuffix>v2div2sf2_mask): ... this.
9082 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
9083 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
9084 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
9086 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
9088 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9090 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9092 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9094 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9096 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
9097 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
9098 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
9099 gen_ufix_truncv8dfv8si2.
9100 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
9101 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
9102 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
9103 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
9104 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
9105 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
9106 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
9107 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
9109 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9113 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9115 2018-05-25 Jim Wilson <jimw@sifive.com>
9117 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9118 * config/riscv/riscv.c (struct machine_function): Add
9119 interrupt_handler_p and attribute_checked_p fields.
9120 (riscv_attribute_table): Add interrupt.
9121 (riscv_interrupt_type_p): New.
9122 (riscv_save_reg_p): Save extra regs for interrupt handler.
9123 (riscv_use_save_libcall): Return false for interrupt handler.
9124 (riscv_first_stack_step): Add forward declaration.
9125 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9126 for interrupt handler with large frame. Use it for saved reg list.
9127 (riscv_expand_prologue): Move flag_stack_usage_info support to
9128 eliminate duplication.
9129 (riscv_expand_epilogue): Generate mret for interrupt handler.
9130 (riscv_epilogue_uses): New.
9131 (riscv_can_use_return_insn): Return false for interrupt handler.
9132 (riscv_function_ok_for_sibcall): Likewise.
9133 (riscv_set_current_function): Add interrupt handler support.
9134 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9135 * config/riscv/riscv.md (UNSPECV_MRET): New.
9137 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9139 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9141 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9143 PR tree-optimization/85712
9144 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9145 this candidate has already been replaced in-situ by a copy.
9147 2018-05-25 Jason Merrill <jason@redhat.com>
9149 PR c++/80485 - inline function non-zero address.
9150 * symtab.c (nonzero_address): Check DECL_COMDAT.
9152 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9155 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9156 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9157 extension of SImode operation. Use const123_operand predicate.
9158 (*saddsi_1): Remove.
9159 (*saddl_se_1): Ditto.
9161 (*ssubl_se_1): Ditto.
9162 * config/alpha/predicates.md (const123_operand): New predicate.
9163 * config/alpha/constraints.md (P): Use IN_RANGE.
9165 2018-05-25 Richard Biener <rguenther@suse.de>
9167 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9169 (ref_maybe_used_by_stmt_p): Likewise.
9170 (stmt_may_clobber_ref_p): Likewise.
9171 (stmt_may_clobber_ref_p_1): Likewise.
9172 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9174 (ref_maybe_used_by_stmt_p): Likewise.
9175 (stmt_may_clobber_ref_p): Likewise.
9176 (stmt_may_clobber_ref_p_1): Likewise.
9177 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9178 the alias oracle to disambiguate DRs with stmts DR analysis
9180 (vect_analyze_data_refs): Do not give up on not analyzable
9181 DRs for BB vectorization. Remove code truncating the dataref
9184 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9187 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9188 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9189 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9191 2018-05-25 Richard Biener <rguenther@suse.de>
9193 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9194 function, combining stmt data ref gathering and fatal analysis
9196 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9197 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9198 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9199 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9200 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9202 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9204 PR tree-optimization/85720
9205 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9206 SCC if all partitions are builtins.
9207 (version_loop_by_alias_check): New parameter. Generate cancelable
9208 runtime alias check if all partitions are builtins.
9209 (distribute_loop): Update call to above function.
9211 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9213 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9214 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9215 (parm_default_def_partition_arg): Ditto.
9216 (set_parm_default_def_partition): Ditto.
9217 (get_parm_default_def_partitions): Ditto and make it static.
9218 (get_undefined_value_partitions): Ditto and make it static.
9219 (remove_ssa_form): Refactor call to init_var_map here.
9220 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9221 computation for loop region.
9222 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9223 (register_default_def): Delete.
9224 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9225 (parm_default_def_partition_arg): Ditto.
9226 (set_parm_default_def_partition): Ditto.
9227 (get_parm_default_def_partitions): Ditto and make it static.
9228 (get_undefined_value_partitions): Ditto and make it static.
9229 (coalesce_with_default, coalesce_with_default): Update comment.
9230 (create_coalesce_list_for_region): New func factored out from
9231 create_outofssa_var_map.
9232 (populate_coalesce_list_for_outofssa): New func factored out from
9233 create_outofssa_var_map and coalesce_ssa_name.
9234 (create_outofssa_var_map): Delete.
9235 (coalesce_ssa_name): Refactor to support live range computation.
9236 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9237 (get_parm_default_def_partitions): Delete.
9238 (get_undefined_value_partitions): Ditto.
9239 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9240 computation for loop region.
9241 (new_tree_live_info, loe_visit_block): Ditto.
9242 (live_worklist, set_var_live_on_entry): Ditto.
9243 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9244 * tree-ssa-live.h (struct _var_map): New fields.
9245 (init_var_map): Change decl.
9246 (region_contains_p): New.
9248 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9250 * tree-ssa-live.h (live_merge_and_clear): Delete.
9252 2018-05-25 Richard Biener <rguenther@suse.de>
9255 * tree-dump.c (dequeue_and_dump): Remove access to removed
9256 operand 2 of a SWITCH_EXPR.
9258 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9260 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9261 multiplication and division.
9262 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9263 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9264 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9265 (cond_udiv_optab, cond_umod_optab): New optabs.
9266 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9267 (IFN_COND_RDIV): New internal functions.
9268 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9269 TRUNC_MOD_EXPR and RDIV_EXPR.
9270 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9271 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9273 (SVE_INT_BINARY): Include mult.
9274 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9275 (optab, sve_int_op): Handle mult.
9276 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9278 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9279 for SVE_INT_BINARY_SD.
9281 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9283 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9284 (optab, sve_int_op): Handle div and udiv.
9285 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9286 for SVE_INT_BINARY_SD.
9287 (*<optab><mode>3): New insn for the same.
9289 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9291 * tree-vect-patterns.c: Include predict.h.
9292 (vect_recog_divmod_pattern): Restrict check for division support
9293 to when optimizing for size.
9295 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9297 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9298 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9299 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9300 (gimple_match_op::set_op): Likewise.
9301 (gimple_resimplify4): Declare.
9302 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9303 (expr::gen_transform): Likewise.
9304 (decision_tree::gen): Generate a simplification routine for 4 operands.
9305 * gimple-match-head.c (gimple_simplify): Add an overload for
9306 4 operands. In the top-level function, handle up to 4 call
9307 arguments and call gimple_resimplify4.
9308 (gimple_resimplify4): New function.
9309 (build_call_internal): Pass a fourth operand.
9310 (maybe_push_to_seq): Likewise.
9311 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9312 Fold VEC_COND_EXPRs of an operation and a default value into
9313 an IFN_COND_* function if possible.
9314 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9316 (SVE_COND_FP_BINARY): Include them.
9317 (optab, sve_fp_op): Handle them.
9318 (SVE_INT_BINARY_REV): New code iterator.
9319 (SVE_COND_FP_BINARY_REV): New int iterator.
9320 (commutative): New int attribute.
9321 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9323 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9325 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9326 (*cond_<optab><mode>): New patterns for reversed operands.
9328 2018-05-25 Richard Biener <rguenther@suse.de>
9330 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9331 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9332 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9333 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9334 * tree-vect-data-refs.c (everywhere): Adjust users.
9335 * tree-vect-loop.c (everywhere): Likewise.
9336 * tree-vect-slp.c (everywhere): Likewise.
9337 * tree-vect-stmts.c (everywhere): Likewise.
9338 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9340 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9342 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9344 (gcc_cv_as_section_exclude): ... this.
9345 Try Solaris as #exclude syntax.
9346 * configure: Regenerate.
9347 * config.in: Regenerate.
9348 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9350 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9351 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9353 * varasm.c (default_elf_asm_named_section): Don't check if
9354 HAVE_GAS_SECTION_EXCLUDE is defined.
9356 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9358 * doc/md.texi: Update the documentation of the cond_* optabs
9359 to mention the new final operand. Fix GET_MODE_NUNITS call.
9360 Describe the scalar case too.
9361 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9362 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9364 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9365 (get_conditional_internal_fn): Update comment.
9366 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9367 accumulator value as a final argument to conditional functions.
9368 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9369 a define_expand and add an "else" operand. Assert for now that
9370 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9371 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9372 (*cond_<optab><mode>): New patterns.
9373 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9374 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9375 (UNSPEC_COND_EOR): Delete.
9376 (optab): Remove associated mappings.
9377 (SVE_INT_BINARY): New code iterator.
9378 (sve_int_op): Remove int attribute and add "minus" to the code
9380 (SVE_COND_INT_OP): Delete.
9381 (SVE_COND_FP_OP): Rename to...
9382 (SVE_COND_FP_BINARY): ...this.
9384 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9386 * optabs.c (can_reuse_operands_p): New function.
9387 (maybe_legitimize_operands): Try to reuse the results for
9390 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9392 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9393 Add {q} suffix to insn mnemonic.
9395 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9397 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9398 (msp430_warn_func_return): New.
9400 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9402 * fold-const.c (tree_nonzero_bits): New function.
9403 * fold-const.h (tree_nonzero_bits): Likewise.
9404 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9405 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9407 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9411 * varasm.c (assemble_alias): Check ifunc_resolver only on
9414 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9417 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9418 when memory input operand is handled.
9420 2018-05-24 Luis Machado <luis.machado@linaro.org>
9422 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9424 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9426 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9428 * match.pd: Delay FMA folds until after vectorization.
9430 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9433 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9434 address check not strict.
9436 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9438 * gimple-match.h (gimple_match_op): New class.
9439 (mprts_hook): Replace parameters with a gimple_match_op *.
9440 (maybe_build_generic_op): Likewise.
9441 (gimple_simplified_result_is_gimple_val): Replace parameters with
9442 a const gimple_match_op *.
9443 (gimple_simplify): Replace code_helper * and tree * parameters with
9444 a gimple_match_op * parameter.
9445 (gimple_resimplify1): Replace code_helper *, tree and tree *
9446 parameters with a gimple_match_op * parameter.
9447 (gimple_resimplify2): Likewise.
9448 (gimple_resimplify3): Likewise.
9449 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9450 parameters with a gimple_match_op * parameter.
9451 * gimple-match-head.c (gimple_simplify): Change prototypes of
9452 auto-generated functions to take a gimple_match_op * instead of
9453 separate code_helper * and tree * parameters. Make the same
9454 change in the top-level overload and update calls to the
9455 gimple_resimplify routines. Update calls to the auto-generated
9456 functions and to maybe_push_res_to_seq in the publicly-facing
9457 operation-specific gimple_simplify overloads.
9458 (gimple_match_op::MAX_NUM_OPS): Define.
9459 (gimple_resimplify1): Replace rcode and ops with a single res_op
9460 parameter. Update call to gimple_simplify.
9461 (gimple_resimplify2): Likewise.
9462 (gimple_resimplify3): Likewise.
9463 (mprts_hook): Replace parameters with a gimple_match_op *.
9464 (maybe_build_generic_op): Likewise.
9465 (build_call_internal): Replace type, nargs and ops with
9466 a gimple_match_op *.
9467 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9468 with a single gimple_match_op *. Update calls to mprts_hook,
9469 build_call_internal and gimple_simplified_result_is_gimple_val.
9470 Factor out code that is common to the tree_code and combined_fn cases.
9471 * genmatch.c (expr::gen_transform): Replace tem_code and
9472 tem_ops with a gimple_match_op called tem_op. Update calls
9473 to the gimple_resimplify functions and maybe_push_res_to_seq.
9474 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9475 res_ops. Update call to the gimple_resimplify functions.
9476 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9477 (decision_tree::gen): Make the functions take a gimple_match_op *
9478 called res_op instead of separate res_code and res_ops parameters.
9479 Update call accordingly.
9480 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9481 and ops with a single res_op parameter. Update calls to
9482 maybe_build_generic_op and maybe_push_res_to_seq.
9483 (fold_stmt_1): Update calls to gimple_simplify and
9484 replace_stmt_with_simplification.
9485 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9486 and gimple_simplified_result_is_gimple_val.
9487 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9489 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9490 with a gimple_match_op *.
9491 (vn_nary_build_or_lookup): Likewise. Update call to
9492 vn_nary_build_or_lookup_1.
9493 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9494 gimple_match_op *. Update calls to the gimple_resimplify routines
9495 and to gimple_simplified_result_is_gimple_val.
9496 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9497 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9498 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9499 (visit_nary_op): Likewise.
9500 (visit_reference_op_load): Likewise.
9502 2018-05-23 Luis Machado <luis.machado@linaro.org>
9504 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9505 modifier for printing the step amount.
9507 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9510 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9513 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9515 * doc/sourcebuild.texi (Endianness): New subsubsection.
9517 2018-05-23 Luis Machado <luis.machado@linaro.org>
9519 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9520 <prefetch_dynamic_strides>: New const bool field.
9521 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9522 prefetch_dynamic_strides.
9523 (exynosm1_prefetch_tune): Likewise.
9524 (thunderxt88_prefetch_tune): Likewise.
9525 (thunderx_prefetch_tune): Likewise.
9526 (thunderx2t99_prefetch_tune): Likewise.
9527 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9529 (aarch64_override_options_internal): Update to set
9530 PARAM_PREFETCH_DYNAMIC_STRIDES.
9531 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9532 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9533 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9534 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9535 prefetch-dynamic-strides setting.
9537 2018-05-23 Luis Machado <luis.machado@linaro.org>
9539 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9540 <minimum_stride>: New const int field.
9541 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9542 minimum_stride field defaulting to -1.
9543 (exynosm1_prefetch_tune): Likewise.
9544 (thunderxt88_prefetch_tune): Likewise.
9545 (thunderx_prefetch_tune): Likewise.
9546 (thunderx2t99_prefetch_tune): Likewise.
9547 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9548 <default_opt_level>: Set to 3.
9549 (aarch64_override_options_internal): Update to set
9550 PARAM_PREFETCH_MINIMUM_STRIDE.
9551 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9552 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9553 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9554 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9555 stride is constant and is below the minimum stride threshold.
9557 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9559 * config/arm/arm-cpus.in (mode26): Delete.
9560 (armv4): Delete mode26 reference.
9561 * config/arm/arm.c (arm_configure_build_target): Delete use of
9564 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9566 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9568 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9569 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9571 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9572 (floatunsdidf2): Ditto.
9574 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9576 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9577 (fixuns_trunc<mode>si2_avx512f): Ditto.
9578 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9579 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9580 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9582 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9584 PR rtl-optimization/79985
9585 * df-scan.c (df_insn_refs_collect): Remove special case for
9586 global registers and asm statements.
9588 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9590 * extend.texi (Global Register Variables): Rewrite the bullet list.
9591 Note that the register is available for allocation. Note that access
9592 via inline asm must use constraints. Add note about async-signal
9593 handlers. Remove paragraph about automagic register selection.
9595 2018-05-23 Richard Biener <rguenther@suse.de>
9597 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9598 of fixed offset from memset VN.
9600 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9602 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9604 (alloc_cand_and_find_basis): Initialize first_interp field.
9605 (slsr_process_mul): Modify first_interp field.
9606 (slsr_process_add): Likewise.
9607 (slsr_process_cast): Modify first_interp field for each new
9609 (slsr_process_copy): Likewise.
9610 (dump_candidate): Dump first_interp field.
9611 (replace_mult_candidate): Process all interpretations, not just
9613 (replace_rhs_if_not_dup): Likewise.
9614 (replace_one_candidate): Likewise.
9616 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9618 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9620 (aarch64_needs_frame_chain): New function.
9621 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9623 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9626 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9627 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9628 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9629 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9630 as true for strict-align.
9631 (aarch64_can_inline_p): Perform checks even when callee has no
9632 attributes to check for strict alignment.
9633 * doc/extend.texi (AArch64 Function Attributes): Document
9635 * doc/invoke.texi: (AArch64 Options): Likewise.
9637 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9639 PR tree-optimization/85853
9640 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9641 the handling of the root of the node to...
9642 (vect_slp_analyze_node_operations_1): ...this new function,
9643 and run the whole thing with the child nodes' def types
9644 set according to their SLP node's def type.
9646 2018-05-23 Richard Biener <rguenther@suse.de>
9649 * tree-data-ref.c (create_runtime_alias_checks): Defer
9650 and ignore overflow warnings.
9652 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9654 PR tree-optimization/85822
9655 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9658 2018-05-23 Richard Biener <rguenther@suse.de>
9660 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9661 memset constants via native_interpret_expr.
9663 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9666 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9668 (cgraph_node::create_alias): Likewise.
9669 (cgraph_node::get_availability): Check ifunc_resolver instead
9670 of looking up ifunc attribute.
9671 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9672 * varasm.c (do_assemble_alias): Likewise.
9673 (assemble_alias): Likewise.
9674 (default_binds_local_p_3): Likewise.
9675 * cgraph.h (cgraph_node): Add ifunc_resolver.
9676 (cgraph_node::only_called_directly_or_aliased_p): Return false
9678 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9680 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9681 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9682 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9683 instead of looking up ifunc attribute.
9685 2018-05-22 Luis Machado <luis.machado@linaro.org>
9687 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9689 2018-05-22 Martin Sebor <msebor@redhat.com>
9692 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9693 only when expasion succeeds.
9694 (expand_builtin_strcmp): Same.
9695 (expand_builtin_strncmp): Same.
9697 2018-05-22 Martin Sebor <msebor@redhat.com>
9699 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9701 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9702 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9704 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9705 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9707 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9708 Add check for legitimate_address.
9709 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9710 (aarch64_swap_ldrstr_operands): New.
9711 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9714 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9715 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9717 * config/aarch64/aarch64.md: New patterns to generate stp
9719 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9720 single words and double words.
9721 (load_pair_sw, load_pair_dw): Likewise.
9722 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9724 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9726 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9727 for different mode ldpstp and add peephole for merged zero stores.
9729 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9731 (aarch64_gen_store_pair): Rename calls to match new patterns.
9732 (aarch64_gen_load_pair): Rename calls to match new patterns.
9733 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9734 (load_pair<DREG:mode><DREG2:mode>): ... This.
9735 (store_pair<mode>): Rename to...
9736 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9737 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9739 (V_INT_EQUIV): Handle SImode.
9740 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9743 2018-05-22 Martin Sebor <msebor@redhat.com>
9746 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9747 or ajust the presumed bound on an operation to avoid unnecessary
9750 2018-05-22 Martin Sebor <msebor@redhat.com>
9752 PR tree-optimization/85826
9753 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9754 assuming that a DECL necesarily has a constant size.
9756 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9759 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9761 2018-05-22 Richard Biener <rguenther@suse.de>
9763 PR tree-optimization/85834
9764 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9765 non-constant and non-zero memset arguments.
9767 2018-05-22 Martin Liska <mliska@suse.cz>
9770 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9772 2018-05-22 Richard Biener <rguenther@suse.de>
9774 PR tree-optimization/85863
9775 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9776 comparisons when vectype is specified.
9777 (vectorizable_condition): Do not specify vectype for
9778 vect_is_simple_cond when SLP vectorizing.
9780 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9783 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9784 define __ibm128 as long double.
9785 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9786 as a distinct type when IEEE 128-bit support is enabled.
9787 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9788 128-bit types to use the correct functions.
9789 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9790 convert between 128-bit floating point types that have different
9791 modes but the same representation, instead of using gen_lowpart to
9793 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9795 (IFKF_reg): New attributes to give the register constraints for
9797 (extend<mode>tf2_internal): New insns to mark an explicit
9798 conversion between 128-bit floating point types that have a
9799 different mode but share the same representation.
9801 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9803 PR tree-optimization/85814
9804 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9805 a null return from get_strinfo when unsharing the next
9806 strinfo in the chain.
9808 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9811 * varasm.c (weak_finish): Clean up weak_decls.
9813 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9815 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9816 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9817 UNSPEC_UADALP values.
9818 * config/aarch64/iterators.md (ABAL): New int iterator.
9821 (sur): Add mappings for the above.
9822 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9824 (aarch64_<sur>abal<mode>_4): Likewise.
9825 (aarch64_<sur>adalp<mode>_3): Likewise.
9826 (<sur>sadv16qi): New define_expand.
9828 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9830 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9831 (*movdf_internal): Ditto.
9832 (*rcpsf2_sse): Ditto.
9833 (*rsqrtsf2_sse): Ditto.
9834 (*sqrt<mode>2_sse): Ditto.
9836 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9838 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9840 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9841 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9842 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9843 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9845 * config/aarch64/arm_neon.h: Likewise.
9846 * config/aarch64/iterators.md (VQ_I): New.
9848 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9850 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9852 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9855 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9857 * config/nds32/constraints.md (S): New constraint.
9858 * config/nds32/nds32.md (call_internal): Use constraint S.
9859 (call_value_internal): Likewise.
9860 (sibcall_internal): Likewise.
9861 (sibcall_value_internal): Likewise.
9863 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9864 Chung-Ju Wu <jasonwucj@gmail.com>
9866 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9869 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9870 Chung-Ju Wu <jasonwucj@gmail.com>
9872 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9873 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9874 (nds32_rtx_costs_impl): Simplify.
9875 (nds32_address_cost_impl): Simplify.
9876 (nds32_init_rtx_costs): New function.
9877 (nds32_rtx_costs_speed_prefer): Likewise.
9878 (nds32_rtx_costs_size_prefer): Likewise.
9879 (nds32_address_cost_speed_prefer): Likewise.
9880 (nds32_address_cost_speed_fwprop): Likewise.
9881 (nds32_address_cost_size_prefer): Likewise.
9882 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9883 * config/nds32/nds32.c (nds32_option_override): Use
9884 nds32_init_rtx_costs function.
9886 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9888 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9889 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9890 (TARGET_PIPELINE_N8): Likewise.
9891 (TARGET_PIPELINE_N10): Likewise.
9892 (TARGET_PIPELINE_N13): Likewise.
9893 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9895 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9897 * config/nds32/nds32-fpu.md: Update copyright year.
9899 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9901 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9903 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9905 * config/nds32/nds32.c
9906 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9907 * config/nds32/nds32.opt (minline-asm-r15): New option.
9909 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9911 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9913 * config/nds32/nds32.md (abssi2): New pattern.
9915 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9917 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9918 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9919 Merge insn pattern from sse_cvtsi2ss<round_name> and
9920 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9921 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9922 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9923 using SWI48 mode iterator.
9924 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9925 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9926 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9927 pattern from sse_cvttss2si<round_saeonly_name>
9928 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9929 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9930 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9931 using SWI48 mode iterator.
9932 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9933 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9934 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9935 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9936 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9937 using SWI48 mode iterator.
9938 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9939 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9940 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9941 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9942 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9943 SWI48 mode iterator.
9944 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9945 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9946 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9947 pattern from sse_cvttsd2si<round_saeonly_name>
9948 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9950 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9952 * config/nds32/nds32-md-auxiliary.c
9953 (nds32_valid_smw_lwm_base_p): Refine.
9954 (nds32_output_smw_single_word): Refine.
9955 (nds32_output_smw_double_word): New.
9956 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9958 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9960 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9961 (nds32_output_stack_pop): Refine.
9962 (nds32_expand_unaligned_load): Refine.
9963 (nds32_expand_unaligned_store): Refine.
9965 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9966 Chung-Ju Wu <jasonwucj@gmail.com>
9968 * config/nds32/constants.md: Add TP_REGNUM constant.
9969 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9970 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9972 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9973 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9974 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9975 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9976 and PIC code generation.
9977 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9979 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9981 * config/nds32/nds32.md: Support TLS and PIC.
9982 * config/nds32/nds32.c: Support TLS and PIC.
9983 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9984 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9987 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9989 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9990 mode with E_ prefix.
9992 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9993 Chung-Ju Wu <jasonwucj@gmail.com>
9995 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9996 * config/nds32/nds32-md-auxiliary.c
9997 (symbolic_reference_mentioned_p): New.
9998 (nds32_legitimize_ict_address): New.
9999 (nds32_expand_ict_move): New.
10000 (nds32_indirect_call_referenced_p): New.
10001 (nds32_symbol_binds_local_p): Delete.
10002 (nds32_long_call_p): Modify.
10003 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
10004 * config/nds32/nds32-protos.h
10005 (symbolic_reference_mentioned_p): Declare.
10006 (nds32_legitimize_ict_address): Declare.
10007 (nds32_expand_ict_move): Declare.
10008 (nds32_indirect_call_referenced_p): Declare.
10009 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
10010 (nds32_relax_group): Use nds32_ict_const_p as condition.
10011 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
10012 (nds32_asm_file_start): Output ict_model directive in asm code.
10013 (nds32_legitimate_address_p): Consider indirect call.
10014 (nds32_print_operand): Consider indirect call.
10015 (nds32_print_operand_address): Consider indirect call.
10016 (nds32_insert_attributes): Handle "indirect_call" attribute.
10017 (TARGET_LEGITIMATE_ADDRESS_P): Define.
10018 (TARGET_LEGITIMATE_CONSTANT_P): Define.
10019 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
10020 (TARGET_DELEGITIMIZE_ADDRESS): Define.
10021 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
10022 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
10023 (TARGET_ICT_MODEL_SMALL): Define.
10024 (TARGET_ICT_MODEL_LARGE): Define.
10025 * config/nds32/nds32.md (movsi): Consider ict model.
10026 (call, call_value): Consider ict model.
10027 (sibcall, sibcall_value): Consider ict model.
10028 * config/nds32/nds32.opt (mict-model): New option.
10029 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
10032 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
10033 Monk Chiang <sh.chiang04@gmail.com>
10034 Jim Wilson <jimw@sifive.com>
10036 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
10037 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
10038 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
10039 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
10040 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
10041 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
10042 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
10043 compute save_libcall_adjustment properly.
10044 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
10045 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
10046 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
10047 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
10048 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
10049 (ABI_SPEC): Handle mabi=ilp32e.
10050 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
10051 (RVE): Add RVE mask.
10052 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
10053 <-march>: Add rv32e as an example.
10055 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
10058 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
10059 (intra_create_variable_infos): Handle C++ constructors.
10061 2018-05-18 Martin Liska <mliska@suse.cz>
10063 * passes.def: Remove a redundant pass.
10065 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
10068 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10070 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10072 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10074 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
10076 (armv2, armv2a, armv3, armv3m): Delete architectures.
10077 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
10078 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10079 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
10081 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
10082 (*mulsidi3adddi): Likewise.
10083 (mulsidi3): Likewise.
10084 (*mulsidi3_nov6): Likewise.
10085 (umulsidi3): Likewise.
10086 (umulsidi3_nov6): Likewise.
10087 (umaddsidi4): Likewise.
10088 (*umulsidi3adddi): Likewise.
10089 (smulsi3_highpart): Likewise.
10090 (*smulsi3_highpart_nov6): Likewise.
10091 (umulsi3_highpart): Likewise.
10092 (*umulsi3_highpart_nov6): Likewise.
10093 * config/arm/arm.h (arm_arch3m): Delete.
10094 * config/arm/arm.c (arm_arch3m): Delete.
10095 (arm_option_override_internal): Update armv3-related comment.
10096 (arm_configure_build_target): Delete use of isa_bit_mode32.
10097 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
10098 (arm_rtx_costs_internal): Delete check of arm_arch3m.
10099 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
10100 (mulsa3): Likewise.
10101 (mulusa3): Likewise.
10102 * config/arm/arm-protos.h (arm_arch3m): Delete.
10103 * config/arm/arm-tables.opt: Regenerate.
10104 * config/arm/arm-tune.md: Likewise.
10105 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
10106 deleted architectures.
10108 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10110 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
10111 (armv5t, armv5te): New features.
10112 (ARMv5, ARMv5e): Delete fgroups.
10113 (ARMv5t, ARMv5te): Adjust for above changes.
10114 (ARMv6m): Likewise.
10115 (armv5, armv5e): Delete arches.
10116 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10118 (*call_reg_arm): Likewise.
10119 (*call_value_reg_armv5): Likewise.
10120 (*call_value_reg_arm): Likewise.
10121 (*call_symbol): Likewise.
10122 (*call_value_symbol): Likewise.
10123 (*sibcall_insn): Likewise.
10124 (*sibcall_value_insn): Likewise.
10125 (clzsi2): Likewise.
10126 (prefetch): Likewise.
10127 (define_split and define_peephole2 dependent on arm_arch5):
10129 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10131 (TARGET_ARM_QBIT): Likewise.
10132 (TARGET_DSP_MULTIPLY): Likewise.
10133 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10134 (arm_arch5, arm_arch5e): Delete.
10135 (arm_arch5t, arm_arch5te): Declare.
10136 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10137 (arm_arch5t): Declare.
10138 (arm_option_reconfigure_globals): Update for the above.
10139 (arm_options_perform_arch_sanity_checks): Update comment, replace
10140 use of arm_arch5 with arm_arch5t.
10141 (use_return_insn): Likewise.
10142 (arm_emit_call_insn): Likewise.
10143 (output_return_instruction): Likewise.
10144 (arm_final_prescan_insn): Likewise.
10145 (arm_coproc_builtin_available): Likewise.
10146 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10147 arm_arch5e with arm_arch5t and arm_arch5te.
10148 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10149 (arm_arch5t, arm_arch5te): Declare.
10150 * config/arm/arm-tables.opt: Regenerate.
10151 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10152 * config/arm/t-multilib: Likewise.
10153 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10154 instead of arm_arch5.
10155 (*call_reg_thumb1): Likewise.
10156 (*call_value_reg_thumb1_v5): Likewise.
10157 (*call_value_reg_thumb1): Likewise.
10158 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10160 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10162 2018-05-18 Martin Liska <mliska@suse.cz>
10164 PR gcov-profile/84846
10165 * doc/gcov.texi: Document -t option of gcov tool.
10167 2018-05-18 Martin Liska <mliska@suse.cz>
10169 PR gcov-profile/84846
10170 * gcov.c (print_usage): Add new -t option.
10171 (process_args): Handle the option.
10172 (generate_results): Use stdout as output when requested by
10175 2018-05-18 Martin Liska <mliska@suse.cz>
10177 PR gcov-profile/84846
10178 * coverage.c (coverage_init): Write PWD to .gcno file.
10179 * doc/gcov.texi: Document how working directory is printed.
10180 * gcov-dump.c (dump_gcov_file): Print PWD.
10181 * gcov.c (output_intermediate_file): Likewise.
10182 (read_graph_file): Read PWD string.
10183 (output_lines): Print PWD.
10185 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10187 PR middle-end/85817
10188 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10189 for retval and return false if all args to phi are zero.
10191 2018-05-18 Richard Biener <rguenther@suse.de>
10193 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10195 (evrp_dom_walker::before_dom_children): Call it.
10197 2018-05-18 Richard Biener <rguenther@suse.de>
10199 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10200 results when processing array refs with variable index.
10202 2018-05-18 Toon Moene <toon@moene.org>
10204 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10205 directly after that of -floop-interchange. Indicate that both
10206 options are enabled by default when specifying -O3.
10208 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10210 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10211 iterator. Delete separate integer-mode vec_set<mode> expander.
10212 (aarch64_simd_vec_setv2di): Delete.
10213 (vec_setv2di): Delete.
10214 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10215 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10216 the "w, r" alternative.
10218 2018-05-18 Martin Liska <mliska@suse.cz>
10220 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10221 * tree-pass.h (make_pass_lower_switch_O0): New function.
10222 * tree-switch-conversion.c (node_has_low_bound): Remove.
10223 (node_has_high_bound): Likewise.
10224 (node_is_bounded): Likewise.
10225 (class pass_lower_switch): Make it a template type and create
10227 (pass_lower_switch::execute): Add template argument.
10228 (make_pass_lower_switch): New function.
10229 (make_pass_lower_switch_O0): New function.
10230 (do_jump_if_equal): Remove.
10231 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10232 all the hard work to tree optimization passes.
10234 2018-05-18 Martin Liska <mliska@suse.cz>
10236 * dbgcnt.c (limit_low): Renamed from limit.
10237 (limit_high): New variable.
10238 (dbg_cnt_is_enabled): Check for upper limit.
10239 (dbg_cnt): Adjust dumping.
10240 (dbg_cnt_set_limit_by_index): Add new argument for high
10242 (dbg_cnt_set_limit_by_name): Likewise.
10243 (dbg_cnt_process_single_pair): Parse new format.
10244 (dbg_cnt_process_opt): Use strtok.
10245 (dbg_cnt_list_all_counters): Remove 'value' and add
10247 * doc/invoke.texi: Document changes.
10249 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10251 * doc/sourcebuild.texi (scalar_all_fma): Document.
10252 * tree.def (FMA_EXPR): Delete.
10253 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10254 * internal-fn.c (ternary_direct): New macro.
10255 (expand_ternary_optab_fn): Likewise.
10256 (direct_ternary_optab_supported_p): Likewise.
10257 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10258 * builtins.c (fold_builtin_fma): Delete.
10259 (fold_builtin_3): Don't call it.
10260 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10261 * expr.c (expand_expr_real_2): Likewise.
10262 * fold-const.c (operand_equal_p): Likewise.
10263 (fold_ternary_loc): Likewise.
10264 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10265 * gimple.c (DEFTREECODE): Likewise.
10266 * gimplify.c (gimplify_expr): Likewise.
10267 * optabs-tree.c (optab_for_tree_code): Likewise.
10268 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10269 * tree-eh.c (operation_could_trap_p): Likewise.
10270 (stmt_could_throw_1_p): Likewise.
10271 * tree-inline.c (estimate_operator_cost): Likewise.
10272 * tree-pretty-print.c (dump_generic_node): Likewise.
10273 (op_code_prio): Likewise.
10274 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10275 * tree-ssa-operands.c (get_expr_operands): Likewise.
10276 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10277 * fold-const-call.h (fold_fma): Delete.
10278 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10279 CFN_FNMA and CFN_FNMS.
10280 (fold_fma): Delete.
10281 * genmatch.c (combined_fn): New enum.
10282 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10283 (commutative_op): New function.
10284 (commutate): Use it. Handle more than 2 operands.
10285 (dt_operand::gen_gimple_expr): Use commutative_op.
10286 (parser::parse_expr): Allow :c to be used with non-binary
10287 operators if the commutative operand is known.
10288 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10289 CFN_FMS, CFN_FNMA and CFN_FNMS.
10290 (backprop::process_assign_use): Remove FMA_EXPR handling.
10291 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10292 (gen_hsa_fma): New function.
10293 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10294 IFN_FNMA and IFN_FNMS.
10295 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10296 * gimple-fold.h (follow_all_ssa_edges): Declare.
10297 * gimple-fold.c (follow_all_ssa_edges): New function.
10298 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10299 gimple_build interface and use follow_all_ssa_edges to fold the result.
10300 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10301 instead of checking for optabs directly.
10302 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10303 rather than FMA_EXPRs.
10304 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10305 call to IFN_FMA instead of an FMA_EXPR.
10307 2018-05-17 Jim Wilson <jimw@sifive.com>
10309 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10310 sign extended promoted subreg, and the range does not have the sign bit
10311 set, then do a sign extend.
10313 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10314 test, check for sign extended subreg and/or constant operands, and
10315 do a sign extend in that case.
10317 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10319 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10320 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10322 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10323 Change logics_shift_reg to logics_shift_imm.
10324 (thunderx2t99_fp_loadpair_basic): Delete.
10325 (thunderx2t99_fp_storepair_basic): Delete.
10326 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10327 (thunderx2t99_asimd_polynomial): Delete.
10328 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10329 and neon_fp_mul_d_scalar_q.
10330 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10331 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10332 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10333 (thunderx2t99_asimd_lut): Add missing tbl types.
10334 (thunderx2t99_asimd_ext): Delete.
10335 (thunderx2t99_asimd_load1_1_mult): Delete.
10336 (thunderx2t99_asimd_load1_2_mult): Delete.
10337 (thunderx2t99_asimd_load1_ldp): New.
10338 (thunderx2t99_asimd_load1): New.
10339 (thunderx2t99_asimd_load2): Add missing *load2* types.
10340 (thunderx2t99_asimd_load3): New.
10341 (thunderx2t99_asimd_load4): New.
10342 (thunderx2t99_asimd_store1_1_mult): Delete.
10343 (thunderx2t99_asimd_store1_2_mult): Delete.
10344 (thunderx2t99_asimd_store2_mult): Delete.
10345 (thunderx2t99_asimd_store2_onelane): Delete.
10346 (thunderx2t99_asimd_store_stp): New.
10347 (thunderx2t99_asimd_store1): New.
10348 (thunderx2t99_asimd_store2): New.
10349 (thunderx2t99_asimd_store3): New.
10350 (thunderx2t99_asimd_store4): New.
10352 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10354 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10355 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10357 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10358 Segher Boessenkool <segher@kernel.crashing.org>
10361 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10364 2018-05-17 Richard Biener <rguenther@suse.de>
10366 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10367 for pruning loop and prune defs feeding only already visited PHIs.
10369 2018-05-17 Richard Biener <rguenther@suse.de>
10371 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10373 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10374 Richard Biener <rguenther@suse.de>
10376 PR tree-optimization/85793
10377 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10378 for VMAT_ELEMENTWISE.
10380 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10382 * internal-fn.h (lookup_internal_fn): Declare
10383 * internal-fn.c (lookup_internal_fn): New function.
10384 * gimple.c (gimple_build_call_from_tree): Handle calls to
10385 internal functions.
10386 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10387 internal function names.
10388 * tree-pretty-print.c (dump_generic_node): Likewise.
10389 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10391 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10393 * gimple-fold.h (gimple_build): Make the function forms take
10394 combined_fn rather than built_in_function.
10395 (gimple_simplify): Likewise.
10396 * gimple-match-head.c (gimple_simplify): Likewise.
10397 * gimple-fold.c (gimple_build): Likewise.
10398 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10399 rather than gimple_build_call_internal.
10400 (get_initial_defs_for_reduction): Likewise.
10401 (vect_create_epilog_for_reduction): Likewise.
10402 (vectorizable_live_operation): Likewise.
10404 2018-05-17 Martin Liska <mliska@suse.cz>
10406 * gimple-ssa-sprintf.c (format_directive): Do not use
10407 space in between 'G_' and '('.
10409 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10412 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10413 even if the mask is not all ones.
10416 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10418 (ix86_gimple_fold_builtin): Likewise.
10421 * config/i386/i386.c: Include tree-vector-builder.h.
10422 (ix86_vector_shift_count): New function.
10423 (ix86_fold_builtin): Fold shift builtins by scalar count.
10424 (ix86_gimple_fold_builtin): Likewise.
10426 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10427 _mm512_setzero): New intrinsics.
10429 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10430 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10432 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10433 code generation for cases where splatting a value is not useful.
10434 * simplify-rtx.c (simplify_ternary_operation): Simplify
10435 vec_merge across a vec_duplicate and a paradoxical subreg forming
10436 a vector mode to a vec_concat.
10438 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10440 * config.gcc: Support "goldmont-plus".
10441 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10443 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10444 PROCESSOR_GOLDMONT_PLUS.
10445 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10446 (processor_target_table): Add "goldmont-plus".
10447 (PTA_GOLDMONT_PLUS): Define.
10448 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10449 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10450 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10451 (fold_builtin_cpu): Add "goldmont-plus".
10452 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10453 (ix86_option_override_internal): Add "goldmont-plus".
10454 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10455 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10456 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10457 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10459 2018-05-17 Richard Biener <rguenther@suse.de>
10461 PR tree-optimization/85757
10462 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10463 remove defs that only feed that PHI from further processing.
10465 2018-05-16 Jim Wilson <jimw@sifive.com>
10467 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10469 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10471 2018-05-16 Mark Wielaard <mark@klomp.org>
10473 * dwarf2out.c (count_index_strings): New function.
10474 (output_indirect_strings): Call count_index_strings and generate
10475 header for dwarf_version >= 5.
10477 2018-05-16 Mark Wielaard <mark@klomp.org>
10479 * dwarf2out.c (dwarf_FORM): New function.
10480 (set_indirect_string): Use dwarf_FORM.
10481 (reset_indirect_string): Likewise.
10482 (size_of_die): Likewise.
10483 (value_format): Likewise.
10484 (output_die): Likewise.
10485 (add_skeleton_AT_string): Likewise.
10486 (output_macinfo_op): Likewise.
10487 (index_string): Likewise.
10488 (output_index_string_offset): Likewise.
10489 (output_index_string): Likewise.
10490 (count_index_strings): Likewise.
10492 2018-05-16 Carl Love <cel@us.ibm.com>
10494 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10495 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10497 2018-05-16 Martin Jambor <mjambor@suse.cz>
10499 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10500 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10502 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10504 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10505 (fnma<mode>4): Likewise.
10506 (fms<mode>4): Likewise.
10507 (fnms<mode>4): Likewise.
10508 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10509 (aarch64_fnma<mode>4): Likewise.
10510 (aarch64_fms<mode>4): Likewise.
10511 (aarch64_fnms<mode>4): Likewise.
10512 (aarch64_fnmadd<mode>4): Likewise.
10514 2018-05-16 Jason Merrill <jason@redhat.com>
10516 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10518 2018-05-16 Richard Biener <rguenther@suse.de>
10520 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10521 (dump_stmt_cost): Declare.
10522 (add_stmt_cost): Dump cost we add.
10523 (add_stmt_costs): New function.
10524 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10525 No longer exported.
10526 (vect_analyze_stmt): Adjust prototype.
10527 (vectorizable_condition): Likewise.
10528 (vectorizable_live_operation): Likewise.
10529 (vectorizable_reduction): Likewise.
10530 (vectorizable_induction): Likewise.
10531 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10532 cost vector to pass to vectorizable_ and record afterwards.
10533 (vect_model_reduction_cost): Take cost vector argument and adjust.
10534 (vect_model_induction_cost): Likewise.
10535 (vectorizable_reduction): Likewise.
10536 (vectorizable_induction): Likewise.
10537 (vectorizable_live_operation): Likewise.
10538 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10539 SLP_TREE_NUMBER_OF_VEC_STMTS.
10540 (vect_analyze_slp_cost_1): Remove.
10541 (vect_analyze_slp_cost): Likewise.
10542 (vect_slp_analyze_node_operations): Take visited args and
10543 a target cost vector. Avoid processing already visited stmt sets.
10544 (vect_slp_analyze_operations): Use a local cost vector to gather
10545 costs and register those of non-discarded instances.
10546 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10547 (vect_schedule_slp_instance): Remove copying of
10548 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10550 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10551 adding cost. Record cost entry location.
10552 (vect_prologue_cost_for_slp_op): Function to compute cost of
10553 a constant or invariant generated for SLP vect in the prologue,
10554 split out from vect_analyze_slp_cost_1.
10555 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10556 (vect_model_promotion_demotion_cost): Likewise.
10557 (vect_model_store_cost): Likewise, make static.
10558 (vect_model_load_cost): Likewise.
10559 (vectorizable_bswap): Add cost vector arg and adjust.
10560 (vectorizable_call): Likewise.
10561 (vectorizable_simd_clone_call): Likewise.
10562 (vectorizable_conversion): Likewise.
10563 (vectorizable_assignment): Likewise.
10564 (vectorizable_shift): Likewise.
10565 (vectorizable_operation): Likewise.
10566 (vectorizable_store): Likewise.
10567 (vectorizable_load): Likewise.
10568 (vectorizable_condition): Likewise.
10569 (vectorizable_comparison): Likewise.
10570 (can_vectorize_live_stmts): Likewise.
10571 (vect_analyze_stmt): Likewise.
10572 (vect_transform_stmt): Adjust calls to vectorizable_*.
10573 * tree-vectorizer.c: Include gimple-pretty-print.h.
10574 (dump_stmt_cost): New function.
10576 2018-05-16 Richard Biener <rguenther@suse.de>
10578 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10579 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10580 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10581 (check_name): New callback.
10582 (dse_classify_store): Track cycles via a visited bitmap of PHI
10583 defs and simplify handling of in-loop and across loop dead stores
10584 and properly fail for loop-variant refs. Handle byte-tracking with
10585 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10588 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10590 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10591 (vect_get_mask_type_for_stmt): Likewise.
10592 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10594 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10595 to determine the statement's vector type and the vector type that
10596 should be used for calculating nunits. Deal with cases in which
10597 the type has to be deferred.
10598 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10599 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10600 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10601 (vect_determine_vf_for_stmt): New functions, split out from...
10602 (vect_determine_vectorization_factor): ...here.
10603 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10604 (vect_get_mask_type_for_stmt): New functions, split out from
10605 vect_determine_vectorization_factor.
10607 2018-05-16 Richard Biener <rguenther@suse.de>
10609 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10610 verify the [VEC_]COND_EXPR embedded comparison.
10612 2018-05-15 Martin Sebor <msebor@redhat.com>
10614 PR tree-optimization/85753
10615 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10616 RECORD_TYPE in addition to ARRAY_TYPE.
10618 2018-05-15 Martin Sebor <msebor@redhat.com>
10620 PR middle-end/85643
10621 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10623 2018-05-15 Richard Biener <rguenther@suse.de>
10625 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10626 add by_clobber_p one. Change algorithm to collect all defs
10627 representing uses we need to walk and try reducing them to
10628 a single one before failing.
10629 (dse_dom_walker::dse_optimize_stmt): Adjust.
10631 2018-05-13 Mark Wielaard <mark@klomp.org>
10633 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10634 (size_of_loc_descr): Likewise.
10635 (output_loc_operands): Likewise.
10636 (output_loc_operands_raw): Likewise.
10637 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10638 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10639 (hash_loc_operands): Likewise.
10640 (compare_loc_operands): Likewise.
10642 2018-05-14 Mark Wielaard <mark@klomp.org>
10644 * dwarf2out.c (count_index_addrs): New function.
10645 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10647 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10649 PR tree-optimization/83648
10650 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10651 return value as malloc candidate.
10653 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10656 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10657 param as true in call to suggest_attribute.
10659 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10661 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10662 -mreadonly-in-sdata.
10664 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10666 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10668 (aarch64_crypto_aesd_fused): Likewise.
10670 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10672 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10673 (movsi_aarch64): Likewise.
10674 (load_pairsi): Likewise.
10675 (load_pairdi): Likewise.
10676 (store_pairsi): Likewise.
10677 (store_pairdi): Likewise.
10678 (load_pairsf): Likewise.
10679 (load_pairdf): Likewise.
10680 (store_pairsf): Likewise.
10681 (store_pairdf): Likewise.
10682 (zero_extend): Likewise.
10683 (trunc): Swap alternatives.
10684 (fcvt_target): Add '?' to prefer w over r.
10686 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10689 * config/i386/i386.md: Disallow non-commutative arithmetics in
10690 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10691 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10692 in the peephole2 before it.
10694 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10696 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10697 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10698 (ix86_handle_option): Handle -mcldemote.
10699 * config.gcc: New header.
10700 * config/i386/cldemoteintrin.h: New file.
10701 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10702 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10704 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10705 OPTION_MASK_ISA_CLDEMOTE.
10706 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10707 (ix86_valid_target_attribute_inner_p): Ditto.
10708 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10709 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10710 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10711 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10712 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10714 * config/i386/i386.opt: Add -mcldemote.
10715 * config/i386/x86intrin.h: New header.
10716 * doc/invoke.texi: Add -mcldemote.
10718 2018-05-14 Richard Biener <rguenther@suse.de>
10720 * doc/match-and-simplify.texi: Adjust :s documentation.
10722 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10724 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10725 intended memcpy size.
10726 (REORDER_45): Likewise.
10728 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10730 * sort.cc: New file.
10731 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10732 * vec.c (qsort_chk): Use gcc_qsort.
10733 * Makefile.in (OBJS-libcommon): Add sort.o.
10734 (build/sort.o): New target. Use it...
10735 (BUILD_RTL): ... here, and...
10736 (build/gencfn-macros): ... here, and...
10737 (build/genmatch): ... here.
10739 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10740 Chung-Ju Wu <jasonwucj@gmail.com>
10742 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10743 * config/nds32/nds32-graywolf.md: New file.
10744 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10747 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10748 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10749 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10750 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10751 * config/nds32/pipelines.md: Include n15 settings.
10753 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10754 Chung-Ju Wu <jasonwucj@gmail.com>
10756 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10757 * config/nds32/nds32-n13.md: New file.
10758 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10759 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10761 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10762 * config/nds32/nds32.md (pipeline_model): Add n13.
10763 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10764 * config/nds32/pipelines.md: Include n13 settings.
10766 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10767 Chung-Ju Wu <jasonwucj@gmail.com>
10769 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10770 * config/nds32/nds32-n10.md: New file.
10771 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10772 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10774 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10775 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10776 * config/nds32/nds32.md (pipeline_model): Add n10.
10777 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10778 * config/nds32/pipelines.md: Include n10 settings.
10780 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10781 Kito Cheng <kito.cheng@gmail.com>
10782 Chung-Ju Wu <jasonwucj@gmail.com>
10784 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10785 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10786 Add enum values for DSP extension instructions.
10787 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10789 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10790 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10791 New code iterators.
10792 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10793 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10794 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10795 * config/nds32/nds32-intrinsic.md: Likewise.
10796 * config/nds32/nds32_intrinsic.h: Likewise.
10797 * config/nds32/nds32-md-auxiliary.c: Likewise.
10798 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10799 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10800 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10801 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10802 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10803 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10804 TYPE_DMAC in switch statement.
10805 * config/nds32/nds32.c: New checking and implementation for DSP
10806 extension instructions.
10807 * config/nds32/nds32.h: Likewise.
10808 * config/nds32/nds32.md: Likewise.
10809 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10810 * config/nds32/predicates.md: Implement new predicates for DSP
10813 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10815 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10816 Reformat alternatives and attributes so it is easier to identify
10817 which constraints/attributes go with which instruction.
10818 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10819 (mov<mode>_softfloat32, FMOVE64): Likewise.
10820 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10821 (mov<mode>_softfloat64, FMOVE64): Likewise.
10823 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10825 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10827 (Basic PowerPC Built-in Functions): The new name of the
10828 subsection previously known as "PowerPC Built-in Functions".
10829 (Basic PowerPC Built-in Functions Available on all Configurations):
10831 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10832 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10833 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10834 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10836 2018-05-11 Martin Jambor <mjambor@suse.cz>
10839 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10842 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10845 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10847 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10849 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10850 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10851 (ix86_handle_option): Handle -mwaitpkg.
10852 * config.gcc: New header.
10853 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10854 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10855 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10857 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10858 OPTION_MASK_ISA_WAITPKG.
10859 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10860 (ix86_option_override_internal): Add PTA_WAITPKG.
10861 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10862 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10863 IX86_BUILTIN_TPAUSE.
10864 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10865 __builtin_ia32_umwait and __builtin_ia32_tpause.
10866 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10867 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10868 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10869 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10870 UNSPECV_TPAUSE): New.
10871 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10872 * config/i386/i386.opt: Add -mwaitpkg.
10873 * config/i386/waitpkgintrin.h: New file.
10874 * config/i386/x86intrin.h: New header.
10875 * doc/invoke.texi: Add -mwaitpkg.
10877 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10880 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10882 (cortex-m0): Use armv6s-m isa.
10883 (cortex-m0plus): Likewise.
10884 (cortex-m1): Likewise.
10885 (cortex-m0.small-multiply): Likewise.
10886 (cortex-m0plus.small-multiply): Likewise.
10887 (cortex-m1.small-multiply): Likewise.
10889 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10890 Jakub Jelinek <jakub@redhat.com>
10892 PR tree-optimization/85692
10893 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10894 source permute as well.
10896 2018-05-11 Martin Liska <mliska@suse.cz>
10899 * doc/extend.texi: Document LLVM style format for no_sanitize
10902 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10904 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10905 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10906 (mode_supports_vsx_dform_quad): Likewise.
10907 (mode_supports_vmx_dform): Move these functions to be next to the
10908 other mode_supports functions.
10909 (mode_supports_dq_form): Likewise.
10910 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10911 mode_supports_dq_form.
10912 (reg_offset_addressing_ok_p): Likewise.
10913 (offsettable_ok_by_alignment): Likewise.
10914 (rs6000_legitimate_offset_address_p): Likewise.
10915 (legitimate_lo_sum_address_p): Likewise.
10916 (rs6000_legitimize_address): Likewise.
10917 (rs6000_legitimize_reload_address): Likewise.
10918 (rs6000_secondary_reload_inner): Likewise.
10919 (rs6000_preferred_reload_class): Likewise.
10920 (rs6000_output_move_128bit): Likewise.
10922 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10924 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10925 Generate SImode target register for null target.
10926 <case IX86_BUILTIN_XGETBV>: Ditto.
10927 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10928 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10930 2018-05-10 Carl Love <cel@us.ibm.com>
10932 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10933 dcbtt and dcbtstt if operands[2] is 0.
10935 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10938 * config/i386/sse.md (usadv64qi): New expander.
10940 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10942 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10943 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10944 -maltivec=be support.
10945 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10946 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10947 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10948 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10949 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10950 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10951 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10952 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10953 altivec_vsumsws): Adjust.
10954 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10955 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10956 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10958 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10959 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10960 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10961 (altivec_lve<VI_char>x): Delete expand.
10962 (*altivec_lve<VI_char>x_internal): Rename to...
10963 (altivec_lve<VI_char>x): ... this.
10964 (altivec_lvxl_<mode>): Delete expand.
10965 (*altivec_lvxl_<mode>_internal): Rename to ...
10966 (altivec_lvxl_<mode>): ... this.
10967 (altivec_stvxl_<mode>): Delete expand.
10968 (*altivec_stvxl_<mode>_internal): Rename to ...
10969 (altivec_stvxl_<mode>): ... this.
10970 (altivec_stve<VI_char>x): Delete expand.
10971 (*altivec_stve<VI_char>x_internal): Rename to ...
10972 (altivec_stve<VI_char>x): ... this.
10973 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10974 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10975 reduc_plus_scal_<mode>): Adjust.
10976 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10978 (rs6000_cpu_cpp_builtins): Adjust.
10979 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10980 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10981 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10982 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10983 -maltivec=be support.
10984 (rs6000_split_vec_extract_var): Adjust.
10985 (rs6000_split_v4si_init): Adjust.
10986 (swap_selector_for_mode): Delete.
10987 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10988 altivec_expand_stvex_be): Delete.
10989 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10990 -maltivec=be support.
10991 (rs6000_gimple_fold_builtin): Ditto.
10992 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10994 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10995 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10996 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10997 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10998 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10999 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
11000 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
11001 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
11002 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
11003 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
11004 anonymous split): Adjust.
11005 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
11006 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
11008 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
11010 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
11011 when --with-gxx-include-dir is also specified.
11012 * configure: Regenerate.
11014 2018-05-09 Jim Wilson <jimw@sifive.com>
11017 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11018 * config/riscv/t-withmultilib: New.
11019 * config/riscv/withmultilib.h: New.
11020 * doc/install.texi: Document RISC-V --with-multilib-list support.
11022 2018-05-09 Richard Biener <rguenther@suse.de>
11024 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
11026 (vect_bb_vectorization_profitable_p): Adjust. Compute
11027 actual scalar cost using the cost vector and the add_stmt_cost
11030 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11032 PR rtl-optimization/85645
11033 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
11034 in the REG_CFA_REGISTER note for LR, don't leave it empty.
11036 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11038 PR rtl-optimization/85645
11039 * shrink-wrap.c (spread_components): Return a boolean saying if
11040 anything was changed.
11041 (try_shrink_wrapping_separate): Iterate spread_components until
11042 nothing changes anymore.
11044 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11046 PR rtl-optimization/85645
11047 * regrename.c (build_def_use): Also kill the chains that include the
11048 destination of a REG_CFA_REGISTER note.
11050 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11052 PR rtl-optimization/85645
11053 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
11054 insn that has a REG_CFA_REGISTER note.
11056 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
11058 * cfgexpand.c (expand_clobber): New function.
11059 (expand_gimple_stmt_1): Use it.
11060 * tree-vect-stmts.c (vect_clobber_variable): New function,
11062 (vectorizable_simd_clone_call): ...here.
11063 (vectorizable_store): Emit a clobber either side of an
11064 IFN_STORE_LANES sequence.
11065 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
11067 2018-05-09 Tom de Vries <tom@codesourcery.com>
11070 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
11071 (define_insn "trap_if_false"): Add exit after trap.
11073 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
11075 PR rtl-optimization/85638
11076 * bb-reorder.c: Include common/common-target.h.
11077 (create_forwarder_block): New function extracted from...
11078 (fix_up_crossing_landing_pad): ...here. Rename into...
11079 (dw2_fix_up_crossing_landing_pad): ...this.
11080 (sjlj_fix_up_crossing_landing_pad): New function.
11081 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
11082 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
11083 from both partitions and exit the loop after one iteration.
11085 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11088 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11090 (Basic PowerPC Built-in Functions): The new name of the
11091 subsection previously known as "PowerPC Built-in Functions".
11092 (Basic PowerPC Built-in Functions Available on all Configurations):
11094 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11096 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11098 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11100 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11103 2018-05-08 Jim Wilson <jimw@sifive.com>
11105 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
11106 (LD_EMUL_SUFFIX): New.
11107 (LINK_SPEC): Use it.
11109 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11111 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11113 (Basic PowerPC Built-in Functions): The new name of the
11114 subsection previously known as "PowerPC Built-in Functions".
11115 (Basic PowerPC Built-in Functions Available on all Configurations):
11117 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11119 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11121 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11123 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11126 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11129 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11130 after cmpelim optimization.
11132 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11134 * config.gcc: Support "goldmont".
11135 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11136 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11137 PROCESSOR_GOLDMONT.
11138 * config/i386/i386.c (m_GOLDMONT): Define.
11139 (processor_target_table): Add "goldmont".
11140 (PTA_GOLDMONT): Define.
11141 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11142 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11143 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11144 (fold_builtin_cpu): Add "goldmont".
11145 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11146 (ix86_option_override_internal): Add "goldmont".
11147 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11148 (processor_type): Add PROCESSOR_GOLDMONT.
11149 * config/i386/i386.md: Add CPU "glm".
11150 * config/i386/glm.md: New file.
11151 * config/i386/x86-tune.def: Add m_GOLDMONT.
11152 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11154 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11157 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11159 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11160 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11161 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11164 * config/i386/i386.c (ix86_fold_builtin): Handle
11165 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11168 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11169 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11171 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11174 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11175 (check_arch): Likewise.
11176 (check_fpu): Return the result rather than printing it.
11177 (end arch): Fix operator precedence.
11178 (end cpu): Likewise.
11179 (END): Print the result from check_fpu.
11181 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11186 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11187 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11188 (*fcmuo<mode>_and): New patterns.
11190 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11192 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11193 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11194 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11195 (cmp_op, sve_imm_con): New code attributes.
11196 (SVE_COND_INT_CMP, imm_con): Delete.
11197 (cmp_op): Remove above unspecs from int attribute.
11198 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11200 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11201 comparison-specific unspecs.
11202 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11203 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11204 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11205 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11206 (*vec_fcm<cmp_op><mode>): Rename to...
11207 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11208 (*vec_fcmuo<mode>): Rename to...
11209 (*fcmuo<mode>): ...this and adjust likewise.
11210 (*pred_fcm<cmp_op><mode>): New pattern.
11211 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11212 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11214 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11216 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11217 (aarch64_emit_sve_predicated_cond): New function.
11218 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11219 (aarch64_emit_unspec_cond_or): Replace with...
11220 (aarch64_emit_sve_or_conds): ...this new function. Use
11221 aarch64_emit_sve_ptrue_op for the individual comparisons and
11222 aarch64_emit_binop to OR them together.
11223 (aarch64_emit_inverted_unspec_cond): Replace with...
11224 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11225 aarch64_emit_sve_ptrue_op for the comparison and
11226 aarch64_emit_unop to invert the result.
11227 (aarch64_expand_sve_vec_cmp_float): Update after the above
11228 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11230 2018-05-07 Nathan Sidwell <nathan@acm.org>
11232 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11233 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11234 (Backwards Compatibility): Likewise.
11236 2018-05-07 Luis Machado <luis.machado@linaro.org>
11240 2018-05-07 Luis Machado <luis.machado@linaro.org>
11242 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11243 <prefetch_dynamic_strides>: New const bool field.
11244 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11245 prefetch_dynamic_strides.
11246 (exynosm1_prefetch_tune): Likewise.
11247 (thunderxt88_prefetch_tune): Likewise.
11248 (thunderx_prefetch_tune): Likewise.
11249 (thunderx2t99_prefetch_tune): Likewise.
11250 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11252 (aarch64_override_options_internal): Update to set
11253 PARAM_PREFETCH_DYNAMIC_STRIDES.
11254 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11255 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11256 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11257 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11258 prefetch-dynamic-strides setting.
11260 2018-05-07 Luis Machado <luis.machado@linaro.org>
11262 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11263 <minimum_stride>: New const int field.
11264 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11265 minimum_stride field.
11266 (exynosm1_prefetch_tune): Likewise.
11267 (thunderxt88_prefetch_tune): Likewise.
11268 (thunderx_prefetch_tune): Likewise.
11269 (thunderx2t99_prefetch_tune): Likewise.
11270 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11271 (aarch64_override_options_internal): Update to set
11272 PARAM_PREFETCH_MINIMUM_STRIDE.
11273 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11274 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11275 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11276 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11277 stride is constant and is below the minimum stride threshold.
11279 2018-05-07 Luis Machado <luis.machado@linaro.org>
11281 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11284 2018-05-07 Luis Machado <luis.machado@linaro.org>
11286 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11287 <prefetch_dynamic_strides>: New const bool field.
11288 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11289 prefetch_dynamic_strides.
11290 (exynosm1_prefetch_tune): Likewise.
11291 (thunderxt88_prefetch_tune): Likewise.
11292 (thunderx_prefetch_tune): Likewise.
11293 (thunderx2t99_prefetch_tune): Likewise.
11294 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11296 (aarch64_override_options_internal): Update to set
11297 PARAM_PREFETCH_DYNAMIC_STRIDES.
11298 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11299 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11300 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11301 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11302 prefetch-dynamic-strides setting.
11304 2018-05-07 Luis Machado <luis.machado@linaro.org>
11306 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11307 <minimum_stride>: New const int field.
11308 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11309 minimum_stride field.
11310 (exynosm1_prefetch_tune): Likewise.
11311 (thunderxt88_prefetch_tune): Likewise.
11312 (thunderx_prefetch_tune): Likewise.
11313 (thunderx2t99_prefetch_tune): Likewise.
11314 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11315 (aarch64_override_options_internal): Update to set
11316 PARAM_PREFETCH_MINIMUM_STRIDE.
11317 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11318 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11319 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11320 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11321 stride is constant and is below the minimum stride threshold.
11323 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11326 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11327 the type is addressable. Don't force op into register if it has
11330 2018-05-05 Roland McGrath <mcgrathr@google.com>
11333 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11334 any section for which we don't know a specific type it should have,
11335 regardless of name. Previously this was done only for the exact
11336 names ".init_array", ".fini_array", and ".preinit_array".
11337 (default_elf_asm_named_section): Add comment about
11338 relationship with default_section_type_flags and SECTION_NOTYPE.
11339 (get_section): Don't consider it a type conflict if one side has
11340 SECTION_NOTYPE and the other doesn't, as long as neither has the
11341 SECTION_BSS et al used in the default_section_type_flags logic.
11343 2018-05-05 Tom de Vries <tom@codesourcery.com>
11346 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11347 (workaround_barsyncs): New function.
11348 (nvptx_reorg): Use workaround_barsyncs.
11349 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11350 (define_expand "nvptx_membar_cta"): New define_expand.
11351 (define_insn "*nvptx_membar_cta"): New insn.
11353 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11355 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11356 To improve optimization opportunities.
11357 * builtin-types.def: The new needed builtin types for the above.
11359 2018-05-04 Richard Biener <rguenther@suse.de>
11361 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11362 * gimple-ssa-store-merging.c
11363 (imm_store_chain_info::output_merged_store): Remove redundant create,
11364 release split_store vector contents on failure.
11365 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11366 scalar stmt vector on cache hit.
11368 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11370 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11372 * config.gcc (powerpc-xilinx-eabi*): Remove.
11373 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11375 (fusion_addis_mem_combo_load): Ditto.
11376 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11378 (rs6000_cpu_cpp_builtins): Ditto.
11379 * config/rs6000/rs6000-linux.c
11380 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11381 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11382 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11384 (rs6000_setup_reg_addr_masks): Ditto.
11385 (rs6000_init_hard_regno_mode_ok): Ditto.
11386 (rs6000_option_override_internal): Ditto.
11387 (legitimate_lo_sum_address_p): Ditto.
11388 (rs6000_legitimize_address): Ditto.
11389 (rs6000_legitimize_reload_address): Ditto.
11390 (rs6000_legitimate_address_p): Ditto.
11391 (abi_v4_pass_in_fpr): Ditto.
11392 (setup_incoming_varargs): Ditto.
11393 (rs6000_gimplify_va_arg): Ditto.
11394 (rs6000_split_multireg_move): Ditto.
11395 (rs6000_savres_strategy): Ditto.
11396 (rs6000_emit_prologue_components): Ditto.
11397 (rs6000_emit_epilogue_components): Ditto.
11398 (rs6000_emit_prologue): Ditto.
11399 (rs6000_emit_epilogue): Ditto.
11400 (rs6000_elf_file_end): Ditto.
11401 (rs6000_function_value): Ditto.
11402 (rs6000_libcall_value): Ditto.
11403 * config/rs6000/rs6000.h: Ditto.
11404 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11405 (TARGET_MINMAX): ... this. New.
11406 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11407 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11408 (*movsi_internal1_single): Delete.
11409 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11410 mfpu=, mxilinx-fpu): Delete.
11411 * config/rs6000/singlefp.h: Delete.
11412 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11413 * config/rs6000/t-rs6000: Ditto.
11414 * config/rs6000/t-xilinx: Delete.
11415 * config/rs6000/titan.md: Adjust for fp_type removal.
11416 * config/rs6000/vsx.md: Remove Xilinx FP support.
11417 (VStype_simple): Delete.
11418 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11419 * config/rs6000/xfpu.h: Delete.
11420 * config/rs6000/xfpu.md: Delete.
11421 * config/rs6000/xilinx.h: Delete.
11422 * config/rs6000/xilinx.opt: Delete.
11423 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11424 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11426 2018-05-04 Tom de Vries <tom@codesourcery.com>
11429 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11432 2018-05-04 Richard Biener <rguenther@suse.de>
11434 PR middle-end/85627
11435 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11436 (expand_complex_div_wide): Likewise.
11437 (expand_complex_operations_1): Likewise.
11438 (expand_complex_libcall): Preserve EH info of the original stmt.
11439 (tree_lower_complex): Handle removed blocks.
11440 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11441 on complex multiplication and division libcall builtins.
11443 2018-05-04 Richard Biener <rguenther@suse.de>
11445 PR middle-end/85574
11446 * fold-const.c (negate_expr_p): Restrict negation of operand
11447 zero of a division to when we know that can happen without
11449 (fold_negate_expr_1): Likewise.
11451 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11454 * real.h (real_nextafter): Declare.
11455 * real.c (real_nextafter): New function.
11456 * fold-const-call.c (fold_const_nextafter): New function.
11457 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11458 CASE_CFN_NEXTTOWARD.
11459 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11460 even when arg1_mode is different from arg0_mode.
11462 2018-05-03 Nathan Sidwell <nathan@acm.org>
11464 * doc/extend.texi (Deprecated Features): Remove
11465 -ffriend-injection.
11466 (Backwards Compatibility): Likewise.
11467 * doc/invoke.texi (C++ Language Options): Likewise.
11468 (C++ Dialect Options): Likewise.
11470 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11473 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11474 _mm512_mask_mullox_epi64): New intrinsics.
11476 2018-05-03 Tom de Vries <tom@codesourcery.com>
11479 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11480 dump files): Add offload-tree.
11482 2018-05-03 Richard Biener <rguenther@suse.de>
11484 PR tree-optimization/85615
11485 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11486 to loops not nested in BBs loop father to avoid creating multi-entry
11489 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11491 PR tree-optimization/70291
11492 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11493 arguments. Change return type to tree. Emit libcall as a new
11494 statement rather than replacing existing one when inplace_p is true.
11495 (expand_complex_multiplication_components): New function.
11496 (expand_complex_multiplication): Expand floating-point complex
11497 multiplication using the above.
11498 (expand_complex_division): Rename inner_type parameter to type.
11499 Update expand_complex_libcall call-site.
11500 (expand_complex_operations_1): Update expand_complex_multiplication
11501 and expand_complex_division call-sites.
11503 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11506 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11507 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11508 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11509 the highest significant bit of the shift count mask is clear. In
11510 check whether and[sq]i3 is needed verify that all significant bits
11511 of the shift count other than the highest are set.
11513 2018-05-02 Tom de Vries <tom@codesourcery.com>
11516 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11517 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11518 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11519 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11520 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11521 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11522 __builtin_goacc_parlevel_size.
11524 2018-05-02 Richard Biener <rguenther@suse.de>
11526 PR tree-optimization/85597
11527 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11528 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11531 2018-05-02 Tom de Vries <tom@codesourcery.com>
11534 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11535 dump files): Add ltrans-tree.
11537 2018-05-02 Tom de Vries <tom@codesourcery.com>
11540 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11541 dump files): Add wpa-ipa.
11543 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11545 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11546 powerpc*-*-linux*paired* target.
11547 * config/rs6000/750cl.h: Delete.
11548 * config/rs6000/paired.h: Delete.
11549 * config/rs6000/paired.md: Delete.
11550 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11552 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11553 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11554 comment. Remove paired float support.
11555 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11556 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11558 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11559 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11561 * config/rs6000/rs6000.c: Remove paired float support.
11562 (paired_expand_vector_init, paired_expand_vector_move,
11563 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11564 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11565 paired_expand_builtin, paired_expand_predicate_builtin,
11566 paired_init_builtins): Delete.
11567 * config/rs6000/rs6000.h: Remove paired float support.
11568 * config/rs6000/rs6000.md: Remove paired float support.
11569 (move_from_CR_ov_bit): Delete.
11570 * config/rs6000/rs6000.opt (mpaired): Delete.
11571 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11572 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11574 2018-05-02 Richard Biener <rguenther@suse.de>
11576 PR middle-end/85567
11577 * gimplify.c (gimplify_save_expr): When in SSA form allow
11578 SAVE_EXPRs to compute to SSA vars.
11580 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11583 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11584 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11585 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11586 clobber operands[2], instead use a new pseudo. Formatting fixes.
11588 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11590 PR tree-optimization/85586
11591 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11592 exit early for statements in the same group if the accesses are
11595 2018-05-02 Tom de Vries <tom@codesourcery.com>
11598 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11601 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11603 PR tree-optimization/85143
11604 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11606 2018-05-01 Tom de Vries <tom@codesourcery.com>
11609 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11610 not found" error message.
11612 2018-05-01 Tom de Vries <tom@codesourcery.com>
11615 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11616 * vec.c (test_ordered_remove_if): New function.
11617 (vec_c_tests): Call test_ordered_remove_if.
11618 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11619 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11620 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11621 VEC_ORDERED_REMOVE_IF.
11623 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11625 PR tree-optimization/82665
11626 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11627 pointer subtraction where arguments come from a memchr call.
11629 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11631 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11632 --push-state --as-needed and --pop-state instead of --as-needed and
11633 --no-as-needed if ld supports it.
11634 * configure: Regenerated.
11637 * doc/install.texi2html: Replace _002d with - and _002a with * in
11638 generated html files using sed.
11640 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11643 * gcc-rich-location.c (blank_line_before_p): New function.
11644 (use_new_line): New function.
11645 (gcc_rich_location::add_fixit_insert_formatted): New function.
11646 * gcc-rich-location.h
11647 (gcc_rich_location::add_fixit_insert_formatted): New function.
11649 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11651 * selftest.c (assert_streq): Rename "expected" and "actual" to
11652 "val1" and "val2". Extend NULL-handling to cover both inputs
11653 symmetrically, while still requiring both to be non-NULL for a pass.
11654 * selftest.h (assert_streq): Rename "expected" and "actual" to
11656 (ASSERT_EQ): Likewise.
11657 (ASSERT_EQ_AT): Likewise.
11658 (ASSERT_KNOWN_EQ): Likewise.
11659 (ASSERT_KNOWN_EQ_AT): Likewise.
11660 (ASSERT_NE): Likewise.
11661 (ASSERT_MAYBE_NE): Likewise.
11662 (ASSERT_MAYBE_NE_AT): Likewise.
11663 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11664 the assertion to pass.
11665 (ASSERT_STREQ_AT): Likewise.
11667 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11669 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11670 interaction with -pie.
11672 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11674 * selftest.h: Fix alphabetization of per-source-file selftest
11677 2018-04-30 Jason Merrill <jason@redhat.com>
11679 PR c++/61982 - dead stores to destroyed objects.
11680 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11683 2018-04-30 Jason Merrill <jason@redhat.com>
11685 * tree.c (build_clobber): New.
11686 * tree.h: Declare it.
11687 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11689 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11691 * diagnostic-show-locus.c (layout::layout): Update for
11692 location_get_source_line returning a char_span.
11693 (struct char_span): Move to input.h.
11694 (struct correction): Update for fields in char_span becoming
11696 (struct source_line): Update for location_get_source_line
11697 returning a char_span.
11698 (layout::print_line): Likewise.
11699 * edit-context.c (edited_file::print_content): Likewise.
11700 (edited_file::print_diff_hunk): Likewise.
11701 (edited_file::print_run_of_changed_lines): Likewise.
11702 (edited_file::get_num_lines): Likewise.
11703 (edited_line::edited_line): Likewise.
11704 * final.c (asm_show_source): Likewise.
11705 * input.c (location_get_source_line): Convert return type
11706 from const char * to char_span, losing the final "line_len"
11708 (dump_location_info): Update for the above.
11709 (get_substring_ranges_for_loc): Likewise. Use a char_span
11710 when handling the literal within the line.
11711 (test_reading_source_line): Update for location_get_source_line
11712 returning a char_span.
11713 * input.h (class char_span): Move here from
11714 diagnostic-show-locus.c, converting from a struct to a class.
11715 Make data members private.
11716 (char_span::operator bool): New.
11717 (char_span::length): New.
11718 (char_span::get_buffer): New.
11719 (char_span::operator[]): New.
11720 (char_span::subspan): Make const.
11721 (char_span::xstrdup): New.
11722 (location_get_source_line): Convert return type from const char *
11723 to char_span, losing the final "line_size" param.
11725 2018-04-30 Jan Hubicka <jh@suse.cz>
11727 * lto-wrapper.c (ltrans_priorities): New static var.
11728 (cmp_priority): New.
11729 (run_gcc): Read priorities and if doing parallel build order
11730 the Makefile by them.
11732 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11734 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11736 2018-04-30 Richard Biener <rguenther@suse.de>
11738 * tree-cfg.c (verify_address): Remove base argument, add
11739 flag whether to check TREE_ADDRESSABLE and do that.
11740 (verify_expr): Remove.
11741 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11742 (verify_gimple_assign_single): Likewise.
11743 (verify_gimple_switch): Likewise.
11744 (verify_expr_location_1): Dereference tp once. Add (disabled)
11745 piece from verify_expr.
11746 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11748 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11750 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11752 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11754 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11755 (small_data_pattern): Likewise.
11756 (arc_rewrite_small_data): Likewise.
11757 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11758 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11759 (get_symbol_alignment): New function.
11760 (legitimate_small_data_address_p): Likewise.
11761 (legitimate_scaled_address): Update, call
11762 legitimate_small_data_address_p.
11763 (output_sdata): New static variable.
11764 (arc_print_operand): Update how we handle small data operands.
11765 (arc_print_operand_address): Likewise.
11766 (arc_legitimate_address_p): Update, use
11767 legitimate_small_data_address_p.
11768 (arc_rewrite_small_data_p): Remove.
11769 (arc_rewrite_small_data_1): Likewise.
11770 (arc_rewrite_small_data): Likewise.
11771 (small_data_pattern): Likewise.
11772 (compact_sda_memory_operand): Update to use
11773 legitimate_small_data_address_p and get_symbol_alignment.
11774 (prepare_move_operands): Don't rewite sdata pattern.
11775 (prepare_extend_operands): Remove.
11776 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11778 (zero_extendqisi2): Likewise.
11779 (zero_extendhisi2): Likewise.
11780 (extendqihi2): Likewise.
11781 (extendqisi2): Likewise.
11782 (extendhisi2): Likewise.
11783 (addsi3): Likewise.
11784 (subsi3): Likewise.
11785 (andsi3): Likewise.
11786 * config/arc/constraints.md (Usd): Change it to memory constraint.
11788 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11790 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11791 as source of std instructions.
11792 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11793 allow 6-bit constants as source for store instructions.
11794 (movdi_insn): Update instruction pattern to allow 6-bit constants
11795 as source for store instructions.
11797 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11799 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11801 2018-04-30 Nathan Sidwell <nathan@acm.org>
11802 Sandra Loosemore <sandra@codesourcery.com>
11804 * dumpfile.c (dump_open): Allow '-' for stdout.
11805 * doc/invoke.texi (Developer Options): Document dump filename
11806 determination early. Document stdin/stdout selection.
11808 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11810 Microblaze Target: PIC data text relative
11812 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11813 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11815 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11816 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11817 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11818 New addressing mode for data-text relative position indepenedent code.
11819 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11820 'ADDRESS_SYMBOLIC_TXT_REL'.
11821 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11822 (microblaze_legitimate_pic_operand): Exclude function calls from
11823 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11824 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11826 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11827 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11828 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11829 for 'address + offset'.
11830 (microblaze_expand_prologue): Add new function prologue call for
11832 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11833 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11834 table in case of TARGET_PIC_DATA_TEXT_REL.
11835 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11836 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11837 Add new macros 'UNSPEC_TEXT',
11838 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11839 + exclude function calls from 'UNSPEC_PLT' in case of data text
11841 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11842 new target hook for generating address diff vector tables in case of
11844 * doc/tm.texi : Regenerate.
11845 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11846 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11847 of addr diff vector generation.
11848 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11849 target hook definition.
11850 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11851 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11852 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11853 option for data text relative.
11855 2018-04-30 Richard Biener <rguenther@suse.de>
11857 * tree-chrec.h (evolution_function_is_constant_p): Remove
11859 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11861 2018-04-30 Richard Biener <rguenther@suse.de>
11864 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11866 2018-04-30 Richard Biener <rguenther@suse.de>
11868 PR tree-optimization/28364
11869 PR tree-optimization/85275
11870 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11871 copying first exit test.
11873 2018-04-28 Mark Wielaard <mark@klomp.org>
11875 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11876 dwarf_version >= 5.
11877 (dwarf_AT): Handle DW_AT_addr_base.
11878 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11880 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11883 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11884 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11885 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11886 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11888 2018-04-28 Richard Biener <rguenther@suse.de>
11890 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11891 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11892 to reflect use. Only add interesting stmts.
11894 2018-04-27 Martin Jambor <mjambor@suse.cz>
11897 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11898 the jump function allows for passing through aggregate values.
11900 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11902 * input.h (in_system_header_at): Convert from macro to inline
11904 (from_macro_expansion_at): Likewise.
11905 (from_macro_definition_at): Likewise.
11907 2018-04-27 Jeff Law <law@redhat.com>
11909 * config.gcc: Mark tile* targets as deprecated/obsolete.
11911 2018-04-27 Richard Biener <rguenther@suse.de>
11913 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11916 2018-04-27 Richard Biener <rguenther@suse.de>
11918 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11920 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11922 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11923 with Yd constraint. Set "preferred_for_speed" attribute from
11924 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11925 with Yd constraint.
11926 (*movdi_internal): Ditto.
11927 (movti_interunit splitters): Remove
11928 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11929 (movdi_interunit splitters): Ditto.
11930 * config/i386/constraints.md (Ye): Remove.
11931 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11933 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11936 * config/aarch64/constraints.md (Usg): Limit to 31.
11937 (Usj): Limit to 63.
11939 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11941 PR tree-optimization/85529
11942 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11943 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11944 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11945 zero extension or masking of the MSB bit.
11946 (optimize_range_tests): Add FIRST_BB argument, pass it through
11947 to optimize_range_tests_var_bound.
11948 (maybe_optimize_range_tests, reassociate_bb): Adjust
11949 optimize_range_tests callers.
11951 2018-04-26 Richard Biener <rguenther@suse.de>
11952 Jakub Jelinek <jakub@redhat.com>
11954 * cgraph.h (symbol_table): Just declare debug method here.
11955 * symtab.c (symbol_table::debug): Define.
11957 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11959 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11961 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11963 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11964 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11965 (*movdi_internal): Substitute Yi and Yj constraint with x
11966 and Ym and Yn constraint with y constraint. Update "isa"
11967 attribute and set "preferred_for_speed" attribute from
11968 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11969 (*movsi_internal): Ditto.
11970 (*movdf_internal): Ditto.
11971 (*movsf_internal): Ditto.
11972 (*zero_extendsidi2): Ditto.
11973 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11974 (sse2_loadld): Ditto.
11975 (*vec_extract<ssevecmodelower>_0): Ditto.
11976 (*vec_extractv4si_0_zext_sse4): Ditto.
11977 (vec_concatv2di): Ditto.
11978 (*vec_dup<mode>): Ditto.
11979 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11980 * config/i386/constraints.md (Yi): Remove.
11985 2018-04-26 Nathan Sidwell <nathan@acm.org>
11987 * dumpfile.c (dump_open): New.
11988 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11989 (dump_finish): Detect stdio/stderr by value not name.
11991 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11993 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11995 2018-04-26 Tom de Vries <tom@codesourcery.com>
11998 * config/nvptx/nvptx.c (verify_neutering_jumps)
11999 (verify_neutering_labels): New function
12000 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12002 2018-04-26 Tom de Vries <tom@codesourcery.com>
12005 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12006 (nvptx_single): Use needs_neutering_p to skip over insns that do not
12009 2018-04-26 Richard Biener <rguenther@suse.de>
12010 Tom de Vries <tom@codesourcery.com>
12013 * lto-streamer-out.c (output_function): Fixup loops if required to match
12014 discovery done in the reader.
12016 2018-04-26 Richard Biener <rguenther@suse.de>
12018 PR tree-optimization/85116
12019 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
12020 have a loop exit from the single latch predecessor. Remove
12021 case of header with just condition.
12022 (ch_base::copy_headers): Exclude infinite loops from any
12024 (pass_ch::execute): Record exits.
12026 2018-04-26 Richard Biener <rguenther@suse.de>
12028 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
12029 prologue cost vector and pass it to vect_get_load_cost.
12030 (vect_get_peeling_costs_all_drs): Likewise.
12031 (vect_peeling_hash_get_lowest_cost): Likewise.
12032 (vect_enhance_data_refs_alignment): Likewise.
12034 2018-04-26 Richard Biener <rguenther@suse.de>
12036 PR middle-end/85450
12037 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
12038 checking of integer<->pointer conversions.
12039 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
12040 sign-/zero-extending pointer types.
12041 (expand_omp_for_static_chunk): Likewise.
12043 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
12044 Jean Lee <xiaoyur347@gmail.com>
12046 * config/mips/mips.c (mips_asan_shadow_offset): New function.
12047 (TARGET_ASAN_SHADOW_OFFSET): Define.
12048 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
12049 true for -fsanitize=address.
12051 2018-04-25 Mark Wielaard <mark@klomp.org>
12053 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
12056 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12058 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
12059 than "alu", remove explicit "memory" and "imm_disp" attributes.
12060 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
12062 PR middle-end/85414
12063 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
12064 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
12065 gen_lowpart_no_emit.
12067 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
12070 * config/i386/i386.c (ix86_expand_builtin): Change memory
12071 operand to XI, extend p0 to Pmode.
12072 * config/i386/i386.md: Change unspec volatile and operand
12073 1 mode to XI, change operand 0 mode to P.
12075 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12077 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
12078 GET_MODE_MASK before any checking.
12079 (nds32_can_use_bset_p): Likewise.
12080 (nds32_can_use_btgl_p): Likewise.
12082 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12084 * config/nds32/nds32-doubleword.md: New define_split pattern for
12085 illegal register number.
12087 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12089 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
12091 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12093 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
12095 2018-04-25 Richard Biener <rguenther@suse.de>
12097 * lto-streamer.h (LTO_major_version): Bump to 8.
12099 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12101 * BASE-VER: Set to 9.0.0.
12103 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
12105 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
12106 in __abskf2 and __powikf2.
12108 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12111 * config/aarch64/constraints.md (Usg, Usj): New constraints.
12112 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
12113 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12114 Use the above on operand 2. Reindent.
12115 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12117 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12120 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12122 * config/i386/i386.opt (mcet): Removed.
12123 * doc/install.texi: Remove -mcet documentation.
12124 * doc/invoke.texi: Likewise.
12126 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12129 * doc/install.texi: Remove -mcet from bootstrap-cet.
12131 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12134 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12135 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12139 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12140 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12141 containing a CONST_VECTOR.
12143 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12145 * doc/install.texi: Update newlib dependency for nvptx.
12147 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12150 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12151 instead of INTVAL when shifting x left.
12153 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12155 PR tree-optimization/85478
12156 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12157 vect_grouped_store_supported for single element vectors.
12159 2018-04-24 Richard Biener <rguenther@suse.de>
12162 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12163 load cost increase to the case of non-constant step.
12165 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12168 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12169 destination if any_malformed_asm.
12171 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12173 PR middle-end/85496
12174 * expr.c (store_field): In the bitfield case, if the value comes from
12175 a function call and is returned in registers by means of a PARALLEL,
12176 do not change the mode of the temporary unless BLKmode and VOIDmode.
12178 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12180 PR rtl-optimization/85423
12181 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12182 dependencies to debug insns when the previous insn is non-debug.
12184 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12186 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12187 enums into a single definition.
12188 (fls): Fix predicates and printing.
12191 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12193 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12194 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12195 and short u6 immediate.
12196 (check_if_valid_sleep_operand): Remove.
12197 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12199 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12201 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12202 flag_always_save_lp condition.
12203 * config/nds32/nds32.opt (malways-save-lp): New option.
12205 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12207 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12208 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12209 * config/nds32/nds32.h
12210 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12211 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12213 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12215 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12216 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12218 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12219 Chung-Ju Wu <jasonwucj@gmail.com>
12221 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12223 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12224 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12226 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12228 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12230 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12232 * config/nds32/nds32-protos.h (nds32_data_alignment,
12233 nds32_local_alignment): Declare.
12234 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12235 nds32_local_alignment): New functions.
12236 (TARGET_CONSTANT_ALIGNMENT): Define.
12237 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12239 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12241 * config/nds32/nds32.c
12242 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12243 (TARGET_MODES_TIEABLE_P): Likewise.
12245 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12247 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12248 level Ofast and Og.
12250 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12251 Chung-Ju Wu <jasonwucj@gmail.com>
12253 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12254 for unaligned access.
12255 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12257 * config/nds32/nds32-intrinsic.md: Likewise.
12258 * config/nds32/nds32_intrinsic.h: Likewise.
12259 * config/nds32/nds32.h (nds32_builtins): Likewise.
12260 * config/nds32/nds32.opt (munaligned-access): New option.
12261 * config/nds32/nds32.c (nds32_asm_file_start): Display
12262 flag_unaligned_access status.
12264 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12266 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12267 -mno-relax is present.
12268 * config/riscv/linux.h (LINK_SPEC): Ditto.
12270 2018-04-20 Martin Sebor <msebor@redhat.com>
12273 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12275 (gimple_fold_builtin_stxcpy_chk): Same.
12276 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12278 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12281 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12282 __powikf2 when long double is IEEE 128-bit.
12284 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12286 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12287 step to make sure stack always aligned.
12289 2018-04-20 Carl Love <cel@us.ibm.com>
12292 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12293 size check for arg0.
12295 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12296 Tom de Vries <tom@codesourcery.com>
12299 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12300 Emit insns for calls too.
12301 (nvptx_find_par): Always look for worker-level predecessor insn.
12302 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12304 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12305 (nvptx_process_pars): Propagate frames for calls.
12307 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12310 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12312 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12313 (ix86_handle_option): Don't handle OPT_mibt.
12314 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12316 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12318 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12319 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12320 (ix86_target_macros): Define __CET__ with flag_cf_protection
12321 for -fcf-protection.
12322 * config/i386/i386.c (isa2_opts): Remove -mibt.
12323 * config/i386/i386.h (TARGET_IBT): Removed.
12324 (TARGET_IBT_P): Likewise.
12325 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12326 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12327 * config/i386/i386.opt (mcet): Update help message.
12328 (mshstk): Likewise.
12330 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12331 -mcet as an alias for -mshstk.
12333 2018-04-20 Richard Biener <rguenther@suse.de>
12335 PR middle-end/85475
12336 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12337 complexity by forcing a single use of the multiply operand.
12339 2018-04-20 Martin Jambor <mjambor@suse.cz>
12342 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12343 recursion dependency to only apply to non-clones.
12345 2018-04-20 Martin Jambor <mjambor@suse.cz>
12348 * ipa-cp.c (create_specialized_node): Check that clones of
12349 self-recursive edges exist during IPA-CP.
12351 2018-04-19 Toon Moene <toon@moene.org>
12353 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12356 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12358 PR tree-optimization/85467
12359 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12360 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12361 VECTOR_CST element to type.
12363 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12366 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12367 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12368 (builtin_longjmp): Likewise.
12369 (save_stack_nonlocal): New pattern.
12370 (restore_stack_nonlocal): Likewise.
12372 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12375 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12376 Replace ASM_OUTPUT_LABEL with fprintf.
12378 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12381 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12382 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12383 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12384 define __IBT__ and __SHSTK__ for -fcf-protection.
12385 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12387 (ix86_trampoline_init): Likewise.
12388 (x86_output_mi_thunk): Likewise.
12389 (ix86_notrack_prefixed_insn_p): Likewise.
12390 (ix86_option_override_internal): Don't disallow -fcf-protection.
12391 * config/i386/i386.md (rdssp<mode>): Also enable for
12393 (incssp<mode>): Likewise.
12394 (nop_endbr): Likewise.
12395 * config/i386/i386.opt (mcet): Change help message to built-in
12398 (mshstk): Likewise.
12399 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12400 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12401 enable CET built-in functions.
12403 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12405 * common/config/i386/i386-common.c
12406 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12407 OPTION_MASK_ISA_MOVDIRI_UNSET,
12408 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12409 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12410 * config.gcc (movdirintrin.h): New header.
12411 * config/i386/cpuid.h (bit_MOVDIRI,
12412 bit_MOVDIR64B): New bits.
12413 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12415 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12416 (VOID, PVOID, PCVOID)): New function types.
12417 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12418 __builtin_ia32_directstoreu_u64,
12419 __builtin_ia32_movdir64b): New builtins.
12420 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12421 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12423 (ix86_valid_target_attribute_inner_p): Ditto.
12424 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12425 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12426 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12427 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12428 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12429 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12430 (movdiri<mode>, movdir64b_<mode>): New.
12431 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12432 * config/i386/immintrin.h: Include movdirintrin.h.
12433 * config/i386/movdirintrin.h: New file.
12434 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12436 2018-04-19 Richard Biener <rguenther@suse.de>
12438 PR middle-end/85455
12439 * cfg.c (clear_bb_flags): When loop state says we have
12440 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12442 2018-04-19 Richard Biener <rguenther@suse.de>
12444 PR tree-optimization/84737
12445 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12446 copying restrict info.
12447 (vect_setup_realignment): Use it.
12448 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12449 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12450 the first DR to all generated stores.
12451 (vectorizable_load): Likewise for loads.
12453 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12455 PR tree-optimization/85446
12456 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12457 the integral and pointer types to have the same precision.
12459 * doc/install.texi: Document --disable-cet being the default and
12462 2018-04-18 Martin Liska <mliska@suse.cz>
12464 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12467 2018-04-18 Martin Liska <mliska@suse.cz>
12470 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12473 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12474 arguments if they are comparable.
12476 2018-04-18 Martin Liska <mliska@suse.cz>
12479 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12482 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12485 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12488 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12489 ENDBR after calling __morestack.
12491 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12494 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12495 by using gcc_base_ver to generate a gcc_driver_version, and use
12496 it when generating GCC_DRIVER_NAME.
12497 * configure: Regenerate.
12499 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12502 * config.gcc: Obsolete powerpc*-*-*spe*.
12504 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12507 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12508 (stabstr_D): Change type of unum from unsigned int to
12509 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12512 2018-04-17 Jim Wilson <jimw@sifive.com>
12515 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12516 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12517 Set arg_pointer_offset after using pretend_args_size.
12519 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12521 PR rtl-optimization/85431
12522 * dse.c (record_store): Ignore zero width stores.
12525 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12526 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12527 __builtin_stack_restore rather than after it.
12528 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12529 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12530 argument instead of virtual_dynamic_stack_rtx.
12532 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12534 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12536 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12537 Add note to error message to explain internal mapping of overloaded
12538 built-in function name to non-overloaded built-in function name.
12539 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12542 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12545 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12546 where the inputs overlap with the output.
12548 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12550 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12551 (=v, v) alternative and explicit "memory" attribute.
12552 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12553 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12555 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12556 "sselog1" type instead of "sselog".
12557 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12558 "sselog". Remove explicit "memory" attribute.
12559 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12560 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12562 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12563 "sselog1" type instead of "sselog". Remove explicit "memory"
12565 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12566 use "sselog1" type instead of "sselog". Remove explicit "memory"
12568 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12569 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12571 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12572 "sselog1" type instead of "sselog". Remove explicit "memory"
12574 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12575 use "sselog1" type instead of "sselog". Remove explicit "memory"
12579 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12581 PR middle-end/85414
12582 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12585 2018-04-17 Martin Jambor <mjambor@suse.cz>
12588 * ipa-cp.c (create_specialized_node): Call
12589 expand_all_artificial_thunks if necessary.
12591 2018-04-17 Martin Liska <mliska@suse.cz>
12594 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12595 in message, remote space in between '_G' and '('.
12597 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12600 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12601 avx512f_vmcmp<mode>3<round_saeonly_name>,
12602 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12603 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12604 avx512f_rndscale<mode><round_saeonly_name>,
12605 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12606 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12607 Use %<iptr>2 instead of %2 for -masm=intel.
12608 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12609 avx512f_vcvttss2usi<round_saeonly_name>,
12610 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12612 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12613 avx512f_vcvttsd2usi<round_saeonly_name>,
12614 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12615 Use %q1 instead of %1 for -masm=intel.
12616 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12617 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12618 of %3 for -masm=intel.
12619 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12621 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12623 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12625 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12626 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12628 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12630 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12631 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12632 %g1 and one with %0 and %1.
12633 (avx512er_vmrcp28<mode><round_saeonly_name>,
12634 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12635 %1 for -masm=intel.
12636 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12637 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12638 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12639 of %0 and %{%4%} for -masm=intel.
12640 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12641 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12642 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12643 order of %0 and %{%5%}%{z%} for -masm=intel.
12645 2018-04-17 Jan Hubicka <jh@suse.cz>
12648 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12650 2018-04-17 Martin Liska <mliska@suse.cz>
12653 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12654 for target_clone error message. Make default implementation
12655 clone to be a local declaration.
12656 (separate_attrs): Add new argument and check for an empty
12658 (expand_target_clones): Handle it.
12659 (ipa_target_clone): Make redirection just for target_clones
12662 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12663 Tom de Vries <tom@codesourcery.com>
12665 PR middle-end/84955
12666 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12667 tiled basic blocks without omp continue statements.
12669 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12672 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12673 vec_extract expression as having side effects to make sure it gets
12676 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12679 * config/i386/i386.c (get_builtin_code_for_version): Check
12682 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12685 * config.gcc: Support "skylake".
12686 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12688 * config/i386/i386.c (m_SKYLAKE): Define.
12689 (processor_target_table): Add "skylake".
12690 (ix86_option_override_internal): Add "skylake".
12691 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12692 PROCESSOR_CANNONLAKE.
12693 (get_builtin_code_for_version): Fix priority for
12694 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12695 PROCESSOR_SKYLAKE-AVX512.
12696 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12697 (processor_type): Add PROCESSOR_SKYLAKE.
12699 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12700 Jason Merrill <jason@redhat.com>
12703 * convert.c (convert_to_integer_1): Use direct recursion for
12704 enumeral types and types with a precision less than the number
12705 of bits in their mode.
12707 2018-04-16 Julia Koval <julia.koval@intel.com>
12710 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12711 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12713 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12716 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12717 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12718 and -mno-direct-move.
12720 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12723 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12724 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12725 Ensure negative shifts result in {0}.
12727 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12729 PR rtl-optimization/79916
12730 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12731 regs (if any) to define how to gnerate SD moves when LRA is in
12734 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12736 PR rtl-optimization/85393
12737 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12738 * except.c (expand_dw2_landing_pad_for_region): Make static.
12739 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12740 a label and unconditional jump to old_bb, rather than
12741 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12744 PR rtl-optimization/85376
12745 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12746 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12747 instead of a specific value.
12749 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12750 Bin Cheng <bin.cheng@arm.com>
12752 PR tree-optimization/82965
12753 PR tree-optimization/83991
12754 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12755 by_profile_only parameter.
12756 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12757 information if the loop was predicted to iterate too many times.
12758 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12760 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12763 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12766 2018-04-13 Martin Liska <mliska@suse.cz>
12767 Jakub Jelinek <jakub@redhat.com>
12769 PR middle-end/81657
12770 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12771 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12772 * builtins.c (expand_builtin_memory_copy_args): Use
12773 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12774 handle dest_addr == pc_rtx.
12776 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12779 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12780 asked to not generate direct moves.
12781 (fix_trunc<mode>si2_stfiwx): Similar.
12782 (fix_trunc<mode>si2_internal): Similar.
12784 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12787 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12788 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12789 lookup if dest in some wider mode is known to be const0_rtx and
12790 if so, record permanent equivalence for it to be ZERO_EXTEND of
12791 the narrower mode destination.
12793 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12795 * lto-streamer-out.c (output_function): Revert 259346.
12796 * omp-expand.c (expand_oacc_for): Likewise.
12798 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12800 PR rtl-optimization/85354
12801 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12802 * sel-sched.c (sel_global_init): ... here.
12804 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12807 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12808 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12809 mode for PE-COFF targets.
12810 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12811 (i386_pe_asm_lto_end): Likewise.
12812 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12813 (TARGET_ASM_LTO_END): Likewise.
12814 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12815 (i386_pe_asm_lto_start): New function.
12816 (i386_pe_asm_lto_end): Likewise.
12818 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12819 Richard Biener <rguenther@suse.de>
12821 PR middle-end/84955
12822 * lto-streamer-out.c (output_function): Fix CFG loop state before
12824 * omp-expand.c (expand_oacc_for): Handle calls to internal
12825 functions like regular functions.
12827 2018-04-12 Richard Biener <rguenther@suse.de>
12830 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12831 for the early LTO debug to properly generate references to it
12832 during DIE emission. Do not re-use that for the skeleton for
12834 (dwarf2out_early_finish): Likewise.
12836 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12839 * config/i386/sse.md
12840 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12841 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12842 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12843 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12844 and output is a reg, avoid creating invalid lowpart subreg, but
12845 instead split into a 512-bit move. Don't split if not AVX512VL,
12846 input is xmm16+ reg and output is a mem.
12847 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12848 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12849 xmm16+ reg and output is a mem.
12851 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12853 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12854 also for flag_dwarf2_cfi_asm.
12856 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12858 PR rtl-optimization/85342
12859 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12860 a bool scalar var inside of the loop instead. Don't try to update
12861 recog_data.operand after failed apply_change_group.
12863 2018-04-12 Tom de Vries <tom@codesourcery.com>
12866 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12867 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12868 array with flexible array member as array without given dimension.
12869 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12870 argument for undefined param to true.
12872 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12875 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12876 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12877 from PowerPC section.
12878 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12879 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12880 help text that is too long.
12881 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12882 help text that is too long.
12883 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12884 help text that is too long.
12886 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12888 * config/alpha/alpha.md (stack_probe_internal): Rename
12889 from "probe_stack". Update all callers.
12891 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12893 PR rtl-optimization/84566
12894 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12895 sched_macro_fuse_insns.
12897 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12900 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12901 (compute_block_dependences): ... from here.
12903 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12905 PR tree-optimization/85331
12906 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12907 from int to HOST_WIDE_INT.
12909 2018-04-11 Martin Jambor <mjambor@suse.cz>
12912 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12913 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12914 not the same as the source val.
12915 (cgraph_edge_brings_value_p): New parameter.
12916 (gather_edges_for_value): Pass destination value to
12917 cgraph_edge_brings_value_p.
12918 (perhaps_add_new_callers): Likewise.
12919 (get_info_about_necessary_edges): Likewise and exclude values brought
12920 only by self-recursive edges.
12921 (create_specialized_node): Redirect only clones of self-calling edges.
12922 (+self_recursive_pass_through_p): New function.
12923 (find_more_scalar_values_for_callers_subset): Use it.
12924 (find_aggregate_values_for_callers_subset): Likewise.
12925 (known_aggs_to_agg_replacement_list): Removed.
12926 (decide_whether_version_node): Re-calculate known constants for all
12927 remaining context clones.
12929 2018-04-11 Richard Biener <rguenther@suse.de>
12932 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12933 from early DWARF output.
12934 (dwarf2out_early_finish): Output line info unconditionally into
12935 early DWARF and add reference to it.
12937 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12940 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12941 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12942 other than V2DFmode using iptr mode attribute.
12943 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12945 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12947 PR rtl-optimization/84659
12948 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12950 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12953 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12955 (output_loc_list): Pass address of a dummy size variable even in the
12956 locview handling loop.
12957 (index_location_lists): Add comment on why skip_loc_list_entry can't
12960 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12963 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12966 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12969 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12970 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12971 and -mstring-compare-inline-limit.
12973 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12976 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12977 for stack clash protection in a register whenever we need it to be in
12980 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12982 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12983 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12985 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12988 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12990 (mlong-double-): Ditto.
12991 * config/rs6000/sysv4.opt (msdata=): Ditto.
12992 (mtls-size=): Ditto.
12994 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12996 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12997 erroneous entries for
12998 "vector int vec_ldl (int, long int *)", and
12999 "vector unsigned int vec_ldl (int, unsigned long int *)".
13000 Add comments and entries for
13001 "vector bool char vec_ldl (int, bool char *)",
13002 "vector bool short vec_ldl (int, bool short *)",
13003 "vector bool int vec_ldl (int, bool int *)",
13004 "vector bool long long vec_ldl (int, bool long long *)",
13005 "vector pixel vec_ldl (int, pixel *)",
13006 "vector long long vec_ldl (int, long long *)",
13007 "vector unsigned long long vec_ldl (int, unsigned long long *)".
13008 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
13009 type tree bool_long_long_type_node and correct definition of
13010 bool_V2DI_type_node to make reference to this new type tree.
13011 (rs6000_mangle_type): Replace erroneous reference to
13012 bool_long_type_node with bool_long_long_type_node.
13013 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
13014 comments to emphasize sign distinctions for char and int types and
13015 replace RS6000_BTI_bool_long constant with
13016 RS6000_BTI_bool_long_long constant. Also add comment to restrict
13017 use of RS6000_BTI_pixel.
13018 (bool_long_type_node): Remove this macro definition.
13019 (bool_long_long_type_node): New macro definition
13021 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13023 PR rtl-optimization/85300
13024 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
13025 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
13026 simplify_unary_operation fails.
13028 2018-04-10 Martin Liska <mliska@suse.cz>
13030 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
13031 cgraph_edge and ipa_ref.
13033 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13037 * config/i386/sse.md
13038 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
13039 computation of the VEC_MERGE selector from mask.
13040 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
13041 Fix decoding of the VEC_MERGE selector into mask.
13043 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13045 PR tree-optimization/85286
13046 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13048 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13050 * final.c (final_1): Set insn_last_address as well as
13051 insn_current_address.
13053 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13056 * explow.c (emit_stack_probe): Call validize_mem on memory location
13057 before passing it to gen_probe_stack. Create address operand and
13058 legitimize it for the probe_stack_address case.
13060 2018-04-09 Jan Hubicka <jh@suse.cz>
13063 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
13064 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
13065 * tree.c (free_lang_data_in_type): Fix handling of binfos;
13067 (free_lang_data): Rebuild type inheritance graph.
13069 2018-04-09 Martin Sebor <msebor@redhat.com>
13071 * invoke.texi (-finline-small-functions): Mention other optimization
13073 (-findirect-inlining, -fpartial-inlining): Same.
13074 (-finline-functions-called-once): Same.
13075 (-freorder-blocks-and-partition): Same.
13077 2018-04-09 Jan Hubicka <jh@suse.cz>
13080 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
13081 jumps; choose last target that matches the criteria (i.e.
13082 no partition changes for non-crossing jumps).
13083 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
13084 support for redirecting crossing jumps to non-crossing.
13086 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
13088 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
13089 also for naked functions.
13091 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
13093 * config/arc/arc.md (add_shift): New pattern.
13094 (add_shift2): Likewise.
13095 (sub_shift): Likewise.
13096 (sub_shift_cmp0_noout): Likewise.
13097 (compare_si_ashiftsi): Likewise.
13098 (xbfu_cmp0_noout): New combine pattern.
13099 (xbfu_cmp0"): Likewise.
13100 (movsi_set_cc_insn): Place the predicable variant first.
13101 (commutative_binary_cmp0_noout): Remove clobber.
13102 (commutative_binary_cmp0): New pattern.
13103 (noncommutative_binary_cmp0): Likewise.
13104 (noncommutative_binary_cmp0_noout): Likewise.
13105 (noncommutative_binary_comparison_result_used): Removed.
13106 (rsub_cmp0): New pattern.
13107 (rsub_cmp0_noout): Likewise.
13108 (extzvsi): Changed, keep only meaningful variants.
13109 (SQH, SEZ): New iterators.
13110 (SQH_postfix): New mode attribute.
13111 (SEZ_prefix): New code attribute.
13112 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
13113 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13114 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13115 of numerical value.
13116 (noncommutative_operator): Check the availability of barrel
13119 2018-04-09 Richard Biener <rguenther@suse.de>
13121 PR tree-optimization/85284
13122 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13123 Only use the niter constraining form of simple_iv when the exit
13124 is always executed.
13126 2018-04-09 Tom de Vries <tom@codesourcery.com>
13129 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13130 (define_expand "*memory_barrier"): New define_expand.
13131 (define_insn "memory_barrier"): New insn.
13133 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13135 PR rtl-optimization/80463
13136 PR rtl-optimization/83972
13137 PR rtl-optimization/83480
13139 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13140 correct producer for the insn.
13141 (tidy_control_flow): Fixup seqnos in case of debug insns.
13143 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13145 PR rtl-optimization/83913
13147 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13148 different sched-times when merging exprs.
13150 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13152 PR rtl-optimization/83962
13154 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13155 tidy_fallthru_edge and tidy_control_flow.
13157 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13159 PR rtl-optimization/83530
13161 * sel-sched.c (force_next_insn): New global variable.
13162 (remove_insn_for_debug): When force_next_insn is true, also leave only
13163 next insn in the ready list.
13164 (sel_sched_region): When the region wasn't scheduled, make another pass
13165 over it with force_next_insn set to 1.
13167 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13169 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13171 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13172 for interrupt control.
13173 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13174 functions for interrupt control.
13175 * config/nds32/nds32-intrinsic.md: Likewise.
13176 * config/nds32/nds32_intrinsic.h: Likewise.
13177 * config/nds32/nds32.h (nds32_builtins): Likewise.
13179 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13181 * config/nds32/nds32.c (nds32_init_machine_status,
13182 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13183 strict_aligned_p field.
13184 (nds32_expand_to_rtl_hook): New function.
13185 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13186 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13188 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13189 Chung-Ju Wu <jasonwucj@gmail.com>
13191 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13192 * config/nds32/nds32-n7.md: New file.
13193 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13194 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13196 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13197 * config/nds32/nds32.md (pipeline_model): Add n7.
13198 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13199 * config/nds32/pipelines.md: Include n7 settings.
13201 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13202 Chung-Ju Wu <jasonwucj@gmail.com>
13204 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13205 * config/nds32/nds32-e8.md: New file.
13206 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13207 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13209 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13210 * config/nds32/nds32.md (pipeline_model): Add e8.
13211 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13212 * config/nds32/pipelines.md: Include e8 settings.
13214 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13215 Chung-Ju Wu <jasonwucj@gmail.com>
13217 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13218 * config/nds32/nds32-n8.md: New file.
13219 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13220 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13222 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13223 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13224 * config/nds32/nds32.md (pipeline_model): Add n8.
13225 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13226 * config/nds32/pipelines.md: Include n8 settings.
13228 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13229 Chung-Ju Wu <jasonwucj@gmail.com>
13231 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13232 * config/nds32/nds32-n9-2r1w.md: New file.
13233 * config/nds32/nds32-n9-3r2w.md: New file.
13234 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13235 nds32_register_ports): New or modify for cpu n9.
13236 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13238 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13239 * config/nds32/nds32-utils.c: New file.
13240 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13241 TARGET_MUL_SLOW): Define.
13242 * config/nds32/nds32.md (pipeline_model): New attribute.
13243 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13244 New options that support cpu n9.
13245 * config/nds32/pipelines.md: Include n9 settings.
13246 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13248 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13250 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13251 information if necessary.
13252 (output_cond_branch_compare_zero): Likewise.
13253 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13254 (nds32_target_alignment): Refine for alignment.
13255 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13256 (FUNCTION_BOUNDARY): Modify.
13257 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13259 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13261 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13263 * config/nds32/constants.md (unspec_volatile_element): Add values for
13264 TLB operation and data prefetch.
13265 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13266 functions for TLB operation and data prefetch.
13267 * config/nds32/nds32-intrinsic.md: Likewise.
13268 * config/nds32/nds32_intrinsic.h: Likewise.
13269 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13270 (nds32_print_operand): Likewise.
13271 * config/nds32/nds32.h (nds32_builtins): Likewise.
13273 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13274 Andrew Pinski <pinsika@gcc.gnu.org>
13276 PR middle-end/82976
13277 * match.pd: Use constant_boolean_node of correct type instead of
13278 boolean_true_node or boolean_false_node for simplifying
13279 pointer comparisons to zero.
13281 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13283 PR tree-optimization/80021
13284 * tree.c (verify_type_variant): Make error call in verify_variant_match
13285 translatable and remove final full stop.
13287 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13289 * config/nds32/constants.md (unspec_volatile_element): Add
13290 UNSPEC_VOLATILE_EH_RETURN.
13291 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13292 nds32_output_stack_pop): Support dwarf exception handling process.
13293 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13294 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13295 exception handling process.
13296 (nds32_compute_stack_frame): Likewise.
13297 (nds32_return_addr_rtx): Likewise.
13298 (nds32_initial_elimination_offset): Likewise.
13299 (nds32_expand_prologue): Likewise.
13300 (nds32_expand_epilogue): Likewise.
13301 (nds32_dynamic_chain_address): New function.
13302 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13303 exception handling.
13304 (DYNAMIC_CHAIN_ADDRESS): Define.
13305 (EH_RETURN_DATA_REGNO): Define.
13306 (EH_RETURN_STACKADJ_RTX): Define.
13307 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13308 patterns for dwarf exception handling.
13310 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13312 * config/nds32/nds32.h: Clean up obsolete macros.
13314 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13316 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13317 Add enum values for particular instructions.
13318 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13319 particular intrinsic functions.
13320 * config/nds32/nds32-intrinsic.md: Likewise.
13321 * config/nds32/nds32_intrinsic.h: Likewise.
13322 * config/nds32/nds32.h (nds32_builtins): Likewise.
13323 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13324 (btst, ave): New patterns for particular instructions.
13326 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13328 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13329 Add enum values for atomic load/store and memory sync.
13330 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13332 * config/nds32/nds32-intrinsic.md: Likewise.
13333 * config/nds32/nds32_intrinsic.h: Likewise.
13334 * config/nds32/nds32.h (nds32_builtins): Likewise.
13336 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13338 PR tree-optimization/85257
13339 * fold-const.c (native_encode_vector): If not all elts could fit
13340 and off is -1, return 0 rather than offset.
13341 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13342 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13343 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13344 adjust buffer in native_interpret_expr call.
13346 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13348 * config/nds32/constants.md (unspec_volatile_element): Add cache
13349 control enum values.
13350 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13351 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13352 * config/nds32/nds32.c (nds32_cctl_names): New.
13353 (nds32_print_operand): Handle cache control register names.
13354 * config/nds32/nds32.h (nds32_builtins): New enum values.
13355 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13357 * config/nds32/nds32.md (type): Add mmu.
13358 * config/nds32/pipelines.md (simple_insn): Add mmu.
13360 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13362 * config/nds32/nds32.md (type): Remove call.
13363 * config/nds32/pipelines.md (simple_insn): Likewise.
13365 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13367 * config/nds32/constants.md (unspec_volatile_element): Add
13368 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13369 UNSPEC_VOLATILE_FMFCFG.
13370 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13371 description for fmfcfg and fmfcsr.
13372 (bdesc_1arg): Add fmtcsr.
13373 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13374 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13375 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13376 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13377 unspec_fmfcfg): New patterns.
13378 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13379 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13380 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13381 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13382 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13383 __nds32__fmfcfg): Define.
13385 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13387 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13388 intrinsic register names.
13389 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13390 intrinsic register enum values and macros.
13392 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13394 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13395 for load/store addressing form.
13396 (nds32_print_operand_address): Likewise.
13398 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13401 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13402 based on LABEL_REF. Remove useless assertion.
13403 (pic_address_needs_scratch): Fix formatting.
13404 (sparc_legitimize_pic_address): Minor tweaks.
13405 (sparc_delegitimize_address): Adjust assertion accordingly.
13406 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13407 into symbolic_operand.
13408 (movsi_high_pic_label_ref): Likewise.
13409 (movsi_lo_sum_pic_label_ref): Likewise.
13410 (movdi_pic_label_ref): Likewise.
13411 (movdi_high_pic_label_ref): Likewise.
13412 (movdi_lo_sum_pic_label_ref): Likewise.
13414 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13416 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13417 custom LIB_SPEC setup.
13419 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13420 Kito Cheng <kito.cheng@gmail.com>
13422 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13423 * config/riscv/freebsd.h: New.
13425 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13427 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13428 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13431 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13432 Kito Cheng <kito.cheng@gmail.com>
13434 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13435 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13436 * config/nds32/nds32-protos.h (nds32_output_call,
13437 nds32_output_return): Declare.
13438 * config/nds32/nds32.md: Refine all the call and return patterns.
13440 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13443 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13444 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13446 PR rtl-optimization/84872
13447 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13448 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13449 EDGE_CROSSING edge.
13451 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13453 * expr.c (copy_blkmode_to_reg): Revert 254862.
13454 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13456 2018-04-06 Richard Biener <rguenther@suse.de>
13458 PR middle-end/85244
13459 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13460 after seeing a component reference with an adjacent field. Treat
13461 refs to arrays at struct end of external decls similar to
13462 refs to unconstrained commons.
13464 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13467 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13468 look through SAVE_EXPRs with non-side-effects argument. Adjust
13470 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13473 2018-04-06 Richard Biener <rguenther@suse.de>
13475 PR middle-end/85180
13476 * alias.c (find_base_term): New wrapper around find_base_term
13477 unwinding CSELIB_VAL_PTR changes.
13478 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13481 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13483 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13485 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13486 constant definitions.
13487 ("nop"): lr 0,0 -> nopr r0
13488 ("nop_lr0", "nop_lr1"): New insn definitions.
13490 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13492 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13493 NDS32_V3PUSH_AVAILABLE_P macro.
13495 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13496 Chung-Ju Wu <jasonwucj@gmail.com>
13498 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13499 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13500 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13501 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13502 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13503 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13504 * config/nds32/constraints.md: New constraints and checking for hard
13505 float configuration.
13506 * config/nds32/iterators.md: New mode iterator and attribute for hard
13507 float configuration.
13508 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13510 * config/nds32/nds32-fpu.md: New file.
13511 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13512 deal with hard float code generation.
13513 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13515 (abi_type, float_reg_number): New enum type.
13516 * config/nds32/nds32-predicates.c: New predicates for hard float.
13517 * config/nds32/nds32-protos.h: Declare functions for hard float.
13518 * config/nds32/nds32.c: Implementation for hard float configuration.
13519 * config/nds32/nds32.h: Definitions for hard float configuration.
13520 * config/nds32/nds32.md: Include hard float machine description and
13521 modify patterns for hard float configuration.
13522 * config/nds32/nds32.opt: New options for hard float configuration.
13523 * config/nds32/predicates.md: New predicates for hard float
13526 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13528 * common/config/nds32/nds32-common.c
13529 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13531 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13533 PR middle-end/85195
13534 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13535 CONSTRUCTOR_ELT (ctor, ...)->value.
13537 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13540 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13542 2018-04-05 Tom de Vries <tom@codesourcery.com>
13545 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13548 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13549 Kito Cheng <kito.cheng@gmail.com>
13551 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13552 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13553 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13554 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13556 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13557 Kito Cheng <kito.cheng@gmail.com>
13559 * config/nds32/constraints.md (Ufe): New memory constraint.
13560 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13561 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13562 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13564 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13565 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13567 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13569 * config/nds32/nds32.md: Use optimize_size in the condition for
13570 alu-shift instructions.
13572 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13574 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13576 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13578 * config/nds32/nds32.md (negsi2): Refine pattern.
13580 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13581 Chung-Ju Wu <jasonwucj@gmail.com>
13583 * config/nds32/iterators.md (shift_rotate): New code iterator.
13584 (shift): New code attribute.
13585 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13586 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13587 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13588 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13589 bit-wise operations.
13590 (andsi3, *andsi3): Ditto.
13591 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13592 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13593 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13594 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13595 nds32_ior_operand, nds32_xor_operand): New predicates.
13597 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13599 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13600 (addsi3, subsi3): ... this.
13602 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13604 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13606 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13608 * config/nds32/nds32.md: Adjust indention.
13610 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13612 * config/nds32/nds32.md (feature): New attribute.
13614 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13616 * config/nds32/nds32.md (subtype): New attribute.
13618 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13621 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13622 expansion to perform a bitwise AND of the argument followed by a
13623 boolean negation of the result.
13625 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13627 PR rtl-optimization/84878
13628 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13629 the basic block. Assert the use reference is not artificial and that
13630 it has an associated insn.
13632 2018-04-04 Michael Matz <matz@suse.de>
13634 * builtins.c (compute_objsize): Pass correct operand
13635 to array_at_struct_end_p.
13637 2018-04-04 Richard Biener <rguenther@suse.de>
13640 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13641 from contexts for DINFO_LEVEL_TERSE and below.
13643 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13645 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13646 resiter_operand condition.
13647 * config/nds32/nds32.md (*move<mode>): Ditto.
13649 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13650 Monk Chiang <sh.chiang04@gmail.com>
13652 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13654 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13656 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13658 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13659 Kito Cheng <kito.cheng@gmail.com>
13661 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13662 nds32_cond_code_str, output_cond_branch,
13663 output_cond_branch_compare_zero, nds32_expand_cbranch,
13664 nds32_expand_cstore, nds32_expand_movcc,
13665 nds32_output_cbranchsi4_equality_zero,
13666 nds32_output_cbranchsi4_equality_reg,
13667 nds32_output_cbranchsi4_equality_reg_or_const_int,
13668 nds32_output_cbranchsi4_greater_less_zero: New functions.
13669 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13670 nds32_expand_cstore, nds32_expand_movcc,
13671 nds32_output_cbranchsi4_equality_zero,
13672 nds32_output_cbranchsi4_equality_reg,
13673 nds32_output_cbranchsi4_equality_reg_or_const_int,
13674 nds32_output_cbranchsi4_greater_less_zero): Declare.
13675 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13676 nds32_rimm11s_operand): New predicates.
13677 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13678 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13681 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13683 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13684 * config/nds32/nds32.md: Ditto.
13685 * config/nds32/pipelines.md: Ditto.
13687 2018-04-04 Richard Biener <rguenther@suse.de>
13689 PR tree-optimization/85168
13690 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13691 propagating abnormals.
13693 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13695 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13697 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13698 Kito Cheng <kito.cheng@gmail.com>
13700 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13701 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13702 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13703 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13704 * config/nds32/nds32.md (sibcall_internal): New.
13705 (sibcall_register): Remove.
13706 (sibcall_immediate): Remove.
13707 (sibcall_value_internal): New.
13708 (sibcall_value_register): Remove.
13709 (sibcall_value_immediate): Remove.
13710 * config/nds32/predicates.md (nds32_general_register_operand): New.
13711 (nds32_call_address_operand): New.
13713 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13715 PR rtl-optimization/85167
13716 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13717 bb_defs if *split_p, instead preinitialize it to NULL.
13719 PR tree-optimization/85156
13720 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13721 evaluating the argument multiple times.
13723 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13725 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13727 (_mm_cvtpd_ps): Likewise.
13728 (_mm_cvttpd_epi32): Likewise.
13729 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13730 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13731 vector, pixel, and bool following altivec.h include.
13733 2018-04-03 Martin Sebor <msebor@redhat.com>
13735 * doc/extend.texi (Common Function Attributes): Clarify.
13736 (const attribute): Likewise.
13737 (pure attribute): Likewise.
13739 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13742 * config/i386/i386.c (ix86_expand_vector_set): Use
13743 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13745 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13747 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13748 instructions when changing rounding bits to preserve precision bits
13749 in the x87 control word.
13751 2018-04-03 Martin Liska <mliska@suse.cz>
13753 PR tree-optimization/82491
13754 * rtl.h (strip_offset_and_add): Replace += suboffset with
13755 poly_uint64 () + suboffset.
13757 2018-03-29 Martin Liska <mliska@suse.cz>
13758 Martin Jambor <mjambor@suse.cz>
13761 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13762 param_type is not an integral or pointer type.
13764 2018-04-03 Richard Biener <rguenther@suse.de>
13766 * sese.h (recompute_all_dominators): Remove.
13768 2018-04-02 Martin Sebor <msebor@redhat.com>
13770 * doc/invoke.texi (-Wrestrict): Fix typos.
13772 2018-04-02 Jim Wilson <jimw@sifive.com>
13774 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13775 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13776 (<optab>di3, <optab>si3_extend): Likewise.
13777 (<optab>si3_mask, <optab>si3_mask_1): New.
13778 (<optab>di3_mask, <optab>di3_mask_1): New.
13779 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13780 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13781 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13783 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13785 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13788 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13790 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13791 (nds32_canonicalize_comparison): New function.
13793 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13794 Kito Cheng <kito.cheng@gmail.com>
13795 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13797 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13798 * config/nds32/constants.md (unspec_volatile_element): Add
13799 UNSPEC_VOLATILE_RELAX_GROUP.
13800 * config/nds32/nds32-relax-opt.c: New file.
13801 * config/nds32/nds32-predicates.c
13802 (nds32_symbol_load_store_p): New function.
13803 * config/nds32/nds32-protos.h
13804 (nds32_symbol_load_store_p): Declare function.
13805 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13806 * config/nds32/nds32.c
13807 (nds32_register_pass): New function to register pass.
13808 (nds32_register_passes): New function to register passes.
13809 * config/nds32/nds32.md (relax_group): New pattern.
13810 * config/nds32/nds32.opt (mrelax-hint): New option.
13811 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13813 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13815 * config/nds32/t-nds32: Modify files dependency.
13817 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13819 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13820 (PROFILE_HOOK): Define its implementation.
13822 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13824 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13825 type and 32-bit size.
13827 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13829 PR middle-end/85090
13830 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13831 (V_128_256): New mode iterator.
13832 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13833 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13834 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13836 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13837 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13839 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13842 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13843 NaN inputs correctly.
13845 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13848 * config/rs6000/vsx.md (??r): New mode attribute.
13849 (*vsx_mov<mode>_64bit): Use it.
13850 (*vsx_mov<mode>_32bit): Likewise.
13852 2018-03-30 Martin Sebor <msebor@redhat.com>
13854 PR tree-optimization/84818
13855 * builtins.c (check_access): Use warning_n.
13857 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13860 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13862 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13865 2018-03-30 Julia Koval <julia.koval@intel.com>
13868 * x86-tune.def (movx, partial_reg_dependency): Enable for
13871 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13873 PR inline-asm/84985
13874 * lra-constraints.c (process_alt_operands): Move setting
13875 this_alternative_matches below.
13877 2018-03-29 Martin Liska <mliska@suse.cz>
13880 * doc/invoke.texi: Document how LTO works with debug info.
13881 Describe auto-load support of binutils. Mention 'x86-64'
13882 as valid option value of -march option.
13884 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13886 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13889 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13890 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13891 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13894 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13897 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13898 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13899 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13900 for RS6000_BTM_POWERPC64.
13901 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13902 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13903 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13906 (DIVDEU): Likewise.
13908 2018-03-28 Carl Love <cel@us.ibm.com>
13911 2017-09-27 Carl Love <cel@us.ibm.com>
13913 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13914 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13915 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13918 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13920 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13921 instead of __vector bool.
13922 (_mm_max_pu8): Likewise.
13923 (_mm_min_pi16): Likewise.
13925 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13928 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13929 (DIVWEUO): Likewise.
13930 (DIVDEO): Likewise.
13931 (DIVDEUO): Likewise.
13932 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13933 DIVWEUO and DIVDEUO.
13934 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13935 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13936 (div_extend): Likewise.
13937 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13939 (__builtin_divweuo): Likewise.
13940 (__builtin_divdeo): Likewise.
13941 (__builtin_divdeuo): Likewise.
13943 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13946 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13947 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13949 PR tree-optimization/82004
13950 * gimple-match-head.c (optimize_pow_to_exp): New function.
13951 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13952 Don't fold to exp if optimize_pow_to_exp is false.
13954 2018-03-28 Martin Liska <mliska@suse.cz>
13957 * calls.c (initialize_argument_information): Fix trailing space.
13958 * common.opt: Fix typo and provide better explanation for
13959 -fsanitize-coverage option.
13960 * config/i386/i386.opt: Fix typo.
13962 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13963 Martin Liska <mliska@suse.cz>
13966 * gimplify.c (asan_poison_variable): Don't do the check for
13967 gimplify_omp_ctxp here.
13968 (gimplify_decl_expr): Do it here.
13969 (gimplify_target_expr): Likewise.
13971 2018-03-28 Martin Liska <mliska@suse.cz>
13974 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13975 chkp_type_bounds_count if MPX is not enabled.
13977 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13979 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13981 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13984 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13985 function to create the function decl for complex long double
13986 multiply and divide for -mabi=ieeelongdouble.
13987 (init_float128_ieee): Call it.
13989 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13992 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13993 -fcf-protection=branch -mibt.
13994 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13996 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13999 * config/arm/arm.c (arm_valid_symbolic_address): Handle
14000 arm_word_relocations.
14002 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
14005 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
14006 extern array declarations.
14008 2018-03-27 Richard Biener <rguenther@suse.de>
14010 PR middle-end/84067
14011 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
14012 explicit single_use checks.
14014 2018-03-27 Richard Biener <rguenther@suse.de>
14016 PR tree-optimization/85082
14017 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
14020 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14022 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
14023 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
14024 Turn on fasynchronous-unwind-tables and funwind-tables.
14026 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
14029 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
14030 (*bmi_blsr_<mode>_ccz): Ditto.
14032 2018-03-26 Tom de Vries <tom@codesourcery.com>
14034 PR tree-optimization/85063
14035 * omp-general.c (offloading_function_p): New function. Factor out
14037 * omp-offload.c (pass_omp_target_link::gate): ... here.
14038 * omp-general.h (offloading_function_p): Declare.
14039 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
14040 with attribute omp declare target for offloading functions.
14042 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
14044 PR tree-optimization/84005
14045 * tree-data-ref.h (get_base_for_alignment): Declare.
14046 * tree-data-ref.c (get_base_for_alignment_1): New function.
14047 (get_base_for_alignment): Likewise.
14048 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
14049 get_base_for_alignment to find a suitable base object, instead
14050 of always using drb->base_address.
14052 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14054 PR inline-asm/85022
14055 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
14056 known size by default.
14058 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
14060 PR inline-asm/85030
14061 * lra-constraints.c (process_alt_operands): Don't match BLKmode
14062 and non BLKmode operands.
14064 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14067 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
14068 Clean up attributes.
14070 2018-03-23 Richard Biener <rguenther@suse.de>
14073 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
14074 we are going to emit early debug for LTO.
14076 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14078 PR inline-asm/85034
14079 * function.c (match_asm_constraints_1): Don't optimize if input
14080 doesn't satisfy general_operand predicate for output's mode.
14082 PR inline-asm/85022
14083 * alias.c (write_dependence_p): Don't require for x_canonicalized
14084 non-VOIDmode if x has VOIDmode.
14087 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
14088 just don't try to optimize it rather than assert it never happens.
14090 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14092 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
14093 macro expansions for definition of ST_INTERNAL_<mode> and
14094 LD_INTERNAL_<mode> builtins.
14095 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
14097 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
14099 (altivec_expand_st_builtin): Likewise.
14100 (altivec_expand_builtin): Remove calls to deleted functions.
14101 (rs6000_address_for_altivec): Delete this function.
14102 * config/rs6000/vector.md: Remove expands for
14103 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
14105 2018-03-22 Sudakshina Das <sudi.das@arm.com>
14108 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
14109 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
14110 re-computing once computed.
14111 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
14112 (arm_init_machine_status): Initialize
14113 machine->static_chain_stack_bytes.
14115 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14118 * doc/extend.texi: Add four new prototypes for vec_ld.
14119 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14120 definitions for more logical presentation.
14121 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14122 entries for V1TI variants of __builtin_altivec_ld builtin.
14123 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14124 handling of V1TI variant of LVX icode pattern.
14125 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14126 (rs6000_gimple_fold_builtin): Likewise.
14127 (altivec_init_builtins): Add code to define
14128 __builtin_altivec_lvx_v1ti function.
14130 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14132 PR inline-asm/84941
14133 * function.c (match_asm_constraints_1): Don't do the optimization
14134 if input isn't a REG, SUBREG, MEM or constant.
14136 2018-03-22 Tom de Vries <tom@codesourcery.com>
14138 PR tree-optimization/84956
14139 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14140 bb_has_abnormal_pred.
14142 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14145 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14146 DECL_INITIAL (decl) to decl at the end.
14147 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14148 adjust the comment.
14150 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14152 * doc/extend.texi (__builtin_tgmath): Document when complex
14153 integer types are treated as _Complex _Float64.
14155 2018-03-21 Tom de Vries <tom@codesourcery.com>
14157 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14159 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14161 PR tree-optimization/84960
14162 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14163 if it is ENTRY block, move them into single succ of ENTRY in that case.
14165 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14167 PR tree-optimization/84811
14168 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14169 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14170 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14171 (known_subrange_p): Update accordingly. Cast each value involved
14172 in the size comparison, rather than casting the result of the
14175 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14177 PR tree-optimization/84982
14178 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14179 by flipping the least significant bit rather than all bits from
14180 bitpos to bitpos + bitsize - 1.
14182 2018-03-21 Nathan Sidwell <nathan@acm.org>
14184 * doc/extend.texi (Deprecated Features): Remove mention of
14185 long-deleted deprecations.
14187 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14190 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14191 * configure: Regenerate.
14193 2018-03-21 Tom de Vries <tom@codesourcery.com>
14195 PR tree-optimization/83126
14196 * tree-parloops.c (num_phis): New function.
14197 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14199 2018-03-21 Nathan Sidwell <nathan@acm.org>
14201 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14202 mention anon-struct/union members and trailing attributes.
14204 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14206 PR tree-optimization/84969
14207 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14208 builtin memset partitions if they set different rhs values.
14210 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14212 PR rtl-optimization/84989
14213 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14214 VEC_DUPLICATE with scalar result mode.
14216 2018-03-21 Martin Liska <mliska@suse.cz>
14219 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14220 not intended return statement.
14222 2018-03-21 Martin Liska <mliska@suse.cz>
14225 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14226 (chkp_find_bound_slots_1): Limit number of iterations.
14228 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14231 * Minor grammar fixes for x86 options.
14233 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14236 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14237 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14239 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14242 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14243 (altivec_lvx_<mode>_1op): Likewise.
14244 (altivec_stvx_<mode>_2op): Likewise.
14245 (altivec_stvx_<mode>_1op): Likewise.
14246 (altivec_lvx_<VM2:mode>): New define_expand.
14247 (altivec_stvx_<VM2:mode>): Likewise.
14248 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14249 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14250 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14251 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14252 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14253 (rs6000_gen_lvx): Likewise.
14254 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14255 (altivec_expand_stv_builtin): Likewise.
14256 (altivec_expand_builtin): Likewise.
14257 * config/rs6000/vector.md: Likewise.
14259 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14262 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14265 2018-03-20 Richard Biener <rguenther@suse.de>
14268 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14269 sign-conversions as zero, fall back to standard scalar_stmt
14272 2018-03-20 Martin Liska <mliska@suse.cz>
14275 * predict.c (rebuild_frequencies): Handle case when we have
14276 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14278 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14281 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14282 flag_section_anchors.
14283 * varasm.c (use_blocks_for_decl_p): Remove hack for
14284 dw2_force_const_mem.
14287 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14289 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14290 be created, use lowpart_subreg of operands[0] rather than operands[0]
14292 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14293 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14294 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14295 and n constraint instead of aarch64_shift_imm_di and Usd.
14296 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14297 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14299 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14302 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14303 to favor GPR over NEON registers.
14304 (<shift>di3_neon): Likewise.
14306 2018-03-20 Tom de Vries <tom@codesourcery.com>
14309 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14310 (nvptx_process_pars): Emit bar.sync asap and alap.
14312 2018-03-20 Tom de Vries <tom@codesourcery.com>
14315 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14316 seen_label if seen_label is already set.
14318 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14321 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14322 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14323 Use 1U instead of 1. Formatting fixes.
14326 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14327 instead of TREE_TYPE (s1) for the return value.
14329 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14331 PR tree-optimization/84946
14332 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14333 bitsize + bitsize in poly_uint64 rather than poly_int64.
14336 * dwarf2asm.c: Include fold-const.c.
14337 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14338 of decl rather than decl itself.
14340 PR rtl-optimization/84643
14341 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14343 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14346 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14347 calling assemble_variable.
14349 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14352 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14353 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14355 2018-03-19 Jim Wilson <jimw@sifive.com>
14358 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14359 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14360 (riscv_first_stack_step): Likewise.
14361 (riscv_option_override): Use STACK_BOUNDARY instead of
14362 MIN_STACK_BOUNDARY.
14363 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14364 MIN_STACK_BOUNDARY.
14365 (BIGGEST_ALIGNMENT): Set to 128.
14366 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14367 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14370 2018-03-19 Richard Biener <rguenther@suse.de>
14372 PR tree-optimization/84933
14373 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14374 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14376 2018-03-19 Richard Biener <rguenther@suse.de>
14378 PR tree-optimization/84859
14379 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14380 (cond_if_else_store_replacement): Perform sinking operation on
14381 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14382 Generalize what a BB with a single eligible store is.
14384 2018-03-19 Richard Biener <rguenther@suse.de>
14386 PR tree-optimization/84929
14387 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14388 chrec_is_positive against non-chrec arg.
14390 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14393 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14395 2018-03-18 Martin Liska <mliska@suse.cz>
14397 PR rtl-optimization/84635
14398 * regrename.c (build_def_use): Use matches_mode only when
14401 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14403 PR tree-optimization/84913
14404 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14405 vectorize chains of COND_EXPRs.
14407 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14409 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14411 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14413 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14415 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14417 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14419 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14420 Kito Cheng <kito.cheng@gmail.com>
14422 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14423 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14424 (nds32_adjust_reg_alloc_order): New function.
14425 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14427 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14429 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14430 nds32_print_operand, nds32_print_operand_address): Use
14431 HOST_WIDE_INT_PRINT_DEC instead.
14433 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14435 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14437 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14440 * config/i386/i386.c (initial_ix86_tune_features,
14441 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14442 unsigned long long.
14443 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14444 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14445 rather than 1u << ix86_tune. Formatting fix.
14446 (ix86_option_override_internal): Change ix86_arch_mask from
14447 unsigned int to unsigned HOST_WIDE_INT, initialize to
14448 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14449 (ix86_function_specific_restore): Likewise.
14451 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14454 * postreload.c (reload_combine_recognize_pattern): Perform
14455 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14456 truncate_int_for_mode the result for the destination's mode.
14459 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14460 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14462 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14465 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14468 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14469 regno_allocno_class_array and sorted_pseudos.
14470 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14471 insns where regno is used.
14473 2018-03-16 Martin Liska <mliska@suse.cz>
14476 * multiple_target.c (create_dispatcher_calls): Redirect
14477 reference in the symbol table.
14479 2018-03-16 Martin Liska <mliska@suse.cz>
14482 * multiple_target.c (create_dispatcher_calls): Redirect also
14485 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14489 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14490 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14491 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14493 2018-03-16 Julia Koval <julia.koval@intel.com>
14495 * doc/invoke.texi (Skylake Server): Add CLWB.
14496 Cannonlake): Remove CLWB.
14498 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14500 PR tree-optimization/84841
14501 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14503 (FLOAT_ONE_CONST_TYPE): Define.
14504 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14505 (sort_by_operand_rank): Put entries with higher constant_type last
14506 rather than first to match comments.
14508 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14510 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14513 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14516 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14517 check and use address of target always.
14519 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14522 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14523 (indirect_thunk_bnd_needed): Likewise.
14524 (indirect_thunks_used): Likewise.
14525 (indirect_thunks_bnd_used): Likewise.
14526 (indirect_return_needed): New.
14527 (indirect_return_bnd_needed): Likewise.
14528 (output_indirect_thunk_function): Add a bool argument for
14530 (output_indirect_thunk_function): Don't generate alias for
14531 function return thunk.
14532 (ix86_code_end): Call output_indirect_thunk_function to generate
14533 function return thunks.
14534 (ix86_output_function_return): Set indirect_return_bnd_needed
14535 and indirect_return_needed instead of indirect_thunk_bnd_needed
14536 and indirect_thunk_needed.
14538 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14540 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14541 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14542 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14544 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14545 Paul Hua <paul.hua.gm@gmail.com>
14548 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14550 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14552 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14553 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14554 resp. SFmode cases.
14556 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14559 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14560 instead of GET_MODE_SIZE when comparing Units.
14562 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14565 * varasm.c (hash_section): Return an unchangeble hash value
14566 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14567 Return !aarch64_can_use_per_function_literal_pools_p ().
14569 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14572 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14573 rather than address of cmode as last argument to prepare_cmp_insn.
14575 2018-03-15 Julia Koval <julia.koval@intel.com>
14577 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14578 F_AVX512VNNI, F_AVX512BITALG): New.
14580 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14583 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14584 insn for floating-point loads and stores.
14586 2018-03-14 Carl Love <cel@us.ibm.com>
14588 * config/rs6000/rs6000-c.c: Add macro definitions for
14589 ALTIVEC_BUILTIN_VEC_PERMXOR.
14590 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14591 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14592 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14593 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14595 * config/doc/extend.texi: Add prototypes for vec_permxor.
14597 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14600 * diagnostic-show-locus.c (class layout_point): Convert m_line
14601 from int to linenum_type.
14602 (line_span::comparator): Use linenum "compare" function when
14603 comparing line numbers.
14604 (test_line_span): New function.
14605 (layout_range::contains_point): Convert param "row" from int to
14607 (layout_range::intersects_line_p): Likewise.
14608 (layout::will_show_line_p): Likewise.
14609 (layout::print_source_line): Likewise.
14610 (layout::should_print_annotation_line_p): Likewise.
14611 (layout::print_annotation_line): Likewise.
14612 (layout::print_leading_fixits): Likewise.
14613 (layout::annotation_line_showed_range_p): Likewise.
14614 (struct line_corrections): Likewise for field m_row.
14615 (line_corrections::line_corrections): Likewise for param "row".
14616 (layout::print_trailing_fixits): Likewise.
14617 (layout::get_state_at_point): Likewise.
14618 (layout::get_x_bound_for_row): Likewise.
14619 (layout::print_line): Likewise.
14620 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14621 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14622 * input.c (selftest::test_linenum_comparisons): New function.
14623 (selftest::input_c_tests): Call it.
14624 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14625 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14626 * selftest.h (ASSERT_GT): New macro.
14627 (ASSERT_GT_AT): New macro.
14628 (ASSERT_LT): New macro.
14629 (ASSERT_LT_AT): New macro.
14631 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14633 PR rtl-optimization/84780
14634 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14636 2018-03-14 Martin Liska <mliska@suse.cz>
14638 * tree.c (record_node_allocation_statistics): Use
14639 get_stats_node_kind.
14640 (get_stats_node_kind): New function extracted from
14641 record_node_allocation_statistics.
14642 (free_node): Use get_stats_node_kind.
14644 2018-03-14 Richard Biener <rguenther@suse.de>
14646 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14647 that the value-set of ANTIC_IN doesn't grow.
14650 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14652 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14653 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14655 2018-03-14 Julia Koval <julia.koval@intel.com>
14657 * config.gcc (icelake-client, icelake-server): New.
14659 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14660 (initial_ix86_arch_features): Ditto.
14661 (PTA_SKYLAKE): Add SGX.
14662 (PTA_ICELAKE): Remove.
14663 (PTA_ICELAKE_CLIENT): New.
14664 (PTA_ICELAKE_SERVER): New.
14665 (ix86_option_override_internal): Split up icelake on icelake client and
14667 (get_builtin_code_for_version): Ditto.
14668 (fold_builtin_cpu): Ditto.
14669 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14670 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14671 * config/i386/i386.h (processor_type): Ditto.
14672 * doc/invoke.texi: Ditto.
14674 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14677 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14678 INTEGER_CST offset, add it together with bitpos / 8 and
14679 sign extend based on POINTER_SIZE.
14683 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14686 * config/i386/constraints.md (Yc): New register constraint.
14687 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14688 Use Yc constraint for alternative 2 of operand 0. Remove
14689 preferred_for_speed attribute.
14691 2018-03-14 Richard Biener <rguenther@suse.de>
14693 PR tree-optimization/84830
14694 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14695 with the old one to avoid oscillations.
14697 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14700 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14702 (assign_by_spills): Return a flag of reload assignment failure.
14703 Do not process the reload assignment failures. Do not spill other
14704 reload pseudos if they has the same reg class. Update n if
14706 (lra_assign): Add a return arg. Set up from the result of
14707 assign_by_spills call.
14708 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14709 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14710 usage_insns if it is not NULL.
14711 (spill_hard_reg_in_range): New function.
14712 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14713 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14714 function prototypes.
14715 (lra_assign): Change prototype.
14716 * lra.c (lra): Add code to deal with fails by splitting hard reg
14719 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14721 * config/riscv/riscv.opt (mrelax): New option.
14722 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14723 "norelax" when riscv_mrelax is disabled.
14724 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14726 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14729 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14730 reassociation for int modes.
14732 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14734 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14735 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14737 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14738 * config/aarch64/aarch64-sve.md
14739 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14740 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14741 (*extend<mode><Vwide>2): Rename to...
14742 (aarch64_sve_extend<mode><Vwide>2): ...this.
14743 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14744 renaming the old pattern to...
14745 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14747 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14748 define_expand, renaming the old pattern to...
14749 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14750 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14751 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14752 account when deciding which SVE instruction the optab should use.
14753 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14755 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14757 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14758 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14759 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14760 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14761 (tlsdesc_small_<mode>): Turn a define_expand and use
14762 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14763 (tlsdesc_small_advsimd_<mode>): ...this.
14764 (tlsdesc_small_sve_<mode>): New pattern.
14766 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14768 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14769 (UNSPEC_UMUL_HIGHPART): New constants.
14770 (MUL_HIGHPART): New int iteraor.
14771 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14772 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14774 (*<su>mul<mode>3_highpart): New define_insn.
14776 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14779 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14782 2018-03-13 Martin Liska <mliska@suse.cz>
14785 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14787 (sem_item_optimizer::~sem_item_optimizer): Release it.
14788 (sem_item_optimizer::merge_classes): Register variable aliases.
14789 (sem_item_optimizer::fixup_pt_set): New function.
14790 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14791 * ipa-icf.h: Declare new variables and functions.
14793 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14795 PR middle-end/84834
14796 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14797 integer_pow2p@2 and test integer_pow2p in condition.
14798 (A < 0 ? C : 0): Similarly for @1.
14800 PR middle-end/84831
14801 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14802 characters starting at p contain '\0' character, don't look beyond
14806 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14807 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14810 * reg-stack.c (change_stack): Change update_end var from int to
14811 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14812 also call set_block_for_insn on the newly added insns and rescan.
14815 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14816 on the last operand.
14819 * tree.c (stabilize_reference_1): Return save_expr (e) for
14820 STATEMENT_LIST even if it doesn't have side-effects.
14822 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14824 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14826 2018-03-12 Renlin Li <renlin.li@arm.com>
14828 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14829 aarch64_output_scalar_simd_mov_immediate.
14831 2018-03-12 Martin Sebor <msebor@redhat.com>
14833 PR tree-optimization/83456
14834 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14835 for perfectly overlapping calls to memcpy.
14836 (gimple_fold_builtin_memory_chk): Same.
14837 (gimple_fold_builtin_strcpy): Handle no-warning.
14838 (gimple_fold_builtin_stxcpy_chk): Same.
14839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14841 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14843 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14844 parameter. Use it for SFmode.
14845 (rs6000_function_arg_advance_1): Adjust.
14846 (rs6000_function_arg): Adjust.
14847 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14849 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14851 PR rtl-optimization/84169
14852 PR rtl-optimization/84780
14853 * combine.c (can_combine_p): Check for a 2-insn combination whether
14854 the destination register is used between the two insns, too.
14856 2018-03-12 Richard Biener <rguenther@suse.de>
14858 PR tree-optimization/84803
14859 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14860 for refs DR analysis didn't process.
14862 2018-03-12 Richard Biener <rguenther@suse.de>
14864 PR tree-optimization/84777
14865 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14866 force-vectorize loops ignore whether we are optimizing for size.
14868 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14870 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14871 (TARGET_MD_ASM_ADJUST): Define.
14873 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14874 Kito Cheng <kito.cheng@gmail.com>
14875 Chung-Ju Wu <jasonwucj@gmail.com>
14877 * config/nds32/nds32.c (nds32_compute_stack_frame,
14878 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14879 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14880 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14881 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14882 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14883 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14884 * config/nds32/nds32.md (prologue, epilogue): Use macro
14885 NDS32_V3PUSH_AVAILABLE_P to do checking.
14887 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14890 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14891 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14892 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14893 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14894 addition of most attributes on !orig_type_die or the attribute not
14895 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14897 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14898 Chung-Ju Wu <jasonwucj@gmail.com>
14900 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14901 __NDS32_VH__ macro.
14902 * config/nds32/nds32.opt (mvh): New option.
14904 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14905 Chung-Ju Wu <jasonwucj@gmail.com>
14907 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14909 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14910 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14913 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14914 Chung-Ju Wu <jasonwucj@gmail.com>
14916 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14918 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14919 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14921 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14922 Kito Cheng <kito.cheng@gmail.com>
14923 Chung-Ju Wu <jasonwucj@gmail.com>
14925 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14926 UNSPEC_FFMISM and UNSPEC_FLMISM.
14927 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14928 for ffb, ffmism and flmism.
14929 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14930 (unspec_ffmism): Ditto.
14931 (unspec_flmism): Ditto.
14932 (nds32_expand_builtin_impl): Check if string extension is available.
14933 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14934 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14936 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14939 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14942 * lra-assigns.c (assign_by_spills): Return a flag of reload
14943 assignment failure. Do not process the reload assignment
14944 failures. Do not spill other reload pseudos if they has the same
14946 (lra_assign): Add a return arg. Set up from the result of
14947 assign_by_spills call.
14948 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14949 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14950 usage_insns if it is not NULL.
14951 (spill_hard_reg_in_range): New function.
14952 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14953 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14954 function prototypes.
14955 (lra_assign): Change prototype.
14956 * lra.c (lra): Add code to deal with fails by splitting hard reg
14959 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14962 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14964 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14967 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14968 (dw_val_node): Add val_symbolic_view.
14969 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14970 (symview_upper_bound): New.
14971 (new_line_info_table): Initialize symviews_since_reset.
14972 (dwarf2out_source_line): Count symviews_since_reset and set
14973 symview_upper_bound.
14974 (dw_val_equal_p): Handle symview.
14975 (add_AT_symview): New.
14976 (print_dw_val): Handle symview.
14977 (attr_checksum, attr_checksum_ordered): Likewise.
14978 (same_dw_val_p, size_of_die): Likewise.
14979 (value_format, output_die): Likewise.
14980 (add_high_low_attributes): Use add_AT_symview for entry_view.
14981 (dwarf2out_finish): Reset symview_upper_bound, clear
14984 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14987 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14988 Add strict argument and use it.
14989 (rs6000_split_multireg_move): Update for new strict argument.
14990 (mem_operand_gpr): Disallow all non-offsettable addresses.
14991 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14993 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14996 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14997 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14998 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
15001 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
15002 decl, use remap_type if we want to use the type.
15004 2018-03-09 Martin Sebor <msebor@redhat.com>
15006 PR tree-optimization/84526
15007 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15009 (builtin_access::generic_overlap): Be prepared to handle non-array
15012 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
15014 PR rtl-optimization/84682
15015 * lra-constraints.c (process_address_1): Check is_address flag
15016 for address constraints.
15017 (process_alt_operands): Likewise.
15018 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
15019 preprocess_constraints.
15020 * recog.h (preprocess_constraints): Add oploc parameter.
15022 * recog.c (preprocess_constraints): Test address_operand for
15023 CT_ADDRESS constraints.
15025 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15028 * lra-assigns.c (assign_by_spills): Return a flag of reload
15029 assignment failure. Do not process the reload assignment
15030 failures. Do not spill other reload pseudos if they has the same
15032 (lra_assign): Add a return arg. Set up from the result of
15033 assign_by_spills call.
15034 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15035 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15036 usage_insns if it is not NULL.
15037 (spill_hard_reg_in_range): New function.
15038 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15039 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15040 function prototypes.
15041 (lra_assign): Change prototype.
15042 * lra.c (lra): Add code to deal with fails by splitting hard reg
15045 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15048 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
15049 Accept complain bool parameter. Only emit errors if it is true.
15050 (arm_parse_cpu_option_name): Likewise.
15051 (arm_target_thumb_only): Adjust callers of the above.
15052 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
15053 prototype to take a default true bool parameter.
15054 (arm_parse_arch_option_name): Likewise.
15056 2018-03-09 David Malcolm <dmalcolm@redhat.com>
15057 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
15061 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
15062 * configure: Regenerate.
15063 * configure.ac ("linker --version-script option"): New.
15064 ("linker soname option"): New.
15066 2018-03-09 Richard Biener <rguenther@suse.de>
15068 PR tree-optimization/84775
15069 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
15070 immediate uses of predicate stmts and mark them modified.
15073 PR tree-optimization/84178
15074 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15076 (version_loop_for_if_conversion): Delay update_ssa call.
15077 (tree_if_conversion): Delay update_ssa until after predicate
15080 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
15083 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
15084 when the function accesses prior frames.
15086 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15089 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
15090 gen_llsym, otherwise call maybe_gen_llsym.
15092 PR inline-asm/84742
15093 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15094 has ',' character inside of it.
15096 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15099 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15100 as clobbering CC_REGNUM.
15102 2018-03-08 Richard Biener <rguenther@suse.de>
15104 PR middle-end/84552
15105 * tree-scalar-evolution.c: Include tree-into-ssa.h.
15106 (follow_copies_to_constant): Do not follow SSA names registered
15109 2018-03-08 Richard Biener <rguenther@suse.de>
15111 PR tree-optimization/84178
15112 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15114 (version_loop_for_if_conversion): Delay update_ssa call.
15115 (tree_if_conversion): Delay update_ssa until after predicate
15118 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15120 PR tree-optimization/84178
15121 * tree-if-conv.c (release_bb_predicate): Remove the
15122 the assertion that the stmts have NULL use_ops.
15123 Discard the statements, asserting that they haven't
15124 yet been added to a BB.
15126 2018-03-08 Richard Biener <rguenther@suse.de>
15128 PR tree-optimization/84746
15129 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15130 (phi_translate): Pass in destination ANTIC_OUT set.
15131 (phi_translate_1): Likewise. For a simplified result lookup
15132 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15133 (phi_translate_set): Adjust.
15134 (do_pre_regular_insertion): Likewise.
15135 (do_pre_partial_partial_insertion): Likewise.
15137 2018-03-08 Martin Liska <mliska@suse.cz>
15139 PR gcov-profile/84735
15140 * doc/gcov.texi: Document usage of profile files.
15141 * gcov-io.h: Document changes in the format.
15143 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15147 * dwarf2out.c (struct dw_line_info_table): Update comments for
15149 (FORCE_RESET_NEXT_VIEW): New.
15150 (FORCE_RESETTING_VIEW_P): New.
15151 (RESETTING_VIEW_P): Check for -1 too.
15152 (ZERO_VIEW_P): Likewise.
15153 (new_line_info_table): Force-reset next view.
15154 (dwarf2out_begin_function): Likewise.
15155 (dwarf2out_source_line): Simplify zero_view_p initialization.
15156 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15157 view directly. Omit view when omitting .loc at line 0.
15159 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15161 PR tree-optimization/84740
15162 * tree-switch-conversion.c (process_switch): Call build_constructors
15163 only if info.phi_count is non-zero.
15165 PR tree-optimization/84739
15166 * tree-tailcall.c (find_tail_calls): Check call arguments against
15167 DECL_ARGUMENTS (current_function_decl) rather than
15168 DECL_ARGUMENTS (func) when checking for tail recursion.
15170 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15172 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15173 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15174 Volker Reichelt's entry and add entries for people that perform
15175 GCC fuzzy testing and report numerous bugs.
15177 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15180 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15181 readonly data in sdata, if that is disabled.
15182 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15183 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15184 -mreadonly-in-sdata option.
15186 2018-03-07 Martin Sebor <msebor@redhat.com>
15188 PR tree-optimization/84468
15189 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15190 basic block when looking for nul assignment.
15192 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15195 * except.h (output_function_exception_table): Adjust prototype.
15196 * except.c (output_function_exception_table): Remove FNNAME parameter
15197 and add SECTION parameter. Ouput one part of the table at a time.
15198 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15199 the first part of the exception table and emit unwind directives.
15200 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15201 (i386_pe_seh_cold_init): Likewise.
15202 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15203 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15204 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15205 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15206 * config/i386/winnt.c: Include except.h.
15207 (struct seh_frame_state): Add reg_offset, after_prologue and
15208 in_cold_section fields.
15209 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15210 (i386_pe_seh_cold_init): New function.
15211 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15212 to seh->in_cold_section.
15213 (seh_emit_push): Record the offset of the push.
15214 (seh_emit_save): Record the offet of the save.
15215 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15216 Test seh->after_prologue to disregard the epilogue.
15217 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15218 (i386_pe_end_cold_function): New function.
15220 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15223 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15224 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15227 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15229 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15230 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15232 PR middle-end/84723
15233 * multiple_target.c: Include tree-inline.h and intl.h.
15234 (expand_target_clones): Diagnose and fail if node->definition and
15235 !tree_versionable_function_p (node->decl).
15237 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15239 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15241 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15242 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15243 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15245 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15248 * combine.c (try_combine): Use reg_or_subregno instead of handling
15249 just paradoxical SUBREGs and REGs.
15251 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15253 * config/arc/arc.c (arc_finalize_pic): Remove function.
15254 (arc_must_save_register): We use single base PIC register, remove
15255 checks to save/restore the PIC register.
15256 (arc_expand_prologue): Likewise.
15257 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15259 (arc_verify_short): Likewise.
15260 (arc_attr_type): Likewise.
15261 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15262 (walk_stores): Likewise.
15263 (arc_address_cost): Make it static.
15264 (arc_verify_short): Likewise.
15265 (branch_dest): Likewise.
15266 (arc_attr_type): Likewise.
15267 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15268 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15269 (arc_final_prescan_insn): Remove inserting the nops due to
15270 hardware hazards. It is done in reorg step.
15271 (insn_length_variant_t): Remove.
15272 (insn_length_parameters_t): Likewise.
15273 (arc_insn_length_parameters): Likewise.
15274 (arc_get_insn_variants): Likewise.
15275 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15277 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15279 PR inline-asm/84683
15280 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15283 PR tree-optimization/84687
15284 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15286 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15288 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15290 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15291 Rename to ppc_speculation_barrier.
15292 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15293 __builtin_ppc_speculation_barrier.
15295 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15298 * combine.c (combine_simplify_rtx): Don't try to simplify if
15299 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15302 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15304 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15305 to 32 bytes when compiling for POWER9.
15307 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15310 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15311 regparm >= 3 with no arg reg available also for calls with
15312 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15315 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15317 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15319 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15322 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15324 2018-03-05 Richard Biener <rguenther@suse.de>
15326 PR tree-optimization/84486
15327 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15328 When inserting a __builtin_assume_aligned call set the LHS
15329 SSA name alignment info accordingly.
15331 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15333 PR tree-optimization/84114
15334 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15335 Avoid reassociation of FLOAT_MODE addition.
15337 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15339 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15340 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15341 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15342 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15343 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15344 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15345 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15347 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15348 __builtin_ia32_wbinvd): New builtins.
15349 (SPECIAL_ARGS2): New.
15350 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15351 (SPECIAL_ARGS2): New.
15352 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15353 (ix86_valid_target_attribute_inner_p): Ditto.
15354 (ix86_init_mmx_sse_builtins): Add special_args2.
15355 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15356 TARGET_WBNOINVD_P): New.
15357 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15358 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15359 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15360 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15361 * config/i386/pconfigintrin.h: New file.
15362 * config/i386/wbnoinvdintrin.h: Ditto.
15363 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15365 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15367 2018-03-05 Richard Biener <rguenther@suse.de>
15369 PR tree-optimization/84670
15370 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15372 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15373 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15374 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15375 doesn't grow if all successors (recursively) were visited at least
15378 2018-03-05 Richard Biener <rguenther@suse.de>
15380 PR tree-optimization/84650
15381 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15382 if executed in the loop pipeline.
15384 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15386 * doc/configfiles.texi (Configuration Files): Move info about
15387 conditionalizing $target-protos.h to...
15388 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15389 differs from $target-protos.h.
15391 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15392 Chung-Ju Wu <jasonwucj@gmail.com>
15394 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15395 * config/nds32/nds32-multiple.md (setmemsi): Define.
15396 * config/nds32/nds32-memory-manipulation.c
15397 (nds32_gen_dup_4_byte_to_word_value): New.
15398 (emit_setmem_word_loop): New.
15399 (emit_setmem_byte_loop): New.
15400 (nds32_expand_setmem_loop): New.
15401 (nds32_expand_setmem_loop_v3m): New.
15402 (nds32_expand_setmem_unroll): New.
15403 (nds32_expand_setmem): New.
15405 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15406 Chung-Ju Wu <jasonwucj@gmail.com>
15408 * config/nds32/nds32-memory-manipulation.c
15409 (nds32_emit_load_store): New.
15410 (nds32_emit_post_inc_load_store): New.
15411 (nds32_emit_mem_move): New.
15412 (nds32_emit_mem_move_block): New.
15413 (nds32_expand_movmemsi_loop_unknown_size): New.
15414 (nds32_expand_movmemsi_loop_known_size): New.
15415 (nds32_expand_movmemsi_loop): New.
15416 (nds32_expand_movmemsi_unroll): New.
15417 (nds32_expand_movmemqi): Rename ...
15418 (nds32_expand_movmemsi): ... to this.
15419 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15420 (movmemsi): ... to this.
15421 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15422 (nds32_expand_movmemsi): ... to this.
15424 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15425 Monk Chiang <sh.chiang04@gmail.com>
15426 Chung-Ju Wu <jasonwucj@gmail.com>
15428 * config/nds32/nds32-protos.h
15429 (nds32_expand_load_multiple): New arguments.
15430 (nds32_expand_store_multiple): Ditto.
15431 (nds32_valid_multiple_load_store): Rename ...
15432 (nds32_valid_multiple_load_store_p): ... to this.
15433 * config/nds32/nds32-memory-manipulation.c
15434 (nds32_expand_load_multiple): Refine implementation.
15435 (nds32_expand_store_multiple): Ditto.
15436 * config/nds32/nds32-multiple.md
15437 (load_multiple): Update nds32_expand_load_multiple interface.
15438 (store_multiple): Update nds32_expand_store_multiple interface.
15439 * config/nds32/nds32-predicates.c
15440 (nds32_valid_multiple_load_store): Rename ...
15441 (nds32_valid_multiple_load_store_p): ... to this and refine
15443 * config/nds32/predicates.md
15444 (nds32_load_multiple_and_update_address_operation): New predicate.
15445 (nds32_store_multiple_and_update_address_operation): New predicate.
15447 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15448 Chung-Ju Wu <jasonwucj@gmail.com>
15450 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15451 (combo): New attribute.
15452 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15454 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15456 * config/nds32/nds32.opt: Change -mcmodel= default value.
15458 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15459 Monk Chiang <sh.chiang04@gmail.com>
15460 Chung-Ju Wu <jasonwucj@gmail.com>
15462 * config/nds32/constants.md (unspec_element): New enum.
15463 * config/nds32/constraints.md (Umw): New constraint.
15464 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15465 * config/nds32/nds32-intrinsic.md: Likewise.
15466 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15467 (nds32_valid_smw_lwm_base_p): New.
15468 (nds32_output_smw_single_word): New.
15469 (nds32_output_lmw_single_word): New.
15470 (nds32_expand_unaligned_load): New.
15471 (nds32_expand_unaligned_store): New.
15472 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15473 (nds32_output_smw_single_word): Declare.
15474 (nds32_output_lmw_single_word): Declare.
15475 (nds32_expand_unaligned_load): Declare.
15476 (nds32_expand_unaligned_store): Declare.
15477 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15478 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15479 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15480 NDS32_BUILTIN_UASTORE_DW.
15481 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15484 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15485 Kito Cheng <kito.cheng@gmail.com>
15486 Chung-Ju Wu <jasonwucj@gmail.com>
15488 * config/nds32/nds32-intrinsic.c
15489 (nds32_expand_builtin_null_ftype_reg): Delete.
15490 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15491 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15492 (nds32_read_argument): New.
15493 (nds32_legitimize_target): Ditto.
15494 (nds32_legitimize_argument): Ditto.
15495 (nds32_check_constant_argument): Ditto.
15496 (nds32_expand_unop_builtin): Ditto.
15497 (nds32_expand_unopimm_builtin): Ditto.
15498 (nds32_expand_binop_builtin): Ditto.
15499 (nds32_builtin_decl_impl): Ditto.
15500 (builtin_description): Ditto.
15501 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15502 (nds32_init_builtins_impl): Ditto.
15503 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15504 (nds32_builtin_decl): New.
15505 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15506 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15508 2018-03-02 Jeff Law <law@redhat.com>
15510 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15511 (redundant_insn, fill_simple_delay_slots): Likewise.
15512 (fill_slots_from_thread): Likewise.
15513 * resource.c (mark_referenced_resources): Likewise.
15514 (mark_set_resources, find_dead_or_set_registers): Likewise.
15516 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15518 * substring-locations.h (format_warning_va): Formatting fix for
15519 ATTRIBUTE_GCC_DIAG.
15520 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15522 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15523 * substring-locations.c: Include intl.h.
15524 (format_warning_va): Turned into small wrapper around
15525 format_warning_n_va, renamed to ...
15526 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15527 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15529 (format_warning_at_substring_n): New function.
15530 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15531 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15532 format_warning_at_substring with just a shorter name instead of
15533 const function pointer.
15534 (fmtwarn_n): New function.
15535 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15536 appropriate, get rid of all the fmtstr temporaries, move conditionals
15537 with G_() wrapped string literals directly into fmtwarn arguments,
15538 cast dir.len to (int), formatting fixes.
15540 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15542 * doc/invoke.texi: Remove "Cilk Plus" references.
15544 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15545 Richard Biener <rguenther@suse.de>
15548 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15549 for error or warning attributes if CALL_FROM_THUNK_P is set.
15552 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15555 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15556 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15559 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15560 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15562 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15563 instead of -1U in last predictors element's probability member.
15565 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15568 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15569 arguments if they are comparable.
15571 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15573 PR tree-optimization/84634
15574 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15575 masks and masked_loop_p with a single loop_masks, making sure it's
15576 null for bb vectorization.
15578 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15580 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15581 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15582 loop->force_vectorize to check whether there is no alias.
15584 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15587 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15589 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15591 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15592 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15593 instead of a loop around prev_real_insn.
15594 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15597 PR inline-asm/84625
15598 * config/i386/i386.c (ix86_print_operand): Use conditional
15599 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15602 2018-03-02 Richard Biener <rguenther@suse.de>
15604 PR tree-optimization/84427
15605 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15606 (bitmap_set_subtract_values): Rewrite to handle multiple
15609 (prune_clobbered_mems): Likewise.
15610 (phi_translate): Take edge instead of pred/phiblock.
15611 (phi_translate_1): Likewise.
15612 (phi_translate_set): Likewise. Insert all translated
15613 exprs for a value into the set, keeping possibly multiple
15614 expressions per value.
15615 (compute_antic_aux): Adjust for phi_translate changes.
15616 When intersecting union the expressions and prune those
15617 not in the final value set, keeping possibly multiple
15618 expressions per value. Do not use value-insertion
15619 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15620 all expressions. Add verification that the value-sets
15621 only shrink during iteration.
15622 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15623 (do_pre_regular_insertion): Likewise.
15624 (do_pre_partial_partial_insertion): Likewise.
15626 2018-03-02 Richard Biener <rguenther@suse.de>
15629 * config/darwin.c (saved_debug_info_level): New static global.
15630 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15631 (darwin_asm_lto_end): Restore debug info generation settings.
15633 2018-03-01 Martin Liska <mliska@suse.cz>
15636 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15637 volatile arguments.
15639 2018-03-01 Richard Biener <rguenther@suse.de>
15642 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15643 type annotation with LTO when debug was disabled at compile-time.
15645 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15647 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15649 (mips_final_postscan_insn): Likewise.
15651 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15653 PR rtl-optimization/84528
15654 * alias.c (init_alias_target): Add commentary.
15655 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15656 a unique base value if the frame pointer is not eliminated
15657 to the stack pointer.
15659 2018-03-01 Tom de Vries <tom@codesourcery.com>
15661 PR rtl-optimization/83327
15662 * lra-int.h (hard_regs_spilled_into): Declare.
15663 * lra.c (hard_regs_spilled_into): Define.
15664 (init_reg_info): Init hard_regs_spilled_into.
15665 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15666 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15667 (process_bb_lives): Handle hard_regs_spilled_into.
15668 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15669 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15671 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15673 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15674 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15675 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15676 * config/rs6000/aix72.h: New file.
15678 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15680 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15681 instead of warning_at with conditional singular and plural messages
15685 * stor-layout.c (update_alignment_for_field): For
15686 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15687 && !DECL_PACKED (field), do the alignment update, just use
15688 only desired_align instead of MAX (type_align, desired_align)
15690 (place_field): Don't do known_align < desired_align handling
15691 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15692 is non-NULL, instead do it after rli->prev_field handling and
15693 only if not within a bitfield word. For DECL_PACKED (field)
15694 use type_align of BITS_PER_UNIT.
15696 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15698 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15699 superfluous parentheses and trailing spaces.
15701 2018-02-28 Richard Biener <rguenther@suse.de>
15703 PR tree-optimization/84584
15704 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15705 SCoPs with fake exit edge.
15707 2018-02-28 Martin Liska <mliska@suse.cz>
15710 * timevar.c (timer::print): Fix format to properly print 100%
15713 2018-02-28 Richard Biener <rguenther@suse.de>
15715 PR middle-end/84607
15716 * genmatch.c (capture_info::walk_match): Do not mark
15717 captured expressions without operands as expr_p given
15718 they act more like predicates and should be subject to
15719 "lost tail" side-effect preserving.
15721 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15723 PR rtl-optimization/81611
15724 * auto-inc-dec.c (attempt_change): Move dead note from
15725 mem_insn if it's the next use of regno
15726 (find_address): Take address use of reg holding
15727 non-incremented value. Add parm to limit search to the named
15729 (merge_in_block): Attempt to use a mem insn that is the next
15730 use of the original regno.
15732 2018-02-27 Martin Sebor <msebor@redhat.com>
15735 * doc/invoke.texi (-Wmissing-attributes): New option.
15736 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15738 2018-02-27 Martin Sebor <msebor@redhat.com>
15740 PR translation/84207
15741 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15742 n argument to unsigned HOST_WIDE_INT.
15743 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15744 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15745 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15746 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15748 2018-02-27 Richard Biener <rguenther@suse.de>
15750 PR tree-optimization/84512
15751 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15752 Do not use the estimate returned from record_stmt_cost for
15753 the scalar iteration cost but sum properly using add_stmt_cost.
15755 2018-02-27 Richard Biener <rguenther@suse.de>
15757 PR tree-optimization/84466
15758 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15759 Adjust last change to less strictly validate use operands.
15761 2018-02-27 Martin Liska <mliska@suse.cz>
15763 PR gcov-profile/84548
15764 * gcov.c (process_file): Allow partial overlap and consider it
15765 also as group functions.
15766 (output_lines): Properly calculate range of lines for a group.
15768 2018-02-27 Martin Liska <mliska@suse.cz>
15770 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15771 'ggc' suffixes. Change first column width.
15772 (timer::print): Fix formatting of the column.
15774 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15776 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15777 preserve inline entry blocks for the sake of debug inline
15778 entry point markers alone.
15779 (remove_unused_locals): Suggest in comments a better place to
15780 force the preservation of inline entry blocks that are
15781 otherwise unused, but do not preserve them.
15783 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15785 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15787 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15790 * config/i386/constraints.md (Bs): Replace
15791 ix86_indirect_branch_register with
15792 TARGET_INDIRECT_BRANCH_REGISTER.
15794 * config/i386/i386.md (indirect_jump): Likewise.
15795 (tablejump): Likewise.
15796 (*sibcall_memory): Likewise.
15797 (*sibcall_value_memory): Likewise.
15798 Peepholes of indirect call and jump via memory: Likewise.
15799 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15800 (*sibcall_value_GOT_32): Likewise.
15801 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15802 (GOT_memory_operand): Likewise.
15803 (call_insn_operand): Likewise.
15804 (sibcall_insn_operand): Likewise.
15805 (GOT32_symbol_operand): Likewise.
15806 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15808 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15810 PR rtl-optimization/83496
15811 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15812 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15813 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15814 redundant insn, if any.
15815 (relax_delay_slots): Likewise.
15816 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15818 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15820 PR tree-optimization/83965
15821 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15822 that grouped statements are part of a reduction chain. Return
15823 true if the statement is not marked as a reduction itself but
15824 is part of a group.
15825 (vect_recog_dot_prod_pattern): Don't check whether the statement
15826 is part of a group here.
15827 (vect_recog_sad_pattern): Likewise.
15828 (vect_recog_widen_sum_pattern): Likewise.
15830 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15833 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15835 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15838 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15840 (ix86_output_indirect_function_return): New prototype.
15841 (ix86_split_simple_return_pop_internal): Likewise.
15842 * config/i386/i386.c (indirect_return_via_cx): New.
15843 (indirect_return_via_cx_bnd): Likewise.
15844 (indirect_thunk_name): Handle return va CX_REG.
15845 (output_indirect_thunk_function): Create alias for
15846 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15847 (ix86_output_indirect_jmp): Remove the bool argument.
15848 (ix86_output_indirect_function_return): New function.
15849 (ix86_split_simple_return_pop_internal): Likewise.
15850 * config/i386/i386.md (*indirect_jump): Don't pass false
15851 to ix86_output_indirect_jmp.
15852 (*tablejump_1): Likewise.
15853 (simple_return_pop_internal): Change it to define_insn_and_split.
15854 Call ix86_split_simple_return_pop_internal to split it for
15855 -mfunction-return=.
15856 (simple_return_indirect_internal): Call
15857 ix86_output_indirect_function_return instead of
15858 ix86_output_indirect_jmp.
15860 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15863 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15864 memset and value initialization afterwards.
15866 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15868 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15870 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15873 * common/config/aarch64/aarch64-common.c
15874 (aarch_option_optimization_table[]): Switch
15875 off fomit-frame-pointer
15877 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15878 Chung-Ju Wu <jasonwucj@gmail.com>
15880 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15882 (store_multiple): Ditto.
15884 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15886 * config.gcc: Add --with-cpu support for nds32 target.
15887 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15888 * config/nds32/nds32.opt: Add -mcpu= option.
15890 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15892 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15893 isel=yes): Warn for these deprecated options.
15895 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15897 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15898 ISA_2_5_MASKS_EMBEDDED.
15900 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15902 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15903 p->max as pointers rather than using iterative_hash_expr.
15905 2018-02-23 Carl Love <cel@us.ibm.com>
15907 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15908 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15910 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15911 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15912 P8V_BUILTIN_VEC_VUNSIGNED2.
15914 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15917 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15918 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15920 * lra-constraints.c (curr_insn_transform): Set up
15921 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15923 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15924 * lra-eliminations.c (spill_pseudos): Ditto.
15925 (process_insn_for_elimination): Ditto.
15926 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15927 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15928 LRA_NON_CLOBBERED_ALT.
15930 2018-02-22 Martin Sebor <msebor@redhat.com>
15932 PR tree-optimization/84480
15933 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15934 to maybe_diag_stxncpy_trunc. Call it.
15935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15936 from gimple_fold_builtin_strcpy. Print inlining stack.
15937 (handle_builtin_stxncpy): Print inlining stack.
15938 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15940 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15943 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15944 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15945 and -fcheck-pointer-bounds are used together.
15946 (indirect_thunk_prefix): New enum.
15947 (indirect_thunk_need_prefix): New function.
15948 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15949 "_nt" instead of "_bnd" for NOTRACK prefix.
15950 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15951 (output_indirect_thunk_function): Likewise.
15953 (ix86_code_end): Update output_indirect_thunk_function calls.
15954 (ix86_output_indirect_branch_via_reg): Replace
15955 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15956 (ix86_output_indirect_branch_via_push): Likewise.
15957 (ix86_output_function_return): Likewise.
15958 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15959 incompatible with -fcf-protection=branch and
15960 -fcheck-pointer-bounds.
15962 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15965 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15966 Change gcc_assert call to output_operand_lossage.
15968 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15970 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15972 2018-02-22 DJ Delorie <dj@redhat.com>
15973 Sebastian Perta <sebastian.perta@renesas.com>
15974 Oleg Endo <olegendo@gcc.gnu.org>
15976 * config/rx/rx.c (rx_rtx_costs): New function.
15977 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15979 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15981 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15983 2018-02-22 Martin Liska <mliska@suse.cz>
15986 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15987 Add "native" as a possible value.
15989 2018-02-22 Martin Liska <mliska@suse.cz>
15992 * config/i386/i386.c (ix86_option_override_internal):
15993 Add "native" as a possible value for -march and -mtune.
15995 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15998 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15999 to all type variants.
16001 PR tree-optimization/84503
16002 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
16003 width as info->bitpos + info->bitsize - start.
16004 (merged_store_group::merge_overlapping): Simplify width computation.
16005 (check_no_overlap): New function.
16006 (imm_store_chain_info::try_coalesce_bswap): Compute expected
16007 start + width and last_order of the group, fail if check_no_overlap
16009 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
16010 to group if check_no_overlap fails.
16012 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16014 * config/rs6000/altivec.md: Delete contraint arguments to
16015 define_expand, define_split, and define_peephole2, and in
16016 define_insn_and_split if always unused.
16017 * config/rs6000/darwin.md: Ditto.
16018 * config/rs6000/dfp.md: Ditto.
16019 * config/rs6000/rs6000.md: Ditto.
16020 * config/rs6000/sync.md: Ditto.
16021 * config/rs6000/vector.md: Ditto.
16022 * config/rs6000/vsx.md: Ditto.
16024 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16026 * config/rs6000/altivec.md: Write output control strings as braced
16027 blocks instead of double-quoted strings.
16028 * config/rs6000/darwin.md: Ditto.
16029 * config/rs6000/rs6000.md: Ditto.
16030 * config/rs6000/vector.md: Ditto.
16031 * config/rs6000/vsx.md: Ditto.
16033 2018-02-21 Jason Merrill <jason@redhat.com>
16035 PR c++/84314 - ICE with templates and fastcall attribute.
16036 * attribs.c (build_type_attribute_qual_variant): Remove assert.
16038 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16040 * ipa-cp.c (determine_versionability): Fix comment typos.
16042 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16045 * ipa-cp.c (determine_versionability): Do not version functions caling
16048 2018-02-21 Martin Liska <mliska@suse.cz>
16051 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
16052 Add "native" as a possible value.
16053 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
16054 the macro when native cpu detection is available.
16056 2018-02-21 Martin Liska <mliska@suse.cz>
16059 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
16060 Add "native" as a possible value.
16061 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
16062 when native cpu detection is available.
16064 2018-02-21 Jakub Jelinek <jakub@redhat.com>
16065 Martin Sebor <msebor@redhat.com>
16067 PR tree-optimization/84478
16068 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
16070 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
16071 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
16072 support which is conservatively correct, for 2 only stay conservative
16073 for maxlen. Formatting and comment capitalization fixes. Add STRICT
16074 argument to the 2 argument get_range_strlen, adjust 6 arg
16075 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
16077 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
16078 (gimple_fold_builtin_strlen): Pass true as last argument to
16081 2018-02-20 Martin Sebor <msebor@redhat.com>
16083 PR middle-end/84095
16084 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
16085 (builtin_memref::set_base_and_offset): Same. Handle inner references.
16086 (builtin_memref::builtin_memref): Factor out parts into
16087 set_base_and_offset and call it.
16089 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16091 PR middle-end/84406
16092 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
16093 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
16094 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
16095 search at the associated MODE_INT.
16097 2018-02-20 Jeff Law <law@redhat.com>
16099 PR middle-end/82123
16100 PR tree-optimization/81592
16101 PR middle-end/79257
16102 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
16103 for range data rather than using global data.
16104 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
16105 range data rather than using global data.
16106 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
16107 pass it to children as needed.
16108 (struct directive::fmtresult): Similarly.
16109 (struct directive::set_width): Similarly.
16110 (struct directive::set_precision): Similarly.
16111 (format_integer, format_directive, parse_directive): Similarly.
16112 (format_none): Accept unnamed vr_values parameter.
16113 (format_percent, format_floating, format_character): Similarly.
16114 (format_string, format_plain): Similarly.
16115 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16116 the EVRP range analyzer for range data rather than using global data.
16117 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16118 gimple-ssa-evrp-analyze.h
16119 (class sprintf_dom_walker): Add after_dom_children member function.
16120 Add evrp_range_analyzer member.
16121 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16122 range analyzer as needed.
16123 (sprintf_dom_walker::after_dom_children): New member function.
16124 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16126 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16127 (evrp_range_analyzer::pop_to_marker): Likewise.
16129 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16131 PR tree-optimization/84419
16132 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16133 with the required type if its current type is compatible but
16136 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16138 PR middle-end/82004
16139 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16140 after vectorization.
16142 2018-02-20 Martin Liska <mliska@suse.cz>
16145 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16146 possible values if we don't have a hint.
16148 2018-02-20 Martin Liska <mliska@suse.cz>
16152 * final.c (shorten_branches): Build align_tab array with one
16154 * opts.c (finish_options): Add alignment option limit check.
16155 (MAX_CODE_ALIGN): Likewise.
16156 (MAX_CODE_ALIGN_VALUE): Likewise.
16157 * doc/invoke.texi: Document maximum allowed option value for
16158 all -falign-* options.
16160 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16163 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16164 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16165 * var-tracking.c (emit_note_insn_var_location): Remove all references
16166 to NOTE_INSN_CALL_ARG_LOCATION.
16167 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16168 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16169 Use copy_rtx_if_shared.
16170 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16171 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16172 (dwarf2out_var_location): Remove handling of
16173 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16175 * final.c (final_scan_insn): Remove all references to
16176 NOTE_INSN_CALL_ARG_LOCATION.
16177 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16178 before dumping final insns.
16179 * except.c (emit_note_eh_region_end): Remove all references to
16180 NOTE_INSN_CALL_ARG_LOCATION.
16181 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16182 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16183 * config/arc/arc.c (hwloop_optimize): Likewise.
16184 * config/arm/arm.c (create_fix_barrier): Likewise.
16185 * config/s390/s390.c (s390_chunkify_start): Likewise.
16186 * config/sh/sh.c (find_barrier): Likewise.
16187 * config/i386/i386.c (rest_of_insert_endbranch,
16188 ix86_seh_fixup_eh_fallthru): Likewise.
16189 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16190 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16191 * config/frv/frv.c (frv_function_prologue): Likewise.
16192 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16194 (note_outside_basic_block_p): Remove all references to
16195 NOTE_INSN_CALL_ARG_LOCATION.
16196 * gengtype.c (adjust_field_rtx_def): Likewise.
16197 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16199 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16200 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16203 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16206 PR tree-optimization/84452
16207 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16208 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16211 2018-02-19 Martin Liska <mliska@suse.cz>
16214 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16216 2018-02-19 Martin Liska <mliska@suse.cz>
16217 Richard Sandiford <richard.sandiford@linaro.org>
16219 PR tree-optimization/82491
16220 * gimple-fold.c (get_base_constructor): Make earlier bail out
16223 2018-02-19 Carl Love <cel@us.ibm.com>
16225 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16226 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16228 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16229 P8V_BUILTIN_VEC_NEG.
16231 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16233 * config/rl78/rl78.md (movdf): New define expand.
16235 2018-02-19 Martin Liska <mliska@suse.cz>
16238 * doc/invoke.texi: Fix typo.
16239 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16241 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16243 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16244 handle rs6000_single_float and rs6000_double_float specially for
16247 2018-02-16 Jeff Law <law@redhat.com>
16249 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16250 (pop_regs): Use it.
16252 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16255 * ipa-inline.c (inline_small_functions): Fix a typo.
16257 2018-02-16 Nathan Sidwell <nathan@acm.org>
16259 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16261 2018-02-16 Carl Love <cel@us.ibm.com>
16263 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16264 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16265 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16266 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16267 expansion to P8V_BUILTIN_VEC_FLOAT2.
16269 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16271 PR rtl-optimization/70023
16272 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16273 src_regno into account.
16275 2018-02-16 Carl Love <cel@us.ibm.com>
16277 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16278 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16279 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16280 * config/rs6000/rs6000.c: Remove case statements for
16281 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16282 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16283 and P9V_BUILTIN_VEC_VINSERT4B.
16284 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16285 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16286 * config/rs6000/vsx.md:
16287 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16290 2018-02-16 Carl Love <cel@us.ibm.com>
16292 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16294 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16296 * config/rs6000/rs6000-c.c: Add the definitions for
16297 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16298 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16299 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16300 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16301 definition for insert4b and define insn *insert3b_internal.
16302 * doc/extend.texi: Add documentation for vec_extract4b.
16304 2018-02-16 Nathan Sidwell <nathan@acm.org>
16306 * doc/extend.texi (Backwards Compatibility): Mention friend
16307 injection. Note for-scope is deprecated.
16308 * doc/invoke.texi (-ffriend-injection): Deprecate.
16310 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16312 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16313 that moved to I2, also allow destinations that are a paradoxical
16314 subreg (instead of a normal reg).
16316 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16319 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16322 2018-02-16 Richard Biener <rguenther@suse.de>
16324 PR tree-optimization/84037
16325 PR tree-optimization/84016
16327 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16328 Adjust vec_construct for the fact we need additional higher latency
16329 128bit inserts for AVX256 and AVX512 vector builds.
16330 (ix86_add_stmt_cost): Scale vector construction cost for
16333 2018-02-16 Richard Biener <rguenther@suse.de>
16335 PR tree-optimization/84417
16336 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16337 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16338 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16340 2018-02-16 Martin Liska <mliska@suse.cz>
16343 * internal-fn.def (ASAN_CHECK): Set proper flags.
16344 (ASAN_MARK): Likewise.
16346 2018-02-16 Julia Koval <julia.koval@intel.com>
16348 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16349 from PTA_CANNONLAKE.
16351 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16354 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16355 Use ++iter rather than iter++ for std::list iterators.
16356 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16357 defer deleting them until all nodes in the forest are processed. Do
16358 free even leaf nodes. Change to_process into auto_vec.
16361 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16362 * vec.h (vec_default_construct): Use memset instead of placement new
16363 if BROKEN_VALUE_INITIALIZATION is defined.
16364 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16365 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16368 PR rtl-optimization/83723
16369 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16370 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16371 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16373 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16375 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16377 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16379 PR rtl-optimization/81443
16380 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16381 from inner REGs to paradoxical SUBREGs.
16383 2018-02-16 Richard Biener <rguenther@suse.de>
16385 PR tree-optimization/84399
16386 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16387 For operands we can analyze at their definition make sure we can
16388 analyze them at each use as well.
16390 2018-02-16 Richard Biener <rguenther@suse.de>
16392 PR tree-optimization/84190
16393 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16394 volatile accesses if the decl isn't volatile.
16396 2018-02-15 Jason Merrill <jason@redhat.com>
16398 PR c++/84314 - ICE with templates and fastcall attribute.
16399 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16400 TYPE_CANONICAL on an existing type.
16402 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16404 PR tree-optimization/84383
16405 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16406 dstoff nor call operand_equal_p if dstbase is NULL.
16408 PR tree-optimization/84334
16409 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16410 also a CONSTANT_CLASS_P, punt.
16412 2018-02-14 Jim Wilson <jimw@sifive.com>
16414 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16415 first SMALL_OPERAND check. New local min_second_step. Move assert
16416 to where locals are set. Add TARGET_RVC support.
16417 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16419 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16421 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16423 2018-02-14 Martin Sebor <msebor@redhat.com>
16425 PR tree-optimization/83698
16426 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16427 arrays constrain the offset range to their bounds.
16428 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16429 (builtin_access::overlap): Avoid setting the size of overlap if it's
16431 (maybe_diag_overlap): Also consider arrays when deciding what values
16432 of offsets to include in diagnostics.
16434 2018-02-14 Martin Sebor <msebor@redhat.com>
16437 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16438 that correspond to the kind of a declaration.
16440 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16443 * config/pa/pa.md: Load address of PIC label using the linkage table
16444 if the label is nonlocal.
16446 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16448 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16449 warning message if user requests -maltivec=be.
16450 * doc/invoke.texi: Document deprecation of -maltivec=be.
16452 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16455 * config/rs6000/rs6000-c.c: Update definitions for
16456 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16457 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16459 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16462 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16463 add _get_ssp intrinsics. Remove argument from
16464 __builtin_ia32_rdssp[d|q].
16465 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16466 * config/i386/i386-builtin.def: Remove argument from
16467 __builtin_ia32_rdssp[d|q].
16468 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16469 ix86_expand_special_args_builtin for _rdssp[d|q].
16470 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16471 Clear register before usage.
16472 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16473 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16475 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16477 PR tree-optimization/84357
16478 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16479 operand 1 of an ARRAY_REF too.
16481 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16484 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16485 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16487 (set_of_reg): New struct.
16488 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16489 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16490 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16492 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16493 Split into bitclr, bitset, bitinvert patterns if appropriate.
16494 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16495 use rx_fuse_in_memory_bitop.
16496 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16497 to named insn, correct maximum insn length.
16499 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16502 * machmode.def: Define a complex mode for PARTIAL_INT.
16503 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16505 * doc/rtl.texi: Document CSPImode.
16506 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16508 (msp430_hard_regno_nregs_with_padding): Likewise.
16510 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16513 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16515 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16517 PR rtl-optimization/84169
16518 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16519 we generated a parallel as new i3 and we split that to new i2 and i3
16520 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16521 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16522 those to i2, not i1. Partially rewrite this scan code.
16524 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16527 * stor-layout.c (place_field): For variable length fields, adjust
16528 offset_align afterwards not just based on the field's alignment,
16529 but also on the size.
16531 PR middle-end/84309
16532 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16533 of exps and logs in the use_exp2 case.
16535 2018-02-13 Jeff Law <law@redhat.com>
16537 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16538 entry for "vector".
16540 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16543 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16547 * common.opt (gas-loc-support, gas-locview-support): New.
16548 (ginline-points, ginternal-reset-location-views): New.
16549 * doc/invoke.texi: Document them. Use @itemx where intended.
16550 (gvariable-location-views): Adjust.
16551 * target.def (reset_location_view): New.
16552 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16553 (TARGET_RESET_LOCATION_VIEW): New.
16554 * doc/tm.texi: Rebuilt.
16555 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16556 (dwarf2out_default_as_locview_support): New.
16557 (output_asm_line_debug_info): Use option variables.
16558 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16559 (output_loc_list): Likewise.
16560 (add_high_low_attributes): Check option variables.
16561 Don't output entry view attribute in strict mode.
16562 (gen_inlined_subroutine_die): Check option variables.
16563 (dwarf2out_inline_entry): Likewise.
16564 (init_sections_and_labels): Likewise.
16565 (dwarf2out_early_finish): Likewise.
16566 (maybe_reset_location_view): New, from...
16567 (dwarf2out_var_location): ... here. Call it.
16568 * debug.h (dwarf2out_default_as_loc_support): Declare.
16569 (dwarf2out_default_as_locview_support): Declare.
16570 * hooks.c (hook_int_rtx_insn_0): New.
16571 * hooks.h (hook_int_rtx_insn_0): Declare.
16572 * toplev.c (process_options): Take -gas-loc-support and
16573 -gas-locview-support from dwarf2out. Enable
16574 -gvariable-location-views by default only with locview
16575 assembler support. Enable -ginternal-reset-location-views by
16576 default only if the target defines the corresponding hook.
16577 Enable -ginline-points by default if location views are
16578 enabled; force it disabled if statement frontiers are
16580 * tree-inline.c (expand_call_inline): Check option variables.
16581 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16583 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16585 PR tree-optimization/84321
16586 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16587 handling. Also check whether the anti-range contains any values
16588 that satisfy the mask; switch to a VR_RANGE if not.
16590 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16593 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16595 2018-02-13 Martin Jambor <mjambor@suse.cz>
16598 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16599 of call statements, also set location of a load to a temporary.
16601 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16603 * config/rl78/rl78.c (add_vector_labels): New function.
16604 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16605 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16606 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16607 which checks that no arguments are passed.
16608 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16609 * doc/extend.texi: Documentation for the new attribute.
16611 2018-02-13 Andreas Schwab <schwab@suse.de>
16613 * config/riscv/linux.h (CPP_SPEC): Define.
16615 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16618 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16619 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16620 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16621 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16622 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16623 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16624 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16625 temporarily for AES and PCLMUL builtins.
16627 PR tree-optimization/84339
16628 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16629 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16632 PR middle-end/84309
16633 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16634 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16635 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16637 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16639 * omp-simd-clone.h: New file.
16640 * omp-simd-clone.c: Include omp-simd-clone.h.
16641 (expand_simd_clones): No longer static.
16642 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16643 cgraph.h and omp-simd-clone.h.
16644 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16645 (vect_recog_widen_shift_pattern): Formatting fix.
16646 (vect_pattern_recog_1): Don't check optab for calls.
16649 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16650 operands[2] into a REG before using gen_lowpart on it.
16652 2018-02-12 Jeff Law <law@redhat.com>
16655 * config/sh/sh.c (find_barrier): Consider a sibling call
16658 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16659 successfully back substituting a reg.
16661 2018-02-12 Richard Biener <rguenther@suse.de>
16663 PR tree-optimization/84037
16664 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16665 parameter, move visited init to caller.
16666 (vect_slp_analyze_operations): Separate cost from validity
16667 check, initialize visited once for all instances.
16668 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16670 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16672 (vect_model_promotion_demotion_cost): Likewise.
16673 (vectorizable_bswap): Guard cost modeling with !slp_node
16674 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16676 (vectorizable_call): Likewise.
16677 (vectorizable_conversion): Likewise.
16678 (vectorizable_assignment): Likewise.
16679 (vectorizable_shift): Likewise.
16680 (vectorizable_operation): Likewise.
16681 (vectorizable_store): Likewise.
16682 (vectorizable_load): Likewise.
16683 (vectorizable_condition): Likewise.
16684 (vectorizable_comparison): Likewise.
16686 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16689 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16690 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16691 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16693 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16695 PR middle-end/83665
16696 * params.def (inline-min-speedup): Increase from 8 to 15.
16697 (max-inline-insns-auto): Decrease from 40 to 30.
16698 * ipa-split.c (consider_split): Add some buffer for function to
16699 be considered inlining candidate.
16700 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16703 2018-02-12 Richard Biener <rguenther@suse.de>
16705 PR tree-optimization/84037
16706 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16707 matched stmts if we cannot swap the non-matched ones.
16709 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16711 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16712 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16713 _mm_maskz_scalef_round_ss): New intrinsics.
16714 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16715 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16716 __builtin_ia32_scalefss_round): Remove.
16717 (__builtin_ia32_scalefsd_mask_round,
16718 __builtin_ia32_scalefss_mask_round): New intrinsics.
16719 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16720 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16721 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16722 "<round_constraint>")): Changed to ...
16723 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16724 "<round_scalar_constraint>")): ... this.
16725 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16726 %0, %1, %2<round_op3>}"): Changed to ...
16727 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16728 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16729 %2<round_scalar_mask_op3>}"): ... this.
16730 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16732 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16734 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16735 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16736 (_mm_maskz_sqrt_round_ss): New intrinsics.
16737 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16738 (__builtin_ia32_sqrtsd_mask_round)
16739 (__builtin_ia32_sqrtss_mask_round): New builtins.
16740 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16741 (__builtin_ia32_sqrtss_round): Remove.
16742 (__builtin_ia32_sqrtsd_mask_round)
16743 (__builtin_ia32_sqrtss_mask_round): New builtins.
16744 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16745 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16746 ((match_operand:VF_128 1 "vector_operand"
16747 "xBm,<round_constraint>")): Changed to ...
16748 ((match_operand:VF_128 1 "vector_operand"
16749 "xBm,<round_scalar_constraint>")): ... this.
16750 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16751 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16752 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16753 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16754 %<iptr>1<round_scalar_mask_op3>}): ... this.
16755 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16756 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16758 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16761 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16762 Cast vec_cmpeq result to correct type.
16763 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16764 Cast vec_cmpgt result to correct type.
16766 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16768 * final.c (final_scan_insn_1): Renamed from...
16769 (final_scan_insn): ... this. New wrapper, to recover
16770 seen from the outermost call in recursive ones.
16771 * config/sparc/sparc.c (output_return): Drop seen from call.
16772 (output_sibcall): Likewise.
16773 * config/visium/visium.c (output_branch): Likewise.
16775 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16777 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16780 2018-02-10 Alan Modra <amodra@gmail.com>
16783 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16784 Specify LR as an input.
16786 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16789 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16790 remove_member_access_dummy_vars): New functions.
16791 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16792 lower_omp_1, execute_lower_omp): Use them.
16794 PR rtl-optimization/84308
16795 * shrink-wrap.c (spread_components): Release todo vector.
16797 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16799 PR rtl-optimization/57193
16800 * ira-color.c (struct allocno_color_data): Add member
16801 conflict_allocno_hard_prefs.
16802 (update_conflict_allocno_hard_prefs): New.
16803 (bucket_allocno_compare_func): Add a preference based on
16804 conflict_allocno_hard_prefs.
16805 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16806 (color_allocnos): Remove a dead code. Initiate
16807 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16809 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16812 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16813 constraint from =wa to wa. Avoid a subreg on the output operand,
16814 instead use a pseudo and subreg it in a move.
16815 (p9_xxbrd_<mode>): Changed to ...
16816 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16817 (p9_xxbrd_v2df): New expander.
16818 (p9_xxbrw_<mode>): Changed to ...
16819 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16820 (p9_xxbrw_v4sf): New expander.
16822 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16824 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16826 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16829 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16830 multiply in 32-bit mode.
16831 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16832 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16835 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16837 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16838 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16839 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16840 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16842 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16845 * dwarf2out.c (is_trivial_indirect_ref): New function.
16846 (dwarf2out_late_global_decl): Do not generate a location
16847 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16848 and that are not defined in the current unit.
16850 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16852 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16853 instead of a libcall for UNORDERED.
16855 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16858 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16859 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16861 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16864 * config/s390/s390.c (s390_set_current_function): Invoke
16865 s390_indirect_branch_settings also if fndecl didn't change.
16867 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16869 * config/rs6000/rs6000.md (blockage): Set length to zero.
16871 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16873 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16875 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16878 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16879 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16883 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16884 PARALLEL incoming that failed vt_get_decl_and_offset check.
16886 PR middle-end/84237
16887 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16888 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16890 (get_variable_section): For decls in named .bss* sections pass true as
16891 second argument to bss_initializer_p.
16893 2018-02-09 Marek Polacek <polacek@redhat.com>
16894 Jakub Jelinek <jakub@redhat.com>
16897 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16898 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16899 Sync some changes from cxx_fold_indirect_ref.
16901 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16903 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16905 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16906 (BLOCK_INLINE_ENTRY_LABEL): New.
16907 (dwarf2out_var_location): Disregard inline entry markers.
16908 (inline_entry_data): New struct.
16909 (inline_entry_data_hasher): New hashtable type.
16910 (inline_entry_data_hasher::hash): New.
16911 (inline_entry_data_hasher::equal): New.
16912 (inline_entry_data_table): New variable.
16913 (add_high_low_attributes): Add DW_AT_entry_pc and
16914 DW_AT_GNU_entry_view attributes if a pending entry is found
16915 in inline_entry_data_table. Add old entry_pc attribute only
16916 if debug nonbinding markers are disabled.
16917 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16918 markers are enabled.
16919 (block_within_block_p, dwarf2out_inline_entry): New.
16920 (dwarf2out_finish): Check that no entries remained in
16921 inline_entry_data_table.
16922 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16923 (final_scan_insn, notice_source_line): Likewise.
16924 (rest_of_clean_state): Skip inline entry markers.
16925 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16927 * gimple.c (gimple_build_debug_inline_entry): New.
16928 * gimple.h (enum gimple_debug_subcode): Add
16929 GIMPLE_DEBUG_INLINE_ENTRY.
16930 (gimple_build_debug_inline_entry): Declare.
16931 (gimple_debug_inline_entry_p): New.
16932 (gimple_debug_nonbind_marker_p): Adjust.
16933 * insn-notes.def (INLINE_ENTRY): New.
16934 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16935 inline entry marker notes.
16936 (print_insn): Likewise.
16937 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16938 (INSN_DEBUG_MARKER_KIND): Likewise.
16939 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16940 * tree-inline.c (expand_call_inline): Build and insert
16941 debug_inline_entry stmt.
16942 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16943 inline entry blocks early, if nonbind markers are enabled.
16944 (dump_scope_block): Dump fragment info.
16945 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16946 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16947 (gimple_build_debug_inline_entry): New.
16948 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16949 Enable/disable inline entry points too.
16950 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16951 (DEBUG_INSN): Describe inline entry markers.
16953 * common.opt (gvariable-location-views): New.
16954 (gvariable-location-views=incompat5): New.
16955 * config.in: Rebuilt.
16956 * configure: Rebuilt.
16957 * configure.ac: Test assembler for view support.
16958 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16959 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16960 * dwarf2out.c (var_loc_view): New typedef.
16961 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16962 (dwarf2out_locviews_in_attribute): New.
16963 (dwarf2out_locviews_in_loclist): New.
16964 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16965 (enum dw_line_info_opcode): Add LI_adv_address.
16966 (struct dw_line_info_table): Add view.
16967 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16968 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16969 (zero_view_p): New variable.
16970 (ZERO_VIEW_P): New macro.
16971 (output_asm_line_debug_info): New.
16972 (struct var_loc_node): Add view.
16973 (add_AT_view_list, AT_loc_list): New.
16974 (add_var_loc_to_decl): Add view param. Test it against last.
16975 (new_loc_list): Add view params. Record them.
16976 (AT_loc_list_ptr): Handle loc and view lists.
16977 (view_list_to_loc_list_val_node): New.
16978 (print_dw_val): Handle dw_val_class_view_list.
16979 (size_of_die): Likewise.
16980 (value_format): Likewise.
16981 (loc_list_has_views): New.
16982 (gen_llsym): Set vl_symbol too.
16983 (maybe_gen_llsym, skip_loc_list_entry): New.
16984 (dwarf2out_maybe_output_loclist_view_pair): New.
16985 (output_loc_list): Output view list or entries too.
16986 (output_view_list_offset): New.
16987 (output_die): Handle dw_val_class_view_list.
16988 (output_dwarf_version): New.
16989 (output_compilation_unit_header): Use it.
16990 (output_skeleton_debug_sections): Likewise.
16991 (output_rnglists, output_line_info): Likewise.
16992 (output_pubnames, output_aranges): Update version comments.
16993 (output_one_line_info_table): Output view numbers in asm comments.
16994 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16995 Call maybe_gen_llsym.
16996 (loc_list_from_tree_1): Adjust.
16997 (add_AT_location_description): Create view list attribute if
16998 needed, check it's absent otherwise.
16999 (convert_cfa_to_fb_loc_list): Adjust.
17000 (maybe_emit_file): Call output_asm_line_debug_info for test.
17001 (dwarf2out_var_location): Reset views as needed. Precompute
17002 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
17003 attribute. Set view.
17004 (new_line_info_table): Reset next view.
17005 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
17006 (dwarf2out_source_line): Likewise. Output view resets and labels to
17007 the assembler, or select appropriate line info opcodes.
17008 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
17009 (optimize_string_length): Catch it. Adjust.
17010 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
17011 dw_val_class_view_list, and remove it if no longer needed.
17012 (hash_loc_list): Hash view numbers.
17013 (loc_list_hasher::equal): Compare them.
17014 (optimize_location_lists): Check whether a view list symbol is
17015 needed, and whether the locview attribute is present, and
17016 whether they match. Remove the locview attribute if no longer
17018 (index_location_lists): Call skip_loc_list_entry for test.
17019 (dwarf2out_finish): Call output_asm_line_debug_info for test.
17020 Use output_dwarf_version.
17021 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
17022 (struct dw_val_node): Add val_view_list.
17023 * final.c (SEEN_NEXT_VIEW): New.
17024 (set_next_view_needed): New.
17025 (clear_next_view_needed): New.
17026 (maybe_output_next_view): New.
17027 (final_start_function): Rename to...
17028 (final_start_function_1): ... this. Take pointer to FIRST,
17029 add SEEN parameter. Emit param bindings in the initial view.
17030 (final_start_function): Reintroduce SEEN-less interface.
17031 (final): Rename to...
17032 (final_1): ... this. Take SEEN parameter. Output final pending
17033 next view at the end.
17034 (final): Reintroduce seen-less interface.
17035 (final_scan_insn): Output pending next view before switching
17036 sections or ending a block. Mark the next view as needed when
17037 outputting variable locations. Notify debug backend of section
17038 changes, and of location view changes.
17039 (rest_of_handle_final): Adjust.
17040 * toplev.c (process_options): Autodetect value for debug variable
17041 location views option. Warn on incompat5 without -gdwarf-5.
17042 * doc/invoke.texi (gvariable-location-views): New.
17043 (gvariable-location-views=incompat5): New.
17044 (gno-variable-location-views): New.
17046 2018-02-08 David Malcolm <dmalcolm@redhat.com>
17048 PR tree-optimization/84136
17049 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
17050 that the result of find_edge is non-NULL.
17052 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17055 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
17056 storing integer register in SImode. Fix cost of 256 and 512
17057 byte aligned SSE register store.
17059 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17061 * config/i386/i386.c (ix86_multiplication_cost): Fix
17062 multiplication cost for TARGET_AVX512DQ.
17064 2018-02-08 Marek Polacek <polacek@redhat.com>
17066 PR tree-optimization/84238
17067 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
17070 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17072 PR tree-optimization/84265
17073 * tree-vect-stmts.c (vectorizable_store): Don't treat
17074 VMAT_CONTIGUOUS accesses as grouped.
17075 (vectorizable_load): Likewise.
17077 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17079 PR tree-optimization/81635
17080 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
17081 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
17082 (test_round_for_mask): New functions.
17083 (wide_int_cc_tests): Call test_round_for_mask.
17084 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
17085 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
17086 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
17087 range returned by get_range_info.
17089 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17092 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
17093 * symtab.c: Include builtins.h
17094 (symtab_node::output_to_lto_symbol_table_p): Move here
17095 from lto-streamer-out.c:output_symbol_p.
17096 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
17097 (output_symbol_p): Move all logic to symtab.c
17098 (produce_symtab): Update.
17100 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17102 * config/s390/s390-opts.h (enum indirect_branch): Define.
17103 * config/s390/s390-protos.h (s390_return_addr_from_memory)
17104 (s390_indirect_branch_via_thunk)
17105 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
17106 (enum s390_indirect_branch_type): Define.
17107 * config/s390/s390.c (struct s390_frame_layout, struct
17108 machine_function): Remove.
17109 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
17110 (indirect_branch_table_label_no, indirect_branch_table_name):
17112 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
17113 (enum s390_indirect_branch_option): Define.
17114 (s390_return_addr_from_memory): New function.
17115 (s390_handle_string_attribute): New function.
17116 (s390_attribute_table): Add new attribute handler.
17117 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17118 (s390_indirect_branch_via_thunk): New function.
17119 (s390_indirect_branch_via_inline_thunk): New function.
17120 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17121 sibling call optimization for non z10 compiles.
17122 (s390_emit_call): Force indirect branch target to be a single
17123 register. Add r1 clobber for non-z10 compiles.
17124 (s390_emit_epilogue): Emit return jump via return_use expander.
17125 (s390_reorg): Handle JUMP_INSNs as execute targets.
17126 (s390_option_override_internal): Perform validity checks for the
17127 new command line options.
17128 (s390_indirect_branch_attrvalue): New function.
17129 (s390_indirect_branch_settings): New function.
17130 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17131 (s390_output_indirect_thunk_function): New function.
17132 (s390_code_end): Implement target hook.
17133 (s390_case_values_threshold): Implement target hook.
17134 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17136 * config/s390/s390.h (struct s390_frame_layout)
17137 (struct machine_function): Move here from s390.c.
17138 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17139 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17140 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17141 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17142 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17143 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17144 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17145 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17146 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17147 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17148 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17149 (mnemonic attribute): Add values which aren't recognized
17151 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17152 pattern for branch conversion. Fix mnemonic attribute.
17153 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17154 indirect branch via thunk if requested.
17155 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17156 ("*indirect_jump"): Disable for branch conversion using out of
17158 ("indirect_jump_via_thunk<mode>_z10")
17159 ("indirect_jump_via_thunk<mode>")
17160 ("indirect_jump_via_inlinethunk<mode>_z10")
17161 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17162 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17163 ("casesi_jump_via_inlinethunk<mode>_z10")
17164 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17165 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17166 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17167 ("*indirect2_jump"): Disable for branch conversion.
17168 ("casesi_jump"): Turn into expander and expand patterns for branch
17170 ("return_use"): New expander.
17171 ("*return"): Emit return via thunk and rename it to ...
17172 ("*return<mode>"): ... this one.
17173 * config/s390/s390.opt: Add new options and and enum for the
17176 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17178 * lra-constraints.c (match_reload): Unconditionally use
17179 gen_lowpart_SUBREG, rather than selecting between that
17180 and equivalent gen_rtx_SUBREG code.
17182 2018-02-08 Richard Biener <rguenther@suse.de>
17184 PR tree-optimization/84233
17185 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17186 changed flag instead of boguously re-using phi_inserted.
17188 2018-02-08 Martin Jambor <mjambor@suse.cz>
17190 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17191 static local variables.
17193 2018-02-08 Richard Biener <rguenther@suse.de>
17195 PR tree-optimization/84278
17196 * tree-vect-stmts.c (vectorizable_store): When looking for
17197 smaller vector types to perform grouped strided loads/stores
17198 make sure the mode is supported by the target.
17199 (vectorizable_load): Likewise.
17201 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17203 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17204 Increase LDP/STP opportunities by adding adjacent callee-saves.
17206 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17208 PR rtl-optimization/84068
17209 PR rtl-optimization/83459
17210 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17212 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17214 PR tree-optimization/84224
17215 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17216 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17217 non-zero arguments.
17219 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17222 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17223 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17224 count, remove one USE.
17226 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17228 * doc/install.texi (Configuration): Document the
17229 --with-long-double-format={ibm,ieee} PowerPC configuration
17233 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17234 Convert from define_expand to be define_insn_and_split. Rework
17235 float/double/_Float128 conversions to QI/HI/SImode to work with
17236 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17237 conversions to QI/HImode types did a store and then a load to
17238 truncate the value. For conversions to VSX registers, don't split
17239 the insn, instead emit the code directly. Use the code iterator
17240 any_fix to combine signed and unsigned conversions.
17241 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17242 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17243 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17244 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17245 (fix_<mode>di2_hw): Likewise.
17246 (fixuns_<mode>di2_hw): Likewise.
17247 (fix_<mode>si2_hw): Likewise.
17248 (fixuns_<mode>si2_hw): Likewise.
17249 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17250 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17251 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17252 fix<uns>_trunc<SFDF:mode>si2_p8.
17253 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17254 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17255 (fix<uns>_<mode>_mem): Likewise.
17256 (fctiw<u>z_<mode>_mem): Likewise.
17257 (fix<uns>_<mode>_mem): Likewise.
17258 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17259 the register allocator from doing a direct move to the GPRs to do
17260 a store, and instead use the ISA 3.0 store byte/half-word from
17261 vector register instruction. For IEEE 128-bit floating point,
17262 also optimize stores of 32-bit ints.
17263 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17265 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17267 * genextract.c (push_pathstr_operand): New function to support
17269 (walk_rtx): Call push_pathstr_operand.
17270 (print_path): Support [a-zA-Z].
17272 2018-02-07 Richard Biener <rguenther@suse.de>
17274 PR tree-optimization/84037
17275 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17276 (cse_and_gimplify_to_preheader): Declare.
17277 (vect_get_place_in_interleaving_chain): Likewise.
17278 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17280 (_loop_vec_info::~_loop_vec_info): Delete it.
17281 (cse_and_gimplify_to_preheader): New function.
17282 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17283 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17284 (vectorizable_load): Likewise. For grouped stores always base
17285 the IV on the first element.
17286 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17287 condition before gimplifying.
17289 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17291 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17292 *DIV_EXPR and *MOD_EXPR.
17294 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17297 * config/i386/i386.c (ix86_option_override_internal): Mask out
17298 the CF_SET bit when checking -fcf-protection.
17300 2018-02-07 Tom de Vries <tom@codesourcery.com>
17303 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17306 2018-02-07 Richard Biener <rguenther@suse.de>
17308 PR tree-optimization/84204
17309 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17312 PR tree-optimization/84205
17313 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17314 special-case isl_ast_op_zdiv_r.
17316 PR tree-optimization/84223
17317 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17318 Only add conditions from within the region.
17319 (gather_bbs::after_dom_children): Adjust.
17321 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17324 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17325 * config/avr/avr.md: Only post-reload split REG-REG moves if
17326 either register is GENERAL_REG_P.
17328 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17330 PR tree-optimization/84235
17331 * tree-ssa-scopedtables.c
17332 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17333 if the subtraction is performed in floating point type where NaNs are
17334 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17335 build 1. Formatting fix.
17337 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17340 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17341 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17342 and skip it regardless of bb boundaries. Use CALL_P macro,
17343 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17344 unnecessarily, formatting fix.
17346 2018-02-06 Michael Collison <michael.collison@arm.com>
17348 * config/arm/thumb2.md:
17349 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17350 (*thumb_mov_notscc): Ditto.
17352 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17355 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17356 unsigned_fix, not "s".
17358 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17360 * configure.ac (gcc_fn_eh_frame_ro): New function.
17361 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17362 correct .eh_frame permissions.
17363 * configure: Regenerate.
17365 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17367 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17368 irrelevant options.
17370 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17372 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17373 Display warning message for -mno-speculate-indirect-jumps.
17375 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17377 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17379 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17381 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17383 PR tree-optimization/84225
17384 * tree-eh.c (find_trapping_overflow): Only call
17385 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17387 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17390 * config/i386/i386.c: Reimplement the check of possible options
17391 -mibt/-mshstk conbination. Change error messages.
17392 * doc/invoke.texi: Fix a typo: remove extra '='.
17394 2018-02-06 Marek Polacek <polacek@redhat.com>
17396 PR tree-optimization/84228
17397 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17399 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17402 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17403 emitted arch directives.
17404 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17405 __ARM_FEATURE_COPROC before changing architectures.
17407 2018-02-06 Richard Biener <rguenther@suse.de>
17409 * config/i386/i386.c (print_reg): Fix typo.
17410 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17412 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17414 * configure: Regenerate.
17416 2018-02-05 Martin Sebor <msebor@redhat.com>
17418 PR tree-optimization/83369
17419 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17422 2018-02-05 Martin Liska <mliska@suse.cz>
17424 * doc/invoke.texi: Cherry-pick upstream r323995.
17426 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17428 * ira.c (ira_init_register_move_cost): Adjust comment.
17430 2018-02-05 Martin Liska <mliska@suse.cz>
17432 PR gcov-profile/84137
17433 * doc/gcov.texi: Fix typo in documentation.
17435 2018-02-05 Martin Liska <mliska@suse.cz>
17437 PR gcov-profile/83879
17438 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17439 using dlopen functionality.
17441 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17443 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17444 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17445 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17446 _mm_maskz_range_round_ss): New intrinsics.
17447 (__builtin_ia32_rangesd128_round)
17448 (__builtin_ia32_rangess128_round): Remove.
17449 (__builtin_ia32_rangesd128_mask_round,
17450 __builtin_ia32_rangess128_mask_round): New builtins.
17451 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17452 __builtin_ia32_rangess128_round): Remove.
17453 (__builtin_ia32_rangesd128_mask_round,
17454 __builtin_ia32_rangess128_mask_round): New builtins.
17455 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17456 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17457 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17458 "<round_saeonly_constraint>")): Changed to ...
17459 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17460 "<round_saeonly_scalar_constraint>")): ... this.
17461 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17462 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17463 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17464 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17465 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17467 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17469 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17471 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17472 Remove all values except native, 8540 and 8548.
17474 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17476 * config/i386/i386.c (ix86_output_function_return): Pass
17477 INVALID_REGNUM, instead of -1, as invalid register number to
17478 indirect_thunk_name and output_indirect_thunk.
17480 2018-02-02 Julia Koval <julia.koval@intel.com>
17482 * config.gcc: Add -march=icelake.
17483 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17484 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17485 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17486 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17487 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17488 (processor_target_table): Add icelake.
17489 (ix86_option_override_internal): Handle new PTAs.
17490 (get_builtin_code_for_version): Handle icelake.
17491 (M_INTEL_COREI7_ICELAKE): New.
17492 (fold_builtin_cpu): Handle icelake.
17493 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17494 * doc/invoke.texi: Add -march=icelake.
17496 2018-02-02 Julia Koval <julia.koval@intel.com>
17498 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17499 to wide_int_bitmask.
17500 * wide-int-bitmask.h: New.
17502 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17505 * config/i386/i386.md: Replace Pmode with word_mode in
17506 builtin_setjmp_setup and builtin_longjmp to support x32.
17508 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17512 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17514 (rs6000_supported_cpu_names): New static variable.
17515 (linux_cpu_translation_table): Likewise.
17516 (elf_platform) <cpu>: Define new static variable and use it.
17517 Translate kernel AT_PLATFORM name to canonical name if needed.
17518 Error if platform name is unknown.
17520 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17523 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17525 2018-02-01 Jeff Law <law@redhat.com>
17528 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17529 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17530 the scratch if RELEASE_VIA_POP is false.
17531 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17532 If we have to save a temporary register, decrement SIZE appropriately.
17533 Pass new arguments to release_scratch_register_on_entry.
17534 (ix86_adjust_stack_and_probe): Likewise.
17535 (ix86_emit_probe_stack_range): Pass new arguments to
17536 release_scratch_register_on_entry.
17538 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17540 PR rtl-optimization/84157
17541 * combine.c (change_zero_ext): Use REG_P predicate in
17542 front of HARD_REGISTER_P predicate.
17544 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17546 * config/avr/avr.c (avr_option_override): Move disabling of
17547 -fdelete-null-pointer-checks to...
17548 * common/config/avr/avr-common.c (avr_option_optimization_table):
17551 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17553 PR tree-optimization/81635
17554 * tree-data-ref.c (split_constant_offset_1): For types that
17555 wrap on overflow, try to use range info to prove that wrapping
17558 2018-02-01 Renlin Li <renlin.li@arm.com>
17561 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17562 TAILCALL_ADDR_REGS.
17563 (aarch64_register_move_cost): Likewise.
17564 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17565 TAILCALL_ADDR_REGS.
17566 (REG_CLASS_NAMES): Likewise.
17567 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17568 TAILCALL_ADDR_REGS. Remove IP registers.
17569 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17571 2018-02-01 Richard Biener <rguenther@suse.de>
17573 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17574 for specifying RPO order and allow NULL for that.
17575 * domwalk.c (dom_walker::dom_walker): Likewise.
17576 (dom_walker::walk): Handle NULL RPO order.
17577 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17579 (rewrite_update_dom_walker): Likewise.
17580 (mark_def_dom_walker): Likewise.
17582 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17584 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17585 (aarch64_maybe_expand_sve_subreg_move): Declare.
17586 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17587 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17589 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17590 that are semantically a reverse operation.
17591 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17592 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17593 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17595 (aarch64_can_change_mode_class): For big-endian, forbid changes
17596 between two SVE modes if they have different element sizes.
17598 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17600 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17601 the TImode handling for big-endian targets.
17603 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17605 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17606 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17608 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17609 Remove BSWAP handing for big-endian targets and use the form of
17610 LD1RQ appropariate for the mode.
17612 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17614 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17615 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17616 duplicated element.
17618 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17621 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17622 check for operands that need to go through aarch64_sve_reload_be.
17624 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17626 PR tree-optimization/81661
17627 PR tree-optimization/84117
17628 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17629 * tree-eh.c: Include gimplify.h.
17630 (find_trapping_overflow, replace_trapping_overflow,
17631 rewrite_to_non_trapping_overflow): New functions.
17632 * tree-vect-loop.c: Include tree-eh.h.
17633 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17634 * tree-data-ref.c: Include tree-eh.h.
17635 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17637 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17639 PR rtl-optimization/84123
17640 * combine.c (change_zero_ext): Check if hard register satisfies
17641 can_change_dest_mode before calling gen_lowpart_SUBREG.
17643 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17646 * ira.c (ira_init_register_move_cost): Remove assert.
17648 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17650 PR rtl-optimization/84071
17651 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17652 * doc/tm.texi: Regenerate.
17654 2018-01-31 Richard Biener <rguenther@suse.de>
17656 PR tree-optimization/84132
17657 * tree-data-ref.c (analyze_miv_subscript): Properly
17658 check whether evolution_function_is_affine_multivariate_p
17659 before calling gcd_of_steps_may_divide_p.
17661 2018-01-31 Julia Koval <julia.koval@intel.com>
17664 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17665 * config/i386/i386.md (rdpid_rex64) New.
17666 (rdpid): Make 32bit only.
17668 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17671 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17672 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17674 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17677 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17679 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17681 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17683 PR rtl-optimization/84071
17684 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17685 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17687 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17689 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17690 (arc_attribute_table): Add 'aux' attribute.
17691 (arc_in_small_data_p): Consider aux like variables.
17692 (arc_is_aux_reg_p): New function.
17693 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17694 (arc_get_aux_arg): New function.
17695 (prepare_move_operands): Handle aux-register access.
17696 (arc_handle_aux_attribute): New function.
17697 * doc/extend.texi (ARC Variable attributes): Add subsection.
17699 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17701 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17702 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17703 (arc_attribute_table): Add 'uncached' attribute.
17704 (arc_print_operand): Print '.di' flag for uncached memory
17706 (arc_in_small_data_p): Do not consider for small data the uncached
17708 (arc_is_uncached_mem_p): New function.
17709 * config/arc/predicates.md (compact_store_memory_operand): Check
17710 for uncached memory accesses.
17711 (nonvol_nonimm_operand): Likewise.
17712 * doc/extend.texi (ARC Type Attribute): New subsection.
17714 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17717 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17718 falign-loops=): Add Optimization flag.
17720 2018-01-30 Jeff Law <law@redhat.com>
17723 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17724 INT_REGISTERS_SAVED. Check it prior to calling
17725 get_scratch_register_on_entry.
17726 (ix86_adjust_stack_and_probe): Similarly.
17727 (ix86_emit_probe_stack_range): Similarly.
17728 (ix86_expand_prologue): Corresponding changes.
17730 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17733 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17734 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17736 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17739 * lra-constraints.c (curr_insn_transform): Process AND in the
17742 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17744 PR rtl-optimization/83986
17745 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17746 dependence against last_pending_memory_flush in addition to
17747 pending_jump_insns.
17749 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17751 PR tree-optimization/81611
17752 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17755 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17758 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17761 2018-01-30 Richard Biener <rguenther@suse.de>
17762 Jakub Jelinek <jakub@redhat.com>
17764 PR tree-optimization/84111
17765 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17766 inner loops added during recursion, as they don't have up-to-date
17769 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17772 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17773 (can_inline_edge_by_limits_p): ... here.
17774 (can_early_inline_edge_p, check_callers,
17775 update_caller_keys, update_callee_keys, recursive_inlining,
17776 add_new_edges_to_heap, speculation_useful_p,
17777 inline_small_functions,
17778 inline_small_functions, flatten_function,
17779 inline_to_all_callers_1): Update.
17781 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17783 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17786 2018-01-30 Richard Biener <rguenther@suse.de>
17788 PR tree-optimization/83008
17789 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17790 invariant and constant vector uses in stmts when they need
17791 more than one stmt.
17793 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17796 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17797 * configure: Regenerate.
17799 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17801 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17803 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17804 Use gen_rtx_REG rather than gen_lowpart.
17806 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17808 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17809 rather than 0 when creating partial subregs.
17811 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17813 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17816 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17819 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17820 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17821 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17822 flags. This restores the settings used before the 2017-07-24.
17823 Turning off pre increment/decrement/modify allows IVOPTS to
17824 optimize DF/SF loops where the index is an int.
17826 2018-01-29 Richard Biener <rguenther@suse.de>
17827 Kelvin Nilsen <kelvin@gcc.gnu.org>
17830 * tree-vect-stmts.c (vectorizable_call): Don't call
17831 targetm.vectorize_builtin_md_vectorized_function if callee is
17834 2018-01-22 Carl Love <cel@us.ibm.com>
17836 * doc/extend.tex: Fix typo in second arg in
17837 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17839 2018-01-29 Richard Biener <rguenther@suse.de>
17841 PR tree-optimization/84086
17842 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17843 (flush_ssaname_freelist): When SSA names were released reset
17844 the SCEV hash table.
17846 2018-01-29 Richard Biener <rguenther@suse.de>
17848 PR tree-optimization/84057
17849 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17850 removed paths when removing edges.
17852 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17854 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17855 -mfunction-return=@var{choice}.
17857 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17859 PR diagnostic/84034
17860 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17861 Handle CR like TAB.
17862 (layout::print_source_line): Likewise.
17863 (test_get_line_width_without_trailing_whitespace): Add test cases.
17865 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17867 PR middle-end/84040
17868 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17871 2018-01-26 Jim Wilson <jimw@sifive.com>
17873 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17875 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17878 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17880 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17881 and CMP + SUB-immediate -> SUBS.
17883 2018-01-26 Martin Sebor <msebor@redhat.com>
17885 PR tree-optimization/83896
17886 * tree-ssa-strlen.c (get_string_len): Rename...
17887 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17888 Avoid assuming length is constant.
17889 (handle_char_store): Use HOST_WIDE_INT for string length.
17891 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17894 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17895 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17897 2018-01-26 Richard Biener <rguenther@suse.de>
17899 PR rtl-optimization/84003
17900 * dse.c (record_store): Only record redundant stores when
17901 the earlier store aliases at least all accesses the later one does.
17903 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17905 PR rtl-optimization/83985
17906 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17907 REG_CFA_RESTORE insns.
17908 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17909 REG_CFA_RESTORE insns here.
17912 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17913 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17915 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17917 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17918 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17919 (arc_init): Likewise.
17920 (arc_override_options): Likewise.
17921 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17923 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17925 * config/arc/arc.h (TARGET_DBNZ): Define.
17926 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17927 properly set the tune attribute.
17928 (dbnz): Use TARGET_DBNZ guard.
17929 * config/arc/arc.opt (mtune): Add core3 option.
17931 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17933 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17934 recognize new pic like addresses.
17935 (arc_delegitimize_address): Clean up.
17937 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17939 * config/arc/arc-arches.def: Option mrf16 valid for all
17941 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17942 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17943 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17944 * config/arc/arc-tables.opt: Regenerate.
17945 * config/arc/arc.c (arc_conditional_register_usage): Handle
17946 reduced register file case.
17947 (arc_file_start): Set must have build attributes.
17948 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17949 mrf16 option value.
17950 * config/arc/arc.opt (mrf16): Add new option.
17951 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17952 * config/arc/genmultilib.awk: Handle new mrf16 option.
17953 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17954 * config/arc/t-multilib: Regenerate.
17955 * doc/invoke.texi (ARC Options): Document mrf16 option.
17957 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17959 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17960 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17961 (arc_attribute_table): Add 'secure_call' attribute.
17962 (arc_print_operand): Print secure call operand.
17963 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17965 (arc_is_secure_call_p): New function. * config/arc/arc.md
17966 (call_i): Add support for sjli instruction.
17967 (call_value_i): Likewise.
17968 * config/arc/constraints.md (Csc): New constraint.
17970 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17971 John Eric Martin <John.Martin@emmicro-us.com>
17973 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17974 * config/arc/arc.c (_arc_jli_section): New struct.
17975 (arc_jli_section): New type.
17976 (rc_jli_sections): New static variable.
17977 (arc_handle_jli_attribute): New function.
17978 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17979 (arc_file_end): New function.
17980 (TARGET_ASM_FILE_END): Define.
17981 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17982 (arc_add_jli_section): New function.
17983 (jli_call_scan): Likewise.
17984 (arc_reorg): Call jli_call_scan.
17985 (arc_output_addsi): Remove 'S' from printing asm operand.
17986 (arc_is_jli_call_p): New function.
17987 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17989 (movhi_insn): Likewise.
17990 (movsi_insn): Likewise.
17991 (movsi_set_cc_insn): Likewise.
17992 (loadqi_update): Likewise.
17993 (load_zeroextendqisi_update): Likewise.
17994 (load_signextendqisi_update): Likewise.
17995 (loadhi_update): Likewise.
17996 (load_zeroextendhisi_update): Likewise.
17997 (load_signextendhisi_update): Likewise.
17998 (loadsi_update): Likewise.
17999 (loadsf_update): Likewise.
18000 (movsicc_insn): Likewise.
18001 (bset_insn): Likewise.
18002 (bxor_insn): Likewise.
18003 (bclr_insn): Likewise.
18004 (bmsk_insn): Likewise.
18005 (bicsi3_insn): Likewise.
18006 (cmpsi_cc_c_insn): Likewise.
18007 (movsi_ne): Likewise.
18008 (movsi_cond_exec): Likewise.
18009 (clrsbsi2): Likewise.
18010 (norm_f): Likewise.
18020 (call_i): Remove 'S' asm letter, add jli instruction.
18021 (call_value_i): Likewise.
18022 * config/arc/arc.op (mjli-always): New option.
18023 * config/arc/constraints.md (Cji): New constraint.
18024 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
18026 (subsf3_fpx): Likewise.
18027 (mulsf3_fpx): Likewise.
18028 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
18030 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
18031 function attrbutes.
18032 * doc/invoke.texi (ARC): Document mjli-always option.
18034 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
18036 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
18037 avoid addition with 0 and use incw and decw where possible.
18039 2018-01-26 Richard Biener <rguenther@suse.de>
18041 PR tree-optimization/81082
18042 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
18043 association if it requires casting to unsigned.
18044 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
18045 from fold_plusminus_mult_expr to catch important cases late when
18046 range info is available.
18048 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18050 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
18051 * configure.ac (hidden_linkonce): New test.
18052 * configure: Regenerate.
18053 * config.in: Regenerate.
18055 2018-01-26 Julia Koval <julia.koval@intel.com>
18057 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
18058 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
18059 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
18060 _mm_mask_bitshuffle_epi64_mask): Fix type.
18061 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
18062 USI_FTYPE_V4DI_V4DI_USI): Remove.
18063 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
18064 __builtin_ia32_vpshufbitqmb256_mask,
18065 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
18066 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
18067 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
18069 2018-01-26 Alan Modra <amodra@gmail.com>
18072 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
18073 UNSPEC_VBPERMQ. Sort other unspecs.
18075 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
18077 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
18079 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
18081 PR middle-end/83055
18082 * predict.c (drop_profile): Do not push/pop cfun; update also
18084 (handle_missing_profiles): Fix logic looking for zero profiles.
18086 2018-01-25 Jakub Jelinek <jakub@redhat.com>
18088 PR middle-end/83977
18089 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
18090 on functions with #pragma omp declare simd or functions with simd
18092 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
18093 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
18094 Remove trailing \n from warning_at calls.
18096 2018-01-25 Tom de Vries <tom@codesourcery.com>
18099 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18100 for neutered workers.
18102 2018-01-24 Joseph Myers <joseph@codesourcery.com>
18105 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
18106 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
18108 2018-01-24 Jeff Law <law@redhat.com>
18111 * i386.c (get_probe_interval): Move to earlier point.
18112 (ix86_compute_frame_layout): If -fstack-clash-protection and
18113 the frame is larger than the probe interval, then use pushes
18114 to save registers rather than reg->mem moves.
18115 (ix86_expand_prologue): Remove conditional for int_registers_saved
18118 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18121 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18122 min/max for never referenced object.
18124 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18126 PR middle-end/83977
18127 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18129 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18130 attributes from DECL_ATTRIBUTES (decl) without affecting
18131 DECL_ATTRIBUTES (current_function_decl).
18132 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18133 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18135 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18137 PR tree-optimization/83979
18138 * fold-const.c (fold_comparison): Use constant_boolean_node
18139 instead of boolean_{true,false}_node.
18141 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18143 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18146 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18148 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18149 Simplify the clause that sets the length attribute.
18150 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18151 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18152 clause that sets the length attribute.
18153 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18155 2018-01-24 Tom de Vries <tom@codesourcery.com>
18158 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18159 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18160 Add strict parameter.
18161 (prevent_branch_around_nothing): Insert dummy insn between branch to
18162 label and label with no ptx insn inbetween.
18163 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18165 2018-01-24 Tom de Vries <tom@codesourcery.com>
18168 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18169 for neutered threads in warp.
18170 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18172 2018-01-24 Richard Biener <rguenther@suse.de>
18174 PR tree-optimization/83176
18175 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18178 2018-01-24 Richard Biener <rguenther@suse.de>
18180 PR tree-optimization/82819
18181 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18182 code generating pluses that are no-ops in the target precision.
18184 2018-01-24 Richard Biener <rguenther@suse.de>
18186 PR middle-end/84000
18187 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18189 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18191 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18192 to merge probabilities.
18193 * predict.c (probably_never_executed): Also mark as cold functions
18194 with global 0 profile and guessed local profile.
18195 * profile-count.c (profile_probability::combine_with_count): New
18197 * profile-count.h (profile_probability::operator*,
18198 profile_probability::operator*=, profile_probability::operator/,
18199 profile_probability::operator/=): Reduce precision to adjusted
18200 and set value to guessed on contradictory divisions.
18201 (profile_probability::combine_with_freq): Remove.
18202 (profile_probability::combine_wiht_count): Declare.
18203 (profile_count::force_nonzero):: Set to adjusted.
18204 (profile_count::probability_in):: Set quality to adjusted.
18205 * tree-ssa-tail-merge.c (replace_block_by): Use
18206 combine_with_count.
18208 2018-01-23 Andrew Waterman <andrew@sifive.com>
18209 Jim Wilson <jimw@sifive.com>
18211 * config/riscv/riscv.c (riscv_stack_boundary): New.
18212 (riscv_option_override): Set riscv_stack_boundary. Handle
18213 riscv_preferred_stack_boundary_arg.
18214 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18215 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18216 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18217 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18218 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18219 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18221 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18224 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18225 of struct ix86_frame.
18226 (ix86_expand_epilogue): Likewise. Add a local variable for
18227 the reg_save_offset field in struct ix86_frame.
18229 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18231 PR tree-optimization/82604
18232 * tree-loop-distribution.c (enum partition_kind): New enum item
18233 PKIND_PARTIAL_MEMSET.
18234 (partition_builtin_p): Support above new enum item.
18235 (generate_code_for_partition): Ditto.
18236 (compute_access_range): Differentiate cases that equality can be
18237 proven at all loops, the innermost loops or no loops.
18238 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18239 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18240 (finalize_partitions, distribute_loop): Don't fuse partition of
18241 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18242 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18243 parloop is enabled.
18245 2018-01-23 Martin Liska <mliska@suse.cz>
18247 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18248 order to ignore the predictor.
18249 (PRED_POLYMORPHIC_CALL): Likewise.
18250 (PRED_RECURSIVE_CALL): Likewise.
18252 2018-01-23 Martin Liska <mliska@suse.cz>
18254 * tree-profile.c (tree_profiling): Print function header to
18255 aware reader which function we are working on.
18256 * value-prof.c (gimple_find_values_to_profile): Do not print
18257 not interesting value histograms.
18259 2018-01-23 Martin Liska <mliska@suse.cz>
18261 * profile-count.h (enum profile_quality): Add
18262 profile_uninitialized as the first value. Do not number values
18263 as they are zero based.
18264 (profile_count::verify): Update sanity check.
18265 (profile_probability::verify): Likewise.
18267 2018-01-23 Nathan Sidwell <nathan@acm.org>
18269 * doc/invoke.texi (ffor-scope): Deprecate.
18271 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18273 PR tree-optimization/83510
18274 * domwalk.c (set_all_edges_as_executable): New function.
18275 (dom_walker::dom_walker): Convert bool param
18276 "skip_unreachable_blocks" to enum reachability. Move setup of
18277 edge flags to set_all_edges_as_executable and only do it when
18278 reachability is REACHABLE_BLOCKS.
18279 * domwalk.h (enum dom_walker::reachability): New enum.
18280 (dom_walker::dom_walker): Convert bool param
18281 "skip_unreachable_blocks" to enum reachability.
18282 (set_all_edges_as_executable): New decl.
18283 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18284 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18286 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18287 but converting true to REACHABLE_BLOCKS.
18288 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18290 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18291 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18292 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18294 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18295 if check_all_array_refs will be called.
18297 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18299 * tree.c (selftest::test_location_wrappers): Add more test
18302 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18304 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18305 (selftest::test_bit_in_range): Likewise.
18307 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18310 * doc/sourcebuild.texi (vect_float): Say that the selector
18311 only describes the situation when -funsafe-math-optimizations is on.
18312 (vect_float_strict): Document.
18314 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18316 PR tree-optimization/83965
18317 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18318 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18319 instead of checking only for a reduction.
18320 (vect_recog_widen_sum_pattern): Likewise.
18322 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18324 * predict.c (probably_never_executed): Only use precise profile info.
18325 (compute_function_frequency): Skip after inlining hack since we now
18326 have quality checking.
18328 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18330 * profile-count.h (profile_probability::very_unlikely,
18331 profile_probability::unlikely, profile_probability::even): Set
18332 precision to guessed.
18334 2018-01-23 Richard Biener <rguenther@suse.de>
18336 PR tree-optimization/83963
18337 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18338 Properly terminate dominator walk when crossing the exit edge not
18339 when visiting its source block.
18341 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18344 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18345 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18347 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18349 PR tree-optimization/83957
18350 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18351 semicolon after for body surrounded by braces.
18353 PR tree-optimization/83081
18354 * profile-count.h (profile_probability::split): New method.
18355 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18356 Use profile_probability::split.
18357 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18358 when splitting a single conditional jump into 2.
18360 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18362 PR tree-optimization/69452
18363 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18366 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18368 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18369 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18370 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18372 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18374 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18376 * config/rl78/rl78.md (movdi): New define_expand.
18377 * config/rl78/rl78.c (rl78_split_movdi): New function.
18379 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18382 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18384 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18385 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18386 128-bit to produce an UNSPEC move to get the double word with the
18387 signbit and then a shift directly to do signbit.
18388 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18389 implementation with a new version that just does either a direct
18390 move or a regular move. Move memory interface to separate insns.
18391 Move insns so they are next to the expander.
18392 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18393 with signbit move. Split big and little endian case.
18394 (signbit<mode>2_dm_mem_le): Likewise.
18395 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18396 (signbit<mode>2_dm2): Likewise.
18398 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18400 * config/rl78/rl78.md (anddi3): New define_expand.
18402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18404 * config/rl78/rl78.md (umindi3): New define_expand.
18406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18408 * config/rl78/rl78.md (smindi3): New define_expand.
18410 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18412 * config/rl78/rl78.md (smaxdi3): New define_expand.
18414 2018-01-22 Carl Love <cel@us.ibm.com>
18416 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18417 LVX_V1TI): Add macro expansion.
18418 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18419 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18420 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18421 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18422 Change check to determine if the instruction is a byte reversing
18423 entry. Fix typo in comment.
18424 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18425 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18426 Add def_builtin calls for new builtins.
18427 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18428 Add define_insn expansion.
18430 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18432 * config/rl78/rl78.md (umaxdi3): New define_expand.
18434 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18436 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18437 for non-QImode registers.
18439 2018-01-22 Richard Biener <rguenther@suse.de>
18441 PR tree-optimization/83963
18442 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18443 including the loop exit block.
18444 (scop_detection::merge_sese): Likewise.
18445 (scop_detection::add_scop): Do it here instead.
18447 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18449 * doc/sourcebuild.texi (arm_softfloat): Document.
18451 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18454 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18455 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18456 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18458 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18459 David Edelsohn <dje.gcc@gmail.com>
18462 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18463 Change "crset eq" to "crset 2".
18464 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18465 (*call_indirect_aix<mode>_nospec): Likewise.
18466 (*call_value_indirect_aix<mode>_nospec): Likewise.
18467 (*call_indirect_elfv2<mode>_nospec): Likewise.
18468 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18469 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18470 change assembly output from . to $.
18471 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18472 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18473 (*tablejump<mode>_internal1_nospec): Likewise.
18475 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18478 * config/sh/sh_optimize_sett_clrt.cc:
18479 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18481 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18483 PR tree-optimization/83940
18484 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18485 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18486 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18487 use it to pass back the definition type.
18488 (vect_check_store_rhs): Likewise rhs_dt_out.
18489 (vect_build_gather_load_calls): Add a mask_dt argument and use
18490 it instead of a call to vect_is_simple_use.
18491 (vectorizable_store): Update calls to vect_check_load_store_mask
18492 and vect_check_store_rhs. Use the dt returned by the latter instead
18493 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18494 instead of calls to vect_is_simple_use. Pass the scalar rather
18495 than the vector operand to vect_is_simple_use when handling
18496 second and subsequent copies of an rhs value.
18497 (vectorizable_load): Update calls to vect_check_load_store_mask
18498 and vect_build_gather_load_calls. Use the cached mask_dt and
18499 gs_info.offset_dt instead of calls to vect_is_simple_use.
18501 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18503 PR middle-end/83945
18504 * tree-emutls.c: Include gimplify.h.
18505 (lower_emutls_2): New function.
18506 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18507 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18508 it before further processing.
18511 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18512 UINTVAL (trueop1) instead of INTVAL (op1).
18514 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18518 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18519 INCOMING_FRAME_SP_OFFSET if not defined.
18520 (scan_trace): Add ENTRY argument. If true and
18521 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18522 emit a note to adjust the CFA offset.
18523 (create_cfi_notes): Adjust scan_trace callers.
18524 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18525 INCOMING_FRAME_SP_OFFSET in the CIE.
18526 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18527 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18529 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18530 * doc/tm.texi: Regenerated.
18532 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18534 PR rtl-optimization/83147
18535 * lra-constraints.c (remove_inheritance_pseudos): Use
18536 lra_substitute_pseudo_within_insn.
18538 2018-01-19 Tom de Vries <tom@codesourcery.com>
18539 Cesar Philippidis <cesar@codesourcery.com>
18542 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18544 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18547 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18548 spaces for function labels.
18550 2018-01-19 Martin Liska <mliska@suse.cz>
18552 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18553 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18554 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18555 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18556 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18557 (PRED_CONST_RETURN): Change from 69 to 65.
18558 (PRED_NULL_RETURN): Change from 91 to 71.
18559 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18560 (PRED_LOOP_GUARD): Change from 66 to 73.
18562 2018-01-19 Martin Liska <mliska@suse.cz>
18564 * predict.c (predict_insn_def): Add new assert.
18565 (struct branch_predictor): Change type to signed integer.
18566 (test_prediction_value_range): Amend test to cover
18567 PROB_UNINITIALIZED.
18568 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18569 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18570 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18571 (PRED_LOOP_IV_COMPARE): Likewise.
18572 * predict.h (PROB_UNINITIALIZED): Define new constant.
18574 2018-01-19 Martin Liska <mliska@suse.cz>
18576 * predict.c (dump_prediction): Add new format for
18577 analyze_brprob.py script which is enabled with -details
18579 * profile-count.h (precise_p): New function.
18581 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18583 PR tree-optimization/83922
18584 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18585 there are no statements that need masking.
18586 (vect_active_double_reduction_p): New function.
18587 (vect_analyze_loop_operations): Use it when handling phis that
18588 are not in the loop header.
18590 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18592 PR tree-optimization/83914
18593 * tree-vect-loop.c (vectorizable_induction): Don't convert
18594 init_expr or apply the peeling adjustment for inductions
18595 that are nested within the vectorized loop.
18597 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18599 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18602 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18606 * function.h (gimplify_parameters): Add gimple_seq * argument.
18607 * function.c: Include gimple.h and options.h.
18608 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18609 for the added local temporaries if needed.
18610 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18611 if there are any parameter cleanups, wrap whole body into a
18612 try/finally with the cleanups.
18614 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18617 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18618 Use GET_MODE_CLASS for scalar floating point.
18620 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18624 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18625 Fix call of call_cgraph_insertion_hooks.
18627 2018-01-18 Martin Sebor <msebor@redhat.com>
18629 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18631 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18634 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18637 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18640 * common.opt: (-ffile-prefix-map): New option.
18641 * opts.c (common_handle_option): Defer it.
18642 * opts-global.c (handle_common_deferred_options): Handle it.
18643 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18644 * file-prefix-map.h: New file.
18645 (remap_debug_filename, add_debug_prefix_map): ...here.
18646 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18647 * final.c (debug_prefix_map, add_debug_prefix_map
18648 remap_debug_filename): Move to...
18649 * file-prefix-map.c: New file.
18650 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18651 generalize, get rid of alloca(), use strrchr() instead of strchr().
18652 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18653 Implement in terms of add_prefix_map().
18654 (remap_macro_filename, remap_debug_filename): Implement in term of
18656 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18657 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18658 * dbxout.c: Include file-prefix-map.h.
18659 * varasm.c: Likewise.
18660 * vmsdbgout.c: Likewise.
18661 * xcoffout.c: Likewise.
18662 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18663 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18664 * doc/invoke.texi (-ffile-prefix-map): Document.
18665 (-fdebug-prefix-map): Update description.
18667 2018-01-18 Martin Liska <mliska@suse.cz>
18669 * config/i386/i386.c (indirect_thunk_name): Document that also
18671 (output_indirect_thunk): Document why both instructions
18672 (pause and lfence) are generated.
18674 2018-01-18 Richard Biener <rguenther@suse.de>
18676 PR tree-optimization/83887
18677 * graphite-scop-detection.c
18678 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18679 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18680 (scop_detection::merge_sese): Re-implement with a flood-fill
18681 algorithm that properly finds a SESE region if it exists.
18683 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18686 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18687 pointer_diff optimizations use view_convert instead of convert.
18689 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18691 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18692 Generate different code for -mno-speculate-indirect-jumps.
18693 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18694 (*call_indirect_aix<mode>): Disable for
18695 -mno-speculate-indirect-jumps.
18696 (*call_indirect_aix<mode>_nospec): New define_insn.
18697 (*call_value_indirect_aix<mode>): Disable for
18698 -mno-speculate-indirect-jumps.
18699 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18700 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18701 -mno-speculate-indirect-jumps.
18702 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18704 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18706 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18707 long double type, set the flags for noting the default long double
18708 type, even if we don't pass or return a long double type.
18710 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18713 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18716 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18718 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18719 support for merge[hl].
18720 (fold_mergehl_helper): New helper function.
18721 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18722 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18723 (altivec_vmrglw_direct): Add xxmrglw insn.
18725 2018-01-17 Andrew Waterman <andrew@sifive.com>
18727 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18728 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18730 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18733 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18734 call the lto_location_cache before reading the
18735 DECL_SOURCE_LOCATION of the types.
18737 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18738 Richard Sandiford <richard.sandiford@linaro.org>
18740 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18741 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18742 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18743 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18744 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18746 * config/aarch64/constraints.md (aarch64_movti_operand):
18748 * config/aarch64/predicates.md (Uti): Add new constraint.
18750 2018-01-17 Carl Love <cel@us.ibm.com>
18752 * config/rs6000/vsx.md (define_expand xl_len_r,
18753 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18754 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18756 (define_expand, define_insn): Move the shift left from the
18757 define_insn to the define_expand for lxvl and stxvl instructions.
18758 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18759 and XL_LEN_R definitions to PURE.
18761 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18763 * config/i386/i386.c (indirect_thunk_name): Declare regno
18764 as unsigned int. Compare regno with INVALID_REGNUM.
18765 (output_indirect_thunk): Ditto.
18766 (output_indirect_thunk_function): Ditto.
18767 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18768 in the call to output_indirect_thunk_function.
18770 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18772 PR middle-end/83884
18773 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18774 rather than the size of inner_type to determine the stack slot size
18775 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18777 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18780 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18783 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18785 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18786 endian Linux systems to optionally enable multilibs for selecting
18787 the long double type if the user configured an explicit type.
18788 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18789 have no long double multilibs if not defined.
18790 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18791 warn if the user used -mabi={ieee,ibm}longdouble and we built
18792 multilibs for long double.
18793 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18794 appropriate multilib option.
18795 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18797 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18798 for building long double multilibs.
18799 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18801 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18803 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18806 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18808 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18811 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18814 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18817 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18819 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18820 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18822 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18824 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18825 different rtl trees depending on TARGET_64BIT.
18826 (rs6000_gen_lvx): Likewise.
18828 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18830 * config/visium/visium.md (nop): Tweak comment.
18831 (hazard_nop): Likewise.
18833 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18835 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18836 -mspeculate-indirect-jumps.
18837 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18838 for -mno-speculate-indirect-jumps.
18839 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18840 (*call_value_indirect_elfv2<mode>): Disable for
18841 -mno-speculate-indirect-jumps.
18842 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18843 (indirect_jump): Emit different RTL for
18844 -mno-speculate-indirect-jumps.
18845 (*indirect_jump<mode>): Disable for
18846 -mno-speculate-indirect-jumps.
18847 (*indirect_jump<mode>_nospec): New define_insn.
18848 (tablejump): Emit different RTL for
18849 -mno-speculate-indirect-jumps.
18850 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18851 (tablejumpsi_nospec): New define_expand.
18852 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18853 (tablejumpdi_nospec): New define_expand.
18854 (*tablejump<mode>_internal1): Disable for
18855 -mno-speculate-indirect-jumps.
18856 (*tablejump<mode>_internal1_nospec): New define_insn.
18857 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18860 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18862 * caller-save.c (insert_save): Drop unnecessary parameter. All
18865 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18866 Richard Biener <rguenth@suse.de>
18869 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18870 return early, inline manually is_gimple_sizepos. Make sure if we
18871 call gimplify_expr we don't end up with a gimple constant.
18872 * tree.c (variably_modified_type_p): Don't return true for
18873 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18874 * gimplify.h (is_gimple_sizepos): Remove.
18876 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18878 PR tree-optimization/83857
18879 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18880 vectorizable_live_operation for pure SLP statements.
18881 (vectorizable_live_operation): Handle PHIs.
18883 2018-01-16 Richard Biener <rguenther@suse.de>
18885 PR tree-optimization/83867
18886 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18887 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18889 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18892 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18893 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18894 If off is not INTEGER_CST, issue a may not be aligned warning
18895 rather than isn't aligned. Use isn%'t rather than isn't.
18896 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18898 <case MULT_EXPR>: Improve the case when bottom and one of the
18899 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18900 operand, in that case check if the other operand is multiple of
18901 bottom divided by the INTEGER_CST operand.
18903 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18906 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18907 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18908 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18909 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18910 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18911 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18912 (pa_function_arg_size): New function.
18914 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18916 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18917 in a separate statement.
18919 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18921 PR tree-optimization/83847
18922 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18923 group gathers and scatters.
18925 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18927 PR rtl-optimization/86620
18928 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18930 PR rtl-optimization/83213
18931 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18932 to last if both are JUMP_INSNs.
18934 PR tree-optimization/83843
18935 * gimple-ssa-store-merging.c
18936 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18937 store_immediate_info for bswap/nop orig_stores.
18939 2018-01-15 Andrew Waterman <andrew@sifive.com>
18941 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18943 <UDIV>: Increase cost if !TARGET_DIV.
18945 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18947 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18948 (define_attr "cr_logical_3op"): New.
18949 (cceq_ior_compare): Adjust.
18950 (cceq_ior_compare_complement): Adjust.
18951 (*cceq_rev_compare): Adjust.
18952 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18953 (is_cracked_insn): Adjust.
18954 (insn_must_be_first_in_group): Adjust.
18955 * config/rs6000/40x.md: Adjust.
18956 * config/rs6000/440.md: Adjust.
18957 * config/rs6000/476.md: Adjust.
18958 * config/rs6000/601.md: Adjust.
18959 * config/rs6000/603.md: Adjust.
18960 * config/rs6000/6xx.md: Adjust.
18961 * config/rs6000/7450.md: Adjust.
18962 * config/rs6000/7xx.md: Adjust.
18963 * config/rs6000/8540.md: Adjust.
18964 * config/rs6000/cell.md: Adjust.
18965 * config/rs6000/e300c2c3.md: Adjust.
18966 * config/rs6000/e500mc.md: Adjust.
18967 * config/rs6000/e500mc64.md: Adjust.
18968 * config/rs6000/e5500.md: Adjust.
18969 * config/rs6000/e6500.md: Adjust.
18970 * config/rs6000/mpc.md: Adjust.
18971 * config/rs6000/power4.md: Adjust.
18972 * config/rs6000/power5.md: Adjust.
18973 * config/rs6000/power6.md: Adjust.
18974 * config/rs6000/power7.md: Adjust.
18975 * config/rs6000/power8.md: Adjust.
18976 * config/rs6000/power9.md: Adjust.
18977 * config/rs6000/rs64.md: Adjust.
18978 * config/rs6000/titan.md: Adjust.
18980 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18982 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18983 ix86_indirect_branch_register logic.
18985 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18987 * config/i386/constraints.md (Bs): Update
18988 ix86_indirect_branch_register check. Don't check
18989 ix86_indirect_branch_register with GOT_memory_operand.
18991 * config/i386/predicates.md (GOT_memory_operand): Don't check
18992 ix86_indirect_branch_register here.
18993 (GOT32_symbol_operand): Likewise.
18995 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18997 * config/i386/predicates.md (constant_call_address_operand):
18998 Rewrite ix86_indirect_branch_register logic.
18999 (sibcall_insn_operand): Likewise.
19001 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19003 * config/i386/constraints.md (Bs): Replace
19004 ix86_indirect_branch_thunk_register with
19005 ix86_indirect_branch_register.
19007 * config/i386/i386.md (indirect_jump): Likewise.
19008 (tablejump): Likewise.
19009 (*sibcall_memory): Likewise.
19010 (*sibcall_value_memory): Likewise.
19011 Peepholes of indirect call and jump via memory: Likewise.
19012 * config/i386/i386.opt: Likewise.
19013 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19014 (GOT_memory_operand): Likewise.
19015 (call_insn_operand): Likewise.
19016 (sibcall_insn_operand): Likewise.
19017 (GOT32_symbol_operand): Likewise.
19019 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19021 PR middle-end/83837
19022 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
19023 type rather than type addr's type points to.
19024 (expand_omp_atomic_mutex): Likewise.
19025 (expand_omp_atomic): Likewise.
19027 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19030 * config/i386/i386.c (output_indirect_thunk_function): Use
19031 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
19032 for __x86_return_thunk.
19034 2018-01-15 Richard Biener <rguenther@suse.de>
19036 PR middle-end/83850
19037 * expmed.c (extract_bit_field_1): Fix typo.
19039 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19042 * config/arm/iterators.md (VF): New mode iterator.
19043 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
19044 Remove integer-related logic from pattern.
19045 (neon_vabd<mode>_3): Likewise.
19047 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19049 PR middle-end/82694
19050 * common.opt (fstrict-overflow): No longer an alias.
19051 (fwrapv-pointer): New option.
19052 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
19053 also for pointer types based on flag_wrapv_pointer.
19054 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
19055 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
19056 opts->x_flag_wrapv got set.
19057 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
19058 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
19059 POINTER_TYPE_OVERFLOW_UNDEFINED.
19060 * match.pd: Likewise in address comparison pattern.
19061 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
19063 2018-01-15 Richard Biener <rguenther@suse.de>
19066 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
19067 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
19068 Reset type names to their identifier if their TYPE_DECL doesn't
19069 have linkage (and thus is used for ODR and devirt).
19070 (save_debug_info_for_decl): Remove.
19071 (save_debug_info_for_type): Likewise.
19072 (add_tree_to_fld_list): Adjust.
19073 * tree-pretty-print.c (dump_generic_node): Make dumping of
19074 type names more robust.
19076 2018-01-15 Richard Biener <rguenther@suse.de>
19078 * BASE-VER: Bump to 8.0.1.
19080 2018-01-14 Martin Sebor <msebor@redhat.com>
19083 * builtins.c (check_access): Avoid warning when the no-warning bit
19086 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
19088 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
19089 * ira-color (allocno_hard_regs_compare): Likewise.
19091 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
19094 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
19095 Use .pushsection/.popsection.
19097 2018-01-14 Martin Sebor <msebor@redhat.com>
19100 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
19102 2018-01-14 Jakub Jelinek <jakub@redhat.com>
19104 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
19105 entry from extra_headers.
19106 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
19107 extra_headers, make the list bitwise identical to the i?86-*-* one.
19109 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19111 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
19112 -mcmodel=large with -mindirect-branch=thunk,
19113 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19114 -mfunction-return=thunk-extern.
19115 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19116 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19117 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19119 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19121 * config/i386/i386.c (print_reg): Print the name of the full
19122 integer register without '%'.
19123 (ix86_print_operand): Handle 'V'.
19124 * doc/extend.texi: Document 'V' modifier.
19126 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19128 * config/i386/constraints.md (Bs): Disallow memory operand for
19129 -mindirect-branch-register.
19131 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19132 (GOT_memory_operand): Likewise.
19133 (call_insn_operand): Likewise.
19134 (sibcall_insn_operand): Likewise.
19135 (GOT32_symbol_operand): Likewise.
19136 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19137 for -mindirect-branch-register.
19138 (tablejump): Likewise.
19139 (*sibcall_memory): Likewise.
19140 (*sibcall_value_memory): Likewise.
19141 Disallow peepholes of indirect call and jump via memory for
19142 -mindirect-branch-register.
19143 (*call_pop): Replace m with Bw.
19144 (*call_value_pop): Likewise.
19145 (*sibcall_pop_memory): Replace m with Bs.
19146 * config/i386/i386.opt (mindirect-branch-register): New option.
19147 * doc/invoke.texi: Document -mindirect-branch-register option.
19149 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19151 * config/i386/i386-protos.h (ix86_output_function_return): New.
19152 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19153 set function_return_type.
19154 (indirect_thunk_name): Add ret_p to indicate thunk for function
19156 (output_indirect_thunk_function): Pass false to
19157 indirect_thunk_name.
19158 (ix86_output_indirect_branch_via_reg): Likewise.
19159 (ix86_output_indirect_branch_via_push): Likewise.
19160 (output_indirect_thunk_function): Create alias for function
19161 return thunk if regno < 0.
19162 (ix86_output_function_return): New function.
19163 (ix86_handle_fndecl_attribute): Handle function_return.
19164 (ix86_attribute_table): Add function_return.
19165 * config/i386/i386.h (machine_function): Add
19166 function_return_type.
19167 * config/i386/i386.md (simple_return_internal): Use
19168 ix86_output_function_return.
19169 (simple_return_internal_long): Likewise.
19170 * config/i386/i386.opt (mfunction-return=): New option.
19171 (indirect_branch): Mention -mfunction-return=.
19172 * doc/extend.texi: Document function_return function attribute.
19173 * doc/invoke.texi: Document -mfunction-return= option.
19175 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19177 * config/i386/i386-opts.h (indirect_branch): New.
19178 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19179 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19180 with local indirect jump when converting indirect call and jump.
19181 (ix86_set_indirect_branch_type): New.
19182 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19183 (indirectlabelno): New.
19184 (indirect_thunk_needed): Likewise.
19185 (indirect_thunk_bnd_needed): Likewise.
19186 (indirect_thunks_used): Likewise.
19187 (indirect_thunks_bnd_used): Likewise.
19188 (INDIRECT_LABEL): Likewise.
19189 (indirect_thunk_name): Likewise.
19190 (output_indirect_thunk): Likewise.
19191 (output_indirect_thunk_function): Likewise.
19192 (ix86_output_indirect_branch_via_reg): Likewise.
19193 (ix86_output_indirect_branch_via_push): Likewise.
19194 (ix86_output_indirect_branch): Likewise.
19195 (ix86_output_indirect_jmp): Likewise.
19196 (ix86_code_end): Call output_indirect_thunk_function if needed.
19197 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19199 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19200 (ix86_attribute_table): Add indirect_branch.
19201 * config/i386/i386.h (machine_function): Add indirect_branch_type
19202 and has_local_indirect_jump.
19203 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19205 (tablejump): Likewise.
19206 (*indirect_jump): Use ix86_output_indirect_jmp.
19207 (*tablejump_1): Likewise.
19208 (simple_return_indirect_internal): Likewise.
19209 * config/i386/i386.opt (mindirect-branch=): New option.
19210 (indirect_branch): New.
19213 (thunk-inline): Likewise.
19214 (thunk-extern): Likewise.
19215 * doc/extend.texi: Document indirect_branch function attribute.
19216 * doc/invoke.texi: Document -mindirect-branch= option.
19218 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19221 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19223 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19225 * ipa-inline.c (want_inline_small_function_p): Return false if
19226 inlining has already failed with CIF_FINAL_ERROR.
19227 (update_caller_keys): Call want_inline_small_function_p before
19229 (update_callee_keys): Likewise.
19231 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19233 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19235 (rs6000_quadword_masked_address_p): Likewise.
19236 (quad_aligned_load_p): Likewise.
19237 (quad_aligned_store_p): Likewise.
19238 (const_load_sequence_p): Add comment to describe the outer-most loop.
19239 (mimic_memory_attributes_and_flags): New function.
19240 (rs6000_gen_stvx): Likewise.
19241 (replace_swapped_aligned_store): Likewise.
19242 (rs6000_gen_lvx): Likewise.
19243 (replace_swapped_aligned_load): Likewise.
19244 (replace_swapped_load_constant): Capitalize argument name in
19245 comment describing this function.
19246 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19247 and stores that access quad-word aligned addresses and replace
19248 with stvx or lvx instructions when appropriate.
19249 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19250 New function prototype.
19251 (rs6000_quadword_masked_address_p): Likewise.
19252 (rs6000_gen_lvx): Likewise.
19253 (rs6000_gen_stvx): Likewise.
19254 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19255 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19256 when memory address is aligned.
19257 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19258 this split to select lvx instruction when memory address is aligned.
19259 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19260 instruction when memory address is aligned.
19261 (*vsx_le_perm_load_v16qi): Likewise.
19262 (four unnamed splitters): Modify to select the stvx instruction
19263 when memory is aligned.
19265 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19267 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19268 which appears in the queue multiple times.
19270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19271 Alan Hayward <alan.hayward@arm.com>
19272 David Sherwood <david.sherwood@arm.com>
19274 * tree-vectorizer.h (vec_lower_bound): New structure.
19275 (_loop_vec_info): Add check_nonzero and lower_bounds.
19276 (LOOP_VINFO_CHECK_NONZERO): New macro.
19277 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19278 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19279 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19280 fields. Make seg_len the distance travelled, not including the
19282 (dr_direction_indicator): Declare.
19283 (dr_zero_step_indicator): Likewise.
19284 (dr_known_forward_stride_p): Likewise.
19285 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19287 (runtime_alias_check_p): Allow runtime alias checks with
19289 (operator ==): Compare access_size and align.
19290 (prune_runtime_alias_test_list): Rework for new distinction between
19291 the access_size and seg_len.
19292 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19294 (get_segment_min_max): New function.
19295 (create_intersect_range_checks): Use it.
19296 (dr_step_indicator): New function.
19297 (dr_direction_indicator): Likewise.
19298 (dr_zero_step_indicator): Likewise.
19299 (dr_known_forward_stride_p): Likewise.
19300 * tree-loop-distribution.c (data_ref_segment_size): Return
19301 DR_STEP * (niters - 1).
19302 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19304 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19305 (vect_preserves_scalar_order_p): New function, split out from...
19306 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19307 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19308 (vect_vfa_access_size): New function.
19309 (vect_vfa_align): Likewise.
19310 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19311 (dump_lower_bound): New function.
19312 (vect_check_lower_bound): Likewise.
19313 (vect_small_gap_p): Likewise.
19314 (vectorizable_with_step_bound_p): Likewise.
19315 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19316 depencies if the vectorization factor is 1. Convert the checks
19317 for nonzero steps into checks on the bounds of DR_STEP. Try using
19318 a bunds check for variable steps if the minimum required step is
19319 relatively small. Update calls to the dr_with_seg_len
19320 constructor and to vect_compile_time_alias.
19321 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19323 (vect_loop_versioning): Call it.
19324 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19326 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19329 Alan Hayward <alan.hayward@arm.com>
19330 David Sherwood <david.sherwood@arm.com>
19332 * doc/sourcebuild.texi (vect_scatter_store): Document.
19333 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19335 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19337 * genopinit.c (main): Add supports_vec_scatter_store and
19338 supports_vec_scatter_store_cached to target_optabs.
19339 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19340 IFN_MASK_SCATTER_STORE.
19341 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19343 * internal-fn.h (internal_store_fn_p): Declare.
19344 (internal_fn_stored_value_index): Likewise.
19345 * internal-fn.c (scatter_store_direct): New macro.
19346 (expand_scatter_store_optab_fn): New function.
19347 (direct_scatter_store_optab_supported_p): New macro.
19348 (internal_store_fn_p): New function.
19349 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19350 IFN_MASK_SCATTER_STORE.
19351 (internal_fn_mask_index): Likewise.
19352 (internal_fn_stored_value_index): New function.
19353 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19354 for scatter stores.
19355 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19356 * optabs-query.c (supports_vec_scatter_store_p): New function.
19357 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19358 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19359 true for scatter stores.
19360 (vect_gather_scatter_fn_p): Handle scatter stores too.
19361 (vect_check_gather_scatter): Consider using scatter stores if
19362 supports_vec_scatter_store_p.
19363 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19364 scatter stores too.
19365 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19366 internal_fn_stored_value_index.
19367 (check_load_store_masking): Handle scatter stores too.
19368 (vect_get_store_rhs): Make public.
19369 (vectorizable_call): Use internal_store_fn_p.
19370 (vectorizable_store): Handle scatter store internal functions.
19371 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19372 when deciding whether the end of the group has been reached.
19373 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19374 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19375 (mask_scatter_store<mode>): New insns.
19377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19378 Alan Hayward <alan.hayward@arm.com>
19379 David Sherwood <david.sherwood@arm.com>
19381 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19382 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19383 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19385 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19386 Use vect_truncate_gather_scatter_offset if we can't treat the
19387 operation as a normal gather load or scatter store.
19388 (get_group_load_store_type): Take the gather_scatter_info
19389 as argument. Try using a gather load or scatter store for
19390 single-element groups.
19391 (get_load_store_type): Update calls to get_group_load_store_type
19392 and vect_use_strided_gather_scatters_p.
19394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19395 Alan Hayward <alan.hayward@arm.com>
19396 David Sherwood <david.sherwood@arm.com>
19398 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19399 optional tree argument.
19400 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19402 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19403 but continue to use the current value as a fallback.
19404 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19405 to compare the updates.
19406 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19407 (get_load_store_type): Use it when handling a strided access.
19408 (vect_get_strided_load_store_ops): New function.
19409 (vect_get_data_ptr_increment): Likewise.
19410 (vectorizable_load): Handle strided gather loads. Always pass
19411 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19414 Alan Hayward <alan.hayward@arm.com>
19415 David Sherwood <david.sherwood@arm.com>
19417 * doc/md.texi (gather_load@var{m}): Document.
19418 (mask_gather_load@var{m}): Likewise.
19419 * genopinit.c (main): Add supports_vec_gather_load and
19420 supports_vec_gather_load_cached to target_optabs.
19421 * optabs-tree.c (init_tree_optimization_optabs): Use
19422 ggc_cleared_alloc to allocate target_optabs.
19423 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19424 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19426 * internal-fn.h (internal_load_fn_p): Declare.
19427 (internal_gather_scatter_fn_p): Likewise.
19428 (internal_fn_mask_index): Likewise.
19429 (internal_gather_scatter_fn_supported_p): Likewise.
19430 * internal-fn.c (gather_load_direct): New macro.
19431 (expand_gather_load_optab_fn): New function.
19432 (direct_gather_load_optab_supported_p): New macro.
19433 (direct_internal_fn_optab): New function.
19434 (internal_load_fn_p): Likewise.
19435 (internal_gather_scatter_fn_p): Likewise.
19436 (internal_fn_mask_index): Likewise.
19437 (internal_gather_scatter_fn_supported_p): Likewise.
19438 * optabs-query.c (supports_at_least_one_mode_p): New function.
19439 (supports_vec_gather_load_p): Likewise.
19440 * optabs-query.h (supports_vec_gather_load_p): Declare.
19441 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19442 and memory_type field.
19443 (NUM_PATTERNS): Bump to 15.
19444 * tree-vect-data-refs.c: Include internal-fn.h.
19445 (vect_gather_scatter_fn_p): New function.
19446 (vect_describe_gather_scatter_call): Likewise.
19447 (vect_check_gather_scatter): Try using internal functions for
19448 gather loads. Recognize existing calls to a gather load function.
19449 (vect_analyze_data_refs): Consider using gather loads if
19450 supports_vec_gather_load_p.
19451 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19452 (vect_get_gather_scatter_offset_type): Likewise.
19453 (vect_convert_mask_for_vectype): Likewise.
19454 (vect_add_conversion_to_patterm): Likewise.
19455 (vect_try_gather_scatter_pattern): Likewise.
19456 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19457 (vect_vect_recog_func_ptrs): Add it.
19458 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19459 internal_fn_mask_index and internal_gather_scatter_fn_p.
19460 (check_load_store_masking): Take the gather_scatter_info as an
19461 argument and handle gather loads.
19462 (vect_get_gather_scatter_ops): New function.
19463 (vectorizable_call): Check internal_load_fn_p.
19464 (vectorizable_load): Likewise. Handle gather load internal
19466 (vectorizable_store): Update call to check_load_store_masking.
19467 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19468 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19469 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19470 (aarch64_gather_scale_operand_d): New predicates.
19471 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19472 (mask_gather_load<mode>): New insns.
19474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19475 Alan Hayward <alan.hayward@arm.com>
19476 David Sherwood <david.sherwood@arm.com>
19478 * optabs.def (fold_left_plus_optab): New optab.
19479 * doc/md.texi (fold_left_plus_@var{m}): Document.
19480 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19481 * internal-fn.c (fold_left_direct): Define.
19482 (expand_fold_left_optab_fn): Likewise.
19483 (direct_fold_left_optab_supported_p): Likewise.
19484 * fold-const-call.c (fold_const_fold_left): New function.
19485 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19486 * tree-parloops.c (valid_reduction_p): New function.
19487 (gather_scalar_reductions): Use it.
19488 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19489 (vect_finish_replace_stmt): Declare.
19490 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19491 (needs_fold_left_reduction_p): New function, split out from...
19492 (vect_is_simple_reduction): ...here. Accept reductions that
19493 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19494 (vect_force_simple_reduction): Also store the reduction type in
19495 the assignment's STMT_VINFO_REDUC_TYPE.
19496 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19497 (merge_with_identity): New function.
19498 (vect_expand_fold_left): Likewise.
19499 (vectorize_fold_left_reduction): Likewise.
19500 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19501 scalar phi in place for it. Check for target support and reject
19502 cases that would reassociate the operation. Defer the transform
19503 phase to vectorize_fold_left_reduction.
19504 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19505 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19506 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19510 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19511 call to ifc_temp_var.
19513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19514 Alan Hayward <alan.hayward@arm.com>
19515 David Sherwood <david.sherwood@arm.com>
19517 * target.def (legitimize_address_displacement): Take the original
19518 offset as a poly_int.
19519 * targhooks.h (default_legitimize_address_displacement): Update
19521 * targhooks.c (default_legitimize_address_displacement): Likewise.
19522 * doc/tm.texi: Regenerate.
19523 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19524 as an argument, moving assert of ad->disp == ad->disp_term to...
19525 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19526 Try calling targetm.legitimize_address_displacement before expanding
19527 the address rather than afterwards, and adjust for the new interface.
19528 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19529 Match the new hook interface. Handle SVE addresses.
19530 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19531 new hook interface.
19533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19535 * Makefile.in (OBJS): Add early-remat.o.
19536 * target.def (select_early_remat_modes): New hook.
19537 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19538 * doc/tm.texi: Regenerate.
19539 * targhooks.h (default_select_early_remat_modes): Declare.
19540 * targhooks.c (default_select_early_remat_modes): New function.
19541 * timevar.def (TV_EARLY_REMAT): New timevar.
19542 * passes.def (pass_early_remat): New pass.
19543 * tree-pass.h (make_pass_early_remat): Declare.
19544 * early-remat.c: New file.
19545 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19547 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19550 Alan Hayward <alan.hayward@arm.com>
19551 David Sherwood <david.sherwood@arm.com>
19553 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19554 vfm1 with a bound_epilog parameter.
19555 (vect_do_peeling): Update calls accordingly, and move the prologue
19556 call earlier in the function. Treat the base bound_epilog as 0 for
19557 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19558 this base when peeling for gaps.
19559 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19560 with fully-masked loops.
19561 (vect_estimate_min_profitable_iters): Handle the single peeled
19562 iteration in that case.
19564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19565 Alan Hayward <alan.hayward@arm.com>
19566 David Sherwood <david.sherwood@arm.com>
19568 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19569 single-element interleaving even if the size is not a power of 2.
19570 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19571 accesses for single-element interleaving if the group size is
19574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19575 Alan Hayward <alan.hayward@arm.com>
19576 David Sherwood <david.sherwood@arm.com>
19578 * doc/md.texi (fold_extract_last_@var{m}): Document.
19579 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19580 * optabs.def (fold_extract_last_optab): New optab.
19581 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19582 * internal-fn.c (fold_extract_direct): New macro.
19583 (expand_fold_extract_optab_fn): Likewise.
19584 (direct_fold_extract_optab_supported_p): Likewise.
19585 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19586 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19587 EXTRACT_LAST_REDUCTION.
19588 (get_initial_def_for_reduction): Do not create an initial vector
19589 for EXTRACT_LAST_REDUCTION reductions.
19590 (vectorizable_reduction): Leave the scalar phi in place for
19591 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19592 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19593 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19594 transform phase to vectorizable_condition.
19595 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19597 (vect_finish_stmt_generation): ...here.
19598 (vect_finish_replace_stmt): New function.
19599 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19600 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19602 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19605 Alan Hayward <alan.hayward@arm.com>
19606 David Sherwood <david.sherwood@arm.com>
19608 * doc/md.texi (extract_last_@var{m}): Document.
19609 * optabs.def (extract_last_optab): New optab.
19610 * internal-fn.def (EXTRACT_LAST): New internal function.
19611 * internal-fn.c (cond_unary_direct): New macro.
19612 (expand_cond_unary_optab_fn): Likewise.
19613 (direct_cond_unary_optab_supported_p): Likewise.
19614 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19615 loops using EXTRACT_LAST.
19616 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19617 (extract_last_<mode>): ...this optab.
19618 (vec_extract<mode><Vel>): Update accordingly.
19620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19621 Alan Hayward <alan.hayward@arm.com>
19622 David Sherwood <david.sherwood@arm.com>
19624 * target.def (empty_mask_is_expensive): New hook.
19625 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19626 * doc/tm.texi: Regenerate.
19627 * targhooks.h (default_empty_mask_is_expensive): Declare.
19628 * targhooks.c (default_empty_mask_is_expensive): New function.
19629 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19630 if the target says that empty masks are expensive.
19631 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19633 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19636 Alan Hayward <alan.hayward@arm.com>
19637 David Sherwood <david.sherwood@arm.com>
19639 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19640 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19641 (vect_use_loop_mask_for_alignment_p): New function.
19642 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19643 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19644 niters_skip argument. Make sure that the first niters_skip elements
19645 of the first iteration are inactive.
19646 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19647 Update call to vect_set_loop_masks_directly.
19648 (get_misalign_in_elems): New function, split out from...
19649 (vect_gen_prolog_loop_niters): ...here.
19650 (vect_update_init_of_dr): Take a code argument that specifies whether
19651 the adjustment should be added or subtracted.
19652 (vect_update_init_of_drs): Likewise.
19653 (vect_prepare_for_masked_peels): New function.
19654 (vect_do_peeling): Skip prologue peeling if we're using a mask
19655 instead. Update call to vect_update_inits_of_drs.
19656 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19658 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19659 alignment. Do not include the number of peeled iterations in
19660 the minimum threshold in that case.
19661 (vectorizable_induction): Adjust the start value down by
19662 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19663 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19664 Take the number of skipped iterations into account when calculating
19666 * tree-vect-stmts.c (vect_gen_while_not): New function.
19668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19669 Alan Hayward <alan.hayward@arm.com>
19670 David Sherwood <david.sherwood@arm.com>
19672 * doc/sourcebuild.texi (vect_fully_masked): Document.
19673 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19674 default value to 0.
19675 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19677 (vect_analyze_loop_2): ...here. Don't check the vectorization
19678 factor against the number of loop iterations if the loop is
19681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19682 Alan Hayward <alan.hayward@arm.com>
19683 David Sherwood <david.sherwood@arm.com>
19685 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19686 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19687 (dump_groups): Update accordingly.
19688 (iv_use::mem_type): New member variable.
19689 (address_p): New function.
19690 (record_use): Add a mem_type argument and initialize the new
19692 (record_group_use): Add a mem_type argument. Use address_p.
19693 Remove obsolete null checks of base_object. Update call to record_use.
19694 (find_interesting_uses_op): Update call to record_group_use.
19695 (find_interesting_uses_cond): Likewise.
19696 (find_interesting_uses_address): Likewise.
19697 (get_mem_type_for_internal_fn): New function.
19698 (find_address_like_use): Likewise.
19699 (find_interesting_uses_stmt): Try find_address_like_use before
19700 calling find_interesting_uses_op.
19701 (addr_offset_valid_p): Use the iv mem_type field as the type
19702 of the addressed memory.
19703 (add_autoinc_candidates): Likewise.
19704 (get_address_cost): Likewise.
19705 (split_small_address_groups_p): Use address_p.
19706 (split_address_groups): Likewise.
19707 (add_iv_candidate_for_use): Likewise.
19708 (autoinc_possible_for_pair): Likewise.
19709 (rewrite_groups): Likewise.
19710 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19711 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19712 (get_alias_ptr_type_for_ptr_address): New function.
19713 (rewrite_use_address): Rewrite address uses in calls that were
19714 identified by find_address_like_use.
19716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19717 Alan Hayward <alan.hayward@arm.com>
19718 David Sherwood <david.sherwood@arm.com>
19720 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19722 * gimple-expr.h (is_gimple_addressable: Likewise.
19723 * gimple-expr.c (is_gimple_address): Likewise.
19724 * internal-fn.c (expand_call_mem_ref): New function.
19725 (expand_mask_load_optab_fn): Use it.
19726 (expand_mask_store_optab_fn): Likewise.
19728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19729 Alan Hayward <alan.hayward@arm.com>
19730 David Sherwood <david.sherwood@arm.com>
19732 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19733 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19734 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19735 (cond_umax@var{mode}): Document.
19736 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19737 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19738 (cond_umin_optab, cond_umax_optab): New optabs.
19739 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19740 (COND_IOR, COND_XOR): New internal functions.
19741 * internal-fn.h (get_conditional_internal_fn): Declare.
19742 * internal-fn.c (cond_binary_direct): New macro.
19743 (expand_cond_binary_optab_fn): Likewise.
19744 (direct_cond_binary_optab_supported_p): Likewise.
19745 (get_conditional_internal_fn): New function.
19746 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19747 Cope with reduction statements that are vectorized as calls rather
19749 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19750 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19751 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19752 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19753 (UNSPEC_COND_EOR): New unspecs.
19754 (optab): Add mappings for them.
19755 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19756 (sve_int_op, sve_fp_op): New int attributes.
19758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19759 Alan Hayward <alan.hayward@arm.com>
19760 David Sherwood <david.sherwood@arm.com>
19762 * optabs.def (while_ult_optab): New optab.
19763 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19764 * internal-fn.def (WHILE_ULT): New internal function.
19765 * internal-fn.h (direct_internal_fn_supported_p): New override
19766 that takes two types as argument.
19767 * internal-fn.c (while_direct): New macro.
19768 (expand_while_optab_fn): New function.
19769 (convert_optab_supported_p): Likewise.
19770 (direct_while_optab_supported_p): New macro.
19771 * wide-int.h (wi::udiv_ceil): New function.
19772 * tree-vectorizer.h (rgroup_masks): New structure.
19773 (vec_loop_masks): New typedef.
19774 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19775 and fully_masked_p.
19776 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19777 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19778 (vect_max_vf): New function.
19779 (slpeel_make_loop_iterate_ntimes): Delete.
19780 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19781 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19782 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19783 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19784 internal-fn.h, stor-layout.h and optabs-query.h.
19785 (vect_set_loop_mask): New function.
19786 (add_preheader_seq): Likewise.
19787 (add_header_seq): Likewise.
19788 (interleave_supported_p): Likewise.
19789 (vect_maybe_permute_loop_masks): Likewise.
19790 (vect_set_loop_masks_directly): Likewise.
19791 (vect_set_loop_condition_masked): Likewise.
19792 (vect_set_loop_condition_unmasked): New function, split out from
19793 slpeel_make_loop_iterate_ntimes.
19794 (slpeel_make_loop_iterate_ntimes): Rename to..
19795 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19796 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19797 (vect_do_peeling): Update call accordingly.
19798 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19800 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19801 mask_compare_type, can_fully_mask_p and fully_masked_p.
19802 (release_vec_loop_masks): New function.
19803 (_loop_vec_info): Use it to free the loop masks.
19804 (can_produce_all_loop_masks_p): New function.
19805 (vect_get_max_nscalars_per_iter): Likewise.
19806 (vect_verify_full_masking): Likewise.
19807 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19808 retries, and free the mask rgroups before retrying. Check loop-wide
19809 reasons for disallowing fully-masked loops. Make the final decision
19810 about whether use a fully-masked loop or not.
19811 (vect_estimate_min_profitable_iters): Do not assume that peeling
19812 for the number of iterations will be needed for fully-masked loops.
19813 (vectorizable_reduction): Disable fully-masked loops.
19814 (vectorizable_live_operation): Likewise.
19815 (vect_halve_mask_nunits): New function.
19816 (vect_double_mask_nunits): Likewise.
19817 (vect_record_loop_mask): Likewise.
19818 (vect_get_loop_mask): Likewise.
19819 (vect_transform_loop): Handle the case in which the final loop
19820 iteration might handle a partial vector. Call vect_set_loop_condition
19821 instead of slpeel_make_loop_iterate_ntimes.
19822 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19823 (check_load_store_masking): New function.
19824 (prepare_load_store_mask): Likewise.
19825 (vectorizable_store): Handle fully-masked loops.
19826 (vectorizable_load): Likewise.
19827 (supportable_widening_operation): Use vect_halve_mask_nunits for
19829 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19830 (vect_gen_while): New function.
19831 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19832 (aarch64_uqdec<mode>): New insn.
19834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19835 Alan Hayward <alan.hayward@arm.com>
19836 David Sherwood <david.sherwood@arm.com>
19838 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19839 (reduc_xor_scal_optab): New optabs.
19840 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19841 (reduc_xor_scal_@var{m}): Document.
19842 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19843 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19844 internal functions.
19845 * fold-const-call.c (fold_const_call): Handle them.
19846 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19847 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19848 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19849 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19850 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19851 (UNSPEC_XORV): New unspecs.
19852 (optab): Add entries for them.
19853 (BITWISEV): New int iterator.
19854 (bit_reduc_op): New int attributes.
19856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19857 Alan Hayward <alan.hayward@arm.com>
19858 David Sherwood <david.sherwood@arm.com>
19860 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19861 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19862 * optabs.def (vec_shl_insert_optab): New optab.
19863 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19864 (duplicate_and_interleave): Likewise.
19865 * tree-vect-loop.c: Include internal-fn.h.
19866 (neutral_op_for_slp_reduction): New function, split out from
19867 get_initial_defs_for_reduction.
19868 (get_initial_def_for_reduction): Handle option 2 for variable-length
19869 vectors by loading the neutral value into a vector and then shifting
19870 the initial value into element 0.
19871 (get_initial_defs_for_reduction): Replace the code argument with
19872 the neutral value calculated by neutral_op_for_slp_reduction.
19873 Use gimple_build_vector for constant-length vectors.
19874 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19875 but the first group_size elements have a neutral value.
19876 Use duplicate_and_interleave otherwise.
19877 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19878 Update call to get_initial_defs_for_reduction. Handle SLP
19879 reductions for variable-length vectors by creating one vector
19880 result for each scalar result, with the elements associated
19881 with other scalar results stubbed out with the neutral value.
19882 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19883 Require IFN_VEC_SHL_INSERT for double reductions on
19884 variable-length vectors, or SLP reductions that have
19885 a neutral value. Require can_duplicate_and_interleave_p
19886 support for variable-length unchained SLP reductions if there
19887 is no neutral value, such as for MIN/MAX reductions. Also require
19888 the number of vector elements to be a multiple of the number of
19889 SLP statements when doing variable-length unchained SLP reductions.
19890 Update call to vect_create_epilog_for_reduction.
19891 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19892 and remove initial values.
19893 (duplicate_and_interleave): Make public.
19894 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19895 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19898 Alan Hayward <alan.hayward@arm.com>
19899 David Sherwood <david.sherwood@arm.com>
19901 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19902 (can_duplicate_and_interleave_p): New function.
19903 (vect_get_and_check_slp_defs): Take the vector of statements
19904 rather than just the current one. Remove excess parentheses.
19905 Restriction rejectinon of vect_constant_def and vect_external_def
19906 for variable-length vectors to boolean types, or types for which
19907 can_duplicate_and_interleave_p is false.
19908 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19909 (duplicate_and_interleave): New function.
19910 (vect_get_constant_vectors): Use gimple_build_vector for
19911 constant-length vectors and suitable variable-length constant
19912 vectors. Use duplicate_and_interleave for other variable-length
19913 vectors. Don't defer the update when inserting new statements.
19915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19916 Alan Hayward <alan.hayward@arm.com>
19917 David Sherwood <david.sherwood@arm.com>
19919 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19920 min_profitable_iters doesn't go negative.
19922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19923 Alan Hayward <alan.hayward@arm.com>
19924 David Sherwood <david.sherwood@arm.com>
19926 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19927 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19928 * optabs.def (vec_mask_load_lanes_optab): New optab.
19929 (vec_mask_store_lanes_optab): Likewise.
19930 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19931 (MASK_STORE_LANES): Likewise.
19932 * internal-fn.c (mask_load_lanes_direct): New macro.
19933 (mask_store_lanes_direct): Likewise.
19934 (expand_mask_load_optab_fn): Handle masked operations.
19935 (expand_mask_load_lanes_optab_fn): New macro.
19936 (expand_mask_store_optab_fn): Handle masked operations.
19937 (expand_mask_store_lanes_optab_fn): New macro.
19938 (direct_mask_load_lanes_optab_supported_p): Likewise.
19939 (direct_mask_store_lanes_optab_supported_p): Likewise.
19940 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19942 (vect_load_lanes_supported): Likewise.
19943 * tree-vect-data-refs.c (strip_conversion): New function.
19944 (can_group_stmts_p): Likewise.
19945 (vect_analyze_data_ref_accesses): Use it instead of checking
19946 for a pair of assignments.
19947 (vect_store_lanes_supported): Take a masked_p parameter.
19948 (vect_load_lanes_supported): Likewise.
19949 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19950 vect_store_lanes_supported and vect_load_lanes_supported.
19951 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19952 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19953 parameter. Don't allow gaps for masked accesses.
19954 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19955 and vect_load_lanes_supported.
19956 (get_load_store_type): Take a masked_p parameter and update
19957 call to get_group_load_store_type.
19958 (vectorizable_store): Update call to get_load_store_type.
19959 Handle IFN_MASK_STORE_LANES.
19960 (vectorizable_load): Update call to get_load_store_type.
19961 Handle IFN_MASK_LOAD_LANES.
19963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19964 Alan Hayward <alan.hayward@arm.com>
19965 David Sherwood <david.sherwood@arm.com>
19967 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19969 * config/aarch64/aarch64-protos.h
19970 (aarch64_sve_struct_memory_operand_p): Declare.
19971 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19972 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19973 (VPRED, vpred): Handle SVE structure modes.
19974 * config/aarch64/constraints.md (Utx): New constraint.
19975 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19976 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19977 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19978 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19979 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19980 structure modes. Split into pieces after RA.
19981 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19982 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19984 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19985 SVE structure modes.
19986 (aarch64_classify_address): Likewise.
19987 (sizetochar): Move earlier in file.
19988 (aarch64_print_operand): Handle SVE register lists.
19989 (aarch64_array_mode): New function.
19990 (aarch64_sve_struct_memory_operand_p): Likewise.
19991 (TARGET_ARRAY_MODE): Redefine.
19993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19994 Alan Hayward <alan.hayward@arm.com>
19995 David Sherwood <david.sherwood@arm.com>
19997 * target.def (array_mode): New target hook.
19998 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19999 * doc/tm.texi: Regenerate.
20000 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
20001 * hooks.c (hook_optmode_mode_uhwi_none): New function.
20002 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
20003 targetm.array_mode.
20004 * stor-layout.c (mode_for_array): Likewise. Support polynomial
20007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20008 Alan Hayward <alan.hayward@arm.com>
20009 David Sherwood <david.sherwood@arm.com>
20011 * fold-const.c (fold_binary_loc): Check the argument types
20012 rather than the result type when testing for a vector operation.
20014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20016 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
20017 * doc/tm.texi: Regenerate.
20019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20020 Alan Hayward <alan.hayward@arm.com>
20021 David Sherwood <david.sherwood@arm.com>
20023 * doc/invoke.texi (-msve-vector-bits=): Document new option.
20024 (sve): Document new AArch64 extension.
20025 * doc/md.texi (w): Extend the description of the AArch64
20026 constraint to include SVE vectors.
20027 (Upl, Upa): Document new AArch64 predicate constraints.
20028 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
20030 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
20031 (msve-vector-bits=): New option.
20032 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
20033 SVE when these are disabled.
20034 (sve): New extension.
20035 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
20036 modes. Adjust their number of units based on aarch64_sve_vg.
20037 (MAX_BITSIZE_MODE_ANY_MODE): Define.
20038 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
20039 aarch64_addr_query_type.
20040 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
20041 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
20042 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
20043 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
20044 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
20045 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
20046 (aarch64_simd_imm_zero_p): Delete.
20047 (aarch64_check_zero_based_sve_index_immediate): Declare.
20048 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20049 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20050 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20051 (aarch64_sve_float_mul_immediate_p): Likewise.
20052 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20053 rather than an rtx.
20054 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
20055 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
20056 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
20057 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
20058 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
20059 (aarch64_regmode_natural_size): Likewise.
20060 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
20061 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
20063 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
20064 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
20065 for VG and the SVE predicate registers.
20066 (V_ALIASES): Add a "z"-prefixed alias.
20067 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
20068 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
20069 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
20070 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
20071 (REG_CLASS_NAMES): Add entries for them.
20072 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
20073 and the predicate registers.
20074 (aarch64_sve_vg): Declare.
20075 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
20076 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
20077 (REGMODE_NATURAL_SIZE): Define.
20078 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
20080 * config/aarch64/aarch64.c: Include cfgrtl.h.
20081 (simd_immediate_info): Add a constructor for series vectors,
20082 and an associated step field.
20083 (aarch64_sve_vg): New variable.
20084 (aarch64_dbx_register_number): Handle VG and the predicate registers.
20085 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
20086 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
20087 (VEC_ANY_DATA, VEC_STRUCT): New constants.
20088 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
20089 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
20090 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
20091 (aarch64_get_mask_mode): New functions.
20092 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
20093 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
20094 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
20095 predicate modes and predicate registers. Explicitly restrict
20096 GPRs to modes of 16 bytes or smaller. Only allow FP registers
20097 to store a vector mode if it is recognized by
20098 aarch64_classify_vector_mode.
20099 (aarch64_regmode_natural_size): New function.
20100 (aarch64_hard_regno_caller_save_mode): Return the original mode
20102 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
20103 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
20104 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
20105 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
20107 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
20108 does not overlap dest if the function is frame-related. Handle
20110 (aarch64_split_add_offset): New function.
20111 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
20112 them aarch64_add_offset.
20113 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20114 and update call to aarch64_sub_sp.
20115 (aarch64_add_cfa_expression): New function.
20116 (aarch64_expand_prologue): Pass extra temporary registers to the
20117 functions above. Handle the case in which we need to emit new
20118 DW_CFA_expressions for registers that were originally saved
20119 relative to the stack pointer, but now have to be expressed
20120 relative to the frame pointer.
20121 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20123 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20124 IP0 and IP1 values for SVE frames.
20125 (aarch64_expand_vec_series): New function.
20126 (aarch64_expand_sve_widened_duplicate): Likewise.
20127 (aarch64_expand_sve_const_vector): Likewise.
20128 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20129 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20130 into the register, rather than emitting a SET directly.
20131 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20132 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20133 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20134 (offset_9bit_signed_scaled_p): New functions.
20135 (aarch64_replicate_bitmask_imm): New function.
20136 (aarch64_bitmask_imm): Use it.
20137 (aarch64_cannot_force_const_mem): Reject expressions involving
20138 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20139 (aarch64_classify_index): Handle SVE indices, by requiring
20140 a plain register index with a scale that matches the element size.
20141 (aarch64_classify_address): Handle SVE addresses. Assert that
20142 the mode of the address is VOIDmode or an integer mode.
20143 Update call to aarch64_classify_symbol.
20144 (aarch64_classify_symbolic_expression): Update call to
20145 aarch64_classify_symbol.
20146 (aarch64_const_vec_all_in_range_p): New function.
20147 (aarch64_print_vector_float_operand): Likewise.
20148 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20149 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20150 and the FP immediates 1.0 and 0.5.
20151 (aarch64_print_address_internal): Handle SVE addresses.
20152 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20153 (aarch64_regno_regclass): Handle predicate registers.
20154 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20156 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20157 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20158 (aarch64_convert_sve_vector_bits): New function.
20159 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20160 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20161 rather than an rtx.
20162 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20163 Handle SVE vector and predicate modes. Accept VL-based constants
20164 that need only one temporary register, and VL offsets that require
20165 no temporary registers.
20166 (aarch64_conditional_register_usage): Mark the predicate registers
20167 as fixed if SVE isn't available.
20168 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20169 Return true for SVE vector and predicate modes.
20170 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20171 rather than an unsigned int. Handle SVE modes.
20172 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20174 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20176 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20177 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20178 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20179 (aarch64_sve_float_mul_immediate_p): New functions.
20180 (aarch64_sve_valid_immediate): New function.
20181 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20182 Explicitly reject structure modes. Check for INDEX constants.
20183 Handle PTRUE and PFALSE constants.
20184 (aarch64_check_zero_based_sve_index_immediate): New function.
20185 (aarch64_simd_imm_zero_p): Delete.
20186 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20187 vector modes. Accept constants in the range of CNT[BHWD].
20188 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20189 ask for an Advanced SIMD mode.
20190 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20191 (aarch64_simd_vector_alignment): Handle SVE predicates.
20192 (aarch64_vectorize_preferred_vector_alignment): New function.
20193 (aarch64_simd_vector_alignment_reachable): Use it instead of
20195 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20196 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20198 (MAX_VECT_LEN): Delete.
20199 (expand_vec_perm_d): Add a vec_flags field.
20200 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20201 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20202 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20204 (aarch64_evpc_rev): Rename to...
20205 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20206 (aarch64_evpc_rev_global): New function.
20207 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20208 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20210 (aarch64_evpc_sve_tbl): New function.
20211 (aarch64_expand_vec_perm_const_1): Update after rename of
20212 aarch64_evpc_rev. Handle SVE permutes too, trying
20213 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20214 than aarch64_evpc_tbl.
20215 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20216 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20217 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20218 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20219 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20220 (aarch64_expand_sve_vcond): New functions.
20221 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20222 of aarch64_vector_mode_p.
20223 (aarch64_dwarf_poly_indeterminate_value): New function.
20224 (aarch64_compute_pressure_classes): Likewise.
20225 (aarch64_can_change_mode_class): Likewise.
20226 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20227 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20228 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20229 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20230 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20231 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20232 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20233 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20235 (Dn, Dl, Dr): Accept const as well as const_vector.
20236 (Dz): Likewise. Compare against CONST0_RTX.
20237 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20238 of "vector" where appropriate.
20239 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20240 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20241 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20242 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20243 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20244 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20245 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20246 (v_int_equiv): Extend to SVE modes.
20247 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20249 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20250 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20251 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20252 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20253 (SVE_COND_FP_CMP): New int iterators.
20254 (perm_hilo): Handle the new unpack unspecs.
20255 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20257 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20258 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20259 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20260 (aarch64_equality_operator, aarch64_constant_vector_operand)
20261 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20262 (aarch64_sve_nonimmediate_operand): Likewise.
20263 (aarch64_sve_general_operand): Likewise.
20264 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20265 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20266 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20267 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20268 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20269 (aarch64_sve_float_arith_immediate): Likewise.
20270 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20271 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20272 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20273 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20274 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20275 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20276 (aarch64_sve_float_arith_operand): Likewise.
20277 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20278 (aarch64_sve_float_mul_operand): Likewise.
20279 (aarch64_sve_vec_perm_operand): Likewise.
20280 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20281 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20282 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20283 as well as const_vector.
20284 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20285 in file. Use CONST0_RTX and CONSTM1_RTX.
20286 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20287 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20288 Use aarch64_simd_imm_zero.
20289 * config/aarch64/aarch64-sve.md: New file.
20290 * config/aarch64/aarch64.md: Include it.
20291 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20292 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20293 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20294 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20295 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20296 (sve): New attribute.
20297 (enabled): Disable instructions with the sve attribute unless
20299 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20300 aarch64_expand_mov_immediate.
20301 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20302 CNT[BHSD] immediates.
20303 (movti): Split CONST_POLY_INT moves into two halves.
20304 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20305 Split additions that need a temporary here if the destination
20306 is the stack pointer.
20307 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20308 (*add<mode>3_poly_1): New instruction.
20309 (set_clobber_cc): New expander.
20311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20313 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20314 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20315 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20316 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20317 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20318 Change innermode from fixed_mode_size to machine_mode.
20319 (simplify_subreg): Update call accordingly. Handle a constant-sized
20320 subreg of a variable-length CONST_VECTOR.
20322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20323 Alan Hayward <alan.hayward@arm.com>
20324 David Sherwood <david.sherwood@arm.com>
20326 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20327 (add_offset_to_base): New function, split out from...
20328 (create_mem_ref): ...here. When handling a scale other than 1,
20329 check first whether the address is valid without the offset.
20330 Add it into the base if so, leaving the index and scale as-is.
20332 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20335 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20336 fold_for_warn before checking if arg2 is INTEGER_CST.
20338 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20340 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20341 (store_multiple_operation): Delete.
20342 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20343 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20344 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20345 guarded by TARGET_STRING.
20346 (rs6000_output_load_multiple): Delete.
20347 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20348 OPTION_MASK_STRING / TARGET_STRING handling.
20349 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20350 (const rs6000_opt_masks) <"string">: Change mask to 0.
20351 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20352 (MASK_STRING): Delete.
20353 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20355 (load_multiple): Delete.
20362 (store_multiple): Delete.
20369 (movmemsi_8reg): Delete.
20370 (corresponding unnamed define_insn): Delete.
20371 (movmemsi_6reg): Delete.
20372 (corresponding unnamed define_insn): Delete.
20373 (movmemsi_4reg): Delete.
20374 (corresponding unnamed define_insn): Delete.
20375 (movmemsi_2reg): Delete.
20376 (corresponding unnamed define_insn): Delete.
20377 (movmemsi_1reg): Delete.
20378 (corresponding unnamed define_insn): Delete.
20379 * config/rs6000/rs6000.opt (mno-string): New.
20380 (mstring): Replace by deprecation warning stub.
20381 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20383 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20385 * regrename.c (regrename_do_replace): If replacing the same
20386 reg multiple times, try to reuse last created gen_raw_REG.
20389 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20390 main to workaround a bug in GDB.
20392 2018-01-12 Tom de Vries <tom@codesourcery.com>
20395 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20397 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20399 PR rtl-optimization/80481
20400 * ira-color.c (get_cap_member): New function.
20401 (allocnos_conflict_by_live_ranges_p): Use it.
20402 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20403 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20405 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20408 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20409 (*saddl_se_1): Ditto.
20410 (*ssubsi_1): Ditto.
20411 (*ssubl_se_1): Ditto.
20413 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20415 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20416 rather than wi::to_widest for DR_INITs.
20417 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20418 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20419 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20421 (vect_analyze_group_access_1): Note that here.
20423 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20425 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20426 polynomial type sizes.
20428 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20430 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20431 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20432 (gimple_add_tmp_var): Likewise.
20434 2018-01-12 Martin Liska <mliska@suse.cz>
20436 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20437 (gimple_alloc_sizes): Likewise.
20438 (dump_gimple_statistics): Use PRIu64 in printf format.
20439 * gimple.h: Change uint64_t to int.
20441 2018-01-12 Martin Liska <mliska@suse.cz>
20443 * tree-core.h: Use uint64_t instead of int.
20444 * tree.c (tree_node_counts): Likewise.
20445 (tree_node_sizes): Likewise.
20446 (dump_tree_statistics): Use PRIu64 in printf format.
20448 2018-01-12 Martin Liska <mliska@suse.cz>
20450 * Makefile.in: As qsort_chk is implemented in vec.c, add
20451 vec.o to linkage of gencfn-macros.
20452 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20453 passing the info to record_node_allocation_statistics.
20454 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20456 * ggc-common.c (struct ggc_usage): Add operator== and use
20457 it in operator< and compare function.
20458 * mem-stats.h (struct mem_usage): Likewise.
20459 * vec.c (struct vec_usage): Remove operator< and compare
20460 function. Can be simply inherited.
20462 2018-01-12 Martin Jambor <mjambor@suse.cz>
20465 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20466 * tree-ssa-math-opts.c: Include domwalk.h.
20467 (convert_mult_to_fma_1): New function.
20468 (fma_transformation_info): New type.
20469 (fma_deferring_state): Likewise.
20470 (cancel_fma_deferring): New function.
20471 (result_of_phi): Likewise.
20472 (last_fma_candidate_feeds_initial_phi): Likewise.
20473 (convert_mult_to_fma): Added deferring logic, split actual
20474 transformation to convert_mult_to_fma_1.
20475 (math_opts_dom_walker): New type.
20476 (math_opts_dom_walker::after_dom_children): New method, body moved
20477 here from pass_optimize_widening_mul::execute, added deferring logic
20479 (pass_optimize_widening_mul::execute): Moved most of code to
20480 math_opts_dom_walker::after_dom_children.
20481 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20482 * config/i386/i386.c (ix86_option_override_internal): Added
20483 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20485 2018-01-12 Richard Biener <rguenther@suse.de>
20488 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20489 inline instance vars.
20491 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20494 * config/rx/rx.c (rx_is_restricted_memory_address):
20495 Handle SUBREG case.
20497 2018-01-12 Richard Biener <rguenther@suse.de>
20499 PR tree-optimization/80846
20500 * target.def (split_reduction): New target hook.
20501 * targhooks.c (default_split_reduction): New function.
20502 * targhooks.h (default_split_reduction): Declare.
20503 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20504 target requests first reduce vectors by combining low and high
20506 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20507 (get_vectype_for_scalar_type_and_size): Export.
20508 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20509 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20510 * doc/tm.texi: Regenerate.
20511 * config/i386/i386.c (ix86_split_reduction): Implement
20512 TARGET_VECTORIZE_SPLIT_REDUCTION.
20514 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20517 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20518 in PIC mode except for TARGET_VXWORKS_RTP.
20519 * config/sparc/sparc.c: Include cfgrtl.h.
20520 (TARGET_INIT_PIC_REG): Define.
20521 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20522 (sparc_pic_register_p): New predicate.
20523 (sparc_legitimate_address_p): Use it.
20524 (sparc_legitimize_pic_address): Likewise.
20525 (sparc_delegitimize_address): Likewise.
20526 (sparc_mode_dependent_address_p): Likewise.
20527 (gen_load_pcrel_sym): Remove 4th parameter.
20528 (load_got_register): Adjust call to above. Remove obsolete stuff.
20529 (sparc_expand_prologue): Do not call load_got_register here.
20530 (sparc_flat_expand_prologue): Likewise.
20531 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20532 (sparc_use_pseudo_pic_reg): New function.
20533 (sparc_init_pic_reg): Likewise.
20534 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20535 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20537 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20539 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20540 Add item for branch_cost.
20542 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20544 PR rtl-optimization/83565
20545 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20546 not extend the result to a larger mode for rotate operations.
20547 (num_sign_bit_copies1): Likewise.
20549 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20552 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20554 Use values-Xc.o for -pedantic.
20555 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20557 2018-01-12 Martin Liska <mliska@suse.cz>
20560 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20562 (possible_polymorphic_call_targets): Use it.
20563 (ipa_devirt): Likewise.
20565 2018-01-12 Martin Liska <mliska@suse.cz>
20567 * profile-count.h (enum profile_quality): Use 0 as invalid
20568 enum value of profile_quality.
20570 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20572 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20573 -mext-string options.
20575 2018-01-12 Richard Biener <rguenther@suse.de>
20577 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20578 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20579 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20581 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20583 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20585 * configure.ac (--with-long-double-format): Add support for the
20586 configuration option to change the default long double format on
20588 * config.gcc (powerpc*-linux*-*): Likewise.
20589 * configure: Regenerate.
20590 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20591 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20592 used without modification.
20594 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20596 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20597 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20598 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20599 MISC_BUILTIN_SPEC_BARRIER.
20600 (rs6000_init_builtins): Likewise.
20601 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20603 (speculation_barrier): New define_insn.
20604 * doc/extend.texi: Document __builtin_speculation_barrier.
20606 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20609 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20610 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20611 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20613 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20614 integral modes instead of "ss" and "sd".
20615 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20616 vectors with 32-bit and 64-bit elements.
20617 (vecdupssescalarmodesuffix): New mode attribute.
20618 (vec_dup<mode>): Use it.
20620 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20623 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20624 frame if argument is passed on stack.
20626 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20629 * ree.c (combine_reaching_defs): Optimize also
20630 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20631 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20633 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20635 PR middle-end/83189
20636 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20638 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20640 PR middle-end/83718
20641 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20642 after they are computed.
20644 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20646 PR tree-optimization/83695
20647 * gimple-loop-linterchange.cc
20648 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20649 reset cached scev information after interchange.
20650 (pass_linterchange::execute): Remove call to scev_reset_htab.
20652 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20654 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20655 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20656 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20657 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20658 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20659 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20660 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20661 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20662 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20663 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20664 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20665 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20666 (V_lane_reg): Likewise.
20667 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20669 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20670 (vfmal_lane_low<mode>_intrinsic,
20671 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20672 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20673 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20674 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20675 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20676 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20678 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20680 * config/arm/arm-cpus.in (fp16fml): New feature.
20681 (ALL_SIMD): Add fp16fml.
20682 (armv8.2-a): Add fp16fml as an option.
20683 (armv8.3-a): Likewise.
20684 (armv8.4-a): Add fp16fml as part of fp16.
20685 * config/arm/arm.h (TARGET_FP16FML): Define.
20686 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20688 * config/arm/arm-modes.def (V2HF): Define.
20689 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20690 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20691 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20692 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20693 vfmsl_low, vfmsl_high): New set of builtins.
20694 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20695 (vfml_op): New code attribute.
20696 (VFMLHALVES): New int iterator.
20697 (VFML, VFMLSEL): New mode attributes.
20698 (V_reg): Define mapping for V2HF.
20699 (V_hi, V_lo): New mode attributes.
20700 (VF_constraint): Likewise.
20701 (vfml_half, vfml_half_selector): New int attributes.
20702 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20704 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20705 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20707 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20708 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20709 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20710 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20712 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20713 Document new effective target and option set.
20715 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20717 * config/arm/arm-cpus.in (armv8_4): New feature.
20718 (ARMv8_4a): New fgroup.
20719 (armv8.4-a): New arch.
20720 * config/arm/arm-tables.opt: Regenerate.
20721 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20722 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20723 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20724 Add matching rules for -march=armv8.4-a and extensions.
20725 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20727 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20730 * config/rx/rx.md (BW): New mode attribute.
20731 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20733 2018-01-11 Richard Biener <rguenther@suse.de>
20735 PR tree-optimization/83435
20736 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20737 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20738 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20740 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20741 Alan Hayward <alan.hayward@arm.com>
20742 David Sherwood <david.sherwood@arm.com>
20744 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20746 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20747 (aarch64_print_address_internal): Use it to check for a zero offset.
20749 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20750 Alan Hayward <alan.hayward@arm.com>
20751 David Sherwood <david.sherwood@arm.com>
20753 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20754 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20755 Return a poly_int64 rather than a HOST_WIDE_INT.
20756 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20757 rather than a HOST_WIDE_INT.
20758 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20759 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20760 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20761 final_offset from HOST_WIDE_INT to poly_int64.
20762 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20763 to_constant when getting the number of units in an Advanced SIMD
20765 (aarch64_builtin_vectorized_function): Check for a constant number
20767 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20769 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20770 attribute instead of GET_MODE_NUNITS.
20771 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20772 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20773 GET_MODE_SIZE for fixed-size registers.
20774 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20775 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20776 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20777 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20778 (aarch64_print_operand, aarch64_print_address_internal)
20779 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20780 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20781 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20782 Handle polynomial GET_MODE_SIZE.
20783 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20784 wider than SImode without modification.
20785 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20786 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20787 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20788 passing and returning SVE modes.
20789 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20790 rather than GEN_INT.
20791 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20792 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20793 (aarch64_allocate_and_probe_stack_space): Likewise.
20794 (aarch64_layout_frame): Cope with polynomial offsets.
20795 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20796 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20797 polynomial offsets.
20798 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20799 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20800 poly_int64 rather than a HOST_WIDE_INT.
20801 (aarch64_get_separate_components, aarch64_process_components)
20802 (aarch64_expand_prologue, aarch64_expand_epilogue)
20803 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20804 (aarch64_anchor_offset): New function, split out from...
20805 (aarch64_legitimize_address): ...here.
20806 (aarch64_builtin_vectorization_cost): Handle polynomial
20807 TYPE_VECTOR_SUBPARTS.
20808 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20810 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20811 number of elements from the PARALLEL rather than the mode.
20812 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20813 rather than GET_MODE_BITSIZE.
20814 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20815 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20816 (aarch64_expand_vec_perm_const_1): Handle polynomial
20817 d->perm.length () and d->perm elements.
20818 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20819 Apply to_constant to d->perm elements.
20820 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20821 polynomial CONST_VECTOR_NUNITS.
20822 (aarch64_move_pointer): Take amount as a poly_int64 rather
20824 (aarch64_progress_pointer): Avoid temporary variable.
20825 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20826 the mode attribute instead of GET_MODE.
20828 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20829 Alan Hayward <alan.hayward@arm.com>
20830 David Sherwood <david.sherwood@arm.com>
20832 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20833 x exists before using it.
20834 (aarch64_add_constant_internal): Rename to...
20835 (aarch64_add_offset_1): ...this. Replace regnum with separate
20836 src and dest rtxes. Handle the case in which they're different,
20837 including when the offset is zero. Replace scratchreg with an rtx.
20838 Use 2 additions if there is no spare register into which we can
20839 move a 16-bit constant.
20840 (aarch64_add_constant): Delete.
20841 (aarch64_add_offset): Replace reg with separate src and dest
20842 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20843 Use aarch64_add_offset_1.
20844 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20845 an rtx rather than an int. Take the delta as a poly_int64
20846 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20847 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20848 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20849 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20850 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20851 and aarch64_add_sp.
20852 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20853 aarch64_add_constant.
20855 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20857 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20858 Use scalar_float_mode.
20860 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20862 * config/aarch64/aarch64-simd.md
20863 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20864 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20865 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20866 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20867 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20868 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20869 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20870 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20871 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20872 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20874 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20877 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20878 targ_options->x_arm_arch_string is non NULL.
20880 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20882 * config/aarch64/aarch64.h
20883 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20885 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20888 * expmed.c (emit_store_flag_force): Swap if const op0
20889 and change VOIDmode to mode of op0.
20891 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20893 PR rtl-optimization/83761
20894 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20895 than bytes to mode_for_size.
20897 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20899 PR middle-end/83189
20900 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20901 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20904 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20906 PR middle-end/83575
20907 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20908 when in layout mode.
20909 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20910 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20913 2018-01-10 Michael Collison <michael.collison@arm.com>
20915 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20916 * config/aarch64/aarch64-option-extension.def: Add
20917 AARCH64_OPT_EXTENSION of 'fp16fml'.
20918 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20919 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20920 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20921 * config/aarch64/constraints.md (Ui7): New constraint.
20922 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20923 (VFMLA_SEL_W): Ditto.
20926 (VFMLA16_LOW): New int iterator.
20927 (VFMLA16_HIGH): Ditto.
20928 (UNSPEC_FMLAL): New unspec.
20929 (UNSPEC_FMLSL): Ditto.
20930 (UNSPEC_FMLAL2): Ditto.
20931 (UNSPEC_FMLSL2): Ditto.
20932 (f16mac): New code attribute.
20933 * config/aarch64/aarch64-simd-builtins.def
20934 (aarch64_fmlal_lowv2sf): Ditto.
20935 (aarch64_fmlsl_lowv2sf): Ditto.
20936 (aarch64_fmlalq_lowv4sf): Ditto.
20937 (aarch64_fmlslq_lowv4sf): Ditto.
20938 (aarch64_fmlal_highv2sf): Ditto.
20939 (aarch64_fmlsl_highv2sf): Ditto.
20940 (aarch64_fmlalq_highv4sf): Ditto.
20941 (aarch64_fmlslq_highv4sf): Ditto.
20942 (aarch64_fmlal_lane_lowv2sf): Ditto.
20943 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20944 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20945 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20946 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20947 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20948 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20949 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20950 (aarch64_fmlal_lane_highv2sf): Ditto.
20951 (aarch64_fmlsl_lane_highv2sf): Ditto.
20952 (aarch64_fmlal_laneq_highv2sf): Ditto.
20953 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20954 (aarch64_fmlalq_lane_highv4sf): Ditto.
20955 (aarch64_fmlsl_lane_highv4sf): Ditto.
20956 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20957 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20958 * config/aarch64/aarch64-simd.md:
20959 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20960 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20961 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20962 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20963 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20964 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20965 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20966 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20967 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20968 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20969 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20970 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20971 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20972 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20973 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20974 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20975 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20976 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20977 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20978 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20979 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20980 (vfmlsl_low_u32): Ditto.
20981 (vfmlalq_low_u32): Ditto.
20982 (vfmlslq_low_u32): Ditto.
20983 (vfmlal_high_u32): Ditto.
20984 (vfmlsl_high_u32): Ditto.
20985 (vfmlalq_high_u32): Ditto.
20986 (vfmlslq_high_u32): Ditto.
20987 (vfmlal_lane_low_u32): Ditto.
20988 (vfmlsl_lane_low_u32): Ditto.
20989 (vfmlal_laneq_low_u32): Ditto.
20990 (vfmlsl_laneq_low_u32): Ditto.
20991 (vfmlalq_lane_low_u32): Ditto.
20992 (vfmlslq_lane_low_u32): Ditto.
20993 (vfmlalq_laneq_low_u32): Ditto.
20994 (vfmlslq_laneq_low_u32): Ditto.
20995 (vfmlal_lane_high_u32): Ditto.
20996 (vfmlsl_lane_high_u32): Ditto.
20997 (vfmlal_laneq_high_u32): Ditto.
20998 (vfmlsl_laneq_high_u32): Ditto.
20999 (vfmlalq_lane_high_u32): Ditto.
21000 (vfmlslq_lane_high_u32): Ditto.
21001 (vfmlalq_laneq_high_u32): Ditto.
21002 (vfmlslq_laneq_high_u32): Ditto.
21003 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
21004 (AARCH64_FL_FOR_ARCH8_4): New.
21005 (AARCH64_ISA_F16FML): New ISA flag.
21006 (TARGET_F16FML): New feature flag for fp16fml.
21007 (doc/invoke.texi): Document new fp16fml option.
21009 2018-01-10 Michael Collison <michael.collison@arm.com>
21011 * config/aarch64/aarch64-builtins.c:
21012 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
21013 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21014 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
21015 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
21016 (AARCH64_ISA_SHA3): New ISA flag.
21017 (TARGET_SHA3): New feature flag for sha3.
21018 * config/aarch64/iterators.md (sha512_op): New int attribute.
21019 (CRYPTO_SHA512): New int iterator.
21020 (UNSPEC_SHA512H): New unspec.
21021 (UNSPEC_SHA512H2): Ditto.
21022 (UNSPEC_SHA512SU0): Ditto.
21023 (UNSPEC_SHA512SU1): Ditto.
21024 * config/aarch64/aarch64-simd-builtins.def
21025 (aarch64_crypto_sha512hqv2di): New builtin.
21026 (aarch64_crypto_sha512h2qv2di): Ditto.
21027 (aarch64_crypto_sha512su0qv2di): Ditto.
21028 (aarch64_crypto_sha512su1qv2di): Ditto.
21029 (aarch64_eor3qv8hi): Ditto.
21030 (aarch64_rax1qv2di): Ditto.
21031 (aarch64_xarqv2di): Ditto.
21032 (aarch64_bcaxqv8hi): Ditto.
21033 * config/aarch64/aarch64-simd.md:
21034 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
21035 (aarch64_crypto_sha512su0qv2di): Ditto.
21036 (aarch64_crypto_sha512su1qv2di): Ditto.
21037 (aarch64_eor3qv8hi): Ditto.
21038 (aarch64_rax1qv2di): Ditto.
21039 (aarch64_xarqv2di): Ditto.
21040 (aarch64_bcaxqv8hi): Ditto.
21041 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
21042 (vsha512h2q_u64): Ditto.
21043 (vsha512su0q_u64): Ditto.
21044 (vsha512su1q_u64): Ditto.
21045 (veor3q_u16): Ditto.
21046 (vrax1q_u64): Ditto.
21047 (vxarq_u64): Ditto.
21048 (vbcaxq_u16): Ditto.
21049 * config/arm/types.md (crypto_sha512): New type attribute.
21050 (crypto_sha3): Ditto.
21051 (doc/invoke.texi): Document new sha3 option.
21053 2018-01-10 Michael Collison <michael.collison@arm.com>
21055 * config/aarch64/aarch64-builtins.c:
21056 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
21057 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21058 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
21059 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
21060 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
21061 (AARCH64_ISA_SM4): New ISA flag.
21062 (TARGET_SM4): New feature flag for sm4.
21063 * config/aarch64/aarch64-simd-builtins.def
21064 (aarch64_sm3ss1qv4si): Ditto.
21065 (aarch64_sm3tt1aq4si): Ditto.
21066 (aarch64_sm3tt1bq4si): Ditto.
21067 (aarch64_sm3tt2aq4si): Ditto.
21068 (aarch64_sm3tt2bq4si): Ditto.
21069 (aarch64_sm3partw1qv4si): Ditto.
21070 (aarch64_sm3partw2qv4si): Ditto.
21071 (aarch64_sm4eqv4si): Ditto.
21072 (aarch64_sm4ekeyqv4si): Ditto.
21073 * config/aarch64/aarch64-simd.md:
21074 (aarch64_sm3ss1qv4si): Ditto.
21075 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
21076 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
21077 (aarch64_sm4eqv4si): Ditto.
21078 (aarch64_sm4ekeyqv4si): Ditto.
21079 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
21080 (sm3part_op): Ditto.
21081 (CRYPTO_SM3TT): Ditto.
21082 (CRYPTO_SM3PART): Ditto.
21083 (UNSPEC_SM3SS1): New unspec.
21084 (UNSPEC_SM3TT1A): Ditto.
21085 (UNSPEC_SM3TT1B): Ditto.
21086 (UNSPEC_SM3TT2A): Ditto.
21087 (UNSPEC_SM3TT2B): Ditto.
21088 (UNSPEC_SM3PARTW1): Ditto.
21089 (UNSPEC_SM3PARTW2): Ditto.
21090 (UNSPEC_SM4E): Ditto.
21091 (UNSPEC_SM4EKEY): Ditto.
21092 * config/aarch64/constraints.md (Ui2): New constraint.
21093 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
21094 * config/arm/types.md (crypto_sm3): New type attribute.
21095 (crypto_sm4): Ditto.
21096 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
21097 (vsm3tt1aq_u32): Ditto.
21098 (vsm3tt1bq_u32): Ditto.
21099 (vsm3tt2aq_u32): Ditto.
21100 (vsm3tt2bq_u32): Ditto.
21101 (vsm3partw1q_u32): Ditto.
21102 (vsm3partw2q_u32): Ditto.
21103 (vsm4eq_u32): Ditto.
21104 (vsm4ekeyq_u32): Ditto.
21105 (doc/invoke.texi): Document new sm4 option.
21107 2018-01-10 Michael Collison <michael.collison@arm.com>
21109 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
21110 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
21111 (AARCH64_FL_FOR_ARCH8_4): New.
21112 (AARCH64_FL_V8_4): New flag.
21113 (doc/invoke.texi): Document new armv8.4-a option.
21115 2018-01-10 Michael Collison <michael.collison@arm.com>
21117 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21118 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21119 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21120 * config/aarch64/aarch64-option-extension.def: Add
21121 AARCH64_OPT_EXTENSION of 'sha2'.
21122 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21123 (crypto): Disable sha2 and aes if crypto disabled.
21124 (crypto): Enable aes and sha2 if enabled.
21125 (simd): Disable sha2 and aes if simd disabled.
21126 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21128 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21129 (TARGET_SHA2): New feature flag for sha2.
21130 (TARGET_AES): New feature flag for aes.
21131 * config/aarch64/aarch64-simd.md:
21132 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21133 conditional on TARGET_AES.
21134 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21135 (aarch64_crypto_sha1hsi): Make pattern conditional
21137 (aarch64_crypto_sha1hv4si): Ditto.
21138 (aarch64_be_crypto_sha1hv4si): Ditto.
21139 (aarch64_crypto_sha1su1v4si): Ditto.
21140 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21141 (aarch64_crypto_sha1su0v4si): Ditto.
21142 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21143 (aarch64_crypto_sha256su0v4si): Ditto.
21144 (aarch64_crypto_sha256su1v4si): Ditto.
21145 (doc/invoke.texi): Document new aes and sha2 options.
21147 2018-01-10 Martin Sebor <msebor@redhat.com>
21149 PR tree-optimization/83781
21150 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21153 2018-01-11 Martin Sebor <msebor@gmail.com>
21154 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21156 PR tree-optimization/83501
21157 PR tree-optimization/81703
21159 * tree-ssa-strlen.c (get_string_cst): Rename...
21160 (get_string_len): ...to this. Handle global constants.
21161 (handle_char_store): Adjust.
21163 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21164 Jim Wilson <jimw@sifive.com>
21166 * config/riscv/riscv-protos.h (riscv_output_return): New.
21167 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21168 (riscv_attribute_table, riscv_output_return),
21169 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21170 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21171 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21173 (riscv_expand_prologue): Add early return for naked function.
21174 (riscv_expand_epilogue): Likewise.
21175 (riscv_function_ok_for_sibcall): Return false for naked function.
21176 (riscv_set_current_function): New.
21177 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21178 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21179 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21180 * doc/extend.texi (RISC-V Function Attributes): New.
21182 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21184 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21185 check for 128-bit long double before checking TCmode.
21186 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21187 128-bit long doubles before checking TFmode or TCmode.
21188 (FLOAT128_IBM_P): Likewise.
21190 2018-01-10 Martin Sebor <msebor@redhat.com>
21192 PR tree-optimization/83671
21193 * builtins.c (c_strlen): Unconditionally return zero for the empty
21195 Use -Warray-bounds for warnings.
21196 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21197 for non-constant array indices with COMPONENT_REF, arrays of
21198 arrays, and pointers to arrays.
21199 (gimple_fold_builtin_strlen): Determine and set length range for
21200 non-constant character arrays.
21202 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21204 PR middle-end/81897
21205 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21208 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21210 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21212 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21215 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21216 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21217 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21218 indexed_or_indirect_operand predicate.
21219 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21220 (*vsx_le_perm_load_v8hi): Likewise.
21221 (*vsx_le_perm_load_v16qi): Likewise.
21222 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21223 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21224 (*vsx_le_perm_store_v8hi): Likewise.
21225 (*vsx_le_perm_store_v16qi): Likewise.
21226 (eight unnamed splitters): Likewise.
21228 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21230 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21231 * config/rs6000/emmintrin.h: Likewise.
21232 * config/rs6000/mmintrin.h: Likewise.
21233 * config/rs6000/xmmintrin.h: Likewise.
21235 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21238 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21240 * tree.c (tree_nop_conversion): Return true for location wrapper
21242 (maybe_wrap_with_location): New function.
21243 (selftest::check_strip_nops): New function.
21244 (selftest::test_location_wrappers): New function.
21245 (selftest::tree_c_tests): Call it.
21246 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21247 (maybe_wrap_with_location): New decl.
21248 (EXPR_LOCATION_WRAPPER_P): New macro.
21249 (location_wrapper_p): New inline function.
21250 (tree_strip_any_location_wrapper): New inline function.
21252 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21255 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21256 stack_realign_offset for the largest alignment of stack slot
21258 (ix86_find_max_used_stack_alignment): New function.
21259 (ix86_finalize_stack_frame_flags): Use it. Set
21260 max_used_stack_alignment if we don't realign stack.
21261 * config/i386/i386.h (machine_function): Add
21262 max_used_stack_alignment.
21264 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21266 * config/arm/arm.opt (-mbranch-cost): New option.
21267 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21270 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21273 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21274 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21276 2018-01-10 Richard Biener <rguenther@suse.de>
21279 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21280 early out so it also covers the case where we have a non-NULL
21283 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21285 PR tree-optimization/83753
21286 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21287 for non-strided grouped accesses if the number of elements is 1.
21289 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21292 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21293 * i386.h (TARGET_USE_GATHER): Define.
21294 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21296 2018-01-10 Martin Liska <mliska@suse.cz>
21299 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21300 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21302 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21303 CLEANUP_NO_PARTITIONING is not set.
21305 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21307 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21308 for vectors, as a partial revert of r254296.
21309 * rtl.h (const_vec_p): Delete.
21310 (const_vec_duplicate_p): Don't test for vector CONSTs.
21311 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21312 * expmed.c (make_tree): Likewise.
21315 * common.md (E, F): Use CONSTANT_P instead of checking for
21317 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21318 checking for CONST_VECTOR.
21320 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21322 PR middle-end/83575
21323 * predict.c (force_edge_cold): Handle in more sane way edges
21324 with no prediction.
21326 2018-01-09 Carl Love <cel@us.ibm.com>
21328 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21330 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21331 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21332 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21333 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21334 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21335 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21336 * config/rs6000/rs6000-protos.h: Add extern defition for
21337 rs6000_generate_float2_double_code.
21338 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21340 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21341 (float2_v2df): Add define_expand.
21343 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21346 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21347 op_mode in the force_to_mode call.
21349 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21351 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21352 instead of checking each element individually.
21353 (aarch64_evpc_uzp): Likewise.
21354 (aarch64_evpc_zip): Likewise.
21355 (aarch64_evpc_ext): Likewise.
21356 (aarch64_evpc_rev): Likewise.
21357 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21358 instead of checking each element individually. Return true without
21360 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21361 whether all selected elements come from the same input, instead of
21362 checking each element individually. Remove calls to gen_rtx_REG,
21363 start_sequence and end_sequence and instead assert that no rtl is
21366 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21368 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21369 order of HIGH and CONST checks.
21371 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21373 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21374 if the destination isn't an SSA_NAME.
21376 2018-01-09 Richard Biener <rguenther@suse.de>
21378 PR tree-optimization/83668
21379 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21381 (canonicalize_loop_form): ... here, renamed from ...
21382 (canonicalize_loop_closed_ssa_form): ... this and amended to
21383 swap successor edges for loop exit blocks to make us use
21384 the RPO order we need for initial schedule generation.
21386 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21388 PR tree-optimization/64811
21389 * match.pd: When optimizing comparisons with Inf, avoid
21390 introducing or losing exceptions from comparisons with NaN.
21392 2018-01-09 Martin Liska <mliska@suse.cz>
21395 * asan.c (shadow_mem_size): Add gcc_assert.
21397 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21399 Don't save registers in main().
21402 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21403 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21404 * config/avr/avr.c (avr_set_current_function): Don't error if
21405 naked, OS_task or OS_main are specified at the same time.
21406 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21408 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21410 * common/config/avr/avr-common.c (avr_option_optimization_table):
21411 Switch on -mmain-is-OS_task for optimizing compilations.
21413 2018-01-09 Richard Biener <rguenther@suse.de>
21415 PR tree-optimization/83572
21416 * graphite.c: Include cfganal.h.
21417 (graphite_transform_loops): Connect infinite loops to exit
21418 and remove fake edges at the end.
21420 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21422 * ipa-inline.c (edge_badness): Revert accidental checkin.
21424 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21427 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21428 symbols; not inline clones.
21430 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21433 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21434 hard registers. Formatting fixes.
21436 PR preprocessor/83722
21437 * gcc.c (try_generate_repro): Pass
21438 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21439 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21442 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21443 Kito Cheng <kito.cheng@gmail.com>
21445 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21446 (riscv_leaf_function_p): Delete.
21447 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21449 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21451 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21453 (do_ifelse): New function.
21454 (do_isel): New function.
21455 (do_sub3): New function.
21456 (do_add3): New function.
21457 (do_load_mask_compare): New function.
21458 (do_overlap_load_compare): New function.
21459 (expand_compare_loop): New function.
21460 (expand_block_compare): Call expand_compare_loop() when appropriate.
21461 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21462 option description.
21463 (-mblock-compare-inline-loop-limit): New option.
21465 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21468 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21469 Reverse order of second and third operands in first alternative.
21470 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21471 of first and second elements in UNSPEC_VPERMR vector.
21472 (altivec_expand_vec_perm_le): Likewise.
21474 2018-01-08 Jeff Law <law@redhat.com>
21476 PR rtl-optimizatin/81308
21477 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21478 (process_switch): If group_case_labels makes a change, then set
21480 (pass_convert_switch::execute): If a switch is converted, then
21481 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21483 PR rtl-optimization/81308
21484 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21487 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21489 PR target/83663 - Revert r255946
21490 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21491 generation for cases where splatting a value is not useful.
21492 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21493 across a vec_duplicate and a paradoxical subreg forming a vector
21494 mode to a vec_concat.
21496 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21498 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21499 -march=armv8.3-a variants.
21500 * config/arm/t-multilib: Likewise.
21501 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21503 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21505 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21507 (cceq_ior_compare_complement): Give it a name so I can use it, and
21508 change boolean_or_operator predicate to boolean_operator so it can
21509 be used to generate a crand.
21510 (eqne): New code iterator.
21511 (bd/bd_neg): New code_attrs.
21512 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21513 a single define_insn.
21514 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21515 decrement (bdnzt/bdnzf/bdzt/bdzf).
21516 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21517 with the new names of the branch decrement patterns, and added the
21518 names of the branch decrement conditional patterns.
21520 2018-01-08 Richard Biener <rguenther@suse.de>
21522 PR tree-optimization/83563
21523 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21526 2018-01-08 Richard Biener <rguenther@suse.de>
21528 PR middle-end/83713
21529 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21531 2018-01-08 Richard Biener <rguenther@suse.de>
21533 PR tree-optimization/83685
21534 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21535 references to abnormals.
21537 2018-01-08 Richard Biener <rguenther@suse.de>
21540 * dwarf2out.c (output_indirect_strings): Handle empty
21541 skeleton_debug_str_hash.
21542 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21544 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21546 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21547 (emit_store_direct): Likewise.
21548 (arc_trampoline_adjust_address): Likewise.
21549 (arc_asm_trampoline_template): New function.
21550 (arc_initialize_trampoline): Use asm_trampoline_template.
21551 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21552 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21553 * config/arc/arc.md (flush_icache): Delete pattern.
21555 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21557 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21558 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21561 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21564 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21565 by not USED_FOR_TARGET.
21566 (make_pass_resolve_sw_modes): Likewise.
21568 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21570 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21573 2018-01-08 Richard Biener <rguenther@suse.de>
21575 PR middle-end/83580
21576 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21578 2018-01-08 Richard Biener <rguenther@suse.de>
21580 PR middle-end/83517
21581 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21583 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21585 PR middle-end/81897
21586 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21587 basic blocks with a small number of successors.
21588 (convert_control_dep_chain_into_preds): Improve handling of
21590 (dump_predicates): Split apart into...
21591 (dump_pred_chain): ...here...
21592 (dump_pred_info): ...and here.
21593 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21594 (can_chain_union_be_invalidated_p): Improve check for invalidation
21596 (uninit_uses_cannot_happen): Avoid unnecessary if
21597 convert_control_dep_chain_into_preds yielded nothing.
21599 2018-01-06 Martin Sebor <msebor@redhat.com>
21601 PR tree-optimization/83640
21602 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21603 subtracting negative offset from size.
21604 (builtin_access::overlap): Adjust offset bounds of the access to fall
21605 within the size of the object if possible.
21607 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21609 PR rtl-optimization/83699
21610 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21611 extract_bit_field_as_subreg to cases in which the extracted
21612 value is also a vector.
21614 * lra-constraints.c (process_alt_operands): Test for the equivalence
21615 substitutions when detecting a possible reload cycle.
21617 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21620 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21621 by default if flag_selective_schedling{,2}. Formatting fixes.
21623 PR rtl-optimization/83682
21624 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21625 if it has non-VECTOR_MODE element mode.
21626 (vec_duplicate_p): Likewise.
21628 PR middle-end/83694
21629 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21630 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21632 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21635 * config/i386/i386-builtin.def
21636 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21637 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21638 Require also OPTION_MASK_ISA_AVX512F in addition to
21639 OPTION_MASK_ISA_GFNI.
21640 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21641 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21642 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21643 to OPTION_MASK_ISA_GFNI.
21644 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21645 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21646 OPTION_MASK_ISA_AVX512BW.
21647 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21648 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21649 addition to OPTION_MASK_ISA_GFNI.
21650 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21651 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21652 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21653 to OPTION_MASK_ISA_GFNI.
21654 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21655 a requirement for all ISAs rather than any of them with a few
21657 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21659 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21660 bitmasks to be enabled with 3 exceptions, instead of requiring any
21661 enabled ISA with lots of exceptions.
21662 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21663 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21664 Change avx512bw in isa attribute to avx512f.
21665 * config/i386/sgxintrin.h: Add license boilerplate.
21666 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21667 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21668 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21669 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21671 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21672 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21673 temporarily sse2 rather than sse if not enabled already.
21676 * config/i386/sse.md (VI248_VLBW): Rename to ...
21677 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21678 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21679 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21680 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21681 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21682 mode iterator instead of VI248_VLBW.
21684 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21686 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21687 (record_modified): Skip clobbers; add debug output.
21688 (param_change_prob): Use sreal frequencies.
21690 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21692 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21693 punt for user-aligned variables.
21695 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21697 * tree-chrec.c (chrec_contains_symbols): Return true for
21700 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21703 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21704 of (x|y) == x for BICS pattern.
21706 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21708 PR tree-optimization/83605
21709 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21710 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21713 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21715 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21716 * config/epiphany/rtems.h: New file.
21718 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21719 Uros Bizjak <ubizjak@gmail.com>
21722 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21723 QIreg_operand instead of register_operand predicate.
21724 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21725 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21726 comments instead of -fmitigate[-_]rop.
21728 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21731 * cgraphunit.c (symbol_table::compile): Switch to text_section
21732 before calling assembly_start debug hook.
21733 * run-rtl-passes.c (run_rtl_passes): Likewise.
21736 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21738 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21739 range_int_cst_p rather than !symbolic_range_p before calling
21740 extract_range_from_multiplicative_op_1.
21742 2018-01-04 Jeff Law <law@redhat.com>
21744 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21745 redundant test in assertion.
21747 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21749 * doc/rtl.texi: Document machine_mode wrapper classes.
21751 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21753 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21754 using tree_to_uhwi.
21756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21758 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21759 the VEC_PERM_EXPR fold to fail.
21761 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21764 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21765 to switched_sections.
21767 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21770 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21771 test for d.testing.
21773 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21776 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21777 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21779 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21782 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21783 is BLKmode and bitpos not zero or mode change is needed.
21785 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21788 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21791 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21794 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21795 instead of MULT rtx. Update all corresponding splitters.
21796 (*saddl_se): Ditto.
21797 (*ssub<modesuffix>): Ditto.
21798 (*ssubl_se): Ditto.
21799 (*cmp_sadd_di): Update split patterns.
21800 (*cmp_sadd_si): Ditto.
21801 (*cmp_sadd_sidi): Ditto.
21802 (*cmp_ssub_di): Ditto.
21803 (*cmp_ssub_si): Ditto.
21804 (*cmp_ssub_sidi): Ditto.
21805 * config/alpha/predicates.md (const23_operand): New predicate.
21806 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21807 Look for ASHIFT, not MULT inner operand.
21808 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21810 2018-01-04 Martin Liska <mliska@suse.cz>
21812 PR gcov-profile/83669
21813 * gcov.c (output_intermediate_file): Add version to intermediate
21815 * doc/gcov.texi: Document new field 'version' in intermediate
21816 file format. Fix location of '-k' option of gcov command.
21818 2018-01-04 Martin Liska <mliska@suse.cz>
21821 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21823 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21825 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21827 2018-01-03 Martin Sebor <msebor@redhat.com>
21829 PR tree-optimization/83655
21830 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21831 checking calls with invalid arguments.
21833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21835 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21836 (vectorizable_mask_load_store): Delete.
21837 (vectorizable_call): Return false for masked loads and stores.
21838 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21839 instead of gimple_assign_rhs1.
21840 (vectorizable_load): Handle IFN_MASK_LOAD.
21841 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21845 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21847 (vectorizable_mask_load_store): ...here.
21848 (vectorizable_load): ...and here.
21850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21852 * tree-vect-stmts.c (vect_build_all_ones_mask)
21853 (vect_build_zero_merge_argument): New functions, split out from...
21854 (vectorizable_load): ...here.
21856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21858 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21860 (vectorizable_mask_load_store): ...here.
21861 (vectorizable_store): ...and here.
21863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21865 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21867 (vectorizable_mask_load_store): ...here.
21869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21871 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21872 (vect_model_store_cost): Take a vec_load_store_type instead of a
21874 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21875 (vect_model_store_cost): Take a vec_load_store_type instead of a
21877 (vectorizable_mask_load_store): Update accordingly.
21878 (vectorizable_store): Likewise.
21879 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21883 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21884 IFN_MASK_LOAD calls here rather than...
21885 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21888 Alan Hayward <alan.hayward@arm.com>
21889 David Sherwood <david.sherwood@arm.com>
21891 * expmed.c (extract_bit_field_1): For vector extracts,
21892 fall back to extract_bit_field_as_subreg if vec_extract
21895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21896 Alan Hayward <alan.hayward@arm.com>
21897 David Sherwood <david.sherwood@arm.com>
21899 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21900 they are variable or constant sized.
21901 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21902 slots for constant-sized data.
21904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21905 Alan Hayward <alan.hayward@arm.com>
21906 David Sherwood <david.sherwood@arm.com>
21908 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21909 handling COND_EXPRs with boolean comparisons, try to find a better
21910 basis for the mask type than the boolean itself.
21912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21914 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21915 is calculated and how it can be overridden.
21916 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21917 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21919 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21923 Alan Hayward <alan.hayward@arm.com>
21924 David Sherwood <david.sherwood@arm.com>
21926 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21927 Remove the mode argument.
21928 (aarch64_simd_valid_immediate): Remove the mode and inverse
21930 * config/aarch64/iterators.md (bitsize): New iterator.
21931 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21932 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21933 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21934 aarch64_simd_valid_immediate.
21935 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21936 (aarch64_reg_or_bic_imm): Likewise.
21937 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21938 with an insn_type enum and msl with a modifier_type enum.
21939 Replace element_width with a scalar_mode. Change the shift
21940 to unsigned int. Add constructors for scalar_float_mode and
21941 scalar_int_mode elements.
21942 (aarch64_vect_float_const_representable_p): Delete.
21943 (aarch64_can_const_movi_rtx_p)
21944 (aarch64_simd_scalar_immediate_valid_for_move)
21945 (aarch64_simd_make_constant): Update call to
21946 aarch64_simd_valid_immediate.
21947 (aarch64_advsimd_valid_immediate_hs): New function.
21948 (aarch64_advsimd_valid_immediate): Likewise.
21949 (aarch64_simd_valid_immediate): Remove mode and inverse
21950 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21951 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21952 and aarch64_float_const_representable_p on the result.
21953 (aarch64_output_simd_mov_immediate): Remove mode argument.
21954 Update call to aarch64_simd_valid_immediate and use of
21955 simd_immediate_info.
21956 (aarch64_output_scalar_simd_mov_immediate): Update call
21959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21960 Alan Hayward <alan.hayward@arm.com>
21961 David Sherwood <david.sherwood@arm.com>
21963 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21964 (mode_nunits): Likewise CONST_MODE_NUNITS.
21965 * machmode.def (ADJUST_NUNITS): Document.
21966 * genmodes.c (mode_data::need_nunits_adj): New field.
21967 (blank_mode): Update accordingly.
21968 (adj_nunits): New variable.
21969 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21971 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21972 listed in adj_nunits.
21973 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21974 listed in adj_nunits. Don't emit case statements for such modes.
21975 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21976 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21977 nothing if adj_nunits is nonnull.
21978 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21979 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21980 (emit_mode_fbit): Update use of print_maybe_const_decl.
21981 (emit_move_size): Likewise. Treat the array as non-const
21983 (emit_mode_adjustments): Handle adj_nunits.
21985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21987 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21988 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21989 (VECTOR_MODES): Use it.
21990 (make_vector_modes): Take the prefix as an argument.
21992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21993 Alan Hayward <alan.hayward@arm.com>
21994 David Sherwood <david.sherwood@arm.com>
21996 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21997 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21998 for MODE_VECTOR_BOOL.
21999 * machmode.def (VECTOR_BOOL_MODE): Document.
22000 * genmodes.c (VECTOR_BOOL_MODE): New macro.
22001 (make_vector_bool_mode): New function.
22002 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
22004 * lto-streamer-in.c (lto_input_mode_table): Likewise.
22005 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
22007 * stor-layout.c (int_mode_for_mode): Likewise.
22008 * tree.c (build_vector_type_for_mode): Likewise.
22009 * varasm.c (output_constant_pool_2): Likewise.
22010 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
22011 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
22012 for MODE_VECTOR_BOOL.
22013 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
22014 of mode class checks.
22015 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
22016 instead of a list of mode class checks.
22017 (expand_vector_scalar_condition): Likewise.
22018 (type_for_widest_vector_mode): Handle BImode as an inner mode.
22020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22021 Alan Hayward <alan.hayward@arm.com>
22022 David Sherwood <david.sherwood@arm.com>
22024 * machmode.h (mode_size): Change from unsigned short to
22026 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
22027 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22028 or if measurement_type is not polynomial.
22029 (fixed_size_mode::includes_p): Check for constant-sized modes.
22030 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
22031 return a poly_uint16 rather than an unsigned short.
22032 (emit_mode_size): Change the type of mode_size from unsigned short
22033 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
22034 (emit_mode_adjustments): Cope with polynomial vector sizes.
22035 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22037 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22039 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
22040 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
22041 * caller-save.c (setup_save_areas): Likewise.
22042 (replace_reg_with_saved_mem): Likewise.
22043 * calls.c (emit_library_call_value_1): Likewise.
22044 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
22045 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
22046 (gen_lowpart_for_combine): Likewise.
22047 * convert.c (convert_to_integer_1): Likewise.
22048 * cse.c (equiv_constant, cse_insn): Likewise.
22049 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
22050 (cselib_subst_to_values): Likewise.
22051 * dce.c (word_dce_process_block): Likewise.
22052 * df-problems.c (df_word_lr_mark_ref): Likewise.
22053 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
22054 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
22055 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
22056 (rtl_for_decl_location): Likewise.
22057 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
22058 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
22059 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
22060 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
22061 (expand_expr_real_1): Likewise.
22062 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
22063 (pad_below): Likewise.
22064 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22065 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
22066 * ira.c (get_subreg_tracking_sizes): Likewise.
22067 * ira-build.c (ira_create_allocno_objects): Likewise.
22068 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
22069 (ira_sort_regnos_for_alter_reg): Likewise.
22070 * ira-costs.c (record_operand_costs): Likewise.
22071 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
22072 (resolve_simple_move): Likewise.
22073 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
22074 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
22075 (lra_constraints): Likewise.
22076 (CONST_POOL_OK_P): Reject variable-sized modes.
22077 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
22078 (add_pseudo_to_slot, lra_spill): Likewise.
22079 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22080 * optabs-query.c (get_best_extraction_insn): Likewise.
22081 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22082 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
22083 (expand_mult_highpart, valid_multiword_target_p): Likewise.
22084 * recog.c (offsettable_address_addr_space_p): Likewise.
22085 * regcprop.c (maybe_mode_change): Likewise.
22086 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
22087 * regrename.c (build_def_use): Likewise.
22088 * regstat.c (dump_reg_info): Likewise.
22089 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
22090 (find_reloads, find_reloads_subreg_address): Likewise.
22091 * reload1.c (eliminate_regs_1): Likewise.
22092 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
22093 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
22094 (simplify_binary_operation_1, simplify_subreg): Likewise.
22095 * targhooks.c (default_function_arg_padding): Likewise.
22096 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
22097 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
22098 (verify_gimple_assign_ternary): Likewise.
22099 * tree-inline.c (estimate_move_cost): Likewise.
22100 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22101 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
22102 (get_address_cost_ainc): Likewise.
22103 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
22104 (vect_supportable_dr_alignment): Likewise.
22105 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22106 (vectorizable_reduction): Likewise.
22107 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
22108 (vectorizable_operation, vectorizable_load): Likewise.
22109 * tree.c (build_same_sized_truth_vector_type): Likewise.
22110 * valtrack.c (cleanup_auto_inc_dec): Likewise.
22111 * var-tracking.c (emit_note_insn_var_location): Likewise.
22112 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
22113 (ADDR_VEC_ALIGN): Likewise.
22115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22116 Alan Hayward <alan.hayward@arm.com>
22117 David Sherwood <david.sherwood@arm.com>
22119 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22121 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22122 or if measurement_type is polynomial.
22123 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22124 * combine.c (make_extraction): Likewise.
22125 * dse.c (find_shift_sequence): Likewise.
22126 * dwarf2out.c (mem_loc_descriptor): Likewise.
22127 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22128 (extract_bit_field, extract_low_bits): Likewise.
22129 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22130 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22131 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22132 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22133 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22134 * reload.c (find_reloads): Likewise.
22135 * reload1.c (alter_reg): Likewise.
22136 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22137 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22138 * tree-if-conv.c (predicate_mem_writes): Likewise.
22139 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22140 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22141 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22142 * valtrack.c (dead_debug_insert_temp): Likewise.
22143 * varasm.c (mergeable_constant_section): Likewise.
22144 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22147 Alan Hayward <alan.hayward@arm.com>
22148 David Sherwood <david.sherwood@arm.com>
22150 * expr.c (expand_assignment): Cope with polynomial mode sizes
22151 when assigning to a CONCAT.
22153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22154 Alan Hayward <alan.hayward@arm.com>
22155 David Sherwood <david.sherwood@arm.com>
22157 * machmode.h (mode_precision): Change from unsigned short to
22159 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22161 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22162 or if measurement_type is not polynomial.
22163 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22164 in which the mode is already known to be a scalar_int_mode.
22165 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22166 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22168 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22169 for GET_MODE_PRECISION.
22170 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22171 for GET_MODE_PRECISION.
22172 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22174 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22175 (expand_field_assignment, make_extraction): Likewise.
22176 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22177 (get_last_value): Likewise.
22178 * convert.c (convert_to_integer_1): Likewise.
22179 * cse.c (cse_insn): Likewise.
22180 * expr.c (expand_expr_real_1): Likewise.
22181 * lra-constraints.c (simplify_operand_subreg): Likewise.
22182 * optabs-query.c (can_atomic_load_p): Likewise.
22183 * optabs.c (expand_atomic_load): Likewise.
22184 (expand_atomic_store): Likewise.
22185 * ree.c (combine_reaching_defs): Likewise.
22186 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22187 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22188 * tree.h (type_has_mode_precision_p): Likewise.
22189 * ubsan.c (instrument_si_overflow): Likewise.
22191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22192 Alan Hayward <alan.hayward@arm.com>
22193 David Sherwood <david.sherwood@arm.com>
22195 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22196 polynomial numbers of units.
22197 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22198 (valid_vector_subparts_p): New function.
22199 (build_vector_type): Remove temporary shim and take the number
22200 of units as a poly_uint64 rather than an int.
22201 (build_opaque_vector_type): Take the number of units as a
22202 poly_uint64 rather than an int.
22203 * tree.c (build_vector_from_ctor): Handle polynomial
22204 TYPE_VECTOR_SUBPARTS.
22205 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22206 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22207 (build_vector_from_val): If the number of units is variable,
22208 use build_vec_duplicate_cst for constant operands and
22209 VEC_DUPLICATE_EXPR otherwise.
22210 (make_vector_type): Remove temporary is_constant ().
22211 (build_vector_type, build_opaque_vector_type): Take the number of
22212 units as a poly_uint64 rather than an int.
22213 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22215 * cfgexpand.c (expand_debug_expr): Likewise.
22216 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22217 (store_constructor, expand_expr_real_1): Likewise.
22218 (const_scalar_mask_from_tree): Likewise.
22219 * fold-const-call.c (fold_const_reduction): Likewise.
22220 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22221 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22222 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22223 (fold_relational_const): Likewise.
22224 (native_interpret_vector): Likewise. Change the size from an
22225 int to an unsigned int.
22226 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22227 TYPE_VECTOR_SUBPARTS.
22228 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22229 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22230 duplicating a non-constant operand into a variable-length vector.
22231 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22232 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22233 * ipa-icf.c (sem_variable::equals): Likewise.
22234 * match.pd: Likewise.
22235 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22236 * print-tree.c (print_node): Likewise.
22237 * stor-layout.c (layout_type): Likewise.
22238 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22239 * tree-cfg.c (verify_gimple_comparison): Likewise.
22240 (verify_gimple_assign_binary): Likewise.
22241 (verify_gimple_assign_ternary): Likewise.
22242 (verify_gimple_assign_single): Likewise.
22243 * tree-pretty-print.c (dump_generic_node): Likewise.
22244 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22245 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22246 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22247 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22248 (vect_shift_permute_load_chain): Likewise.
22249 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22250 (expand_vector_condition, optimize_vector_constructor): Likewise.
22251 (lower_vec_perm, get_compute_type): Likewise.
22252 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22253 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22254 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22255 (vect_recog_mask_conversion_pattern): Likewise.
22256 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22257 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22258 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22259 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22260 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22261 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22262 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22263 (supportable_widening_operation): Likewise.
22264 (supportable_narrowing_operation): Likewise.
22265 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22267 * varasm.c (output_constant): Likewise.
22269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22270 Alan Hayward <alan.hayward@arm.com>
22271 David Sherwood <david.sherwood@arm.com>
22273 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22274 so that both the length == 3 and length != 3 cases set up their
22275 own permute vectors. Add comments explaining why we know the
22276 number of elements is constant.
22277 (vect_permute_load_chain): Likewise.
22279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22280 Alan Hayward <alan.hayward@arm.com>
22281 David Sherwood <david.sherwood@arm.com>
22283 * machmode.h (mode_nunits): Change from unsigned char to
22285 (ONLY_FIXED_SIZE_MODES): New macro.
22286 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22287 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22288 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22290 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22291 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22292 or if measurement_type is not polynomial.
22293 * genmodes.c (ZERO_COEFFS): New macro.
22294 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22296 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22297 Use ZERO_COEFFS when emitting initializers.
22298 * data-streamer.h (bp_pack_poly_value): New function.
22299 (bp_unpack_poly_value): Likewise.
22300 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22301 for GET_MODE_NUNITS.
22302 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22303 for GET_MODE_NUNITS.
22304 * tree.c (make_vector_type): Remove temporary shim and make
22305 the real function take the number of units as a poly_uint64
22306 rather than an int.
22307 (build_vector_type_for_mode): Handle polynomial nunits.
22308 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22309 * emit-rtl.c (const_vec_series_p_1): Likewise.
22310 (gen_rtx_CONST_VECTOR): Likewise.
22311 * fold-const.c (test_vec_duplicate_folding): Likewise.
22312 * genrecog.c (validate_pattern): Likewise.
22313 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22314 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22315 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22316 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22317 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22318 * rtlanal.c (subreg_get_info): Likewise.
22319 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22320 (vect_grouped_load_supported): Likewise.
22321 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22322 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22323 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22324 (simplify_const_unary_operation, simplify_binary_operation_1)
22325 (simplify_const_binary_operation, simplify_ternary_operation)
22326 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22327 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22328 instead of CONST_VECTOR_NUNITS.
22329 * varasm.c (output_constant_pool_2): Likewise.
22330 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22331 explicit-encoded elements in the XVEC for variable-length vectors.
22333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22335 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22338 Alan Hayward <alan.hayward@arm.com>
22339 David Sherwood <david.sherwood@arm.com>
22341 * coretypes.h (fixed_size_mode): Declare.
22342 (fixed_size_mode_pod): New typedef.
22343 * builtins.h (target_builtins::x_apply_args_mode)
22344 (target_builtins::x_apply_result_mode): Change type to
22345 fixed_size_mode_pod.
22346 * builtins.c (apply_args_size, apply_result_size, result_vector)
22347 (expand_builtin_apply_args_1, expand_builtin_apply)
22348 (expand_builtin_return): Update accordingly.
22350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22352 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22353 * cselib.c (cselib_hash_rtx): Likewise.
22354 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22355 CONST_VECTOR encoding.
22357 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22358 Jeff Law <law@redhat.com>
22361 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22362 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22363 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22364 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22367 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22368 explicitly probe *sp in a noreturn function if there were any callee
22369 register saves or frame pointer is needed.
22371 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22374 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22375 BLKmode for ternary, binary or unary expressions.
22378 * var-tracking.c (delete_vta_debug_insn): New inline function.
22379 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22380 insns from get_insns () to NULL instead of each bb separately.
22381 Use delete_vta_debug_insn. No longer static.
22382 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22383 delete_vta_debug_insns callers.
22384 * rtl.h (delete_vta_debug_insns): Declare.
22385 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22386 instead of variable_tracking_main.
22388 2018-01-03 Martin Sebor <msebor@redhat.com>
22390 PR tree-optimization/83603
22391 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22392 arguments past the endof the argument list in functions declared
22393 without a prototype.
22394 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22395 Avoid checking when arguments are null.
22397 2018-01-03 Martin Sebor <msebor@redhat.com>
22400 * doc/extend.texi (attribute const): Fix a typo.
22401 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22402 issuing -Wsuggest-attribute for void functions.
22404 2018-01-03 Martin Sebor <msebor@redhat.com>
22406 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22407 offset_int::from instead of wide_int::to_shwi.
22408 (maybe_diag_overlap): Remove assertion.
22409 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22410 * gimple-ssa-sprintf.c (format_directive): Same.
22411 (parse_directive): Same.
22412 (sprintf_dom_walker::compute_format_length): Same.
22413 (try_substitute_return_value): Same.
22415 2018-01-03 Jeff Law <law@redhat.com>
22417 PR middle-end/83654
22418 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22419 non-constant residual for zero at runtime and avoid probing in
22420 that case. Reorganize code for trailing problem to mirror handling
22423 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22425 PR tree-optimization/83501
22426 * tree-ssa-strlen.c (get_string_cst): New.
22427 (handle_char_store): Call get_string_cst.
22429 2018-01-03 Martin Liska <mliska@suse.cz>
22431 PR tree-optimization/83593
22432 * tree-ssa-strlen.c: Include tree-cfg.h.
22433 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22434 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22435 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22437 (strlen_dom_walker::before_dom_children): Call
22438 gimple_purge_dead_eh_edges. Dump tranformation with details
22440 (strlen_dom_walker::before_dom_children): Update call by adding
22441 new argument cleanup_eh.
22442 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22444 2018-01-03 Martin Liska <mliska@suse.cz>
22447 * cif-code.def (VARIADIC_THUNK): New enum value.
22448 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22451 2018-01-03 Jan Beulich <jbeulich@suse.com>
22453 * sse.md (mov<mode>_internal): Tighten condition for when to use
22454 vmovdqu<ssescalarsize> for TI and OI modes.
22456 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22458 Update copyright years.
22460 2018-01-03 Martin Liska <mliska@suse.cz>
22463 * ipa-visibility.c (function_and_variable_visibility): Skip
22464 functions with noipa attribure.
22466 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22468 * gcc.c (process_command): Update copyright notice dates.
22469 * gcov-dump.c (print_version): Ditto.
22470 * gcov.c (print_version): Ditto.
22471 * gcov-tool.c (print_version): Ditto.
22472 * gengtype.c (create_file): Ditto.
22473 * doc/cpp.texi: Bump @copying's copyright year.
22474 * doc/cppinternals.texi: Ditto.
22475 * doc/gcc.texi: Ditto.
22476 * doc/gccint.texi: Ditto.
22477 * doc/gcov.texi: Ditto.
22478 * doc/install.texi: Ditto.
22479 * doc/invoke.texi: Ditto.
22481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22483 * vector-builder.h (vector_builder::m_full_nelts): Change from
22484 unsigned int to poly_uint64.
22485 (vector_builder::full_nelts): Update prototype accordingly.
22486 (vector_builder::new_vector): Likewise.
22487 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22488 (vector_builder::operator ==): Likewise.
22489 (vector_builder::finalize): Likewise.
22490 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22491 Take the number of elements as a poly_uint64 rather than an
22493 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22494 from unsigned int to poly_uint64.
22495 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22496 (vec_perm_indices::new_vector): Likewise.
22497 (vec_perm_indices::length): Likewise.
22498 (vec_perm_indices::nelts_per_input): Likewise.
22499 (vec_perm_indices::input_nelts): Likewise.
22500 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22501 number of elements per input as a poly_uint64 rather than an
22502 unsigned int. Use the original encoding for variable-length
22503 vectors, rather than clamping each individual element.
22504 For the second and subsequent elements in each pattern,
22505 clamp the step and base before clamping their sum.
22506 (vec_perm_indices::series_p): Handle polynomial element counts.
22507 (vec_perm_indices::all_in_range_p): Likewise.
22508 (vec_perm_indices_to_tree): Likewise.
22509 (vec_perm_indices_to_rtx): Likewise.
22510 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22511 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22512 (tree_vector_builder::new_binary_operation): Handle polynomial
22513 element counts. Return false if we need to know the number
22514 of elements at compile time.
22515 * fold-const.c (fold_vec_perm): Punt if the number of elements
22516 isn't known at compile time.
22518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22520 * vec-perm-indices.h (vec_perm_builder): Change element type
22521 from HOST_WIDE_INT to poly_int64.
22522 (vec_perm_indices::element_type): Update accordingly.
22523 (vec_perm_indices::clamp): Handle polynomial element_types.
22524 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22525 (vec_perm_indices::all_in_range_p): Likewise.
22526 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22528 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22529 polynomial vec_perm_indices element types.
22530 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22531 * fold-const.c (fold_vec_perm): Likewise.
22532 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22533 * tree-vect-generic.c (lower_vec_perm): Likewise.
22534 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22535 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22536 element type to HOST_WIDE_INT.
22538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22539 Alan Hayward <alan.hayward@arm.com>
22540 David Sherwood <david.sherwood@arm.com>
22542 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22543 rather than an int. Use plus_constant.
22544 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22545 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22548 Alan Hayward <alan.hayward@arm.com>
22549 David Sherwood <david.sherwood@arm.com>
22551 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22552 a HOST_WIDE_INT to a poly_int64.
22554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22555 Alan Hayward <alan.hayward@arm.com>
22556 David Sherwood <david.sherwood@arm.com>
22558 * calls.c (load_register_parameters): Cope with polynomial
22559 mode sizes. Require a constant size for BLKmode parameters
22560 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22561 forces a parameter to be padded at the lsb end in order to
22562 fill a complete number of words, require the parameter size
22563 to be ordered wrt UNITS_PER_WORD.
22565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22566 Alan Hayward <alan.hayward@arm.com>
22567 David Sherwood <david.sherwood@arm.com>
22569 * reload1.c (spill_stack_slot_width): Change element type
22570 from unsigned int to poly_uint64_pod.
22571 (alter_reg): Treat mode sizes as polynomial.
22573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22574 Alan Hayward <alan.hayward@arm.com>
22575 David Sherwood <david.sherwood@arm.com>
22577 * reload.c (complex_word_subreg_p): New function.
22578 (reload_inner_reg_of_subreg, push_reload): Use it.
22580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22581 Alan Hayward <alan.hayward@arm.com>
22582 David Sherwood <david.sherwood@arm.com>
22584 * lra-constraints.c (process_alt_operands): Reject matched
22585 operands whose sizes aren't ordered.
22586 (match_reload): Refer to this check here.
22588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22589 Alan Hayward <alan.hayward@arm.com>
22590 David Sherwood <david.sherwood@arm.com>
22592 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22593 that the mode size is in the set {1, 2, 4, 8, 16}.
22595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22596 Alan Hayward <alan.hayward@arm.com>
22597 David Sherwood <david.sherwood@arm.com>
22599 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22600 Use plus_constant instead of gen_rtx_PLUS.
22602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22603 Alan Hayward <alan.hayward@arm.com>
22604 David Sherwood <david.sherwood@arm.com>
22606 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22607 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22608 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22609 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22610 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22611 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22612 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22613 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22614 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22615 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22617 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22618 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22619 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22620 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22621 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22622 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22623 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22624 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22625 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22626 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22628 * expr.c (emit_move_resolve_push): Treat the input and result
22629 of PUSH_ROUNDING as a poly_int64.
22630 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22631 (emit_push_insn): Likewise.
22632 * lra-eliminations.c (mark_not_eliminable): Likewise.
22633 * recog.c (push_operand): Likewise.
22634 * reload1.c (elimination_effects): Likewise.
22635 * rtlanal.c (nonzero_bits1): Likewise.
22636 * calls.c (store_one_arg): Likewise. Require the padding to be
22637 known at compile time.
22639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22640 Alan Hayward <alan.hayward@arm.com>
22641 David Sherwood <david.sherwood@arm.com>
22643 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22644 Use plus_constant instead of gen_rtx_PLUS.
22646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22647 Alan Hayward <alan.hayward@arm.com>
22648 David Sherwood <david.sherwood@arm.com>
22650 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22651 rather than an int.
22653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22654 Alan Hayward <alan.hayward@arm.com>
22655 David Sherwood <david.sherwood@arm.com>
22657 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22658 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22659 via stack temporaries. Treat the mode size as polynomial too.
22661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22662 Alan Hayward <alan.hayward@arm.com>
22663 David Sherwood <david.sherwood@arm.com>
22665 * expr.c (expand_expr_real_2): When handling conversions involving
22666 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22667 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22668 as a poly_uint64 too.
22670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22671 Alan Hayward <alan.hayward@arm.com>
22672 David Sherwood <david.sherwood@arm.com>
22674 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22677 Alan Hayward <alan.hayward@arm.com>
22678 David Sherwood <david.sherwood@arm.com>
22680 * combine.c (can_change_dest_mode): Handle polynomial
22681 REGMODE_NATURAL_SIZE.
22682 * expmed.c (store_bit_field_1): Likewise.
22683 * expr.c (store_constructor): Likewise.
22684 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22685 and polynomial REGMODE_NATURAL_SIZE.
22686 (gen_lowpart_common): Likewise.
22687 * reginfo.c (record_subregs_of_mode): Likewise.
22688 * rtlanal.c (read_modify_subreg_p): Likewise.
22690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22691 Alan Hayward <alan.hayward@arm.com>
22692 David Sherwood <david.sherwood@arm.com>
22694 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22695 numbers of elements.
22697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22698 Alan Hayward <alan.hayward@arm.com>
22699 David Sherwood <david.sherwood@arm.com>
22701 * match.pd: Cope with polynomial numbers of vector elements.
22703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22704 Alan Hayward <alan.hayward@arm.com>
22705 David Sherwood <david.sherwood@arm.com>
22707 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22708 in a POINTER_PLUS_EXPR.
22710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22711 Alan Hayward <alan.hayward@arm.com>
22712 David Sherwood <david.sherwood@arm.com>
22714 * omp-simd-clone.c (simd_clone_subparts): New function.
22715 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22716 (ipa_simd_modify_function_body): Likewise.
22718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22719 Alan Hayward <alan.hayward@arm.com>
22720 David Sherwood <david.sherwood@arm.com>
22722 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22723 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22724 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22725 (expand_vector_condition, vector_element): Likewise.
22726 (subparts_gt): New function.
22727 (get_compute_type): Use subparts_gt.
22728 (count_type_subparts): Delete.
22729 (expand_vector_operations_1): Use subparts_gt instead of
22730 count_type_subparts.
22732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22733 Alan Hayward <alan.hayward@arm.com>
22734 David Sherwood <david.sherwood@arm.com>
22736 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22737 (vect_compile_time_alias): ...this new function. Do the calculation
22738 on poly_ints rather than trees.
22739 (vect_prune_runtime_alias_test_list): Update call accordingly.
22741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22742 Alan Hayward <alan.hayward@arm.com>
22743 David Sherwood <david.sherwood@arm.com>
22745 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22747 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22754 constant and extern definitions for variable-length vectors.
22755 (vect_get_constant_vectors): Note that the number of units
22756 is known to be constant.
22758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22759 Alan Hayward <alan.hayward@arm.com>
22760 David Sherwood <david.sherwood@arm.com>
22762 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22763 of units as polynomial. Choose between WIDE and NARROW based
22766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22767 Alan Hayward <alan.hayward@arm.com>
22768 David Sherwood <david.sherwood@arm.com>
22770 * tree-vect-stmts.c (simd_clone_subparts): New function.
22771 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22774 Alan Hayward <alan.hayward@arm.com>
22775 David Sherwood <david.sherwood@arm.com>
22777 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22778 vectors as polynomial. Use build_index_vector for
22779 IFN_GOMP_SIMD_LANE.
22781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22782 Alan Hayward <alan.hayward@arm.com>
22783 David Sherwood <david.sherwood@arm.com>
22785 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22786 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22787 for variable-length vectors.
22788 (vectorizable_mask_load_store): Treat the number of units as
22789 polynomial, asserting that it is constant if the condition has
22790 already been enforced.
22791 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22798 of units as polynomial. Punt if we can't tell at compile time
22799 which vector contains the final result.
22801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22802 Alan Hayward <alan.hayward@arm.com>
22803 David Sherwood <david.sherwood@arm.com>
22805 * tree-vect-loop.c (vectorizable_induction): Treat the number
22806 of units as polynomial. Punt on SLP inductions. Use an integer
22807 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22808 cast of such a series for variable-length floating-point
22811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22812 Alan Hayward <alan.hayward@arm.com>
22813 David Sherwood <david.sherwood@arm.com>
22815 * tree.h (build_index_vector): Declare.
22816 * tree.c (build_index_vector): New function.
22817 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22818 of units as polynomial, forcibly converting it to a constant if
22819 vectorizable_reduction has already enforced the condition.
22820 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22821 to create a {1,2,3,...} vector.
22822 (vectorizable_reduction): Treat the number of units as polynomial.
22823 Choose vectype_in based on the largest scalar element size rather
22824 than the smallest number of units. Enforce the restrictions
22827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22828 Alan Hayward <alan.hayward@arm.com>
22829 David Sherwood <david.sherwood@arm.com>
22831 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22832 number of units as polynomial.
22834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22835 Alan Hayward <alan.hayward@arm.com>
22836 David Sherwood <david.sherwood@arm.com>
22838 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22839 * target.def (autovectorize_vector_sizes): Return the vector sizes
22840 by pointer, using vector_sizes rather than a bitmask.
22841 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22842 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22843 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22845 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22846 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22847 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22848 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22849 * omp-general.c (omp_max_vf): Likewise.
22850 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22851 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22852 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22853 * tree-vect-slp.c (vect_slp_bb): Likewise.
22854 * doc/tm.texi: Regenerate.
22855 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22857 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22858 the vector size as a poly_uint64 rather than an unsigned int.
22859 (current_vector_size): Change from an unsigned int to a poly_uint64.
22860 (get_vectype_for_scalar_type): Update accordingly.
22861 * tree.h (build_truth_vector_type): Take the size and number of
22862 units as a poly_uint64 rather than an unsigned int.
22863 (build_vector_type): Add a temporary overload that takes
22864 the number of units as a poly_uint64 rather than an unsigned int.
22865 * tree.c (make_vector_type): Likewise.
22866 (build_truth_vector_type): Take the number of units as a poly_uint64
22867 rather than an unsigned int.
22869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22870 Alan Hayward <alan.hayward@arm.com>
22871 David Sherwood <david.sherwood@arm.com>
22873 * target.def (get_mask_mode): Take the number of units and length
22874 as poly_uint64s rather than unsigned ints.
22875 * targhooks.h (default_get_mask_mode): Update accordingly.
22876 * targhooks.c (default_get_mask_mode): Likewise.
22877 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22878 * doc/tm.texi: Regenerate.
22880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22881 Alan Hayward <alan.hayward@arm.com>
22882 David Sherwood <david.sherwood@arm.com>
22884 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22885 * omp-general.c (omp_max_vf): Likewise.
22886 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22887 (expand_omp_simd): Handle polynomial safelen.
22888 * omp-low.c (omplow_simd_context): Add a default constructor.
22889 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22890 (lower_rec_simd_input_clauses): Update accordingly.
22891 (lower_rec_input_clauses): Likewise.
22893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22894 Alan Hayward <alan.hayward@arm.com>
22895 David Sherwood <david.sherwood@arm.com>
22897 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22898 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22899 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22900 (vect_analyze_slp_cost): Likewise.
22901 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22902 (vect_model_load_cost): Likewise.
22904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22905 Alan Hayward <alan.hayward@arm.com>
22906 David Sherwood <david.sherwood@arm.com>
22908 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22909 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22910 from an unsigned int * to a poly_uint64_pod *.
22911 (calculate_unrolling_factor): New function.
22912 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22915 Alan Hayward <alan.hayward@arm.com>
22916 David Sherwood <david.sherwood@arm.com>
22918 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22919 from an unsigned int to a poly_uint64.
22920 (_loop_vec_info::slp_unrolling_factor): Likewise.
22921 (_loop_vec_info::vectorization_factor): Change from an int
22923 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22924 (vect_get_num_vectors): New function.
22925 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22926 (vect_get_num_copies): Use vect_get_num_vectors.
22927 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22928 to an unsigned int *.
22929 (vect_analyze_data_refs): Change min_vf from an int * to a
22931 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22932 than an unsigned HOST_WIDE_INT.
22933 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22934 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22935 to an unsigned int *.
22936 (vect_analyze_data_ref_dependences): Likewise.
22937 (vect_compute_data_ref_alignment): Handle polynomial vf.
22938 (vect_enhance_data_refs_alignment): Likewise.
22939 (vect_prune_runtime_alias_test_list): Likewise.
22940 (vect_shift_permute_load_chain): Likewise.
22941 (vect_supportable_dr_alignment): Likewise.
22942 (dependence_distance_ge_vf): Take the vectorization factor as a
22943 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22944 (vect_analyze_data_refs): Change min_vf from an int * to a
22946 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22947 vfm1 as a poly_uint64 rather than an int. Make the same change
22948 for the returned bound_scalar.
22949 (vect_gen_vector_loop_niters): Handle polynomial vf.
22950 (vect_do_peeling): Likewise. Update call to
22951 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22952 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22954 * tree-vect-loop.c (vect_determine_vectorization_factor)
22955 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22956 (vect_get_known_peeling_cost): Likewise.
22957 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22958 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22959 (vect_transform_loop): Likewise. Use the lowest possible VF when
22960 updating the upper bounds of the loop.
22961 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22962 rather than an int.
22963 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22964 polynomial unroll factors.
22965 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22966 (vect_make_slp_decision): Likewise.
22967 (vect_supported_load_permutation_p): Likewise, and polynomial
22969 (vect_analyze_slp_cost): Handle polynomial vf.
22970 (vect_slp_analyze_node_operations): Likewise.
22971 (vect_slp_analyze_bb_1): Likewise.
22972 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22973 than an unsigned HOST_WIDE_INT.
22974 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22975 (vectorizable_load): Handle polynomial vf.
22976 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22978 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22981 Alan Hayward <alan.hayward@arm.com>
22982 David Sherwood <david.sherwood@arm.com>
22984 * match.pd: Handle bit operations involving three constants
22985 and try to fold one pair.
22987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22989 * tree-vect-loop-manip.c: Include gimple-fold.h.
22990 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22991 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22992 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22993 Add a path that uses a step of VF instead of 1, but disable it
22995 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22996 and niters_no_overflow parameters. Update calls to
22997 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22998 Create a new SSA name if the latter choses to use a ste other
22999 than zero, and return it via niters_vector_mult_vf_var.
23000 * tree-vect-loop.c (vect_transform_loop): Update calls to
23001 vect_do_peeling, vect_gen_vector_loop_niters and
23002 slpeel_make_loop_iterate_ntimes.
23003 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
23004 (vect_gen_vector_loop_niters): Update declarations after above changes.
23006 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
23008 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
23009 128-bit round to integer instructions.
23010 (ceil<mode>2): Likewise.
23011 (btrunc<mode>2): Likewise.
23012 (round<mode>2): Likewise.
23014 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23016 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
23017 unaligned VSX load/store on P8/P9.
23018 (expand_block_clear): Allow the use of unaligned VSX
23019 load/store on P8/P9.
23021 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23023 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
23025 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
23026 swap associated with both a load and a store.
23028 2018-01-02 Andrew Waterman <andrew@sifive.com>
23030 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
23031 * config/riscv/riscv.md (clear_cache): Use it.
23033 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
23035 * web.c: Remove out-of-date comment.
23037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23039 * expr.c (fixup_args_size_notes): Check that any existing
23040 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
23041 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
23042 (emit_single_push_insn): ...here.
23044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23046 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
23047 (const_vector_encoded_nelts): New function.
23048 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
23049 (const_vector_int_elt, const_vector_elt): Declare.
23050 * emit-rtl.c (const_vector_int_elt_1): New function.
23051 (const_vector_elt): Likewise.
23052 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
23053 of CONST_VECTOR_ELT.
23055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23057 * expr.c: Include rtx-vector-builder.h.
23058 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
23059 directly on the tree encoding.
23060 (const_vector_from_tree): Likewise.
23061 * optabs.c: Include rtx-vector-builder.h.
23062 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
23063 sequence of "u" values.
23064 * vec-perm-indices.c: Include rtx-vector-builder.h.
23065 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
23066 directly on the vec_perm_indices encoding.
23068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23070 * doc/rtl.texi (const_vector): Describe new encoding scheme.
23071 * Makefile.in (OBJS): Add rtx-vector-builder.o.
23072 * rtx-vector-builder.h: New file.
23073 * rtx-vector-builder.c: Likewise.
23074 * rtl.h (rtx_def::u2): Add a const_vector field.
23075 (CONST_VECTOR_NPATTERNS): New macro.
23076 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
23077 (CONST_VECTOR_DUPLICATE_P): Likewise.
23078 (CONST_VECTOR_STEPPED_P): Likewise.
23079 (CONST_VECTOR_ENCODED_ELT): Likewise.
23080 (const_vec_duplicate_p): Check for a duplicated vector encoding.
23081 (unwrap_const_vec_duplicate): Likewise.
23082 (const_vec_series_p): Check for a non-duplicated vector encoding.
23083 Say that the function only returns true for integer vectors.
23084 * emit-rtl.c: Include rtx-vector-builder.h.
23085 (gen_const_vec_duplicate_1): Delete.
23086 (gen_const_vector): Call gen_const_vec_duplicate instead of
23087 gen_const_vec_duplicate_1.
23088 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
23089 (gen_const_vec_duplicate): Use rtx_vector_builder.
23090 (gen_const_vec_series): Likewise.
23091 (gen_rtx_CONST_VECTOR): Likewise.
23092 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
23093 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23094 Build a new vector rather than modifying a CONST_VECTOR in-place.
23095 (handle_special_swappables): Update call accordingly.
23096 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
23097 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23098 Build a new vector rather than modifying a CONST_VECTOR in-place.
23099 (handle_special_swappables): Update call accordingly.
23101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23103 * simplify-rtx.c (simplify_const_binary_operation): Use
23104 CONST_VECTOR_ELT instead of XVECEXP.
23106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23108 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
23109 the selector elements to be different from the data elements
23110 if the selector is a VECTOR_CST.
23111 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
23112 ssizetype for the selector.
23114 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23116 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23117 before testing each element individually.
23118 * tree-vect-generic.c (lower_vec_perm): Likewise.
23120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23122 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23123 * selftest-run-tests.c (selftest::run_tests): Call it.
23124 * vector-builder.h (vector_builder::operator ==): New function.
23125 (vector_builder::operator !=): Likewise.
23126 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23127 (vec_perm_indices::all_from_input_p): New function.
23128 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23129 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23130 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23131 instead of reading the VECTOR_CST directly. Detect whether both
23132 vector inputs are the same before constructing the vec_perm_indices,
23133 and update the number of inputs argument accordingly. Use the
23134 utility functions added above. Only construct sel2 if we need to.
23136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23138 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23139 the broadcast of the low byte.
23140 (expand_mult_highpart): Use an explicit encoding for the permutes.
23141 * optabs-query.c (can_mult_highpart_p): Likewise.
23142 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23143 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23144 (vectorizable_bswap): Likewise.
23145 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23146 explicit encoding for the power-of-2 permutes.
23147 (vect_permute_store_chain): Likewise.
23148 (vect_grouped_load_supported): Likewise.
23149 (vect_permute_load_chain): Likewise.
23151 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23153 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23154 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23155 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23156 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23157 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23158 (vect_gen_perm_mask_any): Likewise.
23160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23162 * int-vector-builder.h: New file.
23163 * vec-perm-indices.h: Include int-vector-builder.h.
23164 (vec_perm_indices): Redefine as an int_vector_builder.
23165 (auto_vec_perm_indices): Delete.
23166 (vec_perm_builder): Redefine as a stand-alone class.
23167 (vec_perm_indices::vec_perm_indices): New function.
23168 (vec_perm_indices::clamp): Likewise.
23169 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23170 (vec_perm_indices::new_vector): New function.
23171 (vec_perm_indices::new_expanded_vector): Update for new
23172 vec_perm_indices class.
23173 (vec_perm_indices::rotate_inputs): New function.
23174 (vec_perm_indices::all_in_range_p): Operate directly on the
23175 encoded form, without computing elided elements.
23176 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23177 encoding. Update for new vec_perm_indices class.
23178 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23179 the given vec_perm_builder.
23180 (expand_vec_perm_var): Update vec_perm_builder constructor.
23181 (expand_mult_highpart): Use vec_perm_builder instead of
23182 auto_vec_perm_indices.
23183 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23184 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23185 or double series encoding as appropriate.
23186 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23187 vec_perm_indices instead of auto_vec_perm_indices.
23188 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23189 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23190 (vect_permute_store_chain): Likewise.
23191 (vect_grouped_load_supported): Likewise.
23192 (vect_permute_load_chain): Likewise.
23193 (vect_shift_permute_load_chain): Likewise.
23194 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23195 (vect_transform_slp_perm_load): Likewise.
23196 (vect_schedule_slp_instance): Likewise.
23197 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23198 (vectorizable_mask_load_store): Likewise.
23199 (vectorizable_bswap): Likewise.
23200 (vectorizable_store): Likewise.
23201 (vectorizable_load): Likewise.
23202 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23203 vec_perm_indices instead of auto_vec_perm_indices. Use
23204 tree_to_vec_perm_builder to read the vector from a tree.
23205 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23206 vec_perm_builder instead of a vec_perm_indices.
23207 (have_whole_vector_shift): Use vec_perm_builder and
23208 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23209 truncation to calc_vec_perm_mask_for_shift.
23210 (vect_create_epilog_for_reduction): Likewise.
23211 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23212 from auto_vec_perm_indices to vec_perm_indices.
23213 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23214 instead of changing individual elements.
23215 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23216 the vector in d.perm.
23217 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23218 from auto_vec_perm_indices to vec_perm_indices.
23219 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23220 instead of changing individual elements.
23221 (arm_vectorize_vec_perm_const): Use new_vector to install
23222 the vector in d.perm.
23223 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23224 Update vec_perm_builder constructor.
23225 (rs6000_expand_interleave): Likewise.
23226 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23227 (rs6000_expand_interleave): Likewise.
23229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23231 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23232 to qimode could truncate the indices.
23233 * optabs.c (expand_vec_perm_var): Likewise.
23235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23237 * Makefile.in (OBJS): Add vec-perm-indices.o.
23238 * vec-perm-indices.h: New file.
23239 * vec-perm-indices.c: Likewise.
23240 * target.h (vec_perm_indices): Replace with a forward class
23242 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23243 * optabs.h: Include vec-perm-indices.h.
23244 (expand_vec_perm): Delete.
23245 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23246 (expand_vec_perm_const): Declare.
23247 * target.def (vec_perm_const_ok): Replace with...
23248 (vec_perm_const): ...this new hook.
23249 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23250 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23251 * doc/tm.texi: Regenerate.
23252 * optabs.def (vec_perm_const): Delete.
23253 * doc/md.texi (vec_perm_const): Likewise.
23254 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23255 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23256 expand_vec_perm for constant permutation vectors. Assert that
23257 the mode of variable permutation vectors is the integer equivalent
23258 of the mode that is being permuted.
23259 * optabs-query.h (selector_fits_mode_p): Declare.
23260 * optabs-query.c: Include vec-perm-indices.h.
23261 (selector_fits_mode_p): New function.
23262 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23263 is defined, instead of checking whether the vec_perm_const_optab
23264 exists. Use targetm.vectorize.vec_perm_const instead of
23265 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23266 fit in the vector mode before using a variable permute.
23267 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23268 vec_perm_indices instead of an rtx.
23269 (expand_vec_perm): Replace with...
23270 (expand_vec_perm_const): ...this new function. Take the selector
23271 as a vec_perm_indices rather than an rtx. Also take the mode of
23272 the selector. Update call to shift_amt_for_vec_perm_mask.
23273 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23274 Use vec_perm_indices::new_expanded_vector to expand the original
23275 selector into bytes. Check whether the indices fit in the vector
23276 mode before using a variable permute.
23277 (expand_vec_perm_var): Make global.
23278 (expand_mult_highpart): Use expand_vec_perm_const.
23279 * fold-const.c: Includes vec-perm-indices.h.
23280 * tree-ssa-forwprop.c: Likewise.
23281 * tree-vect-data-refs.c: Likewise.
23282 * tree-vect-generic.c: Likewise.
23283 * tree-vect-loop.c: Likewise.
23284 * tree-vect-slp.c: Likewise.
23285 * tree-vect-stmts.c: Likewise.
23286 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23288 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23289 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23290 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23291 (aarch64_vectorize_vec_perm_const): ...this new function.
23292 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23293 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23294 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23295 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23296 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23297 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23298 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23300 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23301 check for NEON modes.
23302 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23303 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23304 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23305 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23307 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23308 the old VEC_PERM_CONST conditions.
23309 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23310 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23311 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23312 (ia64_vectorize_vec_perm_const_ok): Merge into...
23313 (ia64_vectorize_vec_perm_const): ...this new function.
23314 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23315 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23316 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23317 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23318 * config/mips/mips.c (mips_expand_vec_perm_const)
23319 (mips_vectorize_vec_perm_const_ok): Merge into...
23320 (mips_vectorize_vec_perm_const): ...this new function.
23321 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23322 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23323 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23324 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23325 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23326 (rs6000_expand_vec_perm_const): Delete.
23327 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23329 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23330 (altivec_expand_vec_perm_const_le): Take each operand individually.
23331 Operate on constant selectors rather than rtxes.
23332 (altivec_expand_vec_perm_const): Likewise. Update call to
23333 altivec_expand_vec_perm_const_le.
23334 (rs6000_expand_vec_perm_const): Delete.
23335 (rs6000_vectorize_vec_perm_const_ok): Delete.
23336 (rs6000_vectorize_vec_perm_const): New function.
23337 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23338 an element count and rtx array.
23339 (rs6000_expand_extract_even): Update call accordingly.
23340 (rs6000_expand_interleave): Likewise.
23341 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23342 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23343 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23344 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23345 (rs6000_expand_vec_perm_const): Delete.
23346 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23347 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23348 (altivec_expand_vec_perm_const_le): Take each operand individually.
23349 Operate on constant selectors rather than rtxes.
23350 (altivec_expand_vec_perm_const): Likewise. Update call to
23351 altivec_expand_vec_perm_const_le.
23352 (rs6000_expand_vec_perm_const): Delete.
23353 (rs6000_vectorize_vec_perm_const_ok): Delete.
23354 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23355 reference to the SPE evmerge intructions.
23356 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23357 an element count and rtx array.
23358 (rs6000_expand_extract_even): Update call accordingly.
23359 (rs6000_expand_interleave): Likewise.
23360 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23361 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23363 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23367 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23368 vector mode and that that mode matches the mode of the data
23370 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23371 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23372 directly using expand_vec_perm_1 when forcing selectors into
23374 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23376 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23378 * optabs-query.h (can_vec_perm_p): Delete.
23379 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23380 * optabs-query.c (can_vec_perm_p): Split into...
23381 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23382 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23383 particular selector is valid.
23384 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23385 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23386 (vect_grouped_load_supported): Likewise.
23387 (vect_shift_permute_load_chain): Likewise.
23388 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23389 (vect_transform_slp_perm_load): Likewise.
23390 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23391 (vectorizable_bswap): Likewise.
23392 (vect_gen_perm_mask_checked): Likewise.
23393 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23394 implementations of variable permutation vectors into account
23395 when deciding which selector to use.
23396 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23397 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23398 with a false third argument.
23399 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23400 to test whether the constant selector is valid and can_vec_perm_var_p
23401 to test whether a variable selector is valid.
23403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23405 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23406 * optabs-query.c (can_vec_perm_p): Likewise.
23407 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23408 instead of vec_perm_indices.
23409 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23410 (vect_gen_perm_mask_checked): Likewise,
23411 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23412 (vect_gen_perm_mask_checked): Likewise,
23414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23416 * optabs-query.h (qimode_for_vec_perm): Declare.
23417 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23418 (qimode_for_vec_perm): ...this new function.
23419 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23421 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23423 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23424 does not have a conditional at the top.
23426 2018-01-02 Richard Biener <rguenther@suse.de>
23428 * ipa-inline.c (big_speedup_p): Fix expression.
23430 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23433 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23436 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23440 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23441 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23442 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23443 cond_taken_branch_cost 3->4.
23445 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23447 PR tree-optimization/83581
23448 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23449 TODO_cleanup_cfg if any changes have been made.
23451 PR middle-end/83608
23452 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23453 convert_modes if target mode has the right side, but different mode
23456 PR middle-end/83609
23457 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23458 last argument when extracting from CONCAT. If either from_real or
23459 from_imag is NULL, use expansion through memory. If result is not
23460 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23461 the parts directly to inner mode, if even that fails, use expansion
23464 PR middle-end/83623
23465 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23466 check for bswap in mode rather than HImode and use that in expand_unop
23469 Copyright (C) 2018 Free Software Foundation, Inc.
23471 Copying and distribution of this file, with or without modification,
23472 are permitted in any medium without royalty provided the copyright
23473 notice and this notice are preserved.