1 2018-08-17 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/86841
4 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6 2018-08-17 Martin Liska <mliska@suse.cz>
8 * common.opt: Remove Warn, Init and Report for options with
9 Ignore/Deprecated flag. Warning is done automatically for
11 * config/i386/i386.opt: Likewise.
12 * config/ia64/ia64.opt: Likewise.
13 * config/rs6000/rs6000.opt: Likewise.
14 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
15 Remove usage of flag_check_pointer_bounds.
16 * lto-wrapper.c (merge_and_complain): Do not handle
17 OPT_fcheck_pointer_bounds.
18 (append_compiler_options): Likewise.
19 * opt-functions.awk: Do not handle Deprecated.
20 * optc-gen.awk: Check that Var, Report and Init are not
21 used for an option with Ignore/Deprecated flag.
22 * opts-common.c (decode_cmdline_option): Do not report
24 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
26 * opts.h (struct cl_option): Remove cl_deprecated flag.
27 (CL_ERR_DEPRECATED): Remove error enum value.
29 2018-08-17 Richard Biener <rguenther@suse.de>
32 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
33 across a va-arg-pack using call adjust its return value accordingly.
35 2018-08-16 Martin Sebor <msebor@redhat.com>
37 PR tree-optimization/86853
38 * gimple-ssa-sprintf.c (struct format_result): Rename member.
39 (struct fmtresult): Add member and initialize it in ctors.
40 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
41 (format_string): Handle %S the same as %ls. Set MAYFAIL.
42 (format_directive): Set POSUNDER4K when MAYFAIL is set.
43 (parse_directive): Handle %C same as %c.
44 (sprintf_dom_walker::compute_format_length): Adjust.
45 (is_call_safe): Adjust.
47 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
49 * builtins.c (c_strlen): Add new parameter eltsize. Use it
50 for determining how to count the elements.
51 * builtins.h (c_strlen): Adjust prototype.
52 * expr.c (string_constant): Add new parameter mem_size.
53 Set *mem_size appropriately.
54 * expr.h (string_constant): Adjust protoype.
55 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
56 * gimple-fold.h (get_range_strlen): Adjust prototype.
57 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
58 (format_string): Call get_string_length with eltsize.
60 2018-08-16 David Malcolm <dmalcolm@redhat.com>
62 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
63 to emit the span, rather than setting it as the prefix.
65 2018-08-16 David Malcolm <dmalcolm@redhat.com>
67 * diagnostic-show-locus.c (layout::start_annotation_line): Add
68 "margin_char" parameter, defaulting to space. Use it in place
69 of pp_space for the initial part of the margin.
70 (layout::print_leading_fixits): Use '+' when filling the margin
71 of line-insertion fix-it hints.
73 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
75 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
78 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
80 * config/rs6000/altivec.md: Don't set length attribute to the default
82 * config/rs6000/darwin.md: Ditto.
83 * config/rs6000/dfp.md: Ditto.
84 * config/rs6000/htm.md: Ditto.
85 * config/rs6000/rs6000.md: Ditto.
86 * config/rs6000/sync.md: Ditto.
87 * config/rs6000/vsx.md: Ditto.
89 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
91 * config/rs6000/altivec.md: Don't set length attribute to the default
92 value, for branch instructions.
93 * config/rs6000/darwin.md: Ditto.
94 * config/rs6000/rs6000.md: Ditto.
96 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
98 * config/rs6000/rs6000.md (length): Always define as const_int 4.
99 (unnamed conditional branch define_insn): Set length to 4 or 8
101 (<bd>_<mode>): Similar, for alternative 0.
102 (<bd>tf_<mode>): Ditto.
104 2018-08-16 Tamar Christina <tamar.christina@arm.com>
106 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
108 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
110 * doc/rtl.texi: Replace old RTX class names with new names.
113 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
115 * expmed.h (canonicalize_comparison): New declaration.
116 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
117 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
118 * optabs.c (prepare_cmp_insn): Likewise.
119 * rtl.h (unsigned_condition_p): New function which checks if a
120 comparison operator is unsigned.
122 2018-08-16 Nathan Sidwell <nathan@acm.org>
124 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
125 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
127 2018-08-16 Tamar Christina <tamar.christina@arm.com>
130 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
131 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
132 (mov<mov>): ..this and enable unconditionally.
134 2018-08-16 Tamar Christina <tamar.christina@arm.com>
136 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
138 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
140 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
141 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
142 "Common" with "Target".
144 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
146 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
147 * doc/invoke.texi (mmitigate-rop): Remove.
148 * config/i386/i386.c: Do not include "regrename.h".
149 (ix86_rop_should_change_byte_p, reg_encoded_number)
150 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
152 (ix86_reorg): Remove call to ix86_mitigate_rop.
153 * config/i386/i386.md (attr "modrm_class"): Remove.
154 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
155 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
156 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
158 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
160 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
161 allow folding of mergeh() and mergel() for the float and double types.
162 (fold_mergehl_helper): Rework to handle building a permute tree
165 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
167 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
170 2018-08-15 David Malcolm <dmalcolm@redhat.com>
172 * common.opt (fdiagnostics-show-labels): New option.
173 * diagnostic-show-locus.c (class layout_range): Add field
175 (class layout): Add field "m_show_labels_p".
176 (layout_range::layout_range): Add param "label" and use it to
178 (make_range): Pass in NULL for new "label" param of layout_range's
180 (layout::layout): Initialize m_show_labels_p.
181 (layout::maybe_add_location_range): Pass in loc_range->m_label
182 when constructing layout_range instances.
183 (struct line_label): New struct.
184 (layout::print_any_labels): New member function.
185 (layout::print_line): Call it if label-printing is enabled.
186 (selftest::test_one_liner_labels): New test.
187 (selftest::test_diagnostic_show_locus_one_liner): Call it.
188 * diagnostic.c (diagnostic_initialize): Initialize
189 context->show_labels_p.
190 * diagnostic.h (struct diagnostic_context): Add field
192 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
193 -fno-diagnostics-show-labels.
194 * dwarf2out.c (gen_producer_string): Add
195 OPT_fdiagnostics_show_labels to the ignored options.
196 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
198 (gcc_rich_location::maybe_add_expr): Likewise.
199 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
200 label" param, defaulting to NULL.
201 (gcc_rich_location::add_expr): Add "label" param.
202 (gcc_rich_location::maybe_add_expr): Likewise.
203 (class text_range_label): New class.
204 (class range_label_for_type_mismatch): New class.
205 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
206 of format_warning_va.
207 (fmtwarn_n): Likewise for new params of format_warning_n_va.
208 * lto-wrapper.c (merge_and_complain): Add
209 OPT_fdiagnostics_show_labels to the "pick one setting" options.
210 (append_compiler_options): Likewise to the dropped options.
211 (append_diag_options): Likewise to the passed-on options.
212 * opts.c (common_handle_option): Handle the new option.
213 * selftest-diagnostic.c
214 (test_diagnostic_context::test_diagnostic_context): Enable
216 * substring-locations.c: Include "gcc-rich-location.h".
217 (format_warning_n_va): Add "fmt_label" and "param_label" params
218 and use them as appropriate.
219 (format_warning_va): Add "fmt_label" and "param_label" params,
220 passing them on to format_warning_n_va.
221 (format_warning_at_substring): Likewise.
222 (format_warning_at_substring_n): Likewise.
223 * substring-locations.h (format_warning_va): Add "fmt_label" and
224 "param_label" params.
225 (format_warning_n_va): Likewise.
226 (format_warning_at_substring): Likewise.
227 (format_warning_at_substring_n): Likewise.
228 * toplev.c (general_init): Initialize global_dc->show_labels_p.
230 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
233 * builtins.c (expand_builtin_memcmp): Do not expand the call
234 when overflow is detected.
236 2018-08-15 Martin Sebor <msebor@redhat.com>
238 PR tree-optimization/71625
239 * config/aarch64/aarch64-builtins.c
240 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
242 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
244 * config/s390/s390.c (s390_reorg): Remove loop.
246 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
249 (darwin_function_switched_text_sections): Delete.
250 * gcc/config/darwin.h
251 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
253 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
256 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
257 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
258 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
260 2018-08-15 Martin Liska <mliska@suse.cz>
262 PR tree-optimization/86925
263 * predict.c (expr_expected_value_1): When taking
264 later predictor, assign also probability.
265 Use fold_build2_initializer_loc in order to fold
266 the expression in -frounding-math.
268 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
270 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
272 (expand_vec_perm_1): Try the new method.
274 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
277 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
278 Check whether lra_live_max_point is 0 before dividing.
280 2018-08-14 Martin Sebor <msebor@redhat.com>
282 PR tree-optimization/86650
283 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
284 (vrp_prop::check_mem_ref): Same.
286 2018-08-13 Liu Hao <lh_mouse@126.com>
288 * pretty-print.c (eat_esc_sequence): Swap the foreground and
289 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
290 and clear it thereafter, as it only works for DBCS.
292 2018-08-13 Liu Hao <lh_mouse@126.com>
294 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
295 handle returned by _get_osf_handle().
297 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
299 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
300 for folding vec_perm.
302 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
304 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
305 Add support for gimple-folding of vec_pack() and vec_unpack()
308 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
310 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
311 vec_xst variants to the list.
312 (rs6000_gimple_fold_builtin): Add support for folding unaligned
313 vector loads and stores.
315 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
317 * config.gcc (rs6000-ibm-aix4.x): Delete.
318 (rs6000-ibm-aix5.1): Delete.
319 (rs6000-ibm-aix5.2): Delete.
320 (rs6000-ibm-aix5.3): Delete.
321 * config/rs6000/aix43.h: Delete.
322 * config/rs6000/aix51.h: Delete.
323 * config/rs6000/aix52.h: Delete.
324 * config/rs6000/t-aix43: Delete.
326 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
328 * config/s390/s390.c (s390_decompose_constant_pool_ref):
330 (s390_decompose_address): Factor out constant pool ref
333 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
335 * config/nds32/nds32-predicates.c
336 (nds32_can_use_bclr_p): Change return type as bool.
337 (nds32_can_use_bset_p): Ditto.
338 (nds32_can_use_btgl_p): Ditto.
339 (nds32_can_use_bitci_p): Ditto.
340 * config/nds32/nds32-protos.h
341 (nds32_can_use_bclr_p): Change declaration.
342 (nds32_can_use_bset_p): Ditto.
343 (nds32_can_use_btgl_p): Ditto.
344 (nds32_can_use_bitci_p): Ditto.
346 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
349 Support -msched-prolog-epilog option.
350 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
352 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
354 * common/config/nds32/nds32-common.c
355 (nds32_option_optimization_table): Enalbe -malways-align.
357 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
359 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
361 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
362 OPT_misr_secure_ case.
363 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
364 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
365 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
367 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
368 (nds32_isr_info): New field security_level.
369 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
370 * config/nds32/nds32.md (return_internal): Consider critical attribute.
371 * config/nds32/nds32.opt (misr-secure): New option.
372 * config/nds32/nds32_init.inc: New file.
373 * config/nds32/nds32_isr.h: New file.
375 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
377 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
378 Update comment for atomic instructions.
379 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
380 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
382 (atomic_loaddi): Revise fence expansion to only emit fence prior to
383 load for __ATOMIC_SEQ_CST model.
384 (atomic_loaddi_1): Remove float register target.
385 (atomic_storedi): Handle CONST_INT values.
386 (atomic_storedi_1): Remove float register source. Add special case
388 (memory_barrier): New expander and insn.
390 2018-08-11 Jakub Jelinek <jakub@redhat.com>
392 PR tree-optimization/86835
393 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
394 new_stmt after def_gsi, make sure to insert new_square_stmt after
395 that stmt, not 2 stmts before it.
397 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
400 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
403 2018-08-10 Martin Liska <mliska@suse.cz>
406 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
408 * builtins.c (expand_builtin_expect_with_probability):
410 (expand_builtin_expect_with_probability): New function.
411 (build_builtin_expect_predicate): Add new argumnet probability
412 for BUILT_IN_EXPECT_WITH_PROBABILITY.
413 (fold_builtin_expect):
416 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
417 * builtins.h (fold_builtin_expect): Set new argument.
418 * doc/extend.texi: Document __builtin_expect_with_probability.
419 * doc/invoke.texi: Likewise.
420 * gimple-fold.c (gimple_fold_call): Pass new argument.
421 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
422 also BUILT_IN_EXPECT_WITH_PROBABILITY.
423 * predict.c (get_predictor_value): New function.
424 (expr_expected_value): Add new argument probability. Assume
425 that predictor and probability are always non-null.
426 (expr_expected_value_1): Likewise. For __builtin_expect and
427 __builtin_expect_with_probability set probability. Handle
428 combination in binary expressions.
429 (tree_predict_by_opcode): Simplify code by simply calling
431 (pass_strip_predict_hints::execute): Add handling of
432 BUILT_IN_EXPECT_WITH_PROBABILITY.
433 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
435 * tree.h (DECL_BUILT_IN_P): New function.
437 2018-08-10 Martin Liska <mliska@suse.cz>
439 PR tree-optimization/85799
440 * passes.def: Add argument for pass_strip_predict_hints.
441 * predict.c (class pass_strip_predict_hints): Add new argument
443 (strip_predictor_early): New function.
444 (pass_strip_predict_hints::execute): Call the function to
446 (strip_predict_hints): New function.
447 * predict.def: Fix comment.
449 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
451 * Makefile.in: Clarify which tm.texi to copy over to assert the
452 right to grant a GFDL license for all.
454 2018-08-09 Jeff Law <law@redhat.com>
456 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
459 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
461 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
464 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
466 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
467 reductions for variable-length vectors.
469 2018-08-09 David Malcolm <dmalcolm@redhat.com>
472 * common.opt (fdiagnostics-show-line-numbers): New option.
473 * diagnostic-show-locus.c (class layout): Add fields
474 "m_show_line_numbers_p" and "m_linenum_width";
475 (num_digits): New function.
476 (test_num_digits): New function.
477 (layout::layout): Initialize new fields. Update m_x_offset
478 logic to handle any left margin.
479 (layout::print_source_line): Print line number when requested.
480 (layout::start_annotation_line): New member function.
481 (layout::print_annotation_line): Call it.
482 (layout::print_leading_fixits): Likewise.
483 (layout::print_trailing_fixits): Likewise. Update calls to
484 move_to_column for new parameter.
485 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
486 it to potentially call start_annotation_line.
487 (layout::show_ruler): Call start_annotation_line.
488 (selftest::test_line_numbers_multiline_range): New selftest.
489 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
490 and selftest::test_line_numbers_multiline_range.
491 * diagnostic.c (diagnostic_initialize): Initialize
493 * diagnostic.h (struct diagnostic_context): Add field
494 "show_line_numbers_p".
495 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
496 -fno-diagnostics-show-line-numbers.
497 * dwarf2out.c (gen_producer_string): Add
498 OPT_fdiagnostics_show_line_numbers to the ignored options.
499 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
500 one setting" options.
501 (append_compiler_options): Likewise to the dropped options.
502 (append_diag_options): Likewise to the passed-on options.
503 * opts.c (common_handle_option): Handle the new option.
504 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
506 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
508 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
509 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
510 third argument of type "const signed char" to descriptions of
511 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
512 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
513 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
514 __builtin_bcdsub_ov functions.
516 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
518 PR tree-optimization/86858
519 * tree-vect-loop.c (vect_is_simple_reduction): Restore
520 flow_bb_inside_loop_p calls.
522 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
524 PR tree-optimization/86871
525 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
526 instead of gimple_assign_lhs.
528 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
531 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
532 register constraint to operand 0.
533 (add<mode>3_carryinC): Likewise.
534 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
536 2018-08-09 Martin Liska <mliska@suse.cz>
539 * common.opt: Remove extra line.
541 2018-08-09 Martin Liska <mliska@suse.cz>
543 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
544 at the end of a line, make first letter capital and end up
545 a sentence with a dot.
546 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
547 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
548 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
549 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
550 (PARAM_MAX_ISL_OPERATIONS): Likewise.
551 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
552 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
553 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
554 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
555 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
556 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
557 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
558 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
559 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
560 (PARAM_TREE_REASSOC_WIDTH): Likewise.
561 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
562 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
563 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
565 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
568 * config/s390/s390.c (s390_option_override_internal): Reduce the
569 stack-clash-protection-probe-interval param if it would be too big
572 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
575 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
576 * config/m68k/m68k.c (handle_move_double): Don't call
577 m68k_final_prescan_insn.
578 (m68k_adjust_decorated_operand): Renamed from
579 m68k_final_prescan_insn, remove first and third operand and
581 (print_operand): Call it.
582 (print_operand_address): Call it.
584 2018-08-08 Nathan Sidwell <nathan@acm.org>
586 * diagnostic.c (diagnostic_report_current_module): Use
587 linemap_included_from & linemap_included_from_linemap.
589 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
591 * config/aarch64/aarch64-cores.def: Add phecda core.
592 * config/aarch64/aarch64-tune.md: Regenerate.
593 * doc/invoke.texi: Add phecda core.
595 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
598 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
600 * config/s390/s390.md ("movti"): Add more alternatives for
601 constant to GPR copies.
603 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
605 * config/s390/s390.c: Fix whitespace damage throughout the file.
606 * config/s390/s390.h: Likewise.
607 * config/s390/tpf.h: Likewise.
609 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
611 * config/s390/s390.c (s390_loadrelative_operand_p):
612 Remove TARGET_CPU_ZARCH usages.
613 (s390_rtx_costs): Likewise.
614 (s390_legitimate_constant_p): Likewise.
615 (s390_cannot_force_const_mem): Likewise.
616 (legitimate_reload_constant_p): Likewise.
617 (s390_preferred_reload_class): Likewise.
618 (legitimize_pic_address): Likewise.
619 (legitimize_tls_address): Likewise.
620 (s390_split_branches): Removed.
621 (s390_add_execute): Removed.
622 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
623 (s390_mainpool_start): Likewise.
624 (s390_mainpool_finish): Likewise.
625 (s390_mainpool_cancel): Removed.
626 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
627 (s390_chunkify_cancel): Likewise.
628 (s390_return_addr_rtx): Likewise.
629 (s390_register_info): Remove split_branches_pending_p uages.
630 (s390_optimize_register_info): Likewise.
631 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
632 split_branches_pending_p usages.
633 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
634 (s390_load_got): Likewise.
635 (s390_expand_split_stack_prologue): Likewise.
636 (output_asm_nops): Likewise.
637 (s390_function_profiler): Likewise.
638 (s390_emit_call): Likewise.
639 (s390_conditional_register_usage): Likewise.
640 (s390_optimize_prologue): Likewise.
641 (s390_reorg): Remove TARGET_CPU_ZARCH and
642 split_branches_pending_p usages.
643 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
645 (s390_output_indirect_thunk_function): Likewise.
646 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
647 (TARGET_CPU_ZARCH_P): Removed.
648 (struct machine_function): Remove split_branches_pending_p.
649 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
651 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
653 * common/config/s390/s390-common.c (processor_flags_table):
655 * config.gcc: Remove with_arch/with_tune support.
656 * config/s390/2064.md: Remove cpu attribute comparisons.
657 * config/s390/driver-native.c (s390_host_detect_local_cpu):
659 * config/s390/linux.h (ASM_SPEC):
660 Remove -march support.
661 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
662 Use a table to get an arch level.
663 * config/s390/s390-opts.h (enum processor_type):
666 (processor_table): Remove entries, add arch_level values.
667 (s390_issue_rate): Remove cases.
668 (s390_option_override): Adjust
669 s390_option_override_internal() call.
670 (s390_option_override_internal): Remove deprecation warning.
671 (s390_valid_target_attribute_tree): Adjust
672 s390_option_override_internal() call.
673 * config/s390/s390.h (struct s390_processor):
674 Share with s390-c.c, add arch_level field.
675 * config/s390/s390.md:
676 Remove occurrences in cpu attribute.
677 * config/s390/s390.opt: Remove -march/-mtune support.
678 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
679 * doc/invoke.texi: Remove deprecation warning.
681 2018-08-08 Luis Machado <luis.machado@linaro.org>
683 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
685 (qdf24xx_tunings): Set vector cost structure to
688 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
689 <register_sextend>: Set to 3.
691 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
694 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
695 * config/aarch64/aarch64-simd.md
696 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
697 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
698 (aarch64_frecpx<mode>): ...this new pattern.
699 * config/aarch64/aarch64-simd-builtins.def: Remove comment
700 about aarch64_frecp<FRECP:frecp_suffix><mode>.
702 2018-08-07 Martin Liska <mliska@suse.cz>
705 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
706 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
707 * predict.def (PRED_MALLOC_NONNULL): New predictor.
708 * doc/extend.texi: Document that malloc attribute adds
711 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
714 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
715 Define to speculation_safe_value_not_needed.
717 2018-08-06 Jeff Law <law@redhat.com>
719 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
720 the vr_values instance to cprop_into_stmt.
721 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
722 (cprop_operand): Also query EVRP to determine if OP is a constant.
724 2018-08-06 Nathan Sidwell <nathan@acm.org>
726 * diagnostic.c (diagnostic_report_current_module): Reroll
727 included-at loop. Translate text.
729 2018-08-06 David Malcolm <dmalcolm@redhat.com>
731 * function-tests.c (selftest::test_expansion_to_rtl): Call
732 free_after_compilation.
734 2018-08-06 Alan Hayward <alan.hayward@arm.com>
736 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
738 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
740 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
741 loops with memory block operations from getting unrolled.
743 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
746 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
747 Define to speculation_safe_value_not_needed.
749 2018-08-06 Jeff Law <law@redhat.com>
751 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
754 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
757 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
758 with all enabled __intN types.
760 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
762 2018-08-06 Alan Hayward <alan.hayward@arm.com>
764 * alias.c (record_set): Check for clobber high.
765 * cfgexpand.c (expand_gimple_stmt): Likewise.
766 * combine-stack-adj.c (single_set_for_csa): Likewise.
767 * combine.c (find_single_use_1): Likewise.
768 (set_nonzero_bits_and_sign_copies): Likewise.
769 (get_combine_src_dest): Likewise.
770 (is_parallel_of_n_reg_sets): Likewise.
771 (try_combine): Likewise.
772 (record_dead_and_set_regs_1): Likewise.
773 (reg_dead_at_p_1): Likewise.
774 (reg_dead_at_p): Likewise.
775 * dce.c (deletable_insn_p): Likewise.
776 (mark_nonreg_stores_1): Likewise.
777 (mark_nonreg_stores_2): Likewise.
778 * df-scan.c (df_find_hard_reg_defs): Likewise.
779 (df_uses_record): Likewise.
780 (df_get_call_refs): Likewise.
781 * dwarf2out.c (mem_loc_descriptor): Likewise.
782 * haifa-sched.c (haifa_classify_rtx): Likewise.
783 * ira-build.c (create_insn_allocnos): Likewise.
784 * ira-costs.c (scan_one_insn): Likewise.
785 * ira.c (equiv_init_movable_p): Likewise.
786 (rtx_moveable_p): Likewise.
787 (interesting_dest_for_shprep): Likewise.
788 * jump.c (mark_jump_label_1): Likewise.
789 * postreload-gcse.c (record_opr_changes): Likewise.
790 * postreload.c (reload_cse_simplify): Likewise.
791 (struct reg_use): Add source expr.
792 (reload_combine): Check for clobber high.
793 (reload_combine_note_use): Likewise.
794 (reload_cse_move2add): Likewise.
795 (move2add_note_store): Likewise.
796 * print-rtl.c (print_pattern): Likewise.
797 * recog.c (decode_asm_operands): Likewise.
798 (store_data_bypass_p): Likewise.
799 (if_test_bypass_p): Likewise.
800 * regcprop.c (kill_clobbered_value): Likewise.
801 (kill_set_value): Likewise.
802 * reginfo.c (reg_scan_mark_refs): Likewise.
803 * reload1.c (maybe_fix_stack_asms): Likewise.
804 (eliminate_regs_1): Likewise.
805 (elimination_effects): Likewise.
806 (mark_not_eliminable): Likewise.
807 (scan_paradoxical_subregs): Likewise.
808 (forget_old_reloads_1): Likewise.
809 * reorg.c (find_end_label): Likewise.
810 (try_merge_delay_insns): Likewise.
811 (redundant_insn): Likewise.
812 (own_thread_p): Likewise.
813 (fill_simple_delay_slots): Likewise.
814 (fill_slots_from_thread): Likewise.
815 (dbr_schedule): Likewise.
816 * resource.c (update_live_status): Likewise.
817 (mark_referenced_resources): Likewise.
818 (mark_set_resources): Likewise.
819 * rtl.c (copy_rtx): Likewise.
820 * rtlanal.c (reg_referenced_p): Likewise.
821 (single_set_2): Likewise.
822 (noop_move_p): Likewise.
823 (note_stores): Likewise.
824 * sched-deps.c (sched_analyze_reg): Likewise.
825 (sched_analyze_insn): Likewise.
827 2018-08-06 Alan Hayward <alan.hayward@arm.com>
829 * cse.c (invalidate_reg): New function extracted from...
830 (invalidate): ...here.
831 (canonicalize_insn): Check for clobber high.
832 (invalidate_from_clobbers): invalidate clobber highs.
833 (invalidate_from_sets_and_clobbers): Likewise.
834 (count_reg_usage): Check for clobber high.
835 (insn_live_p): Likewise.
836 * cselib.c (cselib_expand_value_rtx_1):Likewise.
837 (cselib_invalidate_regno): Check for clobber in setter.
838 (cselib_invalidate_rtx): Pass through setter.
839 (cselib_invalidate_rtx_note_stores):
840 (cselib_process_insn): Check for clobber high.
841 * cselib.h (cselib_invalidate_rtx): Add operand.
843 2018-08-06 Alan Hayward <alan.hayward@arm.com>
845 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
846 (mark_not_eliminable): Likewise.
847 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
848 * lra-lives.c (process_bb_lives): Check for clobber high.
849 * lra.c (new_insn_reg): Remember clobber highs.
850 (collect_non_operand_hard_regs): Check for clobber high.
851 (lra_set_insn_recog_data): Likewise.
852 (add_regs_to_insn_regno_info): Likewise.
853 (lra_update_insn_regno_info): Likewise.
855 2018-08-06 Alan Hayward <alan.hayward@arm.com>
857 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
858 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
860 2018-08-06 Alan Hayward <alan.hayward@arm.com>
862 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
863 (copy_insn_1): Likewise.
864 (gen_hard_reg_clobber_high): New gen function.
865 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
866 * genemit.c (gen_exp): Likewise.
867 (gen_emit_seq): Pass through info.
868 (gen_insn): Check for CLOBBER_HIGH.
869 (gen_expand): Pass through info.
870 (gen_split): Likewise.
871 (output_add_clobbers): Likewise.
872 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
873 (remove_clobbers): Likewise.
874 * rtl.h (gen_hard_reg_clobber_high): New declaration.
876 2018-08-06 Alan Hayward <alan.hayward@arm.com>
878 * doc/rtl.texi (clobber_high): Add.
879 (parallel): Add in clobber high
880 * rtl.c (rtl_check_failed_code3): Add function.
881 * rtl.def (CLOBBER_HIGH): Add expression.
882 * rtl.h (RTL_CHECKC3): Add macro.
883 (rtl_check_failed_code3): Add declaration.
886 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
889 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
890 cfun->machine->max_used_stack_alignment if needed.
892 2018-08-04 Martin Sebor <msebor@redhat.com>
894 PR tree-optimization/86571
895 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
898 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
901 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
904 2018-08-03 Jeff Law <law@redhat.com>
907 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
908 Define to speculation_safe_value_not_needed.
910 2018-08-03 David Malcolm <dmalcolm@redhat.com>
912 * doc/gcov.texi (-x): Remove duplicate "to".
913 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
914 (-Wif-not-aligned): Remove duplicate "is".
915 (-flto): Remove duplicate "the".
916 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
918 (MSP430 Options): Remove duplicate "and" from the description
919 of "-mgprel-sec=regexp".
920 (x86 Options): Remove duplicate copies of "vmldLog102" and
921 vmlsLog104 from description of "-mveclibabi=type".
923 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
925 * internal-fn.h (first_commutative_argument): Declare.
926 * internal-fn.c (first_commutative_argument): New function.
927 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
928 restrictions for pattern statements. Use first_commutative_argument
929 to look for commutative operands in calls to internal functions.
931 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
933 * Makefile.in (wide-int-range.o): New.
934 * tree-vrp.c: Move all the wide_int_* functions to...
935 * wide-int-range.cc: ...here.
936 * tree-vrp.h: Move all the wide_int_* prototypes to...
937 * wide-int-range.h: ...here.
939 2018-08-03 Tom de Vries <tdevries@suse.de>
941 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
943 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
944 * except.c (output_function_exception_table): Do early exit if
945 targetm_common.except_unwind_info (&global_options) == UI_NONE.
947 2018-08-03 Martin Liska <mliska@suse.cz>
949 * predict.c (dump_prediction): Change to 2 digits
952 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
954 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
956 (falkor_am_1_gtov_gtov): ... a new insn reservation.
958 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
960 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
961 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
962 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
963 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
964 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
966 2018-08-02 David Malcolm <dmalcolm@redhat.com>
968 * diagnostic-show-locus.c (diagnostic_show_locus): Use
969 pp_take_prefix when saving the existing prefix.
970 * diagnostic.c (diagnostic_append_note): Likewise.
971 * langhooks.c (lhd_print_error_function): Likewise.
972 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
973 param's type. Free the existing prefix.
974 (pp_take_prefix): New function.
975 (pretty_printer::pretty_printer): Drop the prefix parameter.
976 Rename the length parameter to match the comment.
977 (pretty_printer::~pretty_printer): Free the prefix.
978 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
980 (struct pretty_printer): Drop the "const" from "prefix" field's
981 type and clarify memory management.
982 (pp_set_prefix): Drop the "const" from the 2nd param.
983 (pp_take_prefix): New decl.
985 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
987 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
988 for word_mode_ok here instead of passing as argument.
989 (expand_block_compare): Change select_block_compare_mode() call.
990 (expand_strncmp_gpr_sequence): New function.
991 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
993 2018-08-02 Jeff Law <law@redhat.com>
996 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
997 Define to speculation_safe_value_not_needed.
1000 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1001 Define to speculation_safe_value_not_needed.
1003 2018-08-02 Tom de Vries <tdevries@suse.de>
1006 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1007 function. Return UI_TARGET unconditionally.
1008 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1009 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1011 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1013 * genemit.c (print_overload_test): Fix typo.
1015 2018-08-02 Richard Biener <rguenther@suse.de>
1017 PR tree-optimization/86816
1018 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1019 which checks for value availability before querying it.
1020 (gvn_uses_equal): Use it.
1021 (same_succ_hash): Likewise.
1022 (gimple_equal_p): Likewise.
1024 2018-08-02 Nick Clifton <nickc@redhat.com>
1027 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1028 Define to speculation_safe_value_not_needed.
1031 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1032 Define to speculation_safe_value_not_needed.
1035 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1036 Define to speculation_safe_value_not_needed.
1039 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1040 Define to speculation_safe_value_not_needed.
1043 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1044 Define to speculation_safe_value_not_needed.
1047 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1048 Define to speculation_safe_value_not_needed.
1051 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1052 Define to speculation_safe_value_not_needed.
1055 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1056 Define to speculation_safe_value_not_needed.
1059 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1060 speculation_safe_value_not_needed.
1063 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1064 to speculation_safe_value_not_needed.
1066 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1068 * doc/md.texi: Expand the documentation of instruction names
1069 to mention port-local uses. Document '@' in pattern names.
1070 * read-md.h (overloaded_instance, overloaded_name): New structs.
1072 (md_reader::handle_overloaded_name): New member function.
1073 (md_reader::get_overloads): Likewise.
1074 (md_reader::m_first_overload): New member variable.
1075 (md_reader::m_next_overload_ptr): Likewise.
1076 (md_reader::m_overloads_htab): Likewise.
1077 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1078 m_next_overload_ptr and m_overloads_htab.
1079 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1080 (get_mode_token, get_code_token, get_int_token): New functions.
1081 (map_attr_string): Add an optional argument that passes back
1082 the associated iterator.
1083 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1084 (md_reader::handle_overloaded_name, add_overload_instance): New
1086 (apply_iterators): Handle '@' names. Report an error if '@'
1087 is used without iterators.
1088 (initialize_iterators): Initialize the new iterator_group fields.
1089 * genopinit.c (handle_overloaded_code_for)
1090 (handle_overloaded_gen): New functions.
1091 (main): Use them to print declarations of maybe_code_for_* and
1092 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1093 * genemit.c (print_overload_arguments, print_overload_test)
1094 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1095 (main): Use it to print definitions of maybe_code_for_* and
1096 maybe_gen_* functions.
1097 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1098 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1099 instead of explicit mode checks.
1100 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1101 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1102 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1103 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1104 (aarch64_expand_compare_and_swap): Likewise
1105 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1106 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1107 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1108 (aarch64_constant_pool_reload_icode): Delete.
1109 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1110 instead of aarch64_constant_pool_reload_icode. Use
1111 code_for_aarch64_reload_mov instead of explicit mode checks.
1112 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1113 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1114 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1115 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1116 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1117 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1118 (aarch64_atomic_load_op_code): Delete.
1119 (aarch64_emit_atomic_load_op): Likewise.
1120 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1121 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1122 instead of aarch64_emit_atomic_load_op.
1123 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1124 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1125 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1126 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1127 character before the pattern name.
1128 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1129 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1130 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1131 (aarch64_frecps<mode>): Likewise.
1132 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1133 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1134 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1135 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1136 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1138 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1140 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1141 Allow HFmode constants if TARGET_FP_F16INST.
1143 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1146 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1147 No longer check last store for clobber of address register.
1149 2018-08-02 Martin Liska <mliska@suse.cz>
1151 PR gcov-profile/86817
1152 * gcov.c (process_all_functions): New function.
1154 (process_file): Move functions processing to
1155 process_all_functions.
1157 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1159 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1160 "const" to the "gimple *" and "rtx_insn *" parameters.
1161 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1163 (dump_location_t::dump_location_t): Likewise.
1165 2018-08-01 Martin Sebor <msebor@redhat.com>
1167 PR tree-optimization/86650
1168 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1169 rather than a "gcall *". Directly pass the data of interest
1170 to percent_K_format, rather than building a temporary CALL_EXPR
1172 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1173 (gimple_fold_builtin_strncat): Adjust.
1174 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1175 gcall* argument with gimple*.
1176 * gimple-ssa-warn-restrict.c (check_call): Same.
1177 (wrestrict_dom_walker::before_dom_children): Same.
1178 (builtin_access::builtin_access): Same.
1179 (check_bounds_or_overlap): Same
1180 (maybe_diag_overlap): Same.
1181 (maybe_diag_offset_bounds): Same.
1182 * tree-diagnostic.c (default_tree_printer): Move usage of
1183 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1185 * tree-pretty-print.c (percent_K_format): Add argument.
1186 * tree-pretty-print.h: Add argument.
1187 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1188 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1189 (maybe_diag_stxncpy_trunc): Same.
1190 (handle_builtin_stxncpy): Same.
1191 (handle_builtin_strcat): Same.
1193 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1195 * match.pd: Optimise pointer range checks.
1197 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1199 PR tree-optimization/86758
1200 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1201 to remove pattern statements.
1203 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1205 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1206 result of dfs_enumerate_from when constructing stmt_vec_infos,
1207 instead of additionally calling get_loop_body.
1209 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1211 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1213 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1214 When creating an iv, assert that the step is not known to be zero.
1215 (vect_setup_realignment): Update call accordingly.
1216 * tree-vect-stmts.c (vectorizable_store): Likewise.
1217 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1219 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1221 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1222 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1223 (vectorizable_reduction): Likewise.
1224 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1225 (vect_detect_hybrid_slp_stmts): Likewise.
1226 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1228 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1230 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1231 (wide_int_set_zero_nonzero_bits): ...this.
1232 (zero_nonzero_bits_from_vr): Rename to...
1233 (vrp_set_zero_nonzero_bits): ...this.
1234 (extract_range_from_multiplicative_op_1): Abstract wide int
1236 (wide_int_range_multiplicative_op): ...here.
1237 (extract_range_from_binary_expr_1): Extract wide int binary
1238 operations into their own functions.
1239 (wide_int_range_lshift): New.
1240 (wide_int_range_can_optimize_bit_op): New.
1241 (wide_int_range_shift_undefined_p): New.
1242 (wide_int_range_bit_xor): New.
1243 (wide_int_range_bit_ior): New.
1244 (wide_int_range_bit_and): New.
1245 (wide_int_range_trunc_mod): New.
1246 (extract_range_into_wide_ints): New.
1247 (vrp_shift_undefined_p): New.
1248 (extract_range_from_multiplicative_op): New.
1249 (vrp_can_optimize_bit_op): New.
1250 * tree-vrp.h (value_range::dump): New.
1251 (wide_int_range_multiplicative_op): New.
1252 (wide_int_range_lshift):New.
1253 (wide_int_range_shift_undefined_p): New.
1254 (wide_int_range_bit_xor): New.
1255 (wide_int_range_bit_ior): New.
1256 (wide_int_range_bit_and): New.
1257 (wide_int_range_trunc_mod): New.
1258 (zero_nonzero_bits_from_bounds): Rename to...
1259 (wide_int_set_zero_nonzero_bits): ...this.
1260 (zero_nonzero_bits_from_vr): Rename to...
1261 (vrp_set_zero_nonzero_bits): ...this.
1262 (range_easy_mask_min_max): Rename to...
1263 (wide_int_range_can_optimize_bit_op): this.
1264 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1265 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1267 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1269 * tree-vectorizer.h (vect_orig_stmt): New function.
1270 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1271 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1272 (vect_create_epilog_for_reduction): Likewise.
1273 (vectorizable_live_operation): Likewise.
1274 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1275 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1276 * tree-vect-stmts.c (vectorizable_call): Likewise.
1277 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1279 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1281 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1283 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1284 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1285 (vect_transform_loop): Likewise.
1286 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1288 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1290 * tree-vectorizer.h (vect_schedule_slp): Return void.
1291 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1292 (vect_schedule_slp): Likewise.
1294 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1296 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1298 (vect_transform_loop): Update calls accordingly. Schedule SLP
1299 instances before the main loop, if any exist.
1301 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1303 PR tree-optimization/86749
1304 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1305 If the lhs is used in a COND_EXPR, check that it is being used
1306 as the "then" or "else" value.
1308 2018-08-01 Tom de Vries <tdevries@suse.de>
1311 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1312 speculation_safe_value_not_needed.
1314 2018-08-01 Richard Biener <rguenther@suse.de>
1316 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1319 2018-08-01 Martin Liska <mliska@suse.cz>
1321 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1322 format how successful transformation is dumped.
1323 (gimple_mod_pow2_value_transform): Likewise.
1324 (gimple_mod_subtract_transform): Likewise.
1325 (gimple_stringops_transform): Likewise.
1327 2018-08-01 Martin Liska <mliska@suse.cz>
1330 * value-prof.c (interesting_stringop_to_profile_p):
1331 Simplify the code and add BUILT_IN_MEMMOVE.
1332 (gimple_stringops_transform): Likewise.
1334 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1336 * config/aarch64/aarch64-simd.md
1337 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1338 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1339 use GPI iterator instead of SI mode.
1341 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1343 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1344 rs6000_speculation_barrier.
1345 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1346 new barrier pattern name.
1348 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1350 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1351 (speculation_barrier): New insn.
1353 2018-08-01 Richard Biener <rguenther@suse.de>
1356 * graphite.h: Include isl/id.h and isl/space.h to allow build
1359 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1362 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1363 mode for COFF targets.
1364 * defaults.h (TARGET_COFF): Define.
1365 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1366 TARGET_COFF): Define.
1367 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1368 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1369 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1371 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1373 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1374 * gimple-streamer-out.c (output_bb): Save it.
1375 * lto-streamer-in.c (input_struct_function_base): Restore
1376 instance discriminator if available. Create map on demand.
1377 * lto-streamer-out.c (output_struct_function_base): Save it if
1379 * final.c (decl_to_instance_map): Document LTO strategy.
1381 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1382 Olivier Hainque <hainque@adacore.com>
1384 * debug.h (decl_to_instance_map_t): New type.
1385 (decl_to_instance_map): Declare.
1386 (maybe_create_decl_to_instance_map): New inline function.
1387 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1388 to track basic block discriminators.
1389 (final_start_function_1): Initialize them.
1390 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1392 (decl_to_instance_map): New variable.
1393 (map_decl_to_instance, maybe_set_discriminator): New functions.
1394 (notice_source_line): Set discriminator.
1396 2018-07-31 Ian Lance Taylor <iant@golang.org>
1398 * targhooks.c (default_have_speculation_safe_value): Add
1401 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1403 * dump-context.h: Include "pretty-print.h".
1404 (dump_context::refresh_dumps_are_enabled): New decl.
1405 (dump_context::emit_item): New decl.
1406 (class dump_context): Add fields "m_test_pp" and
1408 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1409 (temp_dump_context::get_dumped_text): New decl.
1410 (class temp_dump_context): Add field "m_pp".
1411 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1412 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1414 (set_dump_file): Update for above change.
1415 (set_alt_dump_file): Likewise.
1416 (dump_loc): New overload, taking a pretty_printer *.
1417 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1418 to any test pretty-printer.
1419 (make_item_for_dump_gimple_stmt): New function, adapted from
1420 optinfo::add_gimple_stmt.
1421 (dump_context::dump_gimple_stmt): Call it, and use the result,
1422 eliminating the direct usage of dump_file and alt_dump_file in
1423 favor of indirectly using them via emit_item.
1424 (make_item_for_dump_gimple_expr): New function, adapted from
1425 optinfo::add_gimple_expr.
1426 (dump_context::dump_gimple_expr): Call it, and use the result,
1427 eliminating the direct usage of dump_file and alt_dump_file in
1428 favor of indirectly using them via emit_item.
1429 (make_item_for_dump_generic_expr): New function, adapted from
1431 (dump_context::dump_generic_expr): Call it, and use the result,
1432 eliminating the direct usage of dump_file and alt_dump_file in
1433 favor of indirectly using them via emit_item.
1434 (make_item_for_dump_printf_va): New function, adapted from
1435 optinfo::add_printf_va.
1436 (make_item_for_dump_printf): New function.
1437 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1438 and use the result, eliminating the direct usage of dump_file and
1439 alt_dump_file in favor of indirectly using them via emit_item.
1440 (make_item_for_dump_dec): New function.
1441 (dump_context::dump_dec): Call it, and use the result,
1442 eliminating the direct usage of dump_file and alt_dump_file in
1443 favor of indirectly using them via emit_item.
1444 (make_item_for_dump_symtab_node): New function, adapted from
1445 optinfo::add_symtab_node.
1446 (dump_context::dump_symtab_node): Call it, and use the result,
1447 eliminating the direct usage of dump_file and alt_dump_file in
1448 favor of indirectly using them via emit_item.
1449 (dump_context::begin_scope): Reimplement, avoiding direct usage
1450 of dump_file and alt_dump_file in favor of indirectly using them
1452 (dump_context::emit_item): New member function.
1453 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1454 Set up test pretty-printer on the underlying context. Call
1455 refresh_dumps_are_enabled.
1456 (temp_dump_context::~temp_dump_context): Call
1457 refresh_dumps_are_enabled.
1458 (temp_dump_context::get_dumped_text): New member function.
1459 (selftest::verify_dumped_text): New function.
1460 (ASSERT_DUMPED_TEXT_EQ): New macro.
1461 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1462 and then without optinfo enabled. Add uses of
1463 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1464 * dumpfile.h: Update comment for the dump_* API.
1465 * optinfo-emit-json.cc
1466 (selftest::test_building_json_from_dump_calls): Update for new
1467 param for temp_dump_context ctor.
1468 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1469 and "m_owned" field.
1470 (optinfo_item::~optinfo_item): Likewise.
1471 (optinfo::add_item): New member function.
1472 (optinfo::emit): Update comment.
1473 (optinfo::add_string): Delete.
1474 (optinfo::add_printf): Delete.
1475 (optinfo::add_printf_va): Delete.
1476 (optinfo::add_gimple_stmt): Delete.
1477 (optinfo::add_gimple_expr): Delete.
1478 (optinfo::add_tree): Delete.
1479 (optinfo::add_symtab_node): Delete.
1480 (optinfo::add_dec): Delete.
1481 * optinfo.h (class dump_context): New forward decl.
1482 (optinfo::add_item): New decl.
1483 (optinfo::add_string): Delete.
1484 (optinfo::add_printf): Delete.
1485 (optinfo::add_printf_va): Delete.
1486 (optinfo::add_gimple_stmt): Delete.
1487 (optinfo::add_gimple_expr): Delete.
1488 (optinfo::add_tree): Delete.
1489 (optinfo::add_symtab_node): Delete.
1490 (optinfo::add_dec): Delete.
1491 (optinfo::add_poly_int): Delete.
1492 (optinfo_item::optinfo_item): Remove "owned" param.
1493 (class optinfo_item): Remove field "m_owned".
1495 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1498 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1499 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1500 requested variable alignment.
1501 (expand_one_ssa_partition): Likewise.
1502 (expand_one_var): Likewise.
1504 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1506 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1507 to speculation_safe_value_not_needed.
1509 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1511 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1512 * targhooks.c (speculation_safe_value_not_needed): New function.
1513 * target.def (have_speculation_safe_value): Update documentation.
1514 * doc/tm.texi: Regenerated.
1516 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1518 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1519 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1521 (despeculate_copy<ALLI:mode>_insn): New insn.
1522 (despeculate_copyti_insn): New insn.
1523 (despeculate_simple<ALLI:mode>): New insn
1524 (despeculate_simpleti): New insn.
1525 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1527 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1528 aarch64_speculation_safe_value.
1529 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1531 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1533 * config/aarch64/aarch64-speculation.cc: New file.
1534 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1535 before pass_reorder_blocks.
1536 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1538 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1539 X14 and X15 when tracking speculation.
1540 * config/aarch64/aarch64.md (register name constants): Add
1541 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1542 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1543 (speculation_barrier): New insn attribute.
1544 (cmp<mode>): Allow SP in comparisons.
1545 (speculation_tracker): New insn.
1546 (speculation_barrier): Add speculation_barrier attribute.
1547 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1548 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1549 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1551 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1553 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1554 aarch64_track_speculation is true.
1555 (tb<optab><mode>1): Likewise.
1556 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1557 generate CB[N]Z when tracking speculation.
1558 (aarch64_split_compare_and_swap): Likewise.
1559 (aarch64_split_atomic_op): Likewise.
1561 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1563 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1565 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1567 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1568 (speculation_barrier): New insn.
1570 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1572 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1573 * config/arm/arm.md (speculation_barrier): New expand.
1574 (speculation_barrier_insn): New pattern.
1576 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1578 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1579 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1580 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1581 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1583 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1584 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1585 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1586 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1587 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1588 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1589 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1590 * builtins.c (expand_speculation_safe_value): New function.
1591 (expand_builtin): Call it.
1592 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1593 * doc/extend.texi: Document __builtin_speculation_safe_value.
1594 * doc/md.texi: Document "speculation_barrier" pattern.
1595 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1596 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1597 * doc/tm.texi: Regenerated.
1598 * target.def (have_speculation_safe_value, speculation_safe_value): New
1600 * targhooks.c (default_have_speculation_safe_value): New function.
1601 (default_speculation_safe_value): New function.
1602 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1603 (default_speculation_safe_value): Add prototype.
1605 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1607 * dump-context.h (dump_context::dump_loc): New decl.
1608 * dumpfile.c (dump_context::dump_loc): New member function.
1609 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1610 and dump_gimple_stmt.
1611 (dump_context::dump_gimple_expr_loc): Likewise, using
1613 (dump_context::dump_generic_expr_loc): Likewise, using
1615 (dump_context::dump_printf_loc_va): Likewise, using
1617 (dump_context::begin_scope): Explicitly using the global function
1618 "dump_loc", rather than the member function.
1620 2018-07-31 Martin Sebor <msebor@redhat.com>
1622 PR tree-optimization/86741
1623 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1625 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1627 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1628 depend on whether prefetch instructions will be emitted or not.
1629 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1630 will be emitted or not.
1631 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1632 (TARGET_SETMEM_PFD): New macros.
1634 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1636 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1637 (NULL_STMT_VEC_INFO): Delete.
1638 (stmt_vec_info::operator*): Likewise.
1639 (stmt_vec_info::operator gimple *): Likewise.
1640 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1641 of NULL_STMT_VEC_INFO.
1642 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1643 (vect_reassociating_reduction_p): Likewise.
1644 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1645 (vectorizable_store): Likewise.
1646 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1647 (vec_info::free_stmt_vec_infos): Likewise.
1649 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1651 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1652 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1654 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1656 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1657 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1658 (vec_info::free_stmt_vec_info): New private member functions.
1659 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1660 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1661 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1662 set_stmt_vec_info_vec and free_stmt_vec_infos.
1663 * tree-vect-loop.c (_loop_vec_info): Remove call to
1664 set_stmt_vec_info_vec.
1665 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1666 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1667 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1668 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1669 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1670 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1671 with the clearing in the xcalloc.
1672 (stmt_vec_info_vec): Delete.
1673 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1674 (vectorize_loops): Likewise.
1675 (vec_info::~vec_info): Remove argument from call to
1676 free_stmt_vec_infos.
1677 (vec_info::add_stmt): Remove vinfo argument from call to
1680 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1682 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1683 rather than a gimple stmt.
1684 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1685 information for pattern statements when passed the original
1686 statement; instead wait to be passed the pattern statement itself.
1687 Don't call set_vinfo_for_stmt here.
1688 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1689 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1690 stmt_vec_infos here.
1691 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1692 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1693 stmt_vec_infos entry.
1695 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1697 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1698 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1699 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1700 * tree-vect-stmts.c (vectorizable_call): Likewise.
1701 (vectorizable_simd_clone_call): Likewise.
1703 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1705 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1706 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1707 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1708 * tree-vect-loop.c (vect_transform_loop): Likewise.
1709 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1710 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1712 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1714 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1715 (vect_dr_stmt): Delete.
1716 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1717 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1719 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1720 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1721 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1722 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1723 (vect_analyze_data_refs): Likewise.
1724 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1726 (vect_find_same_alignment_drs): Likewise.
1727 (vect_slp_analyze_node_dependences): Update calls accordingly.
1728 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1729 instead of DR_VECT_AUX.
1730 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1731 of a vector data references. Use vec_info::lookup_dr instead of
1733 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1734 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1735 instead of DR_VECT_AUX.
1737 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1739 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1741 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1743 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1744 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1745 (vect_gen_prolog_loop_niters): Likewise.
1747 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1749 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1750 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1751 (vect_known_alignment_in_bytes, vect_dr_behavior)
1752 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1753 instead of data_references. Update calls to other routines for
1754 which the same change has been made.
1755 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1756 dr_vec_infos instead of stmt_vec_infos.
1757 (vect_analyze_data_ref_dependence): Update call accordingly.
1758 (vect_slp_analyze_data_ref_dependence)
1759 (vect_record_base_alignments): Use DR_VECT_AUX.
1760 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1761 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1762 (vector_alignment_reachable_p, vect_get_data_access_cost)
1763 (vect_peeling_supportable, vect_analyze_group_access_1)
1764 (vect_analyze_group_access, vect_analyze_data_ref_access)
1765 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1766 (vect_compile_time_alias, vect_small_gap_p)
1767 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1768 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1769 instead of data_references. Update calls to other routines for
1770 which the same change has been made.
1771 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1772 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1773 (vect_slp_analyze_and_verify_node_alignment)
1774 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1775 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1776 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1778 (_vect_peel_info::dr): Replace with...
1779 (_vect_peel_info::dr_info): ...this new field.
1780 (vect_peeling_hash_get_most_frequent)
1781 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1782 (vect_peeling_hash_get_lowest_cost):
1783 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1784 routines for which the same change has been made.
1785 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1787 * tree-vect-loop-manip.c (get_misalign_in_elems)
1788 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1790 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1791 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1792 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1793 (get_group_load_store_type, get_negative_load_store_type)
1794 (vect_get_data_ptr_increment, vectorizable_store)
1795 (vectorizable_load): Likewise.
1796 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1797 Update calls to other routines for which the same change has been made.
1799 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1801 * tree-vectorizer.h (vec_info::move_dr): New member function.
1802 (dataref_aux): Rename to...
1803 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1804 (_stmt_vec_info::dr_aux): Update accordingly.
1805 (_stmt_vec_info::data_ref_info): Delete.
1806 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1807 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1808 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1809 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1811 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1812 (STMT_VINFO_DR_INFO): New macro.
1813 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1814 (set_dr_misalignment): Update after rename of dataref_aux.
1815 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1816 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1817 initialization of STMT_VINFO_DATA_REF.
1818 * tree-vectorizer.c (vec_info::move_dr): New function.
1819 * tree-vect-patterns.c (vect_recog_bool_pattern)
1820 (vect_recog_mask_conversion_pattern)
1821 (vect_recog_gather_scatter_pattern): Use it.
1822 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1823 the "dr" and "stmt" fields of dr_vec_info instead of
1824 STMT_VINFO_DATA_REF.
1826 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1828 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1829 (is_pattern_stmt_p): Use it.
1830 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1831 on pattern statements.
1833 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1835 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1836 original stmt as a stmt_vec_info rather than a gimple stmt.
1837 (vect_pattern_recog_1): Take the statement directly as a
1838 stmt_vec_info, rather than via a gimple_stmt_iterator.
1839 Update call to vect_mark_pattern_stmts.
1840 (vect_pattern_recog): Update calls accordingly.
1842 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1844 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1845 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1846 a vect_def_type for the first argument.
1847 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1848 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1849 operand if it isn't defined by a vectorized statement.
1850 (vect_build_gather_load_calls): Remove the mask_dt argument and
1851 update calls to vect_get_vec_def_for_stmt_copy.
1852 (vectorizable_bswap): Likewise the dt argument.
1853 (vectorizable_call): Update calls to vectorizable_bswap and
1854 vect_get_vec_def_for_stmt_copy.
1855 (vectorizable_simd_clone_call, vectorizable_assignment)
1856 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1857 (vectorizable_comparison): Update calls to
1858 vect_get_vec_def_for_stmt_copy.
1859 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1861 (vect_get_loop_based_defs): Remove dt argument and update call
1862 to vect_get_vec_def_for_stmt_copy.
1863 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1864 and vect_get_vec_def_for_stmt_copy.
1865 (vectorizable_load): Update calls to vect_build_gather_load_calls
1866 and vect_get_vec_def_for_stmt_copy.
1867 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1868 (vectorizable_reduction, vectorizable_live_operation): Update calls
1869 to vect_get_vec_def_for_stmt_copy.
1871 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1873 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1874 and gimple stmt arguments with a stmt_vec_info.
1875 (vect_record_base_alignments): Update calls accordingly.
1876 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1877 and gimple stmt arguments with a stmt_vec_info.
1878 (vect_build_slp_tree_1): Remove vinfo argument and update call
1879 to vect_record_max_nunits.
1880 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1881 and vect_record_max_nunits.
1883 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1885 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1886 file and take a stmt_vec_info instead of a gimple stmt.
1887 (supportable_widening_operation, vect_finish_replace_stmt)
1888 (vect_finish_stmt_generation, vect_get_store_rhs)
1889 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1890 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1891 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1892 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1893 (vect_create_data_ref_ptr, bump_vector_ptr)
1894 (vect_permute_store_chain, vect_setup_realignment)
1895 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1896 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1897 (vectorizable_reduction, vectorizable_induction)
1898 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1899 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1900 than gimple stmts as arguments.
1901 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1902 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1903 (can_group_stmts_p, vect_check_gather_scatter)
1904 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1905 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1906 (vect_permute_load_chain, vect_shift_permute_load_chain)
1907 (vect_transform_grouped_load)
1908 (vect_record_grouped_load_vectors): Likewise.
1909 * tree-vect-loop.c (vect_fixup_reduc_chain)
1910 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1911 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1912 (vectorizable_reduction, vectorizable_induction)
1913 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1914 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1915 (vect_get_load_store_mask): Likewise.
1916 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1917 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1918 * tree-vect-stmts.c (vect_mark_relevant)
1919 (is_simple_and_all_uses_invariant)
1920 (exist_non_indexing_operands_for_use_p, process_use)
1921 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1922 (vect_get_vec_def_for_operand, vect_get_vec_defs)
1923 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1924 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1925 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1926 (get_negative_load_store_type, get_load_store_type)
1927 (vect_check_load_store_mask, vect_check_store_rhs)
1928 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1929 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1930 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1931 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1932 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1933 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1934 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1935 (supportable_widening_operation): Likewise.
1937 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1939 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1940 a stmt_vec_info instead of a gcall.
1941 (vect_check_gather_scatter): Update call accordingly.
1942 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1944 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1945 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1946 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1947 instead of a gimple stmt.
1948 (vect_transform_loop): Update calls accordingly.
1949 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1950 stmt_vec_infos instead of gimple stmts.
1951 (vect_analyze_slp_instance): Update use accordingly.
1952 * tree-vect-stmts.c (read_vector_array, write_vector_array)
1953 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1954 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1955 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1956 (vect_gen_widened_results_half, vect_get_loop_based_defs)
1957 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1958 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1959 down to subroutines.
1961 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1963 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1964 of the worklist from a vector of gimple stmts to a vector of
1966 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1967 (vect_mark_stmts_to_be_vectorized): Likewise
1969 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1971 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1972 statement before passing it to vect_analyze_stmt.
1973 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1974 the chain of phi vector definitions. Track the exit phi via its
1976 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1977 STMT_VINFO_REDUC_DEF.
1978 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1979 stmt_vec_infos to handle the statement chains.
1980 (vect_get_slp_defs): Record the first statement in the node
1981 using a stmt_vec_info.
1982 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1983 statements here and pass their stmt_vec_info down to subroutines.
1984 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1985 down to vect_finish_stmt_generation.
1986 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1987 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1988 stmt_vec_infos to subroutines.
1989 (vect_remove_stores): Use stmt_vec_infos to handle the statement
1992 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1994 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1995 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1996 (vect_permute_store_chain, vect_setup_realignment)
1997 (vect_permute_load_chain, vect_shift_permute_load_chain)
1998 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1999 stmts internally, and when passing values to other vectorizer routines.
2000 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2001 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2002 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2003 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2004 (vectorizable_reduction, vectorizable_induction)
2005 (vectorizable_live_operation, vect_transform_loop_stmt)
2006 (vect_transform_loop): Likewise.
2007 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2008 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2009 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2010 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2011 (vect_slp_analyze_node_operations_1): Likewise.
2012 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2013 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2014 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2015 (vect_finish_stmt_generation_1, get_group_load_store_type)
2016 (get_load_store_type, vect_build_gather_load_calls)
2017 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2018 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2019 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2020 (vectorizable_store, vectorizable_load, vectorizable_condition)
2021 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2022 (supportable_widening_operation): Likewise.
2023 (vect_get_vector_types_for_stmt): Likewise.
2024 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2026 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2028 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2029 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2030 (vect_permute_store_chain, vect_permute_load_chain)
2031 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2032 repeated stmt_vec_info lookups.
2033 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2034 (vect_update_ivs_after_vectorizer): Likewise.
2035 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2036 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2037 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2038 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2039 (vect_bb_slp_scalar_cost): Likewise.
2040 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2042 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2044 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2045 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2046 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2047 to get gassigns and gcalls, rather than operating on generc gimple
2049 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2050 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2051 (vectorizable_load, vect_analyze_stmt): Likewise.
2052 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2056 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2057 return stmt_vec_infos rather than gimple stmts. Do not accept
2059 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2061 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2062 Update use of get_later_stmt.
2063 (vect_get_constant_vectors): Update call accordingly.
2064 (vect_schedule_slp_instance): Likewise
2065 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2066 (vect_slp_analyze_instance_dependence): Likewise.
2067 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2069 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2071 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2072 (stmt_info_for_cost::stmt_info): ...this new field.
2073 (add_stmt_costs): Update accordingly.
2074 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2075 (vect_get_known_peeling_cost): Likewise.
2076 (vect_estimate_min_profitable_iters): Likewise.
2077 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2081 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2082 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2083 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2085 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2087 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2089 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2090 a gimple stmt to a stmt_vec_info.
2091 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2093 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2095 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2096 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2097 (_loop_vec_info::reduction_chains): Likewise.
2098 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2100 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2102 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2104 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2105 a gimple stmt to a stmt_vec_info.
2106 (_stmt_vec_info::next_element): Likewise.
2107 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2108 (vect_slp_analyze_and_verify_node_alignment)
2109 (vect_analyze_group_access_1, vect_analyze_group_access)
2110 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2111 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2112 (vect_supportable_dr_alignment): Update accordingly.
2113 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2114 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2115 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2116 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2117 * tree-vect-slp.c (vect_build_slp_tree_1)
2118 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2119 (vect_split_slp_store_group, vect_analyze_slp_instance)
2120 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2121 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2122 (get_group_load_store_type, get_load_store_type)
2123 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2124 (vect_transform_stmt, vect_remove_stores): Likewise.
2126 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2128 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2130 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2131 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2132 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2133 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2134 (vector_alignment_reachable_p, vect_get_data_access_cost)
2135 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2136 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2137 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2138 (vect_analyze_group_access_1, vect_analyze_group_access)
2139 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2140 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2141 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2142 result of vect_dr_stmt and use the stmt_vec_info instead of
2143 the associated gimple stmt.
2144 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2145 (vect_gen_prolog_loop_niters): Likewise.
2146 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2148 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2150 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2151 to a vec<stmt_vec_info>.
2152 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2153 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2155 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2156 to a vec<stmt_vec_info>.
2157 (bst_traits::value_type, bst_traits::value_type): Likewise.
2158 (bst_traits::hash): Update accordingly.
2159 (vect_get_and_check_slp_defs): Change the stmts parameter from
2160 a vec<gimple *> to a vec<stmt_vec_info>.
2161 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2162 (vect_build_slp_tree): Likewise.
2163 (vect_build_slp_tree_2): Likewise. Update uses of
2164 SLP_TREE_SCALAR_STMTS.
2165 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2166 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2167 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2168 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2169 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2170 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2171 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2172 (vect_get_constant_vectors, vect_get_slp_defs)
2173 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2174 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2175 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2176 instead of gimple stmts.
2177 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2178 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2179 (vect_slp_analyze_instance_dependence): Update uses of
2180 SLP_TREE_SCALAR_STMTS.
2181 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2182 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2183 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2184 (get_initial_defs_for_reduction): Likewise.
2185 (vect_create_epilog_for_reduction): Likewise.
2186 (vectorize_fold_left_reduction): Likewise.
2187 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2188 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2189 (can_vectorize_live_stmts): Likewise.
2191 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2193 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2194 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2195 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2197 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2199 (gather_scalar_reductions): Update after above interface changes.
2200 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2201 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2203 (vect_force_simple_reduction): Likewise.
2204 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2205 LOOP_VINFO_REDUCTIONS.
2206 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2208 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2210 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2211 a gimple stmt to a stmt_vec_info.
2212 * tree-vect-loop.c (vect_active_double_reduction_p)
2213 (vect_force_simple_reduction, vectorizable_reduction): Update
2216 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2218 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2219 vec<gimple *> to a vec<stmt_vec_info>.
2220 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2221 the reduction_phis argument from a vec<gimple *> to a
2223 (vectorizable_reduction): Likewise the phis local variable that
2224 is passed to vect_create_epilog_for_reduction. Update for new type
2225 of SLP_TREE_VEC_STMTS.
2226 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2227 (vectorizable_live_operation): Likewise.
2228 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2229 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2231 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2233 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2234 a gimple stmt to a stmt_vec_info.
2235 (vectorizable_condition, vectorizable_live_operation)
2236 (vectorizable_reduction, vectorizable_induction): Pass back the
2237 vectorized statement as a stmt_vec_info.
2238 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2239 use of STMT_VINFO_VEC_STMT.
2240 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2241 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2242 as stmt_vec_infos rather than gimple stmts.
2243 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2245 (vectorizable_live_operation): Likewise.
2246 (vectorizable_reduction, vectorizable_induction): Likewise,
2247 updating use of STMT_VINFO_VEC_STMT.
2248 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2249 of STMT_VINFO_VEC_STMT.
2250 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2251 (vectorizable_simd_clone_call, vectorizable_conversion)
2252 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2253 (vectorizable_store, vectorizable_load, vectorizable_condition)
2254 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2255 from a gimple stmt to a stmt_vec_info.
2256 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2257 pointer to a stmt_vec_info to the vectorizable_* routines.
2259 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2261 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2262 a gimple stmt to a stmt_vec_info.
2263 (is_pattern_stmt_p): Update accordingly.
2264 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2265 (vect_record_grouped_load_vectors): Likewise.
2266 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2267 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2268 (vect_model_reduction_cost): Likewise.
2269 (vect_create_epilog_for_reduction): Likewise.
2270 (vectorizable_reduction, vectorizable_induction): Likewise.
2271 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2272 Return the stmt_vec_info for the pattern statement.
2273 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2274 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2275 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2276 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2277 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2278 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2279 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2280 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2282 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2284 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2285 (vect_finish_stmt_generation): Likewise.
2286 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2287 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2288 (vect_build_gather_load_calls): Use the return value of the above
2289 functions instead of a separate call to vinfo_for_stmt. Use narrow
2290 scopes for the input gimple stmt and wider scopes for the associated
2291 stmt_vec_info. Use vec_info::lookup_def when setting these
2292 stmt_vec_infos from an SSA_NAME definition.
2293 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2294 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2295 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2296 (vectorizable_store, vectorizable_load, vectorizable_condition)
2297 (vectorizable_comparison): Likewise.
2298 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2299 (vectorizable_reduction): Likewise.
2301 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2303 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2304 stmt_vec_info * parameter before the optional gimple **.
2305 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2306 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2307 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2308 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2309 (vectorizable_live_operation): Likewise.
2310 * tree-vect-patterns.c (type_conversion_p): Likewise.
2311 (vect_look_through_possible_promotion): Likewise.
2312 (vect_recog_rotate_pattern): Likewise.
2313 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2315 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2317 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2318 a typedef to a wrapper class.
2319 (NULL_STMT_VEC_INFO): New macro.
2320 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2321 (stmt_vec_info::operator*): New function.
2322 (stmt_vec_info::operator gimple *): Likewise.
2323 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2324 (add_stmt_costs): Likewise.
2325 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2326 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2327 (vect_get_known_peeling_cost): Likewise.
2328 (vect_estimate_min_profitable_iters): Likewise.
2329 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2330 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2331 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2332 (vectorizable_store, free_stmt_vec_infos): Likewise.
2333 (new_stmt_vec_info): Change return type of xcalloc to
2336 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2338 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2339 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2340 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2341 a single_imm_use-based sequence.
2342 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2344 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2346 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2347 * tree-vectorizer.c (vec_info::lookup_def): New function.
2348 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2349 (vect_widened_op_tree): Likewise.
2350 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2351 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2352 (vectorizable_reduction): Likewise.
2353 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2355 (vect_is_slp_reduction): Update calls accordingly. Use
2356 vec_info::lookup_def.
2357 (vect_is_simple_reduction): Likewise
2358 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2360 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2362 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2363 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2364 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2366 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2367 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2368 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2369 (vect_is_slp_reduction, vectorizable_induction)
2370 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2371 * tree-vect-patterns.c (vect_init_pattern_stmt):
2372 (vect_determine_min_output_precision_1, vect_determine_precisions)
2373 (vect_pattern_recog): Likewise.
2374 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2375 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2376 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2377 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2378 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2379 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2380 info field from a loop to a loop_vec_info.
2382 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2384 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2385 (vec_info::add_stmt): Declare.
2386 * tree-vectorizer.c (vec_info::add_stmt): New function.
2387 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2388 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2389 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2390 (vectorizable_induction): Likewise.
2391 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2392 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2393 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2394 (vectorizable_load): Likewise.
2395 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2396 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2397 (vect_recog_gather_scatter_pattern): Likewise.
2398 (append_pattern_def_seq): Likewise. Remove a check that is
2399 performed by add_stmt itself.
2401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2403 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2404 which make_ssa_name was called with new_stmt before new_stmt
2407 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2409 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2411 (vect_is_slp_reduction): ...here...
2412 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2413 that are already known to be false.
2415 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2417 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2418 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2419 STMT_VINFO_NUM_SLP_USES when it's true.
2420 (vect_free_slp_instance): Add a final_p parameter and pass it to
2422 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2423 (vect_analyze_slp_instance): Likewise.
2424 (vect_slp_analyze_operations): Likewise.
2425 (vect_slp_analyze_bb_1): Likewise.
2426 * tree-vectorizer.c (vec_info): Likewise.
2427 * tree-vect-loop.c (vect_transform_loop): Likewise.
2429 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2431 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2432 function is not called for second and subsequent members of
2435 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2437 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2438 cases for nested loops from here to ...
2439 (vect_create_epilog_for_reduction): ...here. Only call
2440 vect_is_simple_use for inner-loop reductions.
2442 2018-07-31 Martin Liska <mliska@suse.cz>
2444 PR gcov-profile/85338
2445 PR gcov-profile/85350
2446 PR gcov-profile/85372
2447 * profile.c (struct location_triplet): New.
2448 (struct location_triplet_hash): Likewise.
2449 (output_location): Do not output a BB that
2450 is already recorded for a line.
2451 (branch_prob): Use streamed_locations.
2453 2018-07-31 Martin Liska <mliska@suse.cz>
2455 PR gcov-profile/85370
2456 * coverage.c (coverage_begin_function): Do not mark target
2457 clones as artificial functions.
2459 2018-07-31 Martin Liska <mliska@suse.cz>
2461 PR gcov-profile/83813
2462 PR gcov-profile/84758
2463 PR gcov-profile/85217
2464 PR gcov-profile/85332
2465 * profile.c (branch_prob): Do not record GOTO expressions
2466 for GIMPLE statements which locations are already streamed.
2468 2018-07-31 Olivier Hainque <hainque@adacore.com>
2470 * gcc.c (handle_spec_function): Accept a soft_matched_part
2471 argument, as do_spec_1. Pass it down to ...
2472 (eval_spec_function): Accept a soft_matched_part argument,
2473 and pass it down to ...
2474 (do_spec_2): Accept a soft_matched_part argument, and pass
2475 it down to do_spec_1.
2476 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2477 (handle_braces): Update call to handle_spec_function.
2478 (driver::set_up_specs): Update calls to do_spec_2.
2479 (compare_debug_dump_opt_spec_function): Likewise.
2480 (compare_debug_self_opt_spec_function): Likewise.
2482 2018-07-31 Olivier Hainque <hainque@adacore.com>
2484 * common.opt (nolibc): New option.
2485 * doc/invoke.texi (Link Options): Document it.
2486 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2487 * config/alpha/linux.h: Likewise.
2488 * config/arc/elf.h: Likewise.
2489 * config/arm/uclinux-elf.h: Likewise.
2490 * config/arm/unknown-elf.h: Likewise.
2491 * config/avr/avrlibc.h: Likewise.
2492 * config/bfin/bfin.h: Likewise.
2493 * config/bfin/linux.h: Likewise.
2494 * config/bfin/uclinux.h: Likewise.
2495 * config/darwin.h: Likewise.
2496 * config/darwin10.h: Likewise.
2497 * config/darwin12.h: Likewise.
2498 * config/gnu-user.h: Likewise.
2499 * config/lm32/uclinux-elf.h: Likewise.
2500 * config/pa/pa-hpux11.h: Likewise.
2501 * config/pa/pa64-hpux.h: Likewise.
2502 * config/sparc/sparc.h: Likewise.
2504 2018-07-31 Olivier Hainque <hainque@adacore.com>
2506 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2507 undefined variables.
2509 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2512 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2515 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2517 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2518 terminated string literal.
2520 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2522 PR rtl-optimization/85160
2523 * combine.c (is_just_move): New function.
2524 (try_combine): Allow combining two instructions into two if neither of
2525 the original instructions was a move.
2527 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2530 * doc/extend.texi (Global Register Variables): Discourage use of type
2532 (Local Register Variables): Likewise.
2534 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2536 PR tree-optimization/86506
2537 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2539 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2542 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2543 hard_regno, make sure no insn between `from` and `to` clobbers it.
2545 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2546 Tom de Vries <tdevries@suse.de>
2548 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2549 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2550 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2551 PTX_DEFAULT_RUNTIME_DIM.
2552 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2554 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2556 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2557 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2558 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2560 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2562 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2564 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2565 predicate to reflect toc fusion has been deleted.
2566 (toc_fusion_mem_raw): Delete, no longer used.
2567 (toc_fusion_mem_wrapped): Likewise.
2568 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2570 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2571 Delete, no longer used.
2572 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2573 meant to be used for toc fusion.
2574 (rs6000_debug_print_mode): Delete toc fusion debugging.
2575 (rs6000_debug_reg_global): Likewise.
2576 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2577 fusion and secondary reload support that were never used.
2578 (rs6000_option_override_internal): Delete TOC fusion, that was only
2579 partially defined, and it did not work unless you also used the
2581 (rs6000_legitimate_address_p): Delete TOC fusion support.
2582 (rs6000_opt_masks): Likewise.
2583 (fusion_wrap_memory_address): Delete function, no longer used.
2584 (fusion_split_address); Delete TOC fusion support.
2585 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2586 longer used with toc fusion being deleted.
2587 (TARGET_TOC_FUSION_FP): Likewise.
2588 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2590 (toc fusion spliter): Delete TOC fusion support.
2591 (toc_fusionload_<mode>): Likewise.
2592 (toc_fusionload_di): Likewise.
2593 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2594 longer needs to be named. Rename predicate to delete TOC fusion.
2595 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2596 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2597 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2598 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2599 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2601 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2603 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2604 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2605 __int128 in built-in function prototypes.
2606 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2607 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2609 2018-07-27 Martin Sebor <msebor@redhat.com>
2611 PR tree-optimization/86696
2612 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2613 types, including enums.
2614 (handle_char_store): Be prepared for the above function to fail.
2616 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2618 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2619 when optimization level is lower than 2 or optimize for size.
2621 2018-07-26 Martin Sebor <msebor@redhat.com>
2623 PR tree-optimization/86043
2624 PR tree-optimization/86042
2625 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2626 (get_string_cst_length): Rename...
2627 (get_min_string_length): ...to this. Add argument.
2628 (handle_char_store): Extend to handle multi-character stores by
2630 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2631 * tree.h (initializer_zerop): Add argument.
2633 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2636 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2637 declare target to variables if they have always,{to,from,tofrom} map
2640 2018-07-26 Martin Liska <mliska@suse.cz>
2643 * lto-wrapper.c: Add linker_output as prefix
2644 for ltrans_output_file.
2646 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2648 PR rtl-optimization/85805
2649 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2650 value for hard registers if that was written in the same mode.
2652 2018-07-26 Martin Liska <mliska@suse.cz>
2654 PR gcov-profile/86536
2655 * gcov.c (format_gcov): Use printf format %.*f directly
2656 and do not handle special values.
2658 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2660 * common/config/arc/arc-common.c (arc_option_optimization_table):
2661 Update default optimizations for size.
2663 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2665 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2666 * config/arc/constraints.md (CfZ): New constraint.
2667 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2668 (subsf3_fpu): Likewise.
2669 (cmpsf_fpu): Likewise.
2670 (cmpsf_fpu_uneq): Likewise.
2672 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2674 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2676 (arc_is_uncached_mem_p): uncached applies to both the variable and
2679 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2681 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2684 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2686 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2687 field "m_scopes" from vec to auto_vec.
2689 2018-07-25 Martin Liska <mliska@suse.cz>
2691 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2694 2018-07-25 Richard Biener <rguenther@suse.de>
2697 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2698 special wrt context_die late.
2699 (gen_subprogram_die): Re-use DIEs in local scope.
2701 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2703 PR tree-optimization/86644
2704 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2706 2018-07-25 Martin Liska <mliska@suse.cz>
2709 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2710 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2712 2018-07-25 Martin Liska <mliska@suse.cz>
2715 * params.def: Explain ASan abbreviation and provide
2716 a documentation link.
2718 2018-07-24 Martin Sebor <msebor@redhat.com>
2720 PR tree-optimization/86622
2721 PR tree-optimization/86532
2722 * builtins.h (string_length): Declare.
2723 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2724 (check_access): Be prepared for non-constant length ranges.
2725 (string_length): Make extern.
2726 * expr.c (string_constant): Only handle the minor non-constant
2727 array index. Use string_constant to compute the length of
2728 a generic string constant.
2730 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2732 PR tree-optimization/86618
2733 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2734 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2736 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2738 PR tree-optimization/86636
2739 * json.cc (json::object::set): Fix comment. Add assertions.
2740 (json::array::append): Move here from json.h. Add comment and an
2742 (json::string::string): Likewise.
2743 * json.h (json::array::append): Move to json.cc.
2744 (json::string::string): Likewise.
2745 * optinfo-emit-json.cc
2746 (optrecord_json_writer::impl_location_to_json): Assert that we
2747 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2748 wrapper around it. Expand the location once, rather than three
2750 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2751 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2753 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2754 places. Fix some overlong lines.
2756 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2758 * config/aarch64/aarch64-simd.md
2759 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2760 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2761 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2762 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2763 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2764 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2765 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2766 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2767 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2769 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2772 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2773 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2774 and abs_d == d, do the power of two handling if profitable.
2776 2018-07-24 Richard Biener <rguenther@suse.de>
2778 * match.pd: Add BIT_FIELD_REF canonicalizations.
2780 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2783 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2785 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2787 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2788 terminated STRING_CST object.
2790 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2792 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2794 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2796 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2797 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2798 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2799 the elements into a register.
2800 (rs6000_split_v4si_init_di_reg): Delete.
2801 (rs6000_split_v4si_init): Delete.
2802 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2803 (vsx_init_v4si): Rewrite as a define_expand.
2805 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2807 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2808 zero_extend argument from memory): New.
2810 2018-07-22 Martin Sebor <msebor@redhat.com>
2813 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2814 diagnosing calls with unknown arguments unless -Walloca-larger-than
2815 is restricted to less than PTRDIFF_MAX bytes.
2817 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2819 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2821 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2823 * pretty-print.c (text_info::set_location): Remove redundant
2824 "line_table" parameter from call to rich_location::set_range.
2826 2018-07-20 Martin Sebor <msebor@redhat.com>
2829 * builtins.c (expand_builtin_alloca): Adjust.
2830 * calls.c (alloc_max_size): Simplify.
2831 * cgraphunit.c (cgraph_node::expand): Adjust.
2832 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2834 (frame_larger_than_size): Same.
2835 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2836 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2838 * doc/invoke.texi (GCC Command Options): Document option arguments.
2839 Explain byte-size arguments and suffixes.
2840 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2841 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2842 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2843 * doc/options.texi (UInteger): Expand.
2844 (Host_Wide_Int, ByteSize): Document new properties.
2845 * final.c (final_start_function_1): Include sizes in an error message.
2846 * function.c (frame_offset_overflow): Same.
2847 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2848 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2849 Diagnose unbounded alloca calls only for limits of less than
2851 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2852 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2854 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2855 only for limits of less than PTRDIFF_MAX.
2856 * langhooks-def.h (lhd_handle_option): Change function argument
2858 * langhooks.c (lhd_handle_option): Same.
2859 * langhooks.h (handle_option): Same.
2860 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2862 (var_type, var_type_struct): Same.
2863 (var_set): Handle ByteSize flag.
2864 * optc-gen.awk: Add comments to output to ease debugging. Make
2865 use of HOST_WIDE_INT where appropriate.
2866 * opts-gen-save.awk: Use %lx to format unsigned long.
2867 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2868 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2869 arguments. Parse bytes-size suffixes.
2870 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2871 (enum_value_to_arg): Same.
2872 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2873 (handle_option): Adjust.
2874 (generate_option): Change function argument to HOST_WIDE_INT.
2875 (cmdline_handle_error): Adjust.
2876 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2877 (set_option): Change function argument to HOST_WIDE_INT.
2878 (option_enabled): Handle cl_host_wide_int.
2879 (get_option_state): Handle CLVC_SIZE.
2880 (control_warning_option): Same.
2881 * opts.c (common_handle_option): Change function argument to
2882 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2883 OPT_Wvla_larger_than_.
2884 * opts.h (enum cl_var_type): Add an enumerator.
2885 * stor-layout.c (layout_decl): Print a more meaningful warning.
2886 * toplev.c (output_stack_usage): Adjust.
2888 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
2890 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2891 call to inline_expand_builtin_string_cmp.
2892 (expand_builtin_strcmp): Likewise.
2893 (expand_builtin_strncmp): Likewise.
2894 (inline_string_cmp): Delete the last parameter, change char_type_node
2895 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2897 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2898 the inlining expansion on target where the type of the call has same or
2899 narrower precision than unsigned char.
2901 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2903 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2904 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2905 * common.opt (fsave-optimization-record): New option.
2906 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2907 * doc/invoke.texi (-fsave-optimization-record): New option.
2908 * dumpfile.c: Include "optinfo-emit-json.h".
2909 (struct kv_pair): Move to coretypes.h.
2910 (optgroup_options): Make non-static.
2911 (dump_context::end_scope): Call
2912 optimization_records_maybe_pop_dump_scope.
2913 * dumpfile.h (optgroup_options): New decl.
2914 * json.cc: New file.
2916 * optinfo-emit-json.cc: New file.
2917 * optinfo-emit-json.h: New file.
2918 * optinfo.cc: Include "optinfo-emit-json.h".
2919 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2920 (optinfo_enabled_p): Check optimization_records_enabled_p.
2921 (optinfo_wants_inlining_info_p): Likewise.
2922 * optinfo.h: Update comment.
2923 * profile-count.c (profile_quality_as_string): New function.
2924 * profile-count.h (profile_quality_as_string): New decl.
2925 (profile_count::quality): New accessor.
2926 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2927 and optinfo_emit_json_cc_tests.
2928 * selftest.h (selftest::json_cc_tests): New decl.
2929 (selftest::optinfo_emit_json_cc_tests): New decl.
2930 * toplev.c: Include "optinfo-emit-json.h".
2931 (compile_file): Call optimization_records_finish.
2932 (do_compile): Call optimization_records_start.
2933 * tree-ssa-live.c: Include optinfo.h.
2934 (remove_unused_scope_block_p): Retain inlining information if
2935 optinfo_wants_inlining_info_p returns true.
2937 2018-07-20 Richard Biener <rguenther@suse.de>
2940 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2941 to cover -flto-partition=none.
2943 2018-07-20 Martin Liska <mliska@suse.cz>
2945 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2946 (get_decl_source_range): Remove unused function.
2948 2018-07-20 Richard Biener <rguenther@suse.de>
2950 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2951 (struct vn_phi_s): Likewise.
2952 (struct vn_reference_s): Likewise.
2953 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2954 for searching the slot of an entry known to be in the hash itself.
2955 (vn_phi_hasher::equal): Likewise.
2956 (vn_reference_hasher::equal): Likewise.
2957 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2959 (optimistic_info, current_info): Remove, keeping only valid_info.
2960 (vn_reference_lookup_1): Remove fallback lookup.
2961 (vn_reference_lookup_2): Likewise.
2962 (vn_nary_op_lookup_1): Likewise.
2963 (vn_phi_lookup): Likewise.
2964 (vn_nary_build_or_lookup_1): Make sure to not chain the built
2966 (vn_reference_insert): Adjust, chain the inserted hash element
2967 at last_inserted_ref.
2968 (vn_reference_insert_pieces): Likewise.
2969 (visit_reference_op_call): Likewise.
2970 (vn_nary_op_insert_into): Chain the inserted hash element at
2972 (vn_nary_op_insert_pieces): Adjust.
2973 (vn_nary_op_insert): Likewise.
2974 (vn_nary_op_insert_stmt): Likewise.
2975 (vn_phi_insert): Adjust, chain the inserted hash element at
2977 (process_scc): Remove clearing and copying the optimistic
2978 table. Instead remove elements inserted during an optimistic
2979 iteration from the single table we maintain.
2980 (init_scc_vn): Adjust.
2981 (free_scc_vn): Likewise.
2982 (sccvn_dom_walker::record_cond): Likewise.
2983 (sccvn_dom_walker::after_dom_children): Likewise.
2985 2018-07-19 Martin Sebor <msebor@redhat.com>
2987 PR tree-optimization/84047
2988 PR tree-optimization/83776
2989 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2990 (check_array_bounds): Call it.
2992 2018-07-19 Martin Sebor <msebor@redhat.com>
2994 * align.h (align_flags): Use member initialization.
2996 2018-07-19 David Malcolm <dmalcolm@redhat.com>
2998 * Makefile.in (OBJS): Add optinfo.o.
2999 * coretypes.h (class symtab_node): New forward decl.
3000 (struct cgraph_node): New forward decl.
3001 (class varpool_node): New forward decl.
3002 * dump-context.h: New file.
3003 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3005 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3006 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3007 (set_alt_dump_file): Likewise.
3008 (dump_context::~dump_context): New dtor.
3009 (dump_gimple_stmt): Move implementation to...
3010 (dump_context::dump_gimple_stmt): ...this new member function.
3011 Add the stmt to any pending optinfo, creating one if need be.
3012 (dump_gimple_stmt_loc): Move implementation to...
3013 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3014 Start a new optinfo and add the stmt to it.
3015 (dump_gimple_expr): Move implementation to...
3016 (dump_context::dump_gimple_expr): ...this new member function.
3017 Add the stmt to any pending optinfo, creating one if need be.
3018 (dump_gimple_expr_loc): Move implementation to...
3019 (dump_context::dump_gimple_expr_loc): ...this new member function.
3020 Start a new optinfo and add the stmt to it.
3021 (dump_generic_expr): Move implementation to...
3022 (dump_context::dump_generic_expr): ...this new member function.
3023 Add the tree to any pending optinfo, creating one if need be.
3024 (dump_generic_expr_loc): Move implementation to...
3025 (dump_context::dump_generic_expr_loc): ...this new member
3026 function. Add the tree to any pending optinfo, creating one if
3028 (dump_printf): Move implementation to...
3029 (dump_context::dump_printf_va): ...this new member function. Add
3030 the text to any pending optinfo, creating one if need be.
3031 (dump_printf_loc): Move implementation to...
3032 (dump_context::dump_printf_loc_va): ...this new member function.
3033 Start a new optinfo and add the stmt to it.
3034 (dump_dec): Move implementation to...
3035 (dump_context::dump_dec): ...this new member function. Add the
3036 value to any pending optinfo, creating one if need be.
3037 (dump_context::dump_symtab_node): New member function.
3038 (dump_context::get_scope_depth): New member function.
3039 (dump_context::begin_scope): New member function.
3040 (dump_context::end_scope): New member function.
3041 (dump_context::ensure_pending_optinfo): New member function.
3042 (dump_context::begin_next_optinfo): New member function.
3043 (dump_context::end_any_optinfo): New member function.
3044 (dump_context::s_current): New global.
3045 (dump_context::s_default): New global.
3046 (dump_scope_depth): Delete global.
3047 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3048 (dump_symtab_node): New function.
3049 (get_dump_scope_depth): Reimplement in terms of dump_context.
3050 (dump_begin_scope): Likewise.
3051 (dump_end_scope): Likewise.
3052 (selftest::temp_dump_context::temp_dump_context): New ctor.
3053 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3054 (selftest::verify_item): New function.
3055 (ASSERT_IS_TEXT): New macro.
3056 (ASSERT_IS_TREE): New macro.
3057 (ASSERT_IS_GIMPLE): New macro.
3058 (selftest::test_capture_of_dump_calls): New test.
3059 (selftest::dumpfile_c_tests): Call it.
3060 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3061 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3062 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3063 descriptive comment.
3064 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3065 (dump_node, dump_bb): Move these unrelated decls.
3066 (class dump_manager): Add leading comment.
3067 * optinfo.cc: New file.
3068 * optinfo.h: New file.
3070 2018-07-19 Michael Collison <michael.collison@arm.com>
3071 Richard Henderson <rth@redhat.com>
3073 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3074 (subti): Handle op1 zero.
3075 (subvti4, usub4ti4): New.
3076 (*sub<GPI>3_compare1_imm): New.
3077 (sub<GPI>3_carryinCV): New.
3078 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3079 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3081 2018-07-19 Michael Collison <michael.collison@arm.com>
3082 Richard Henderson <rth@redhat.com>
3084 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3085 (addti3): Create simpler code if low part is already known to be 0.
3086 (addvti4, uaddvti4): New.
3087 (*add<GPI>3_compareC_cconly_imm): New.
3088 (*add<GPI>3_compareC_cconly): New.
3089 (*add<GPI>3_compareC_imm): New.
3090 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3091 handle constants within this pattern..
3092 (*add<GPI>3_compareV_cconly_imm): New.
3093 (*add<GPI>3_compareV_cconly): New.
3094 (*add<GPI>3_compareV_imm): New.
3095 (add<GPI>3_compareV): New.
3096 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3097 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3098 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3099 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3100 with 'comparison' operator.
3101 (*add<GPI>3_compareV_cconly_imm): Ditto.
3102 (*add<GPI>3_compareV_cconly): Ditto.
3103 (*add<GPI>3_compareV_imm): Ditto.
3104 (add<GPI>3_compareV): Ditto.
3105 (add<mode>3_carryinC): Ditto.
3106 (*add<mode>3_carryinC_zero): Ditto.
3107 (*add<mode>3_carryinC): Ditto.
3108 (add<mode>3_carryinV): Ditto.
3109 (*add<mode>3_carryinV_zero): Ditto.
3110 (*add<mode>3_carryinV): Ditto.
3112 2018-07-19 Michael Collison <michael.collison@arm.com>
3113 Richard Henderson <rth@redhat.com>
3115 * config/aarch64/aarch64-modes.def (CC_V): New.
3116 * config/aarch64/aarch64-protos.h
3117 (aarch64_addti_scratch_regs): Declare
3118 (aarch64_subvti_scratch_regs): Declare.
3119 (aarch64_expand_subvti): Declare.
3120 (aarch64_gen_unlikely_cbranch): Declare
3121 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3122 for signed overflow using CC_Vmode.
3123 (aarch64_get_condition_code_1): Handle CC_Vmode.
3124 (aarch64_gen_unlikely_cbranch): New function.
3125 (aarch64_addti_scratch_regs): New function.
3126 (aarch64_subvti_scratch_regs): New function.
3127 (aarch64_expand_subvti): New function.
3129 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3131 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3133 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3134 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3137 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3140 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3141 address check not strict.
3143 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3145 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3147 (store_pair_lanes<mode>): Likewise.
3148 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3149 enum value 'ADDR_QUERY_LDP_STP_N'.
3150 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3151 (aarch64_print_address_internal): Add declaration.
3152 (aarch64_print_ldpstp_address): Remove.
3153 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3154 (aarch64_print_operand): Change printing of 'y'.
3155 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3156 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3157 'true' rather than '1'.
3158 * config/aarch64/constraints.md (Uml): Likewise.
3159 (Uml): Rename to Umn.
3162 2018-07-19 Richard Biener <rguenther@suse.de>
3164 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3166 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3167 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3168 (vn_reference_hasher): Likewise.
3169 (struct vn_tables_s): Remove obstack and alloc-pool members.
3170 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3171 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3172 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3173 (vn_reference_insert_pieces): Likewise.
3174 (alloc_vn_nary_op_noinit): Adjust.
3175 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3176 (vn_phi_eq): Adjust.
3177 (shared_lookup_phiargs): Remove.
3178 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3179 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3180 (visit_reference_op_call): Likewise.
3181 (copy_nary, copy_phi, copy_reference): Remove.
3182 (process_scc): Rewind the obstack when iterating. Do not
3183 copy the elements to valid_info but just move them from one
3184 hashtable to the other.
3185 (allocate_vn_table): Adjust.
3186 (free_vn_table): Likewise.
3187 (init_scc_vn): Likewise.
3188 (free_scc_vn): Likewise.
3190 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3193 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3194 indirect_return as function type attribute.
3195 (ix86_attribute_table): Change indirect_return to function
3197 * doc/extend.texi: Update indirect_return attribute.
3199 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3201 * wide-int.h (widest2_int): New.
3202 * gimple-fold.c (arith_overflowed_p): Use it.
3203 * tree.h (widest2_int_cst): New.
3204 * tree-vrp.c (wide_int_binop_overflow): Rename from
3205 vrp_int_const_binop.
3206 Rewrite to work on trees.
3207 (extract_range_from_multiplicative_op_1): Abstract code to...
3208 (wide_int_range_min_max): ...here.
3209 (wide_int_range_cross_product): ...and here.
3210 (extract_range_from_binary_expr_1): Abstract overflow code to...
3211 (wide_int_range_mult_wrapping): ...here.
3212 * tree-vrp.h (wide_int_range_cross_product): New.
3213 (wide_int_range_mult_wrapping): New.
3215 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3216 Julia Koval <julia.koval@intel.com>
3218 * config/i386/x86-tune-costs.h (skylake_memcpy,
3219 skylake_memset): Replace rep_prefix with unrolling for size 512.
3221 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3224 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3225 comparision with EQ_EXPR in last stmt.
3227 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3229 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3230 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3231 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3232 previously known as "PowerPC AltiVec Built-in Functions". Move
3233 some material to new subsubsections "PowerPC AltiVec Built-in
3234 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3236 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3237 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3238 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3239 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3241 2018-07-18 Richard Biener <rguenther@suse.de>
3243 PR tree-optimization/86557
3244 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3247 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3249 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3251 2018-07-17 Jeff Law <law@redhat.com>
3253 * config/arm/arm.c (get_label_padding): Update for recent
3254 changes to label_to_alignment.
3256 PR tree-optimization/86010
3257 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3259 * config/mips/mips.c (vr4130_align_insns): Update for recent
3260 changes to label_to_alignment.
3262 * config/frv/frv.c (frv_label_align): Update for recent changes
3263 to label_to_alignment.
3265 * config/nios2/nios2.c (nios2_label_align): Update for recent
3266 changes which dropped ALIGN_LABELS_LOG.
3268 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3270 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3271 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3273 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3275 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3276 deprecated align_labels_log.
3278 2018-07-17 Richard Biener <rguenther@suse.de>
3281 * dwarf2out.c (init_sections_and_labels): Always generate
3282 a debug_line_str_section for early LTO debug.
3283 (dwarf2out_finish): Reset debug_line_str_hash output early.
3284 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3286 (dwarf2out_early_finish): Output debug_line_str.
3288 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3290 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3291 index register on z196 or later.
3293 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3295 * config/s390/s390.c (s390_default_align): Set default function
3297 (s390_override_options_after_change): Call s390_default align.
3298 (s390_option_override_internal): Call s390_default_align.
3299 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3301 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3304 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3305 to _looptemp_ clauses, other than the first two.
3307 2018-07-17 Martin Liska <mliska@suse.cz>
3309 * opts.c: Do not enable OPT_falign_* for -Os.
3311 2018-07-17 Martin Liska <mliska@suse.cz>
3313 * align.h (MAX_CODE_ALIGN): New.
3314 (MAX_CODE_ALIGN_VALUE): New.
3315 * common/config/i386/i386-common.c (ix86_handle_option):
3316 (MAX_CODE_ALIGN): Moved to align.h.
3317 * final.c (MAX_CODE_ALIGN): Likewise.
3318 * opts.c (parse_and_check_align_values):
3319 (MAX_CODE_ALIGN): Likewise.
3320 (MAX_CODE_ALIGN_VALUE): Likewise.
3322 2018-07-17 Martin Liska <mliska@suse.cz>
3324 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3325 in order to fulfil coding style.
3326 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3327 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3328 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3329 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3330 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3331 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3332 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3333 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3334 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3336 2018-07-17 Martin Liska <mliska@suse.cz>
3338 * align.h: New file.
3339 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3341 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3342 align_flags of label_to_alignment.
3343 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3345 * config/m68k/m68k.c: Do not use removed align_labels_value and
3347 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3348 (LOOP_ALIGN): Likewise.
3349 (LABEL_ALIGN): Likewise.
3350 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3351 Remove not used macro.
3352 (rs6000_loop_align): Change return type to align_flags.
3353 (rs6000_loop_align_max_skip): Remove.
3354 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3355 Change return type to align_flags.
3356 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3357 Remove not used macro.
3358 (rs6000_loop_align): Change return type to align_flags.
3359 (rs6000_loop_align_max_skip): Remove.
3360 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3361 * config/rx/rx-protos.h (rx_align_for_label): Make it
3363 * config/rx/rx.c (rx_align_for_label): Change return type
3365 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3367 into align_flags class.
3368 (LABEL_ALIGN): Likewise.
3369 (LOOP_ALIGN): Likewise.
3370 * config/s390/s390.c (s390_label_align): Use align_flags
3372 (s390_asm_output_function_label): Likewise.
3373 * config/sh/sh.c (sh_override_options_after_change):
3374 Use align_flags class directly without macros.
3375 (find_barrier): Likewise.
3376 (barrier_align): Likewise.
3377 (sh_loop_align): Likewise.
3378 * config/spu/spu.c (spu_option_override):
3379 Use align_flags_tuple::get_value instead of removed macros.
3380 (spu_sched_init): Likewise.
3381 * config/spu/spu.h (GTY): Likewise.
3382 * config/visium/visium.c (visium_option_override):
3383 Set "8" as default secondary alignment.
3384 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3385 in order to guarantee secondary alignment of 8.
3386 * coretypes.h: Include align.h header file.
3387 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3388 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3389 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3390 * doc/tm.texi.in: Likewise.
3391 * final.c (struct label_alignment): Remove not used structure.
3392 (LABEL_ALIGN): Change type to align_flags.
3393 (LOOP_ALIGN): Likewise.
3394 (JUMP_ALIGN): Likewise.
3395 (default_loop_align_max_skip): Remove.
3396 (default_label_align_max_skip): Likewise.
3397 (default_jump_align_max_skip): Likewise.
3398 (default_label_align_after_barrier_max_skip):
3399 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3400 (LABEL_TO_MAX_SKIP): Remove.
3401 (label_to_alignment): Return align_flags type instead of integer.
3402 (label_to_max_skip): Remove.
3403 (align_fuzz): Use align_flags type.
3404 (compute_alignments): Use align_flags type and use align_flags::max
3405 to combine multiple alignments.
3406 (grow_label_align): Grow vec instead of C array.
3407 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3408 (shorten_branches): Use align_flags type and use align_flags::max
3409 to combine multiple alignments.
3410 (final_scan_insn_1): Remove usage of secondary alignment that comes
3411 from label alignment, but instead use proper secondary alignment
3412 which is computed in grow_label_align.
3413 * flags.h (struct align_flags_tuple): Move to align.h.
3414 (struct align_flags): Likewise.
3415 (state_align_loops): Rename to align_loops.
3416 (state_align_jumps): Rename to align_jumps.
3417 (state_align_labels): Rename to align_labels.
3418 (state_align_functions): Rename to align_functions.
3419 (align_loops_log): Remove.
3420 (align_jumps_log): Remove.
3421 (align_labels_log): Remove.
3422 (align_functions_log): Remove.
3423 (align_loops_max_skip): Remove.
3424 (align_jumps_max_skip): Remove.
3425 (align_labels_max_skip): Remove.
3426 (align_functions_max_skip): Remove.
3427 (align_loops_value): Remove.
3428 (align_jumps_value): Remove.
3429 (align_labels_value): Remove.
3430 (align_functions_value): Remove.
3431 * output.h (label_to_alignment): Change return type to align_flags.
3432 (label_to_max_skip): Remove.
3433 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3434 jump_align_max_skip macros.
3435 * targhooks.h (default_loop_align_max_skip): Remove.
3436 (default_label_align_max_skip): Likewise.
3437 (default_jump_align_max_skip): Likewise.
3438 (default_label_align_after_barrier_max_skip): Remove.
3439 * toplev.c (read_log_maxskip): Use ::normalize function.
3440 (parse_N_M): Remove not used argument and also call ::normalize.
3441 (parse_alignment_opts): Do not pass unused arguments.
3442 * varasm.c (assemble_start_function): Use directly align_functions
3443 instead of removed macros.
3444 * system.h: Do not poison removed macros.
3446 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3449 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3450 and cond temporaries don't have reference type if iterator has
3451 pointer type. For init use &for_pre_body instead of pre_p if
3452 for_pre_body is non-empty.
3454 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3456 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3457 double-double modes to SFmode directly directly.
3458 (trunc<mode>sf2_fprs): Delete.
3460 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3462 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3463 for conversions between IFmode and the decimal floating point modes.
3464 (init_float128_ieee): Use the correct names for conversions between
3465 KFmode and the decimal floating point modes.
3467 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3469 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3470 for the conversions between TDmode and IFmode.
3471 (init_float128_ieee): Use more correct names for the conversions
3472 between TDmode and KFmode.
3474 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3476 PR tree-optimization/86526
3477 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3478 (inline_expand_builtin_string_cmp): Likewise.
3479 (inline_string_cmp): Likewise. Use c_readstr instead of
3480 builtin_memcpy_read_str. Add unit_mode temporary.
3482 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3485 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3486 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3488 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3490 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3491 Alphabetize prototypes of built-in functions, separating out
3492 built-in functions that are listed in this section but should be
3493 described elsewhere.
3495 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3498 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3499 conditional move when trapping comparison was split to a
3500 non-trapping one (and vice versa).
3502 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3504 * config/s390/s390.c (s390_function_profiler): Generate nops
3505 instead of profiler call sequences.
3506 * config/s390/s390.opt: Add the new option.
3508 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3510 * config/s390/s390.c (s390_function_profiler): Generate
3511 __mcount_loc section.
3512 * config/s390/s390.opt: Add the new option.
3514 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3516 * common.opt: Add the new warning.
3517 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3518 %r0,__fentry__" when -mfentry is specified.
3519 (s390_option_override_internal): Disallow -mfentry for 31-bit
3521 * config/s390/s390.opt: Add the new option.
3523 2018-07-16 Richard Biener <rguenther@suse.de>
3526 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3527 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3529 2018-07-16 Martin Liska <mliska@suse.cz>
3532 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3535 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3537 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3539 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3541 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3543 * config/arc/arc-cpus.def (hs4x): New cpu.
3545 * config/arc/arc-tables.opt: Regenerate.
3546 * config/arc/arc.c (arc_sched_issue_rate): New function.
3547 (TARGET_SCHED_ISSUE_RATE): Define.
3548 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3549 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3551 (attr tune): Add ARCHS4x tune values.
3552 (attr tune_dspmpy): Define.
3553 (*tst): Correct instruction type.
3554 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3555 * config/arc/arcHS4x.md: New file.
3556 * config/arc/fpu.md: Update instruction type attributes.
3557 * config/arc/t-multilib: Regenerate.
3559 2018-07-16 Tom de Vries <tdevries@suse.de>
3562 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3564 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3566 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3569 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3571 PR tree-optimization/86514
3572 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3573 conversion to a boolean type from a type with greater precision.
3575 2018-07-16 Tom de Vries <tdevries@suse.de>
3577 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3578 dump_flags request TDF_SLIM.
3580 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3582 * fold-const.c (int_const_binop_1): Abstract...
3583 (wide_int_binop): ...wide int code here.
3584 (poly_int_binop): ...poly int code here.
3585 Abstract the rest of int_const_binop_1 into int_const_binop.
3586 * fold-const.h (wide_int_binop): New.
3587 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3588 Remove useless PLUS/MINUS_EXPR case.
3589 (zero_nonzero_bits_from_vr): Move wide int code...
3590 (zero_nonzero_bits_from_bounds): ...here.
3591 (extract_range_from_binary_expr_1): Move mask optimization code...
3592 (range_easy_mask_min_max): ...here.
3593 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3594 (range_easy_mask_min_max): New.
3596 2018-07-15 Jeff Law <law@redhat.com>
3599 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3602 2018-07-14 Jim Wilson <jimw@sifive.com>
3604 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3606 2018-07-14 Paul Koning <ni1d@arrl.net>
3608 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3610 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3612 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3615 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3616 Steve Munroe <munroesj52@gmail.com>
3618 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3619 (_mm_andnot_si128): Likewise.
3620 (_mm_or_si128): Likewise.
3621 (_mm_xor_si128): Likewise.
3623 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3626 * builtins.c (expand_builtin_memcmp): Inline the calls first
3627 when result_eq is false.
3628 (expand_builtin_strcmp): Inline the calls first.
3629 (expand_builtin_strncmp): Likewise.
3630 (inline_string_cmp): New routine. Expand a string compare
3631 call by using a sequence of char comparison.
3632 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3633 a call to str(n)cmp/memcmp.
3634 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3636 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3638 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3640 * config/arm/driver-arm.c: Include arm-native.h.
3641 (host_detect_local_cpu): Use auto-generated data tables.
3642 (vendors, arm_cpu_table): Delete. Move part information to ...
3643 * config/arm/arm-cpus.in: ... here.
3644 * config/arm/parsecpu.awk (gen_native): New function.
3645 (vendor, part): New CPU fields.
3646 (END): Add support for building the native CPU detection tables.
3647 * config/arm/t-arm (arm-native.h): Add build rule.
3648 (driver-arm.o): Add dependency on arm-native.h.
3650 2018-07-13 Richard Biener <rguenther@suse.de>
3653 * match.pd (addr1 - addr2): Allow either of the operand to
3656 2018-07-13 Tom de Vries <tdevries@suse.de>
3658 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3661 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3663 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3664 arrays instead of numbered variables.
3666 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3668 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3669 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3671 2018-07-13 Richard Biener <rguenther@suse.de>
3674 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3675 instead of get_context_die.
3677 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3678 Richard Biener <rguenther@suse.de>
3681 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3682 that the loop latch destination where phi is defined.
3684 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3686 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3687 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3688 (riscv_set_current_function): Extract getting interrupt type to new
3690 (riscv_get_interrupt_type): New function.
3691 (riscv_merge_decl_attributes): New function, checking interrupt type is
3693 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3695 2018-07-12 Paul Koning <ni1d@arrl.net>
3697 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3700 2018-07-12 Paul Koning <ni1d@arrl.net>
3702 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3703 zero reference, add doloop_end instead.
3704 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3705 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3706 detail for doloop_end.
3708 2018-07-12 Martin Sebor <msebor@redhat.com>
3711 * attribs.c (decl_attributes): Reject conflicting attributes before
3712 calling attribute handlers.
3714 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3716 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3718 (gcc::dump_manager::get_dump_file_name): likewise.
3719 (dump_begin): Likewise.
3720 * dumpfile.h (dump_begin): Update prototype.
3721 (gcc::dump_manager::get_dump_file_name,
3722 gcc::dump_manager::get_dump_file_name): Update prototype.
3724 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3726 * internal-fn.h (vectorizable_internal_fn_p): New function.
3727 * tree-vect-slp.c (compatible_calls_p): Likewise.
3728 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3729 to internal functions.
3730 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3732 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3734 * fold-const.h (inverse_conditions_p): Declare.
3735 * fold-const.c (inverse_conditions_p): New function.
3736 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3737 that test the inverse condition of a conditional internal function.
3738 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3739 * internal-fn.c (internal_fn_mask_index): Handle conditional
3741 (vectorized_internal_fn_supported_p): New function.
3742 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3743 (any_pred_load_store): Replace with...
3744 (need_to_predicate): ...this new variable.
3745 (redundant_ssa_names): New variable.
3746 (ifcvt_can_use_mask_load_store): Move initial checks to...
3747 (ifcvt_can_predicate): ...this new function. Handle tree codes
3748 for which a conditional internal function exists.
3749 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3750 instead of ifcvt_can_use_mask_load_store. Update after variable
3752 (predicate_load_or_store): New function, split out from
3753 predicate_mem_writes.
3754 (check_redundant_cond_expr): New function.
3755 (value_available_p): Likewise.
3756 (predicate_rhs_code): Likewise.
3757 (predicate_mem_writes): Rename to...
3758 (predicate_statements): ...this. Use predicate_load_or_store
3759 and predicate_rhs_code.
3760 (combine_blocks, tree_if_conversion): Update after above name changes.
3761 (ifcvt_local_dce): Handle redundant_ssa_names.
3762 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3763 general conditional functions.
3764 * tree-vect-stmts.c (vectorizable_call): Likewise.
3766 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3767 Alan Hayward <alan.hayward@arm.com>
3768 David Sherwood <david.sherwood@arm.com>
3770 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3771 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3772 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3773 plus and minus and convert them into IFN_COND_FMA-based sequences.
3774 (convert_mult_to_fma): Handle conditional plus and minus.
3776 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3778 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3779 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3780 (cond_fnms_optab): New optabs.
3781 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3783 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3784 * internal-fn.h (get_conditional_internal_fn): Declare.
3785 (get_unconditional_internal_fn): Likewise.
3786 * internal-fn.c (cond_ternary_direct): New macro.
3787 (expand_cond_ternary_optab_fn): Likewise.
3788 (direct_cond_ternary_optab_supported_p): Likewise.
3789 (FOR_EACH_COND_FN_PAIR): Likewise.
3790 (get_conditional_internal_fn): New function.
3791 (get_unconditional_internal_fn): Likewise.
3792 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3793 (gimple_match_op::gimple_match_op): Add a new overload for 5
3795 (gimple_match_op::set_op): Likewise.
3796 (gimple_resimplify5): Declare.
3797 * genmatch.c (decision_tree::gen): Generate simplifications for
3799 * gimple-match-head.c (gimple_simplify): Define an overload for
3800 5 operands. Handle calls with 5 arguments in the top-level overload.
3801 (convert_conditional_op): Handle conversions from unconditional
3802 internal functions to conditional ones.
3803 (gimple_resimplify5): New function.
3804 (build_call_internal): Pass a fifth operand.
3805 (maybe_push_res_to_seq): Likewise.
3806 (try_conditional_simplification): Try converting conditional
3807 internal functions to unconditional internal functions.
3808 Handle 3-operand unconditional forms.
3809 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3810 Define ternary equivalents of the current rules for binary conditional
3812 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3814 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3815 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3816 (optab): Handle them.
3817 (SVE_COND_FP_TERNARY): New int iterator.
3818 (sve_fmla_op, sve_fmad_op): New int attributes.
3819 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3820 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3821 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3823 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3825 * target.def (preferred_else_value): New target hook.
3826 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3827 * doc/tm.texi: Regenerate.
3828 * targhooks.h (default_preferred_else_value): Declare.
3829 * targhooks.c (default_preferred_else_value): New function.
3830 * internal-fn.h (conditional_internal_fn_code): Declare.
3831 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3832 (get_conditional_internal_fn): Use it.
3833 (conditional_internal_fn_code): New function.
3834 * gimple-match.h (gimple_match_cond): New struct.
3835 (gimple_match_op): Add a cond member function.
3836 (gimple_match_op::gimple_match_op): Update all forms to take a
3838 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3839 for the suboperation, but don't specify a particular else_value.
3840 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3841 (visit_nary_op, visit_reference_op_load): Pass
3842 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3843 * gimple-match-head.c: Include tree-eh.h
3844 (convert_conditional_op): New function.
3845 (maybe_resimplify_conditional_op): Likewise.
3846 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3847 (gimple_resimplify2): Likewise.
3848 (gimple_resimplify3): Likewise.
3849 (gimple_resimplify4): Likewise.
3850 (maybe_push_res_to_seq): Return null for conditional operations.
3851 (try_conditional_simplification): New function.
3852 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3854 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3856 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3858 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3860 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3862 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3864 (hash_tree): Do not hash DECL_FCONTEXT
3865 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3866 Do not stream DECL_FCONTEXT.
3867 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3868 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3870 2018-07-12 Richard Biener <rguenther@suse.de>
3873 * dwarf2out.c (gen_block_die): Only output blocks when they have
3874 at least one !DECL_IGNORED_P variable.
3876 2018-07-12 Richard Biener <rguenther@suse.de>
3879 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3880 Remove -mieee-fp handling.
3882 2018-07-12 Richard Biener <rguenther@suse.de>
3884 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3885 left-over from last patch.
3887 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3889 PR tree-optimization/86492
3890 * gimple-ssa-store-merging.c
3891 (imm_store_chain_info::coalesce_immediate_stores): Call
3892 check_no_overlap even for the merge_overlapping case. Formatting fix.
3894 2018-07-12 Richard Biener <rguenther@suse.de>
3897 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3898 move possibly trapping operations into the conditional.
3900 2018-07-12 Richard Biener <rguenther@suse.de>
3902 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3903 (vn_lookup_simplify_result): Remove recursion limit applied
3905 (vn_nary_build_or_lookup_1): Adjust.
3906 (try_to_simplify): Likewise.
3907 * gimple-match-head.c (gimple_resimplify1): Instead apply one
3909 (gimple_resimplify2): Likewise.
3910 (gimple_resimplify3): Likewise.
3911 (gimple_resimplify4): Likewise.
3913 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3915 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3916 Use __mmask64 type instead of __mmask8 for __M argument.
3917 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3918 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3920 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3921 __mmask16 for __M argument.
3922 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3923 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3924 to __mmask16 instead of __mmask8.
3925 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3926 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3927 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3928 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3929 instead of __mmask16 for __U argument.
3930 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3931 __mmask16 instead of __mmask8 for __U argument.
3932 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3934 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3936 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3938 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3940 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3942 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3944 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3946 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3948 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3950 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3952 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3953 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3954 return type as well as __M argument type and all casts from __mmask8
3956 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3957 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3958 return type as well as __M argument type and all casts from __mmask8
3960 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3961 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3962 return type as well as __M argument type and all casts from __mmask8
3964 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3965 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3966 return type as well as __M argument type and all casts from __mmask8
3968 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3969 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3972 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3974 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3975 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3978 2018-07-11 Paul Koning <ni1d@arrl.net>
3980 * doc/md.texi (define_subst): Document how multiple occurrences of
3981 the same argument in the replacement pattern are handled.
3983 2018-07-11 Paul Koning <ni1d@arrl.net>
3985 * doc/extend.texi (Common Variable Attributes): Move "mode" into
3987 (Common Type Attributes): Add "mode" attribute.
3989 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
3991 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3992 stream DECL_ORIGINAL_TYPE.
3993 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3994 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3995 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3996 Do not walk original type.
3997 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3999 (write_ts_decl_non_common_tree_pointers): Do not stream
4001 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4002 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4004 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4006 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4007 thread twice from the same starting edge.
4009 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4011 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4012 * gimple.c (gimple_call_nonnull_result_p): ...here...
4013 (gimple_call_nonnull_arg): ...and here.
4014 * gimple.h (gimple_call_nonnull_result_p): New.
4015 (gimple_call_nonnull_arg): New.
4017 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4019 * config/arm/arm-cpus.in: Move information from fpu field of each
4020 cpu definition to the isa field.
4021 * config/arm/parsecpu.awk (fpu): Delete match rule.
4022 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4024 2018-07-11 Richard Biener <rguenther@suse.de>
4027 * dwarf2out.c (init_sections_and_labels): Use
4028 output_asm_line_debug_info consistently.
4029 (dwarf2out_early_finish): Likewise.
4030 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4033 2018-07-11 Richard Biener <rguenther@suse.de>
4035 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4036 Rework father_bb setting in a way to avoid propagating constants
4037 multiple times on a loop body.
4039 2018-07-10 Mark Wielaard <mark@klomp.org>
4042 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4044 2018-07-10 Richard Biener <rguenther@suse.de>
4046 * hash-map.h (hash_map::iterator::operator*): Return
4047 references to key and value.
4049 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4052 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4053 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4054 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4055 (gimplify_omp_for): For composite loops, move outer
4056 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4057 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4058 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4059 TREE_LIST for both the original class iterator and the "last" helper
4060 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4061 loop, remember has_decl_expr from outer composite loops for the
4062 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4064 2018-07-09 Martin Sebor <msebor@redhat.com>
4068 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4069 accessing implicitly initialized array elements.
4070 * expr.c (string_constant): Handle string initializers of
4071 character arrays within aggregates.
4072 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4073 Store element offset. As a special case, handle zero size.
4074 (fold_nonarray_ctor_reference): Same.
4075 (fold_ctor_reference): Add argument. Store subobject offset.
4076 * gimple-fold.h (fold_ctor_reference): Add argument.
4078 2018-07-09 Paul Koning <ni1d@arrl.net>
4080 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4081 (pdp11_insn_cost): New function.
4082 (pdp11_md_asm_adjust): New function.
4083 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4084 (pdp11_rtx_costs): Update to match machine better.
4085 (output_addr_const_pdp11): Correct format mismatch warnings.
4086 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4087 * config/pdp11/pdp11.md: General change to add base_cost and/or
4088 length attributes for use by new pdp11_insn_cost function.
4089 (MIN_BRANCH): Correct definition.
4091 (doloop_end): Use standard pattern name for looping pattern.
4092 (doloop_end_nocc): New.
4093 (movsf): Add another constraint alternative.
4094 (zero_extendqihi2): Add constraint alternatives for not in place
4096 (zero_extendhisi2): Remove.
4097 (shift patterns): Add CC handling variants.
4101 (define_peephole2): New peephole to recognize mov that sets CC for
4104 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4106 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4107 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4110 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4113 * real.c (real_nextafter): Return true if result is denormal.
4115 2018-07-09 Martin Liska <mliska@suse.cz>
4117 * common.opt: Add back wrongly removed attribute.
4119 2018-07-09 Richard Biener <rguenther@suse.de>
4122 * dwarf2out.c (gen_block_die): For an early generated DIE
4123 always output high/low PC attributes.
4125 2018-07-09 Tom de Vries <tdevries@suse.de>
4127 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4128 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4129 onto VAR_DECL with abstract origin.
4131 2018-07-07 Jim Wilson <jimw@sifive.com>
4133 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4135 2018-07-07 Tom de Vries <tdevries@suse.de>
4137 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4139 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4141 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4143 (combine_bound): Use wide-int overflow calculation instead of
4145 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4147 * fold-const.c (int_const_binop_2): Same.
4148 (extract_muldiv_1): Same.
4149 (fold_div_compare): Same.
4150 (fold_abs_const): Same.
4152 * poly-int.h (add): Same.
4156 * predict.c (predict_iv_comparison): Same.
4157 * profile-count.c (slow_safe_scale_64bit): Same.
4158 * simplify-rtx.c (simplify_const_binary_operation): Same.
4159 * tree-chrec.c (tree_fold_binomial): Same.
4160 * tree-data-ref.c (split_constant_offset_1): Same.
4161 * tree-if-conv.c (idx_within_array_bound): Same.
4162 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4163 * tree-ssa-phiopt.c (minmax_replacement): Same.
4164 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4165 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4166 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4167 * wide-int.cc (wi::add_large): Same.
4168 (wi::mul_internal): Same.
4169 (wi::sub_large): Same.
4170 (wi::divmod_internal): Same.
4171 * wide-int.h: Change overflow type to overflow_type for neg, add,
4172 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4173 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4174 mul_internal, divmod_internal.
4175 (overflow_type): New enum.
4176 (accumulate_overflow): New.
4178 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4180 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4181 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4183 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4185 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4186 argument is checked for zero before entering loop, avoid checking again.
4188 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4190 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4191 * gimplify.h (generic_expr_could_trap_p): Likwise.
4192 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4194 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4196 PR tree-optimization/86401
4197 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4198 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4199 (fold_bit_and_mask): ... here. New helper function for match.pd.
4200 * fold-const.h (fold_bit_and_mask): Declare.
4201 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4203 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4206 * target.def (translate_mode_attribute): New hook.
4207 * targhooks.h (default_translate_mode_attribute): Declare.
4208 * targhooks.c (default_translate_mode_attribute): New function.
4209 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4210 * doc/tm.texi: Regenerate.
4211 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4212 (rs6000_translate_mode_attribute): New function.
4214 2018-07-06 Paul Koning <ni1d@arrl.net>
4216 * doc/md.texi (define_split): Document DONE and FAIL.
4217 (define_peephole2): Ditto.
4219 2018-07-05 Jeff Law <law@redhat.com>
4221 PR tree-optimization/86010
4222 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4223 both the head and tail of mem* and str* calls.
4225 2018-07-05 Jim Wilson <jimw@sifive.com>
4227 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4229 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4231 * config/aarch64/aarch64-simd.md: correct flags text for
4232 MIN_EXPR replacement.
4234 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4236 * configure: Regenerated.
4238 2018-07-05 Carl Love <cel@us.ibm.com>
4240 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4241 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4242 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4243 VSX_BUILTIN_DOUBLEL_V4SF.
4245 2018-07-05 Martin Sebor <msebor@redhat.com>
4248 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4249 than its domain to compute its the upper bound of a char array.
4251 2018-07-05 Nathan Sidwell <nathan@acm.org>
4253 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4254 * doc/cpp.texi: Update comment.
4255 * doc/tm.texi: Rebuilt.
4256 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4257 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4258 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4259 headers affected by extern "C".
4260 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4261 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4262 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4263 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4264 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4265 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4266 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4267 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4268 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4269 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4270 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4271 config/sparc/sp64-elf.h, config/spu/spu.h,
4272 config/stormy16/stormy16.h, config/v850/v850.h,
4273 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4274 define NO_IMPLICIT_EXTERN_C.
4275 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4277 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4280 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4281 instead of GET_MODE_SIZE when comparing Units.
4283 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4286 * rtlanal.c (set_noop_p): Constrain on mode change,
4287 include hard-reg-set.h
4289 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4291 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4293 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4296 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4299 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4301 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4303 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4306 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4308 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4310 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4313 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4315 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4317 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4318 neg_*_op* variables.
4320 2018-07-04 Martin Liska <mliska@suse.cz>
4322 * tree-switch-conversion.c: Define
4323 max_ratio_for_speed and max_ratio_for_size constants.
4325 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4326 Martin Liska <mliska@suse.cz>
4331 * common.opt: Rename align options with 'str_' prefix.
4332 * common/config/i386/i386-common.c (set_malign_value): New
4334 (ix86_handle_option): Use it to set -falign-* options/
4335 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4336 type from int to string.
4337 * config/aarch64/aarch64.c: Update default values from int
4339 * config/alpha/alpha.c (alpha_override_options_after_change):
4341 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4342 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4343 max skip conditionally.
4344 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4345 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4346 max skip conditionally.
4347 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4348 max skip conditionally.
4349 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4350 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4351 max skip conditionally.
4352 * config/i386/i386.c (struct ptt): Change type from int to
4354 (ix86_default_align): Set default values.
4355 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4356 max skip conditionally.
4357 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4358 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4359 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4360 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4361 max skip conditionally.
4362 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4363 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4364 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4365 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4366 max skip conditionally.
4367 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4368 * config/ia64/ia64.c (ia64_option_override): Set default values
4369 for alignment options.
4370 * config/m68k/m68k.c: Handle new str_align_* options.
4371 * config/mips/mips.c (mips_set_compression_mode): Change
4373 (mips_option_override): Set default values for options.
4374 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4376 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4378 * config/rx/rx.c (rx_option_override): Likewise.
4379 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4380 (LABEL_ALIGN): Use align_labels_log.
4381 (LOOP_ALIGN): Use align_loops_align.
4382 * config/s390/s390.c (s390_asm_output_function_label): Use new
4384 * config/sh/sh.c (sh_override_options_after_change):
4385 Change type of constants.
4386 * config/spu/spu.c (spu_sched_init): Likewise.
4387 * config/sparc/sparc.c (sparc_option_override): Set default
4389 * config/visium/visium.c (visium_option_override): Likewise.
4390 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4391 emit p2align format with last argument if it's not needed.
4392 * doc/invoke.texi: Document extended format of -falign-*.
4393 * final.c: Use align_labels alignment.
4394 * flags.h (struct target_flag_state): Change type to use
4396 (struct align_flags_tuple): New.
4397 (struct align_flags): Likewise.
4398 (align_loops_log): Redefine macro to use new types.
4399 (align_loops_max_skip): Redefine macro to use new types.
4400 (align_jumps_log): Redefine macro to use new types.
4401 (align_jumps_max_skip): Redefine macro to use new types.
4402 (align_labels_log): Redefine macro to use new types.
4403 (align_labels_max_skip): Redefine macro to use new types.
4404 (align_functions_log): Redefine macro to use new types.
4405 (align_loops): Redefine macro to use new types.
4406 (align_jumps): Redefine macro to use new types.
4407 (align_labels): Redefine macro to use new types.
4408 (align_functions): Redefine macro to use new types.
4409 (align_functions_max_skip): Redefine macro to use new types.
4410 (align_loops_value): New macro.
4411 (align_jumps_value): New macro.
4412 (align_labels_value): New macro.
4413 (align_functions_value): New macro.
4414 * function.c (invoke_set_current_function_hook): Propagate
4415 alignment values from flags to global variables default in
4417 * ipa-icf.c (sem_function::equals_wpa): Use
4418 cl_optimization_option_eq instead of memcmp.
4419 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4421 (cl_optimization_stream_in): Likewise.
4422 * optc-save-gen.awk: Support strings in cl_optimization.
4423 * opth-gen.awk: Likewise.
4424 * opts.c (finish_options): Remove error checking of invalid
4426 (MAX_CODE_ALIGN): Remove.
4427 (MAX_CODE_ALIGN_VALUE): Likewise.
4428 (parse_and_check_align_values): New function.
4429 (check_alignment_argument): Likewise.
4430 (common_handle_option): Use check_alignment_argument.
4431 * opts.h (parse_and_check_align_values): Declare.
4432 * toplev.c (init_alignments): Remove.
4433 (read_log_maxskip): New.
4434 (parse_N_M): Likewise.
4435 (parse_alignment_opts): Likewise.
4436 (backend_init_target): Remove usage of init_alignments.
4437 * toplev.h (parse_alignment_opts): Declare.
4438 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4440 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4441 * tree.c (cl_option_hasher::equal): New.
4442 * varasm.c: Use new global macros.
4444 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4446 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4447 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4448 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4449 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4450 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4451 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4452 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4453 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4454 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4455 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4456 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4457 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4458 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4460 2018-07-04 Martin Liska <mliska@suse.cz>
4461 Jonathan Wakely <jwakely@redhat.com>
4463 * coverage.c: Use correct type.
4464 * doc/invoke.texi: Language correction.
4466 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4469 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4470 ENDBRANCH for non-tail call which may return via indirect branch.
4471 * doc/extend.texi: Document indirect_return attribute.
4473 2018-07-03 Martin Sebor <msebor@redhat.com>
4475 PR tree-optimization/86274
4476 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4478 (format_floating): Correct handling of infinities and NaNs.
4480 2018-07-03 Martin Sebor <msebor@redhat.com>
4482 * print-tree.c (print_real_cst): New function.
4483 (print_node_brief): Call it.
4484 (print_node): Ditto.
4486 2018-07-03 Jeff Law <law@redhat.com>
4488 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4489 into a single pattern.
4491 * config/h8300/h8300.md (ors code_iterator): New.
4492 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4493 a single pattern and single splitter.
4494 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4495 (iorqi3_1, xorqi3_1): Likewise.
4496 (iorqi3, xorqi3 expanders): Similarly.
4498 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4499 (movmd_internal) into a single pattern using the P mode iterator.
4500 (movmd splitters): Similarly.
4501 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4502 (movsd splitters): Similarly.
4504 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4505 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4506 right table based on the size of the operand.
4507 * config/h8300/h8300.md (length_table): Corresponding changes. All
4508 references to "addb", "addw" and "addl" changed to "add".
4509 (btst patterns): Merge two variants into a single pattern.
4510 (tstqi, tsthi): Likewise.
4511 (addhi3_incdec, addsi3_incdec): Likewise.
4512 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4513 (mulhi3, mulsi3): Likewise.
4514 (udivhi3, udivsi3): Likewise.
4515 (divhi3, divsi3): Likewise.
4516 (andorqi3, andorhi3, andorsi3): Likewise.
4518 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4521 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4522 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4524 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4526 PR tree-optimization/85694
4527 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4528 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4530 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4531 (<u>avg<mode>3_ceil): New patterns.
4533 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4535 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4536 scan-tree-dump directive.
4537 * gcc.dg/vect/slp-perm-2.c: Likewise.
4538 * gcc.dg/vect/slp-perm-3.c: Likewise.
4539 * gcc.dg/vect/slp-perm-5.c: Likewise.
4540 * gcc.dg/vect/slp-perm-6.c: Likewise.
4541 * gcc.dg/vect/slp-perm-7.c: Likewise.
4542 * gcc.dg/vect/slp-perm-8.c: Likewise.
4544 2018-07-03 Marek Polacek <polacek@redhat.com>
4547 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4549 2018-07-03 Richard Biener <rguenther@suse.de>
4552 * tree-ssa-structalias.c (find_func_clobbers): Properly
4553 handle indirect calls.
4555 2018-07-03 Jeff Law <law@redhat.com>
4557 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4558 (shifts): New code iterator.
4559 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4560 expander. Fix HImode handling on H8/SX.
4561 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4562 (subqi3, subhi3, subsi3 expanders): Likewise.
4563 (andqi3, andhi3, andsi3 expanders): Likewise.
4564 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4565 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4566 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4567 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4568 (zero_extendqihi2, zero_extendqisi2): Likewise.
4569 (extendqihi2, extendqisi2): Likewise.
4570 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4571 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4572 (rotlqi3_1, rotlhi3_1): Likewise.
4573 (logicalhi3_sn, logicalsi3_sn): Likewise.
4574 (logicalhi3, logicalsi3): Likewise.
4576 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4578 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4579 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4580 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4581 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4582 type to append_pattern_def_seq instead of creating a stmt_vec_info
4584 (build_mask_conversion): Likewise. Remove vinfo argument.
4585 (vect_add_conversion_to_patterm): Likewise, renaming to...
4586 (vect_add_conversion_to_pattern): ...this.
4587 (vect_recog_mask_conversion_pattern): Update call to
4588 build_mask_conversion. Pass the vector type to
4589 append_pattern_def_seq here too.
4590 (vect_recog_gather_scatter_pattern): Update call to
4591 vect_add_conversion_to_pattern.
4593 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4595 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4596 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4597 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4598 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4599 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4600 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4601 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4602 append_pattern_def_seq instead of new_pattern_def_seq.
4603 (vect_recog_divmod_pattern): Do both of the above.
4604 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4607 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4609 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4610 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4611 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4612 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4613 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4614 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4615 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4616 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4617 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4618 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4619 parameter with a single stmt_vec_info.
4620 (vect_recog_func_ptr): Likewise.
4621 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4622 (vect_try_gather_scatter_pattern): ...this.
4623 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4624 the stmt_vec_info of the statement to be matched. Don't clear
4625 STMT_VINFO_RELATED_STMT.
4626 (vect_pattern_recog): Update call accordingly.
4628 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4630 PR tree-optimization/85694
4631 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4632 (uavgM3_ceil): Document new optabs.
4633 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4634 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4636 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4637 (savg_ceil_optab): New optabs.
4638 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4639 (vect_vect_recog_func_ptrs): Add it.
4640 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4641 constant directly from the associated lhs.
4643 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4645 * tree-vect-patterns.c (vect_split_statement): New function.
4646 (vect_convert_input): Use it to try to split an existing cast.
4648 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4650 * poly-int.h (print_hex): New function.
4651 * dumpfile.h (dump_dec, dump_hex): Declare.
4652 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4653 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4654 min_input_precision, operation_precision and operation_sign.
4655 * tree-vect-patterns.c (vect_get_range_info): New function.
4656 (vect_same_loop_or_bb_p, vect_single_imm_use)
4657 (vect_operation_fits_smaller_type): Delete.
4658 (vect_look_through_possible_promotion): Add an optional
4659 single_use_p parameter.
4660 (vect_recog_over_widening_pattern): Rewrite to use new
4661 stmt_vec_info infomration. Handle one operation at a time.
4662 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4663 (vect_truncatable_operation_p, vect_set_operation_type)
4664 (vect_set_min_input_precision): New functions.
4665 (vect_determine_min_output_precision_1): Likewise.
4666 (vect_determine_min_output_precision): Likewise.
4667 (vect_determine_precisions_from_range): Likewise.
4668 (vect_determine_precisions_from_users): Likewise.
4669 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4670 (vect_vect_recog_func_ptrs): Put over_widening first.
4672 (vect_pattern_recog): Call vect_determine_precisions.
4674 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4676 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4677 statements that have been replaced by further pattern statements.
4678 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4680 2018-07-03 Richard Biener <rguenther@suse.de>
4682 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4683 always set *dt. Dump vectype in vectype overload.
4684 * dumpfile.h (dump_gimple_expr): New function.
4685 (dump_gimple_expr_loc): Likewise.
4686 * dumpfile.c (dump_gimple_expr): New function.
4687 (dump_gimple_expr_loc): Likewise.
4689 2018-07-02 Jeff Law <law@redhat.com>
4691 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4692 the H8/300, H8/300H and H8/S variants into a single pattern.
4693 (movhi_h8300, movqi_h8300hs): Similarly.
4694 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4695 (QHI mode iterator): New.
4697 * config/h8300/h8300.md: Remove trailing whitespace.
4699 2018-07-02 Jim Wilson <jimw@sifive.com>
4701 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4702 instead of emit_insn for interrupt returns.
4703 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4704 (riscv_sret, riscv_uret): Likewise.
4706 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4708 * pretty-print.c (selftest::test_pp_format): Move save and restore
4709 of quotes to class auto_fix_quotes, and add an instance.
4710 * selftest.c: Include "intl.h".
4711 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4712 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4713 * selftest.h (selftest::auto_fix_quotes): New class.
4715 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4717 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4718 (aarch64_sve_prepare_conditional_op): Remove.
4719 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4720 Allow aarch64_simd_reg_or_zero as select operand; remove
4721 the aarch64_sve_prepare_conditional_op call.
4722 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4723 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4724 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4725 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4726 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4727 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4728 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4729 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4730 and a splitters to match all of the *_any patterns.
4731 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4733 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4734 (SVE_COND_FP_BINARY_REV): Remove.
4735 (sve_int_op_rev, sve_fp_op_rev): New.
4736 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4737 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4738 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4739 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4740 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4741 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4742 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4743 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4744 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4746 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4747 Remove match_dup 1 from the inner unspec.
4748 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4750 * config/aarch64/aarch64.md (movprfx): New attr.
4751 (length): Default movprfx to 8.
4752 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4753 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4754 (*<su>mul<SVE_I>3_highpart): Likewise.
4755 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4756 (*v<ASHIFT><SVE_I>3): Likewise.
4757 (*<su><MAXMIN><SVE_I>3): Likewise.
4758 (*<su><MAXMIN><SVE_F>3): Likewise.
4759 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4760 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4761 (*div<SVE_F>4): Likewise.
4763 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4765 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4768 2018-07-02 Richard Biener <rguenther@suse.de>
4770 PR tree-optimization/86363
4771 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4772 memset argument refers to a non-variable address.
4774 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4776 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4777 {PLUS,MINUS}_EXPR code to...
4778 (adjust_symbolic_bound): ...here,
4779 (combine_bound): ...here,
4780 (set_value_range_with_overflow): ...and here.
4782 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4784 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4786 (extract_range_from_abs_expr): ...here.
4788 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4790 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4791 -fno-omit-frame-pointer when not optimizing.
4793 2018-07-02 Martin Liska <mliska@suse.cz>
4796 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4797 (propagate_nothrow): Likewise.
4799 2018-07-02 Martin Liska <mliska@suse.cz>
4802 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4804 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4806 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4807 function in r262149, changing "loc" param from source_location to
4808 const dump_location_t &.
4809 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4810 declaration, as above.
4812 2018-07-01 Paul Koning <ni1d@arrl.net>
4814 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4815 -munit-asm, -mgnu-asm, -mdec-asm.
4816 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4817 (pdp11_output_labelref): New.
4818 (pdp11_output_def): New.
4819 (pdp11_output_addr_vec_elt): New.
4820 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4821 %# and %@ format codes.
4822 (pdp11_option_override): New.
4823 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4824 (pdp11_output_ident): New.
4825 (pdp11_asm_named_section): New.
4826 (pdp11_asm_init_sections): New.
4827 (pdp11_file_start): New.
4828 (pdp11_file_end): New.
4829 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4830 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4831 %o, like %c but octal.
4832 (pdp11_option_override): New.
4833 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4835 (DATA_SECTION_ASM_OP): Ditto.
4836 (READONLY_DATA_SECTION_ASM_OP): New.
4837 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4838 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4839 (ASM_OUTPUT_LABELREF): Ditto.
4840 (ASM_OUTPUT_DEF): Ditto.
4841 (ASM_OUTPUT_EXTERNAL): New.
4842 (ASM_OUTPUT_SOURCE_FILENAME): New.
4843 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4844 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4845 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4846 %# and %@ format codes.
4847 * config/pdp11/pdp11.opt (mgnu-asm): New.
4848 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4849 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4850 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4852 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4854 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4855 dereferencing path[] beyond its length.
4857 (debug_all_paths): New.
4858 (rewire_first_differing_edge): New.
4859 (adjust_paths_after_duplication): New.
4860 (duplicate_thread_path): Call adjust_paths_after_duplication.
4862 (thread_through_all_blocks): Add new argument to
4863 duplicate_thread_path.
4865 2018-06-30 Jim Wilson <jimw@sifive.com>
4867 * config/riscv/predicates.md (p2m1_shift_operand): New.
4868 (high_mask_shift_operand): New.
4869 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4870 pattern using p2m1_shift_operand.
4871 (lshsi3_zero_extend_3+2): New combiner pattern using
4872 high_mask_shift_operand.
4874 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4876 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4878 (vect_recog_rotate_pattern): ...here.
4879 (vect_convert_input): Try to insert casts of invariants in the
4881 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4882 preheader to be empty.
4884 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4886 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4887 vector type. If given, install it in the new statement's
4889 (vect_element_precision): New function.
4890 (vect_unpromoted_value): New struct.
4891 (vect_unpromoted_value::vect_unpromoted_value): New function.
4892 (vect_unpromoted_value::set_op): Likewise.
4893 (vect_look_through_possible_promotion): Likewise.
4894 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4895 (vect_widened_op_tree, vect_convert_input): Likewise.
4896 (vect_convert_inputs, vect_convert_output): Likewise.
4897 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4898 to handle the optional cast of the multiplication result and
4899 vect_widened_op_tree to detect the widened multiplication itself.
4900 Do not require the input and output of promotion casts to have
4901 the same sign, but base the signedness of the operation on the
4902 input rather than the result. If the pattern includes two
4903 promotions, check that those promotions have the same sign.
4904 Do not restrict the MULT_EXPR handling to a double-width result;
4905 handle quadruple-width results and wider. Use vect_convert_inputs
4906 to convert the inputs to the common type.
4907 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
4908 to handle the optional cast of the ABS result. Also allow a sign
4909 change or a sign extension between the ABS and MINUS.
4910 Use vect_widened_op_tree to detect the widened subtraction and use
4911 vect_convert_inputs to convert the inputs to the common type.
4912 (vect_handle_widen_op_by_const): Delete.
4913 (vect_recog_widen_op_pattern): New function.
4914 (vect_recog_widen_mult_pattern): Use it.
4915 (vect_recog_widen_shift_pattern): Likewise.
4916 (vect_recog_widen_sum_pattern): Use
4917 vect_look_through_possible_promotion to handle the promoted
4920 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4922 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4923 the containing gimple_seq *.
4924 * gimple-iterator.h (gsi_for_stmt): Declare it.
4925 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4926 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4927 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4928 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4929 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4931 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4933 (vect_mark_pattern_stmts): ...here. Handle cases in which the
4934 statement being replaced is part of an existing pattern
4935 definition sequence, inserting the new pattern statements before
4937 (vect_pattern_recog_1): Don't return a bool. If the statement
4938 is already part of a pattern, instead apply pattern matching
4939 to the pattern definition statements. Don't clear the
4940 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4941 (vect_pattern_recog): Don't break after the first match;
4942 continue processing the pattern definition statements instead.
4943 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4945 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4947 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4948 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4949 (vect_recog_widen_sum_pattern): Use it.
4951 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4953 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4954 phi is not a pattern statement and has not been replaced by
4955 a pattern statement.
4956 * tree-vect-patterns.c (type_conversion_p): Don't check
4957 STMT_VINFO_IN_PATTERN_P.
4958 (vect_recog_vector_vector_shift_pattern): Likewise.
4959 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4960 the pattern statement rather than the original statement; check
4961 directly for a WIDEN_MULT_EXPR here.
4962 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4963 vect_is_simple_use to return the pattern statement rather
4964 than the original statement; use is_pattern_stmt_p to check
4965 for such a pattern statement.
4966 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4967 to return the pattern statement rather than the original statement;
4968 don't do the same transformation here.
4969 (vect_is_simple_use): If the defining statement has been replaced
4970 by a pattern statement, return the pattern statement instead.
4971 Remove the corresponding (local) transformation from the vectype
4974 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4976 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4977 end and default to null.
4978 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4979 (vectorizable_reduction): Update calls accordingly, dropping the
4980 gimple ** argument if the passed-back statement isn't needed.
4981 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4982 (vect_recog_rotate_pattern): Likewise.
4983 (vect_recog_mask_conversion_pattern): Likewise.
4984 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4985 (vect_mask_constant_operand_p): Likewise.
4986 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4987 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4988 (get_group_load_store_type, get_load_store_type): Likewise.
4989 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4990 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4991 (vectorizable_conversion, vectorizable_assignment): Likewise.
4992 (vectorizable_shift, vectorizable_operation): Likewise.
4993 (vectorizable_store, vect_is_simple_cond): Likewise.
4994 (vectorizable_condition, vectorizable_comparison): Likewise.
4995 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4996 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4997 and move it to the end. Cope with null def_stmt_outs.
4999 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5001 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5003 2018-06-29 Jeff Law <law@redhat.com>
5005 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5006 displacements for TARGET_V850E2V3 and newer.
5007 (TARGET_LRA_P): Remove. Defaults to LRA now.
5008 * config/v850/v850.md (sign23byte_load): Remove.
5009 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5010 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5012 2018-06-29 Martin Liska <mliska@suse.cz>
5015 * coverage.c (coverage_init): Mangle full path name.
5016 * doc/invoke.texi: Document the change.
5017 * gcov-io.c (mangle_path): New.
5018 * gcov-io.h (mangle_path): Likewise.
5019 * gcov.c (mangle_name): Use mangle_path for path mangling.
5021 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5023 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5024 if starting source register is not even.
5026 2018-06-29 Martin Liska <mliska@suse.cz>
5028 PR tree-optimization/86263
5029 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5030 Make edge redirection.
5032 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5034 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5035 (dump_scope_depth): New variable.
5036 (get_dump_scope_depth): New function.
5037 (dump_begin_scope): New function.
5038 (dump_end_scope): New function.
5039 * dumpfile.h (get_dump_scope_depth): New declaration.
5040 (dump_begin_scope): New declaration.
5041 (dump_end_scope): New declaration.
5042 (class auto_dump_scope): New class.
5043 (AUTO_DUMP_SCOPE): New macro.
5044 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5047 2018-06-29 Richard Biener <rguenther@suse.de>
5049 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5050 compute_all_dependences succeeds.
5051 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5052 exceed --param loop-max-datarefs-for-datadeps.
5054 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5056 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5058 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5061 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5062 alternative 0 in preferred_for_speed attribute.
5064 2018-06-28 Paul Koning <ni1d@arrl.net>
5066 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5067 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5068 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5069 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5070 for shift insn patterns.
5072 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5074 * cgraph.c (cgraph_node::get_body): Replace assignments to
5075 "dump_file" with calls to set_dump_file.
5076 * dumpfile.c (alt_dump_file): Make static, and group with...
5077 (alt_flags): ...this definition.
5078 (dumps_are_enabled): New variable.
5079 (refresh_dumps_are_enabled): New function.
5080 (set_dump_file): New function.
5081 (set_alt_dump_file): New function.
5082 (gcc::dump_manager::dump_start): Replace assignments to
5083 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5085 (gcc::dump_manager::dump_finish): Likewise.
5086 * dumpfile.h (alt_dump_file): Delete decl.
5087 (dumps_are_enabled): New variable decl.
5088 (set_dump_file): New function decl.
5089 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5091 * tree-nested.c (lower_nested_functions): Replace assignments to
5092 "dump_file" with calls to set_dump_file.
5094 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5096 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5097 goto_locus of each outgoing edge of each basic block.
5099 2018-06-28 Richard Biener <rguenther@suse.de>
5101 * dwarf2out.c (decl_scope_table): Remove.
5102 (push_decl_scope): Likewise.
5103 (pop_decl_scope): Likewise.
5104 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5105 (gen_struct_or_union_type_die): Likewise.
5106 (gen_tagged_type_die): Likewise.
5107 (dwarf2out_init): Do not initialize decl_scope_table.
5108 (dwarf2out_c_finalize): Do not free it.
5110 2018-06-28 Richard Biener <rguenther@suse.de>
5112 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5113 deciding whether to not re-use a DIE.
5115 2018-06-28 Richard Biener <rguenther@suse.de>
5117 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5118 DW_AT_abstract_origin attribute.
5120 2018-06-28 Martin Liska <mliska@suse.cz>
5122 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5123 Use newly introduced constants.
5124 * tree-switch-conversion.h (struct jump_table_cluster):
5125 Define max_ratio_for_size and max_ratio_for_speed.
5127 2018-06-28 Martin Liska <mliska@suse.cz>
5129 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5130 Add new checking assert to catch invalid state.
5131 (jump_table_cluster::can_be_handled): Handle single case
5133 (jump_table_cluster::is_beneficial): Bail out for such case.
5134 (bit_test_cluster::find_bit_tests):
5135 Add new checking assert to catch invalid state.
5136 (bit_test_cluster::can_be_handled): Handle single case
5138 (bit_test_cluster::is_beneficial): Bail out for such case.
5139 (switch_decision_tree::analyze_switch_statement):
5142 2018-06-28 Martin Liska <mliska@suse.cz>
5144 * common.opt: Introduce -completion option.
5145 * gcc.c (driver_handle_option): Handle it.
5146 (driver::main): Print completions if completion
5148 * opt-suggestions.c (option_proposer::get_completions):
5150 (option_proposer::suggest_completion): Likewise.
5151 (option_proposer::find_param_completions): Likewise.
5152 (verify_autocompletions): Likewise.
5153 (test_completion_valid_options): Likewise.
5154 (test_completion_valid_params): Likewise.
5155 (in_completion_p): Likewise.
5156 (empty_completion_p): Likewise.
5157 (test_completion_partial_match): Likewise.
5158 (test_completion_garbage): Likewise.
5159 (opt_proposer_c_tests): Likewise.
5160 * opt-suggestions.h: Declare new functions.
5161 * opts.c (common_handle_option): Handle OPT__completion_.
5162 * selftest-run-tests.c (selftest::run_tests): Add
5163 opt_proposer_c_tests.
5164 * selftest.c (assert_str_startswith): New.
5165 * selftest.h (assert_str_startswith): Likewise.
5166 (opt_proposer_c_tests): New.
5167 (ASSERT_STR_STARTSWITH): Likewise.
5169 2018-06-28 Martin Liska <mliska@suse.cz>
5171 * Makefile.in: Add opt-suggestions.o.
5172 * gcc-main.c: Include opt-suggestions.h.
5173 * gcc.c (driver::driver): Likewise.
5174 (driver::~driver): Remove m_option_suggestions.
5175 (driver::build_option_suggestions): Moved to option_proposer.
5176 (driver::suggest_option): Likewise.
5177 (driver::handle_unrecognized_options): Use option_proposer.
5178 * gcc.h (class driver): Add new memver m_option_proposer.
5179 * opt-suggestions.c: New file.
5180 * opt-suggestions.h: New file.
5182 2018-06-28 Martin Liska <mliska@suse.cz>
5184 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5185 (auto_string_vec::~auto_string_vec): Likewise.
5187 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5189 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5190 prevent_decl_creation_for_types fields up and add reset_location field.
5191 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5192 statement if id->reset_location is true.
5193 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5194 id->reset_location is true.
5195 (copy_phis_for_bb): Force input_location on the arguments if
5196 id->reset_location is true.
5197 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5198 is set on the function to be inlined.
5200 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5202 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5204 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5206 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5207 registers for Pmode.
5208 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5209 hard registers for the clobbered pseudo.
5211 2018-06-27 Paul Koning <ni1d@arrl.net>
5213 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5214 mutually exclusive options.
5215 * config/pdp11/constraints.md (h): New constraint.
5216 (O): Update definition to match shift code generation.
5217 (D): New constraint.
5218 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5220 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5222 (output_jump): Change arguments.
5223 (pdp11_fixed_cc_regs): New function.
5224 (pdp11_cc_mode): Ditto.
5225 (pdp11_expand_shift): Ditto.
5226 (pdp11_assemble_shift): Ditto.
5227 (pdp11_small_shift): Ditto.
5228 (pdp11_branch_cost): Remove.
5229 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5231 (pdp11_register_move_cost): Update for CC registers.
5232 (pdp11_rtx_costs): Add case for LSHIFTRT.
5233 (pdp11_output_jump): Add CCNZ mode conditional branches.
5234 (notice_update_cc_on_set): Remove.
5235 (pdp11_cc_mode): New function.
5236 (simple_memory_operand): Correct pre/post decrement case.
5237 (no_side_effect_operand): New function.
5238 (pdp11_regno_reg_class): Add CC_REGS class.
5239 (pdp11_fixed_cc_regs): New function.
5240 (pdp11_small_shift): New function.
5241 (pdp11_expand_shift): New function to expand shift insns.
5242 (pdp11_assemble_shift): New function to output shifts.
5243 (pdp11_branch_cost): Remove.
5244 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5245 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5246 (WCHAR_TYPE): Ditto.
5247 (PTRDIFF_TYPE): Ditto.
5248 (ADJUST_INSN_LENGTH): New macro.
5249 (FIXED_REGISTERS): Add CC registers.
5250 (CALL_USED_REGISTERS): Ditto.
5252 (REG_CLASS_NAMES): Ditto.
5253 (REG_CLASS_CONTENTS): Ditto.
5254 (SELECT_CC_MODE): Use new function.
5255 (TARGET_FLAGS_REGNUM): New macro.
5256 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5257 (cc0_reg_rtx): Remove.
5258 (CC_STATUS_MDEP): Remove.
5259 (CC_STATUS_MDEFP_INIT): Remove.
5260 (CC_IN_FPU): Remove.
5261 (NOTICE_UPDATE_CC): Remove.
5262 (REGISTER_NAMES): Add CC registers.
5263 (BRANCH_COST): Change to constant 1.
5264 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5266 * config/pdp11/pdp11.opt (mbcopy): Remove.
5267 (mbcopy-builtin): Remove.
5268 (mbranch-cheap): Remove.
5269 (mbranch-expensive): Remove.
5270 * config/pdp11/predicates.md (expand_shift_operand): Update to
5271 match shift code generation.
5272 (ccnz_operator): New predicate.
5273 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5274 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5275 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5277 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5278 description of O constraint.
5280 2018-06-27 Jeff Law <law@redhat.com>
5281 Austin Law <austinklaw@gmail.com>
5283 * config/v850/v850.md (addsi3_set_flags): New pattern.
5284 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5285 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5286 (zero_extendhisi2_v850_set_flags): Likewise.
5287 (zero_extendqisi2_v850_set_flags): Likewise.
5288 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5289 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5290 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5292 * config/v850/v850-protos.h (notice_update_cc): Remove.
5293 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5294 (v850_print_operand): Handle 'D' and "d".
5295 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5296 Add handling of arithmetic/logical operations compared against zero.
5297 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5298 Do not look at v850_compare_op, instead get mode from last argument.
5299 (v850_gen_compare): Remove
5300 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5301 after reload for prologue insns.
5302 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5304 (construct_save_jarl): Likewise.
5305 (TARGET_FLAGS_REGNUM): Define.
5306 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5307 (NOTICE_UPDATE_CC): Remove.
5308 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5309 than cc0. Conditionalize on reload_completed.
5310 (cmpsi_insn, setfcc_insn): Likewise.
5311 (tst1 splitter): Turn into define_and_split which sets the flags
5313 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5314 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5315 (cstoresf4, cstoredf4): Clobber the flags.
5316 (cmpsi, cmpsf, cmpdf): Remove expanders.
5317 (setf_insn): Remove pattern.
5318 (addsi3): Turn into define_and_split which clobbers the flags after
5319 reload and a suitable pattern (addsi3_clobber_flags) for use after
5321 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5322 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5323 (ashrsi3, ashrsi3_v850e2): Likewise.
5324 (bins): Clobber the flags.
5325 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5326 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5327 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5328 (call_value_internal_short, call_value_internal_long): Likewise.
5329 (callt_save_interrupt, callt_return_interrupt): Likewise.
5330 (save_interrupt, return_interrupt): Likewise.
5331 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5332 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5333 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5334 (All FP comparisons): Only allow after reload has completed.
5336 (divh, divhu): Tweak output template.
5337 (branch_z_normal, branch_z_invert): Remove
5338 (branch_nz_normal, branch_nz_invert): Likewise.
5339 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5341 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5342 * config/v850/v850.c (notice_update_cc): Remove.
5343 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5344 (CC_NO_CARRY): Likewise.
5345 (NOTICE_UPDATE_CC): Define to nothing.
5346 * config/v850/v850.md: Remove block comment on cc0 handling
5347 Remove "cc" attribute from all patterns. Remove cc_status handling
5348 from all patterns. Minor formatting fixes.
5350 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5352 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5353 (cortex-a76.cortex-a55): Likewise.
5354 * config/aarch64/aarch64-tune.md: Regenerate.
5355 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5356 cortex-a76.cortex-a55.
5358 2018-06-27 Jeff Law <law@redhat.com>
5360 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5361 (MULTILIB_DIRNAMES): Similarly.
5363 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5365 * gimple.h (gimple_return_retbnd): Delete.
5366 (gimple_return_set_retbnd): Likewise.
5367 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5368 gimple_return_set_retbnd.
5369 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5370 gimple_return_retbnd and adjust.
5371 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5372 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5373 Explicitly return NULL in a couple more cases. Move assertion
5374 on debug statements and remove unreachable code.
5375 (reset_debug_binding): Do not test id->retbnd.
5376 (expand_call_inline): Do not set it.
5378 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5380 * configure.ac: Add --disable-gcov option.
5381 * configure: Regenerate.
5382 * Makefile.in: Honour @enable_gcov@.
5383 * doc/install.texi: Document --disable-gcov.
5385 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5387 * config/arm/arm-cpus.in (cortex-a76): New entry.
5388 (cortex-a76.cortex-a55): Likewise.
5389 * config/arm/arm-tables.opt: Regenerate.
5390 * config/arm/arm-tune.md: Likewise.
5391 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5392 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5393 cortex-a76.cortex-a55.
5395 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5398 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5400 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5402 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5404 (EPILOGUE_USES): Likewise.
5406 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5408 * tree-inline.c (remap_location): New function extracted from...
5409 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5410 (copy_phis_for_bb): ...here. Call remap_location.
5411 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5413 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5415 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5416 unaligned vsx for 16B memset.
5418 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5421 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5422 ieee128_float_type_node to long_double_type_node unless
5423 TARGET_LONG_DOUBLE_128 is set.
5425 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5427 * cfgloop.c (get_loop_location): Convert return type from
5428 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5429 by implicit construction from rtx_insn *, and using
5430 dump_user_location_t::from_function_decl for the fallback case.
5431 * cfgloop.h (get_loop_location): Convert return type from
5432 location_t to dump_user_location_t.
5433 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5434 dump_printf_loc to pass in a dump_location_t rather than a
5435 location_t, via the gimple stmt.
5436 * coverage.c (get_coverage_counts): Update calls to
5437 dump_printf_loc to pass in dump_location_t rather than a
5439 * doc/optinfo.texi (Dump types): Convert example of
5440 dump_printf_loc from taking "locus" to taking "insn". Update
5441 description of the "_loc" calls to cover dump_location_t.
5442 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5444 (dump_user_location_t::dump_user_location_t): New constructors,
5445 from gimple *stmt and rtx_insn *.
5446 (dump_user_location_t::from_function_decl): New function.
5447 (dump_loc): Make static.
5448 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5449 const dump_location_t &.
5450 (dump_generic_expr_loc): Delete.
5451 (dump_printf_loc): Convert param "loc" from location_t to
5452 const dump_location_t &.
5453 (selftest::test_impl_location): New function.
5454 (selftest::dumpfile_c_tests): New function.
5455 * dumpfile.h: Include "profile-count.h".
5456 (class dump_user_location_t): New class.
5457 (struct dump_impl_location_t): New struct.
5458 (class dump_location_t): New class.
5459 (dump_printf_loc): Convert 2nd param from source_location to
5460 const dump_location_t &.
5461 (dump_generic_expr_loc): Delete.
5462 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5463 const dump_location_t &.
5464 * gimple-fold.c (fold_gimple_assign): Update call to
5465 dump_printf_loc to pass in a dump_location_t rather than a
5466 location_t, via the gimple stmt.
5467 (gimple_fold_call): Likewise.
5468 * gimple-loop-interchange.cc
5469 (loop_cand::analyze_iloop_reduction_var): Update for change to
5470 check_reduction_path.
5471 (tree_loop_interchange::interchange): Update for change to
5473 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5474 change in return-type of find_loop_location.
5475 (graphite_regenerate_ast_isl): Likewise.
5476 * graphite-optimize-isl.c (optimize_isl): Likewise.
5477 * graphite.c (graphite_transform_loops): Likewise.
5478 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5479 pass in a dump_location_t rather than a location_t, via the
5481 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5482 * ipa.c (walk_polymorphic_call_targets): Likewise.
5483 * loop-unroll.c (report_unroll): Convert "locus" param from
5484 location_t to dump_location_t.
5485 (decide_unrolling): Update for change to get_loop_location's
5487 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5488 location_t to dump_user_location_t.
5489 (grid_find_single_omp_among_assignments_1): Updates calls to
5490 dump_printf_loc to pass in a dump_location_t rather than a
5491 location_t, via the gimple stmt.
5492 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5493 location_t to dump_location_t. Updates calls to dump_printf_loc
5494 to pass in a dump_location_t rather than a location_t, via the
5496 (grid_inner_loop_gridifiable_p): Likewise.
5497 (grid_dist_follows_simple_pattern): Likewise.
5498 (grid_gfor_follows_tiling_pattern): Likewise.
5499 (grid_target_follows_gridifiable_pattern): Likewise.
5500 (grid_attempt_target_gridification): Convert initialization
5501 of local "grid" from memset to zero-initialization; FIXME: does
5502 this require C++11? Update call to dump_printf_loc to pass in a
5503 optinfo_location rather than a location_t, via the gimple stmt.
5504 * profile.c (read_profile_edge_counts): Updates call to
5505 dump_printf_loc to pass in a dump_location_t rather than a
5507 (compute_branch_probabilities): Likewise.
5508 * selftest-run-tests.c (selftest::run_tests): Call
5510 * selftest.h (dumpfile_c_tests): New decl.
5511 * tree-loop-distribution.c (pass_loop_distribution::execute):
5512 Update for change in return type of find_loop_location.
5513 * tree-parloops.c (parallelize_loops): Likewise.
5514 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5515 "locus" from location_t to dump_user_location_t.
5516 (canonicalize_loop_induction_variables): Likewise.
5517 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5518 for change in return type of find_loop_location.
5519 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5520 to dump_printf_loc to pass in a dump_location_t rather than a
5521 location_t, via the stmt.
5522 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5524 * tree-vect-loop-manip.c (find_loop_location): Convert return
5525 type from source_location to dump_user_location_t.
5526 (vect_do_peeling): Update for above change.
5527 (vect_loop_versioning): Update for change in type of
5529 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5530 from location_t to dump_user_location_t.
5531 (vect_estimate_min_profitable_iters): Update for change in type
5533 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5534 location_t to dump_location_t.
5535 (vect_slp_bb): Update for change in type of vect_location.
5536 * tree-vectorizer.c (vect_location): Convert from source_location
5537 to dump_user_location_t.
5538 (try_vectorize_loop_1): Update for change in vect_location's type.
5539 (vectorize_loops): Likewise.
5540 (increase_alignment): Likewise.
5541 * tree-vectorizer.h (vect_location): Convert from source_location
5542 to dump_user_location_t.
5543 (find_loop_location): Convert return type from source_location to
5544 dump_user_location_t.
5545 (check_reduction_path): Convert 1st param from location_t to
5546 dump_user_location_t.
5547 * value-prof.c (check_counter): Update call to dump_printf_loc to
5548 pass in a dump_user_location_t rather than a location_t; update
5549 call to error_at for change in type of "locus".
5550 (check_ic_target): Update call to dump_printf_loc to
5551 pass in a dump_user_location_t rather than a location_t, via the
5554 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5556 * config/s390/s390.h (enum processor_flags): Do not use
5557 default tune parameter when -march was specified.
5559 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5562 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5563 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5566 2018-06-26 Richard Biener <rguenther@suse.de>
5568 PR tree-optimization/86287
5570 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5571 (vect_analyze_loop): Initialize n_stmts.
5573 2018-06-26 Richard Biener <rguenther@suse.de>
5576 * fold-const.c (fold_convertible_p): Pointer extension
5579 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5582 * dwarf2out.c (loc_list_has_views): Adjust comments.
5583 (dw_loc_list): Split single cross-partition range with
5586 2018-06-25 Jeff Law <law@redhat.com>
5588 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5589 on -mbig-switch by default.
5591 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5593 (const_float_0_operand): Remove unused predicate.
5594 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5595 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5596 (recipsf2): New expander. Original pattern now called
5598 (recipdf2, recipdf2_insn): Similarly.
5599 (rsqrtsf2, rsqrtsf2_insn): Similarly
5600 (rsqrtdf2, rsqrtdf2_insn): Similarly
5602 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5604 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5605 Simplify logic for FreeBSD (twice).
5607 2018-06-25 Martin Sebor <msebor@redhat.com>
5609 PR tree-optimization/86204
5610 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5611 a strnlen result if it's less than the length of the string.
5613 2018-06-25 Martin Sebor <msebor@redhat.com>
5615 PR tree-optimization/85700
5616 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5617 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5618 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5620 2018-06-25 Martin Sebor <msebor@redhat.com>
5622 * doc/extend.texi (Zero-length arrays): Update and clarify.
5624 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5626 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5627 added IEEE/IBM long double multilib support on PowerPC little
5628 endian Linux systems.
5629 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5630 (MULTILIB_DEFAULTS): Likewise.
5631 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5633 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5634 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5635 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5637 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5640 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5641 (REORDER_45): Likewise.
5643 2018-06-25 Jeff Law <law@redhat.com>
5645 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5646 dividend to 32 bits. Adjust length.
5647 (udivmodhi4): Cleanup output template. Fix length.
5649 2018-06-25 Carl Love <cel@us.ibm.com>
5651 * config/rs6000/vsx.md: Change word selector to prefered location.
5653 2018-06-25 Richard Biener <rguenther@suse.de>
5655 PR tree-optimization/86304
5656 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5657 epilogue-if-converted loops as well.
5659 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5661 * lto-section-out.c (lto_begin_section): Do not print section
5662 name for noaddr and unnumbered dumps.
5664 2018-06-25 Richard Biener <rguenther@suse.de>
5666 * tree-vectorizer.h (struct vec_info_shared): New structure
5667 with parts split out from struct vec_info and loop_nest from
5668 struct _loop_vec_info.
5669 (struct vec_info): Adjust accordingly.
5670 (struct _loop_vec_info): Likewise.
5671 (LOOP_VINFO_LOOP_NEST): Adjust.
5672 (LOOP_VINFO_DATAREFS): Likewise.
5673 (LOOP_VINFO_DDRS): Likewise.
5674 (struct _bb_vec_info): Likewise.
5675 (BB_VINFO_DATAREFS): Likewise.
5676 (BB_VINFO_DDRS): Likewise.
5677 (struct _stmt_vec_info): Add dr_aux member.
5678 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5679 (DR_MISALIGNMENT_UNINITIALIZED): New.
5680 (set_dr_misalignment): Adjust.
5681 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5682 (vect_analyze_loop): Adjust prototype.
5683 (vect_analyze_loop_form): Likewise.
5684 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5685 Compute dependences lazily.
5686 (vect_record_base_alignments): Use shared datarefs/ddrs.
5687 (vect_verify_datarefs_alignment): Likewise.
5688 (vect_analyze_data_refs_alignment): Likewise.
5689 (vect_analyze_data_ref_accesses): Likewise.
5690 (vect_analyze_data_refs): Likewise.
5691 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5692 constructor parameter for shared part.
5693 (vect_analyze_loop_form): Pass in shared part and adjust.
5694 (vect_analyze_loop_2): Pass in storage for the number of
5695 stmts. Move loop nest finding to the caller. Compute
5697 (vect_analyze_loop): Pass in shared part.
5698 (vect_transform_loop): Verify shared datarefs are unchanged.
5699 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5700 constructor parameter for shared part.
5701 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5702 (vect_slp_bb): Verify shared datarefs are unchanged before
5704 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5706 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5707 DR_MISALIGNMENT_UNINITIALIZED.
5708 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5709 parameter for shared part.
5710 (vec_info::~vec_info): Adjust.
5711 (vec_info_shared::vec_info_shared): New.
5712 (vec_info_shared::~vec_info_shared): Likewise.
5713 (vec_info_shared::save_datarefs): Likewise.
5714 (vec_info_shared::check_datarefs): Likewise.
5715 (try_vectorize_loop_1): Construct shared part live for analyses
5716 of a single loop for multiple vector sizes.
5717 * tree-parloops.c (gather_scalar_reductions): Adjust.
5719 2018-06-25 Richard Biener <rguenther@suse.de>
5721 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5722 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5723 (vect_analyze_data_refs): Remove similar code from here and
5724 simplify accordingly.
5726 2018-06-25 Richard Biener <rguenther@suse.de>
5728 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5729 for reverse storage order accesses rather than asserting
5730 they cannot happen here.
5732 2018-06-25 Tom de Vries <tdevries@suse.de>
5735 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5736 Use data16 instead of .byte for insn prefix.
5738 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5741 * parser.c (make_char_string_pack): Pass this literal chars
5742 through cpp_interpret_string.
5743 (cp_parser_userdef_numeric_literal): Check the result of
5744 make_char_string_pack.
5746 2018-06-24 Maya Rashish <coypu@sdf.org>
5748 * ginclude/stddef.h: Simplify conditions around avoiding
5749 re-definition of __size_t.
5751 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5753 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5754 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5756 2018-06-22 Maya Rashish <coypu@sdf.org>
5758 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5759 made to OpenBSD and NetBSD through the years.
5761 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5763 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5764 behavior of vec_pack (vector double, vector double) to match
5765 behavior of vec_float2 (vector double, vector double).
5767 2018-06-22 Olivier Hainque <hainque@adacore.com>
5769 * gimplify.c (gimplify_function_tree): Prevent creation
5770 of a trampoline for the address of the current function
5771 passed to entry/exit instrumentation hooks.
5773 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5776 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5779 2018-06-22 Martin Liska <mliska@suse.cz>
5781 PR tree-optimization/86263
5782 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5783 Bail out if is_enabled is false.
5784 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5786 (jump_table_cluster::is_enabled): New function.
5788 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5790 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5791 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5792 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5793 (lto_input_ts_binfo_tree_pointers): Likewise.
5794 * tree-streamer-out.c (streamer_write_tree_bitfields,
5795 write_ts_binfo_tree_pointers): Likewise.
5796 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5798 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5800 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5802 2018-06-22 Martin Liska <mliska@suse.cz>
5804 * symbol-summary.h (get): Make it pure and inline move
5805 functionality from ::get function.
5806 (get): Remove and inline into ::get and ::get_create.
5807 (get_create): Move code from ::get function.
5809 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5812 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5813 -x assembler-with-cpp.
5815 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5817 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5818 _FILE_OFFSET_BITS=64 for C++.
5820 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5822 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5823 conversion insn that shows up when pr85657-3.c is compiled using
5824 IEEE 128-bit long double.
5825 (neg<mode>2_internal): Use the correct mode to check whether the
5826 mode is IBM extended.
5827 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5828 multiply and divide external functions from being created more
5831 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5833 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5835 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5836 the edge can be forwarded.
5837 (cfg_layout_merge_blocks): Likewise.
5839 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5841 * except.c (finish_eh_generation): Commit edge insertions only after
5842 the EH edges have been redirected from post-landing to landing pads.
5844 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5846 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5847 create_tmp_var_for to create the FRAME decl.
5848 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5850 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5852 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5853 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5854 debug statement when resetting its value.
5855 (expand_call_inline): Copy the locus of the call onto the assignment
5856 of the return value, if any. Use local variable in more cases.
5858 2018-06-21 Martin Liska <mliska@suse.cz>
5860 * ipa-pure-const.c (propagate_nothrow): Use
5861 funct_state_summaries->get.
5862 (dump_malloc_lattice): Likewise.
5863 (propagate_malloc): Likewise.
5865 2018-06-21 Richard Biener <rguenther@suse.de>
5867 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5868 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5869 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5871 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5872 BLOCK_ABSTRACT_ORIGIN unconditionally.
5874 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5876 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5878 * ipa-reference.c (ipa_reference_c_finalize): Delete
5879 ipa_ref_opt_sum_summaries and set it to NULL.
5881 2018-06-21 Tom de Vries <tdevries@suse.de>
5883 PR tree-optimization/85859
5884 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5885 test with comment from bb_no_side_effects_p.
5887 2018-06-21 Richard Biener <rguenther@suse.de>
5889 PR tree-optimization/86232
5890 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5891 max for constant niter.
5893 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5895 * config/aarch64/aarch64-simd.md
5896 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5898 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5900 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5901 Make opernads of the unspec commutative.
5903 2018-06-21 Richard Biener <rguenther@suse.de>
5905 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5906 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5907 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5908 (vect_analyze_data_ref_dependence): Re-order checks to deal with
5910 (vect_record_base_alignments): Do not record base alignment
5911 for gathers or scatters.
5912 (vect_compute_data_ref_alignment): Drop return value that is always
5913 true. Bail out early for gathers or scatters.
5914 (vect_enhance_data_refs_alignment): Bail out early for gathers
5916 (vect_find_same_alignment_drs): Likewise.
5917 (vect_analyze_data_refs_alignment): Remove dead code.
5918 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5919 (vect_analyze_data_refs): For possible gathers or scatters do
5920 not create an alternate DR, just check their possible validity
5921 and mark them. Adjust DECL_NONALIASED handling to not rely
5923 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5924 update inits of gathers or scatters.
5925 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5926 Also copy gather/scatter flag to pattern vinfo.
5928 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
5930 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5931 behavior of vec_packsu (vector unsigned long long, vector unsigned
5932 long long) to match behavior of vec_packs with same signature.
5934 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
5935 Thomas Schwinge <thomas@codesourcery.com>
5936 Cesar Philippidis <cesar@codesourcery.com>
5938 * gimplify.c (gimplify_scan_omp_clauses): Add support for
5939 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5940 (gimplify_adjust_omp_clauses): Likewise.
5941 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5942 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5943 (gimplify_omp_target_update): Update handling of acc update and
5945 * omp-low.c (install_var_field): Remove unused parameter
5946 base_pointers_restrict.
5947 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5948 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5950 (omp_target_base_pointers_restrict_p): Delete.
5951 (scan_omp_target): Update call to scan_sharing_clauses.
5952 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5954 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5955 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5956 (convert_local_omp_clauses): Likewise.
5957 * tree-pretty-print.c (dump_omp_clause): Likewise.
5958 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
5960 (omp_clause_code_name): Likewise.
5962 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5965 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5968 PR tree-optimization/86231
5969 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5970 anti-range don't overwrite *vr0min before using it to compute *vr0max.
5972 2018-06-20 Tom de Vries <tdevries@suse.de>
5974 PR tree-optimization/86097
5975 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5976 iv type if signedness of iv type is not the same as that of *nit.
5978 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5980 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
5981 EDGE_EH edges, verify they are all EDGE_EH.
5983 2018-06-20 Maya Rashish <coypu@sdf.org>
5985 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5987 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5989 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5990 * config/aarch64/aarch64.c (xgene1_tunings): Add
5991 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5992 (aarch64_mode_valid_for_sched_fusion_p):
5993 Allow 16-byte modes.
5994 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5995 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5997 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5999 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6000 * config/aarch64/iterators.md (VQ2): New mode iterator.
6002 2018-06-20 Martin Liska <mliska@suse.cz>
6004 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6005 Change default ratio from 10 to 8.
6007 2018-06-20 Martin Liska <mliska@suse.cz>
6009 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6011 (bit_test_cluster::find_bit_tests): Likewise.
6012 (switch_decision_tree::analyze_switch_statement): Find clusters.
6013 * tree-switch-conversion.h (struct jump_table_cluster): Document
6016 2018-06-20 Martin Liska <mliska@suse.cz>
6018 * tree-switch-conversion.c (switch_conversion::collect):
6019 Record m_uniq property.
6020 (switch_conversion::expand): Bail out for special conditions.
6021 (group_cluster::~group_cluster): New.
6022 (group_cluster::group_cluster): Likewise.
6023 (group_cluster::dump): Likewise.
6024 (jump_table_cluster::emit): New.
6025 (switch_decision_tree::fix_phi_operands_for_edges): New.
6026 (struct case_node): Remove struct.
6027 (jump_table_cluster::can_be_handled): New.
6028 (case_values_threshold): Moved to header.
6029 (reset_out_edges_aux): Likewise.
6030 (jump_table_cluster::is_beneficial): New.
6031 (bit_test_cluster::can_be_handled): Likewise.
6032 (add_case_node): Remove.
6033 (bit_test_cluster::is_beneficial): New.
6034 (case_bit_test::cmp): New.
6035 (bit_test_cluster::emit): New.
6036 (expand_switch_as_decision_tree_p): Remove.
6037 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6038 (fix_phi_operands_for_edge): Likewise.
6039 (switch_decision_tree::analyze_switch_statement): New.
6040 (compute_cases_per_edge): Move ...
6041 (switch_decision_tree::compute_cases_per_edge): ... here.
6042 (try_switch_expansion): Likewise.
6043 (switch_decision_tree::try_switch_expansion): Likewise.
6044 (record_phi_operand_mapping): Likewise.
6045 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6046 (emit_case_decision_tree): Likewise.
6047 (switch_decision_tree::emit): Likewise.
6048 (balance_case_nodes): Likewise.
6049 (switch_decision_tree::balance_case_nodes): Likewise.
6050 (dump_case_nodes): Likewise.
6051 (switch_decision_tree::dump_case_nodes): Likewise.
6052 (emit_jump): Likewise.
6053 (switch_decision_tree::emit_jump): Likewise.
6054 (emit_cmp_and_jump_insns): Likewise.
6055 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6056 (emit_case_nodes): Likewise.
6057 (switch_decision_tree::emit_case_nodes): Likewise.
6058 (conditional_probability): Remove.
6059 * tree-switch-conversion.h (enum cluster_type): New.
6061 (struct cluster): Likewise.
6062 (cluster::cluster): Likewise.
6063 (struct simple_cluster): Likewise.
6064 (simple_cluster::simple_cluster): Likewise.
6065 (struct group_cluster): Likewise.
6066 (struct jump_table_cluster): Likewise.
6067 (struct bit_test_cluster): Likewise.
6068 (struct min_cluster_item): Likewise.
6069 (struct case_tree_node): Likewise.
6070 (case_tree_node::case_tree_node): Likewise.
6071 (jump_table_cluster::case_values_threshold): Likewise.
6072 (struct case_bit_test): Likewise.
6073 (struct switch_decision_tree): Likewise.
6074 (struct switch_conversion): Likewise.
6075 (switch_decision_tree::reset_out_edges_aux): Likewise.
6077 2018-06-20 Martin Liska <mliska@suse.cz>
6079 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6080 (hoist_edge_and_branch_if_true): Likewise.
6081 (expand_switch_using_bit_tests_p): Likewise.
6082 (struct case_bit_test): Likewise.
6083 (case_bit_test_cmp): Likewise.
6084 (emit_case_bit_tests): Likewise.
6085 (switch_conversion::switch_conversion): New class.
6086 (struct switch_conv_info): Remove old struct.
6087 (collect_switch_conv_info): More to ...
6088 (switch_conversion::collect): ... this.
6089 (check_range): Likewise.
6090 (switch_conversion::check_range): Likewise.
6091 (check_all_empty_except_final): Likewise.
6092 (switch_conversion::check_all_empty_except_final): Likewise.
6093 (check_final_bb): Likewise.
6094 (switch_conversion::check_final_bb): Likewise.
6095 (create_temp_arrays): Likewise.
6096 (switch_conversion::create_temp_arrays): Likewise.
6097 (free_temp_arrays): Likewise.
6098 (gather_default_values): Likewise.
6099 (switch_conversion::gather_default_values): Likewise.
6100 (build_constructors): Likewise.
6101 (switch_conversion::build_constructors): Likewise.
6102 (constructor_contains_same_values_p): Likewise.
6103 (switch_conversion::contains_same_values_p): Likewise.
6104 (array_value_type): Likewise.
6105 (switch_conversion::array_value_type): Likewise.
6106 (build_one_array): Likewise.
6107 (switch_conversion::build_one_array): Likewise.
6108 (build_arrays): Likewise.
6109 (switch_conversion::build_arrays): Likewise.
6110 (gen_def_assigns): Likewise.
6111 (switch_conversion::gen_def_assigns): Likewise.
6112 (prune_bbs): Likewise.
6113 (switch_conversion::prune_bbs): Likewise.
6114 (fix_phi_nodes): Likewise.
6115 (switch_conversion::fix_phi_nodes): Likewise.
6116 (gen_inbound_check): Likewise.
6117 (switch_conversion::gen_inbound_check): Likewise.
6118 (process_switch): Use the newly created class.
6119 (switch_conversion::expand): New.
6120 (switch_conversion::~switch_conversion): New.
6121 * tree-switch-conversion.h: New file.
6123 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6125 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6126 tree-vect-patterns.c.
6127 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6128 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6129 (vect_recog_sad_pattern): Likewise.
6130 (vect_recog_widen_sum_pattern): Likewise.
6131 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6132 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6133 (vect_recog_rotate_pattern): Likewise.
6134 (vect_recog_mult_pattern): Likewise.
6135 (vect_recog_vector_vector_shift_pattern): Likewise.
6136 (vect_recog_divmod_pattern): Likewise.
6137 (vect_recog_mixed_size_cond_pattern): Likewise.
6138 (vect_recog_bool_pattern): Likewise.
6139 (vect_recog_mask_conversion_pattern): Likewise.
6140 (vect_try_gather_scatter_pattern): Likewise.
6141 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6142 (vect_recog_over_widening_pattern): Likewise.
6143 (vect_recog_gather_scatter_pattern): Likewise.
6144 (vect_recog_func_ptr): Move from tree-vectorizer.h
6145 (vect_vect_recog_func_ptrs): Move further down the file.
6146 (vect_recog_func): Likewise. Remove the third argument.
6147 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6148 (vect_pattern_recog_1): Expect the pattern function to do any
6149 necessary target tests. Also expect it to provide a vector type.
6150 Remove the type_in handling.
6152 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6154 * tree-vect-patterns.c (vect_pattern_detected): New function.
6155 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6156 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6157 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6158 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6159 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6160 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6161 (vect_recog_mask_conversion_pattern)
6162 (vect_try_gather_scatter_pattern): Likewise.
6164 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6166 * tree-vect-patterns.c (vect_get_internal_def): New function.
6167 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6168 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6169 (search_type_for_mask_1): Use it.
6171 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6173 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6174 redundant WIDEN_SUM_EXPR handling.
6175 (vect_recog_sad_pattern): Likewise.
6177 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6179 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6180 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6181 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6182 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6183 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6185 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6187 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6188 use the stmt_vec_info of the original bb statement for the
6189 new zero assignment, even if the call is part of a pattern.
6191 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6193 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6194 that the sequence is attached to the original statement rather
6195 than the pattern statement.
6196 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6197 PATTERN_DEF_SEQ from the original statement rather than
6198 the main pattern statement.
6199 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6200 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6201 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6203 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6205 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6206 definition statements before the early exit for statements that aren't
6208 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6210 (vect_transform_loop): ...here. Process pattern definition
6211 statements without first checking whether the main pattern
6212 statement is live or relevant.
6214 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6216 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6217 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6219 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6221 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6222 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6223 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6224 (expand_block_compare): Change select_block_compare_mode call.
6225 (expand_strncmp_align_check): Use new functions, fix comment.
6226 (emit_final_str_compare_gpr): New function.
6227 (expand_strn_compare): Refactor and clean up code.
6228 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6230 2018-06-19 Tony Reix <tony.reix@atos.com>
6231 Damien Bergamini <damien.bergamini@atos.com>
6232 David Edelsohn <dje.gcc@gmail.com>
6234 * collect2.c (static_obj): New variable.
6235 (static_libs): New variable.
6236 (is_in_list): Uncomment declaration.
6237 (main): Track AIX libraries linked statically.
6238 (is_in_list): Uncomment definition.
6239 (scan_prog_file): Don't add AIX shared libraries initializer
6240 to constructor list if linking statically.
6242 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6244 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6246 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6248 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6250 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6253 2018-06-19 Martin Liska <mliska@suse.cz>
6255 * config/i386/i386.c (ix86_can_inline_p): Do not use
6256 ipa_fn_summaries::get_create.
6257 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6259 (devirtualization_time_bonus): Likewise.
6260 (ipcp_propagate_stage): Likewise.
6261 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6262 (edge_set_predicate): Likewise.
6263 (evaluate_conditions_for_known_args): Likewise.
6264 (evaluate_properties_for_edge): Likewise.
6265 (ipa_call_summary::reset): Tranform to ...
6266 (ipa_call_summary::~ipa_call_summary): ... this.
6267 (ipa_fn_summary::reset): Transform to ...
6268 (ipa_fn_summary::~ipa_fn_summary): ... this.
6269 (ipa_fn_summary_t::remove): Rename to ...
6270 (ipa_fn_summary_t::remove_callees): ... this.
6271 (ipa_fn_summary_t::duplicate): Use placement new
6272 instead of memory copy.
6273 (ipa_call_summary_t::duplicate): Likewise.
6274 (ipa_call_summary_t::remove): Remove.
6275 (dump_ipa_call_summary): Change get_create to get.
6276 (ipa_dump_fn_summary): Dump only when summary exists.
6277 (analyze_function_body): Use symbol_summary::get instead
6279 (compute_fn_summary): Likewise.
6280 (estimate_edge_devirt_benefit): Likewise.
6281 (estimate_edge_size_and_time): Likewise.
6282 (inline_update_callee_summaries): Likewise.
6283 (remap_edge_change_prob): Likewise.
6284 (remap_edge_summaries): Likewise.
6285 (ipa_merge_fn_summary_after_inlining): Likewise.
6286 (write_ipa_call_summary): Likewise.
6287 (ipa_fn_summary_write): Likewise.
6288 (ipa_free_fn_summary): Likewise.
6289 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6290 (struct ipa_call_summary): Likewise.
6291 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6293 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6294 (estimate_size_after_inlining): Likewise.
6295 (estimate_growth): Likewise.
6296 (growth_likely_positive): Likewise.
6297 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6298 (inline_call): Likewise.
6299 * ipa-inline.c (caller_growth_limits): Likewise.
6300 (can_inline_edge_p): Likewise.
6301 (can_inline_edge_by_limits_p): Likewise.
6302 (compute_uninlined_call_time): Likewise.
6303 (compute_inlined_call_time): Likewise.
6304 (want_inline_small_function_p): Likewise.
6305 (edge_badness): Likewise.
6306 (update_caller_keys): Likewise.
6307 (update_callee_keys): Likewise.
6308 (inline_small_functions): Likewise.
6309 (inline_to_all_callers_1): Likewise.
6310 (dump_overall_stats): Likewise.
6311 (early_inline_small_functions): Likewise.
6312 (early_inliner): Likewise.
6313 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6314 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6315 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6316 * ipa-split.c (execute_split_functions): Likewise.
6317 * symbol-summary.h: Likewise.
6318 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6320 2018-06-19 Richard Biener <rguenther@suse.de>
6322 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6323 (vectorize_loops): ... here. Fix dbgcnt handling.
6324 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6326 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6329 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6330 ieee128 argument takes up only one (vector) register, not two (floating
6333 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6335 * gimplify.c (gimplify_init_constructor): Really never clear for an
6336 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6338 2018-06-19 Richard Biener <rguenther@suse.de>
6340 PR tree-optimization/86179
6341 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6342 after failed recognition.
6344 2018-06-18 Martin Sebor <msebor@redhat.com>
6347 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6348 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6349 Handle integer subtraction.
6350 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6351 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6353 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6355 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6356 param from rtx to rtx_insn *.
6357 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6359 (frv_ifcvt_modify_insn): Likwise.
6360 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6361 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6362 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6363 as_a <rtx_insn *> cast to local "unprotected_region" once
6364 it's been established that it's not NULL or pc_rtx.
6365 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6366 param "sethi" from rtx to rtx_insn *.
6367 (nds32_group_float_insns): Likewise for param "insn".
6368 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6370 (vax_output_int_subtract): Likewise.
6371 * config/vax/vax.c (vax_output_int_add): Likewise for param
6373 (vax_output_int_subtract): Likewise.
6374 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6375 (emit_pattern_after): Likewise for param "after".
6376 (emit_insn_after): Likewise.
6377 (emit_jump_insn_after): Likewise.
6378 (emit_call_insn_after): Likewise.
6379 (emit_debug_insn_after): Likewise.
6380 (emit_pattern_before): Likewise for param "before".
6381 (emit_insn_before): Likewise.
6382 (emit_jump_insn_before): Likewise.
6383 * final.c (get_insn_template): Likewise for param "insn", removing
6385 * output.h (get_insn_template): Likewise for 2nd param.
6386 * rtl.h (emit_insn_before): Likewise.
6387 (emit_jump_insn_before): Likewise.
6388 (emit_debug_insn_before_noloc): Likewise.
6389 (emit_insn_after): Likewise.
6390 (emit_jump_insn_after): Likewise.
6391 (emit_call_insn_after): Likewise.
6392 (emit_debug_insn_after): Likewise.
6393 (set_insn_deleted): Likewise for param.
6395 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6398 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6399 floating point modes, so that IFmode is numerically greater than
6400 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6401 to declare the ordering. This prevents IFmode from being
6402 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6403 machine. Include rs6000-modes.h to share the fractional values
6404 between genmodes* and the rest of the compiler.
6408 * config/rs6000/rs6000-modes.h: New file.
6409 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6410 meaning of rs6000_long_double_size so that 126..128 selects an
6411 appropriate 128-bit floating point type.
6412 (rs6000_option_override_internal): Likewise.
6413 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6414 (TARGET_LONG_DOUBLE_128): Change the meaning of
6415 rs6000_long_double_size so that 126..128 selects an appropriate
6416 128-bit floating point type.
6417 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6418 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6419 source and destination to match the standard usage.
6420 (truncifkf2): Likewise.
6421 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6422 ISA 2.07 to use an explicit clobber, instead of passing in a
6424 (copysign<mode>3_soft): Likewise.
6426 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6428 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6429 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6430 (vect_slp_analyze_instance_dependence): Likewise.
6431 (vect_enhance_data_refs_alignment): Likewise.
6432 (vect_analyze_data_refs_alignment): Likewise.
6433 (vect_slp_analyze_and_verify_instance_alignment
6434 (vect_analyze_data_ref_accesses): Likewise.
6435 (vect_prune_runtime_alias_test_list): Likewise.
6436 (vect_analyze_data_refs): Likewise.
6437 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6438 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6439 (vect_analyze_scalar_cycles_1): Likewise.
6440 (vect_get_loop_niters): Likewise.
6441 (vect_analyze_loop_form_1): Likewise.
6442 (vect_update_vf_for_slp): Likewise.
6443 (vect_analyze_loop_operations): Likewise.
6444 (vect_analyze_loop): Likewise.
6445 (vectorizable_induction): Likewise.
6446 (vect_transform_loop): Likewise.
6447 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6448 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6449 (vect_make_slp_decision): Likewise.
6450 (vect_detect_hybrid_slp): Likewise.
6451 (vect_slp_analyze_operations): Likewise.
6452 (vect_slp_bb): Likewise.
6453 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6454 (vectorizable_bswap): Likewise.
6455 (vectorizable_call): Likewise.
6456 (vectorizable_simd_clone_call): Likewise.
6457 (vectorizable_conversion): Likewise.
6458 (vectorizable_assignment): Likewise.
6459 (vectorizable_shift): Likewise.
6460 (vectorizable_operation): Likewise.
6461 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6463 2018-06-18 Martin Sebor <msebor@redhat.com>
6465 PR tree-optimization/81384
6466 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6467 * builtins.c (expand_builtin_strnlen): New function.
6468 (expand_builtin): Call it.
6469 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6470 * builtins.def (BUILT_IN_STRNLEN): New.
6471 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6472 Warn for bounds in excess of maximum object size.
6473 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6474 single-value ranges. Handle strnlen.
6475 (handle_builtin_strlen): Handle strnlen.
6476 (strlen_check_and_optimize_stmt): Same.
6477 * doc/extend.texi (Other Builtins): Document strnlen.
6479 2018-06-18 Maya Rashish <coypu@sdf.org>
6481 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6482 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6483 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6485 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6487 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6489 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6491 * tree.c (escaped_string::escape): Replace cast to char * by
6492 const_cast<char *> (unescaped).
6494 2018-06-18 Nick Clifton <nickc@redhat.com>
6497 * tree.c (escaped_string): New class. Converts an unescaped
6498 string into its escaped equivalent.
6499 (warn_deprecated_use): Use the new class to convert the
6500 deprecation message, if present.
6501 (test_escaped_strings): New self test.
6502 (test_c_tests): Add test_escaped_strings.
6503 * doc/extend.texi (deprecated): Add a note that the
6504 deprecation message is affected by the -fmessage-length
6505 option, and that control characters will be escaped.
6506 (#pragma GCC error): Document this pragma.
6507 (#pragma GCC warning): Likewise.
6508 * doc/invoke.texi (-fmessage-length): Document this option's
6509 effect on the #warning and #error preprocessor directives and
6510 the deprecated attribute.
6512 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6514 * tree.c (decl_value_expr_lookup): Revert latest change.
6515 (decl_value_expr_insert): Likewise.
6517 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6519 * gimplify.c (nonlocal_vlas): Delete.
6520 (nonlocal_vla_vars): Likewise.
6521 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6523 (gimplify_body): Do not create and destroy nonlocal_vlas.
6524 * tree-nested.c: Include diagnostic.h.
6525 (use_pointer_in_frame): Tweak.
6526 (lookup_field_for_decl): Add assertion and declare the transformation.
6527 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6528 internal error when the reference is in a wrong context. Do not
6529 create a debug decl by default.
6530 (note_nonlocal_block_vlas): Delete.
6531 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6532 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6533 create a debug decl by default.
6534 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6535 call is in a wrong context.
6536 (fixup_vla_decls): New function.
6537 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6538 debug variables were created.
6539 * tree.c (decl_value_expr_lookup): Add checking assertion.
6540 (decl_value_expr_insert): Likewise.
6542 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6545 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6546 * tree-scalar-evolution.c (interpret_expr): Likewise.
6547 (expression_expensive_p): Likewise.
6548 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6549 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6550 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6551 (ssa_defined_by_minus_one_stmt_p): New.
6553 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6556 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6557 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6558 * dojump.c (do_jump): Likewise.
6559 * expr.c (expand_expr_real_2): Check operand type's sign.
6560 * fold-const.c (const_unop): Handle ABSU_EXPR.
6561 (fold_abs_const): Likewise.
6562 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6563 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6564 (strip_sign_op_1): Likesise.
6565 * match.pd: Add new pattern to generate ABSU_EXPR.
6566 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6567 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6568 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6569 * tree-inline.c (estimate_operator_cost): Likewise.
6570 * tree-pretty-print.c (dump_generic_node): Likewise.
6571 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6572 * tree.def (ABSU_EXPR): New.
6574 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6577 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6578 documented as preserved for backward compatibility only.
6579 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6581 PR rtl-optimization/86108
6582 * bb-reorder.c (create_forwarder_block): Renamed to ...
6583 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6584 jump from new landing pad to the second part.
6585 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6588 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6591 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6592 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6593 Only call store_expr for halves if the mode is the same.
6596 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6597 Fix up comment formatting.
6599 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6601 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6602 (typed_splay_tree::closure,
6603 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6604 (typed_splay_tree::typed_splay_tree,
6605 typed_splay_tree::operator =): Declared private.
6606 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6607 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6608 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6609 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6610 typed_splay_tree::splay_tree_splay,
6611 typed_splay_tree::splay_tree_foreach_helper,
6612 typed_splay_tree::splay_tree_insert,
6613 typed_splay_tree::splay_tree_remove,
6614 typed_splay_tree::splay_tree_lookup,
6615 typed_splay_tree::splay_tree_predecessor,
6616 typed_splay_tree::splay_tree_successor,
6617 typed_splay_tree::splay_tree_min,
6618 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6619 (typed_splay_tree::root, typed_splay_tree::comp,
6620 typed_splay_tree::delete_key,
6621 typed_splay_tree::delete_value): New data members.
6622 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6623 typed_splay_tree::remove.
6625 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6627 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6628 -mginv and -mno-ginv to the assembler.
6629 * config/mips/mips.opt (-mcrc): New option.
6631 * doc/invoke.text (-mcrc): Document.
6634 2018-06-15 Nick Clifton <nickc@redhat.com>
6637 * tree.c (escaped_string): New class. Converts an unescaped
6638 string into its escaped equivalent.
6639 (warn_deprecated_use): Use the new class to convert the
6640 deprecation message, if present.
6641 (test_escaped_strings): New self test.
6642 (test_c_tests): Add test_escaped_strings.
6643 * doc/extend.texi (deprecated): Add a note that the
6644 deprecation message is affected by the -fmessage-length
6645 option, and that control characters will be escaped.
6646 (#pragma GCC error): Document this pragma.
6647 (#pragma GCC warning): Likewise.
6648 * doc/invoke.texi (-fmessage-length): Document this option's
6649 effect on the #warning and #error preprocessor directives and
6650 the deprecated attribute.
6652 2018-06-15 Richard Biener <rguenther@suse.de>
6654 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6655 here, also noting vector size used.
6656 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6657 size used in MSG_OPTIMIZED_LOCATIONS dump.
6658 (pass_slp_vectorize::execute): Adjust.
6660 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6663 * config/arc/arc.c (arc_return_address_register): Fix
6666 2018-06-15 Richard Biener <rguenther@suse.de>
6669 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6670 leave useless conversion stripping to force_gimple_operand_gsi.
6671 (gimplify_build2): Likewise.
6672 (gimplify_build1): Likewise.
6674 2018-06-15 Richard Biener <rguenther@suse.de>
6677 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6678 before adjusting their block.
6680 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6682 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6683 multilibs for *-*-rtems*.
6684 * config/riscv/t-rtems: New file.
6686 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6689 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6690 unsigned_type_for returns NULL.
6693 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6694 subregs of multi-word pseudos unless the float mode has word size.
6696 2018-06-14 Richard Biener <rguenther@suse.de>
6699 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6700 duplicate and harmful type_hash_canon.
6701 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6703 2018-06-14 Richard Biener <rguenther@suse.de>
6706 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6709 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6711 * config/rtems.h (STDINT_LONG32): Define.
6713 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6714 Prachi Godbole <prachi.godbole@imgtec.com>
6716 * config/mips/mips-cpus.def: Define P6600.
6717 * config/mips/mips-tables.opt: Regenerate.
6718 * config/mips/mips.c (mips_ucbranch_type): New enum.
6719 (mips_rtx_cost_data): Add support for P6600.
6720 (mips_issue_rate): Likewise.
6721 (mips_multipass_dfa_lookahead): Likewise.
6722 (mips_avoid_hazard): Likewise.
6723 (mips_reorg_process_insns): Likewise.
6724 (mips_classify_branch_p6600): New function.
6725 * config/mips/mips.h (TUNE_P6600): New define.
6726 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6727 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6728 * config/mips/mips.md: Include p6600.md.
6729 (processor): Add p6600.
6730 * config/mips/p6600.md: New file.
6731 * doc/invoke.texi: Add p6600 to supported architectures.
6733 2018-06-13 Martin Sebor <msebor@redhat.com>
6735 PR tree-optimization/86114
6736 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6738 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6740 2018-06-13 Richard Biener <rguenther@suse.de>
6742 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6743 Properly set vector type of the intermediate stmt.
6744 * tree-vect-stmts.c (vectorizable_operation): The destination
6745 var always has vectype_out type.
6747 2018-06-13 Jeff Law <law@redhat.com>
6749 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6750 integer 0 for argument to print_rtl_with_bb.
6751 (rl78_reorg): Likewise.
6753 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6755 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6756 from rtx to rtx_insn *.
6757 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6759 (add_sched_insns_for_speculation): Likewise for local "target",
6760 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6761 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6762 from rtx_insn ** to rtx_code_label **.
6763 (reorg_emit_nops): Likewise.
6764 (c6x_reorg): Likewise for local "call_labels".
6765 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6767 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6768 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6769 the loops over LABEL_REFS.
6770 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6772 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6773 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6774 (split_branches): Strengthen local "olabel" from rtx to
6775 rtx_insn *, adding a safe_as_a cast.
6776 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6778 (add_insn_after): Likewise for first two params.
6779 (add_insn_before): Likewise.
6780 (remove_insn): Likewise for param.
6781 (emit_pattern_before_noloc): Likewise for second and third params.
6782 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6783 (emit_call_insn_before_noloc): Likewise.
6784 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6786 (emit_barrier_before): Likewise.
6787 (emit_label_before): Strengthen "label" param from "rtx" to
6788 "rtx_code_label *". Strengthen "before" param from "rtx" to
6790 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6792 (emit_pattern_after_noloc): Likewise.
6793 (emit_insn_after_noloc): Likewise.
6794 (emit_jump_insn_after_noloc): Likewise.
6795 (emit_call_insn_after_noloc): Likewise.
6796 (emit_debug_insn_after_noloc): Likewise.
6797 (emit_barrier_after): Likewise.
6798 (emit_label_after): Likewise for both params.
6799 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6800 "loc" param from "int" to "location_t".
6801 (emit_insn_after_setloc): Likewise.
6802 (emit_jump_insn_after_setloc): Likewise.
6803 (emit_call_insn_after_setloc): Likewise.
6804 (emit_debug_insn_after_setloc): Likewise.
6805 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6806 "loc" param from "int" to "location_t".
6807 (emit_pattern_before): Convert NULL_RTX to NULL.
6808 (emit_insn_before_setloc): Convert "loc" param from "int" to
6810 (emit_jump_insn_before_setloc): Likewise.
6811 (emit_call_insn_before_setloc): Likewise.
6812 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6813 rtx_insn *. Convert "loc" param from "int" to "location_t".
6814 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6815 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6816 Convert 3rd param from "int" to "location_t".
6817 (emit_barrier_before, emit_barrier_after, next_real_insn):
6818 Strengthen param from rtx to rtx_insn *.
6819 (emit_label_before): Strengthen 1st param from "rtx" to
6820 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6822 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6823 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6824 Strengthen 2nd param from "rtx" to "rtx_insn *".
6825 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6826 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6827 Likewise. Convert 3rd param from "int" to "location_t".
6828 (emit_label_after): Strengthen 1st param from "rtx" to
6830 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6832 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6833 from "rtx" to "rtx_insn *".
6835 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6837 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6838 bodies streamed in with -Q.
6839 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6840 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6841 * gimple-streamer-out.c: Include gimple-pretty-print.h
6842 (output_bb): Dump stmts streamed.
6843 * lto-section-out.c: Include print-tree.h
6844 (lto_begin_section): Dump sections created.
6845 (lto_output_decl_index): Dump decl encoded.
6846 * lto-streamer-out.c: Include print-tree.h
6847 (create_output_block): Dump output block created.
6848 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6849 (output_function): Dump function output.
6850 (output_constructor): Dump constructor streamed.
6851 (write_global_stream): Output indexes encoded.
6852 (produce_asm_for_decls): Dump streams encoded.
6853 * lto-streamer.c (streamer_dump_file): New global var.
6854 * lto-streamer.h (streamer_dump_file): Declare.
6855 * passes.c (ipa_write_summaries): Initialize streamer dump.
6856 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6859 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6862 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6863 offsets for register save directives. Emit a second batch of save
6864 directives, if need be, when the function accesses prior frames.
6866 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6868 * config/arc/fpu.md (fmasf4): Force operand to register.
6869 (fnmasf4): Likewise.
6871 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6873 * config/arc/arc-protos.h (arc_pad_return): Remove.
6874 * config/arc/arc.c (machine_function): Remove force_short_suffix
6876 (arc_print_operand): Adjust printing of '&'.
6877 (arc_verify_short): Remove conditional printing of short suffix.
6878 (arc_final_prescan_insn): Remove reference to size_reason.
6879 (pad_return): New function.
6880 (arc_reorg): Call pad_return.
6881 (arc_pad_return): Remove.
6882 (arc_init_machine_status): Remove reference to force_short_suffix.
6883 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6884 (attr length): When attribute iscompact is true force to 2
6885 regardless; in the case of maybe check if we want to force the
6886 instruction to have 4 bytes length.
6887 (nopv): Change it to generate 4 byte long nop as well.
6888 (blockage): New pattern.
6889 (simple_return): Remove call to arc_pad_return.
6890 (p_return_i): Likewise.
6892 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6894 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6896 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6898 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6901 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6903 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6904 for ARC700 and ARCv2.
6906 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
6909 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6910 operands[2] instead of operands[1].
6913 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6915 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6916 case, check whether the outer register overlaps an unallocatable
6917 register, not just whether it fits the required class.
6919 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6921 * poly-int.h (can_div_trunc_p): Add new overload in which all values
6923 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6924 (memrefs_conflict_p): Likewise.
6925 (init_alias_analysis): Likewise.
6926 * cfgexpand.c (expand_debug_expr): Likewise.
6927 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6928 * cse.c (fold_rtx): Likewise.
6929 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6930 * expr.c (emit_block_move_hints): Likewise.
6931 (clear_storage_hints, push_block, emit_push_insn): Likewise.
6932 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6933 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6934 (emit_group_store): Likewise.
6935 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
6936 to read the PRE/POST_MODIFY increment.
6937 * calls.c (store_one_arg): Use strip_offset.
6938 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6940 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6942 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6943 (simplify_binary_operation_1): Extend CONST_INT handling to
6945 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6946 than a HOST_WIDE_INT.
6947 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6949 (adjust_mems, add_stores): Update accodingly.
6950 (vt_canonicalize_addr): Track polynomial offsets.
6951 (emit_note_insn_var_location): Likewise.
6952 (vt_add_function_parameter): Likewise.
6953 (vt_initialize): Likewise.
6955 2018-06-12 Jeff Law <law@redhat.com>
6957 * config.gcc (alpha*-*-freebsd*): Remove.
6958 * config/alpha/freebsd.h: Remove.
6960 2018-06-12 David Malcolm <dmalcolm@redhat.com>
6963 * spellcheck-tree.c (levenshtein_distance): Rename to...
6964 (get_edit_distance): ...this, and update for underlying renaming.
6965 * spellcheck-tree.h (levenshtein_distance): Rename to...
6966 (get_edit_distance): ...this.
6967 * spellcheck.c (levenshtein_distance): Rename to...
6968 (get_edit_distance): ...this. Convert from Levenshtein distance
6969 to Damerau-Levenshtein distance by supporting transpositions of
6970 adjacent characters. Rename "v1" to "v_next" and "v0" to
6972 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6973 (selftest::test_edit_distance_unit_test_oneway): ...this, and
6974 update for underlying renaming.
6975 (selftest::levenshtein_distance_unit_test): Rename to...
6976 (selftest::test_get_edit_distance_unit): ...this, and update for
6977 underlying renaming.
6978 (selftest::test_find_closest_string): Add example from PR 69968
6979 where transposition helps
6980 (selftest::test_metric_conditions): Update for renaming.
6981 (selftest::test_metric_conditions): Likewise.
6982 (selftest::spellcheck_c_tests): Likewise.
6983 * spellcheck.h (levenshtein_distance): Rename both overloads to...
6984 (get_edit_distance): ...this.
6985 (best_match::consider): Update for renaming.
6987 2018-06-12 Martin Sebor <msebor@redhat.com>
6989 PR tree-optimization/85259
6990 * builtins.c (compute_objsize): Handle constant offsets.
6991 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6992 true iff a warning has been issued.
6993 * gimple.h (gimple_nonartificial_location): New function.
6994 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6995 gimple_nonartificial_location and handle -Wno-system-headers.
6996 (handle_builtin_stxncpy): Same.
6998 2018-06-12 Martin Sebor <msebor@redhat.com>
7001 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7003 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7005 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7006 BUILTIN_VEC_XST entries for pointer to double and long long.
7008 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7011 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7013 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7016 2018-06-12 Martin Liska <mliska@suse.cz>
7018 * doc/options.texi: Document IntegerRange.
7020 2018-06-12 Martin Liska <mliska@suse.cz>
7022 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7023 * opt-functions.awk: Handle Deprecated flag.
7024 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7026 (read_cmdline_option): Report warning for a deprecated option.
7027 * opts.h (struct cl_option): Add new field cl_deprecated.
7028 (CL_ERR_DEPRECATED): New.
7030 2018-06-12 Martin Liska <mliska@suse.cz>
7032 * doc/options.texi: Document Deprecated option flag.
7034 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7036 * config/arc/arc-arch.h (arc_extras): New enum.
7037 (arc_cpu_t):Add field extra.
7038 (arc_cpu_types): Consider the extras.
7039 * config/arc/arc-cpus.def: Add extras info.
7040 * config/arc/arc-opts.h (processor_type): Consider extra field.
7041 * config/arc/arc.c (arc_override_options): Handle extra field.
7043 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7045 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7046 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7048 * config/arc/arc.c (arc_init): Update pic warning.
7049 (irq_range): Update irq range parsing warnings.
7050 (arc_override_options): Update various warning messages.
7051 (arc_handle_aux_attribute): Likewise.
7053 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7055 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7057 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7059 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7060 compiler messages directives.
7062 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7064 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7065 * config/mips/mips-tables.opt: Regenerate.
7066 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7068 * doc/invoke.texi: Document -march=i6500.
7070 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7072 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7073 (i6400_gpmul): Add cpu_unit.
7074 (i6400_gpdiv): Likewise.
7075 (i6400_msa_add_d): Update reservations.
7076 (i6400_msa_int_add) Likewise.
7077 (i6400_msa_short_logic3) Likewise.
7078 (i6400_msa_short_logic2) Likewise.
7079 (i6400_msa_short_logic) Likewise.
7080 (i6400_msa_move) Likewise.
7081 (i6400_msa_cmp) Likewise.
7082 (i6400_msa_short_float2) Likewise.
7083 (i6400_msa_div_d) Likewise.
7084 (i6400_msa_long_logic1) Likewise.
7085 (i6400_msa_long_logic2) Likewise.
7086 (i6400_msa_mult) Likewise.
7087 (i6400_msa_long_float2) Likewise.
7088 (i6400_msa_long_float4) Likewise.
7089 (i6400_msa_long_float5) Likewise.
7090 (i6400_msa_long_float8) Likewise.
7091 (i6400_fpu_fadd): Include frint type.
7092 (i6400_fpu_store): New define_insn_reservation.
7093 (i6400_fpu_load): Likewise.
7094 (i6400_fpu_move): Likewise.
7095 (i6400_fpu_fcmp): Likewise.
7096 (i6400_fpu_fmadd): Likewise.
7097 (i6400_int_mult): Include imul3nc type and update reservation.
7098 (i6400_int_div): Include idiv3 type and update reservation.
7099 (i6400_int_load): Update to check type not move_type.
7100 (i6400_int_store): Likewise.
7101 (i6400_int_prefetch): Set zero latency.
7103 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7105 * gcc.c: Document new %@{...} sequence.
7106 (LINK_COMMAND_SPEC): Use it for the -L switches.
7107 (cpp_unique_options): Use it for the -I switches.
7108 (at_file_argbuf): New global variable.
7109 (in_at_file): Likewise.
7110 (alloc_args): Create at_file_argbuf.
7111 (clear_args): Truncate at_file_argbuf.
7112 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7113 (open_at_file): New function.
7114 (close_at_file): Likewise.
7115 (create_at_file): Delete.
7116 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7119 (validate_switches_from_spec): Deal with %@{...} sequence.
7120 (validate_switches): Likewise.
7121 (driver::finalize): Call clear_args.
7123 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7125 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7127 2018-06-11 Martin Sebor <msebor@redhat.com>
7129 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7130 (Wno-class-memaccess): Add @opindex.
7131 (Wno-templates, Wno-multiple-inheritance): Same.
7132 (Wno-virtual-inheritance, Wno-namespaces): Same.
7133 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7134 (Wno-format-overflow, Wno-format-truncation): Same.
7135 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7136 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7137 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7138 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7139 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7140 (Wno-misspelled-isr): Same.
7142 2018-06-11 Martin Sebor <msebor@redhat.com>
7144 * PR tree-optimization/86083
7145 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7147 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7149 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7151 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7154 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7155 on the correct operand.
7156 (*movdi_internal64): Ditto.
7158 2018-06-11 Martin Liska <mliska@suse.cz>
7160 PR tree-optimization/86089
7161 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7163 2018-06-11 Julia Koval <julia.koval@intel.com>
7165 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7166 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7167 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7169 2018-06-11 Olivier Hainque <hainque@adacore.com>
7171 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7172 for Ada with strict dwarf2.
7174 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7177 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7180 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7182 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7184 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7186 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7189 2018-06-08 Martin Liska <mliska@suse.cz>
7191 * tree-cfg.h (debug_function): Fix argument type to match
7194 2018-06-08 Martin Liska <mliska@suse.cz>
7196 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7197 Remove usage of MPX-related (and removed) fields.
7198 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7200 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7202 * cfg.c (debug): Use TDF_NONE rather than 0.
7203 * cfghooks.c (debug): Likewise.
7204 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7205 (struct dump_option_value_info): Convert to...
7206 (struct kv_pair): ...this template type.
7207 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7209 (optinfo_verbosity_options): Likewise.
7210 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7212 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7213 than int for "optgroup_flags" param.
7214 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7216 (dump_dec): Likewise.
7217 (dump_finish): Use TDF_NONE rather than 0.
7218 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7219 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7220 than 0. Update for change to option_ptr.
7221 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7222 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7223 0. Update for changes to optinfo_verbosity_options and
7225 (opt_info_switch_p): Convert optgroup_flags from int to
7227 (dump_basic_block): Use dump_flags_t rather than int
7228 for "dump_kind" param.
7229 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7230 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7231 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7232 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7233 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7234 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7235 TDF_NONE): Convert from macros to...
7236 (enum dump_flag): ...this new enum.
7237 (dump_flags_t): Update to use enum.
7238 (operator|, operator&, operator~, operator|=, operator&=):
7239 Implement for dump_flags_t.
7240 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7241 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7242 Convert from macros to...
7243 (enum optgroup_flag): ...this new enum.
7244 (optgroup_flags_t): New typedef.
7245 (operator|, operator|=): Implement for optgroup_flags_t.
7246 (struct dump_file_info): Convert field "alt_flags" to
7247 dump_flags_t. Convert field "optgroup_flags" to
7249 (dump_basic_block): Use dump_flags_t rather than int for param.
7250 (dump_generic_expr_loc): Likewise.
7251 (dump_dec): Likewise.
7252 (dump_register): Convert param "optgroup_flags" to
7254 (opt_info_enable_passes): Likewise.
7255 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7257 * gimple-pretty-print.c (debug): Likewise.
7258 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7259 (merged_store_group::apply_stores): Likewise.
7260 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7261 * gimple.c (verify_gimple_pp): Likewise.
7262 * graphite-poly.c (print_pbb_body): Likewise.
7263 * passes.c (pass_manager::register_one_dump_file): Convert
7264 local "optgroup_flags" to optgroup_flags_t.
7265 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7267 (debug_body): Likewise.
7268 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7269 to optgroup_flags_t.
7270 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7272 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7273 (convert_mult_to_fma): Likewise.
7274 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7275 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7276 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7277 "dump_kind" to dump_flags_t.
7279 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7281 * config/rs6000/rs6000.c (min, max): Delete.
7283 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7285 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7286 -mabi=spe and -mabi=no-spe.
7288 2018-06-08 Martin Liska <mliska@suse.cz>
7290 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7291 where we expect an existing summary.
7293 2018-06-08 Martin Liska <mliska@suse.cz>
7295 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7296 * ipa-inline.h (estimate_edge_growth): Likewise.
7298 2018-06-08 Martin Liska <mliska@suse.cz>
7300 * cgraph.c (function_version_hasher::hash): Use
7301 cgraph_node::get_uid ().
7302 (function_version_hasher::equal):
7303 * cgraph.h (cgraph_node::get_uid): New method.
7304 * ipa-inline.c (update_caller_keys): Use
7305 cgraph_node::get_uid ().
7306 (update_callee_keys): Likewise.
7307 * ipa-utils.c (searchc): Likewise.
7308 (ipa_reduced_postorder): Likewise.
7309 * lto-cgraph.c (input_node): Likewise.
7310 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7311 * symbol-summary.h (symtab_insertion): Likewise.
7312 (symtab_removal): Likewise.
7313 (symtab_duplication): Likewise.
7314 * tree-pretty-print.c (dump_function_header): Likewise.
7315 * tree-sra.c (convert_callers_for_node): Likewise.
7317 2018-06-08 Martin Liska <mliska@suse.cz>
7319 * cgraph.c (symbol_table::create_edge): Always assign a new
7321 (symbol_table::free_edge): Do not recycle numbers.
7322 * cgraph.h (cgraph_edge::get): New method.
7323 * symbol-summary.h (symtab_removal): Use it.
7324 (symtab_duplication): Likewise.
7325 (call_summary::hashable_uid): Remove.
7327 2018-06-08 Martin Liska <mliska@suse.cz>
7329 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7330 (initialize_growth_caches): Remove.
7331 (free_growth_caches): Likewise.
7332 (do_estimate_edge_time): Use edge_growth_cache.
7333 (do_estimate_edge_size): Likewise.
7334 (do_estimate_edge_hints): Likewise.
7335 * ipa-inline.c (reset_edge_caches): Likewise.
7336 (recursive_inlining): Likewise.
7337 (inline_small_functions): Likewise.
7338 * ipa-inline.h (initialize_growth_caches): Remove.
7339 (estimate_edge_size): Likewise.
7340 (estimate_edge_time): Likewise.
7341 (estimate_edge_hints): Likewise.
7342 (reset_edge_growth_cache): Likewise.
7343 * symbol-summary.h (call_summary::remove): New method.
7345 2018-06-08 Martin Liska <mliska@suse.cz>
7347 * ipa-cp.c (class edge_clone_summary): New summary.
7348 (grow_edge_clone_vectors): Remove.
7349 (ipcp_edge_duplication_hook): Remove.
7350 (class edge_clone_summary_t): New call_summary class.
7351 (ipcp_edge_removal_hook): Remove.
7352 (edge_clone_summary_t::duplicate): New function.
7353 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7354 (create_specialized_node): Likewise.
7355 (ipcp_driver): Initialize edge_clone_summaries and do not
7358 2018-06-08 Martin Liska <mliska@suse.cz>
7360 * symbol-summary.h (get): New function.
7361 (call_summary::m_initialize_when_cloning): New class member.
7363 2018-06-08 Martin Liska <mliska@suse.cz>
7365 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7366 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7367 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7368 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7369 (remove_cgraph_node_from_order): Use the removed_nodes set.
7370 (do_per_function_toporder): Likwise.
7371 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7372 instead of summary_uid.
7373 (symtab_removal): Likewise.
7374 (symtab_duplication): Likewise.
7376 2018-06-08 Martin Liska <mliska@suse.cz>
7378 * ipa-cp.c (ipcp_store_bits_results): Use
7379 ipcp_transformation_sum.
7380 (ipcp_store_vr_results): Likewise.
7381 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7383 (ipcp_transformation_initialize): ... this.
7384 (ipa_set_node_agg_value_chain):
7385 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7386 (write_ipcp_transformation_info): Likewise.
7387 (read_ipcp_transformation_info): Likewise.
7388 (ipcp_update_bits): Likewise.
7389 (ipcp_update_vr): Likewise.
7390 (ipcp_transform_function): Likewise.
7391 * ipa-prop.h: Rename ipcp_transformation_summary to
7392 ipcp_transformation.
7393 (class ipcp_transformation_t): New function summary.
7394 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7395 (ipa_get_agg_replacements_for_node): Likewise.
7397 2018-06-08 Martin Liska <mliska@suse.cz>
7399 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7401 (class funct_state_summary_t): New function_summary class.
7402 (has_function_state): Remove.
7403 (get_function_state): Likewise.
7404 (set_function_state): Likewise.
7405 (add_new_function): Likewise.
7406 (funct_state_summary_t::insert): New function.
7407 (duplicate_node_data): Remove.
7408 (remove_node_data): Remove.
7409 (funct_state_summary_t::duplicate): New function.
7410 (register_hooks): Create new funct_state_summaries.
7411 (pure_const_generate_summary): Use it.
7412 (pure_const_write_summary): Likewise.
7413 (pure_const_read_summary): Likewise.
7414 (propagate_pure_const): Likewise.
7415 (propagate_nothrow): Likewise.
7416 (dump_malloc_lattice): Likewise.
7417 (propagate_malloc): Likewise.
7418 (execute): Do not register hooks, just remove summary
7420 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7423 2018-06-08 Martin Liska <mliska@suse.cz>
7425 * ipa-reference.c (remove_node_data): Remove.
7426 (duplicate_node_data): Likewise.
7427 (class ipa_ref_var_info_summary_t): New class.
7428 (class ipa_ref_opt_summary_t): Likewise.
7429 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7430 (get_reference_optimization_summary): Use
7431 ipa_ref_opt_sum_summaries.
7432 (set_reference_vars_info): Remove.
7433 (set_reference_optimization_summary): Likewise.
7434 (ipa_init): Create summaries.
7435 (init_function_info): Use function summary.
7436 (ipa_ref_opt_summary_t::duplicate): New function.
7437 (ipa_ref_opt_summary_t::remove): New function.
7438 (get_read_write_all_from_node): Fix GNU coding style.
7439 (propagate): Use function summary.
7440 (write_node_summary_p): Fix GNU coding style.
7441 (stream_out_bitmap): Likewise.
7442 (ipa_reference_read_optimization_summary): Use function summary.
7443 (ipa_reference_c_finalize): Do not release hooks.
7445 2018-06-08 Martin Liska <mliska@suse.cz>
7447 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7448 (analyze_function_body): Extract multiple calls of get_create.
7449 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7450 * ipa-inline.c (recursive_inlining): Use ::get method.
7451 * ipa-inline.h (estimate_edge_growth): Likewise.
7453 2018-06-08 Martin Liska <mliska@suse.cz>
7455 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7457 (hsa_function_summary::hsa_function_summary): Use the new enum
7459 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7460 * hsa-gen.c (hsa_get_host_function): Likewise.
7461 (get_brig_function_name): Likewise.
7462 * ipa-hsa.c (process_hsa_functions): Likewise.
7463 (ipa_hsa_write_summary): Likewise.
7464 * symbol-summary.h (symtab_duplication): Use ::get function/
7465 (get): New function.
7467 2018-06-08 Martin Liska <mliska@suse.cz>
7469 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7471 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7472 (hsa_register_kernel): Likewise.
7473 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7474 * hsa-gen.c (hsa_get_host_function): Likewise.
7475 (get_brig_function_name): Likewise.
7476 (generate_hsa): Likewise.
7477 (pass_gen_hsail::execute): Likewise.
7478 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7479 (devirtualization_time_bonus): Likewise.
7480 (ipcp_propagate_stage): Likewise.
7481 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7482 (edge_set_predicate): Likewise.
7483 (evaluate_conditions_for_known_args): Likewise.
7484 (evaluate_properties_for_edge): Likewise.
7485 (ipa_fn_summary::reset): Likewise.
7486 (ipa_fn_summary_t::duplicate): Likewise.
7487 (dump_ipa_call_summary): Likewise.
7488 (ipa_dump_fn_summary): Likewise.
7489 (analyze_function_body): Likewise.
7490 (compute_fn_summary): Likewise.
7491 (estimate_edge_devirt_benefit): Likewise.
7492 (estimate_edge_size_and_time): Likewise.
7493 (estimate_calls_size_and_time): Likewise.
7494 (estimate_node_size_and_time): Likewise.
7495 (inline_update_callee_summaries): Likewise.
7496 (remap_edge_change_prob): Likewise.
7497 (remap_edge_summaries): Likewise.
7498 (ipa_merge_fn_summary_after_inlining): Likewise.
7499 (ipa_update_overall_fn_summary): Likewise.
7500 (read_ipa_call_summary): Likewise.
7501 (inline_read_section): Likewise.
7502 (write_ipa_call_summary): Likewise.
7503 (ipa_fn_summary_write): Likewise.
7504 (ipa_free_fn_summary): Likewise.
7505 * ipa-hsa.c (process_hsa_functions): Likewise.
7506 (ipa_hsa_write_summary): Likewise.
7507 (ipa_hsa_read_section): Likewise.
7508 * ipa-icf.c (sem_function::merge): Likewise.
7509 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7510 (do_estimate_edge_time): Likewise.
7511 (estimate_size_after_inlining): Likewise.
7512 (estimate_growth): Likewise.
7513 (growth_likely_positive): Likewise.
7514 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7515 (inline_call): Likewise.
7516 * ipa-inline.c (caller_growth_limits): Likewise.
7517 (can_inline_edge_p): Likewise.
7518 (can_inline_edge_by_limits_p): Likewise.
7519 (compute_uninlined_call_time): Likewise.
7520 (compute_inlined_call_time): Likewise.
7521 (want_inline_small_function_p): Likewise.
7522 (edge_badness): Likewise.
7523 (update_caller_keys): Likewise.
7524 (update_callee_keys): Likewise.
7525 (recursive_inlining): Likewise.
7526 (inline_small_functions): Likewise.
7527 (inline_to_all_callers_1): Likewise.
7528 (dump_overall_stats): Likewise.
7529 (early_inline_small_functions): Likewise.
7530 (early_inliner): Likewise.
7531 * ipa-inline.h (estimate_edge_growth): Likewise.
7532 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7533 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7534 * ipa-prop.h (IPA_NODE_REF): Likewise.
7535 (IPA_EDGE_REF): Likewise.
7536 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7537 (propagate_malloc): Likewise.
7538 * ipa-split.c (execute_split_functions): Likewise.
7539 * symbol-summary.h: Rename get to get_create.
7541 (get_create): Likewise.
7542 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7544 2018-06-08 Martin Liska <mliska@suse.cz>
7546 * symbol-summary.h (release): Move definition out of class
7548 (symtab_removal): Likewise.
7549 (symtab_duplication): Likewise.
7551 2018-06-08 Martin Liska <mliska@suse.cz>
7553 * symbol-summary.h (function_summary): Move constructor
7554 implementation out of class declaration.
7555 (release): Likewise.
7556 (symtab_insertion): Likewise.
7557 (symtab_removal): Likewise.
7558 (symtab_duplication): Likewise.
7561 2018-06-08 Martin Liska <mliska@suse.cz>
7563 * Makefile.in: Remove support for MPX (macros, related functions,
7564 fields in cgraph_node, ...).
7565 * builtin-types.def (BT_BND): Likewise.
7566 (BT_FN_BND_CONST_PTR): Likewise.
7567 (BT_FN_CONST_PTR_BND): Likewise.
7568 (BT_FN_VOID_PTR_BND): Likewise.
7569 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7570 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7571 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7572 (expand_builtin_mempcpy_with_bounds): Likewise.
7573 (expand_builtin_memset_with_bounds): Likewise.
7574 (expand_builtin_memset_args): Likewise.
7575 (std_expand_builtin_va_start): Likewise.
7576 (expand_builtin): Likewise.
7577 (expand_builtin_with_bounds): Likewise.
7578 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7579 (DEF_LIB_BUILTIN_CHKP): Likewise.
7580 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7581 (DEF_CHKP_BUILTIN): Likewise.
7582 (BUILT_IN_MEMCPY): Likewise.
7583 (BUILT_IN_MEMMOVE): Likewise.
7584 (BUILT_IN_MEMPCPY): Likewise.
7585 (BUILT_IN_MEMSET): Likewise.
7586 (BUILT_IN_STPCPY): Likewise.
7587 (BUILT_IN_STRCAT): Likewise.
7588 (BUILT_IN_STRCHR): Likewise.
7589 (BUILT_IN_STRCPY): Likewise.
7590 (BUILT_IN_STRLEN): Likewise.
7591 (BUILT_IN_MEMCPY_CHK): Likewise.
7592 (BUILT_IN_MEMMOVE_CHK): Likewise.
7593 (BUILT_IN_MEMPCPY_CHK): Likewise.
7594 (BUILT_IN_MEMSET_CHK): Likewise.
7595 (BUILT_IN_STPCPY_CHK): Likewise.
7596 (BUILT_IN_STRCAT_CHK): Likewise.
7597 (BUILT_IN_STRCPY_CHK): Likewise.
7598 * calls.c (store_bounds): Likewise.
7599 (emit_call_1): Likewise.
7600 (special_function_p): Likewise.
7601 (maybe_warn_nonstring_arg): Likewise.
7602 (initialize_argument_information): Likewise.
7603 (finalize_must_preallocate): Likewise.
7604 (compute_argument_addresses): Likewise.
7605 (expand_call): Likewise.
7606 * cfgexpand.c (expand_call_stmt): Likewise.
7607 (expand_return): Likewise.
7608 (expand_gimple_stmt_1): Likewise.
7609 (pass_expand::execute): Likewise.
7610 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7611 (cgraph_node::remove): Likewise.
7612 (cgraph_node::dump): Likewise.
7613 (cgraph_node::verify_node): Likewise.
7614 * cgraph.h (chkp_function_instrumented_p): Likewise.
7615 (symtab_node::get_alias_target): Likewise.
7616 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7617 (cgraph_local_p): Likewise.
7618 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7619 (cgraph_edge::rebuild_references): Likewise.
7620 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7621 (walk_polymorphic_call_targets): Likewise.
7622 (cgraph_node::expand_thunk): Likewise.
7623 (symbol_table::output_weakrefs): Likewise.
7624 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7625 (ix86_handle_option): Likewise.
7626 * config/i386/constraints.md: Likewise.
7627 * config/i386/i386-builtin-types.def (BND): Likewise.
7631 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7632 (BDESC_FIRST): Likewise.
7634 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7635 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7636 * config/i386/i386.c (enum reg_class): Likewise.
7637 (ix86_target_string): Likewise.
7638 (ix86_option_override_internal): Likewise.
7639 (ix86_conditional_register_usage): Likewise.
7640 (ix86_valid_target_attribute_inner_p): Likewise.
7641 (ix86_set_indirect_branch_type): Likewise.
7642 (ix86_set_current_function): Likewise.
7643 (ix86_function_arg_regno_p): Likewise.
7644 (init_cumulative_args): Likewise.
7645 (ix86_function_arg_advance): Likewise.
7646 (ix86_function_arg): Likewise.
7647 (ix86_pass_by_reference): Likewise.
7648 (ix86_function_value_regno_p): Likewise.
7649 (ix86_function_value_1): Likewise.
7650 (ix86_function_value_bounds): Likewise.
7651 (ix86_return_in_memory): Likewise.
7652 (ix86_setup_incoming_vararg_bounds): Likewise.
7653 (ix86_va_start): Likewise.
7654 (indirect_thunk_need_prefix): Likewise.
7655 (print_reg): Likewise.
7656 (ix86_print_operand): Likewise.
7657 (ix86_expand_call): Likewise.
7658 (ix86_output_function_return): Likewise.
7659 (reg_encoded_number): Likewise.
7660 (BDESC_VERIFYS): Likewise.
7661 (ix86_init_mpx_builtins): Likewise.
7662 (ix86_init_builtins): Likewise.
7663 (ix86_emit_cmove): Likewise.
7664 (ix86_emit_move_max): Likewise.
7665 (ix86_expand_builtin): Likewise.
7666 (ix86_builtin_mpx_function): Likewise.
7667 (ix86_get_arg_address_for_bt): Likewise.
7668 (ix86_load_bounds): Likewise.
7669 (ix86_store_bounds): Likewise.
7670 (ix86_load_returned_bounds): Likewise.
7671 (ix86_store_returned_bounds): Likewise.
7672 (ix86_class_likely_spilled_p): Likewise.
7673 (ix86_hard_regno_mode_ok): Likewise.
7674 (x86_order_regs_for_local_alloc): Likewise.
7675 (ix86_mitigate_rop): Likewise.
7676 (ix86_bnd_prefixed_insn_p): Likewise.
7677 (ix86_mpx_bound_mode): Likewise.
7678 (ix86_make_bounds_constant): Likewise.
7679 (ix86_initialize_bounds): Likewise.
7680 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7681 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7682 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7683 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7684 (TARGET_CHKP_BOUND_MODE): Likewise.
7685 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7686 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7687 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7688 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7689 * config/i386/i386.h (TARGET_MPX): Likewise.
7690 (TARGET_MPX_P): Likewise.
7691 (VALID_BND_REG_MODE): Likewise.
7692 (FIRST_BND_REG): Likewise.
7693 (LAST_BND_REG): Likewise.
7694 (enum reg_class): Likewise.
7695 (BND_REG_P): Likewise.
7696 (BND_REGNO_P): Likewise.
7697 (BNDmode): Likewise.
7698 (ADJUST_INSN_LENGTH): Likewise.
7699 * config/i386/i386.md: Likewise.
7700 * config/i386/i386.opt: Likewise.
7701 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7702 (defined): Likewise.
7703 (LINK_MPX): Likewise.
7704 (MPX_SPEC): Likewise.
7705 (LIBMPX_SPEC): Likewise.
7706 (LIBMPXWRAPPERS_SPEC): Likewise.
7707 (CHKP_SPEC): Likewise.
7708 * config/i386/predicates.md: Likewise.
7709 * dbxout.c (dbxout_type): Likewise.
7710 * doc/extend.texi: Likewise.
7711 * doc/invoke.texi: Likewise.
7712 * doc/md.texi: Likewise.
7713 * doc/tm.texi: Likewise.
7714 * doc/tm.texi.in: Likewise.
7715 * dwarf2out.c (is_base_type): Likewise.
7716 (gen_formal_types_die): Likewise.
7717 (gen_subprogram_die): Likewise.
7718 (gen_type_die_with_usage): Likewise.
7719 (gen_decl_die): Likewise.
7720 (dwarf2out_late_global_decl): Likewise.
7721 * expr.c (expand_assignment): Likewise.
7722 (emit_storent_insn): Likewise.
7723 (store_expr_with_bounds): Likewise.
7724 (store_expr): Likewise.
7725 (expand_expr_real_1): Likewise.
7726 * expr.h (store_expr_with_bounds): Likewise.
7727 * function.c (use_register_for_decl): Likewise.
7728 (struct bounds_parm_data): Likewise.
7729 (assign_parms_augmented_arg_list): Likewise.
7730 (assign_parm_find_entry_rtl): Likewise.
7731 (assign_parm_is_stack_parm): Likewise.
7732 (assign_parm_load_bounds): Likewise.
7733 (assign_bounds): Likewise.
7734 (assign_parms): Likewise.
7735 (expand_function_start): Likewise.
7736 * gcc.c (CHKP_SPEC): Likewise.
7737 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7738 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7739 (wrestrict_dom_walker::check_call): Likewise.
7740 * gimple.c (gimple_build_call_from_tree): Likewise.
7741 * gimple.h (enum gf_mask): Likewise.
7742 (gimple_call_with_bounds_p): Likewise.
7743 (gimple_call_set_with_bounds): Likewise.
7744 * gimplify.c (gimplify_init_constructor): Likewise.
7745 * ipa-cp.c (initialize_node_lattices): Likewise.
7746 (propagate_constants_across_call): Likewise.
7747 (find_more_scalar_values_for_callers_subset): Likewise.
7748 * ipa-hsa.c (process_hsa_functions): Likewise.
7749 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7750 * ipa-icf.c (sem_function::merge): Likewise.
7751 * ipa-inline.c (early_inliner): Likewise.
7752 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7753 (warn_function_cold): Likewise.
7754 (propagate_pure_const): Likewise.
7755 * ipa-ref.h (enum GTY): Likewise.
7756 * ipa-split.c (find_retbnd): Likewise.
7757 (consider_split): Likewise.
7758 (split_function): Likewise.
7759 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7760 * ipa.c (walk_polymorphic_call_targets): Likewise.
7761 (symbol_table::remove_unreachable_nodes): Likewise.
7762 (process_references): Likewise.
7763 (cgraph_build_static_cdtor_1): Likewise.
7764 * lto-cgraph.c (lto_output_node): Likewise.
7765 (output_refs): Likewise.
7766 (compute_ltrans_boundary): Likewise.
7767 (input_overwrite_node): Likewise.
7768 (input_node): Likewise.
7769 (input_cgraph_1): Likewise.
7770 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7771 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7772 (class pass_chkp_instrumentation_passes): Likewise.
7773 (make_pass_chkp_instrumentation_passes): Likewise.
7774 * passes.def: Likewise.
7775 * rtl.h (struct GTY): Likewise.
7776 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7777 * stor-layout.c (layout_type): Likewise.
7778 * symtab.c: Likewise.
7779 * target.def: Likewise.
7780 * targhooks.c (default_chkp_bound_type): Likewise.
7781 (default_chkp_bound_mode): Likewise.
7782 (default_builtin_chkp_function): Likewise.
7783 (default_chkp_function_value_bounds): Likewise.
7784 (default_chkp_make_bounds_constant): Likewise.
7785 (default_chkp_initialize_bounds): Likewise.
7786 * targhooks.h (default_chkp_bound_type): Likewise.
7787 (default_chkp_bound_mode): Likewise.
7788 (default_builtin_chkp_function): Likewise.
7789 (default_chkp_function_value_bounds): Likewise.
7790 (default_chkp_make_bounds_constant): Likewise.
7791 (default_chkp_initialize_bounds): Likewise.
7792 * toplev.c (compile_file): Likewise.
7793 (process_options): Likewise.
7794 * tree-core.h (DEF_BUILTIN): Likewise.
7795 (DEF_BUILTIN_CHKP): Likewise.
7796 * tree-inline.c (declare_return_variable): Likewise.
7797 (remap_gimple_stmt): Likewise.
7798 (copy_bb): Likewise.
7799 (initialize_inlined_parameters): Likewise.
7800 (expand_call_inline): Likewise.
7801 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7802 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7803 (make_pass_ipa_chkp_produce_thunks): Likewise.
7804 (make_pass_chkp): Likewise.
7805 (make_pass_chkp_opt): Likewise.
7806 (make_pass_chkp_instrumentation_passes): Likewise.
7807 * tree-pretty-print.c (dump_generic_node): Likewise.
7808 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7809 * tree-ssa-dce.c (propagate_necessity): Likewise.
7810 (eliminate_unnecessary_stmts): Likewise.
7811 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7812 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7813 * tree-ssa-sccvn.h: Likewise.
7814 * tree-ssa-strlen.c (get_string_length): Likewise.
7815 (valid_builtin_call): Likewise.
7816 (adjust_last_stmt): Likewise.
7817 (handle_builtin_strchr): Likewise.
7818 (handle_builtin_strcpy): Likewise.
7819 (handle_builtin_stxncpy): Likewise.
7820 (handle_builtin_memcpy): Likewise.
7821 (handle_builtin_strcat): Likewise.
7822 (strlen_check_and_optimize_stmt): Likewise.
7823 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7824 * tree-streamer-in.c: Likewise.
7825 * tree-streamer.c (record_common_node): Likewise.
7826 * tree.c (tree_code_size): Likewise.
7827 (wide_int_to_tree_1): Likewise.
7828 (type_contains_placeholder_1): Likewise.
7829 (build_common_tree_nodes): Likewise.
7830 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7831 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7832 (POINTER_BOUNDS_P): Likewise.
7833 (BOUNDED_TYPE_P): Likewise.
7834 (BOUNDED_P): Likewise.
7835 (CALL_WITH_BOUNDS_P): Likewise.
7836 (pointer_bounds_type_node): Likewise.
7837 * value-prof.c (gimple_ic): Likewise.
7838 * var-tracking.c (vt_add_function_parameters): Likewise.
7839 * varasm.c (make_decl_rtl): Likewise.
7840 (assemble_start_function): Likewise.
7841 (output_constant): Likewise.
7842 (maybe_assemble_visibility): Likewise.
7843 * varpool.c (ctor_for_folding): Likewise.
7844 * chkp-builtins.def: Remove.
7845 * ipa-chkp.c: Remove.
7846 * ipa-chkp.h: Remove.
7847 * rtl-chkp.c: Remove.
7848 * rtl-chkp.h: Remove.
7849 * tree-chkp-opt.c: Remove.
7850 * tree-chkp.c: Remove.
7851 * tree-chkp.h: Remove.
7853 2018-06-07 Carl Love <cel@us.ibm.com>
7855 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7856 vextract_fp_from_shortl): Add BE support.
7858 2018-06-07 Paul Koning <ni1d@arrl.net>
7860 * compare-elim.c (try_merge_compare): Don't merge compare if
7861 address contains a side effect.
7862 (try_eliminate_compare): Likewise.
7864 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7866 * config.gcc: Support "tremont".
7867 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7868 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7870 * config/i386/i386.c (m_TREMONT): Define.
7871 (processor_target_table): Add "tremont".
7872 (PTA_TREMONT): Define.
7873 (ix86_lea_outperforms): Add TARGET_TREMONT.
7874 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7875 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7876 and M_INTEL_GOLDMONT_PLUS.
7877 (fold_builtin_cpu): Add "tremont".
7878 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7879 (ix86_option_override_internal): Add "tremont".
7880 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7881 (processor_type): Add PROCESSOR_TREMONT.
7882 * config/i386/x86-tune.def: Add m_TREMONT.
7883 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7885 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7887 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7888 symbol defined for msp430i* devices to be lower case.
7890 2018-06-07 Richard Biener <rguenther@suse.de>
7892 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7893 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
7894 Properly wrap signed arithmetic if overflow wraps.
7896 2018-06-07 Jakub Jelinek <jakub@redhat.com>
7898 PR tree-optimization/69615
7899 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7900 of a cast from a same precision integral SSA_NAME in a bb dominated
7901 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7902 cast to utype if rhs2 has already a compatible type.
7904 2018-06-07 Richard Biener <rguenther@suse.de>
7906 PR tree-optimization/85935
7907 * graphite-scop-detection.c (find_params_in_bb): Analyze
7908 condition operands with respect to the correct loop. Assert
7909 the analysis doesn't fail.
7911 2018-06-04 Carl Love <cel@us.ibm.com>
7913 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7914 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
7915 as it is slightly cheaper.
7916 (first_match_or_eos_index_<mode>):
7917 Calculate index using natural element order.
7918 (first_match_index_<mode>):
7919 Calculate index using natural element order.
7920 (first_match_or_eos_index_<mode>):
7921 Calculate index using natural order.
7922 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7923 for BE and LE modes.
7924 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7925 P9V_BUILTIN_VCLZLSBB_V16QI.
7926 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7929 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7931 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7932 indentation and line wrap for many prototypes. Add missing
7933 @smallexample directives around block of prototypes for vec_xl and
7936 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
7938 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7939 track if we pass or return IEEE 128-bit floating point.
7940 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7941 C++ mangling that is compatible with GCC 8.1.
7942 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7943 (init_cumulative_args): Note if we pass or return IEEE 128-bit
7944 floating point types.
7945 (rs6000_function_arg_advance_1): Likewise.
7946 (rs6000_mangle_type): Optionally generate mangled names that match
7947 what GCC 8.1 generated for IEEE 128-bit floating point types.
7948 (rs6000_globalize_decl_name): If we have an external function that
7949 passes or returns IEEE 128-bit types, generate a weak reference
7950 from the mangled name used in GCC 8.1 to the current mangled
7952 (rs6000_init_builtins): Make __ibm128 use the long double type if
7953 long double is IBM extended double. Make __float128 use the long
7954 double type if long double is IEEE 128-bit.
7957 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7958 macro for __ibm128 built-in functions.
7959 (PACK_IF): Add __ibm128 pack/unpack functions.
7960 (UNPACK_IF): Likewise.
7961 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7962 enable long double built-in functions if long double is IEEE
7963 128-bit floating point.
7964 (rs6000_invalid_builtin): Update long double built-in function
7966 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7967 functions, adjust the built-in function to use the long double
7968 built-in function if __ibm128 and long double are the same type.
7969 * doc/extend.texi (PowerPC builtins): Update documention for
7970 __builtin_{,un}pack_longdouble. Add documentation for
7971 __builtin_{,un}pack_ibm128.
7973 2018-06-06 Jim Wilson <jimw@sifive.com>
7975 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7976 (struct machine_function): New field interrupt_mode.
7977 (riscv_handle_type_attribute): New function. Add forward declaration.
7978 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7979 (riscv_expand_epilogue): Check interrupt_mode field.
7980 (riscv_set_current_function): Check interrupt attribute args and
7981 set interrupt_mode field.
7982 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7983 (riscv_sret, riscv_uret): New.
7984 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7985 new arguments to interrupt attribute.
7987 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
7990 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7991 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7993 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7995 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7996 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7997 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7998 several redundant entries.
8000 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8002 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8003 type from "rtx" to "rtx_insn *".
8004 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8005 for local "call_insn", removing cast.
8006 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8008 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8010 PR tree-optimization/86066
8011 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8012 for BIT_INSERT_EXPR stores.
8014 2018-06-06 Richard Biener <rguenther@suse.de>
8016 PR tree-optimization/86062
8017 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8018 component refs ontop
8019 of to be offsetted base.
8021 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8023 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8024 to be static and remove check on interrupt attribute name.
8026 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8028 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8029 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8031 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8034 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8036 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8037 Remove second argument, change how error is called.
8038 (aarch64_layout_arg): Remove second argument from
8039 aarch64_err_no_fpadvsimd call.
8040 (aarch64_init_cumulative_args): Ditto.
8041 (aarch64_gimplify_va_arg_expr): Ditto.
8042 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8044 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8046 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8047 (*simple_return_indirect_internal<mode>): Rename from
8048 simple_return_indirect_internal. Use W mode iterator.
8049 (rstorssp): New expander.
8050 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8051 (clrssbsy): New expander.
8052 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8054 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8056 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8058 (cmse_check_pointed_object): Likewise.
8060 2018-06-05 Martin Liska <mliska@suse.cz>
8062 PR gcov-profile/47618
8063 * doc/invoke.texi: Document how -fprofile-dir format
8066 2018-06-05 Richard Biener <rguenther@suse.de>
8068 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8069 removal pretend DOM info isn't available so we do not update
8070 it and only remove edges, not dominated blocks. Actually free
8071 DOM info in case we removed something. Remove unreachable blocks.
8072 (mfb_keep_latches): Work with either DOM info or marked backedges.
8073 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8074 first. Mark backedges if DOM info isn't available.
8075 (Re-)compute DOM info after cleanup_control_flow_pre.
8077 2018-06-05 Richard Biener <rguenther@suse.de>
8079 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8080 (locus_discrim_hasher::hash): Adjust.
8081 (locus_discrim_hasher::equal): Likewise.
8082 (next_discriminator_for_locus): Work on line directly.
8083 (same_line_p): Pass in expanded locus1 as well.
8084 (assign_discriminators): Avoid redundant location expansions.
8086 2018-06-05 Richard Biener <rguenther@suse.de>
8088 PR tree-optimization/86046
8089 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8090 if required after clearing TREE_ADDRESSABLE.
8092 2018-06-05 Richard Biener <rguenther@suse.de>
8094 PR tree-optimization/86047
8095 * tree-ssa-loop.c (for_each_index): Glob handling of all
8096 decls and constants and really handle all of them.
8098 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8101 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8102 qualifier_void_pointer and qualifier_const_void_pointer.
8103 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8104 (arm_init_builtins): Handle the above.
8105 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8106 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8107 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8110 2018-06-05 Martin Liska <mliska@suse.cz>
8112 * auto-profile.c (read_autofdo_file): Do not use
8113 gcov_ctr_summary struct.
8114 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8115 * coverage.c (struct counts_entry): Likewise.
8116 (read_counts_file): Read just single summary entry.
8117 (get_coverage_counts): Use gcov_summary struct.
8118 * coverage.h (get_coverage_counts): Likewise.
8119 * gcov-dump.c (dump_working_sets): Likewise.
8120 (tag_summary): Dump just single summary.
8121 * gcov-io.c (gcov_write_summary): Write just histogram
8123 (gcov_read_summary): Read just single summary.
8124 (compute_working_sets): Use gcov_summary struct.
8125 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8126 of GCOV_COUNTERS_SUMMABLE.
8127 (GCOV_COUNTERS_SUMMABLE): Remove.
8128 (GCOV_FIRST_VALUE_COUNTER): Replace with
8129 GCOV_COUNTER_V_INTERVAL.
8130 (struct gcov_ctr_summary): Remove.
8131 (struct gcov_summary): Directly use fields of former
8133 (compute_working_sets): Use gcov_summary struct.
8134 * gcov.c (read_count_file): Do not use ctrs fields.
8135 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8137 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8139 * profile.c: Likewise.
8140 * profile.h: Likewise.
8142 2018-06-05 Martin Liska <mliska@suse.cz>
8144 PR gcov-profile/84846
8145 * gcov.c (output_lines): Print working directory only
8146 in intermediate format.
8148 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8150 * config/s390/s390-builtin-types.def: Add void function type.
8151 * config/s390/s390-builtins.def: Use the function type for the
8154 2018-06-04 Jim Wilson <jimw@sifive.com>
8156 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8158 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8159 and maybe_eh_return. Change regno to unsigned int. Use new args to
8160 handle EH_RETURN_DATA_REGNO registers properly.
8161 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8162 (riscv_expand_epilogue): Update comment. Change argument name and
8163 type. Update code to use new name and type. Pass new args to
8164 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8166 * config/riscv/riscv.md (NORMAL_RETURN): New.
8167 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8168 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8169 (eh_return): Call gen_eh_return_internal and emit barrier.
8170 (eh_return_internal): Call riscv_expand_epilogue.
8172 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8174 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8175 bit_insertion field and declare can_be_merged_into method.
8176 (merged_store_group::can_be_merged_into): New method.
8177 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8178 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8179 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8181 2018-06-04 Richard Biener <rguenther@suse.de>
8183 PR tree-optimization/85955
8184 * builtins.c (fold_builtin_sincos): Convert pointers to
8185 destination to appropriate type before dereferencing.
8187 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8189 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8191 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8193 * expr.c (expand_expr_real_1): Force the operand into memory if
8194 its TYPE_MODE is BLKmode and if there is no integer mode for
8195 the number of bits being extracted.
8197 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8201 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8202 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8204 2018-06-04 Richard Biener <rguenther@suse.de>
8206 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8207 (cleanup_tree_cfg_noloop): ... single caller. Do
8208 start_recording_case_labels later.
8210 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8212 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8213 to _IMMINTRIN_H_INCLUDED.
8214 * config/i386/pconfigintrin.h: Ditto.
8215 * config/i386/waitpkgintrin.h: Ditto.
8216 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8217 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8218 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8219 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8220 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8221 waitpkgintrin.h and cldemoteintrin.h.
8223 2018-06-04 Richard Biener <rguenther@suse.de>
8225 PR tree-optimization/86038
8226 * tracer.c (find_best_successor): Check probability for
8227 being initialized, bail out if not.
8229 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8232 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8233 of bits to ignore when comparing architectures.
8235 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8237 PR tree-optimization/69615
8238 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8239 maximum or minimum of the type, try to merge it also as if
8240 range1 is + [-, x - 1] or + [x + 1, -].
8243 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8245 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8247 PR tree-optimization/86034
8248 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8249 the unsigned bitfield type in a bit insertion sequence if it does not
8250 have a larger precision than the bitfield size.
8251 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8253 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8255 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8257 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8259 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8260 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8261 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8262 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8264 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8266 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8267 Disable -fdelete-null-pointer-checks for ELF toolchain.
8269 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8270 Kito Cheng <kito.cheng@gmail.com>
8272 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8273 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8274 (nds32*-*-*): Add glibc and uclibc conditions.
8275 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8276 (TARGET_EXCEPT_UNWIND_INFO): Define.
8277 * config/nds32/elf.h: New file.
8278 * config/nds32/linux.h: New file.
8279 * config/nds32/nds32-elf.opt: New file.
8280 * config/nds32/nds32-linux.opt: New file.
8281 * config/nds32/nds32-fp-as-gp.c
8282 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8283 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8285 (nds32_asm_file_end): Ditto.
8286 (nds32_print_operand): Ditto.
8287 (nds32_insert_attributes): Ditto.
8288 (nds32_init_libfuncs): New function.
8289 (TARGET_HAVE_TLS): Define.
8290 (TARGET_INIT_LIBFUNCS): Define.
8291 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8293 (TARGET_ELF): Apply different mcmodel setting.
8294 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8295 been migrated into elf.h and linux.h files.
8296 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8297 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8298 (mcmodel): The content has been migrated into nds32-elf.opt and
8299 nds32-linux.opt files.
8300 * config/nds32/t-elf: New file.
8301 * config/nds32/t-linux: New file.
8303 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8304 Shiva Chen <shiva0217@gmail.com>
8306 * config/nds32/constants.md (unspec_volatile_element): Add
8307 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8308 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8310 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8311 (make_pass_nds32_fp_as_gp): Declare.
8312 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8314 (nds32_asm_function_end_prologue): Remove unused asm output.
8315 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8316 (nds32_asm_file_start): Output necessary fp_as_gp information.
8317 (nds32_option_override): Adjust register usage.
8318 (nds32_expand_prologue): Consider fp_as_gp situation.
8319 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8320 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8323 (simple_return): Ditto.
8324 (omit_fp_begin): Output special directive for fp_as_gp.
8325 (omit_fp_end): Output special directive for fp_as_gp.
8326 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8327 mforbid-fp-as-gp): New options.
8329 2018-06-01 Mark Wielaard <mark@klomp.org>
8331 * dwarf2out.c (dwarf2out_finish): Remove generation of
8332 DW_AT_loclists_base.
8334 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8336 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8337 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8338 (struct merged_store_group): Add bit_insertion field.
8339 (dump_char_array): Use standard hexadecimal format.
8340 (merged_store_group::merged_store_group): Set bit_insertion to false.
8341 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8342 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8343 also print the mask in the dump file.
8344 (pass_store_merging::gate): Minor tweak.
8345 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8346 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8347 stores with INTEGER_CST stores.
8348 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8349 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8350 and use it throughout. Generate bit insertion sequences if need be.
8351 (pass_store_merging::process_store): Remove redundant condition.
8352 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8354 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8356 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8357 the 128-bit floating point types. Fix function comment.
8359 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8361 * config/aarch64/aarch64-simd.md
8362 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8364 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8367 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8369 PR tree-optimization/85989
8370 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8372 (backprop::intersect_uses): Check it when deciding whether this
8373 is a backedge reference.
8374 (backprop::process_block): Add each phi to m_visited_phis
8375 after visiting it, then clear it at the end.
8377 2018-06-01 Richard Biener <rguenther@suse.de>
8379 * tree-vectorizer.h (vect_dr_stmt): New function.
8380 (vect_get_load_cost): Adjust.
8381 (vect_get_store_cost): Likewise.
8382 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8383 Use vect_dr_stmt instead of DR_SMTT.
8384 (vect_record_base_alignments): Likewise.
8385 (vect_calculate_target_alignment): Likewise.
8386 (vect_compute_data_ref_alignment): Likewise and make static.
8387 (vect_update_misalignment_for_peel): Likewise.
8388 (vect_verify_datarefs_alignment): Likewise.
8389 (vector_alignment_reachable_p): Likewise.
8390 (vect_get_data_access_cost): Likewise. Pass down
8391 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8392 (vect_get_peeling_costs_all_drs): Likewise.
8393 (vect_peeling_hash_get_lowest_cost): Likewise.
8394 (vect_enhance_data_refs_alignment): Likewise.
8395 (vect_find_same_alignment_drs): Likewise.
8396 (vect_analyze_data_refs_alignment): Likewise.
8397 (vect_analyze_group_access_1): Likewise.
8398 (vect_analyze_group_access): Likewise.
8399 (vect_analyze_data_ref_access): Likewise.
8400 (vect_analyze_data_ref_accesses): Likewise.
8401 (vect_vfa_segment_size): Likewise.
8402 (vect_small_gap_p): Likewise.
8403 (vectorizable_with_step_bound_p): Likewise.
8404 (vect_prune_runtime_alias_test_list): Likewise.
8405 (vect_analyze_data_refs): Likewise.
8406 (vect_supportable_dr_alignment): Likewise.
8407 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8408 (vect_gen_prolog_loop_niters): Likewise.
8409 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8410 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8412 (vect_recog_mask_conversion_pattern): Likewise.
8413 (vect_try_gather_scatter_pattern): Likewise.
8414 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8415 to vect_get_store_cost.
8416 (vect_get_store_cost): Get stmt_info instead of DR.
8417 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8418 (vect_get_load_cost): Get stmt_info instead of DR.
8420 2018-06-01 Richard Biener <rguenther@suse.de>
8423 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8424 vars wrapped in MEM_REFs.
8426 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8428 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8429 Fix subreg tests so that we only return a choice between
8430 GENERAL_REGS and FP_REGS if the original classes included both.
8432 2018-06-01 Richard Biener <rguenther@suse.de>
8435 * tree-ssa-structalias.c (get_function_part_constraint):
8436 Handle NULL fi->decl.
8437 (find_func_aliases_for_call): Properly handle indirect
8438 fi from direct call.
8439 (find_func_clobbers): Likewise.
8440 (ipa_pta_execute): Likewise.
8441 (create_variable_info_for): For functions that are ifunc_resolver
8442 resolve to a varinfo that contains the result of the resolver call.
8443 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8446 2018-05-31 Michael Collison <michael.collison@arm.com>
8448 * config/aarch64/aarch64.md:
8449 (*fix_to_zero_extenddfdi2): New pattern.
8450 * gcc.target/aarch64/fix_extend1.c: New testcase.
8452 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8456 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8457 and BUILT_IN_STRNCMP_EQ.
8458 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8459 BUILT_IN_STRNCMP_EQ.
8460 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8461 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8462 (gimple_fold_builtin): Likewise.
8463 * tree-ssa-strlen.c (compute_string_length): New function.
8464 (determine_min_obsize): New function.
8465 (handle_builtin_string_cmp): New function to handle calls to
8466 string compare functions.
8467 (strlen_optimize_stmt): Add handling to builtin string compare
8469 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8470 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8471 * tree.c (build_common_builtin_nodes): Add new defines of
8472 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8474 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8477 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8478 functions with naked attribute.
8480 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8482 * config/i386/sse.md (avx_vec_concat<mode>):
8483 Substitute concat_tg_mode mode attribute with xtg_mode.
8484 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8485 (concat_tg_mode): Remove mode attribute.
8487 2018-05-31 Martin Sebor <msebor@redhat.com>
8490 * calls.c (alloc_max_size): Correct a logic error/typo.
8491 Treat excessive arguments as infinite. Warn for invalid arguments.
8492 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8494 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8497 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8498 and movx for Haswell.
8500 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8501 Cesar Philippidis <cesar@codesourcery.com>
8504 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8505 when emitting error on private/firstprivate reductions.
8506 * omp-low.c (lower_omp_target): Avoid reference-type processing
8507 on pointers for firstprivate clause.
8509 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8511 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8514 * config/aarch64/aarch64-simd.md
8515 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8516 (aarch64_ld1_x3_<mode>): Likewise
8517 (aarch64_st1x2<VALLDIF:mode>): Likewise
8518 (aarch64_st1_x2_<mode>): Likewise
8519 (aarch64_st1x3<VALLDIF:mode>): Likewise
8520 (aarch64_st1_x3_<mode>): Likewise
8521 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8522 (vld1_s8_x3): Likewise.
8523 (vld1_u16_x3): Likewise.
8524 (vld1_s16_x3): Likewise.
8525 (vld1_u32_x3): Likewise.
8526 (vld1_s32_x3): Likewise.
8527 (vld1_u64_x3): Likewise.
8528 (vld1_s64_x3): Likewise.
8529 (vld1_f16_x3): Likewise.
8530 (vld1_f32_x3): Likewise.
8531 (vld1_f64_x3): Likewise.
8532 (vld1_p8_x3): Likewise.
8533 (vld1_p16_x3): Likewise.
8534 (vld1_p64_x3): Likewise.
8535 (vld1q_u8_x3): Likewise.
8536 (vld1q_s8_x3): Likewise.
8537 (vld1q_u16_x3): Likewise.
8538 (vld1q_s16_x3): Likewise.
8539 (vld1q_u32_x3): Likewise.
8540 (vld1q_s32_x3): Likewise.
8541 (vld1q_u64_x3): Likewise.
8542 (vld1q_s64_x3): Likewise.
8543 (vld1q_f16_x3): Likewise.
8544 (vld1q_f32_x3): Likewise.
8545 (vld1q_f64_x3): Likewise.
8546 (vld1q_p8_x3): Likewise.
8547 (vld1q_p16_x3): Likewise.
8548 (vld1q_p64_x3): Likewise.
8549 (vst1_s64_x2): Likewise.
8550 (vst1_u64_x2): Likewise.
8551 (vst1_f64_x2): Likewise.
8552 (vst1_s8_x2): Likewise.
8553 (vst1_p8_x2): Likewise.
8554 (vst1_s16_x2): Likewise.
8555 (vst1_p16_x2): Likewise.
8556 (vst1_s32_x2): Likewise.
8557 (vst1_u8_x2): Likewise.
8558 (vst1_u16_x2): Likewise.
8559 (vst1_u32_x2): Likewise.
8560 (vst1_f16_x2): Likewise.
8561 (vst1_f32_x2): Likewise.
8562 (vst1_p64_x2): Likewise.
8563 (vst1q_s8_x2): Likewise.
8564 (vst1q_p8_x2): Likewise.
8565 (vst1q_s16_x2): Likewise.
8566 (vst1q_p16_x2): Likewise.
8567 (vst1q_s32_x2): Likewise.
8568 (vst1q_s64_x2): Likewise.
8569 (vst1q_u8_x2): Likewise.
8570 (vst1q_u16_x2): Likewise.
8571 (vst1q_u32_x2): Likewise.
8572 (vst1q_u64_x2): Likewise.
8573 (vst1q_f16_x2): Likewise.
8574 (vst1q_f32_x2): Likewise.
8575 (vst1q_f64_x2): Likewise.
8576 (vst1q_p64_x2): Likewise.
8577 (vst1_s64_x3): Likewise.
8578 (vst1_u64_x3): Likewise.
8579 (vst1_f64_x3): Likewise.
8580 (vst1_s8_x3): Likewise.
8581 (vst1_p8_x3): Likewise.
8582 (vst1_s16_x3): Likewise.
8583 (vst1_p16_x3): Likewise.
8584 (vst1_s32_x3): Likewise.
8585 (vst1_u8_x3): Likewise.
8586 (vst1_u16_x3): Likewise.
8587 (vst1_u32_x3): Likewise.
8588 (vst1_f16_x3): Likewise.
8589 (vst1_f32_x3): Likewise.
8590 (vst1_p64_x3): Likewise.
8591 (vst1q_s8_x3): Likewise.
8592 (vst1q_p8_x3): Likewise.
8593 (vst1q_s16_x3): Likewise.
8594 (vst1q_p16_x3): Likewise.
8595 (vst1q_s32_x3): Likewise.
8596 (vst1q_s64_x3): Likewise.
8597 (vst1q_u8_x3): Likewise.
8598 (vst1q_u16_x3): Likewise.
8599 (vst1q_u32_x3): Likewise.
8600 (vst1q_u64_x3): Likewise.
8601 (vst1q_f16_x3): Likewise.
8602 (vst1q_f32_x3): Likewise.
8603 (vst1q_f64_x3): Likewise.
8604 (vst1q_p64_x3): Likewise.
8606 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8608 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8609 user_label_prefix to name.
8611 * tree-core.h: Update comment about the format of NAME string
8612 passed to handler in attribute_spec.
8614 * config/msp430/msp430.md: Remove erroneous subreg expression from
8615 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8616 zero_extend{q,h}isi2.
8618 2018-05-30 Borislav Petkov <bp@suse.de>
8620 * doc/extend.texi: Document some architecture specific
8621 constraints and sort entries.
8623 2018-05-30 Martin Sebor <msebor@redhat.com>
8626 * builtins.c (expand_builtin_stpcpy_1): New function.
8627 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8628 only if the former succeeds.
8630 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8632 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8635 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8637 * doc/invoke.texi (-flinker-output): Document
8639 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8641 * passes.c (ipa_write_summaries): Only modify statements if body
8643 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8644 incrementally linking.
8645 (ipa_passes): Likewise.
8646 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8647 pass down resolution info.
8648 * common.opt (flag_incremental_link): Update info.
8649 * gcc.c (plugin specs): Turn flinker-output=* to
8650 -plugin-opt=-linker-output-known
8651 * toplev.c (compile_file): Also cut compilation when doing incremental
8653 * flag-types. (enum lto_partition_model): Add
8654 LTO_LINKER_OUTPUT_NOLTOREL.
8655 (invoke.texi): Add -flinker-output docs.
8656 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8657 link same way as WPA; do not stream in dead initializers.
8659 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8660 darf2out_register_external_decl): Support incremental link.
8662 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8664 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8666 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8668 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8669 it down to simple_object_copy_lto_debug_sections.
8670 (run_gcc): Determine incremental LTO link time and configure
8671 lto1 into non-wpa mode, disable renaming of debug sections.
8673 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8675 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8676 descriptions of various incorrectly documented functions.
8678 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8681 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8682 address check not strict.
8684 2018-05-30 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/85964
8687 * tracer.c (better_p): Drop initialized count check, we only
8688 call the function with initialized counts now.
8689 (find_best_successor): Do find a best edge if one
8690 has uninitialized count.
8691 (find_best_predecessor): Likewise. Do BB frequency check only
8692 if count is initialized.
8694 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8696 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8697 (aarch64_ldrstr_offset_compare): New.
8698 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8699 load/store orderings.
8700 (aarch64_gen_adjusted_ldpstp): Likewise.
8702 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8704 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8705 Check for subset of GENERAL_REGS and FP_REGS.
8706 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8709 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8711 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8712 and wi::to_poly_offset. Add the current offset and then check
8713 whether the sum fits, rather than using an unchecked addition of
8714 a checked term. Check for a shwi rather than a uhwi.
8715 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8716 (store_constructor): Use poly_int_tree_p.
8717 (expand_expr_real_1): Likewise.
8718 * function.c (assign_temp): Likewise.
8719 * fold-const.c (const_binop): Use poly_int_tree_p and
8721 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8723 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8724 to_poly_offset for MEM offsets.
8725 * ipa-icf.c (sem_variable::equals): Likewise.
8726 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8727 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8728 wi::to_poly_offset for BIT_FIELD_REF offsets.
8729 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8731 * var-tracking.c (emit_note_insn_var_location): Use
8732 tree_to_poly_uint64.
8734 2018-05-29 Jim Wilson <jimw@sifive.com>
8736 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8738 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8741 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8742 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8744 (sse4_1_round<mode>2): Use nonimmediate_operand
8745 for operand 1 predicate.
8747 2018-05-29 Martin Sebor <msebor@redhat.com>
8748 Richard Biener <rguenther@suse.de>
8751 * calls.c (get_size_range): Call determine_value_range instead
8752 of get_value_range..
8753 * tree-vrp.h (determine_value_range): Declared new function.
8754 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8756 2018-05-29 Richard Biener <rguenther@suse.de>
8758 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8759 sure to use non-pattern stmts for get_earlier_stmt arguments.
8760 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8761 called on pattern stmts.
8762 (get_later_stmt): Likewise.
8764 2018-05-29 Martin Liska <mliska@suse.cz>
8766 PR gcov-profile/85759
8767 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8770 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8772 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8774 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8778 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8779 VEC_PACK_FLOAT_EXPR): New tree codes.
8780 * tree-pretty-print.c (op_code_prio): Handle
8781 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8782 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8783 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8784 * tree-inline.c (estimate_operator_cost): Likewise.
8785 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8786 * fold-const.c (const_binop): Likewise.
8787 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8788 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8789 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8790 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8791 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8792 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8793 * expr.c (expand_expr_real_2): Likewise.
8794 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8795 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8796 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8798 * optabs.c (expand_widen_pattern_expr): For
8799 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8800 sign from result type rather than operand's type.
8801 (expand_binop_directly): For vec_packu_float_optab and
8802 vec_packs_float_optab allow result type to be different from operand's
8804 * optabs-tree.c (optab_for_tree_code): Handle
8805 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8806 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8807 * tree-vect-generic.c (expand_vector_operations_1): Handle
8808 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8809 VEC_PACK_FLOAT_EXPR.
8810 * tree-vect-stmts.c (supportable_widening_operation): Handle
8812 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8813 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8814 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8815 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8816 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8818 (vec_pack<floatprefix>_float_<mode>): New expander.
8819 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8821 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8822 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8823 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8824 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8825 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8827 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8828 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8829 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8830 VEC_PACK_FLOAT_EXPR): Document.
8832 2018-05-29 Richard Biener <rguenther@suse.de>
8834 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8836 (stmt_vec_info_vec): Make pointer.
8837 (init_stmt_vec_info_vec): Remove.
8838 (free_stmt_vec_info_vec): Likewise.
8839 (set_stmt_vec_info_vec): New function.
8840 (free_stmt_vec_infos): Likewise.
8841 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8842 (set_vinfo_for_stmt): Likewise.
8843 (get_earlier_stmt): Likewise.
8844 (get_later_stmt): Likewise.
8845 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8846 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8847 (vec_info::~vec_info): Free stmt_vec_infos.
8848 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8849 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8850 (pass_slp_vectorize::execute): Likewise.
8851 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8852 (free_stmt_vec_info_vec): Likewise.
8853 (set_stmt_vec_info_vec): New function.
8854 (free_stmt_vec_infos): Likewise.
8855 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8856 the global stmt_vec_info_vec.
8857 * tree-parloops.c (gather_scalar_reductions): Use
8858 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8861 2018-05-29 Richard Biener <rguenther@suse.de>
8863 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8865 2018-05-29 Martin Liska <mliska@suse.cz>
8866 David Malcolm <dmalcolm@redhat.com>
8868 * vec.c (test_reverse): New.
8869 (vec_c_tests): Add new test.
8870 * vec.h (vl_ptr>::reverse): New function.
8872 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8874 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8876 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8879 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8881 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8883 2018-05-28 Richard Biener <rguenther@suse.de>
8885 PR tree-optimization/85933
8886 * tree-vect-data-refs.c (vect_record_base_alignments): Only
8887 look at stmts marked as vectorizable.
8889 2018-05-28 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/85934
8892 * tree-vect-generic.c (expand_vector_operations_1): Hoist
8893 vector boolean check before scalar optimization.
8895 2018-05-28 Jakub Jelinek <jakub@redhat.com>
8897 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8900 2018-05-28 Mark Wielaard <mark@klomp.org>
8902 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8903 if it is an expression containing a minus sign.
8905 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
8907 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8909 2018-05-27 Paul Koning <ni1d@arrl.net>
8911 * config/pdp11/pdp11.md (truncsihi2): Remove.
8913 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
8914 Chung-Ju Wu <jasonwucj@gmail.com>
8916 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8918 (unaligned_store_dw): Ditto.
8919 * config/nds32/nds32-memory-manipulation.c
8920 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8921 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8922 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8923 (emit_setmem_word_loop): Rename to ...
8924 (emit_setmem_doubleword_loop): ... this.
8925 (nds32_gen_dup_4_byte_to_word_value): New function.
8926 (nds32_gen_dup_8_byte_to_double_word_value): New function.
8927 (nds32_expand_setmem_loop): Refine implementation.
8928 (nds32_expand_setmem_loop_v3m): Ditto.
8929 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8932 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8934 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8936 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8938 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8939 (nds32_init_machine_status): Initialize machine->attr_naked_p and
8940 machine->attr_no_prologue_p.
8941 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8942 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8943 (nds32_expand_epilogue): Consider attr_naked_p.
8944 (nds32_expand_epilogue_v3pop): Likewise.
8945 (nds32_can_use_return_insn): Likewise.
8946 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8947 attr_no_prologue_p fields.
8948 * config/nds32/nds32.opt (mret-in-naked-func): New option.
8950 2018-05-27 Jakub Jelinek <jakub@redhat.com>
8953 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8955 * config/i386/sse.md
8956 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8958 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8960 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8962 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8964 (*<floatsuffix>floatv2div2sf2): Rename to ...
8965 (*float<floatunssuffix>v2div2sf2): ... this.
8966 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8967 (float<floatunssuffix>v2div2sf2_mask): ... this.
8968 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8969 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8970 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8972 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8974 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8976 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8978 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8980 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8982 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8983 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8984 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8985 gen_ufix_truncv8dfv8si2.
8986 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8987 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8988 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8989 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8990 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8991 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8992 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8993 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8995 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8999 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9001 2018-05-25 Jim Wilson <jimw@sifive.com>
9003 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9004 * config/riscv/riscv.c (struct machine_function): Add
9005 interrupt_handler_p and attribute_checked_p fields.
9006 (riscv_attribute_table): Add interrupt.
9007 (riscv_interrupt_type_p): New.
9008 (riscv_save_reg_p): Save extra regs for interrupt handler.
9009 (riscv_use_save_libcall): Return false for interrupt handler.
9010 (riscv_first_stack_step): Add forward declaration.
9011 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9012 for interrupt handler with large frame. Use it for saved reg list.
9013 (riscv_expand_prologue): Move flag_stack_usage_info support to
9014 eliminate duplication.
9015 (riscv_expand_epilogue): Generate mret for interrupt handler.
9016 (riscv_epilogue_uses): New.
9017 (riscv_can_use_return_insn): Return false for interrupt handler.
9018 (riscv_function_ok_for_sibcall): Likewise.
9019 (riscv_set_current_function): Add interrupt handler support.
9020 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9021 * config/riscv/riscv.md (UNSPECV_MRET): New.
9023 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9025 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9027 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9029 PR tree-optimization/85712
9030 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9031 this candidate has already been replaced in-situ by a copy.
9033 2018-05-25 Jason Merrill <jason@redhat.com>
9035 PR c++/80485 - inline function non-zero address.
9036 * symtab.c (nonzero_address): Check DECL_COMDAT.
9038 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9041 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9042 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9043 extension of SImode operation. Use const123_operand predicate.
9044 (*saddsi_1): Remove.
9045 (*saddl_se_1): Ditto.
9047 (*ssubl_se_1): Ditto.
9048 * config/alpha/predicates.md (const123_operand): New predicate.
9049 * config/alpha/constraints.md (P): Use IN_RANGE.
9051 2018-05-25 Richard Biener <rguenther@suse.de>
9053 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9055 (ref_maybe_used_by_stmt_p): Likewise.
9056 (stmt_may_clobber_ref_p): Likewise.
9057 (stmt_may_clobber_ref_p_1): Likewise.
9058 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9060 (ref_maybe_used_by_stmt_p): Likewise.
9061 (stmt_may_clobber_ref_p): Likewise.
9062 (stmt_may_clobber_ref_p_1): Likewise.
9063 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9064 the alias oracle to disambiguate DRs with stmts DR analysis
9066 (vect_analyze_data_refs): Do not give up on not analyzable
9067 DRs for BB vectorization. Remove code truncating the dataref
9070 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9073 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9074 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9075 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9077 2018-05-25 Richard Biener <rguenther@suse.de>
9079 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9080 function, combining stmt data ref gathering and fatal analysis
9082 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9083 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9084 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9085 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9086 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9088 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9090 PR tree-optimization/85720
9091 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9092 SCC if all partitions are builtins.
9093 (version_loop_by_alias_check): New parameter. Generate cancelable
9094 runtime alias check if all partitions are builtins.
9095 (distribute_loop): Update call to above function.
9097 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9099 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9100 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9101 (parm_default_def_partition_arg): Ditto.
9102 (set_parm_default_def_partition): Ditto.
9103 (get_parm_default_def_partitions): Ditto and make it static.
9104 (get_undefined_value_partitions): Ditto and make it static.
9105 (remove_ssa_form): Refactor call to init_var_map here.
9106 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9107 computation for loop region.
9108 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9109 (register_default_def): Delete.
9110 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9111 (parm_default_def_partition_arg): Ditto.
9112 (set_parm_default_def_partition): Ditto.
9113 (get_parm_default_def_partitions): Ditto and make it static.
9114 (get_undefined_value_partitions): Ditto and make it static.
9115 (coalesce_with_default, coalesce_with_default): Update comment.
9116 (create_coalesce_list_for_region): New func factored out from
9117 create_outofssa_var_map.
9118 (populate_coalesce_list_for_outofssa): New func factored out from
9119 create_outofssa_var_map and coalesce_ssa_name.
9120 (create_outofssa_var_map): Delete.
9121 (coalesce_ssa_name): Refactor to support live range computation.
9122 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9123 (get_parm_default_def_partitions): Delete.
9124 (get_undefined_value_partitions): Ditto.
9125 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9126 computation for loop region.
9127 (new_tree_live_info, loe_visit_block): Ditto.
9128 (live_worklist, set_var_live_on_entry): Ditto.
9129 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9130 * tree-ssa-live.h (struct _var_map): New fields.
9131 (init_var_map): Change decl.
9132 (region_contains_p): New.
9134 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9136 * tree-ssa-live.h (live_merge_and_clear): Delete.
9138 2018-05-25 Richard Biener <rguenther@suse.de>
9141 * tree-dump.c (dequeue_and_dump): Remove access to removed
9142 operand 2 of a SWITCH_EXPR.
9144 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9146 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9147 multiplication and division.
9148 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9149 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9150 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9151 (cond_udiv_optab, cond_umod_optab): New optabs.
9152 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9153 (IFN_COND_RDIV): New internal functions.
9154 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9155 TRUNC_MOD_EXPR and RDIV_EXPR.
9156 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9157 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9159 (SVE_INT_BINARY): Include mult.
9160 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9161 (optab, sve_int_op): Handle mult.
9162 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9164 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9165 for SVE_INT_BINARY_SD.
9167 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9169 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9170 (optab, sve_int_op): Handle div and udiv.
9171 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9172 for SVE_INT_BINARY_SD.
9173 (*<optab><mode>3): New insn for the same.
9175 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9177 * tree-vect-patterns.c: Include predict.h.
9178 (vect_recog_divmod_pattern): Restrict check for division support
9179 to when optimizing for size.
9181 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9183 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9184 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9185 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9186 (gimple_match_op::set_op): Likewise.
9187 (gimple_resimplify4): Declare.
9188 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9189 (expr::gen_transform): Likewise.
9190 (decision_tree::gen): Generate a simplification routine for 4 operands.
9191 * gimple-match-head.c (gimple_simplify): Add an overload for
9192 4 operands. In the top-level function, handle up to 4 call
9193 arguments and call gimple_resimplify4.
9194 (gimple_resimplify4): New function.
9195 (build_call_internal): Pass a fourth operand.
9196 (maybe_push_to_seq): Likewise.
9197 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9198 Fold VEC_COND_EXPRs of an operation and a default value into
9199 an IFN_COND_* function if possible.
9200 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9202 (SVE_COND_FP_BINARY): Include them.
9203 (optab, sve_fp_op): Handle them.
9204 (SVE_INT_BINARY_REV): New code iterator.
9205 (SVE_COND_FP_BINARY_REV): New int iterator.
9206 (commutative): New int attribute.
9207 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9209 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9211 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9212 (*cond_<optab><mode>): New patterns for reversed operands.
9214 2018-05-25 Richard Biener <rguenther@suse.de>
9216 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9217 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9218 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9219 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9220 * tree-vect-data-refs.c (everywhere): Adjust users.
9221 * tree-vect-loop.c (everywhere): Likewise.
9222 * tree-vect-slp.c (everywhere): Likewise.
9223 * tree-vect-stmts.c (everywhere): Likewise.
9224 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9226 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9228 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9230 (gcc_cv_as_section_exclude): ... this.
9231 Try Solaris as #exclude syntax.
9232 * configure: Regenerate.
9233 * config.in: Regenerate.
9234 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9236 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9237 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9239 * varasm.c (default_elf_asm_named_section): Don't check if
9240 HAVE_GAS_SECTION_EXCLUDE is defined.
9242 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9244 * doc/md.texi: Update the documentation of the cond_* optabs
9245 to mention the new final operand. Fix GET_MODE_NUNITS call.
9246 Describe the scalar case too.
9247 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9248 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9250 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9251 (get_conditional_internal_fn): Update comment.
9252 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9253 accumulator value as a final argument to conditional functions.
9254 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9255 a define_expand and add an "else" operand. Assert for now that
9256 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9257 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9258 (*cond_<optab><mode>): New patterns.
9259 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9260 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9261 (UNSPEC_COND_EOR): Delete.
9262 (optab): Remove associated mappings.
9263 (SVE_INT_BINARY): New code iterator.
9264 (sve_int_op): Remove int attribute and add "minus" to the code
9266 (SVE_COND_INT_OP): Delete.
9267 (SVE_COND_FP_OP): Rename to...
9268 (SVE_COND_FP_BINARY): ...this.
9270 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9272 * optabs.c (can_reuse_operands_p): New function.
9273 (maybe_legitimize_operands): Try to reuse the results for
9276 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9278 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9279 Add {q} suffix to insn mnemonic.
9281 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9283 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9284 (msp430_warn_func_return): New.
9286 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9288 * fold-const.c (tree_nonzero_bits): New function.
9289 * fold-const.h (tree_nonzero_bits): Likewise.
9290 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9291 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9293 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9297 * varasm.c (assemble_alias): Check ifunc_resolver only on
9300 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9303 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9304 when memory input operand is handled.
9306 2018-05-24 Luis Machado <luis.machado@linaro.org>
9308 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9310 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9312 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9314 * match.pd: Delay FMA folds until after vectorization.
9316 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9319 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9320 address check not strict.
9322 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9324 * gimple-match.h (gimple_match_op): New class.
9325 (mprts_hook): Replace parameters with a gimple_match_op *.
9326 (maybe_build_generic_op): Likewise.
9327 (gimple_simplified_result_is_gimple_val): Replace parameters with
9328 a const gimple_match_op *.
9329 (gimple_simplify): Replace code_helper * and tree * parameters with
9330 a gimple_match_op * parameter.
9331 (gimple_resimplify1): Replace code_helper *, tree and tree *
9332 parameters with a gimple_match_op * parameter.
9333 (gimple_resimplify2): Likewise.
9334 (gimple_resimplify3): Likewise.
9335 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9336 parameters with a gimple_match_op * parameter.
9337 * gimple-match-head.c (gimple_simplify): Change prototypes of
9338 auto-generated functions to take a gimple_match_op * instead of
9339 separate code_helper * and tree * parameters. Make the same
9340 change in the top-level overload and update calls to the
9341 gimple_resimplify routines. Update calls to the auto-generated
9342 functions and to maybe_push_res_to_seq in the publicly-facing
9343 operation-specific gimple_simplify overloads.
9344 (gimple_match_op::MAX_NUM_OPS): Define.
9345 (gimple_resimplify1): Replace rcode and ops with a single res_op
9346 parameter. Update call to gimple_simplify.
9347 (gimple_resimplify2): Likewise.
9348 (gimple_resimplify3): Likewise.
9349 (mprts_hook): Replace parameters with a gimple_match_op *.
9350 (maybe_build_generic_op): Likewise.
9351 (build_call_internal): Replace type, nargs and ops with
9352 a gimple_match_op *.
9353 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9354 with a single gimple_match_op *. Update calls to mprts_hook,
9355 build_call_internal and gimple_simplified_result_is_gimple_val.
9356 Factor out code that is common to the tree_code and combined_fn cases.
9357 * genmatch.c (expr::gen_transform): Replace tem_code and
9358 tem_ops with a gimple_match_op called tem_op. Update calls
9359 to the gimple_resimplify functions and maybe_push_res_to_seq.
9360 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9361 res_ops. Update call to the gimple_resimplify functions.
9362 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9363 (decision_tree::gen): Make the functions take a gimple_match_op *
9364 called res_op instead of separate res_code and res_ops parameters.
9365 Update call accordingly.
9366 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9367 and ops with a single res_op parameter. Update calls to
9368 maybe_build_generic_op and maybe_push_res_to_seq.
9369 (fold_stmt_1): Update calls to gimple_simplify and
9370 replace_stmt_with_simplification.
9371 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9372 and gimple_simplified_result_is_gimple_val.
9373 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9375 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9376 with a gimple_match_op *.
9377 (vn_nary_build_or_lookup): Likewise. Update call to
9378 vn_nary_build_or_lookup_1.
9379 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9380 gimple_match_op *. Update calls to the gimple_resimplify routines
9381 and to gimple_simplified_result_is_gimple_val.
9382 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9383 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9384 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9385 (visit_nary_op): Likewise.
9386 (visit_reference_op_load): Likewise.
9388 2018-05-23 Luis Machado <luis.machado@linaro.org>
9390 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9391 modifier for printing the step amount.
9393 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9396 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9399 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9401 * doc/sourcebuild.texi (Endianness): New subsubsection.
9403 2018-05-23 Luis Machado <luis.machado@linaro.org>
9405 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9406 <prefetch_dynamic_strides>: New const bool field.
9407 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9408 prefetch_dynamic_strides.
9409 (exynosm1_prefetch_tune): Likewise.
9410 (thunderxt88_prefetch_tune): Likewise.
9411 (thunderx_prefetch_tune): Likewise.
9412 (thunderx2t99_prefetch_tune): Likewise.
9413 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9415 (aarch64_override_options_internal): Update to set
9416 PARAM_PREFETCH_DYNAMIC_STRIDES.
9417 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9418 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9419 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9420 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9421 prefetch-dynamic-strides setting.
9423 2018-05-23 Luis Machado <luis.machado@linaro.org>
9425 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9426 <minimum_stride>: New const int field.
9427 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9428 minimum_stride field defaulting to -1.
9429 (exynosm1_prefetch_tune): Likewise.
9430 (thunderxt88_prefetch_tune): Likewise.
9431 (thunderx_prefetch_tune): Likewise.
9432 (thunderx2t99_prefetch_tune): Likewise.
9433 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9434 <default_opt_level>: Set to 3.
9435 (aarch64_override_options_internal): Update to set
9436 PARAM_PREFETCH_MINIMUM_STRIDE.
9437 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9438 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9439 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9440 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9441 stride is constant and is below the minimum stride threshold.
9443 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9445 * config/arm/arm-cpus.in (mode26): Delete.
9446 (armv4): Delete mode26 reference.
9447 * config/arm/arm.c (arm_configure_build_target): Delete use of
9450 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9452 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9454 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9455 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9457 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9458 (floatunsdidf2): Ditto.
9460 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9462 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9463 (fixuns_trunc<mode>si2_avx512f): Ditto.
9464 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9465 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9466 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9468 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9470 PR rtl-optimization/79985
9471 * df-scan.c (df_insn_refs_collect): Remove special case for
9472 global registers and asm statements.
9474 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9476 * extend.texi (Global Register Variables): Rewrite the bullet list.
9477 Note that the register is available for allocation. Note that access
9478 via inline asm must use constraints. Add note about async-signal
9479 handlers. Remove paragraph about automagic register selection.
9481 2018-05-23 Richard Biener <rguenther@suse.de>
9483 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9484 of fixed offset from memset VN.
9486 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9488 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9490 (alloc_cand_and_find_basis): Initialize first_interp field.
9491 (slsr_process_mul): Modify first_interp field.
9492 (slsr_process_add): Likewise.
9493 (slsr_process_cast): Modify first_interp field for each new
9495 (slsr_process_copy): Likewise.
9496 (dump_candidate): Dump first_interp field.
9497 (replace_mult_candidate): Process all interpretations, not just
9499 (replace_rhs_if_not_dup): Likewise.
9500 (replace_one_candidate): Likewise.
9502 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9504 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9506 (aarch64_needs_frame_chain): New function.
9507 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9509 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9512 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9513 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9514 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9515 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9516 as true for strict-align.
9517 (aarch64_can_inline_p): Perform checks even when callee has no
9518 attributes to check for strict alignment.
9519 * doc/extend.texi (AArch64 Function Attributes): Document
9521 * doc/invoke.texi: (AArch64 Options): Likewise.
9523 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9525 PR tree-optimization/85853
9526 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9527 the handling of the root of the node to...
9528 (vect_slp_analyze_node_operations_1): ...this new function,
9529 and run the whole thing with the child nodes' def types
9530 set according to their SLP node's def type.
9532 2018-05-23 Richard Biener <rguenther@suse.de>
9535 * tree-data-ref.c (create_runtime_alias_checks): Defer
9536 and ignore overflow warnings.
9538 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9540 PR tree-optimization/85822
9541 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9544 2018-05-23 Richard Biener <rguenther@suse.de>
9546 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9547 memset constants via native_interpret_expr.
9549 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9552 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9554 (cgraph_node::create_alias): Likewise.
9555 (cgraph_node::get_availability): Check ifunc_resolver instead
9556 of looking up ifunc attribute.
9557 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9558 * varasm.c (do_assemble_alias): Likewise.
9559 (assemble_alias): Likewise.
9560 (default_binds_local_p_3): Likewise.
9561 * cgraph.h (cgraph_node): Add ifunc_resolver.
9562 (cgraph_node::only_called_directly_or_aliased_p): Return false
9564 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9566 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9567 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9568 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9569 instead of looking up ifunc attribute.
9571 2018-05-22 Luis Machado <luis.machado@linaro.org>
9573 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9575 2018-05-22 Martin Sebor <msebor@redhat.com>
9578 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9579 only when expasion succeeds.
9580 (expand_builtin_strcmp): Same.
9581 (expand_builtin_strncmp): Same.
9583 2018-05-22 Martin Sebor <msebor@redhat.com>
9585 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9587 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9588 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9590 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9591 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9593 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9594 Add check for legitimate_address.
9595 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9596 (aarch64_swap_ldrstr_operands): New.
9597 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9600 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9601 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9603 * config/aarch64/aarch64.md: New patterns to generate stp
9605 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9606 single words and double words.
9607 (load_pair_sw, load_pair_dw): Likewise.
9608 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9610 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9612 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9613 for different mode ldpstp and add peephole for merged zero stores.
9615 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9617 (aarch64_gen_store_pair): Rename calls to match new patterns.
9618 (aarch64_gen_load_pair): Rename calls to match new patterns.
9619 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9620 (load_pair<DREG:mode><DREG2:mode>): ... This.
9621 (store_pair<mode>): Rename to...
9622 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9623 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9625 (V_INT_EQUIV): Handle SImode.
9626 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9629 2018-05-22 Martin Sebor <msebor@redhat.com>
9632 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9633 or ajust the presumed bound on an operation to avoid unnecessary
9636 2018-05-22 Martin Sebor <msebor@redhat.com>
9638 PR tree-optimization/85826
9639 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9640 assuming that a DECL necesarily has a constant size.
9642 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9645 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9647 2018-05-22 Richard Biener <rguenther@suse.de>
9649 PR tree-optimization/85834
9650 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9651 non-constant and non-zero memset arguments.
9653 2018-05-22 Martin Liska <mliska@suse.cz>
9656 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9658 2018-05-22 Richard Biener <rguenther@suse.de>
9660 PR tree-optimization/85863
9661 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9662 comparisons when vectype is specified.
9663 (vectorizable_condition): Do not specify vectype for
9664 vect_is_simple_cond when SLP vectorizing.
9666 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9669 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9670 define __ibm128 as long double.
9671 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9672 as a distinct type when IEEE 128-bit support is enabled.
9673 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9674 128-bit types to use the correct functions.
9675 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9676 convert between 128-bit floating point types that have different
9677 modes but the same representation, instead of using gen_lowpart to
9679 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9681 (IFKF_reg): New attributes to give the register constraints for
9683 (extend<mode>tf2_internal): New insns to mark an explicit
9684 conversion between 128-bit floating point types that have a
9685 different mode but share the same representation.
9687 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9689 PR tree-optimization/85814
9690 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9691 a null return from get_strinfo when unsharing the next
9692 strinfo in the chain.
9694 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9697 * varasm.c (weak_finish): Clean up weak_decls.
9699 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9701 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9702 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9703 UNSPEC_UADALP values.
9704 * config/aarch64/iterators.md (ABAL): New int iterator.
9707 (sur): Add mappings for the above.
9708 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9710 (aarch64_<sur>abal<mode>_4): Likewise.
9711 (aarch64_<sur>adalp<mode>_3): Likewise.
9712 (<sur>sadv16qi): New define_expand.
9714 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9716 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9717 (*movdf_internal): Ditto.
9718 (*rcpsf2_sse): Ditto.
9719 (*rsqrtsf2_sse): Ditto.
9720 (*sqrt<mode>2_sse): Ditto.
9722 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9724 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9726 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9727 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9728 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9729 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9731 * config/aarch64/arm_neon.h: Likewise.
9732 * config/aarch64/iterators.md (VQ_I): New.
9734 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9736 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9738 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9741 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9743 * config/nds32/constraints.md (S): New constraint.
9744 * config/nds32/nds32.md (call_internal): Use constraint S.
9745 (call_value_internal): Likewise.
9746 (sibcall_internal): Likewise.
9747 (sibcall_value_internal): Likewise.
9749 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9750 Chung-Ju Wu <jasonwucj@gmail.com>
9752 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9755 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9756 Chung-Ju Wu <jasonwucj@gmail.com>
9758 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9759 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9760 (nds32_rtx_costs_impl): Simplify.
9761 (nds32_address_cost_impl): Simplify.
9762 (nds32_init_rtx_costs): New function.
9763 (nds32_rtx_costs_speed_prefer): Likewise.
9764 (nds32_rtx_costs_size_prefer): Likewise.
9765 (nds32_address_cost_speed_prefer): Likewise.
9766 (nds32_address_cost_speed_fwprop): Likewise.
9767 (nds32_address_cost_size_prefer): Likewise.
9768 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9769 * config/nds32/nds32.c (nds32_option_override): Use
9770 nds32_init_rtx_costs function.
9772 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9774 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9775 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9776 (TARGET_PIPELINE_N8): Likewise.
9777 (TARGET_PIPELINE_N10): Likewise.
9778 (TARGET_PIPELINE_N13): Likewise.
9779 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9781 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9783 * config/nds32/nds32-fpu.md: Update copyright year.
9785 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9787 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9789 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9791 * config/nds32/nds32.c
9792 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9793 * config/nds32/nds32.opt (minline-asm-r15): New option.
9795 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9797 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9799 * config/nds32/nds32.md (abssi2): New pattern.
9801 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9803 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9804 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9805 Merge insn pattern from sse_cvtsi2ss<round_name> and
9806 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9807 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9808 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9809 using SWI48 mode iterator.
9810 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9811 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9812 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9813 pattern from sse_cvttss2si<round_saeonly_name>
9814 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9815 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9816 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9817 using SWI48 mode iterator.
9818 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9819 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9820 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9821 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9822 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9823 using SWI48 mode iterator.
9824 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9825 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9826 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9827 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9828 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9829 SWI48 mode iterator.
9830 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9831 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9832 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9833 pattern from sse_cvttsd2si<round_saeonly_name>
9834 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9836 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9838 * config/nds32/nds32-md-auxiliary.c
9839 (nds32_valid_smw_lwm_base_p): Refine.
9840 (nds32_output_smw_single_word): Refine.
9841 (nds32_output_smw_double_word): New.
9842 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9844 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9846 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9847 (nds32_output_stack_pop): Refine.
9848 (nds32_expand_unaligned_load): Refine.
9849 (nds32_expand_unaligned_store): Refine.
9851 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9852 Chung-Ju Wu <jasonwucj@gmail.com>
9854 * config/nds32/constants.md: Add TP_REGNUM constant.
9855 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9856 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9858 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9859 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9860 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9861 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9862 and PIC code generation.
9863 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9865 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9867 * config/nds32/nds32.md: Support TLS and PIC.
9868 * config/nds32/nds32.c: Support TLS and PIC.
9869 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9870 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9873 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9875 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9876 mode with E_ prefix.
9878 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9879 Chung-Ju Wu <jasonwucj@gmail.com>
9881 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9882 * config/nds32/nds32-md-auxiliary.c
9883 (symbolic_reference_mentioned_p): New.
9884 (nds32_legitimize_ict_address): New.
9885 (nds32_expand_ict_move): New.
9886 (nds32_indirect_call_referenced_p): New.
9887 (nds32_symbol_binds_local_p): Delete.
9888 (nds32_long_call_p): Modify.
9889 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9890 * config/nds32/nds32-protos.h
9891 (symbolic_reference_mentioned_p): Declare.
9892 (nds32_legitimize_ict_address): Declare.
9893 (nds32_expand_ict_move): Declare.
9894 (nds32_indirect_call_referenced_p): Declare.
9895 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9896 (nds32_relax_group): Use nds32_ict_const_p as condition.
9897 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9898 (nds32_asm_file_start): Output ict_model directive in asm code.
9899 (nds32_legitimate_address_p): Consider indirect call.
9900 (nds32_print_operand): Consider indirect call.
9901 (nds32_print_operand_address): Consider indirect call.
9902 (nds32_insert_attributes): Handle "indirect_call" attribute.
9903 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9904 (TARGET_LEGITIMATE_CONSTANT_P): Define.
9905 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9906 (TARGET_DELEGITIMIZE_ADDRESS): Define.
9907 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9908 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9909 (TARGET_ICT_MODEL_SMALL): Define.
9910 (TARGET_ICT_MODEL_LARGE): Define.
9911 * config/nds32/nds32.md (movsi): Consider ict model.
9912 (call, call_value): Consider ict model.
9913 (sibcall, sibcall_value): Consider ict model.
9914 * config/nds32/nds32.opt (mict-model): New option.
9915 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9918 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
9919 Monk Chiang <sh.chiang04@gmail.com>
9920 Jim Wilson <jimw@sifive.com>
9922 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9923 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
9924 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9925 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9926 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
9927 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9928 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9929 compute save_libcall_adjustment properly.
9930 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9931 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9932 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9933 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9934 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9935 (ABI_SPEC): Handle mabi=ilp32e.
9936 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9937 (RVE): Add RVE mask.
9938 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9939 <-march>: Add rv32e as an example.
9941 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
9944 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9945 (intra_create_variable_infos): Handle C++ constructors.
9947 2018-05-18 Martin Liska <mliska@suse.cz>
9949 * passes.def: Remove a redundant pass.
9951 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
9954 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9956 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9958 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9960 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9962 (armv2, armv2a, armv3, armv3m): Delete architectures.
9963 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9964 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9965 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9967 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9968 (*mulsidi3adddi): Likewise.
9969 (mulsidi3): Likewise.
9970 (*mulsidi3_nov6): Likewise.
9971 (umulsidi3): Likewise.
9972 (umulsidi3_nov6): Likewise.
9973 (umaddsidi4): Likewise.
9974 (*umulsidi3adddi): Likewise.
9975 (smulsi3_highpart): Likewise.
9976 (*smulsi3_highpart_nov6): Likewise.
9977 (umulsi3_highpart): Likewise.
9978 (*umulsi3_highpart_nov6): Likewise.
9979 * config/arm/arm.h (arm_arch3m): Delete.
9980 * config/arm/arm.c (arm_arch3m): Delete.
9981 (arm_option_override_internal): Update armv3-related comment.
9982 (arm_configure_build_target): Delete use of isa_bit_mode32.
9983 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9984 (arm_rtx_costs_internal): Delete check of arm_arch3m.
9985 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9987 (mulusa3): Likewise.
9988 * config/arm/arm-protos.h (arm_arch3m): Delete.
9989 * config/arm/arm-tables.opt: Regenerate.
9990 * config/arm/arm-tune.md: Likewise.
9991 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9992 deleted architectures.
9994 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9996 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9997 (armv5t, armv5te): New features.
9998 (ARMv5, ARMv5e): Delete fgroups.
9999 (ARMv5t, ARMv5te): Adjust for above changes.
10000 (ARMv6m): Likewise.
10001 (armv5, armv5e): Delete arches.
10002 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10004 (*call_reg_arm): Likewise.
10005 (*call_value_reg_armv5): Likewise.
10006 (*call_value_reg_arm): Likewise.
10007 (*call_symbol): Likewise.
10008 (*call_value_symbol): Likewise.
10009 (*sibcall_insn): Likewise.
10010 (*sibcall_value_insn): Likewise.
10011 (clzsi2): Likewise.
10012 (prefetch): Likewise.
10013 (define_split and define_peephole2 dependent on arm_arch5):
10015 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10017 (TARGET_ARM_QBIT): Likewise.
10018 (TARGET_DSP_MULTIPLY): Likewise.
10019 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10020 (arm_arch5, arm_arch5e): Delete.
10021 (arm_arch5t, arm_arch5te): Declare.
10022 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10023 (arm_arch5t): Declare.
10024 (arm_option_reconfigure_globals): Update for the above.
10025 (arm_options_perform_arch_sanity_checks): Update comment, replace
10026 use of arm_arch5 with arm_arch5t.
10027 (use_return_insn): Likewise.
10028 (arm_emit_call_insn): Likewise.
10029 (output_return_instruction): Likewise.
10030 (arm_final_prescan_insn): Likewise.
10031 (arm_coproc_builtin_available): Likewise.
10032 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10033 arm_arch5e with arm_arch5t and arm_arch5te.
10034 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10035 (arm_arch5t, arm_arch5te): Declare.
10036 * config/arm/arm-tables.opt: Regenerate.
10037 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10038 * config/arm/t-multilib: Likewise.
10039 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10040 instead of arm_arch5.
10041 (*call_reg_thumb1): Likewise.
10042 (*call_value_reg_thumb1_v5): Likewise.
10043 (*call_value_reg_thumb1): Likewise.
10044 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10046 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10048 2018-05-18 Martin Liska <mliska@suse.cz>
10050 PR gcov-profile/84846
10051 * doc/gcov.texi: Document -t option of gcov tool.
10053 2018-05-18 Martin Liska <mliska@suse.cz>
10055 PR gcov-profile/84846
10056 * gcov.c (print_usage): Add new -t option.
10057 (process_args): Handle the option.
10058 (generate_results): Use stdout as output when requested by
10061 2018-05-18 Martin Liska <mliska@suse.cz>
10063 PR gcov-profile/84846
10064 * coverage.c (coverage_init): Write PWD to .gcno file.
10065 * doc/gcov.texi: Document how working directory is printed.
10066 * gcov-dump.c (dump_gcov_file): Print PWD.
10067 * gcov.c (output_intermediate_file): Likewise.
10068 (read_graph_file): Read PWD string.
10069 (output_lines): Print PWD.
10071 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10073 PR middle-end/85817
10074 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10075 for retval and return false if all args to phi are zero.
10077 2018-05-18 Richard Biener <rguenther@suse.de>
10079 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10081 (evrp_dom_walker::before_dom_children): Call it.
10083 2018-05-18 Richard Biener <rguenther@suse.de>
10085 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10086 results when processing array refs with variable index.
10088 2018-05-18 Toon Moene <toon@moene.org>
10090 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10091 directly after that of -floop-interchange. Indicate that both
10092 options are enabled by default when specifying -O3.
10094 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10096 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10097 iterator. Delete separate integer-mode vec_set<mode> expander.
10098 (aarch64_simd_vec_setv2di): Delete.
10099 (vec_setv2di): Delete.
10100 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10101 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10102 the "w, r" alternative.
10104 2018-05-18 Martin Liska <mliska@suse.cz>
10106 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10107 * tree-pass.h (make_pass_lower_switch_O0): New function.
10108 * tree-switch-conversion.c (node_has_low_bound): Remove.
10109 (node_has_high_bound): Likewise.
10110 (node_is_bounded): Likewise.
10111 (class pass_lower_switch): Make it a template type and create
10113 (pass_lower_switch::execute): Add template argument.
10114 (make_pass_lower_switch): New function.
10115 (make_pass_lower_switch_O0): New function.
10116 (do_jump_if_equal): Remove.
10117 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10118 all the hard work to tree optimization passes.
10120 2018-05-18 Martin Liska <mliska@suse.cz>
10122 * dbgcnt.c (limit_low): Renamed from limit.
10123 (limit_high): New variable.
10124 (dbg_cnt_is_enabled): Check for upper limit.
10125 (dbg_cnt): Adjust dumping.
10126 (dbg_cnt_set_limit_by_index): Add new argument for high
10128 (dbg_cnt_set_limit_by_name): Likewise.
10129 (dbg_cnt_process_single_pair): Parse new format.
10130 (dbg_cnt_process_opt): Use strtok.
10131 (dbg_cnt_list_all_counters): Remove 'value' and add
10133 * doc/invoke.texi: Document changes.
10135 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10137 * doc/sourcebuild.texi (scalar_all_fma): Document.
10138 * tree.def (FMA_EXPR): Delete.
10139 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10140 * internal-fn.c (ternary_direct): New macro.
10141 (expand_ternary_optab_fn): Likewise.
10142 (direct_ternary_optab_supported_p): Likewise.
10143 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10144 * builtins.c (fold_builtin_fma): Delete.
10145 (fold_builtin_3): Don't call it.
10146 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10147 * expr.c (expand_expr_real_2): Likewise.
10148 * fold-const.c (operand_equal_p): Likewise.
10149 (fold_ternary_loc): Likewise.
10150 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10151 * gimple.c (DEFTREECODE): Likewise.
10152 * gimplify.c (gimplify_expr): Likewise.
10153 * optabs-tree.c (optab_for_tree_code): Likewise.
10154 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10155 * tree-eh.c (operation_could_trap_p): Likewise.
10156 (stmt_could_throw_1_p): Likewise.
10157 * tree-inline.c (estimate_operator_cost): Likewise.
10158 * tree-pretty-print.c (dump_generic_node): Likewise.
10159 (op_code_prio): Likewise.
10160 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10161 * tree-ssa-operands.c (get_expr_operands): Likewise.
10162 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10163 * fold-const-call.h (fold_fma): Delete.
10164 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10165 CFN_FNMA and CFN_FNMS.
10166 (fold_fma): Delete.
10167 * genmatch.c (combined_fn): New enum.
10168 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10169 (commutative_op): New function.
10170 (commutate): Use it. Handle more than 2 operands.
10171 (dt_operand::gen_gimple_expr): Use commutative_op.
10172 (parser::parse_expr): Allow :c to be used with non-binary
10173 operators if the commutative operand is known.
10174 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10175 CFN_FMS, CFN_FNMA and CFN_FNMS.
10176 (backprop::process_assign_use): Remove FMA_EXPR handling.
10177 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10178 (gen_hsa_fma): New function.
10179 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10180 IFN_FNMA and IFN_FNMS.
10181 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10182 * gimple-fold.h (follow_all_ssa_edges): Declare.
10183 * gimple-fold.c (follow_all_ssa_edges): New function.
10184 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10185 gimple_build interface and use follow_all_ssa_edges to fold the result.
10186 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10187 instead of checking for optabs directly.
10188 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10189 rather than FMA_EXPRs.
10190 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10191 call to IFN_FMA instead of an FMA_EXPR.
10193 2018-05-17 Jim Wilson <jimw@sifive.com>
10195 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10196 sign extended promoted subreg, and the range does not have the sign bit
10197 set, then do a sign extend.
10199 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10200 test, check for sign extended subreg and/or constant operands, and
10201 do a sign extend in that case.
10203 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10205 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10206 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10208 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10209 Change logics_shift_reg to logics_shift_imm.
10210 (thunderx2t99_fp_loadpair_basic): Delete.
10211 (thunderx2t99_fp_storepair_basic): Delete.
10212 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10213 (thunderx2t99_asimd_polynomial): Delete.
10214 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10215 and neon_fp_mul_d_scalar_q.
10216 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10217 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10218 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10219 (thunderx2t99_asimd_lut): Add missing tbl types.
10220 (thunderx2t99_asimd_ext): Delete.
10221 (thunderx2t99_asimd_load1_1_mult): Delete.
10222 (thunderx2t99_asimd_load1_2_mult): Delete.
10223 (thunderx2t99_asimd_load1_ldp): New.
10224 (thunderx2t99_asimd_load1): New.
10225 (thunderx2t99_asimd_load2): Add missing *load2* types.
10226 (thunderx2t99_asimd_load3): New.
10227 (thunderx2t99_asimd_load4): New.
10228 (thunderx2t99_asimd_store1_1_mult): Delete.
10229 (thunderx2t99_asimd_store1_2_mult): Delete.
10230 (thunderx2t99_asimd_store2_mult): Delete.
10231 (thunderx2t99_asimd_store2_onelane): Delete.
10232 (thunderx2t99_asimd_store_stp): New.
10233 (thunderx2t99_asimd_store1): New.
10234 (thunderx2t99_asimd_store2): New.
10235 (thunderx2t99_asimd_store3): New.
10236 (thunderx2t99_asimd_store4): New.
10238 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10240 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10241 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10243 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10244 Segher Boessenkool <segher@kernel.crashing.org>
10247 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10250 2018-05-17 Richard Biener <rguenther@suse.de>
10252 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10253 for pruning loop and prune defs feeding only already visited PHIs.
10255 2018-05-17 Richard Biener <rguenther@suse.de>
10257 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10259 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10260 Richard Biener <rguenther@suse.de>
10262 PR tree-optimization/85793
10263 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10264 for VMAT_ELEMENTWISE.
10266 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10268 * internal-fn.h (lookup_internal_fn): Declare
10269 * internal-fn.c (lookup_internal_fn): New function.
10270 * gimple.c (gimple_build_call_from_tree): Handle calls to
10271 internal functions.
10272 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10273 internal function names.
10274 * tree-pretty-print.c (dump_generic_node): Likewise.
10275 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10277 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10279 * gimple-fold.h (gimple_build): Make the function forms take
10280 combined_fn rather than built_in_function.
10281 (gimple_simplify): Likewise.
10282 * gimple-match-head.c (gimple_simplify): Likewise.
10283 * gimple-fold.c (gimple_build): Likewise.
10284 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10285 rather than gimple_build_call_internal.
10286 (get_initial_defs_for_reduction): Likewise.
10287 (vect_create_epilog_for_reduction): Likewise.
10288 (vectorizable_live_operation): Likewise.
10290 2018-05-17 Martin Liska <mliska@suse.cz>
10292 * gimple-ssa-sprintf.c (format_directive): Do not use
10293 space in between 'G_' and '('.
10295 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10298 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10299 even if the mask is not all ones.
10302 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10304 (ix86_gimple_fold_builtin): Likewise.
10307 * config/i386/i386.c: Include tree-vector-builder.h.
10308 (ix86_vector_shift_count): New function.
10309 (ix86_fold_builtin): Fold shift builtins by scalar count.
10310 (ix86_gimple_fold_builtin): Likewise.
10312 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10313 _mm512_setzero): New intrinsics.
10315 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10316 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10318 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10319 code generation for cases where splatting a value is not useful.
10320 * simplify-rtx.c (simplify_ternary_operation): Simplify
10321 vec_merge across a vec_duplicate and a paradoxical subreg forming
10322 a vector mode to a vec_concat.
10324 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10326 * config.gcc: Support "goldmont-plus".
10327 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10329 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10330 PROCESSOR_GOLDMONT_PLUS.
10331 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10332 (processor_target_table): Add "goldmont-plus".
10333 (PTA_GOLDMONT_PLUS): Define.
10334 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10335 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10336 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10337 (fold_builtin_cpu): Add "goldmont-plus".
10338 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10339 (ix86_option_override_internal): Add "goldmont-plus".
10340 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10341 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10342 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10343 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10345 2018-05-17 Richard Biener <rguenther@suse.de>
10347 PR tree-optimization/85757
10348 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10349 remove defs that only feed that PHI from further processing.
10351 2018-05-16 Jim Wilson <jimw@sifive.com>
10353 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10355 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10357 2018-05-16 Mark Wielaard <mark@klomp.org>
10359 * dwarf2out.c (count_index_strings): New function.
10360 (output_indirect_strings): Call count_index_strings and generate
10361 header for dwarf_version >= 5.
10363 2018-05-16 Mark Wielaard <mark@klomp.org>
10365 * dwarf2out.c (dwarf_FORM): New function.
10366 (set_indirect_string): Use dwarf_FORM.
10367 (reset_indirect_string): Likewise.
10368 (size_of_die): Likewise.
10369 (value_format): Likewise.
10370 (output_die): Likewise.
10371 (add_skeleton_AT_string): Likewise.
10372 (output_macinfo_op): Likewise.
10373 (index_string): Likewise.
10374 (output_index_string_offset): Likewise.
10375 (output_index_string): Likewise.
10376 (count_index_strings): Likewise.
10378 2018-05-16 Carl Love <cel@us.ibm.com>
10380 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10381 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10383 2018-05-16 Martin Jambor <mjambor@suse.cz>
10385 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10386 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10388 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10390 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10391 (fnma<mode>4): Likewise.
10392 (fms<mode>4): Likewise.
10393 (fnms<mode>4): Likewise.
10394 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10395 (aarch64_fnma<mode>4): Likewise.
10396 (aarch64_fms<mode>4): Likewise.
10397 (aarch64_fnms<mode>4): Likewise.
10398 (aarch64_fnmadd<mode>4): Likewise.
10400 2018-05-16 Jason Merrill <jason@redhat.com>
10402 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10404 2018-05-16 Richard Biener <rguenther@suse.de>
10406 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10407 (dump_stmt_cost): Declare.
10408 (add_stmt_cost): Dump cost we add.
10409 (add_stmt_costs): New function.
10410 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10411 No longer exported.
10412 (vect_analyze_stmt): Adjust prototype.
10413 (vectorizable_condition): Likewise.
10414 (vectorizable_live_operation): Likewise.
10415 (vectorizable_reduction): Likewise.
10416 (vectorizable_induction): Likewise.
10417 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10418 cost vector to pass to vectorizable_ and record afterwards.
10419 (vect_model_reduction_cost): Take cost vector argument and adjust.
10420 (vect_model_induction_cost): Likewise.
10421 (vectorizable_reduction): Likewise.
10422 (vectorizable_induction): Likewise.
10423 (vectorizable_live_operation): Likewise.
10424 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10425 SLP_TREE_NUMBER_OF_VEC_STMTS.
10426 (vect_analyze_slp_cost_1): Remove.
10427 (vect_analyze_slp_cost): Likewise.
10428 (vect_slp_analyze_node_operations): Take visited args and
10429 a target cost vector. Avoid processing already visited stmt sets.
10430 (vect_slp_analyze_operations): Use a local cost vector to gather
10431 costs and register those of non-discarded instances.
10432 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10433 (vect_schedule_slp_instance): Remove copying of
10434 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10436 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10437 adding cost. Record cost entry location.
10438 (vect_prologue_cost_for_slp_op): Function to compute cost of
10439 a constant or invariant generated for SLP vect in the prologue,
10440 split out from vect_analyze_slp_cost_1.
10441 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10442 (vect_model_promotion_demotion_cost): Likewise.
10443 (vect_model_store_cost): Likewise, make static.
10444 (vect_model_load_cost): Likewise.
10445 (vectorizable_bswap): Add cost vector arg and adjust.
10446 (vectorizable_call): Likewise.
10447 (vectorizable_simd_clone_call): Likewise.
10448 (vectorizable_conversion): Likewise.
10449 (vectorizable_assignment): Likewise.
10450 (vectorizable_shift): Likewise.
10451 (vectorizable_operation): Likewise.
10452 (vectorizable_store): Likewise.
10453 (vectorizable_load): Likewise.
10454 (vectorizable_condition): Likewise.
10455 (vectorizable_comparison): Likewise.
10456 (can_vectorize_live_stmts): Likewise.
10457 (vect_analyze_stmt): Likewise.
10458 (vect_transform_stmt): Adjust calls to vectorizable_*.
10459 * tree-vectorizer.c: Include gimple-pretty-print.h.
10460 (dump_stmt_cost): New function.
10462 2018-05-16 Richard Biener <rguenther@suse.de>
10464 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10465 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10466 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10467 (check_name): New callback.
10468 (dse_classify_store): Track cycles via a visited bitmap of PHI
10469 defs and simplify handling of in-loop and across loop dead stores
10470 and properly fail for loop-variant refs. Handle byte-tracking with
10471 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10474 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10476 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10477 (vect_get_mask_type_for_stmt): Likewise.
10478 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10480 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10481 to determine the statement's vector type and the vector type that
10482 should be used for calculating nunits. Deal with cases in which
10483 the type has to be deferred.
10484 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10485 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10486 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10487 (vect_determine_vf_for_stmt): New functions, split out from...
10488 (vect_determine_vectorization_factor): ...here.
10489 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10490 (vect_get_mask_type_for_stmt): New functions, split out from
10491 vect_determine_vectorization_factor.
10493 2018-05-16 Richard Biener <rguenther@suse.de>
10495 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10496 verify the [VEC_]COND_EXPR embedded comparison.
10498 2018-05-15 Martin Sebor <msebor@redhat.com>
10500 PR tree-optimization/85753
10501 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10502 RECORD_TYPE in addition to ARRAY_TYPE.
10504 2018-05-15 Martin Sebor <msebor@redhat.com>
10506 PR middle-end/85643
10507 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10509 2018-05-15 Richard Biener <rguenther@suse.de>
10511 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10512 add by_clobber_p one. Change algorithm to collect all defs
10513 representing uses we need to walk and try reducing them to
10514 a single one before failing.
10515 (dse_dom_walker::dse_optimize_stmt): Adjust.
10517 2018-05-13 Mark Wielaard <mark@klomp.org>
10519 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10520 (size_of_loc_descr): Likewise.
10521 (output_loc_operands): Likewise.
10522 (output_loc_operands_raw): Likewise.
10523 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10524 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10525 (hash_loc_operands): Likewise.
10526 (compare_loc_operands): Likewise.
10528 2018-05-14 Mark Wielaard <mark@klomp.org>
10530 * dwarf2out.c (count_index_addrs): New function.
10531 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10533 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10535 PR tree-optimization/83648
10536 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10537 return value as malloc candidate.
10539 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10542 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10543 param as true in call to suggest_attribute.
10545 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10547 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10548 -mreadonly-in-sdata.
10550 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10552 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10554 (aarch64_crypto_aesd_fused): Likewise.
10556 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10558 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10559 (movsi_aarch64): Likewise.
10560 (load_pairsi): Likewise.
10561 (load_pairdi): Likewise.
10562 (store_pairsi): Likewise.
10563 (store_pairdi): Likewise.
10564 (load_pairsf): Likewise.
10565 (load_pairdf): Likewise.
10566 (store_pairsf): Likewise.
10567 (store_pairdf): Likewise.
10568 (zero_extend): Likewise.
10569 (trunc): Swap alternatives.
10570 (fcvt_target): Add '?' to prefer w over r.
10572 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10575 * config/i386/i386.md: Disallow non-commutative arithmetics in
10576 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10577 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10578 in the peephole2 before it.
10580 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10582 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10583 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10584 (ix86_handle_option): Handle -mcldemote.
10585 * config.gcc: New header.
10586 * config/i386/cldemoteintrin.h: New file.
10587 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10588 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10590 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10591 OPTION_MASK_ISA_CLDEMOTE.
10592 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10593 (ix86_valid_target_attribute_inner_p): Ditto.
10594 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10595 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10596 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10597 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10598 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10600 * config/i386/i386.opt: Add -mcldemote.
10601 * config/i386/x86intrin.h: New header.
10602 * doc/invoke.texi: Add -mcldemote.
10604 2018-05-14 Richard Biener <rguenther@suse.de>
10606 * doc/match-and-simplify.texi: Adjust :s documentation.
10608 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10610 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10611 intended memcpy size.
10612 (REORDER_45): Likewise.
10614 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10616 * sort.cc: New file.
10617 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10618 * vec.c (qsort_chk): Use gcc_qsort.
10619 * Makefile.in (OBJS-libcommon): Add sort.o.
10620 (build/sort.o): New target. Use it...
10621 (BUILD_RTL): ... here, and...
10622 (build/gencfn-macros): ... here, and...
10623 (build/genmatch): ... here.
10625 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10626 Chung-Ju Wu <jasonwucj@gmail.com>
10628 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10629 * config/nds32/nds32-graywolf.md: New file.
10630 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10631 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10633 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10634 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10635 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10636 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10637 * config/nds32/pipelines.md: Include n15 settings.
10639 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10640 Chung-Ju Wu <jasonwucj@gmail.com>
10642 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10643 * config/nds32/nds32-n13.md: New file.
10644 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10645 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10647 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10648 * config/nds32/nds32.md (pipeline_model): Add n13.
10649 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10650 * config/nds32/pipelines.md: Include n13 settings.
10652 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10653 Chung-Ju Wu <jasonwucj@gmail.com>
10655 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10656 * config/nds32/nds32-n10.md: New file.
10657 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10658 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10660 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10661 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10662 * config/nds32/nds32.md (pipeline_model): Add n10.
10663 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10664 * config/nds32/pipelines.md: Include n10 settings.
10666 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10667 Kito Cheng <kito.cheng@gmail.com>
10668 Chung-Ju Wu <jasonwucj@gmail.com>
10670 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10671 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10672 Add enum values for DSP extension instructions.
10673 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10675 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10676 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10677 New code iterators.
10678 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10679 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10680 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10681 * config/nds32/nds32-intrinsic.md: Likewise.
10682 * config/nds32/nds32_intrinsic.h: Likewise.
10683 * config/nds32/nds32-md-auxiliary.c: Likewise.
10684 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10685 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10686 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10687 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10688 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10689 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10690 TYPE_DMAC in switch statement.
10691 * config/nds32/nds32.c: New checking and implementation for DSP
10692 extension instructions.
10693 * config/nds32/nds32.h: Likewise.
10694 * config/nds32/nds32.md: Likewise.
10695 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10696 * config/nds32/predicates.md: Implement new predicates for DSP
10699 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10701 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10702 Reformat alternatives and attributes so it is easier to identify
10703 which constraints/attributes go with which instruction.
10704 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10705 (mov<mode>_softfloat32, FMOVE64): Likewise.
10706 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10707 (mov<mode>_softfloat64, FMOVE64): Likewise.
10709 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10711 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10713 (Basic PowerPC Built-in Functions): The new name of the
10714 subsection previously known as "PowerPC Built-in Functions".
10715 (Basic PowerPC Built-in Functions Available on all Configurations):
10717 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10718 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10719 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10720 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10722 2018-05-11 Martin Jambor <mjambor@suse.cz>
10725 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10728 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10731 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10733 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10735 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10736 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10737 (ix86_handle_option): Handle -mwaitpkg.
10738 * config.gcc: New header.
10739 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10740 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10741 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10743 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10744 OPTION_MASK_ISA_WAITPKG.
10745 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10746 (ix86_option_override_internal): Add PTA_WAITPKG.
10747 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10748 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10749 IX86_BUILTIN_TPAUSE.
10750 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10751 __builtin_ia32_umwait and __builtin_ia32_tpause.
10752 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10753 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10754 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10755 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10756 UNSPECV_TPAUSE): New.
10757 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10758 * config/i386/i386.opt: Add -mwaitpkg.
10759 * config/i386/waitpkgintrin.h: New file.
10760 * config/i386/x86intrin.h: New header.
10761 * doc/invoke.texi: Add -mwaitpkg.
10763 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10766 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10768 (cortex-m0): Use armv6s-m isa.
10769 (cortex-m0plus): Likewise.
10770 (cortex-m1): Likewise.
10771 (cortex-m0.small-multiply): Likewise.
10772 (cortex-m0plus.small-multiply): Likewise.
10773 (cortex-m1.small-multiply): Likewise.
10775 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10776 Jakub Jelinek <jakub@redhat.com>
10778 PR tree-optimization/85692
10779 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10780 source permute as well.
10782 2018-05-11 Martin Liska <mliska@suse.cz>
10785 * doc/extend.texi: Document LLVM style format for no_sanitize
10788 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10790 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10791 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10792 (mode_supports_vsx_dform_quad): Likewise.
10793 (mode_supports_vmx_dform): Move these functions to be next to the
10794 other mode_supports functions.
10795 (mode_supports_dq_form): Likewise.
10796 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10797 mode_supports_dq_form.
10798 (reg_offset_addressing_ok_p): Likewise.
10799 (offsettable_ok_by_alignment): Likewise.
10800 (rs6000_legitimate_offset_address_p): Likewise.
10801 (legitimate_lo_sum_address_p): Likewise.
10802 (rs6000_legitimize_address): Likewise.
10803 (rs6000_legitimize_reload_address): Likewise.
10804 (rs6000_secondary_reload_inner): Likewise.
10805 (rs6000_preferred_reload_class): Likewise.
10806 (rs6000_output_move_128bit): Likewise.
10808 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10810 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10811 Generate SImode target register for null target.
10812 <case IX86_BUILTIN_XGETBV>: Ditto.
10813 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10814 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10816 2018-05-10 Carl Love <cel@us.ibm.com>
10818 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10819 dcbtt and dcbtstt if operands[2] is 0.
10821 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10824 * config/i386/sse.md (usadv64qi): New expander.
10826 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10828 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10829 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10830 -maltivec=be support.
10831 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10832 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10833 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10834 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10835 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10836 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10837 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10838 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10839 altivec_vsumsws): Adjust.
10840 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10841 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10842 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10844 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10845 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10846 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10847 (altivec_lve<VI_char>x): Delete expand.
10848 (*altivec_lve<VI_char>x_internal): Rename to...
10849 (altivec_lve<VI_char>x): ... this.
10850 (altivec_lvxl_<mode>): Delete expand.
10851 (*altivec_lvxl_<mode>_internal): Rename to ...
10852 (altivec_lvxl_<mode>): ... this.
10853 (altivec_stvxl_<mode>): Delete expand.
10854 (*altivec_stvxl_<mode>_internal): Rename to ...
10855 (altivec_stvxl_<mode>): ... this.
10856 (altivec_stve<VI_char>x): Delete expand.
10857 (*altivec_stve<VI_char>x_internal): Rename to ...
10858 (altivec_stve<VI_char>x): ... this.
10859 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10860 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10861 reduc_plus_scal_<mode>): Adjust.
10862 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10864 (rs6000_cpu_cpp_builtins): Adjust.
10865 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10866 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10867 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10868 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10869 -maltivec=be support.
10870 (rs6000_split_vec_extract_var): Adjust.
10871 (rs6000_split_v4si_init): Adjust.
10872 (swap_selector_for_mode): Delete.
10873 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10874 altivec_expand_stvex_be): Delete.
10875 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10876 -maltivec=be support.
10877 (rs6000_gimple_fold_builtin): Ditto.
10878 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10880 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10881 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10882 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10883 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10884 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10885 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10886 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10887 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10888 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10889 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10890 anonymous split): Adjust.
10891 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10892 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10894 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
10896 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10897 when --with-gxx-include-dir is also specified.
10898 * configure: Regenerate.
10900 2018-05-09 Jim Wilson <jimw@sifive.com>
10903 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10904 * config/riscv/t-withmultilib: New.
10905 * config/riscv/withmultilib.h: New.
10906 * doc/install.texi: Document RISC-V --with-multilib-list support.
10908 2018-05-09 Richard Biener <rguenther@suse.de>
10910 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10912 (vect_bb_vectorization_profitable_p): Adjust. Compute
10913 actual scalar cost using the cost vector and the add_stmt_cost
10916 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10918 PR rtl-optimization/85645
10919 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10920 in the REG_CFA_REGISTER note for LR, don't leave it empty.
10922 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10924 PR rtl-optimization/85645
10925 * shrink-wrap.c (spread_components): Return a boolean saying if
10926 anything was changed.
10927 (try_shrink_wrapping_separate): Iterate spread_components until
10928 nothing changes anymore.
10930 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10932 PR rtl-optimization/85645
10933 * regrename.c (build_def_use): Also kill the chains that include the
10934 destination of a REG_CFA_REGISTER note.
10936 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10938 PR rtl-optimization/85645
10939 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10940 insn that has a REG_CFA_REGISTER note.
10942 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
10944 * cfgexpand.c (expand_clobber): New function.
10945 (expand_gimple_stmt_1): Use it.
10946 * tree-vect-stmts.c (vect_clobber_variable): New function,
10948 (vectorizable_simd_clone_call): ...here.
10949 (vectorizable_store): Emit a clobber either side of an
10950 IFN_STORE_LANES sequence.
10951 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10953 2018-05-09 Tom de Vries <tom@codesourcery.com>
10956 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10957 (define_insn "trap_if_false"): Add exit after trap.
10959 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
10961 PR rtl-optimization/85638
10962 * bb-reorder.c: Include common/common-target.h.
10963 (create_forwarder_block): New function extracted from...
10964 (fix_up_crossing_landing_pad): ...here. Rename into...
10965 (dw2_fix_up_crossing_landing_pad): ...this.
10966 (sjlj_fix_up_crossing_landing_pad): New function.
10967 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10968 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10969 from both partitions and exit the loop after one iteration.
10971 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10974 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10976 (Basic PowerPC Built-in Functions): The new name of the
10977 subsection previously known as "PowerPC Built-in Functions".
10978 (Basic PowerPC Built-in Functions Available on all Configurations):
10980 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10982 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10984 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10986 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10989 2018-05-08 Jim Wilson <jimw@sifive.com>
10991 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10992 (LD_EMUL_SUFFIX): New.
10993 (LINK_SPEC): Use it.
10995 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10997 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10999 (Basic PowerPC Built-in Functions): The new name of the
11000 subsection previously known as "PowerPC Built-in Functions".
11001 (Basic PowerPC Built-in Functions Available on all Configurations):
11003 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11005 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11007 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11009 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11012 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11015 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11016 after cmpelim optimization.
11018 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11020 * config.gcc: Support "goldmont".
11021 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11022 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11023 PROCESSOR_GOLDMONT.
11024 * config/i386/i386.c (m_GOLDMONT): Define.
11025 (processor_target_table): Add "goldmont".
11026 (PTA_GOLDMONT): Define.
11027 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11028 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11029 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11030 (fold_builtin_cpu): Add "goldmont".
11031 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11032 (ix86_option_override_internal): Add "goldmont".
11033 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11034 (processor_type): Add PROCESSOR_GOLDMONT.
11035 * config/i386/i386.md: Add CPU "glm".
11036 * config/i386/glm.md: New file.
11037 * config/i386/x86-tune.def: Add m_GOLDMONT.
11038 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11040 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11043 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11045 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11046 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11047 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11050 * config/i386/i386.c (ix86_fold_builtin): Handle
11051 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11054 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11055 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11057 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11060 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11061 (check_arch): Likewise.
11062 (check_fpu): Return the result rather than printing it.
11063 (end arch): Fix operator precedence.
11064 (end cpu): Likewise.
11065 (END): Print the result from check_fpu.
11067 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11068 Alan Hayward <alan.hayward@arm.com>
11069 David Sherwood <david.sherwood@arm.com>
11071 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11072 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11073 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11074 (*fcmuo<mode>_and): New patterns.
11076 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11078 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11079 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11080 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11081 (cmp_op, sve_imm_con): New code attributes.
11082 (SVE_COND_INT_CMP, imm_con): Delete.
11083 (cmp_op): Remove above unspecs from int attribute.
11084 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11086 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11087 comparison-specific unspecs.
11088 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11089 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11090 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11091 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11092 (*vec_fcm<cmp_op><mode>): Rename to...
11093 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11094 (*vec_fcmuo<mode>): Rename to...
11095 (*fcmuo<mode>): ...this and adjust likewise.
11096 (*pred_fcm<cmp_op><mode>): New pattern.
11097 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11098 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11100 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11102 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11103 (aarch64_emit_sve_predicated_cond): New function.
11104 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11105 (aarch64_emit_unspec_cond_or): Replace with...
11106 (aarch64_emit_sve_or_conds): ...this new function. Use
11107 aarch64_emit_sve_ptrue_op for the individual comparisons and
11108 aarch64_emit_binop to OR them together.
11109 (aarch64_emit_inverted_unspec_cond): Replace with...
11110 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11111 aarch64_emit_sve_ptrue_op for the comparison and
11112 aarch64_emit_unop to invert the result.
11113 (aarch64_expand_sve_vec_cmp_float): Update after the above
11114 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11116 2018-05-07 Nathan Sidwell <nathan@acm.org>
11118 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11119 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11120 (Backwards Compatibility): Likewise.
11122 2018-05-07 Luis Machado <luis.machado@linaro.org>
11126 2018-05-07 Luis Machado <luis.machado@linaro.org>
11128 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11129 <prefetch_dynamic_strides>: New const bool field.
11130 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11131 prefetch_dynamic_strides.
11132 (exynosm1_prefetch_tune): Likewise.
11133 (thunderxt88_prefetch_tune): Likewise.
11134 (thunderx_prefetch_tune): Likewise.
11135 (thunderx2t99_prefetch_tune): Likewise.
11136 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11138 (aarch64_override_options_internal): Update to set
11139 PARAM_PREFETCH_DYNAMIC_STRIDES.
11140 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11141 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11142 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11143 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11144 prefetch-dynamic-strides setting.
11146 2018-05-07 Luis Machado <luis.machado@linaro.org>
11148 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11149 <minimum_stride>: New const int field.
11150 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11151 minimum_stride field.
11152 (exynosm1_prefetch_tune): Likewise.
11153 (thunderxt88_prefetch_tune): Likewise.
11154 (thunderx_prefetch_tune): Likewise.
11155 (thunderx2t99_prefetch_tune): Likewise.
11156 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11157 (aarch64_override_options_internal): Update to set
11158 PARAM_PREFETCH_MINIMUM_STRIDE.
11159 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11160 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11161 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11162 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11163 stride is constant and is below the minimum stride threshold.
11165 2018-05-07 Luis Machado <luis.machado@linaro.org>
11167 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11170 2018-05-07 Luis Machado <luis.machado@linaro.org>
11172 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11173 <prefetch_dynamic_strides>: New const bool field.
11174 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11175 prefetch_dynamic_strides.
11176 (exynosm1_prefetch_tune): Likewise.
11177 (thunderxt88_prefetch_tune): Likewise.
11178 (thunderx_prefetch_tune): Likewise.
11179 (thunderx2t99_prefetch_tune): Likewise.
11180 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11182 (aarch64_override_options_internal): Update to set
11183 PARAM_PREFETCH_DYNAMIC_STRIDES.
11184 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11185 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11186 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11187 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11188 prefetch-dynamic-strides setting.
11190 2018-05-07 Luis Machado <luis.machado@linaro.org>
11192 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11193 <minimum_stride>: New const int field.
11194 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11195 minimum_stride field.
11196 (exynosm1_prefetch_tune): Likewise.
11197 (thunderxt88_prefetch_tune): Likewise.
11198 (thunderx_prefetch_tune): Likewise.
11199 (thunderx2t99_prefetch_tune): Likewise.
11200 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11201 (aarch64_override_options_internal): Update to set
11202 PARAM_PREFETCH_MINIMUM_STRIDE.
11203 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11204 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11205 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11206 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11207 stride is constant and is below the minimum stride threshold.
11209 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11212 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11213 the type is addressable. Don't force op into register if it has
11216 2018-05-05 Roland McGrath <mcgrathr@google.com>
11219 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11220 any section for which we don't know a specific type it should have,
11221 regardless of name. Previously this was done only for the exact
11222 names ".init_array", ".fini_array", and ".preinit_array".
11223 (default_elf_asm_named_section): Add comment about
11224 relationship with default_section_type_flags and SECTION_NOTYPE.
11225 (get_section): Don't consider it a type conflict if one side has
11226 SECTION_NOTYPE and the other doesn't, as long as neither has the
11227 SECTION_BSS et al used in the default_section_type_flags logic.
11229 2018-05-05 Tom de Vries <tom@codesourcery.com>
11232 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11233 (workaround_barsyncs): New function.
11234 (nvptx_reorg): Use workaround_barsyncs.
11235 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11236 (define_expand "nvptx_membar_cta"): New define_expand.
11237 (define_insn "*nvptx_membar_cta"): New insn.
11239 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11241 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11242 To improve optimization opportunities.
11243 * builtin-types.def: The new needed builtin types for the above.
11245 2018-05-04 Richard Biener <rguenther@suse.de>
11247 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11248 * gimple-ssa-store-merging.c
11249 (imm_store_chain_info::output_merged_store): Remove redundant create,
11250 release split_store vector contents on failure.
11251 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11252 scalar stmt vector on cache hit.
11254 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11256 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11258 * config.gcc (powerpc-xilinx-eabi*): Remove.
11259 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11261 (fusion_addis_mem_combo_load): Ditto.
11262 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11264 (rs6000_cpu_cpp_builtins): Ditto.
11265 * config/rs6000/rs6000-linux.c
11266 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11267 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11268 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11270 (rs6000_setup_reg_addr_masks): Ditto.
11271 (rs6000_init_hard_regno_mode_ok): Ditto.
11272 (rs6000_option_override_internal): Ditto.
11273 (legitimate_lo_sum_address_p): Ditto.
11274 (rs6000_legitimize_address): Ditto.
11275 (rs6000_legitimize_reload_address): Ditto.
11276 (rs6000_legitimate_address_p): Ditto.
11277 (abi_v4_pass_in_fpr): Ditto.
11278 (setup_incoming_varargs): Ditto.
11279 (rs6000_gimplify_va_arg): Ditto.
11280 (rs6000_split_multireg_move): Ditto.
11281 (rs6000_savres_strategy): Ditto.
11282 (rs6000_emit_prologue_components): Ditto.
11283 (rs6000_emit_epilogue_components): Ditto.
11284 (rs6000_emit_prologue): Ditto.
11285 (rs6000_emit_epilogue): Ditto.
11286 (rs6000_elf_file_end): Ditto.
11287 (rs6000_function_value): Ditto.
11288 (rs6000_libcall_value): Ditto.
11289 * config/rs6000/rs6000.h: Ditto.
11290 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11291 (TARGET_MINMAX): ... this. New.
11292 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11293 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11294 (*movsi_internal1_single): Delete.
11295 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11296 mfpu=, mxilinx-fpu): Delete.
11297 * config/rs6000/singlefp.h: Delete.
11298 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11299 * config/rs6000/t-rs6000: Ditto.
11300 * config/rs6000/t-xilinx: Delete.
11301 * config/rs6000/titan.md: Adjust for fp_type removal.
11302 * config/rs6000/vsx.md: Remove Xilinx FP support.
11303 (VStype_simple): Delete.
11304 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11305 * config/rs6000/xfpu.h: Delete.
11306 * config/rs6000/xfpu.md: Delete.
11307 * config/rs6000/xilinx.h: Delete.
11308 * config/rs6000/xilinx.opt: Delete.
11309 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11310 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11312 2018-05-04 Tom de Vries <tom@codesourcery.com>
11315 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11318 2018-05-04 Richard Biener <rguenther@suse.de>
11320 PR middle-end/85627
11321 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11322 (expand_complex_div_wide): Likewise.
11323 (expand_complex_operations_1): Likewise.
11324 (expand_complex_libcall): Preserve EH info of the original stmt.
11325 (tree_lower_complex): Handle removed blocks.
11326 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11327 on complex multiplication and division libcall builtins.
11329 2018-05-04 Richard Biener <rguenther@suse.de>
11331 PR middle-end/85574
11332 * fold-const.c (negate_expr_p): Restrict negation of operand
11333 zero of a division to when we know that can happen without
11335 (fold_negate_expr_1): Likewise.
11337 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11340 * real.h (real_nextafter): Declare.
11341 * real.c (real_nextafter): New function.
11342 * fold-const-call.c (fold_const_nextafter): New function.
11343 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11344 CASE_CFN_NEXTTOWARD.
11345 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11346 even when arg1_mode is different from arg0_mode.
11348 2018-05-03 Nathan Sidwell <nathan@acm.org>
11350 * doc/extend.texi (Deprecated Features): Remove
11351 -ffriend-injection.
11352 (Backwards Compatibility): Likewise.
11353 * doc/invoke.texi (C++ Language Options): Likewise.
11354 (C++ Dialect Options): Likewise.
11356 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11359 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11360 _mm512_mask_mullox_epi64): New intrinsics.
11362 2018-05-03 Tom de Vries <tom@codesourcery.com>
11365 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11366 dump files): Add offload-tree.
11368 2018-05-03 Richard Biener <rguenther@suse.de>
11370 PR tree-optimization/85615
11371 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11372 to loops not nested in BBs loop father to avoid creating multi-entry
11375 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11377 PR tree-optimization/70291
11378 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11379 arguments. Change return type to tree. Emit libcall as a new
11380 statement rather than replacing existing one when inplace_p is true.
11381 (expand_complex_multiplication_components): New function.
11382 (expand_complex_multiplication): Expand floating-point complex
11383 multiplication using the above.
11384 (expand_complex_division): Rename inner_type parameter to type.
11385 Update expand_complex_libcall call-site.
11386 (expand_complex_operations_1): Update expand_complex_multiplication
11387 and expand_complex_division call-sites.
11389 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11392 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11393 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11394 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11395 the highest significant bit of the shift count mask is clear. In
11396 check whether and[sq]i3 is needed verify that all significant bits
11397 of the shift count other than the highest are set.
11399 2018-05-02 Tom de Vries <tom@codesourcery.com>
11402 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11403 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11404 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11405 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11406 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11407 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11408 __builtin_goacc_parlevel_size.
11410 2018-05-02 Richard Biener <rguenther@suse.de>
11412 PR tree-optimization/85597
11413 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11414 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11417 2018-05-02 Tom de Vries <tom@codesourcery.com>
11420 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11421 dump files): Add ltrans-tree.
11423 2018-05-02 Tom de Vries <tom@codesourcery.com>
11426 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11427 dump files): Add wpa-ipa.
11429 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11431 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11432 powerpc*-*-linux*paired* target.
11433 * config/rs6000/750cl.h: Delete.
11434 * config/rs6000/paired.h: Delete.
11435 * config/rs6000/paired.md: Delete.
11436 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11438 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11439 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11440 comment. Remove paired float support.
11441 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11442 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11444 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11445 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11447 * config/rs6000/rs6000.c: Remove paired float support.
11448 (paired_expand_vector_init, paired_expand_vector_move,
11449 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11450 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11451 paired_expand_builtin, paired_expand_predicate_builtin,
11452 paired_init_builtins): Delete.
11453 * config/rs6000/rs6000.h: Remove paired float support.
11454 * config/rs6000/rs6000.md: Remove paired float support.
11455 (move_from_CR_ov_bit): Delete.
11456 * config/rs6000/rs6000.opt (mpaired): Delete.
11457 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11458 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11460 2018-05-02 Richard Biener <rguenther@suse.de>
11462 PR middle-end/85567
11463 * gimplify.c (gimplify_save_expr): When in SSA form allow
11464 SAVE_EXPRs to compute to SSA vars.
11466 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11469 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11470 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11471 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11472 clobber operands[2], instead use a new pseudo. Formatting fixes.
11474 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11476 PR tree-optimization/85586
11477 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11478 exit early for statements in the same group if the accesses are
11481 2018-05-02 Tom de Vries <tom@codesourcery.com>
11484 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11487 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11489 PR tree-optimization/85143
11490 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11492 2018-05-01 Tom de Vries <tom@codesourcery.com>
11495 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11496 not found" error message.
11498 2018-05-01 Tom de Vries <tom@codesourcery.com>
11501 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11502 * vec.c (test_ordered_remove_if): New function.
11503 (vec_c_tests): Call test_ordered_remove_if.
11504 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11505 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11506 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11507 VEC_ORDERED_REMOVE_IF.
11509 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11511 PR tree-optimization/82665
11512 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11513 pointer subtraction where arguments come from a memchr call.
11515 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11517 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11518 --push-state --as-needed and --pop-state instead of --as-needed and
11519 --no-as-needed if ld supports it.
11520 * configure: Regenerated.
11523 * doc/install.texi2html: Replace _002d with - and _002a with * in
11524 generated html files using sed.
11526 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11529 * gcc-rich-location.c (blank_line_before_p): New function.
11530 (use_new_line): New function.
11531 (gcc_rich_location::add_fixit_insert_formatted): New function.
11532 * gcc-rich-location.h
11533 (gcc_rich_location::add_fixit_insert_formatted): New function.
11535 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11537 * selftest.c (assert_streq): Rename "expected" and "actual" to
11538 "val1" and "val2". Extend NULL-handling to cover both inputs
11539 symmetrically, while still requiring both to be non-NULL for a pass.
11540 * selftest.h (assert_streq): Rename "expected" and "actual" to
11542 (ASSERT_EQ): Likewise.
11543 (ASSERT_EQ_AT): Likewise.
11544 (ASSERT_KNOWN_EQ): Likewise.
11545 (ASSERT_KNOWN_EQ_AT): Likewise.
11546 (ASSERT_NE): Likewise.
11547 (ASSERT_MAYBE_NE): Likewise.
11548 (ASSERT_MAYBE_NE_AT): Likewise.
11549 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11550 the assertion to pass.
11551 (ASSERT_STREQ_AT): Likewise.
11553 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11555 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11556 interaction with -pie.
11558 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11560 * selftest.h: Fix alphabetization of per-source-file selftest
11563 2018-04-30 Jason Merrill <jason@redhat.com>
11565 PR c++/61982 - dead stores to destroyed objects.
11566 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11569 2018-04-30 Jason Merrill <jason@redhat.com>
11571 * tree.c (build_clobber): New.
11572 * tree.h: Declare it.
11573 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11575 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11577 * diagnostic-show-locus.c (layout::layout): Update for
11578 location_get_source_line returning a char_span.
11579 (struct char_span): Move to input.h.
11580 (struct correction): Update for fields in char_span becoming
11582 (struct source_line): Update for location_get_source_line
11583 returning a char_span.
11584 (layout::print_line): Likewise.
11585 * edit-context.c (edited_file::print_content): Likewise.
11586 (edited_file::print_diff_hunk): Likewise.
11587 (edited_file::print_run_of_changed_lines): Likewise.
11588 (edited_file::get_num_lines): Likewise.
11589 (edited_line::edited_line): Likewise.
11590 * final.c (asm_show_source): Likewise.
11591 * input.c (location_get_source_line): Convert return type
11592 from const char * to char_span, losing the final "line_len"
11594 (dump_location_info): Update for the above.
11595 (get_substring_ranges_for_loc): Likewise. Use a char_span
11596 when handling the literal within the line.
11597 (test_reading_source_line): Update for location_get_source_line
11598 returning a char_span.
11599 * input.h (class char_span): Move here from
11600 diagnostic-show-locus.c, converting from a struct to a class.
11601 Make data members private.
11602 (char_span::operator bool): New.
11603 (char_span::length): New.
11604 (char_span::get_buffer): New.
11605 (char_span::operator[]): New.
11606 (char_span::subspan): Make const.
11607 (char_span::xstrdup): New.
11608 (location_get_source_line): Convert return type from const char *
11609 to char_span, losing the final "line_size" param.
11611 2018-04-30 Jan Hubicka <jh@suse.cz>
11613 * lto-wrapper.c (ltrans_priorities): New static var.
11614 (cmp_priority): New.
11615 (run_gcc): Read priorities and if doing parallel build order
11616 the Makefile by them.
11618 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11620 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11622 2018-04-30 Richard Biener <rguenther@suse.de>
11624 * tree-cfg.c (verify_address): Remove base argument, add
11625 flag whether to check TREE_ADDRESSABLE and do that.
11626 (verify_expr): Remove.
11627 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11628 (verify_gimple_assign_single): Likewise.
11629 (verify_gimple_switch): Likewise.
11630 (verify_expr_location_1): Dereference tp once. Add (disabled)
11631 piece from verify_expr.
11632 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11634 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11636 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11638 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11640 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11641 (small_data_pattern): Likewise.
11642 (arc_rewrite_small_data): Likewise.
11643 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11644 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11645 (get_symbol_alignment): New function.
11646 (legitimate_small_data_address_p): Likewise.
11647 (legitimate_scaled_address): Update, call
11648 legitimate_small_data_address_p.
11649 (output_sdata): New static variable.
11650 (arc_print_operand): Update how we handle small data operands.
11651 (arc_print_operand_address): Likewise.
11652 (arc_legitimate_address_p): Update, use
11653 legitimate_small_data_address_p.
11654 (arc_rewrite_small_data_p): Remove.
11655 (arc_rewrite_small_data_1): Likewise.
11656 (arc_rewrite_small_data): Likewise.
11657 (small_data_pattern): Likewise.
11658 (compact_sda_memory_operand): Update to use
11659 legitimate_small_data_address_p and get_symbol_alignment.
11660 (prepare_move_operands): Don't rewite sdata pattern.
11661 (prepare_extend_operands): Remove.
11662 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11664 (zero_extendqisi2): Likewise.
11665 (zero_extendhisi2): Likewise.
11666 (extendqihi2): Likewise.
11667 (extendqisi2): Likewise.
11668 (extendhisi2): Likewise.
11669 (addsi3): Likewise.
11670 (subsi3): Likewise.
11671 (andsi3): Likewise.
11672 * config/arc/constraints.md (Usd): Change it to memory constraint.
11674 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11676 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11677 as source of std instructions.
11678 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11679 allow 6-bit constants as source for store instructions.
11680 (movdi_insn): Update instruction pattern to allow 6-bit constants
11681 as source for store instructions.
11683 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11685 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11687 2018-04-30 Nathan Sidwell <nathan@acm.org>
11688 Sandra Loosemore <sandra@codesourcery.com>
11690 * dumpfile.c (dump_open): Allow '-' for stdout.
11691 * doc/invoke.texi (Developer Options): Document dump filename
11692 determination early. Document stdin/stdout selection.
11694 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11696 Microblaze Target: PIC data text relative
11698 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11699 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11701 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11702 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11703 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11704 New addressing mode for data-text relative position indepenedent code.
11705 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11706 'ADDRESS_SYMBOLIC_TXT_REL'.
11707 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11708 (microblaze_legitimate_pic_operand): Exclude function calls from
11709 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11710 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11712 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11713 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11714 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11715 for 'address + offset'.
11716 (microblaze_expand_prologue): Add new function prologue call for
11718 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11719 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11720 table in case of TARGET_PIC_DATA_TEXT_REL.
11721 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11722 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11723 Add new macros 'UNSPEC_TEXT',
11724 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11725 + exclude function calls from 'UNSPEC_PLT' in case of data text
11727 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11728 new target hook for generating address diff vector tables in case of
11730 * doc/tm.texi : Regenerate.
11731 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11732 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11733 of addr diff vector generation.
11734 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11735 target hook definition.
11736 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11737 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11738 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11739 option for data text relative.
11741 2018-04-30 Richard Biener <rguenther@suse.de>
11743 * tree-chrec.h (evolution_function_is_constant_p): Remove
11745 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11747 2018-04-30 Richard Biener <rguenther@suse.de>
11750 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11752 2018-04-30 Richard Biener <rguenther@suse.de>
11754 PR tree-optimization/28364
11755 PR tree-optimization/85275
11756 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11757 copying first exit test.
11759 2018-04-28 Mark Wielaard <mark@klomp.org>
11761 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11762 dwarf_version >= 5.
11763 (dwarf_AT): Handle DW_AT_addr_base.
11764 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11766 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11769 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11770 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11771 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11772 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11774 2018-04-28 Richard Biener <rguenther@suse.de>
11776 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11777 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11778 to reflect use. Only add interesting stmts.
11780 2018-04-27 Martin Jambor <mjambor@suse.cz>
11783 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11784 the jump function allows for passing through aggregate values.
11786 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11788 * input.h (in_system_header_at): Convert from macro to inline
11790 (from_macro_expansion_at): Likewise.
11791 (from_macro_definition_at): Likewise.
11793 2018-04-27 Jeff Law <law@redhat.com>
11795 * config.gcc: Mark tile* targets as deprecated/obsolete.
11797 2018-04-27 Richard Biener <rguenther@suse.de>
11799 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11802 2018-04-27 Richard Biener <rguenther@suse.de>
11804 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11806 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11808 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11809 with Yd constraint. Set "preferred_for_speed" attribute from
11810 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11811 with Yd constraint.
11812 (*movdi_internal): Ditto.
11813 (movti_interunit splitters): Remove
11814 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11815 (movdi_interunit splitters): Ditto.
11816 * config/i386/constraints.md (Ye): Remove.
11817 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11819 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11822 * config/aarch64/constraints.md (Usg): Limit to 31.
11823 (Usj): Limit to 63.
11825 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11827 PR tree-optimization/85529
11828 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11829 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11830 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11831 zero extension or masking of the MSB bit.
11832 (optimize_range_tests): Add FIRST_BB argument, pass it through
11833 to optimize_range_tests_var_bound.
11834 (maybe_optimize_range_tests, reassociate_bb): Adjust
11835 optimize_range_tests callers.
11837 2018-04-26 Richard Biener <rguenther@suse.de>
11838 Jakub Jelinek <jakub@redhat.com>
11840 * cgraph.h (symbol_table): Just declare debug method here.
11841 * symtab.c (symbol_table::debug): Define.
11843 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11845 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11847 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11849 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11850 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11851 (*movdi_internal): Substitute Yi and Yj constraint with x
11852 and Ym and Yn constraint with y constraint. Update "isa"
11853 attribute and set "preferred_for_speed" attribute from
11854 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11855 (*movsi_internal): Ditto.
11856 (*movdf_internal): Ditto.
11857 (*movsf_internal): Ditto.
11858 (*zero_extendsidi2): Ditto.
11859 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11860 (sse2_loadld): Ditto.
11861 (*vec_extract<ssevecmodelower>_0): Ditto.
11862 (*vec_extractv4si_0_zext_sse4): Ditto.
11863 (vec_concatv2di): Ditto.
11864 (*vec_dup<mode>): Ditto.
11865 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11866 * config/i386/constraints.md (Yi): Remove.
11871 2018-04-26 Nathan Sidwell <nathan@acm.org>
11873 * dumpfile.c (dump_open): New.
11874 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11875 (dump_finish): Detect stdio/stderr by value not name.
11877 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11879 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11881 2018-04-26 Tom de Vries <tom@codesourcery.com>
11884 * config/nvptx/nvptx.c (verify_neutering_jumps)
11885 (verify_neutering_labels): New function
11886 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11888 2018-04-26 Tom de Vries <tom@codesourcery.com>
11891 * config/nvptx/nvptx.c (needs_neutering_p): New function.
11892 (nvptx_single): Use needs_neutering_p to skip over insns that do not
11895 2018-04-26 Richard Biener <rguenther@suse.de>
11896 Tom de Vries <tom@codesourcery.com>
11899 * lto-streamer-out.c (output_function): Fixup loops if required to match
11900 discovery done in the reader.
11902 2018-04-26 Richard Biener <rguenther@suse.de>
11904 PR tree-optimization/85116
11905 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11906 have a loop exit from the single latch predecessor. Remove
11907 case of header with just condition.
11908 (ch_base::copy_headers): Exclude infinite loops from any
11910 (pass_ch::execute): Record exits.
11912 2018-04-26 Richard Biener <rguenther@suse.de>
11914 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11915 prologue cost vector and pass it to vect_get_load_cost.
11916 (vect_get_peeling_costs_all_drs): Likewise.
11917 (vect_peeling_hash_get_lowest_cost): Likewise.
11918 (vect_enhance_data_refs_alignment): Likewise.
11920 2018-04-26 Richard Biener <rguenther@suse.de>
11922 PR middle-end/85450
11923 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11924 checking of integer<->pointer conversions.
11925 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11926 sign-/zero-extending pointer types.
11927 (expand_omp_for_static_chunk): Likewise.
11929 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
11930 Jean Lee <xiaoyur347@gmail.com>
11932 * config/mips/mips.c (mips_asan_shadow_offset): New function.
11933 (TARGET_ASAN_SHADOW_OFFSET): Define.
11934 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11935 true for -fsanitize=address.
11937 2018-04-25 Mark Wielaard <mark@klomp.org>
11939 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11942 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11944 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11945 than "alu", remove explicit "memory" and "imm_disp" attributes.
11946 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11948 PR middle-end/85414
11949 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11950 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11951 gen_lowpart_no_emit.
11953 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
11956 * config/i386/i386.c (ix86_expand_builtin): Change memory
11957 operand to XI, extend p0 to Pmode.
11958 * config/i386/i386.md: Change unspec volatile and operand
11959 1 mode to XI, change operand 0 mode to P.
11961 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11963 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11964 GET_MODE_MASK before any checking.
11965 (nds32_can_use_bset_p): Likewise.
11966 (nds32_can_use_btgl_p): Likewise.
11968 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11970 * config/nds32/nds32-doubleword.md: New define_split pattern for
11971 illegal register number.
11973 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11975 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11977 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11979 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11981 2018-04-25 Richard Biener <rguenther@suse.de>
11983 * lto-streamer.h (LTO_major_version): Bump to 8.
11985 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11987 * BASE-VER: Set to 9.0.0.
11989 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
11991 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11992 in __abskf2 and __powikf2.
11994 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11997 * config/aarch64/constraints.md (Usg, Usj): New constraints.
11998 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11999 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12000 Use the above on operand 2. Reindent.
12001 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12003 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12006 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12008 * config/i386/i386.opt (mcet): Removed.
12009 * doc/install.texi: Remove -mcet documentation.
12010 * doc/invoke.texi: Likewise.
12012 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12015 * doc/install.texi: Remove -mcet from bootstrap-cet.
12017 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12020 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12021 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12025 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12026 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12027 containing a CONST_VECTOR.
12029 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12031 * doc/install.texi: Update newlib dependency for nvptx.
12033 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12036 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12037 instead of INTVAL when shifting x left.
12039 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12041 PR tree-optimization/85478
12042 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12043 vect_grouped_store_supported for single element vectors.
12045 2018-04-24 Richard Biener <rguenther@suse.de>
12048 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12049 load cost increase to the case of non-constant step.
12051 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12054 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12055 destination if any_malformed_asm.
12057 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12059 PR middle-end/85496
12060 * expr.c (store_field): In the bitfield case, if the value comes from
12061 a function call and is returned in registers by means of a PARALLEL,
12062 do not change the mode of the temporary unless BLKmode and VOIDmode.
12064 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12066 PR rtl-optimization/85423
12067 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12068 dependencies to debug insns when the previous insn is non-debug.
12070 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12072 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12073 enums into a single definition.
12074 (fls): Fix predicates and printing.
12077 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12079 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12080 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12081 and short u6 immediate.
12082 (check_if_valid_sleep_operand): Remove.
12083 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12085 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12087 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12088 flag_always_save_lp condition.
12089 * config/nds32/nds32.opt (malways-save-lp): New option.
12091 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12093 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12094 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12095 * config/nds32/nds32.h
12096 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12097 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12099 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12101 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12102 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12104 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12105 Chung-Ju Wu <jasonwucj@gmail.com>
12107 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12109 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12110 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12112 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12114 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12116 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12118 * config/nds32/nds32-protos.h (nds32_data_alignment,
12119 nds32_local_alignment): Declare.
12120 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12121 nds32_local_alignment): New functions.
12122 (TARGET_CONSTANT_ALIGNMENT): Define.
12123 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12125 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12127 * config/nds32/nds32.c
12128 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12129 (TARGET_MODES_TIEABLE_P): Likewise.
12131 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12133 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12134 level Ofast and Og.
12136 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12137 Chung-Ju Wu <jasonwucj@gmail.com>
12139 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12140 for unaligned access.
12141 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12143 * config/nds32/nds32-intrinsic.md: Likewise.
12144 * config/nds32/nds32_intrinsic.h: Likewise.
12145 * config/nds32/nds32.h (nds32_builtins): Likewise.
12146 * config/nds32/nds32.opt (munaligned-access): New option.
12147 * config/nds32/nds32.c (nds32_asm_file_start): Display
12148 flag_unaligned_access status.
12150 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12152 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12153 -mno-relax is present.
12154 * config/riscv/linux.h (LINK_SPEC): Ditto.
12156 2018-04-20 Martin Sebor <msebor@redhat.com>
12159 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12161 (gimple_fold_builtin_stxcpy_chk): Same.
12162 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12164 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12167 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12168 __powikf2 when long double is IEEE 128-bit.
12170 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12172 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12173 step to make sure stack always aligned.
12175 2018-04-20 Carl Love <cel@us.ibm.com>
12178 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12179 size check for arg0.
12181 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12182 Tom de Vries <tom@codesourcery.com>
12185 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12186 Emit insns for calls too.
12187 (nvptx_find_par): Always look for worker-level predecessor insn.
12188 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12190 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12191 (nvptx_process_pars): Propagate frames for calls.
12193 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12196 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12198 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12199 (ix86_handle_option): Don't handle OPT_mibt.
12200 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12202 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12204 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12205 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12206 (ix86_target_macros): Define __CET__ with flag_cf_protection
12207 for -fcf-protection.
12208 * config/i386/i386.c (isa2_opts): Remove -mibt.
12209 * config/i386/i386.h (TARGET_IBT): Removed.
12210 (TARGET_IBT_P): Likewise.
12211 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12212 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12213 * config/i386/i386.opt (mcet): Update help message.
12214 (mshstk): Likewise.
12216 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12217 -mcet as an alias for -mshstk.
12219 2018-04-20 Richard Biener <rguenther@suse.de>
12221 PR middle-end/85475
12222 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12223 complexity by forcing a single use of the multiply operand.
12225 2018-04-20 Martin Jambor <mjambor@suse.cz>
12228 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12229 recursion dependency to only apply to non-clones.
12231 2018-04-20 Martin Jambor <mjambor@suse.cz>
12234 * ipa-cp.c (create_specialized_node): Check that clones of
12235 self-recursive edges exist during IPA-CP.
12237 2018-04-19 Toon Moene <toon@moene.org>
12239 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12242 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12244 PR tree-optimization/85467
12245 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12246 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12247 VECTOR_CST element to type.
12249 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12252 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12253 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12254 (builtin_longjmp): Likewise.
12255 (save_stack_nonlocal): New pattern.
12256 (restore_stack_nonlocal): Likewise.
12258 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12261 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12262 Replace ASM_OUTPUT_LABEL with fprintf.
12264 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12267 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12268 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12269 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12270 define __IBT__ and __SHSTK__ for -fcf-protection.
12271 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12273 (ix86_trampoline_init): Likewise.
12274 (x86_output_mi_thunk): Likewise.
12275 (ix86_notrack_prefixed_insn_p): Likewise.
12276 (ix86_option_override_internal): Don't disallow -fcf-protection.
12277 * config/i386/i386.md (rdssp<mode>): Also enable for
12279 (incssp<mode>): Likewise.
12280 (nop_endbr): Likewise.
12281 * config/i386/i386.opt (mcet): Change help message to built-in
12284 (mshstk): Likewise.
12285 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12286 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12287 enable CET built-in functions.
12289 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12291 * common/config/i386/i386-common.c
12292 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12293 OPTION_MASK_ISA_MOVDIRI_UNSET,
12294 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12295 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12296 * config.gcc (movdirintrin.h): New header.
12297 * config/i386/cpuid.h (bit_MOVDIRI,
12298 bit_MOVDIR64B): New bits.
12299 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12301 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12302 (VOID, PVOID, PCVOID)): New function types.
12303 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12304 __builtin_ia32_directstoreu_u64,
12305 __builtin_ia32_movdir64b): New builtins.
12306 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12307 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12309 (ix86_valid_target_attribute_inner_p): Ditto.
12310 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12311 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12312 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12313 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12314 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12315 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12316 (movdiri<mode>, movdir64b_<mode>): New.
12317 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12318 * config/i386/immintrin.h: Include movdirintrin.h.
12319 * config/i386/movdirintrin.h: New file.
12320 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12322 2018-04-19 Richard Biener <rguenther@suse.de>
12324 PR middle-end/85455
12325 * cfg.c (clear_bb_flags): When loop state says we have
12326 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12328 2018-04-19 Richard Biener <rguenther@suse.de>
12330 PR tree-optimization/84737
12331 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12332 copying restrict info.
12333 (vect_setup_realignment): Use it.
12334 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12335 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12336 the first DR to all generated stores.
12337 (vectorizable_load): Likewise for loads.
12339 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12341 PR tree-optimization/85446
12342 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12343 the integral and pointer types to have the same precision.
12345 * doc/install.texi: Document --disable-cet being the default and
12348 2018-04-18 Martin Liska <mliska@suse.cz>
12350 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12353 2018-04-18 Martin Liska <mliska@suse.cz>
12356 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12359 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12360 arguments if they are comparable.
12362 2018-04-18 Martin Liska <mliska@suse.cz>
12365 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12368 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12371 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12374 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12375 ENDBR after calling __morestack.
12377 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12380 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12381 by using gcc_base_ver to generate a gcc_driver_version, and use
12382 it when generating GCC_DRIVER_NAME.
12383 * configure: Regenerate.
12385 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12388 * config.gcc: Obsolete powerpc*-*-*spe*.
12390 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12393 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12394 (stabstr_D): Change type of unum from unsigned int to
12395 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12398 2018-04-17 Jim Wilson <jimw@sifive.com>
12401 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12402 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12403 Set arg_pointer_offset after using pretend_args_size.
12405 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12407 PR rtl-optimization/85431
12408 * dse.c (record_store): Ignore zero width stores.
12411 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12412 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12413 __builtin_stack_restore rather than after it.
12414 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12415 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12416 argument instead of virtual_dynamic_stack_rtx.
12418 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12420 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12422 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12423 Add note to error message to explain internal mapping of overloaded
12424 built-in function name to non-overloaded built-in function name.
12425 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12428 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12431 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12432 where the inputs overlap with the output.
12434 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12436 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12437 (=v, v) alternative and explicit "memory" attribute.
12438 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12439 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12441 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12442 "sselog1" type instead of "sselog".
12443 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12444 "sselog". Remove explicit "memory" attribute.
12445 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12446 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12448 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12449 "sselog1" type instead of "sselog". Remove explicit "memory"
12451 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12452 use "sselog1" type instead of "sselog". Remove explicit "memory"
12454 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12455 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12457 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12458 "sselog1" type instead of "sselog". Remove explicit "memory"
12460 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12461 use "sselog1" type instead of "sselog". Remove explicit "memory"
12465 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12467 PR middle-end/85414
12468 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12471 2018-04-17 Martin Jambor <mjambor@suse.cz>
12474 * ipa-cp.c (create_specialized_node): Call
12475 expand_all_artificial_thunks if necessary.
12477 2018-04-17 Martin Liska <mliska@suse.cz>
12480 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12481 in message, remote space in between '_G' and '('.
12483 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12486 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12487 avx512f_vmcmp<mode>3<round_saeonly_name>,
12488 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12489 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12490 avx512f_rndscale<mode><round_saeonly_name>,
12491 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12492 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12493 Use %<iptr>2 instead of %2 for -masm=intel.
12494 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12495 avx512f_vcvttss2usi<round_saeonly_name>,
12496 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12498 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12499 avx512f_vcvttsd2usi<round_saeonly_name>,
12500 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12501 Use %q1 instead of %1 for -masm=intel.
12502 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12503 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12504 of %3 for -masm=intel.
12505 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12507 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12509 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12511 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12512 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12514 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12516 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12517 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12518 %g1 and one with %0 and %1.
12519 (avx512er_vmrcp28<mode><round_saeonly_name>,
12520 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12521 %1 for -masm=intel.
12522 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12523 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12524 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12525 of %0 and %{%4%} for -masm=intel.
12526 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12527 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12528 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12529 order of %0 and %{%5%}%{z%} for -masm=intel.
12531 2018-04-17 Jan Hubicka <jh@suse.cz>
12534 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12536 2018-04-17 Martin Liska <mliska@suse.cz>
12539 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12540 for target_clone error message. Make default implementation
12541 clone to be a local declaration.
12542 (separate_attrs): Add new argument and check for an empty
12544 (expand_target_clones): Handle it.
12545 (ipa_target_clone): Make redirection just for target_clones
12548 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12549 Tom de Vries <tom@codesourcery.com>
12551 PR middle-end/84955
12552 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12553 tiled basic blocks without omp continue statements.
12555 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12558 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12559 vec_extract expression as having side effects to make sure it gets
12562 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12565 * config/i386/i386.c (get_builtin_code_for_version): Check
12568 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12571 * config.gcc: Support "skylake".
12572 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12574 * config/i386/i386.c (m_SKYLAKE): Define.
12575 (processor_target_table): Add "skylake".
12576 (ix86_option_override_internal): Add "skylake".
12577 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12578 PROCESSOR_CANNONLAKE.
12579 (get_builtin_code_for_version): Fix priority for
12580 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12581 PROCESSOR_SKYLAKE-AVX512.
12582 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12583 (processor_type): Add PROCESSOR_SKYLAKE.
12585 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12586 Jason Merrill <jason@redhat.com>
12589 * convert.c (convert_to_integer_1): Use direct recursion for
12590 enumeral types and types with a precision less than the number
12591 of bits in their mode.
12593 2018-04-16 Julia Koval <julia.koval@intel.com>
12596 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12597 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12599 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12602 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12603 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12604 and -mno-direct-move.
12606 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12609 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12610 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12611 Ensure negative shifts result in {0}.
12613 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12615 PR rtl-optimization/79916
12616 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12617 regs (if any) to define how to gnerate SD moves when LRA is in
12620 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12622 PR rtl-optimization/85393
12623 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12624 * except.c (expand_dw2_landing_pad_for_region): Make static.
12625 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12626 a label and unconditional jump to old_bb, rather than
12627 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12630 PR rtl-optimization/85376
12631 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12632 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12633 instead of a specific value.
12635 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12636 Bin Cheng <bin.cheng@arm.com>
12638 PR tree-optimization/82965
12639 PR tree-optimization/83991
12640 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12641 by_profile_only parameter.
12642 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12643 information if the loop was predicted to iterate too many times.
12644 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12646 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12649 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12652 2018-04-13 Martin Liska <mliska@suse.cz>
12653 Jakub Jelinek <jakub@redhat.com>
12655 PR middle-end/81657
12656 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12657 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12658 * builtins.c (expand_builtin_memory_copy_args): Use
12659 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12660 handle dest_addr == pc_rtx.
12662 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12665 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12666 asked to not generate direct moves.
12667 (fix_trunc<mode>si2_stfiwx): Similar.
12668 (fix_trunc<mode>si2_internal): Similar.
12670 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12673 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12674 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12675 lookup if dest in some wider mode is known to be const0_rtx and
12676 if so, record permanent equivalence for it to be ZERO_EXTEND of
12677 the narrower mode destination.
12679 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12681 * lto-streamer-out.c (output_function): Revert 259346.
12682 * omp-expand.c (expand_oacc_for): Likewise.
12684 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12686 PR rtl-optimization/85354
12687 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12688 * sel-sched.c (sel_global_init): ... here.
12690 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12693 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12694 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12695 mode for PE-COFF targets.
12696 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12697 (i386_pe_asm_lto_end): Likewise.
12698 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12699 (TARGET_ASM_LTO_END): Likewise.
12700 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12701 (i386_pe_asm_lto_start): New function.
12702 (i386_pe_asm_lto_end): Likewise.
12704 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12705 Richard Biener <rguenther@suse.de>
12707 PR middle-end/84955
12708 * lto-streamer-out.c (output_function): Fix CFG loop state before
12710 * omp-expand.c (expand_oacc_for): Handle calls to internal
12711 functions like regular functions.
12713 2018-04-12 Richard Biener <rguenther@suse.de>
12716 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12717 for the early LTO debug to properly generate references to it
12718 during DIE emission. Do not re-use that for the skeleton for
12720 (dwarf2out_early_finish): Likewise.
12722 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12725 * config/i386/sse.md
12726 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12727 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12728 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12729 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12730 and output is a reg, avoid creating invalid lowpart subreg, but
12731 instead split into a 512-bit move. Don't split if not AVX512VL,
12732 input is xmm16+ reg and output is a mem.
12733 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12734 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12735 xmm16+ reg and output is a mem.
12737 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12739 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12740 also for flag_dwarf2_cfi_asm.
12742 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12744 PR rtl-optimization/85342
12745 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12746 a bool scalar var inside of the loop instead. Don't try to update
12747 recog_data.operand after failed apply_change_group.
12749 2018-04-12 Tom de Vries <tom@codesourcery.com>
12752 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12753 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12754 array with flexible array member as array without given dimension.
12755 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12756 argument for undefined param to true.
12758 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12761 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12762 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12763 from PowerPC section.
12764 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12765 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12766 help text that is too long.
12767 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12768 help text that is too long.
12769 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12770 help text that is too long.
12772 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12774 * config/alpha/alpha.md (stack_probe_internal): Rename
12775 from "probe_stack". Update all callers.
12777 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12779 PR rtl-optimization/84566
12780 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12781 sched_macro_fuse_insns.
12783 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12786 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12787 (compute_block_dependences): ... from here.
12789 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12791 PR tree-optimization/85331
12792 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12793 from int to HOST_WIDE_INT.
12795 2018-04-11 Martin Jambor <mjambor@suse.cz>
12798 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12799 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12800 not the same as the source val.
12801 (cgraph_edge_brings_value_p): New parameter.
12802 (gather_edges_for_value): Pass destination value to
12803 cgraph_edge_brings_value_p.
12804 (perhaps_add_new_callers): Likewise.
12805 (get_info_about_necessary_edges): Likewise and exclude values brought
12806 only by self-recursive edges.
12807 (create_specialized_node): Redirect only clones of self-calling edges.
12808 (+self_recursive_pass_through_p): New function.
12809 (find_more_scalar_values_for_callers_subset): Use it.
12810 (find_aggregate_values_for_callers_subset): Likewise.
12811 (known_aggs_to_agg_replacement_list): Removed.
12812 (decide_whether_version_node): Re-calculate known constants for all
12813 remaining context clones.
12815 2018-04-11 Richard Biener <rguenther@suse.de>
12818 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12819 from early DWARF output.
12820 (dwarf2out_early_finish): Output line info unconditionally into
12821 early DWARF and add reference to it.
12823 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12826 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12827 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12828 other than V2DFmode using iptr mode attribute.
12829 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12831 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12833 PR rtl-optimization/84659
12834 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12836 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12839 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12841 (output_loc_list): Pass address of a dummy size variable even in the
12842 locview handling loop.
12843 (index_location_lists): Add comment on why skip_loc_list_entry can't
12846 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12849 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12852 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12855 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12856 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12857 and -mstring-compare-inline-limit.
12859 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12862 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12863 for stack clash protection in a register whenever we need it to be in
12866 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12868 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12869 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12871 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12874 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12876 (mlong-double-): Ditto.
12877 * config/rs6000/sysv4.opt (msdata=): Ditto.
12878 (mtls-size=): Ditto.
12880 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12882 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12883 erroneous entries for
12884 "vector int vec_ldl (int, long int *)", and
12885 "vector unsigned int vec_ldl (int, unsigned long int *)".
12886 Add comments and entries for
12887 "vector bool char vec_ldl (int, bool char *)",
12888 "vector bool short vec_ldl (int, bool short *)",
12889 "vector bool int vec_ldl (int, bool int *)",
12890 "vector bool long long vec_ldl (int, bool long long *)",
12891 "vector pixel vec_ldl (int, pixel *)",
12892 "vector long long vec_ldl (int, long long *)",
12893 "vector unsigned long long vec_ldl (int, unsigned long long *)".
12894 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12895 type tree bool_long_long_type_node and correct definition of
12896 bool_V2DI_type_node to make reference to this new type tree.
12897 (rs6000_mangle_type): Replace erroneous reference to
12898 bool_long_type_node with bool_long_long_type_node.
12899 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12900 comments to emphasize sign distinctions for char and int types and
12901 replace RS6000_BTI_bool_long constant with
12902 RS6000_BTI_bool_long_long constant. Also add comment to restrict
12903 use of RS6000_BTI_pixel.
12904 (bool_long_type_node): Remove this macro definition.
12905 (bool_long_long_type_node): New macro definition
12907 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12909 PR rtl-optimization/85300
12910 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12911 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12912 simplify_unary_operation fails.
12914 2018-04-10 Martin Liska <mliska@suse.cz>
12916 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12917 cgraph_edge and ipa_ref.
12919 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12923 * config/i386/sse.md
12924 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12925 computation of the VEC_MERGE selector from mask.
12926 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12927 Fix decoding of the VEC_MERGE selector into mask.
12929 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12931 PR tree-optimization/85286
12932 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12934 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12936 * final.c (final_1): Set insn_last_address as well as
12937 insn_current_address.
12939 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12942 * explow.c (emit_stack_probe): Call validize_mem on memory location
12943 before passing it to gen_probe_stack. Create address operand and
12944 legitimize it for the probe_stack_address case.
12946 2018-04-09 Jan Hubicka <jh@suse.cz>
12949 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12950 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12951 * tree.c (free_lang_data_in_type): Fix handling of binfos;
12953 (free_lang_data): Rebuild type inheritance graph.
12955 2018-04-09 Martin Sebor <msebor@redhat.com>
12957 * invoke.texi (-finline-small-functions): Mention other optimization
12959 (-findirect-inlining, -fpartial-inlining): Same.
12960 (-finline-functions-called-once): Same.
12961 (-freorder-blocks-and-partition): Same.
12963 2018-04-09 Jan Hubicka <jh@suse.cz>
12966 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12967 jumps; choose last target that matches the criteria (i.e.
12968 no partition changes for non-crossing jumps).
12969 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12970 support for redirecting crossing jumps to non-crossing.
12972 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
12974 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12975 also for naked functions.
12977 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
12979 * config/arc/arc.md (add_shift): New pattern.
12980 (add_shift2): Likewise.
12981 (sub_shift): Likewise.
12982 (sub_shift_cmp0_noout): Likewise.
12983 (compare_si_ashiftsi): Likewise.
12984 (xbfu_cmp0_noout): New combine pattern.
12985 (xbfu_cmp0"): Likewise.
12986 (movsi_set_cc_insn): Place the predicable variant first.
12987 (commutative_binary_cmp0_noout): Remove clobber.
12988 (commutative_binary_cmp0): New pattern.
12989 (noncommutative_binary_cmp0): Likewise.
12990 (noncommutative_binary_cmp0_noout): Likewise.
12991 (noncommutative_binary_comparison_result_used): Removed.
12992 (rsub_cmp0): New pattern.
12993 (rsub_cmp0_noout): Likewise.
12994 (extzvsi): Changed, keep only meaningful variants.
12995 (SQH, SEZ): New iterators.
12996 (SQH_postfix): New mode attribute.
12997 (SEZ_prefix): New code attribute.
12998 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12999 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13000 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13001 of numerical value.
13002 (noncommutative_operator): Check the availability of barrel
13005 2018-04-09 Richard Biener <rguenther@suse.de>
13007 PR tree-optimization/85284
13008 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13009 Only use the niter constraining form of simple_iv when the exit
13010 is always executed.
13012 2018-04-09 Tom de Vries <tom@codesourcery.com>
13015 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13016 (define_expand "*memory_barrier"): New define_expand.
13017 (define_insn "memory_barrier"): New insn.
13019 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13021 PR rtl-optimization/80463
13022 PR rtl-optimization/83972
13023 PR rtl-optimization/83480
13025 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13026 correct producer for the insn.
13027 (tidy_control_flow): Fixup seqnos in case of debug insns.
13029 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13031 PR rtl-optimization/83913
13033 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13034 different sched-times when merging exprs.
13036 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13038 PR rtl-optimization/83962
13040 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13041 tidy_fallthru_edge and tidy_control_flow.
13043 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13045 PR rtl-optimization/83530
13047 * sel-sched.c (force_next_insn): New global variable.
13048 (remove_insn_for_debug): When force_next_insn is true, also leave only
13049 next insn in the ready list.
13050 (sel_sched_region): When the region wasn't scheduled, make another pass
13051 over it with force_next_insn set to 1.
13053 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13055 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13057 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13058 for interrupt control.
13059 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13060 functions for interrupt control.
13061 * config/nds32/nds32-intrinsic.md: Likewise.
13062 * config/nds32/nds32_intrinsic.h: Likewise.
13063 * config/nds32/nds32.h (nds32_builtins): Likewise.
13065 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13067 * config/nds32/nds32.c (nds32_init_machine_status,
13068 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13069 strict_aligned_p field.
13070 (nds32_expand_to_rtl_hook): New function.
13071 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13072 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13074 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13075 Chung-Ju Wu <jasonwucj@gmail.com>
13077 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13078 * config/nds32/nds32-n7.md: New file.
13079 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13080 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13082 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13083 * config/nds32/nds32.md (pipeline_model): Add n7.
13084 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13085 * config/nds32/pipelines.md: Include n7 settings.
13087 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13088 Chung-Ju Wu <jasonwucj@gmail.com>
13090 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13091 * config/nds32/nds32-e8.md: New file.
13092 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13093 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13095 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13096 * config/nds32/nds32.md (pipeline_model): Add e8.
13097 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13098 * config/nds32/pipelines.md: Include e8 settings.
13100 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13101 Chung-Ju Wu <jasonwucj@gmail.com>
13103 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13104 * config/nds32/nds32-n8.md: New file.
13105 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13106 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13108 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13109 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13110 * config/nds32/nds32.md (pipeline_model): Add n8.
13111 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13112 * config/nds32/pipelines.md: Include n8 settings.
13114 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13115 Chung-Ju Wu <jasonwucj@gmail.com>
13117 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13118 * config/nds32/nds32-n9-2r1w.md: New file.
13119 * config/nds32/nds32-n9-3r2w.md: New file.
13120 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13121 nds32_register_ports): New or modify for cpu n9.
13122 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13124 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13125 * config/nds32/nds32-utils.c: New file.
13126 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13127 TARGET_MUL_SLOW): Define.
13128 * config/nds32/nds32.md (pipeline_model): New attribute.
13129 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13130 New options that support cpu n9.
13131 * config/nds32/pipelines.md: Include n9 settings.
13132 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13134 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13136 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13137 information if necessary.
13138 (output_cond_branch_compare_zero): Likewise.
13139 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13140 (nds32_target_alignment): Refine for alignment.
13141 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13142 (FUNCTION_BOUNDARY): Modify.
13143 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13145 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13147 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13149 * config/nds32/constants.md (unspec_volatile_element): Add values for
13150 TLB operation and data prefetch.
13151 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13152 functions for TLB operation and data prefetch.
13153 * config/nds32/nds32-intrinsic.md: Likewise.
13154 * config/nds32/nds32_intrinsic.h: Likewise.
13155 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13156 (nds32_print_operand): Likewise.
13157 * config/nds32/nds32.h (nds32_builtins): Likewise.
13159 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13160 Andrew Pinski <pinsika@gcc.gnu.org>
13162 PR middle-end/82976
13163 * match.pd: Use constant_boolean_node of correct type instead of
13164 boolean_true_node or boolean_false_node for simplifying
13165 pointer comparisons to zero.
13167 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13169 PR tree-optimization/80021
13170 * tree.c (verify_type_variant): Make error call in verify_variant_match
13171 translatable and remove final full stop.
13173 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13175 * config/nds32/constants.md (unspec_volatile_element): Add
13176 UNSPEC_VOLATILE_EH_RETURN.
13177 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13178 nds32_output_stack_pop): Support dwarf exception handling process.
13179 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13180 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13181 exception handling process.
13182 (nds32_compute_stack_frame): Likewise.
13183 (nds32_return_addr_rtx): Likewise.
13184 (nds32_initial_elimination_offset): Likewise.
13185 (nds32_expand_prologue): Likewise.
13186 (nds32_expand_epilogue): Likewise.
13187 (nds32_dynamic_chain_address): New function.
13188 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13189 exception handling.
13190 (DYNAMIC_CHAIN_ADDRESS): Define.
13191 (EH_RETURN_DATA_REGNO): Define.
13192 (EH_RETURN_STACKADJ_RTX): Define.
13193 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13194 patterns for dwarf exception handling.
13196 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13198 * config/nds32/nds32.h: Clean up obsolete macros.
13200 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13202 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13203 Add enum values for particular instructions.
13204 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13205 particular intrinsic functions.
13206 * config/nds32/nds32-intrinsic.md: Likewise.
13207 * config/nds32/nds32_intrinsic.h: Likewise.
13208 * config/nds32/nds32.h (nds32_builtins): Likewise.
13209 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13210 (btst, ave): New patterns for particular instructions.
13212 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13214 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13215 Add enum values for atomic load/store and memory sync.
13216 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13218 * config/nds32/nds32-intrinsic.md: Likewise.
13219 * config/nds32/nds32_intrinsic.h: Likewise.
13220 * config/nds32/nds32.h (nds32_builtins): Likewise.
13222 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13224 PR tree-optimization/85257
13225 * fold-const.c (native_encode_vector): If not all elts could fit
13226 and off is -1, return 0 rather than offset.
13227 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13228 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13229 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13230 adjust buffer in native_interpret_expr call.
13232 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13234 * config/nds32/constants.md (unspec_volatile_element): Add cache
13235 control enum values.
13236 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13237 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13238 * config/nds32/nds32.c (nds32_cctl_names): New.
13239 (nds32_print_operand): Handle cache control register names.
13240 * config/nds32/nds32.h (nds32_builtins): New enum values.
13241 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13243 * config/nds32/nds32.md (type): Add mmu.
13244 * config/nds32/pipelines.md (simple_insn): Add mmu.
13246 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13248 * config/nds32/nds32.md (type): Remove call.
13249 * config/nds32/pipelines.md (simple_insn): Likewise.
13251 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13253 * config/nds32/constants.md (unspec_volatile_element): Add
13254 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13255 UNSPEC_VOLATILE_FMFCFG.
13256 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13257 description for fmfcfg and fmfcsr.
13258 (bdesc_1arg): Add fmtcsr.
13259 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13260 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13261 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13262 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13263 unspec_fmfcfg): New patterns.
13264 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13265 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13266 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13267 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13268 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13269 __nds32__fmfcfg): Define.
13271 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13273 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13274 intrinsic register names.
13275 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13276 intrinsic register enum values and macros.
13278 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13280 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13281 for load/store addressing form.
13282 (nds32_print_operand_address): Likewise.
13284 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13287 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13288 based on LABEL_REF. Remove useless assertion.
13289 (pic_address_needs_scratch): Fix formatting.
13290 (sparc_legitimize_pic_address): Minor tweaks.
13291 (sparc_delegitimize_address): Adjust assertion accordingly.
13292 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13293 into symbolic_operand.
13294 (movsi_high_pic_label_ref): Likewise.
13295 (movsi_lo_sum_pic_label_ref): Likewise.
13296 (movdi_pic_label_ref): Likewise.
13297 (movdi_high_pic_label_ref): Likewise.
13298 (movdi_lo_sum_pic_label_ref): Likewise.
13300 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13302 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13303 custom LIB_SPEC setup.
13305 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13306 Kito Cheng <kito.cheng@gmail.com>
13308 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13309 * config/riscv/freebsd.h: New.
13311 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13313 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13314 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13317 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13318 Kito Cheng <kito.cheng@gmail.com>
13320 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13321 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13322 * config/nds32/nds32-protos.h (nds32_output_call,
13323 nds32_output_return): Declare.
13324 * config/nds32/nds32.md: Refine all the call and return patterns.
13326 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13329 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13330 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13332 PR rtl-optimization/84872
13333 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13334 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13335 EDGE_CROSSING edge.
13337 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13339 * expr.c (copy_blkmode_to_reg): Revert 254862.
13340 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13342 2018-04-06 Richard Biener <rguenther@suse.de>
13344 PR middle-end/85244
13345 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13346 after seeing a component reference with an adjacent field. Treat
13347 refs to arrays at struct end of external decls similar to
13348 refs to unconstrained commons.
13350 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13353 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13354 look through SAVE_EXPRs with non-side-effects argument. Adjust
13356 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13359 2018-04-06 Richard Biener <rguenther@suse.de>
13361 PR middle-end/85180
13362 * alias.c (find_base_term): New wrapper around find_base_term
13363 unwinding CSELIB_VAL_PTR changes.
13364 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13367 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13369 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13371 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13372 constant definitions.
13373 ("nop"): lr 0,0 -> nopr r0
13374 ("nop_lr0", "nop_lr1"): New insn definitions.
13376 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13378 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13379 NDS32_V3PUSH_AVAILABLE_P macro.
13381 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13382 Chung-Ju Wu <jasonwucj@gmail.com>
13384 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13385 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13386 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13387 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13388 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13389 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13390 * config/nds32/constraints.md: New constraints and checking for hard
13391 float configuration.
13392 * config/nds32/iterators.md: New mode iterator and attribute for hard
13393 float configuration.
13394 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13396 * config/nds32/nds32-fpu.md: New file.
13397 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13398 deal with hard float code generation.
13399 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13401 (abi_type, float_reg_number): New enum type.
13402 * config/nds32/nds32-predicates.c: New predicates for hard float.
13403 * config/nds32/nds32-protos.h: Declare functions for hard float.
13404 * config/nds32/nds32.c: Implementation for hard float configuration.
13405 * config/nds32/nds32.h: Definitions for hard float configuration.
13406 * config/nds32/nds32.md: Include hard float machine description and
13407 modify patterns for hard float configuration.
13408 * config/nds32/nds32.opt: New options for hard float configuration.
13409 * config/nds32/predicates.md: New predicates for hard float
13412 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13414 * common/config/nds32/nds32-common.c
13415 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13417 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13419 PR middle-end/85195
13420 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13421 CONSTRUCTOR_ELT (ctor, ...)->value.
13423 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13426 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13428 2018-04-05 Tom de Vries <tom@codesourcery.com>
13431 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13434 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13435 Kito Cheng <kito.cheng@gmail.com>
13437 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13438 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13439 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13440 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13442 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13443 Kito Cheng <kito.cheng@gmail.com>
13445 * config/nds32/constraints.md (Ufe): New memory constraint.
13446 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13447 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13448 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13450 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13451 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13453 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13455 * config/nds32/nds32.md: Use optimize_size in the condition for
13456 alu-shift instructions.
13458 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13460 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13462 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13464 * config/nds32/nds32.md (negsi2): Refine pattern.
13466 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13467 Chung-Ju Wu <jasonwucj@gmail.com>
13469 * config/nds32/iterators.md (shift_rotate): New code iterator.
13470 (shift): New code attribute.
13471 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13472 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13473 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13474 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13475 bit-wise operations.
13476 (andsi3, *andsi3): Ditto.
13477 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13478 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13479 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13480 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13481 nds32_ior_operand, nds32_xor_operand): New predicates.
13483 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13485 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13486 (addsi3, subsi3): ... this.
13488 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13490 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13492 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13494 * config/nds32/nds32.md: Adjust indention.
13496 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13498 * config/nds32/nds32.md (feature): New attribute.
13500 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13502 * config/nds32/nds32.md (subtype): New attribute.
13504 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13507 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13508 expansion to perform a bitwise AND of the argument followed by a
13509 boolean negation of the result.
13511 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13513 PR rtl-optimization/84878
13514 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13515 the basic block. Assert the use reference is not artificial and that
13516 it has an associated insn.
13518 2018-04-04 Michael Matz <matz@suse.de>
13520 * builtins.c (compute_objsize): Pass correct operand
13521 to array_at_struct_end_p.
13523 2018-04-04 Richard Biener <rguenther@suse.de>
13526 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13527 from contexts for DINFO_LEVEL_TERSE and below.
13529 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13531 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13532 resiter_operand condition.
13533 * config/nds32/nds32.md (*move<mode>): Ditto.
13535 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13536 Monk Chiang <sh.chiang04@gmail.com>
13538 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13540 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13542 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13544 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13545 Kito Cheng <kito.cheng@gmail.com>
13547 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13548 nds32_cond_code_str, output_cond_branch,
13549 output_cond_branch_compare_zero, nds32_expand_cbranch,
13550 nds32_expand_cstore, nds32_expand_movcc,
13551 nds32_output_cbranchsi4_equality_zero,
13552 nds32_output_cbranchsi4_equality_reg,
13553 nds32_output_cbranchsi4_equality_reg_or_const_int,
13554 nds32_output_cbranchsi4_greater_less_zero: New functions.
13555 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13556 nds32_expand_cstore, nds32_expand_movcc,
13557 nds32_output_cbranchsi4_equality_zero,
13558 nds32_output_cbranchsi4_equality_reg,
13559 nds32_output_cbranchsi4_equality_reg_or_const_int,
13560 nds32_output_cbranchsi4_greater_less_zero): Declare.
13561 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13562 nds32_rimm11s_operand): New predicates.
13563 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13564 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13567 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13569 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13570 * config/nds32/nds32.md: Ditto.
13571 * config/nds32/pipelines.md: Ditto.
13573 2018-04-04 Richard Biener <rguenther@suse.de>
13575 PR tree-optimization/85168
13576 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13577 propagating abnormals.
13579 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13581 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13583 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13584 Kito Cheng <kito.cheng@gmail.com>
13586 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13587 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13588 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13589 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13590 * config/nds32/nds32.md (sibcall_internal): New.
13591 (sibcall_register): Remove.
13592 (sibcall_immediate): Remove.
13593 (sibcall_value_internal): New.
13594 (sibcall_value_register): Remove.
13595 (sibcall_value_immediate): Remove.
13596 * config/nds32/predicates.md (nds32_general_register_operand): New.
13597 (nds32_call_address_operand): New.
13599 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13601 PR rtl-optimization/85167
13602 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13603 bb_defs if *split_p, instead preinitialize it to NULL.
13605 PR tree-optimization/85156
13606 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13607 evaluating the argument multiple times.
13609 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13611 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13613 (_mm_cvtpd_ps): Likewise.
13614 (_mm_cvttpd_epi32): Likewise.
13615 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13616 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13617 vector, pixel, and bool following altivec.h include.
13619 2018-04-03 Martin Sebor <msebor@redhat.com>
13621 * doc/extend.texi (Common Function Attributes): Clarify.
13622 (const attribute): Likewise.
13623 (pure attribute): Likewise.
13625 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13628 * config/i386/i386.c (ix86_expand_vector_set): Use
13629 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13631 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13633 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13634 instructions when changing rounding bits to preserve precision bits
13635 in the x87 control word.
13637 2018-04-03 Martin Liska <mliska@suse.cz>
13639 PR tree-optimization/82491
13640 * rtl.h (strip_offset_and_add): Replace += suboffset with
13641 poly_uint64 () + suboffset.
13643 2018-03-29 Martin Liska <mliska@suse.cz>
13644 Martin Jambor <mjambor@suse.cz>
13647 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13648 param_type is not an integral or pointer type.
13650 2018-04-03 Richard Biener <rguenther@suse.de>
13652 * sese.h (recompute_all_dominators): Remove.
13654 2018-04-02 Martin Sebor <msebor@redhat.com>
13656 * doc/invoke.texi (-Wrestrict): Fix typos.
13658 2018-04-02 Jim Wilson <jimw@sifive.com>
13660 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13661 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13662 (<optab>di3, <optab>si3_extend): Likewise.
13663 (<optab>si3_mask, <optab>si3_mask_1): New.
13664 (<optab>di3_mask, <optab>di3_mask_1): New.
13665 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13666 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13667 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13669 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13671 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13674 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13676 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13677 (nds32_canonicalize_comparison): New function.
13679 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13680 Kito Cheng <kito.cheng@gmail.com>
13681 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13683 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13684 * config/nds32/constants.md (unspec_volatile_element): Add
13685 UNSPEC_VOLATILE_RELAX_GROUP.
13686 * config/nds32/nds32-relax-opt.c: New file.
13687 * config/nds32/nds32-predicates.c
13688 (nds32_symbol_load_store_p): New function.
13689 * config/nds32/nds32-protos.h
13690 (nds32_symbol_load_store_p): Declare function.
13691 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13692 * config/nds32/nds32.c
13693 (nds32_register_pass): New function to register pass.
13694 (nds32_register_passes): New function to register passes.
13695 * config/nds32/nds32.md (relax_group): New pattern.
13696 * config/nds32/nds32.opt (mrelax-hint): New option.
13697 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13699 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13701 * config/nds32/t-nds32: Modify files dependency.
13703 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13705 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13706 (PROFILE_HOOK): Define its implementation.
13708 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13710 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13711 type and 32-bit size.
13713 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13715 PR middle-end/85090
13716 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13717 (V_128_256): New mode iterator.
13718 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13719 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13720 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13722 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13723 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13725 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13728 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13729 NaN inputs correctly.
13731 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13734 * config/rs6000/vsx.md (??r): New mode attribute.
13735 (*vsx_mov<mode>_64bit): Use it.
13736 (*vsx_mov<mode>_32bit): Likewise.
13738 2018-03-30 Martin Sebor <msebor@redhat.com>
13740 PR tree-optimization/84818
13741 * builtins.c (check_access): Use warning_n.
13743 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13746 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13748 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13751 2018-03-30 Julia Koval <julia.koval@intel.com>
13754 * x86-tune.def (movx, partial_reg_dependency): Enable for
13757 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13759 PR inline-asm/84985
13760 * lra-constraints.c (process_alt_operands): Move setting
13761 this_alternative_matches below.
13763 2018-03-29 Martin Liska <mliska@suse.cz>
13766 * doc/invoke.texi: Document how LTO works with debug info.
13767 Describe auto-load support of binutils. Mention 'x86-64'
13768 as valid option value of -march option.
13770 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13772 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13775 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13776 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13777 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13780 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13783 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13784 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13785 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13786 for RS6000_BTM_POWERPC64.
13787 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13788 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13789 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13792 (DIVDEU): Likewise.
13794 2018-03-28 Carl Love <cel@us.ibm.com>
13797 2017-09-27 Carl Love <cel@us.ibm.com>
13799 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13800 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13801 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13804 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13806 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13807 instead of __vector bool.
13808 (_mm_max_pu8): Likewise.
13809 (_mm_min_pi16): Likewise.
13811 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13814 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13815 (DIVWEUO): Likewise.
13816 (DIVDEO): Likewise.
13817 (DIVDEUO): Likewise.
13818 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13819 DIVWEUO and DIVDEUO.
13820 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13821 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13822 (div_extend): Likewise.
13823 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13825 (__builtin_divweuo): Likewise.
13826 (__builtin_divdeo): Likewise.
13827 (__builtin_divdeuo): Likewise.
13829 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13832 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13833 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13835 PR tree-optimization/82004
13836 * gimple-match-head.c (optimize_pow_to_exp): New function.
13837 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13838 Don't fold to exp if optimize_pow_to_exp is false.
13840 2018-03-28 Martin Liska <mliska@suse.cz>
13843 * calls.c (initialize_argument_information): Fix trailing space.
13844 * common.opt: Fix typo and provide better explanation for
13845 -fsanitize-coverage option.
13846 * config/i386/i386.opt: Fix typo.
13848 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13849 Martin Liska <mliska@suse.cz>
13852 * gimplify.c (asan_poison_variable): Don't do the check for
13853 gimplify_omp_ctxp here.
13854 (gimplify_decl_expr): Do it here.
13855 (gimplify_target_expr): Likewise.
13857 2018-03-28 Martin Liska <mliska@suse.cz>
13860 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13861 chkp_type_bounds_count if MPX is not enabled.
13863 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13865 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13867 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13870 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13871 function to create the function decl for complex long double
13872 multiply and divide for -mabi=ieeelongdouble.
13873 (init_float128_ieee): Call it.
13875 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13878 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13879 -fcf-protection=branch -mibt.
13880 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13882 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13885 * config/arm/arm.c (arm_valid_symbolic_address): Handle
13886 arm_word_relocations.
13888 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
13891 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13892 extern array declarations.
13894 2018-03-27 Richard Biener <rguenther@suse.de>
13896 PR middle-end/84067
13897 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13898 explicit single_use checks.
13900 2018-03-27 Richard Biener <rguenther@suse.de>
13902 PR tree-optimization/85082
13903 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13906 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13908 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13909 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13910 Turn on fasynchronous-unwind-tables and funwind-tables.
13912 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
13915 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13916 (*bmi_blsr_<mode>_ccz): Ditto.
13918 2018-03-26 Tom de Vries <tom@codesourcery.com>
13920 PR tree-optimization/85063
13921 * omp-general.c (offloading_function_p): New function. Factor out
13923 * omp-offload.c (pass_omp_target_link::gate): ... here.
13924 * omp-general.h (offloading_function_p): Declare.
13925 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13926 with attribute omp declare target for offloading functions.
13928 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
13930 PR tree-optimization/84005
13931 * tree-data-ref.h (get_base_for_alignment): Declare.
13932 * tree-data-ref.c (get_base_for_alignment_1): New function.
13933 (get_base_for_alignment): Likewise.
13934 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13935 get_base_for_alignment to find a suitable base object, instead
13936 of always using drb->base_address.
13938 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13940 PR inline-asm/85022
13941 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13942 known size by default.
13944 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
13946 PR inline-asm/85030
13947 * lra-constraints.c (process_alt_operands): Don't match BLKmode
13948 and non BLKmode operands.
13950 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13953 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13954 Clean up attributes.
13956 2018-03-23 Richard Biener <rguenther@suse.de>
13959 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13960 we are going to emit early debug for LTO.
13962 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13964 PR inline-asm/85034
13965 * function.c (match_asm_constraints_1): Don't optimize if input
13966 doesn't satisfy general_operand predicate for output's mode.
13968 PR inline-asm/85022
13969 * alias.c (write_dependence_p): Don't require for x_canonicalized
13970 non-VOIDmode if x has VOIDmode.
13973 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13974 just don't try to optimize it rather than assert it never happens.
13976 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13978 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13979 macro expansions for definition of ST_INTERNAL_<mode> and
13980 LD_INTERNAL_<mode> builtins.
13981 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13983 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13985 (altivec_expand_st_builtin): Likewise.
13986 (altivec_expand_builtin): Remove calls to deleted functions.
13987 (rs6000_address_for_altivec): Delete this function.
13988 * config/rs6000/vector.md: Remove expands for
13989 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13991 2018-03-22 Sudakshina Das <sudi.das@arm.com>
13994 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13995 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13996 re-computing once computed.
13997 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13998 (arm_init_machine_status): Initialize
13999 machine->static_chain_stack_bytes.
14001 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14004 * doc/extend.texi: Add four new prototypes for vec_ld.
14005 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14006 definitions for more logical presentation.
14007 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14008 entries for V1TI variants of __builtin_altivec_ld builtin.
14009 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14010 handling of V1TI variant of LVX icode pattern.
14011 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14012 (rs6000_gimple_fold_builtin): Likewise.
14013 (altivec_init_builtins): Add code to define
14014 __builtin_altivec_lvx_v1ti function.
14016 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14018 PR inline-asm/84941
14019 * function.c (match_asm_constraints_1): Don't do the optimization
14020 if input isn't a REG, SUBREG, MEM or constant.
14022 2018-03-22 Tom de Vries <tom@codesourcery.com>
14024 PR tree-optimization/84956
14025 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14026 bb_has_abnormal_pred.
14028 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14031 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14032 DECL_INITIAL (decl) to decl at the end.
14033 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14034 adjust the comment.
14036 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14038 * doc/extend.texi (__builtin_tgmath): Document when complex
14039 integer types are treated as _Complex _Float64.
14041 2018-03-21 Tom de Vries <tom@codesourcery.com>
14043 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14045 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14047 PR tree-optimization/84960
14048 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14049 if it is ENTRY block, move them into single succ of ENTRY in that case.
14051 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14053 PR tree-optimization/84811
14054 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14055 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14056 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14057 (known_subrange_p): Update accordingly. Cast each value involved
14058 in the size comparison, rather than casting the result of the
14061 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14063 PR tree-optimization/84982
14064 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14065 by flipping the least significant bit rather than all bits from
14066 bitpos to bitpos + bitsize - 1.
14068 2018-03-21 Nathan Sidwell <nathan@acm.org>
14070 * doc/extend.texi (Deprecated Features): Remove mention of
14071 long-deleted deprecations.
14073 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14076 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14077 * configure: Regenerate.
14079 2018-03-21 Tom de Vries <tom@codesourcery.com>
14081 PR tree-optimization/83126
14082 * tree-parloops.c (num_phis): New function.
14083 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14085 2018-03-21 Nathan Sidwell <nathan@acm.org>
14087 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14088 mention anon-struct/union members and trailing attributes.
14090 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14092 PR tree-optimization/84969
14093 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14094 builtin memset partitions if they set different rhs values.
14096 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14098 PR rtl-optimization/84989
14099 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14100 VEC_DUPLICATE with scalar result mode.
14102 2018-03-21 Martin Liska <mliska@suse.cz>
14105 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14106 not intended return statement.
14108 2018-03-21 Martin Liska <mliska@suse.cz>
14111 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14112 (chkp_find_bound_slots_1): Limit number of iterations.
14114 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14117 * Minor grammar fixes for x86 options.
14119 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14122 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14123 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14125 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14128 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14129 (altivec_lvx_<mode>_1op): Likewise.
14130 (altivec_stvx_<mode>_2op): Likewise.
14131 (altivec_stvx_<mode>_1op): Likewise.
14132 (altivec_lvx_<VM2:mode>): New define_expand.
14133 (altivec_stvx_<VM2:mode>): Likewise.
14134 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14135 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14136 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14137 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14138 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14139 (rs6000_gen_lvx): Likewise.
14140 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14141 (altivec_expand_stv_builtin): Likewise.
14142 (altivec_expand_builtin): Likewise.
14143 * config/rs6000/vector.md: Likewise.
14145 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14148 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14151 2018-03-20 Richard Biener <rguenther@suse.de>
14154 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14155 sign-conversions as zero, fall back to standard scalar_stmt
14158 2018-03-20 Martin Liska <mliska@suse.cz>
14161 * predict.c (rebuild_frequencies): Handle case when we have
14162 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14164 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14167 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14168 flag_section_anchors.
14169 * varasm.c (use_blocks_for_decl_p): Remove hack for
14170 dw2_force_const_mem.
14173 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14175 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14176 be created, use lowpart_subreg of operands[0] rather than operands[0]
14178 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14179 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14180 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14181 and n constraint instead of aarch64_shift_imm_di and Usd.
14182 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14183 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14185 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14188 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14189 to favor GPR over NEON registers.
14190 (<shift>di3_neon): Likewise.
14192 2018-03-20 Tom de Vries <tom@codesourcery.com>
14195 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14196 (nvptx_process_pars): Emit bar.sync asap and alap.
14198 2018-03-20 Tom de Vries <tom@codesourcery.com>
14201 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14202 seen_label if seen_label is already set.
14204 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14207 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14208 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14209 Use 1U instead of 1. Formatting fixes.
14212 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14213 instead of TREE_TYPE (s1) for the return value.
14215 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14217 PR tree-optimization/84946
14218 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14219 bitsize + bitsize in poly_uint64 rather than poly_int64.
14222 * dwarf2asm.c: Include fold-const.c.
14223 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14224 of decl rather than decl itself.
14226 PR rtl-optimization/84643
14227 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14229 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14232 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14233 calling assemble_variable.
14235 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14238 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14239 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14241 2018-03-19 Jim Wilson <jimw@sifive.com>
14244 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14245 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14246 (riscv_first_stack_step): Likewise.
14247 (riscv_option_override): Use STACK_BOUNDARY instead of
14248 MIN_STACK_BOUNDARY.
14249 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14250 MIN_STACK_BOUNDARY.
14251 (BIGGEST_ALIGNMENT): Set to 128.
14252 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14253 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14256 2018-03-19 Richard Biener <rguenther@suse.de>
14258 PR tree-optimization/84933
14259 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14260 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14262 2018-03-19 Richard Biener <rguenther@suse.de>
14264 PR tree-optimization/84859
14265 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14266 (cond_if_else_store_replacement): Perform sinking operation on
14267 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14268 Generalize what a BB with a single eligible store is.
14270 2018-03-19 Richard Biener <rguenther@suse.de>
14272 PR tree-optimization/84929
14273 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14274 chrec_is_positive against non-chrec arg.
14276 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14279 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14281 2018-03-18 Martin Liska <mliska@suse.cz>
14283 PR rtl-optimization/84635
14284 * regrename.c (build_def_use): Use matches_mode only when
14287 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14289 PR tree-optimization/84913
14290 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14291 vectorize chains of COND_EXPRs.
14293 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14295 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14297 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14299 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14301 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14303 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14305 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14306 Kito Cheng <kito.cheng@gmail.com>
14308 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14309 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14310 (nds32_adjust_reg_alloc_order): New function.
14311 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14313 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14315 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14316 nds32_print_operand, nds32_print_operand_address): Use
14317 HOST_WIDE_INT_PRINT_DEC instead.
14319 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14321 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14323 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14326 * config/i386/i386.c (initial_ix86_tune_features,
14327 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14328 unsigned long long.
14329 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14330 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14331 rather than 1u << ix86_tune. Formatting fix.
14332 (ix86_option_override_internal): Change ix86_arch_mask from
14333 unsigned int to unsigned HOST_WIDE_INT, initialize to
14334 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14335 (ix86_function_specific_restore): Likewise.
14337 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14340 * postreload.c (reload_combine_recognize_pattern): Perform
14341 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14342 truncate_int_for_mode the result for the destination's mode.
14345 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14346 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14348 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14351 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14354 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14355 regno_allocno_class_array and sorted_pseudos.
14356 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14357 insns where regno is used.
14359 2018-03-16 Martin Liska <mliska@suse.cz>
14362 * multiple_target.c (create_dispatcher_calls): Redirect
14363 reference in the symbol table.
14365 2018-03-16 Martin Liska <mliska@suse.cz>
14368 * multiple_target.c (create_dispatcher_calls): Redirect also
14371 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14375 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14376 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14377 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14379 2018-03-16 Julia Koval <julia.koval@intel.com>
14381 * doc/invoke.texi (Skylake Server): Add CLWB.
14382 Cannonlake): Remove CLWB.
14384 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14386 PR tree-optimization/84841
14387 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14389 (FLOAT_ONE_CONST_TYPE): Define.
14390 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14391 (sort_by_operand_rank): Put entries with higher constant_type last
14392 rather than first to match comments.
14394 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14396 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14399 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14402 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14403 check and use address of target always.
14405 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14408 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14409 (indirect_thunk_bnd_needed): Likewise.
14410 (indirect_thunks_used): Likewise.
14411 (indirect_thunks_bnd_used): Likewise.
14412 (indirect_return_needed): New.
14413 (indirect_return_bnd_needed): Likewise.
14414 (output_indirect_thunk_function): Add a bool argument for
14416 (output_indirect_thunk_function): Don't generate alias for
14417 function return thunk.
14418 (ix86_code_end): Call output_indirect_thunk_function to generate
14419 function return thunks.
14420 (ix86_output_function_return): Set indirect_return_bnd_needed
14421 and indirect_return_needed instead of indirect_thunk_bnd_needed
14422 and indirect_thunk_needed.
14424 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14426 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14427 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14428 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14430 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14431 Paul Hua <paul.hua.gm@gmail.com>
14434 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14436 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14438 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14439 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14440 resp. SFmode cases.
14442 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14445 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14446 instead of GET_MODE_SIZE when comparing Units.
14448 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14451 * varasm.c (hash_section): Return an unchangeble hash value
14452 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14453 Return !aarch64_can_use_per_function_literal_pools_p ().
14455 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14458 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14459 rather than address of cmode as last argument to prepare_cmp_insn.
14461 2018-03-15 Julia Koval <julia.koval@intel.com>
14463 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14464 F_AVX512VNNI, F_AVX512BITALG): New.
14466 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14469 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14470 insn for floating-point loads and stores.
14472 2018-03-14 Carl Love <cel@us.ibm.com>
14474 * config/rs6000/rs6000-c.c: Add macro definitions for
14475 ALTIVEC_BUILTIN_VEC_PERMXOR.
14476 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14477 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14478 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14479 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14481 * config/doc/extend.texi: Add prototypes for vec_permxor.
14483 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14486 * diagnostic-show-locus.c (class layout_point): Convert m_line
14487 from int to linenum_type.
14488 (line_span::comparator): Use linenum "compare" function when
14489 comparing line numbers.
14490 (test_line_span): New function.
14491 (layout_range::contains_point): Convert param "row" from int to
14493 (layout_range::intersects_line_p): Likewise.
14494 (layout::will_show_line_p): Likewise.
14495 (layout::print_source_line): Likewise.
14496 (layout::should_print_annotation_line_p): Likewise.
14497 (layout::print_annotation_line): Likewise.
14498 (layout::print_leading_fixits): Likewise.
14499 (layout::annotation_line_showed_range_p): Likewise.
14500 (struct line_corrections): Likewise for field m_row.
14501 (line_corrections::line_corrections): Likewise for param "row".
14502 (layout::print_trailing_fixits): Likewise.
14503 (layout::get_state_at_point): Likewise.
14504 (layout::get_x_bound_for_row): Likewise.
14505 (layout::print_line): Likewise.
14506 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14507 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14508 * input.c (selftest::test_linenum_comparisons): New function.
14509 (selftest::input_c_tests): Call it.
14510 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14511 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14512 * selftest.h (ASSERT_GT): New macro.
14513 (ASSERT_GT_AT): New macro.
14514 (ASSERT_LT): New macro.
14515 (ASSERT_LT_AT): New macro.
14517 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14519 PR rtl-optimization/84780
14520 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14522 2018-03-14 Martin Liska <mliska@suse.cz>
14524 * tree.c (record_node_allocation_statistics): Use
14525 get_stats_node_kind.
14526 (get_stats_node_kind): New function extracted from
14527 record_node_allocation_statistics.
14528 (free_node): Use get_stats_node_kind.
14530 2018-03-14 Richard Biener <rguenther@suse.de>
14532 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14533 that the value-set of ANTIC_IN doesn't grow.
14536 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14538 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14539 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14541 2018-03-14 Julia Koval <julia.koval@intel.com>
14543 * config.gcc (icelake-client, icelake-server): New.
14545 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14546 (initial_ix86_arch_features): Ditto.
14547 (PTA_SKYLAKE): Add SGX.
14548 (PTA_ICELAKE): Remove.
14549 (PTA_ICELAKE_CLIENT): New.
14550 (PTA_ICELAKE_SERVER): New.
14551 (ix86_option_override_internal): Split up icelake on icelake client and
14553 (get_builtin_code_for_version): Ditto.
14554 (fold_builtin_cpu): Ditto.
14555 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14556 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14557 * config/i386/i386.h (processor_type): Ditto.
14558 * doc/invoke.texi: Ditto.
14560 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14563 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14564 INTEGER_CST offset, add it together with bitpos / 8 and
14565 sign extend based on POINTER_SIZE.
14569 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14572 * config/i386/constraints.md (Yc): New register constraint.
14573 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14574 Use Yc constraint for alternative 2 of operand 0. Remove
14575 preferred_for_speed attribute.
14577 2018-03-14 Richard Biener <rguenther@suse.de>
14579 PR tree-optimization/84830
14580 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14581 with the old one to avoid oscillations.
14583 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14586 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14588 (assign_by_spills): Return a flag of reload assignment failure.
14589 Do not process the reload assignment failures. Do not spill other
14590 reload pseudos if they has the same reg class. Update n if
14592 (lra_assign): Add a return arg. Set up from the result of
14593 assign_by_spills call.
14594 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14595 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14596 usage_insns if it is not NULL.
14597 (spill_hard_reg_in_range): New function.
14598 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14599 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14600 function prototypes.
14601 (lra_assign): Change prototype.
14602 * lra.c (lra): Add code to deal with fails by splitting hard reg
14605 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14607 * config/riscv/riscv.opt (mrelax): New option.
14608 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14609 "norelax" when riscv_mrelax is disabled.
14610 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14612 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14615 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14616 reassociation for int modes.
14618 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14620 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14621 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14623 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14624 * config/aarch64/aarch64-sve.md
14625 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14626 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14627 (*extend<mode><Vwide>2): Rename to...
14628 (aarch64_sve_extend<mode><Vwide>2): ...this.
14629 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14630 renaming the old pattern to...
14631 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14633 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14634 define_expand, renaming the old pattern to...
14635 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14636 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14637 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14638 account when deciding which SVE instruction the optab should use.
14639 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14641 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14643 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14644 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14645 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14646 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14647 (tlsdesc_small_<mode>): Turn a define_expand and use
14648 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14649 (tlsdesc_small_advsimd_<mode>): ...this.
14650 (tlsdesc_small_sve_<mode>): New pattern.
14652 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14654 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14655 (UNSPEC_UMUL_HIGHPART): New constants.
14656 (MUL_HIGHPART): New int iteraor.
14657 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14658 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14660 (*<su>mul<mode>3_highpart): New define_insn.
14662 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14665 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14668 2018-03-13 Martin Liska <mliska@suse.cz>
14671 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14673 (sem_item_optimizer::~sem_item_optimizer): Release it.
14674 (sem_item_optimizer::merge_classes): Register variable aliases.
14675 (sem_item_optimizer::fixup_pt_set): New function.
14676 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14677 * ipa-icf.h: Declare new variables and functions.
14679 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14681 PR middle-end/84834
14682 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14683 integer_pow2p@2 and test integer_pow2p in condition.
14684 (A < 0 ? C : 0): Similarly for @1.
14686 PR middle-end/84831
14687 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14688 characters starting at p contain '\0' character, don't look beyond
14692 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14693 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14696 * reg-stack.c (change_stack): Change update_end var from int to
14697 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14698 also call set_block_for_insn on the newly added insns and rescan.
14701 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14702 on the last operand.
14705 * tree.c (stabilize_reference_1): Return save_expr (e) for
14706 STATEMENT_LIST even if it doesn't have side-effects.
14708 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14710 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14712 2018-03-12 Renlin Li <renlin.li@arm.com>
14714 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14715 aarch64_output_scalar_simd_mov_immediate.
14717 2018-03-12 Martin Sebor <msebor@redhat.com>
14719 PR tree-optimization/83456
14720 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14721 for perfectly overlapping calls to memcpy.
14722 (gimple_fold_builtin_memory_chk): Same.
14723 (gimple_fold_builtin_strcpy): Handle no-warning.
14724 (gimple_fold_builtin_stxcpy_chk): Same.
14725 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14727 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14729 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14730 parameter. Use it for SFmode.
14731 (rs6000_function_arg_advance_1): Adjust.
14732 (rs6000_function_arg): Adjust.
14733 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14735 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14737 PR rtl-optimization/84169
14738 PR rtl-optimization/84780
14739 * combine.c (can_combine_p): Check for a 2-insn combination whether
14740 the destination register is used between the two insns, too.
14742 2018-03-12 Richard Biener <rguenther@suse.de>
14744 PR tree-optimization/84803
14745 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14746 for refs DR analysis didn't process.
14748 2018-03-12 Richard Biener <rguenther@suse.de>
14750 PR tree-optimization/84777
14751 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14752 force-vectorize loops ignore whether we are optimizing for size.
14754 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14756 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14757 (TARGET_MD_ASM_ADJUST): Define.
14759 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14760 Kito Cheng <kito.cheng@gmail.com>
14761 Chung-Ju Wu <jasonwucj@gmail.com>
14763 * config/nds32/nds32.c (nds32_compute_stack_frame,
14764 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14765 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14766 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14767 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14768 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14769 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14770 * config/nds32/nds32.md (prologue, epilogue): Use macro
14771 NDS32_V3PUSH_AVAILABLE_P to do checking.
14773 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14776 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14777 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14778 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14779 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14780 addition of most attributes on !orig_type_die or the attribute not
14781 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14783 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14784 Chung-Ju Wu <jasonwucj@gmail.com>
14786 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14787 __NDS32_VH__ macro.
14788 * config/nds32/nds32.opt (mvh): New option.
14790 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14791 Chung-Ju Wu <jasonwucj@gmail.com>
14793 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14795 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14796 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14799 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14800 Chung-Ju Wu <jasonwucj@gmail.com>
14802 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14804 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14805 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14807 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14808 Kito Cheng <kito.cheng@gmail.com>
14809 Chung-Ju Wu <jasonwucj@gmail.com>
14811 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14812 UNSPEC_FFMISM and UNSPEC_FLMISM.
14813 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14814 for ffb, ffmism and flmism.
14815 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14816 (unspec_ffmism): Ditto.
14817 (unspec_flmism): Ditto.
14818 (nds32_expand_builtin_impl): Check if string extension is available.
14819 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14820 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14822 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14825 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14828 * lra-assigns.c (assign_by_spills): Return a flag of reload
14829 assignment failure. Do not process the reload assignment
14830 failures. Do not spill other reload pseudos if they has the same
14832 (lra_assign): Add a return arg. Set up from the result of
14833 assign_by_spills call.
14834 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14835 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14836 usage_insns if it is not NULL.
14837 (spill_hard_reg_in_range): New function.
14838 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14839 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14840 function prototypes.
14841 (lra_assign): Change prototype.
14842 * lra.c (lra): Add code to deal with fails by splitting hard reg
14845 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14848 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14850 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14853 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14854 (dw_val_node): Add val_symbolic_view.
14855 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14856 (symview_upper_bound): New.
14857 (new_line_info_table): Initialize symviews_since_reset.
14858 (dwarf2out_source_line): Count symviews_since_reset and set
14859 symview_upper_bound.
14860 (dw_val_equal_p): Handle symview.
14861 (add_AT_symview): New.
14862 (print_dw_val): Handle symview.
14863 (attr_checksum, attr_checksum_ordered): Likewise.
14864 (same_dw_val_p, size_of_die): Likewise.
14865 (value_format, output_die): Likewise.
14866 (add_high_low_attributes): Use add_AT_symview for entry_view.
14867 (dwarf2out_finish): Reset symview_upper_bound, clear
14870 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14873 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14874 Add strict argument and use it.
14875 (rs6000_split_multireg_move): Update for new strict argument.
14876 (mem_operand_gpr): Disallow all non-offsettable addresses.
14877 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14879 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14882 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14883 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14884 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14887 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14888 decl, use remap_type if we want to use the type.
14890 2018-03-09 Martin Sebor <msebor@redhat.com>
14892 PR tree-optimization/84526
14893 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14895 (builtin_access::generic_overlap): Be prepared to handle non-array
14898 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
14900 PR rtl-optimization/84682
14901 * lra-constraints.c (process_address_1): Check is_address flag
14902 for address constraints.
14903 (process_alt_operands): Likewise.
14904 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14905 preprocess_constraints.
14906 * recog.h (preprocess_constraints): Add oploc parameter.
14908 * recog.c (preprocess_constraints): Test address_operand for
14909 CT_ADDRESS constraints.
14911 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14914 * lra-assigns.c (assign_by_spills): Return a flag of reload
14915 assignment failure. Do not process the reload assignment
14916 failures. Do not spill other reload pseudos if they has the same
14918 (lra_assign): Add a return arg. Set up from the result of
14919 assign_by_spills call.
14920 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14921 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14922 usage_insns if it is not NULL.
14923 (spill_hard_reg_in_range): New function.
14924 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14925 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14926 function prototypes.
14927 (lra_assign): Change prototype.
14928 * lra.c (lra): Add code to deal with fails by splitting hard reg
14931 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14934 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14935 Accept complain bool parameter. Only emit errors if it is true.
14936 (arm_parse_cpu_option_name): Likewise.
14937 (arm_target_thumb_only): Adjust callers of the above.
14938 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14939 prototype to take a default true bool parameter.
14940 (arm_parse_arch_option_name): Likewise.
14942 2018-03-09 David Malcolm <dmalcolm@redhat.com>
14943 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
14947 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14948 * configure: Regenerate.
14949 * configure.ac ("linker --version-script option"): New.
14950 ("linker soname option"): New.
14952 2018-03-09 Richard Biener <rguenther@suse.de>
14954 PR tree-optimization/84775
14955 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14956 immediate uses of predicate stmts and mark them modified.
14959 PR tree-optimization/84178
14960 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14962 (version_loop_for_if_conversion): Delay update_ssa call.
14963 (tree_if_conversion): Delay update_ssa until after predicate
14966 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
14969 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14970 when the function accesses prior frames.
14972 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14975 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14976 gen_llsym, otherwise call maybe_gen_llsym.
14978 PR inline-asm/84742
14979 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14980 has ',' character inside of it.
14982 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14985 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14986 as clobbering CC_REGNUM.
14988 2018-03-08 Richard Biener <rguenther@suse.de>
14990 PR middle-end/84552
14991 * tree-scalar-evolution.c: Include tree-into-ssa.h.
14992 (follow_copies_to_constant): Do not follow SSA names registered
14995 2018-03-08 Richard Biener <rguenther@suse.de>
14997 PR tree-optimization/84178
14998 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15000 (version_loop_for_if_conversion): Delay update_ssa call.
15001 (tree_if_conversion): Delay update_ssa until after predicate
15004 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15006 PR tree-optimization/84178
15007 * tree-if-conv.c (release_bb_predicate): Remove the
15008 the assertion that the stmts have NULL use_ops.
15009 Discard the statements, asserting that they haven't
15010 yet been added to a BB.
15012 2018-03-08 Richard Biener <rguenther@suse.de>
15014 PR tree-optimization/84746
15015 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15016 (phi_translate): Pass in destination ANTIC_OUT set.
15017 (phi_translate_1): Likewise. For a simplified result lookup
15018 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15019 (phi_translate_set): Adjust.
15020 (do_pre_regular_insertion): Likewise.
15021 (do_pre_partial_partial_insertion): Likewise.
15023 2018-03-08 Martin Liska <mliska@suse.cz>
15025 PR gcov-profile/84735
15026 * doc/gcov.texi: Document usage of profile files.
15027 * gcov-io.h: Document changes in the format.
15029 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15033 * dwarf2out.c (struct dw_line_info_table): Update comments for
15035 (FORCE_RESET_NEXT_VIEW): New.
15036 (FORCE_RESETTING_VIEW_P): New.
15037 (RESETTING_VIEW_P): Check for -1 too.
15038 (ZERO_VIEW_P): Likewise.
15039 (new_line_info_table): Force-reset next view.
15040 (dwarf2out_begin_function): Likewise.
15041 (dwarf2out_source_line): Simplify zero_view_p initialization.
15042 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15043 view directly. Omit view when omitting .loc at line 0.
15045 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15047 PR tree-optimization/84740
15048 * tree-switch-conversion.c (process_switch): Call build_constructors
15049 only if info.phi_count is non-zero.
15051 PR tree-optimization/84739
15052 * tree-tailcall.c (find_tail_calls): Check call arguments against
15053 DECL_ARGUMENTS (current_function_decl) rather than
15054 DECL_ARGUMENTS (func) when checking for tail recursion.
15056 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15058 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15059 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15060 Volker Reichelt's entry and add entries for people that perform
15061 GCC fuzzy testing and report numerous bugs.
15063 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15066 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15067 readonly data in sdata, if that is disabled.
15068 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15069 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15070 -mreadonly-in-sdata option.
15072 2018-03-07 Martin Sebor <msebor@redhat.com>
15074 PR tree-optimization/84468
15075 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15076 basic block when looking for nul assignment.
15078 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15081 * except.h (output_function_exception_table): Adjust prototype.
15082 * except.c (output_function_exception_table): Remove FNNAME parameter
15083 and add SECTION parameter. Ouput one part of the table at a time.
15084 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15085 the first part of the exception table and emit unwind directives.
15086 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15087 (i386_pe_seh_cold_init): Likewise.
15088 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15089 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15090 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15091 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15092 * config/i386/winnt.c: Include except.h.
15093 (struct seh_frame_state): Add reg_offset, after_prologue and
15094 in_cold_section fields.
15095 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15096 (i386_pe_seh_cold_init): New function.
15097 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15098 to seh->in_cold_section.
15099 (seh_emit_push): Record the offset of the push.
15100 (seh_emit_save): Record the offet of the save.
15101 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15102 Test seh->after_prologue to disregard the epilogue.
15103 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15104 (i386_pe_end_cold_function): New function.
15106 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15109 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15110 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15113 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15115 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15116 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15118 PR middle-end/84723
15119 * multiple_target.c: Include tree-inline.h and intl.h.
15120 (expand_target_clones): Diagnose and fail if node->definition and
15121 !tree_versionable_function_p (node->decl).
15123 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15125 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15127 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15128 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15129 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15131 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15134 * combine.c (try_combine): Use reg_or_subregno instead of handling
15135 just paradoxical SUBREGs and REGs.
15137 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15139 * config/arc/arc.c (arc_finalize_pic): Remove function.
15140 (arc_must_save_register): We use single base PIC register, remove
15141 checks to save/restore the PIC register.
15142 (arc_expand_prologue): Likewise.
15143 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15145 (arc_verify_short): Likewise.
15146 (arc_attr_type): Likewise.
15147 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15148 (walk_stores): Likewise.
15149 (arc_address_cost): Make it static.
15150 (arc_verify_short): Likewise.
15151 (branch_dest): Likewise.
15152 (arc_attr_type): Likewise.
15153 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15154 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15155 (arc_final_prescan_insn): Remove inserting the nops due to
15156 hardware hazards. It is done in reorg step.
15157 (insn_length_variant_t): Remove.
15158 (insn_length_parameters_t): Likewise.
15159 (arc_insn_length_parameters): Likewise.
15160 (arc_get_insn_variants): Likewise.
15161 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15163 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15165 PR inline-asm/84683
15166 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15169 PR tree-optimization/84687
15170 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15172 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15174 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15176 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15177 Rename to ppc_speculation_barrier.
15178 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15179 __builtin_ppc_speculation_barrier.
15181 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15184 * combine.c (combine_simplify_rtx): Don't try to simplify if
15185 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15188 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15190 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15191 to 32 bytes when compiling for POWER9.
15193 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15196 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15197 regparm >= 3 with no arg reg available also for calls with
15198 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15201 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15203 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15205 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15208 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15210 2018-03-05 Richard Biener <rguenther@suse.de>
15212 PR tree-optimization/84486
15213 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15214 When inserting a __builtin_assume_aligned call set the LHS
15215 SSA name alignment info accordingly.
15217 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15219 PR tree-optimization/84114
15220 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15221 Avoid reassociation of FLOAT_MODE addition.
15223 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15225 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15226 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15227 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15228 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15229 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15230 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15231 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15233 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15234 __builtin_ia32_wbinvd): New builtins.
15235 (SPECIAL_ARGS2): New.
15236 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15237 (SPECIAL_ARGS2): New.
15238 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15239 (ix86_valid_target_attribute_inner_p): Ditto.
15240 (ix86_init_mmx_sse_builtins): Add special_args2.
15241 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15242 TARGET_WBNOINVD_P): New.
15243 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15244 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15245 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15246 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15247 * config/i386/pconfigintrin.h: New file.
15248 * config/i386/wbnoinvdintrin.h: Ditto.
15249 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15251 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15253 2018-03-05 Richard Biener <rguenther@suse.de>
15255 PR tree-optimization/84670
15256 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15258 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15259 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15260 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15261 doesn't grow if all successors (recursively) were visited at least
15264 2018-03-05 Richard Biener <rguenther@suse.de>
15266 PR tree-optimization/84650
15267 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15268 if executed in the loop pipeline.
15270 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15272 * doc/configfiles.texi (Configuration Files): Move info about
15273 conditionalizing $target-protos.h to...
15274 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15275 differs from $target-protos.h.
15277 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15278 Chung-Ju Wu <jasonwucj@gmail.com>
15280 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15281 * config/nds32/nds32-multiple.md (setmemsi): Define.
15282 * config/nds32/nds32-memory-manipulation.c
15283 (nds32_gen_dup_4_byte_to_word_value): New.
15284 (emit_setmem_word_loop): New.
15285 (emit_setmem_byte_loop): New.
15286 (nds32_expand_setmem_loop): New.
15287 (nds32_expand_setmem_loop_v3m): New.
15288 (nds32_expand_setmem_unroll): New.
15289 (nds32_expand_setmem): New.
15291 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15292 Chung-Ju Wu <jasonwucj@gmail.com>
15294 * config/nds32/nds32-memory-manipulation.c
15295 (nds32_emit_load_store): New.
15296 (nds32_emit_post_inc_load_store): New.
15297 (nds32_emit_mem_move): New.
15298 (nds32_emit_mem_move_block): New.
15299 (nds32_expand_movmemsi_loop_unknown_size): New.
15300 (nds32_expand_movmemsi_loop_known_size): New.
15301 (nds32_expand_movmemsi_loop): New.
15302 (nds32_expand_movmemsi_unroll): New.
15303 (nds32_expand_movmemqi): Rename ...
15304 (nds32_expand_movmemsi): ... to this.
15305 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15306 (movmemsi): ... to this.
15307 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15308 (nds32_expand_movmemsi): ... to this.
15310 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15311 Monk Chiang <sh.chiang04@gmail.com>
15312 Chung-Ju Wu <jasonwucj@gmail.com>
15314 * config/nds32/nds32-protos.h
15315 (nds32_expand_load_multiple): New arguments.
15316 (nds32_expand_store_multiple): Ditto.
15317 (nds32_valid_multiple_load_store): Rename ...
15318 (nds32_valid_multiple_load_store_p): ... to this.
15319 * config/nds32/nds32-memory-manipulation.c
15320 (nds32_expand_load_multiple): Refine implementation.
15321 (nds32_expand_store_multiple): Ditto.
15322 * config/nds32/nds32-multiple.md
15323 (load_multiple): Update nds32_expand_load_multiple interface.
15324 (store_multiple): Update nds32_expand_store_multiple interface.
15325 * config/nds32/nds32-predicates.c
15326 (nds32_valid_multiple_load_store): Rename ...
15327 (nds32_valid_multiple_load_store_p): ... to this and refine
15329 * config/nds32/predicates.md
15330 (nds32_load_multiple_and_update_address_operation): New predicate.
15331 (nds32_store_multiple_and_update_address_operation): New predicate.
15333 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15334 Chung-Ju Wu <jasonwucj@gmail.com>
15336 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15337 (combo): New attribute.
15338 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15340 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15342 * config/nds32/nds32.opt: Change -mcmodel= default value.
15344 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15345 Monk Chiang <sh.chiang04@gmail.com>
15346 Chung-Ju Wu <jasonwucj@gmail.com>
15348 * config/nds32/constants.md (unspec_element): New enum.
15349 * config/nds32/constraints.md (Umw): New constraint.
15350 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15351 * config/nds32/nds32-intrinsic.md: Likewise.
15352 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15353 (nds32_valid_smw_lwm_base_p): New.
15354 (nds32_output_smw_single_word): New.
15355 (nds32_output_lmw_single_word): New.
15356 (nds32_expand_unaligned_load): New.
15357 (nds32_expand_unaligned_store): New.
15358 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15359 (nds32_output_smw_single_word): Declare.
15360 (nds32_output_lmw_single_word): Declare.
15361 (nds32_expand_unaligned_load): Declare.
15362 (nds32_expand_unaligned_store): Declare.
15363 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15364 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15365 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15366 NDS32_BUILTIN_UASTORE_DW.
15367 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15370 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15371 Kito Cheng <kito.cheng@gmail.com>
15372 Chung-Ju Wu <jasonwucj@gmail.com>
15374 * config/nds32/nds32-intrinsic.c
15375 (nds32_expand_builtin_null_ftype_reg): Delete.
15376 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15377 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15378 (nds32_read_argument): New.
15379 (nds32_legitimize_target): Ditto.
15380 (nds32_legitimize_argument): Ditto.
15381 (nds32_check_constant_argument): Ditto.
15382 (nds32_expand_unop_builtin): Ditto.
15383 (nds32_expand_unopimm_builtin): Ditto.
15384 (nds32_expand_binop_builtin): Ditto.
15385 (nds32_builtin_decl_impl): Ditto.
15386 (builtin_description): Ditto.
15387 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15388 (nds32_init_builtins_impl): Ditto.
15389 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15390 (nds32_builtin_decl): New.
15391 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15392 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15394 2018-03-02 Jeff Law <law@redhat.com>
15396 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15397 (redundant_insn, fill_simple_delay_slots): Likewise.
15398 (fill_slots_from_thread): Likewise.
15399 * resource.c (mark_referenced_resources): Likewise.
15400 (mark_set_resources, find_dead_or_set_registers): Likewise.
15402 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15404 * substring-locations.h (format_warning_va): Formatting fix for
15405 ATTRIBUTE_GCC_DIAG.
15406 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15408 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15409 * substring-locations.c: Include intl.h.
15410 (format_warning_va): Turned into small wrapper around
15411 format_warning_n_va, renamed to ...
15412 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15413 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15415 (format_warning_at_substring_n): New function.
15416 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15417 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15418 format_warning_at_substring with just a shorter name instead of
15419 const function pointer.
15420 (fmtwarn_n): New function.
15421 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15422 appropriate, get rid of all the fmtstr temporaries, move conditionals
15423 with G_() wrapped string literals directly into fmtwarn arguments,
15424 cast dir.len to (int), formatting fixes.
15426 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15428 * doc/invoke.texi: Remove "Cilk Plus" references.
15430 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15431 Richard Biener <rguenther@suse.de>
15434 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15435 for error or warning attributes if CALL_FROM_THUNK_P is set.
15438 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15441 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15442 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15445 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15446 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15448 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15449 instead of -1U in last predictors element's probability member.
15451 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15454 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15455 arguments if they are comparable.
15457 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15459 PR tree-optimization/84634
15460 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15461 masks and masked_loop_p with a single loop_masks, making sure it's
15462 null for bb vectorization.
15464 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15466 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15467 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15468 loop->force_vectorize to check whether there is no alias.
15470 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15473 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15475 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15477 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15478 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15479 instead of a loop around prev_real_insn.
15480 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15483 PR inline-asm/84625
15484 * config/i386/i386.c (ix86_print_operand): Use conditional
15485 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15488 2018-03-02 Richard Biener <rguenther@suse.de>
15490 PR tree-optimization/84427
15491 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15492 (bitmap_set_subtract_values): Rewrite to handle multiple
15495 (prune_clobbered_mems): Likewise.
15496 (phi_translate): Take edge instead of pred/phiblock.
15497 (phi_translate_1): Likewise.
15498 (phi_translate_set): Likewise. Insert all translated
15499 exprs for a value into the set, keeping possibly multiple
15500 expressions per value.
15501 (compute_antic_aux): Adjust for phi_translate changes.
15502 When intersecting union the expressions and prune those
15503 not in the final value set, keeping possibly multiple
15504 expressions per value. Do not use value-insertion
15505 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15506 all expressions. Add verification that the value-sets
15507 only shrink during iteration.
15508 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15509 (do_pre_regular_insertion): Likewise.
15510 (do_pre_partial_partial_insertion): Likewise.
15512 2018-03-02 Richard Biener <rguenther@suse.de>
15515 * config/darwin.c (saved_debug_info_level): New static global.
15516 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15517 (darwin_asm_lto_end): Restore debug info generation settings.
15519 2018-03-01 Martin Liska <mliska@suse.cz>
15522 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15523 volatile arguments.
15525 2018-03-01 Richard Biener <rguenther@suse.de>
15528 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15529 type annotation with LTO when debug was disabled at compile-time.
15531 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15533 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15535 (mips_final_postscan_insn): Likewise.
15537 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15539 PR rtl-optimization/84528
15540 * alias.c (init_alias_target): Add commentary.
15541 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15542 a unique base value if the frame pointer is not eliminated
15543 to the stack pointer.
15545 2018-03-01 Tom de Vries <tom@codesourcery.com>
15547 PR rtl-optimization/83327
15548 * lra-int.h (hard_regs_spilled_into): Declare.
15549 * lra.c (hard_regs_spilled_into): Define.
15550 (init_reg_info): Init hard_regs_spilled_into.
15551 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15552 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15553 (process_bb_lives): Handle hard_regs_spilled_into.
15554 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15555 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15557 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15559 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15560 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15561 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15562 * config/rs6000/aix72.h: New file.
15564 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15566 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15567 instead of warning_at with conditional singular and plural messages
15571 * stor-layout.c (update_alignment_for_field): For
15572 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15573 && !DECL_PACKED (field), do the alignment update, just use
15574 only desired_align instead of MAX (type_align, desired_align)
15576 (place_field): Don't do known_align < desired_align handling
15577 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15578 is non-NULL, instead do it after rli->prev_field handling and
15579 only if not within a bitfield word. For DECL_PACKED (field)
15580 use type_align of BITS_PER_UNIT.
15582 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15584 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15585 superfluous parentheses and trailing spaces.
15587 2018-02-28 Richard Biener <rguenther@suse.de>
15589 PR tree-optimization/84584
15590 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15591 SCoPs with fake exit edge.
15593 2018-02-28 Martin Liska <mliska@suse.cz>
15596 * timevar.c (timer::print): Fix format to properly print 100%
15599 2018-02-28 Richard Biener <rguenther@suse.de>
15601 PR middle-end/84607
15602 * genmatch.c (capture_info::walk_match): Do not mark
15603 captured expressions without operands as expr_p given
15604 they act more like predicates and should be subject to
15605 "lost tail" side-effect preserving.
15607 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15609 PR rtl-optimization/81611
15610 * auto-inc-dec.c (attempt_change): Move dead note from
15611 mem_insn if it's the next use of regno
15612 (find_address): Take address use of reg holding
15613 non-incremented value. Add parm to limit search to the named
15615 (merge_in_block): Attempt to use a mem insn that is the next
15616 use of the original regno.
15618 2018-02-27 Martin Sebor <msebor@redhat.com>
15621 * doc/invoke.texi (-Wmissing-attributes): New option.
15622 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15624 2018-02-27 Martin Sebor <msebor@redhat.com>
15626 PR translation/84207
15627 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15628 n argument to unsigned HOST_WIDE_INT.
15629 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15630 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15631 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15632 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15634 2018-02-27 Richard Biener <rguenther@suse.de>
15636 PR tree-optimization/84512
15637 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15638 Do not use the estimate returned from record_stmt_cost for
15639 the scalar iteration cost but sum properly using add_stmt_cost.
15641 2018-02-27 Richard Biener <rguenther@suse.de>
15643 PR tree-optimization/84466
15644 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15645 Adjust last change to less strictly validate use operands.
15647 2018-02-27 Martin Liska <mliska@suse.cz>
15649 PR gcov-profile/84548
15650 * gcov.c (process_file): Allow partial overlap and consider it
15651 also as group functions.
15652 (output_lines): Properly calculate range of lines for a group.
15654 2018-02-27 Martin Liska <mliska@suse.cz>
15656 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15657 'ggc' suffixes. Change first column width.
15658 (timer::print): Fix formatting of the column.
15660 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15662 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15663 preserve inline entry blocks for the sake of debug inline
15664 entry point markers alone.
15665 (remove_unused_locals): Suggest in comments a better place to
15666 force the preservation of inline entry blocks that are
15667 otherwise unused, but do not preserve them.
15669 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15671 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15673 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15676 * config/i386/constraints.md (Bs): Replace
15677 ix86_indirect_branch_register with
15678 TARGET_INDIRECT_BRANCH_REGISTER.
15680 * config/i386/i386.md (indirect_jump): Likewise.
15681 (tablejump): Likewise.
15682 (*sibcall_memory): Likewise.
15683 (*sibcall_value_memory): Likewise.
15684 Peepholes of indirect call and jump via memory: Likewise.
15685 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15686 (*sibcall_value_GOT_32): Likewise.
15687 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15688 (GOT_memory_operand): Likewise.
15689 (call_insn_operand): Likewise.
15690 (sibcall_insn_operand): Likewise.
15691 (GOT32_symbol_operand): Likewise.
15692 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15694 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15696 PR rtl-optimization/83496
15697 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15698 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15699 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15700 redundant insn, if any.
15701 (relax_delay_slots): Likewise.
15702 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15704 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15706 PR tree-optimization/83965
15707 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15708 that grouped statements are part of a reduction chain. Return
15709 true if the statement is not marked as a reduction itself but
15710 is part of a group.
15711 (vect_recog_dot_prod_pattern): Don't check whether the statement
15712 is part of a group here.
15713 (vect_recog_sad_pattern): Likewise.
15714 (vect_recog_widen_sum_pattern): Likewise.
15716 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15719 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15721 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15724 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15726 (ix86_output_indirect_function_return): New prototype.
15727 (ix86_split_simple_return_pop_internal): Likewise.
15728 * config/i386/i386.c (indirect_return_via_cx): New.
15729 (indirect_return_via_cx_bnd): Likewise.
15730 (indirect_thunk_name): Handle return va CX_REG.
15731 (output_indirect_thunk_function): Create alias for
15732 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15733 (ix86_output_indirect_jmp): Remove the bool argument.
15734 (ix86_output_indirect_function_return): New function.
15735 (ix86_split_simple_return_pop_internal): Likewise.
15736 * config/i386/i386.md (*indirect_jump): Don't pass false
15737 to ix86_output_indirect_jmp.
15738 (*tablejump_1): Likewise.
15739 (simple_return_pop_internal): Change it to define_insn_and_split.
15740 Call ix86_split_simple_return_pop_internal to split it for
15741 -mfunction-return=.
15742 (simple_return_indirect_internal): Call
15743 ix86_output_indirect_function_return instead of
15744 ix86_output_indirect_jmp.
15746 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15749 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15750 memset and value initialization afterwards.
15752 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15754 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15756 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15759 * common/config/aarch64/aarch64-common.c
15760 (aarch_option_optimization_table[]): Switch
15761 off fomit-frame-pointer
15763 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15764 Chung-Ju Wu <jasonwucj@gmail.com>
15766 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15768 (store_multiple): Ditto.
15770 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15772 * config.gcc: Add --with-cpu support for nds32 target.
15773 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15774 * config/nds32/nds32.opt: Add -mcpu= option.
15776 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15778 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15779 isel=yes): Warn for these deprecated options.
15781 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15783 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15784 ISA_2_5_MASKS_EMBEDDED.
15786 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15788 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15789 p->max as pointers rather than using iterative_hash_expr.
15791 2018-02-23 Carl Love <cel@us.ibm.com>
15793 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15794 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15796 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15797 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15798 P8V_BUILTIN_VEC_VUNSIGNED2.
15800 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15803 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15804 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15806 * lra-constraints.c (curr_insn_transform): Set up
15807 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15809 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15810 * lra-eliminations.c (spill_pseudos): Ditto.
15811 (process_insn_for_elimination): Ditto.
15812 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15813 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15814 LRA_NON_CLOBBERED_ALT.
15816 2018-02-22 Martin Sebor <msebor@redhat.com>
15818 PR tree-optimization/84480
15819 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15820 to maybe_diag_stxncpy_trunc. Call it.
15821 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15822 from gimple_fold_builtin_strcpy. Print inlining stack.
15823 (handle_builtin_stxncpy): Print inlining stack.
15824 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15826 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15829 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15830 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15831 and -fcheck-pointer-bounds are used together.
15832 (indirect_thunk_prefix): New enum.
15833 (indirect_thunk_need_prefix): New function.
15834 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15835 "_nt" instead of "_bnd" for NOTRACK prefix.
15836 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15837 (output_indirect_thunk_function): Likewise.
15839 (ix86_code_end): Update output_indirect_thunk_function calls.
15840 (ix86_output_indirect_branch_via_reg): Replace
15841 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15842 (ix86_output_indirect_branch_via_push): Likewise.
15843 (ix86_output_function_return): Likewise.
15844 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15845 incompatible with -fcf-protection=branch and
15846 -fcheck-pointer-bounds.
15848 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15851 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15852 Change gcc_assert call to output_operand_lossage.
15854 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15856 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15858 2018-02-22 DJ Delorie <dj@redhat.com>
15859 Sebastian Perta <sebastian.perta@renesas.com>
15860 Oleg Endo <olegendo@gcc.gnu.org>
15862 * config/rx/rx.c (rx_rtx_costs): New function.
15863 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15865 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15867 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15869 2018-02-22 Martin Liska <mliska@suse.cz>
15872 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15873 Add "native" as a possible value.
15875 2018-02-22 Martin Liska <mliska@suse.cz>
15878 * config/i386/i386.c (ix86_option_override_internal):
15879 Add "native" as a possible value for -march and -mtune.
15881 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15884 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15885 to all type variants.
15887 PR tree-optimization/84503
15888 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15889 width as info->bitpos + info->bitsize - start.
15890 (merged_store_group::merge_overlapping): Simplify width computation.
15891 (check_no_overlap): New function.
15892 (imm_store_chain_info::try_coalesce_bswap): Compute expected
15893 start + width and last_order of the group, fail if check_no_overlap
15895 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15896 to group if check_no_overlap fails.
15898 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15900 * config/rs6000/altivec.md: Delete contraint arguments to
15901 define_expand, define_split, and define_peephole2, and in
15902 define_insn_and_split if always unused.
15903 * config/rs6000/darwin.md: Ditto.
15904 * config/rs6000/dfp.md: Ditto.
15905 * config/rs6000/rs6000.md: Ditto.
15906 * config/rs6000/sync.md: Ditto.
15907 * config/rs6000/vector.md: Ditto.
15908 * config/rs6000/vsx.md: Ditto.
15910 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15912 * config/rs6000/altivec.md: Write output control strings as braced
15913 blocks instead of double-quoted strings.
15914 * config/rs6000/darwin.md: Ditto.
15915 * config/rs6000/rs6000.md: Ditto.
15916 * config/rs6000/vector.md: Ditto.
15917 * config/rs6000/vsx.md: Ditto.
15919 2018-02-21 Jason Merrill <jason@redhat.com>
15921 PR c++/84314 - ICE with templates and fastcall attribute.
15922 * attribs.c (build_type_attribute_qual_variant): Remove assert.
15924 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15926 * ipa-cp.c (determine_versionability): Fix comment typos.
15928 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15931 * ipa-cp.c (determine_versionability): Do not version functions caling
15934 2018-02-21 Martin Liska <mliska@suse.cz>
15937 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15938 Add "native" as a possible value.
15939 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
15940 the macro when native cpu detection is available.
15942 2018-02-21 Martin Liska <mliska@suse.cz>
15945 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15946 Add "native" as a possible value.
15947 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15948 when native cpu detection is available.
15950 2018-02-21 Jakub Jelinek <jakub@redhat.com>
15951 Martin Sebor <msebor@redhat.com>
15953 PR tree-optimization/84478
15954 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15956 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15957 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15958 support which is conservatively correct, for 2 only stay conservative
15959 for maxlen. Formatting and comment capitalization fixes. Add STRICT
15960 argument to the 2 argument get_range_strlen, adjust 6 arg
15961 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15963 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15964 (gimple_fold_builtin_strlen): Pass true as last argument to
15967 2018-02-20 Martin Sebor <msebor@redhat.com>
15969 PR middle-end/84095
15970 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15971 (builtin_memref::set_base_and_offset): Same. Handle inner references.
15972 (builtin_memref::builtin_memref): Factor out parts into
15973 set_base_and_offset and call it.
15975 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15977 PR middle-end/84406
15978 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15979 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15980 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
15981 search at the associated MODE_INT.
15983 2018-02-20 Jeff Law <law@redhat.com>
15985 PR middle-end/82123
15986 PR tree-optimization/81592
15987 PR middle-end/79257
15988 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15989 for range data rather than using global data.
15990 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15991 range data rather than using global data.
15992 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15993 pass it to children as needed.
15994 (struct directive::fmtresult): Similarly.
15995 (struct directive::set_width): Similarly.
15996 (struct directive::set_precision): Similarly.
15997 (format_integer, format_directive, parse_directive): Similarly.
15998 (format_none): Accept unnamed vr_values parameter.
15999 (format_percent, format_floating, format_character): Similarly.
16000 (format_string, format_plain): Similarly.
16001 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16002 the EVRP range analyzer for range data rather than using global data.
16003 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16004 gimple-ssa-evrp-analyze.h
16005 (class sprintf_dom_walker): Add after_dom_children member function.
16006 Add evrp_range_analyzer member.
16007 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16008 range analyzer as needed.
16009 (sprintf_dom_walker::after_dom_children): New member function.
16010 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16012 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16013 (evrp_range_analyzer::pop_to_marker): Likewise.
16015 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16017 PR tree-optimization/84419
16018 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16019 with the required type if its current type is compatible but
16022 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16024 PR middle-end/82004
16025 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16026 after vectorization.
16028 2018-02-20 Martin Liska <mliska@suse.cz>
16031 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16032 possible values if we don't have a hint.
16034 2018-02-20 Martin Liska <mliska@suse.cz>
16038 * final.c (shorten_branches): Build align_tab array with one
16040 * opts.c (finish_options): Add alignment option limit check.
16041 (MAX_CODE_ALIGN): Likewise.
16042 (MAX_CODE_ALIGN_VALUE): Likewise.
16043 * doc/invoke.texi: Document maximum allowed option value for
16044 all -falign-* options.
16046 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16049 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16050 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16051 * var-tracking.c (emit_note_insn_var_location): Remove all references
16052 to NOTE_INSN_CALL_ARG_LOCATION.
16053 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16054 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16055 Use copy_rtx_if_shared.
16056 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16057 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16058 (dwarf2out_var_location): Remove handling of
16059 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16061 * final.c (final_scan_insn): Remove all references to
16062 NOTE_INSN_CALL_ARG_LOCATION.
16063 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16064 before dumping final insns.
16065 * except.c (emit_note_eh_region_end): Remove all references to
16066 NOTE_INSN_CALL_ARG_LOCATION.
16067 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16068 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16069 * config/arc/arc.c (hwloop_optimize): Likewise.
16070 * config/arm/arm.c (create_fix_barrier): Likewise.
16071 * config/s390/s390.c (s390_chunkify_start): Likewise.
16072 * config/sh/sh.c (find_barrier): Likewise.
16073 * config/i386/i386.c (rest_of_insert_endbranch,
16074 ix86_seh_fixup_eh_fallthru): Likewise.
16075 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16076 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16077 * config/frv/frv.c (frv_function_prologue): Likewise.
16078 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16080 (note_outside_basic_block_p): Remove all references to
16081 NOTE_INSN_CALL_ARG_LOCATION.
16082 * gengtype.c (adjust_field_rtx_def): Likewise.
16083 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16085 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16086 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16089 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16092 PR tree-optimization/84452
16093 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16094 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16097 2018-02-19 Martin Liska <mliska@suse.cz>
16100 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16102 2018-02-19 Martin Liska <mliska@suse.cz>
16103 Richard Sandiford <richard.sandiford@linaro.org>
16105 PR tree-optimization/82491
16106 * gimple-fold.c (get_base_constructor): Make earlier bail out
16109 2018-02-19 Carl Love <cel@us.ibm.com>
16111 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16112 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16114 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16115 P8V_BUILTIN_VEC_NEG.
16117 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16119 * config/rl78/rl78.md (movdf): New define expand.
16121 2018-02-19 Martin Liska <mliska@suse.cz>
16124 * doc/invoke.texi: Fix typo.
16125 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16127 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16129 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16130 handle rs6000_single_float and rs6000_double_float specially for
16133 2018-02-16 Jeff Law <law@redhat.com>
16135 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16136 (pop_regs): Use it.
16138 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16141 * ipa-inline.c (inline_small_functions): Fix a typo.
16143 2018-02-16 Nathan Sidwell <nathan@acm.org>
16145 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16147 2018-02-16 Carl Love <cel@us.ibm.com>
16149 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16150 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16151 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16152 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16153 expansion to P8V_BUILTIN_VEC_FLOAT2.
16155 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16157 PR rtl-optimization/70023
16158 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16159 src_regno into account.
16161 2018-02-16 Carl Love <cel@us.ibm.com>
16163 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16164 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16165 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16166 * config/rs6000/rs6000.c: Remove case statements for
16167 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16168 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16169 and P9V_BUILTIN_VEC_VINSERT4B.
16170 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16171 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16172 * config/rs6000/vsx.md:
16173 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16176 2018-02-16 Carl Love <cel@us.ibm.com>
16178 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16180 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16182 * config/rs6000/rs6000-c.c: Add the definitions for
16183 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16184 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16185 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16186 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16187 definition for insert4b and define insn *insert3b_internal.
16188 * doc/extend.texi: Add documentation for vec_extract4b.
16190 2018-02-16 Nathan Sidwell <nathan@acm.org>
16192 * doc/extend.texi (Backwards Compatibility): Mention friend
16193 injection. Note for-scope is deprecated.
16194 * doc/invoke.texi (-ffriend-injection): Deprecate.
16196 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16198 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16199 that moved to I2, also allow destinations that are a paradoxical
16200 subreg (instead of a normal reg).
16202 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16205 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16208 2018-02-16 Richard Biener <rguenther@suse.de>
16210 PR tree-optimization/84037
16211 PR tree-optimization/84016
16213 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16214 Adjust vec_construct for the fact we need additional higher latency
16215 128bit inserts for AVX256 and AVX512 vector builds.
16216 (ix86_add_stmt_cost): Scale vector construction cost for
16219 2018-02-16 Richard Biener <rguenther@suse.de>
16221 PR tree-optimization/84417
16222 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16223 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16224 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16226 2018-02-16 Martin Liska <mliska@suse.cz>
16229 * internal-fn.def (ASAN_CHECK): Set proper flags.
16230 (ASAN_MARK): Likewise.
16232 2018-02-16 Julia Koval <julia.koval@intel.com>
16234 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16235 from PTA_CANNONLAKE.
16237 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16240 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16241 Use ++iter rather than iter++ for std::list iterators.
16242 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16243 defer deleting them until all nodes in the forest are processed. Do
16244 free even leaf nodes. Change to_process into auto_vec.
16247 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16248 * vec.h (vec_default_construct): Use memset instead of placement new
16249 if BROKEN_VALUE_INITIALIZATION is defined.
16250 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16251 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16254 PR rtl-optimization/83723
16255 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16256 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16257 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16259 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16261 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16263 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16265 PR rtl-optimization/81443
16266 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16267 from inner REGs to paradoxical SUBREGs.
16269 2018-02-16 Richard Biener <rguenther@suse.de>
16271 PR tree-optimization/84399
16272 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16273 For operands we can analyze at their definition make sure we can
16274 analyze them at each use as well.
16276 2018-02-16 Richard Biener <rguenther@suse.de>
16278 PR tree-optimization/84190
16279 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16280 volatile accesses if the decl isn't volatile.
16282 2018-02-15 Jason Merrill <jason@redhat.com>
16284 PR c++/84314 - ICE with templates and fastcall attribute.
16285 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16286 TYPE_CANONICAL on an existing type.
16288 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16290 PR tree-optimization/84383
16291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16292 dstoff nor call operand_equal_p if dstbase is NULL.
16294 PR tree-optimization/84334
16295 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16296 also a CONSTANT_CLASS_P, punt.
16298 2018-02-14 Jim Wilson <jimw@sifive.com>
16300 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16301 first SMALL_OPERAND check. New local min_second_step. Move assert
16302 to where locals are set. Add TARGET_RVC support.
16303 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16305 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16307 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16309 2018-02-14 Martin Sebor <msebor@redhat.com>
16311 PR tree-optimization/83698
16312 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16313 arrays constrain the offset range to their bounds.
16314 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16315 (builtin_access::overlap): Avoid setting the size of overlap if it's
16317 (maybe_diag_overlap): Also consider arrays when deciding what values
16318 of offsets to include in diagnostics.
16320 2018-02-14 Martin Sebor <msebor@redhat.com>
16323 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16324 that correspond to the kind of a declaration.
16326 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16329 * config/pa/pa.md: Load address of PIC label using the linkage table
16330 if the label is nonlocal.
16332 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16334 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16335 warning message if user requests -maltivec=be.
16336 * doc/invoke.texi: Document deprecation of -maltivec=be.
16338 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16341 * config/rs6000/rs6000-c.c: Update definitions for
16342 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16343 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16345 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16348 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16349 add _get_ssp intrinsics. Remove argument from
16350 __builtin_ia32_rdssp[d|q].
16351 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16352 * config/i386/i386-builtin.def: Remove argument from
16353 __builtin_ia32_rdssp[d|q].
16354 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16355 ix86_expand_special_args_builtin for _rdssp[d|q].
16356 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16357 Clear register before usage.
16358 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16359 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16361 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16363 PR tree-optimization/84357
16364 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16365 operand 1 of an ARRAY_REF too.
16367 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16370 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16371 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16373 (set_of_reg): New struct.
16374 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16375 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16376 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16378 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16379 Split into bitclr, bitset, bitinvert patterns if appropriate.
16380 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16381 use rx_fuse_in_memory_bitop.
16382 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16383 to named insn, correct maximum insn length.
16385 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16388 * machmode.def: Define a complex mode for PARTIAL_INT.
16389 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16391 * doc/rtl.texi: Document CSPImode.
16392 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16394 (msp430_hard_regno_nregs_with_padding): Likewise.
16396 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16399 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16401 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16403 PR rtl-optimization/84169
16404 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16405 we generated a parallel as new i3 and we split that to new i2 and i3
16406 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16407 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16408 those to i2, not i1. Partially rewrite this scan code.
16410 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16413 * stor-layout.c (place_field): For variable length fields, adjust
16414 offset_align afterwards not just based on the field's alignment,
16415 but also on the size.
16417 PR middle-end/84309
16418 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16419 of exps and logs in the use_exp2 case.
16421 2018-02-13 Jeff Law <law@redhat.com>
16423 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16424 entry for "vector".
16426 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16429 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16433 * common.opt (gas-loc-support, gas-locview-support): New.
16434 (ginline-points, ginternal-reset-location-views): New.
16435 * doc/invoke.texi: Document them. Use @itemx where intended.
16436 (gvariable-location-views): Adjust.
16437 * target.def (reset_location_view): New.
16438 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16439 (TARGET_RESET_LOCATION_VIEW): New.
16440 * doc/tm.texi: Rebuilt.
16441 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16442 (dwarf2out_default_as_locview_support): New.
16443 (output_asm_line_debug_info): Use option variables.
16444 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16445 (output_loc_list): Likewise.
16446 (add_high_low_attributes): Check option variables.
16447 Don't output entry view attribute in strict mode.
16448 (gen_inlined_subroutine_die): Check option variables.
16449 (dwarf2out_inline_entry): Likewise.
16450 (init_sections_and_labels): Likewise.
16451 (dwarf2out_early_finish): Likewise.
16452 (maybe_reset_location_view): New, from...
16453 (dwarf2out_var_location): ... here. Call it.
16454 * debug.h (dwarf2out_default_as_loc_support): Declare.
16455 (dwarf2out_default_as_locview_support): Declare.
16456 * hooks.c (hook_int_rtx_insn_0): New.
16457 * hooks.h (hook_int_rtx_insn_0): Declare.
16458 * toplev.c (process_options): Take -gas-loc-support and
16459 -gas-locview-support from dwarf2out. Enable
16460 -gvariable-location-views by default only with locview
16461 assembler support. Enable -ginternal-reset-location-views by
16462 default only if the target defines the corresponding hook.
16463 Enable -ginline-points by default if location views are
16464 enabled; force it disabled if statement frontiers are
16466 * tree-inline.c (expand_call_inline): Check option variables.
16467 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16469 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16471 PR tree-optimization/84321
16472 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16473 handling. Also check whether the anti-range contains any values
16474 that satisfy the mask; switch to a VR_RANGE if not.
16476 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16479 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16481 2018-02-13 Martin Jambor <mjambor@suse.cz>
16484 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16485 of call statements, also set location of a load to a temporary.
16487 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16489 * config/rl78/rl78.c (add_vector_labels): New function.
16490 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16491 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16492 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16493 which checks that no arguments are passed.
16494 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16495 * doc/extend.texi: Documentation for the new attribute.
16497 2018-02-13 Andreas Schwab <schwab@suse.de>
16499 * config/riscv/linux.h (CPP_SPEC): Define.
16501 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16504 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16505 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16506 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16507 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16508 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16509 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16510 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16511 temporarily for AES and PCLMUL builtins.
16513 PR tree-optimization/84339
16514 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16515 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16518 PR middle-end/84309
16519 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16520 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16521 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16523 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16525 * omp-simd-clone.h: New file.
16526 * omp-simd-clone.c: Include omp-simd-clone.h.
16527 (expand_simd_clones): No longer static.
16528 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16529 cgraph.h and omp-simd-clone.h.
16530 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16531 (vect_recog_widen_shift_pattern): Formatting fix.
16532 (vect_pattern_recog_1): Don't check optab for calls.
16535 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16536 operands[2] into a REG before using gen_lowpart on it.
16538 2018-02-12 Jeff Law <law@redhat.com>
16541 * config/sh/sh.c (find_barrier): Consider a sibling call
16544 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16545 successfully back substituting a reg.
16547 2018-02-12 Richard Biener <rguenther@suse.de>
16549 PR tree-optimization/84037
16550 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16551 parameter, move visited init to caller.
16552 (vect_slp_analyze_operations): Separate cost from validity
16553 check, initialize visited once for all instances.
16554 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16556 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16558 (vect_model_promotion_demotion_cost): Likewise.
16559 (vectorizable_bswap): Guard cost modeling with !slp_node
16560 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16562 (vectorizable_call): Likewise.
16563 (vectorizable_conversion): Likewise.
16564 (vectorizable_assignment): Likewise.
16565 (vectorizable_shift): Likewise.
16566 (vectorizable_operation): Likewise.
16567 (vectorizable_store): Likewise.
16568 (vectorizable_load): Likewise.
16569 (vectorizable_condition): Likewise.
16570 (vectorizable_comparison): Likewise.
16572 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16575 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16576 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16577 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16579 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16581 PR middle-end/83665
16582 * params.def (inline-min-speedup): Increase from 8 to 15.
16583 (max-inline-insns-auto): Decrease from 40 to 30.
16584 * ipa-split.c (consider_split): Add some buffer for function to
16585 be considered inlining candidate.
16586 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16589 2018-02-12 Richard Biener <rguenther@suse.de>
16591 PR tree-optimization/84037
16592 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16593 matched stmts if we cannot swap the non-matched ones.
16595 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16597 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16598 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16599 _mm_maskz_scalef_round_ss): New intrinsics.
16600 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16601 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16602 __builtin_ia32_scalefss_round): Remove.
16603 (__builtin_ia32_scalefsd_mask_round,
16604 __builtin_ia32_scalefss_mask_round): New intrinsics.
16605 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16606 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16607 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16608 "<round_constraint>")): Changed to ...
16609 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16610 "<round_scalar_constraint>")): ... this.
16611 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16612 %0, %1, %2<round_op3>}"): Changed to ...
16613 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16614 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16615 %2<round_scalar_mask_op3>}"): ... this.
16616 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16618 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16620 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16621 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16622 (_mm_maskz_sqrt_round_ss): New intrinsics.
16623 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16624 (__builtin_ia32_sqrtsd_mask_round)
16625 (__builtin_ia32_sqrtss_mask_round): New builtins.
16626 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16627 (__builtin_ia32_sqrtss_round): Remove.
16628 (__builtin_ia32_sqrtsd_mask_round)
16629 (__builtin_ia32_sqrtss_mask_round): New builtins.
16630 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16631 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16632 ((match_operand:VF_128 1 "vector_operand"
16633 "xBm,<round_constraint>")): Changed to ...
16634 ((match_operand:VF_128 1 "vector_operand"
16635 "xBm,<round_scalar_constraint>")): ... this.
16636 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16637 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16638 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16639 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16640 %<iptr>1<round_scalar_mask_op3>}): ... this.
16641 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16642 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16644 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16647 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16648 Cast vec_cmpeq result to correct type.
16649 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16650 Cast vec_cmpgt result to correct type.
16652 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16654 * final.c (final_scan_insn_1): Renamed from...
16655 (final_scan_insn): ... this. New wrapper, to recover
16656 seen from the outermost call in recursive ones.
16657 * config/sparc/sparc.c (output_return): Drop seen from call.
16658 (output_sibcall): Likewise.
16659 * config/visium/visium.c (output_branch): Likewise.
16661 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16663 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16666 2018-02-10 Alan Modra <amodra@gmail.com>
16669 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16670 Specify LR as an input.
16672 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16675 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16676 remove_member_access_dummy_vars): New functions.
16677 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16678 lower_omp_1, execute_lower_omp): Use them.
16680 PR rtl-optimization/84308
16681 * shrink-wrap.c (spread_components): Release todo vector.
16683 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16685 PR rtl-optimization/57193
16686 * ira-color.c (struct allocno_color_data): Add member
16687 conflict_allocno_hard_prefs.
16688 (update_conflict_allocno_hard_prefs): New.
16689 (bucket_allocno_compare_func): Add a preference based on
16690 conflict_allocno_hard_prefs.
16691 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16692 (color_allocnos): Remove a dead code. Initiate
16693 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16695 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16698 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16699 constraint from =wa to wa. Avoid a subreg on the output operand,
16700 instead use a pseudo and subreg it in a move.
16701 (p9_xxbrd_<mode>): Changed to ...
16702 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16703 (p9_xxbrd_v2df): New expander.
16704 (p9_xxbrw_<mode>): Changed to ...
16705 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16706 (p9_xxbrw_v4sf): New expander.
16708 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16710 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16712 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16715 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16716 multiply in 32-bit mode.
16717 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16718 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16721 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16723 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16724 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16725 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16726 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16728 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16731 * dwarf2out.c (is_trivial_indirect_ref): New function.
16732 (dwarf2out_late_global_decl): Do not generate a location
16733 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16734 and that are not defined in the current unit.
16736 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16738 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16739 instead of a libcall for UNORDERED.
16741 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16744 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16745 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16747 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16750 * config/s390/s390.c (s390_set_current_function): Invoke
16751 s390_indirect_branch_settings also if fndecl didn't change.
16753 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16755 * config/rs6000/rs6000.md (blockage): Set length to zero.
16757 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16759 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16761 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16764 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16765 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16769 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16770 PARALLEL incoming that failed vt_get_decl_and_offset check.
16772 PR middle-end/84237
16773 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16774 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16776 (get_variable_section): For decls in named .bss* sections pass true as
16777 second argument to bss_initializer_p.
16779 2018-02-09 Marek Polacek <polacek@redhat.com>
16780 Jakub Jelinek <jakub@redhat.com>
16783 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16784 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16785 Sync some changes from cxx_fold_indirect_ref.
16787 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16789 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16791 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16792 (BLOCK_INLINE_ENTRY_LABEL): New.
16793 (dwarf2out_var_location): Disregard inline entry markers.
16794 (inline_entry_data): New struct.
16795 (inline_entry_data_hasher): New hashtable type.
16796 (inline_entry_data_hasher::hash): New.
16797 (inline_entry_data_hasher::equal): New.
16798 (inline_entry_data_table): New variable.
16799 (add_high_low_attributes): Add DW_AT_entry_pc and
16800 DW_AT_GNU_entry_view attributes if a pending entry is found
16801 in inline_entry_data_table. Add old entry_pc attribute only
16802 if debug nonbinding markers are disabled.
16803 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16804 markers are enabled.
16805 (block_within_block_p, dwarf2out_inline_entry): New.
16806 (dwarf2out_finish): Check that no entries remained in
16807 inline_entry_data_table.
16808 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16809 (final_scan_insn, notice_source_line): Likewise.
16810 (rest_of_clean_state): Skip inline entry markers.
16811 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16813 * gimple.c (gimple_build_debug_inline_entry): New.
16814 * gimple.h (enum gimple_debug_subcode): Add
16815 GIMPLE_DEBUG_INLINE_ENTRY.
16816 (gimple_build_debug_inline_entry): Declare.
16817 (gimple_debug_inline_entry_p): New.
16818 (gimple_debug_nonbind_marker_p): Adjust.
16819 * insn-notes.def (INLINE_ENTRY): New.
16820 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16821 inline entry marker notes.
16822 (print_insn): Likewise.
16823 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16824 (INSN_DEBUG_MARKER_KIND): Likewise.
16825 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16826 * tree-inline.c (expand_call_inline): Build and insert
16827 debug_inline_entry stmt.
16828 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16829 inline entry blocks early, if nonbind markers are enabled.
16830 (dump_scope_block): Dump fragment info.
16831 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16832 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16833 (gimple_build_debug_inline_entry): New.
16834 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16835 Enable/disable inline entry points too.
16836 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16837 (DEBUG_INSN): Describe inline entry markers.
16839 * common.opt (gvariable-location-views): New.
16840 (gvariable-location-views=incompat5): New.
16841 * config.in: Rebuilt.
16842 * configure: Rebuilt.
16843 * configure.ac: Test assembler for view support.
16844 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16845 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16846 * dwarf2out.c (var_loc_view): New typedef.
16847 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16848 (dwarf2out_locviews_in_attribute): New.
16849 (dwarf2out_locviews_in_loclist): New.
16850 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16851 (enum dw_line_info_opcode): Add LI_adv_address.
16852 (struct dw_line_info_table): Add view.
16853 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16854 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16855 (zero_view_p): New variable.
16856 (ZERO_VIEW_P): New macro.
16857 (output_asm_line_debug_info): New.
16858 (struct var_loc_node): Add view.
16859 (add_AT_view_list, AT_loc_list): New.
16860 (add_var_loc_to_decl): Add view param. Test it against last.
16861 (new_loc_list): Add view params. Record them.
16862 (AT_loc_list_ptr): Handle loc and view lists.
16863 (view_list_to_loc_list_val_node): New.
16864 (print_dw_val): Handle dw_val_class_view_list.
16865 (size_of_die): Likewise.
16866 (value_format): Likewise.
16867 (loc_list_has_views): New.
16868 (gen_llsym): Set vl_symbol too.
16869 (maybe_gen_llsym, skip_loc_list_entry): New.
16870 (dwarf2out_maybe_output_loclist_view_pair): New.
16871 (output_loc_list): Output view list or entries too.
16872 (output_view_list_offset): New.
16873 (output_die): Handle dw_val_class_view_list.
16874 (output_dwarf_version): New.
16875 (output_compilation_unit_header): Use it.
16876 (output_skeleton_debug_sections): Likewise.
16877 (output_rnglists, output_line_info): Likewise.
16878 (output_pubnames, output_aranges): Update version comments.
16879 (output_one_line_info_table): Output view numbers in asm comments.
16880 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16881 Call maybe_gen_llsym.
16882 (loc_list_from_tree_1): Adjust.
16883 (add_AT_location_description): Create view list attribute if
16884 needed, check it's absent otherwise.
16885 (convert_cfa_to_fb_loc_list): Adjust.
16886 (maybe_emit_file): Call output_asm_line_debug_info for test.
16887 (dwarf2out_var_location): Reset views as needed. Precompute
16888 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
16889 attribute. Set view.
16890 (new_line_info_table): Reset next view.
16891 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16892 (dwarf2out_source_line): Likewise. Output view resets and labels to
16893 the assembler, or select appropriate line info opcodes.
16894 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16895 (optimize_string_length): Catch it. Adjust.
16896 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
16897 dw_val_class_view_list, and remove it if no longer needed.
16898 (hash_loc_list): Hash view numbers.
16899 (loc_list_hasher::equal): Compare them.
16900 (optimize_location_lists): Check whether a view list symbol is
16901 needed, and whether the locview attribute is present, and
16902 whether they match. Remove the locview attribute if no longer
16904 (index_location_lists): Call skip_loc_list_entry for test.
16905 (dwarf2out_finish): Call output_asm_line_debug_info for test.
16906 Use output_dwarf_version.
16907 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16908 (struct dw_val_node): Add val_view_list.
16909 * final.c (SEEN_NEXT_VIEW): New.
16910 (set_next_view_needed): New.
16911 (clear_next_view_needed): New.
16912 (maybe_output_next_view): New.
16913 (final_start_function): Rename to...
16914 (final_start_function_1): ... this. Take pointer to FIRST,
16915 add SEEN parameter. Emit param bindings in the initial view.
16916 (final_start_function): Reintroduce SEEN-less interface.
16917 (final): Rename to...
16918 (final_1): ... this. Take SEEN parameter. Output final pending
16919 next view at the end.
16920 (final): Reintroduce seen-less interface.
16921 (final_scan_insn): Output pending next view before switching
16922 sections or ending a block. Mark the next view as needed when
16923 outputting variable locations. Notify debug backend of section
16924 changes, and of location view changes.
16925 (rest_of_handle_final): Adjust.
16926 * toplev.c (process_options): Autodetect value for debug variable
16927 location views option. Warn on incompat5 without -gdwarf-5.
16928 * doc/invoke.texi (gvariable-location-views): New.
16929 (gvariable-location-views=incompat5): New.
16930 (gno-variable-location-views): New.
16932 2018-02-08 David Malcolm <dmalcolm@redhat.com>
16934 PR tree-optimization/84136
16935 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16936 that the result of find_edge is non-NULL.
16938 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16941 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16942 storing integer register in SImode. Fix cost of 256 and 512
16943 byte aligned SSE register store.
16945 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16947 * config/i386/i386.c (ix86_multiplication_cost): Fix
16948 multiplication cost for TARGET_AVX512DQ.
16950 2018-02-08 Marek Polacek <polacek@redhat.com>
16952 PR tree-optimization/84238
16953 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16956 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16958 PR tree-optimization/84265
16959 * tree-vect-stmts.c (vectorizable_store): Don't treat
16960 VMAT_CONTIGUOUS accesses as grouped.
16961 (vectorizable_load): Likewise.
16963 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16965 PR tree-optimization/81635
16966 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16967 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16968 (test_round_for_mask): New functions.
16969 (wide_int_cc_tests): Call test_round_for_mask.
16970 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16971 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16972 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16973 range returned by get_range_info.
16975 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16978 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16979 * symtab.c: Include builtins.h
16980 (symtab_node::output_to_lto_symbol_table_p): Move here
16981 from lto-streamer-out.c:output_symbol_p.
16982 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16983 (output_symbol_p): Move all logic to symtab.c
16984 (produce_symtab): Update.
16986 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16988 * config/s390/s390-opts.h (enum indirect_branch): Define.
16989 * config/s390/s390-protos.h (s390_return_addr_from_memory)
16990 (s390_indirect_branch_via_thunk)
16991 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16992 (enum s390_indirect_branch_type): Define.
16993 * config/s390/s390.c (struct s390_frame_layout, struct
16994 machine_function): Remove.
16995 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16996 (indirect_branch_table_label_no, indirect_branch_table_name):
16998 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16999 (enum s390_indirect_branch_option): Define.
17000 (s390_return_addr_from_memory): New function.
17001 (s390_handle_string_attribute): New function.
17002 (s390_attribute_table): Add new attribute handler.
17003 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17004 (s390_indirect_branch_via_thunk): New function.
17005 (s390_indirect_branch_via_inline_thunk): New function.
17006 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17007 sibling call optimization for non z10 compiles.
17008 (s390_emit_call): Force indirect branch target to be a single
17009 register. Add r1 clobber for non-z10 compiles.
17010 (s390_emit_epilogue): Emit return jump via return_use expander.
17011 (s390_reorg): Handle JUMP_INSNs as execute targets.
17012 (s390_option_override_internal): Perform validity checks for the
17013 new command line options.
17014 (s390_indirect_branch_attrvalue): New function.
17015 (s390_indirect_branch_settings): New function.
17016 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17017 (s390_output_indirect_thunk_function): New function.
17018 (s390_code_end): Implement target hook.
17019 (s390_case_values_threshold): Implement target hook.
17020 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17022 * config/s390/s390.h (struct s390_frame_layout)
17023 (struct machine_function): Move here from s390.c.
17024 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17025 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17026 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17027 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17028 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17029 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17030 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17031 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17032 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17033 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17034 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17035 (mnemonic attribute): Add values which aren't recognized
17037 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17038 pattern for branch conversion. Fix mnemonic attribute.
17039 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17040 indirect branch via thunk if requested.
17041 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17042 ("*indirect_jump"): Disable for branch conversion using out of
17044 ("indirect_jump_via_thunk<mode>_z10")
17045 ("indirect_jump_via_thunk<mode>")
17046 ("indirect_jump_via_inlinethunk<mode>_z10")
17047 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17048 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17049 ("casesi_jump_via_inlinethunk<mode>_z10")
17050 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17051 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17052 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17053 ("*indirect2_jump"): Disable for branch conversion.
17054 ("casesi_jump"): Turn into expander and expand patterns for branch
17056 ("return_use"): New expander.
17057 ("*return"): Emit return via thunk and rename it to ...
17058 ("*return<mode>"): ... this one.
17059 * config/s390/s390.opt: Add new options and and enum for the
17062 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17064 * lra-constraints.c (match_reload): Unconditionally use
17065 gen_lowpart_SUBREG, rather than selecting between that
17066 and equivalent gen_rtx_SUBREG code.
17068 2018-02-08 Richard Biener <rguenther@suse.de>
17070 PR tree-optimization/84233
17071 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17072 changed flag instead of boguously re-using phi_inserted.
17074 2018-02-08 Martin Jambor <mjambor@suse.cz>
17076 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17077 static local variables.
17079 2018-02-08 Richard Biener <rguenther@suse.de>
17081 PR tree-optimization/84278
17082 * tree-vect-stmts.c (vectorizable_store): When looking for
17083 smaller vector types to perform grouped strided loads/stores
17084 make sure the mode is supported by the target.
17085 (vectorizable_load): Likewise.
17087 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17089 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17090 Increase LDP/STP opportunities by adding adjacent callee-saves.
17092 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17094 PR rtl-optimization/84068
17095 PR rtl-optimization/83459
17096 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17098 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17100 PR tree-optimization/84224
17101 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17102 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17103 non-zero arguments.
17105 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17108 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17109 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17110 count, remove one USE.
17112 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17114 * doc/install.texi (Configuration): Document the
17115 --with-long-double-format={ibm,ieee} PowerPC configuration
17119 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17120 Convert from define_expand to be define_insn_and_split. Rework
17121 float/double/_Float128 conversions to QI/HI/SImode to work with
17122 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17123 conversions to QI/HImode types did a store and then a load to
17124 truncate the value. For conversions to VSX registers, don't split
17125 the insn, instead emit the code directly. Use the code iterator
17126 any_fix to combine signed and unsigned conversions.
17127 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17128 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17129 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17130 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17131 (fix_<mode>di2_hw): Likewise.
17132 (fixuns_<mode>di2_hw): Likewise.
17133 (fix_<mode>si2_hw): Likewise.
17134 (fixuns_<mode>si2_hw): Likewise.
17135 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17136 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17137 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17138 fix<uns>_trunc<SFDF:mode>si2_p8.
17139 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17140 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17141 (fix<uns>_<mode>_mem): Likewise.
17142 (fctiw<u>z_<mode>_mem): Likewise.
17143 (fix<uns>_<mode>_mem): Likewise.
17144 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17145 the register allocator from doing a direct move to the GPRs to do
17146 a store, and instead use the ISA 3.0 store byte/half-word from
17147 vector register instruction. For IEEE 128-bit floating point,
17148 also optimize stores of 32-bit ints.
17149 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17151 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17153 * genextract.c (push_pathstr_operand): New function to support
17155 (walk_rtx): Call push_pathstr_operand.
17156 (print_path): Support [a-zA-Z].
17158 2018-02-07 Richard Biener <rguenther@suse.de>
17160 PR tree-optimization/84037
17161 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17162 (cse_and_gimplify_to_preheader): Declare.
17163 (vect_get_place_in_interleaving_chain): Likewise.
17164 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17166 (_loop_vec_info::~_loop_vec_info): Delete it.
17167 (cse_and_gimplify_to_preheader): New function.
17168 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17169 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17170 (vectorizable_load): Likewise. For grouped stores always base
17171 the IV on the first element.
17172 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17173 condition before gimplifying.
17175 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17177 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17178 *DIV_EXPR and *MOD_EXPR.
17180 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17183 * config/i386/i386.c (ix86_option_override_internal): Mask out
17184 the CF_SET bit when checking -fcf-protection.
17186 2018-02-07 Tom de Vries <tom@codesourcery.com>
17189 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17192 2018-02-07 Richard Biener <rguenther@suse.de>
17194 PR tree-optimization/84204
17195 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17198 PR tree-optimization/84205
17199 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17200 special-case isl_ast_op_zdiv_r.
17202 PR tree-optimization/84223
17203 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17204 Only add conditions from within the region.
17205 (gather_bbs::after_dom_children): Adjust.
17207 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17210 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17211 * config/avr/avr.md: Only post-reload split REG-REG moves if
17212 either register is GENERAL_REG_P.
17214 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17216 PR tree-optimization/84235
17217 * tree-ssa-scopedtables.c
17218 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17219 if the subtraction is performed in floating point type where NaNs are
17220 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17221 build 1. Formatting fix.
17223 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17226 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17227 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17228 and skip it regardless of bb boundaries. Use CALL_P macro,
17229 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17230 unnecessarily, formatting fix.
17232 2018-02-06 Michael Collison <michael.collison@arm.com>
17234 * config/arm/thumb2.md:
17235 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17236 (*thumb_mov_notscc): Ditto.
17238 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17241 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17242 unsigned_fix, not "s".
17244 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17246 * configure.ac (gcc_fn_eh_frame_ro): New function.
17247 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17248 correct .eh_frame permissions.
17249 * configure: Regenerate.
17251 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17253 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17254 irrelevant options.
17256 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17258 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17259 Display warning message for -mno-speculate-indirect-jumps.
17261 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17263 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17265 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17267 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17269 PR tree-optimization/84225
17270 * tree-eh.c (find_trapping_overflow): Only call
17271 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17273 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17276 * config/i386/i386.c: Reimplement the check of possible options
17277 -mibt/-mshstk conbination. Change error messages.
17278 * doc/invoke.texi: Fix a typo: remove extra '='.
17280 2018-02-06 Marek Polacek <polacek@redhat.com>
17282 PR tree-optimization/84228
17283 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17285 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17288 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17289 emitted arch directives.
17290 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17291 __ARM_FEATURE_COPROC before changing architectures.
17293 2018-02-06 Richard Biener <rguenther@suse.de>
17295 * config/i386/i386.c (print_reg): Fix typo.
17296 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17298 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17300 * configure: Regenerate.
17302 2018-02-05 Martin Sebor <msebor@redhat.com>
17304 PR tree-optimization/83369
17305 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17308 2018-02-05 Martin Liska <mliska@suse.cz>
17310 * doc/invoke.texi: Cherry-pick upstream r323995.
17312 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17314 * ira.c (ira_init_register_move_cost): Adjust comment.
17316 2018-02-05 Martin Liska <mliska@suse.cz>
17318 PR gcov-profile/84137
17319 * doc/gcov.texi: Fix typo in documentation.
17321 2018-02-05 Martin Liska <mliska@suse.cz>
17323 PR gcov-profile/83879
17324 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17325 using dlopen functionality.
17327 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17329 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17330 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17331 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17332 _mm_maskz_range_round_ss): New intrinsics.
17333 (__builtin_ia32_rangesd128_round)
17334 (__builtin_ia32_rangess128_round): Remove.
17335 (__builtin_ia32_rangesd128_mask_round,
17336 __builtin_ia32_rangess128_mask_round): New builtins.
17337 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17338 __builtin_ia32_rangess128_round): Remove.
17339 (__builtin_ia32_rangesd128_mask_round,
17340 __builtin_ia32_rangess128_mask_round): New builtins.
17341 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17342 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17343 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17344 "<round_saeonly_constraint>")): Changed to ...
17345 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17346 "<round_saeonly_scalar_constraint>")): ... this.
17347 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17348 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17349 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17350 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17351 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17353 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17355 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17357 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17358 Remove all values except native, 8540 and 8548.
17360 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17362 * config/i386/i386.c (ix86_output_function_return): Pass
17363 INVALID_REGNUM, instead of -1, as invalid register number to
17364 indirect_thunk_name and output_indirect_thunk.
17366 2018-02-02 Julia Koval <julia.koval@intel.com>
17368 * config.gcc: Add -march=icelake.
17369 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17370 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17371 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17372 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17373 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17374 (processor_target_table): Add icelake.
17375 (ix86_option_override_internal): Handle new PTAs.
17376 (get_builtin_code_for_version): Handle icelake.
17377 (M_INTEL_COREI7_ICELAKE): New.
17378 (fold_builtin_cpu): Handle icelake.
17379 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17380 * doc/invoke.texi: Add -march=icelake.
17382 2018-02-02 Julia Koval <julia.koval@intel.com>
17384 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17385 to wide_int_bitmask.
17386 * wide-int-bitmask.h: New.
17388 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17391 * config/i386/i386.md: Replace Pmode with word_mode in
17392 builtin_setjmp_setup and builtin_longjmp to support x32.
17394 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17398 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17400 (rs6000_supported_cpu_names): New static variable.
17401 (linux_cpu_translation_table): Likewise.
17402 (elf_platform) <cpu>: Define new static variable and use it.
17403 Translate kernel AT_PLATFORM name to canonical name if needed.
17404 Error if platform name is unknown.
17406 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17409 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17411 2018-02-01 Jeff Law <law@redhat.com>
17414 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17415 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17416 the scratch if RELEASE_VIA_POP is false.
17417 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17418 If we have to save a temporary register, decrement SIZE appropriately.
17419 Pass new arguments to release_scratch_register_on_entry.
17420 (ix86_adjust_stack_and_probe): Likewise.
17421 (ix86_emit_probe_stack_range): Pass new arguments to
17422 release_scratch_register_on_entry.
17424 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17426 PR rtl-optimization/84157
17427 * combine.c (change_zero_ext): Use REG_P predicate in
17428 front of HARD_REGISTER_P predicate.
17430 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17432 * config/avr/avr.c (avr_option_override): Move disabling of
17433 -fdelete-null-pointer-checks to...
17434 * common/config/avr/avr-common.c (avr_option_optimization_table):
17437 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17439 PR tree-optimization/81635
17440 * tree-data-ref.c (split_constant_offset_1): For types that
17441 wrap on overflow, try to use range info to prove that wrapping
17444 2018-02-01 Renlin Li <renlin.li@arm.com>
17447 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17448 TAILCALL_ADDR_REGS.
17449 (aarch64_register_move_cost): Likewise.
17450 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17451 TAILCALL_ADDR_REGS.
17452 (REG_CLASS_NAMES): Likewise.
17453 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17454 TAILCALL_ADDR_REGS. Remove IP registers.
17455 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17457 2018-02-01 Richard Biener <rguenther@suse.de>
17459 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17460 for specifying RPO order and allow NULL for that.
17461 * domwalk.c (dom_walker::dom_walker): Likewise.
17462 (dom_walker::walk): Handle NULL RPO order.
17463 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17465 (rewrite_update_dom_walker): Likewise.
17466 (mark_def_dom_walker): Likewise.
17468 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17470 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17471 (aarch64_maybe_expand_sve_subreg_move): Declare.
17472 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17473 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17475 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17476 that are semantically a reverse operation.
17477 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17478 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17479 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17481 (aarch64_can_change_mode_class): For big-endian, forbid changes
17482 between two SVE modes if they have different element sizes.
17484 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17486 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17487 the TImode handling for big-endian targets.
17489 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17491 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17492 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17494 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17495 Remove BSWAP handing for big-endian targets and use the form of
17496 LD1RQ appropariate for the mode.
17498 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17500 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17501 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17502 duplicated element.
17504 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17507 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17508 check for operands that need to go through aarch64_sve_reload_be.
17510 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17512 PR tree-optimization/81661
17513 PR tree-optimization/84117
17514 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17515 * tree-eh.c: Include gimplify.h.
17516 (find_trapping_overflow, replace_trapping_overflow,
17517 rewrite_to_non_trapping_overflow): New functions.
17518 * tree-vect-loop.c: Include tree-eh.h.
17519 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17520 * tree-data-ref.c: Include tree-eh.h.
17521 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17523 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17525 PR rtl-optimization/84123
17526 * combine.c (change_zero_ext): Check if hard register satisfies
17527 can_change_dest_mode before calling gen_lowpart_SUBREG.
17529 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17532 * ira.c (ira_init_register_move_cost): Remove assert.
17534 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17536 PR rtl-optimization/84071
17537 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17538 * doc/tm.texi: Regenerate.
17540 2018-01-31 Richard Biener <rguenther@suse.de>
17542 PR tree-optimization/84132
17543 * tree-data-ref.c (analyze_miv_subscript): Properly
17544 check whether evolution_function_is_affine_multivariate_p
17545 before calling gcd_of_steps_may_divide_p.
17547 2018-01-31 Julia Koval <julia.koval@intel.com>
17550 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17551 * config/i386/i386.md (rdpid_rex64) New.
17552 (rdpid): Make 32bit only.
17554 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17557 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17558 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17560 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17563 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17565 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17567 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17569 PR rtl-optimization/84071
17570 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17571 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17573 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17575 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17576 (arc_attribute_table): Add 'aux' attribute.
17577 (arc_in_small_data_p): Consider aux like variables.
17578 (arc_is_aux_reg_p): New function.
17579 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17580 (arc_get_aux_arg): New function.
17581 (prepare_move_operands): Handle aux-register access.
17582 (arc_handle_aux_attribute): New function.
17583 * doc/extend.texi (ARC Variable attributes): Add subsection.
17585 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17587 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17588 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17589 (arc_attribute_table): Add 'uncached' attribute.
17590 (arc_print_operand): Print '.di' flag for uncached memory
17592 (arc_in_small_data_p): Do not consider for small data the uncached
17594 (arc_is_uncached_mem_p): New function.
17595 * config/arc/predicates.md (compact_store_memory_operand): Check
17596 for uncached memory accesses.
17597 (nonvol_nonimm_operand): Likewise.
17598 * doc/extend.texi (ARC Type Attribute): New subsection.
17600 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17603 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17604 falign-loops=): Add Optimization flag.
17606 2018-01-30 Jeff Law <law@redhat.com>
17609 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17610 INT_REGISTERS_SAVED. Check it prior to calling
17611 get_scratch_register_on_entry.
17612 (ix86_adjust_stack_and_probe): Similarly.
17613 (ix86_emit_probe_stack_range): Similarly.
17614 (ix86_expand_prologue): Corresponding changes.
17616 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17619 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17620 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17622 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17625 * lra-constraints.c (curr_insn_transform): Process AND in the
17628 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17630 PR rtl-optimization/83986
17631 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17632 dependence against last_pending_memory_flush in addition to
17633 pending_jump_insns.
17635 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17637 PR tree-optimization/81611
17638 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17641 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17644 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17647 2018-01-30 Richard Biener <rguenther@suse.de>
17648 Jakub Jelinek <jakub@redhat.com>
17650 PR tree-optimization/84111
17651 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17652 inner loops added during recursion, as they don't have up-to-date
17655 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17658 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17659 (can_inline_edge_by_limits_p): ... here.
17660 (can_early_inline_edge_p, check_callers,
17661 update_caller_keys, update_callee_keys, recursive_inlining,
17662 add_new_edges_to_heap, speculation_useful_p,
17663 inline_small_functions,
17664 inline_small_functions, flatten_function,
17665 inline_to_all_callers_1): Update.
17667 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17669 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17672 2018-01-30 Richard Biener <rguenther@suse.de>
17674 PR tree-optimization/83008
17675 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17676 invariant and constant vector uses in stmts when they need
17677 more than one stmt.
17679 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17682 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17683 * configure: Regenerate.
17685 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17687 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17689 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17690 Use gen_rtx_REG rather than gen_lowpart.
17692 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17694 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17695 rather than 0 when creating partial subregs.
17697 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17699 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17702 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17705 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17706 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17707 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17708 flags. This restores the settings used before the 2017-07-24.
17709 Turning off pre increment/decrement/modify allows IVOPTS to
17710 optimize DF/SF loops where the index is an int.
17712 2018-01-29 Richard Biener <rguenther@suse.de>
17713 Kelvin Nilsen <kelvin@gcc.gnu.org>
17716 * tree-vect-stmts.c (vectorizable_call): Don't call
17717 targetm.vectorize_builtin_md_vectorized_function if callee is
17720 2018-01-22 Carl Love <cel@us.ibm.com>
17722 * doc/extend.tex: Fix typo in second arg in
17723 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17725 2018-01-29 Richard Biener <rguenther@suse.de>
17727 PR tree-optimization/84086
17728 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17729 (flush_ssaname_freelist): When SSA names were released reset
17730 the SCEV hash table.
17732 2018-01-29 Richard Biener <rguenther@suse.de>
17734 PR tree-optimization/84057
17735 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17736 removed paths when removing edges.
17738 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17740 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17741 -mfunction-return=@var{choice}.
17743 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17745 PR diagnostic/84034
17746 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17747 Handle CR like TAB.
17748 (layout::print_source_line): Likewise.
17749 (test_get_line_width_without_trailing_whitespace): Add test cases.
17751 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17753 PR middle-end/84040
17754 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17757 2018-01-26 Jim Wilson <jimw@sifive.com>
17759 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17761 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17764 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17766 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17767 and CMP + SUB-immediate -> SUBS.
17769 2018-01-26 Martin Sebor <msebor@redhat.com>
17771 PR tree-optimization/83896
17772 * tree-ssa-strlen.c (get_string_len): Rename...
17773 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17774 Avoid assuming length is constant.
17775 (handle_char_store): Use HOST_WIDE_INT for string length.
17777 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17780 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17781 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17783 2018-01-26 Richard Biener <rguenther@suse.de>
17785 PR rtl-optimization/84003
17786 * dse.c (record_store): Only record redundant stores when
17787 the earlier store aliases at least all accesses the later one does.
17789 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17791 PR rtl-optimization/83985
17792 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17793 REG_CFA_RESTORE insns.
17794 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17795 REG_CFA_RESTORE insns here.
17798 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17799 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17801 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17803 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17804 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17805 (arc_init): Likewise.
17806 (arc_override_options): Likewise.
17807 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17809 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17811 * config/arc/arc.h (TARGET_DBNZ): Define.
17812 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17813 properly set the tune attribute.
17814 (dbnz): Use TARGET_DBNZ guard.
17815 * config/arc/arc.opt (mtune): Add core3 option.
17817 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17819 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17820 recognize new pic like addresses.
17821 (arc_delegitimize_address): Clean up.
17823 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17825 * config/arc/arc-arches.def: Option mrf16 valid for all
17827 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17828 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17829 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17830 * config/arc/arc-tables.opt: Regenerate.
17831 * config/arc/arc.c (arc_conditional_register_usage): Handle
17832 reduced register file case.
17833 (arc_file_start): Set must have build attributes.
17834 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17835 mrf16 option value.
17836 * config/arc/arc.opt (mrf16): Add new option.
17837 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17838 * config/arc/genmultilib.awk: Handle new mrf16 option.
17839 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17840 * config/arc/t-multilib: Regenerate.
17841 * doc/invoke.texi (ARC Options): Document mrf16 option.
17843 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17845 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17846 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17847 (arc_attribute_table): Add 'secure_call' attribute.
17848 (arc_print_operand): Print secure call operand.
17849 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17851 (arc_is_secure_call_p): New function. * config/arc/arc.md
17852 (call_i): Add support for sjli instruction.
17853 (call_value_i): Likewise.
17854 * config/arc/constraints.md (Csc): New constraint.
17856 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17857 John Eric Martin <John.Martin@emmicro-us.com>
17859 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17860 * config/arc/arc.c (_arc_jli_section): New struct.
17861 (arc_jli_section): New type.
17862 (rc_jli_sections): New static variable.
17863 (arc_handle_jli_attribute): New function.
17864 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17865 (arc_file_end): New function.
17866 (TARGET_ASM_FILE_END): Define.
17867 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17868 (arc_add_jli_section): New function.
17869 (jli_call_scan): Likewise.
17870 (arc_reorg): Call jli_call_scan.
17871 (arc_output_addsi): Remove 'S' from printing asm operand.
17872 (arc_is_jli_call_p): New function.
17873 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17875 (movhi_insn): Likewise.
17876 (movsi_insn): Likewise.
17877 (movsi_set_cc_insn): Likewise.
17878 (loadqi_update): Likewise.
17879 (load_zeroextendqisi_update): Likewise.
17880 (load_signextendqisi_update): Likewise.
17881 (loadhi_update): Likewise.
17882 (load_zeroextendhisi_update): Likewise.
17883 (load_signextendhisi_update): Likewise.
17884 (loadsi_update): Likewise.
17885 (loadsf_update): Likewise.
17886 (movsicc_insn): Likewise.
17887 (bset_insn): Likewise.
17888 (bxor_insn): Likewise.
17889 (bclr_insn): Likewise.
17890 (bmsk_insn): Likewise.
17891 (bicsi3_insn): Likewise.
17892 (cmpsi_cc_c_insn): Likewise.
17893 (movsi_ne): Likewise.
17894 (movsi_cond_exec): Likewise.
17895 (clrsbsi2): Likewise.
17896 (norm_f): Likewise.
17906 (call_i): Remove 'S' asm letter, add jli instruction.
17907 (call_value_i): Likewise.
17908 * config/arc/arc.op (mjli-always): New option.
17909 * config/arc/constraints.md (Cji): New constraint.
17910 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17912 (subsf3_fpx): Likewise.
17913 (mulsf3_fpx): Likewise.
17914 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17916 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17917 function attrbutes.
17918 * doc/invoke.texi (ARC): Document mjli-always option.
17920 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
17922 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17923 avoid addition with 0 and use incw and decw where possible.
17925 2018-01-26 Richard Biener <rguenther@suse.de>
17927 PR tree-optimization/81082
17928 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17929 association if it requires casting to unsigned.
17930 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17931 from fold_plusminus_mult_expr to catch important cases late when
17932 range info is available.
17934 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17936 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17937 * configure.ac (hidden_linkonce): New test.
17938 * configure: Regenerate.
17939 * config.in: Regenerate.
17941 2018-01-26 Julia Koval <julia.koval@intel.com>
17943 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17944 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17945 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17946 _mm_mask_bitshuffle_epi64_mask): Fix type.
17947 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17948 USI_FTYPE_V4DI_V4DI_USI): Remove.
17949 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17950 __builtin_ia32_vpshufbitqmb256_mask,
17951 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17952 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17953 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17955 2018-01-26 Alan Modra <amodra@gmail.com>
17958 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17959 UNSPEC_VBPERMQ. Sort other unspecs.
17961 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
17963 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17965 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
17967 PR middle-end/83055
17968 * predict.c (drop_profile): Do not push/pop cfun; update also
17970 (handle_missing_profiles): Fix logic looking for zero profiles.
17972 2018-01-25 Jakub Jelinek <jakub@redhat.com>
17974 PR middle-end/83977
17975 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17976 on functions with #pragma omp declare simd or functions with simd
17978 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17979 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17980 Remove trailing \n from warning_at calls.
17982 2018-01-25 Tom de Vries <tom@codesourcery.com>
17985 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17986 for neutered workers.
17988 2018-01-24 Joseph Myers <joseph@codesourcery.com>
17991 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17992 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17994 2018-01-24 Jeff Law <law@redhat.com>
17997 * i386.c (get_probe_interval): Move to earlier point.
17998 (ix86_compute_frame_layout): If -fstack-clash-protection and
17999 the frame is larger than the probe interval, then use pushes
18000 to save registers rather than reg->mem moves.
18001 (ix86_expand_prologue): Remove conditional for int_registers_saved
18004 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18007 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18008 min/max for never referenced object.
18010 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18012 PR middle-end/83977
18013 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18015 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18016 attributes from DECL_ATTRIBUTES (decl) without affecting
18017 DECL_ATTRIBUTES (current_function_decl).
18018 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18019 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18021 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18023 PR tree-optimization/83979
18024 * fold-const.c (fold_comparison): Use constant_boolean_node
18025 instead of boolean_{true,false}_node.
18027 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18029 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18032 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18034 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18035 Simplify the clause that sets the length attribute.
18036 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18037 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18038 clause that sets the length attribute.
18039 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18041 2018-01-24 Tom de Vries <tom@codesourcery.com>
18044 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18045 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18046 Add strict parameter.
18047 (prevent_branch_around_nothing): Insert dummy insn between branch to
18048 label and label with no ptx insn inbetween.
18049 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18051 2018-01-24 Tom de Vries <tom@codesourcery.com>
18054 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18055 for neutered threads in warp.
18056 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18058 2018-01-24 Richard Biener <rguenther@suse.de>
18060 PR tree-optimization/83176
18061 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18064 2018-01-24 Richard Biener <rguenther@suse.de>
18066 PR tree-optimization/82819
18067 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18068 code generating pluses that are no-ops in the target precision.
18070 2018-01-24 Richard Biener <rguenther@suse.de>
18072 PR middle-end/84000
18073 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18075 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18077 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18078 to merge probabilities.
18079 * predict.c (probably_never_executed): Also mark as cold functions
18080 with global 0 profile and guessed local profile.
18081 * profile-count.c (profile_probability::combine_with_count): New
18083 * profile-count.h (profile_probability::operator*,
18084 profile_probability::operator*=, profile_probability::operator/,
18085 profile_probability::operator/=): Reduce precision to adjusted
18086 and set value to guessed on contradictory divisions.
18087 (profile_probability::combine_with_freq): Remove.
18088 (profile_probability::combine_wiht_count): Declare.
18089 (profile_count::force_nonzero):: Set to adjusted.
18090 (profile_count::probability_in):: Set quality to adjusted.
18091 * tree-ssa-tail-merge.c (replace_block_by): Use
18092 combine_with_count.
18094 2018-01-23 Andrew Waterman <andrew@sifive.com>
18095 Jim Wilson <jimw@sifive.com>
18097 * config/riscv/riscv.c (riscv_stack_boundary): New.
18098 (riscv_option_override): Set riscv_stack_boundary. Handle
18099 riscv_preferred_stack_boundary_arg.
18100 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18101 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18102 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18103 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18104 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18105 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18107 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18110 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18111 of struct ix86_frame.
18112 (ix86_expand_epilogue): Likewise. Add a local variable for
18113 the reg_save_offset field in struct ix86_frame.
18115 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18117 PR tree-optimization/82604
18118 * tree-loop-distribution.c (enum partition_kind): New enum item
18119 PKIND_PARTIAL_MEMSET.
18120 (partition_builtin_p): Support above new enum item.
18121 (generate_code_for_partition): Ditto.
18122 (compute_access_range): Differentiate cases that equality can be
18123 proven at all loops, the innermost loops or no loops.
18124 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18125 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18126 (finalize_partitions, distribute_loop): Don't fuse partition of
18127 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18128 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18129 parloop is enabled.
18131 2018-01-23 Martin Liska <mliska@suse.cz>
18133 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18134 order to ignore the predictor.
18135 (PRED_POLYMORPHIC_CALL): Likewise.
18136 (PRED_RECURSIVE_CALL): Likewise.
18138 2018-01-23 Martin Liska <mliska@suse.cz>
18140 * tree-profile.c (tree_profiling): Print function header to
18141 aware reader which function we are working on.
18142 * value-prof.c (gimple_find_values_to_profile): Do not print
18143 not interesting value histograms.
18145 2018-01-23 Martin Liska <mliska@suse.cz>
18147 * profile-count.h (enum profile_quality): Add
18148 profile_uninitialized as the first value. Do not number values
18149 as they are zero based.
18150 (profile_count::verify): Update sanity check.
18151 (profile_probability::verify): Likewise.
18153 2018-01-23 Nathan Sidwell <nathan@acm.org>
18155 * doc/invoke.texi (ffor-scope): Deprecate.
18157 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18159 PR tree-optimization/83510
18160 * domwalk.c (set_all_edges_as_executable): New function.
18161 (dom_walker::dom_walker): Convert bool param
18162 "skip_unreachable_blocks" to enum reachability. Move setup of
18163 edge flags to set_all_edges_as_executable and only do it when
18164 reachability is REACHABLE_BLOCKS.
18165 * domwalk.h (enum dom_walker::reachability): New enum.
18166 (dom_walker::dom_walker): Convert bool param
18167 "skip_unreachable_blocks" to enum reachability.
18168 (set_all_edges_as_executable): New decl.
18169 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18170 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18172 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18173 but converting true to REACHABLE_BLOCKS.
18174 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18176 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18177 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18178 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18180 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18181 if check_all_array_refs will be called.
18183 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18185 * tree.c (selftest::test_location_wrappers): Add more test
18188 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18190 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18191 (selftest::test_bit_in_range): Likewise.
18193 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18196 * doc/sourcebuild.texi (vect_float): Say that the selector
18197 only describes the situation when -funsafe-math-optimizations is on.
18198 (vect_float_strict): Document.
18200 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18202 PR tree-optimization/83965
18203 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18204 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18205 instead of checking only for a reduction.
18206 (vect_recog_widen_sum_pattern): Likewise.
18208 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18210 * predict.c (probably_never_executed): Only use precise profile info.
18211 (compute_function_frequency): Skip after inlining hack since we now
18212 have quality checking.
18214 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18216 * profile-count.h (profile_probability::very_unlikely,
18217 profile_probability::unlikely, profile_probability::even): Set
18218 precision to guessed.
18220 2018-01-23 Richard Biener <rguenther@suse.de>
18222 PR tree-optimization/83963
18223 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18224 Properly terminate dominator walk when crossing the exit edge not
18225 when visiting its source block.
18227 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18230 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18231 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18233 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18235 PR tree-optimization/83957
18236 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18237 semicolon after for body surrounded by braces.
18239 PR tree-optimization/83081
18240 * profile-count.h (profile_probability::split): New method.
18241 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18242 Use profile_probability::split.
18243 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18244 when splitting a single conditional jump into 2.
18246 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18248 PR tree-optimization/69452
18249 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18252 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18254 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18255 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18256 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18258 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18260 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18262 * config/rl78/rl78.md (movdi): New define_expand.
18263 * config/rl78/rl78.c (rl78_split_movdi): New function.
18265 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18268 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18270 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18271 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18272 128-bit to produce an UNSPEC move to get the double word with the
18273 signbit and then a shift directly to do signbit.
18274 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18275 implementation with a new version that just does either a direct
18276 move or a regular move. Move memory interface to separate insns.
18277 Move insns so they are next to the expander.
18278 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18279 with signbit move. Split big and little endian case.
18280 (signbit<mode>2_dm_mem_le): Likewise.
18281 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18282 (signbit<mode>2_dm2): Likewise.
18284 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18286 * config/rl78/rl78.md (anddi3): New define_expand.
18288 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18290 * config/rl78/rl78.md (umindi3): New define_expand.
18292 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18294 * config/rl78/rl78.md (smindi3): New define_expand.
18296 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18298 * config/rl78/rl78.md (smaxdi3): New define_expand.
18300 2018-01-22 Carl Love <cel@us.ibm.com>
18302 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18303 LVX_V1TI): Add macro expansion.
18304 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18305 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18306 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18307 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18308 Change check to determine if the instruction is a byte reversing
18309 entry. Fix typo in comment.
18310 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18311 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18312 Add def_builtin calls for new builtins.
18313 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18314 Add define_insn expansion.
18316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18318 * config/rl78/rl78.md (umaxdi3): New define_expand.
18320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18322 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18323 for non-QImode registers.
18325 2018-01-22 Richard Biener <rguenther@suse.de>
18327 PR tree-optimization/83963
18328 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18329 including the loop exit block.
18330 (scop_detection::merge_sese): Likewise.
18331 (scop_detection::add_scop): Do it here instead.
18333 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18335 * doc/sourcebuild.texi (arm_softfloat): Document.
18337 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18340 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18341 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18342 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18344 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18345 David Edelsohn <dje.gcc@gmail.com>
18348 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18349 Change "crset eq" to "crset 2".
18350 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18351 (*call_indirect_aix<mode>_nospec): Likewise.
18352 (*call_value_indirect_aix<mode>_nospec): Likewise.
18353 (*call_indirect_elfv2<mode>_nospec): Likewise.
18354 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18355 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18356 change assembly output from . to $.
18357 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18358 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18359 (*tablejump<mode>_internal1_nospec): Likewise.
18361 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18364 * config/sh/sh_optimize_sett_clrt.cc:
18365 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18367 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18369 PR tree-optimization/83940
18370 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18371 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18372 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18373 use it to pass back the definition type.
18374 (vect_check_store_rhs): Likewise rhs_dt_out.
18375 (vect_build_gather_load_calls): Add a mask_dt argument and use
18376 it instead of a call to vect_is_simple_use.
18377 (vectorizable_store): Update calls to vect_check_load_store_mask
18378 and vect_check_store_rhs. Use the dt returned by the latter instead
18379 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18380 instead of calls to vect_is_simple_use. Pass the scalar rather
18381 than the vector operand to vect_is_simple_use when handling
18382 second and subsequent copies of an rhs value.
18383 (vectorizable_load): Update calls to vect_check_load_store_mask
18384 and vect_build_gather_load_calls. Use the cached mask_dt and
18385 gs_info.offset_dt instead of calls to vect_is_simple_use.
18387 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18389 PR middle-end/83945
18390 * tree-emutls.c: Include gimplify.h.
18391 (lower_emutls_2): New function.
18392 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18393 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18394 it before further processing.
18397 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18398 UINTVAL (trueop1) instead of INTVAL (op1).
18400 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18404 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18405 INCOMING_FRAME_SP_OFFSET if not defined.
18406 (scan_trace): Add ENTRY argument. If true and
18407 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18408 emit a note to adjust the CFA offset.
18409 (create_cfi_notes): Adjust scan_trace callers.
18410 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18411 INCOMING_FRAME_SP_OFFSET in the CIE.
18412 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18413 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18415 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18416 * doc/tm.texi: Regenerated.
18418 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18420 PR rtl-optimization/83147
18421 * lra-constraints.c (remove_inheritance_pseudos): Use
18422 lra_substitute_pseudo_within_insn.
18424 2018-01-19 Tom de Vries <tom@codesourcery.com>
18425 Cesar Philippidis <cesar@codesourcery.com>
18428 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18430 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18433 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18434 spaces for function labels.
18436 2018-01-19 Martin Liska <mliska@suse.cz>
18438 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18439 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18440 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18441 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18442 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18443 (PRED_CONST_RETURN): Change from 69 to 65.
18444 (PRED_NULL_RETURN): Change from 91 to 71.
18445 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18446 (PRED_LOOP_GUARD): Change from 66 to 73.
18448 2018-01-19 Martin Liska <mliska@suse.cz>
18450 * predict.c (predict_insn_def): Add new assert.
18451 (struct branch_predictor): Change type to signed integer.
18452 (test_prediction_value_range): Amend test to cover
18453 PROB_UNINITIALIZED.
18454 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18455 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18456 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18457 (PRED_LOOP_IV_COMPARE): Likewise.
18458 * predict.h (PROB_UNINITIALIZED): Define new constant.
18460 2018-01-19 Martin Liska <mliska@suse.cz>
18462 * predict.c (dump_prediction): Add new format for
18463 analyze_brprob.py script which is enabled with -details
18465 * profile-count.h (precise_p): New function.
18467 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18469 PR tree-optimization/83922
18470 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18471 there are no statements that need masking.
18472 (vect_active_double_reduction_p): New function.
18473 (vect_analyze_loop_operations): Use it when handling phis that
18474 are not in the loop header.
18476 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18478 PR tree-optimization/83914
18479 * tree-vect-loop.c (vectorizable_induction): Don't convert
18480 init_expr or apply the peeling adjustment for inductions
18481 that are nested within the vectorized loop.
18483 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18485 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18488 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18492 * function.h (gimplify_parameters): Add gimple_seq * argument.
18493 * function.c: Include gimple.h and options.h.
18494 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18495 for the added local temporaries if needed.
18496 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18497 if there are any parameter cleanups, wrap whole body into a
18498 try/finally with the cleanups.
18500 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18503 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18504 Use GET_MODE_CLASS for scalar floating point.
18506 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18510 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18511 Fix call of call_cgraph_insertion_hooks.
18513 2018-01-18 Martin Sebor <msebor@redhat.com>
18515 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18517 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18520 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18523 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18526 * common.opt: (-ffile-prefix-map): New option.
18527 * opts.c (common_handle_option): Defer it.
18528 * opts-global.c (handle_common_deferred_options): Handle it.
18529 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18530 * file-prefix-map.h: New file.
18531 (remap_debug_filename, add_debug_prefix_map): ...here.
18532 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18533 * final.c (debug_prefix_map, add_debug_prefix_map
18534 remap_debug_filename): Move to...
18535 * file-prefix-map.c: New file.
18536 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18537 generalize, get rid of alloca(), use strrchr() instead of strchr().
18538 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18539 Implement in terms of add_prefix_map().
18540 (remap_macro_filename, remap_debug_filename): Implement in term of
18542 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18543 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18544 * dbxout.c: Include file-prefix-map.h.
18545 * varasm.c: Likewise.
18546 * vmsdbgout.c: Likewise.
18547 * xcoffout.c: Likewise.
18548 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18549 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18550 * doc/invoke.texi (-ffile-prefix-map): Document.
18551 (-fdebug-prefix-map): Update description.
18553 2018-01-18 Martin Liska <mliska@suse.cz>
18555 * config/i386/i386.c (indirect_thunk_name): Document that also
18557 (output_indirect_thunk): Document why both instructions
18558 (pause and lfence) are generated.
18560 2018-01-18 Richard Biener <rguenther@suse.de>
18562 PR tree-optimization/83887
18563 * graphite-scop-detection.c
18564 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18565 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18566 (scop_detection::merge_sese): Re-implement with a flood-fill
18567 algorithm that properly finds a SESE region if it exists.
18569 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18572 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18573 pointer_diff optimizations use view_convert instead of convert.
18575 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18577 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18578 Generate different code for -mno-speculate-indirect-jumps.
18579 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18580 (*call_indirect_aix<mode>): Disable for
18581 -mno-speculate-indirect-jumps.
18582 (*call_indirect_aix<mode>_nospec): New define_insn.
18583 (*call_value_indirect_aix<mode>): Disable for
18584 -mno-speculate-indirect-jumps.
18585 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18586 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18587 -mno-speculate-indirect-jumps.
18588 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18590 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18592 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18593 long double type, set the flags for noting the default long double
18594 type, even if we don't pass or return a long double type.
18596 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18599 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18602 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18604 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18605 support for merge[hl].
18606 (fold_mergehl_helper): New helper function.
18607 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18608 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18609 (altivec_vmrglw_direct): Add xxmrglw insn.
18611 2018-01-17 Andrew Waterman <andrew@sifive.com>
18613 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18614 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18616 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18619 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18620 call the lto_location_cache before reading the
18621 DECL_SOURCE_LOCATION of the types.
18623 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18624 Richard Sandiford <richard.sandiford@linaro.org>
18626 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18627 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18628 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18629 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18630 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18632 * config/aarch64/constraints.md (aarch64_movti_operand):
18634 * config/aarch64/predicates.md (Uti): Add new constraint.
18636 2018-01-17 Carl Love <cel@us.ibm.com>
18638 * config/rs6000/vsx.md (define_expand xl_len_r,
18639 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18640 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18642 (define_expand, define_insn): Move the shift left from the
18643 define_insn to the define_expand for lxvl and stxvl instructions.
18644 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18645 and XL_LEN_R definitions to PURE.
18647 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18649 * config/i386/i386.c (indirect_thunk_name): Declare regno
18650 as unsigned int. Compare regno with INVALID_REGNUM.
18651 (output_indirect_thunk): Ditto.
18652 (output_indirect_thunk_function): Ditto.
18653 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18654 in the call to output_indirect_thunk_function.
18656 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18658 PR middle-end/83884
18659 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18660 rather than the size of inner_type to determine the stack slot size
18661 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18663 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18666 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18669 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18671 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18672 endian Linux systems to optionally enable multilibs for selecting
18673 the long double type if the user configured an explicit type.
18674 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18675 have no long double multilibs if not defined.
18676 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18677 warn if the user used -mabi={ieee,ibm}longdouble and we built
18678 multilibs for long double.
18679 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18680 appropriate multilib option.
18681 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18683 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18684 for building long double multilibs.
18685 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18687 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18689 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18692 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18694 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18697 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18700 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18703 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18705 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18706 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18708 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18710 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18711 different rtl trees depending on TARGET_64BIT.
18712 (rs6000_gen_lvx): Likewise.
18714 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18716 * config/visium/visium.md (nop): Tweak comment.
18717 (hazard_nop): Likewise.
18719 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18721 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18722 -mspeculate-indirect-jumps.
18723 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18724 for -mno-speculate-indirect-jumps.
18725 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18726 (*call_value_indirect_elfv2<mode>): Disable for
18727 -mno-speculate-indirect-jumps.
18728 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18729 (indirect_jump): Emit different RTL for
18730 -mno-speculate-indirect-jumps.
18731 (*indirect_jump<mode>): Disable for
18732 -mno-speculate-indirect-jumps.
18733 (*indirect_jump<mode>_nospec): New define_insn.
18734 (tablejump): Emit different RTL for
18735 -mno-speculate-indirect-jumps.
18736 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18737 (tablejumpsi_nospec): New define_expand.
18738 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18739 (tablejumpdi_nospec): New define_expand.
18740 (*tablejump<mode>_internal1): Disable for
18741 -mno-speculate-indirect-jumps.
18742 (*tablejump<mode>_internal1_nospec): New define_insn.
18743 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18746 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18748 * caller-save.c (insert_save): Drop unnecessary parameter. All
18751 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18752 Richard Biener <rguenth@suse.de>
18755 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18756 return early, inline manually is_gimple_sizepos. Make sure if we
18757 call gimplify_expr we don't end up with a gimple constant.
18758 * tree.c (variably_modified_type_p): Don't return true for
18759 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18760 * gimplify.h (is_gimple_sizepos): Remove.
18762 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18764 PR tree-optimization/83857
18765 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18766 vectorizable_live_operation for pure SLP statements.
18767 (vectorizable_live_operation): Handle PHIs.
18769 2018-01-16 Richard Biener <rguenther@suse.de>
18771 PR tree-optimization/83867
18772 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18773 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18775 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18778 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18779 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18780 If off is not INTEGER_CST, issue a may not be aligned warning
18781 rather than isn't aligned. Use isn%'t rather than isn't.
18782 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18784 <case MULT_EXPR>: Improve the case when bottom and one of the
18785 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18786 operand, in that case check if the other operand is multiple of
18787 bottom divided by the INTEGER_CST operand.
18789 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18792 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18793 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18794 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18795 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18796 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18797 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18798 (pa_function_arg_size): New function.
18800 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18802 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18803 in a separate statement.
18805 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18807 PR tree-optimization/83847
18808 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18809 group gathers and scatters.
18811 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18813 PR rtl-optimization/86620
18814 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18816 PR rtl-optimization/83213
18817 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18818 to last if both are JUMP_INSNs.
18820 PR tree-optimization/83843
18821 * gimple-ssa-store-merging.c
18822 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18823 store_immediate_info for bswap/nop orig_stores.
18825 2018-01-15 Andrew Waterman <andrew@sifive.com>
18827 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18829 <UDIV>: Increase cost if !TARGET_DIV.
18831 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18833 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18834 (define_attr "cr_logical_3op"): New.
18835 (cceq_ior_compare): Adjust.
18836 (cceq_ior_compare_complement): Adjust.
18837 (*cceq_rev_compare): Adjust.
18838 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18839 (is_cracked_insn): Adjust.
18840 (insn_must_be_first_in_group): Adjust.
18841 * config/rs6000/40x.md: Adjust.
18842 * config/rs6000/440.md: Adjust.
18843 * config/rs6000/476.md: Adjust.
18844 * config/rs6000/601.md: Adjust.
18845 * config/rs6000/603.md: Adjust.
18846 * config/rs6000/6xx.md: Adjust.
18847 * config/rs6000/7450.md: Adjust.
18848 * config/rs6000/7xx.md: Adjust.
18849 * config/rs6000/8540.md: Adjust.
18850 * config/rs6000/cell.md: Adjust.
18851 * config/rs6000/e300c2c3.md: Adjust.
18852 * config/rs6000/e500mc.md: Adjust.
18853 * config/rs6000/e500mc64.md: Adjust.
18854 * config/rs6000/e5500.md: Adjust.
18855 * config/rs6000/e6500.md: Adjust.
18856 * config/rs6000/mpc.md: Adjust.
18857 * config/rs6000/power4.md: Adjust.
18858 * config/rs6000/power5.md: Adjust.
18859 * config/rs6000/power6.md: Adjust.
18860 * config/rs6000/power7.md: Adjust.
18861 * config/rs6000/power8.md: Adjust.
18862 * config/rs6000/power9.md: Adjust.
18863 * config/rs6000/rs64.md: Adjust.
18864 * config/rs6000/titan.md: Adjust.
18866 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18868 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18869 ix86_indirect_branch_register logic.
18871 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18873 * config/i386/constraints.md (Bs): Update
18874 ix86_indirect_branch_register check. Don't check
18875 ix86_indirect_branch_register with GOT_memory_operand.
18877 * config/i386/predicates.md (GOT_memory_operand): Don't check
18878 ix86_indirect_branch_register here.
18879 (GOT32_symbol_operand): Likewise.
18881 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18883 * config/i386/predicates.md (constant_call_address_operand):
18884 Rewrite ix86_indirect_branch_register logic.
18885 (sibcall_insn_operand): Likewise.
18887 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18889 * config/i386/constraints.md (Bs): Replace
18890 ix86_indirect_branch_thunk_register with
18891 ix86_indirect_branch_register.
18893 * config/i386/i386.md (indirect_jump): Likewise.
18894 (tablejump): Likewise.
18895 (*sibcall_memory): Likewise.
18896 (*sibcall_value_memory): Likewise.
18897 Peepholes of indirect call and jump via memory: Likewise.
18898 * config/i386/i386.opt: Likewise.
18899 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18900 (GOT_memory_operand): Likewise.
18901 (call_insn_operand): Likewise.
18902 (sibcall_insn_operand): Likewise.
18903 (GOT32_symbol_operand): Likewise.
18905 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18907 PR middle-end/83837
18908 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18909 type rather than type addr's type points to.
18910 (expand_omp_atomic_mutex): Likewise.
18911 (expand_omp_atomic): Likewise.
18913 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18916 * config/i386/i386.c (output_indirect_thunk_function): Use
18917 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18918 for __x86_return_thunk.
18920 2018-01-15 Richard Biener <rguenther@suse.de>
18922 PR middle-end/83850
18923 * expmed.c (extract_bit_field_1): Fix typo.
18925 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18928 * config/arm/iterators.md (VF): New mode iterator.
18929 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18930 Remove integer-related logic from pattern.
18931 (neon_vabd<mode>_3): Likewise.
18933 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18935 PR middle-end/82694
18936 * common.opt (fstrict-overflow): No longer an alias.
18937 (fwrapv-pointer): New option.
18938 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18939 also for pointer types based on flag_wrapv_pointer.
18940 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18941 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18942 opts->x_flag_wrapv got set.
18943 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18944 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18945 POINTER_TYPE_OVERFLOW_UNDEFINED.
18946 * match.pd: Likewise in address comparison pattern.
18947 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18949 2018-01-15 Richard Biener <rguenther@suse.de>
18952 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18953 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
18954 Reset type names to their identifier if their TYPE_DECL doesn't
18955 have linkage (and thus is used for ODR and devirt).
18956 (save_debug_info_for_decl): Remove.
18957 (save_debug_info_for_type): Likewise.
18958 (add_tree_to_fld_list): Adjust.
18959 * tree-pretty-print.c (dump_generic_node): Make dumping of
18960 type names more robust.
18962 2018-01-15 Richard Biener <rguenther@suse.de>
18964 * BASE-VER: Bump to 8.0.1.
18966 2018-01-14 Martin Sebor <msebor@redhat.com>
18969 * builtins.c (check_access): Avoid warning when the no-warning bit
18972 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
18974 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18975 * ira-color (allocno_hard_regs_compare): Likewise.
18977 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
18980 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18981 Use .pushsection/.popsection.
18983 2018-01-14 Martin Sebor <msebor@redhat.com>
18986 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18988 2018-01-14 Jakub Jelinek <jakub@redhat.com>
18990 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18991 entry from extra_headers.
18992 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18993 extra_headers, make the list bitwise identical to the i?86-*-* one.
18995 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18997 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18998 -mcmodel=large with -mindirect-branch=thunk,
18999 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19000 -mfunction-return=thunk-extern.
19001 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19002 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19003 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19005 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19007 * config/i386/i386.c (print_reg): Print the name of the full
19008 integer register without '%'.
19009 (ix86_print_operand): Handle 'V'.
19010 * doc/extend.texi: Document 'V' modifier.
19012 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19014 * config/i386/constraints.md (Bs): Disallow memory operand for
19015 -mindirect-branch-register.
19017 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19018 (GOT_memory_operand): Likewise.
19019 (call_insn_operand): Likewise.
19020 (sibcall_insn_operand): Likewise.
19021 (GOT32_symbol_operand): Likewise.
19022 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19023 for -mindirect-branch-register.
19024 (tablejump): Likewise.
19025 (*sibcall_memory): Likewise.
19026 (*sibcall_value_memory): Likewise.
19027 Disallow peepholes of indirect call and jump via memory for
19028 -mindirect-branch-register.
19029 (*call_pop): Replace m with Bw.
19030 (*call_value_pop): Likewise.
19031 (*sibcall_pop_memory): Replace m with Bs.
19032 * config/i386/i386.opt (mindirect-branch-register): New option.
19033 * doc/invoke.texi: Document -mindirect-branch-register option.
19035 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19037 * config/i386/i386-protos.h (ix86_output_function_return): New.
19038 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19039 set function_return_type.
19040 (indirect_thunk_name): Add ret_p to indicate thunk for function
19042 (output_indirect_thunk_function): Pass false to
19043 indirect_thunk_name.
19044 (ix86_output_indirect_branch_via_reg): Likewise.
19045 (ix86_output_indirect_branch_via_push): Likewise.
19046 (output_indirect_thunk_function): Create alias for function
19047 return thunk if regno < 0.
19048 (ix86_output_function_return): New function.
19049 (ix86_handle_fndecl_attribute): Handle function_return.
19050 (ix86_attribute_table): Add function_return.
19051 * config/i386/i386.h (machine_function): Add
19052 function_return_type.
19053 * config/i386/i386.md (simple_return_internal): Use
19054 ix86_output_function_return.
19055 (simple_return_internal_long): Likewise.
19056 * config/i386/i386.opt (mfunction-return=): New option.
19057 (indirect_branch): Mention -mfunction-return=.
19058 * doc/extend.texi: Document function_return function attribute.
19059 * doc/invoke.texi: Document -mfunction-return= option.
19061 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19063 * config/i386/i386-opts.h (indirect_branch): New.
19064 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19065 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19066 with local indirect jump when converting indirect call and jump.
19067 (ix86_set_indirect_branch_type): New.
19068 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19069 (indirectlabelno): New.
19070 (indirect_thunk_needed): Likewise.
19071 (indirect_thunk_bnd_needed): Likewise.
19072 (indirect_thunks_used): Likewise.
19073 (indirect_thunks_bnd_used): Likewise.
19074 (INDIRECT_LABEL): Likewise.
19075 (indirect_thunk_name): Likewise.
19076 (output_indirect_thunk): Likewise.
19077 (output_indirect_thunk_function): Likewise.
19078 (ix86_output_indirect_branch_via_reg): Likewise.
19079 (ix86_output_indirect_branch_via_push): Likewise.
19080 (ix86_output_indirect_branch): Likewise.
19081 (ix86_output_indirect_jmp): Likewise.
19082 (ix86_code_end): Call output_indirect_thunk_function if needed.
19083 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19085 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19086 (ix86_attribute_table): Add indirect_branch.
19087 * config/i386/i386.h (machine_function): Add indirect_branch_type
19088 and has_local_indirect_jump.
19089 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19091 (tablejump): Likewise.
19092 (*indirect_jump): Use ix86_output_indirect_jmp.
19093 (*tablejump_1): Likewise.
19094 (simple_return_indirect_internal): Likewise.
19095 * config/i386/i386.opt (mindirect-branch=): New option.
19096 (indirect_branch): New.
19099 (thunk-inline): Likewise.
19100 (thunk-extern): Likewise.
19101 * doc/extend.texi: Document indirect_branch function attribute.
19102 * doc/invoke.texi: Document -mindirect-branch= option.
19104 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19107 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19109 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19111 * ipa-inline.c (want_inline_small_function_p): Return false if
19112 inlining has already failed with CIF_FINAL_ERROR.
19113 (update_caller_keys): Call want_inline_small_function_p before
19115 (update_callee_keys): Likewise.
19117 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19119 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19121 (rs6000_quadword_masked_address_p): Likewise.
19122 (quad_aligned_load_p): Likewise.
19123 (quad_aligned_store_p): Likewise.
19124 (const_load_sequence_p): Add comment to describe the outer-most loop.
19125 (mimic_memory_attributes_and_flags): New function.
19126 (rs6000_gen_stvx): Likewise.
19127 (replace_swapped_aligned_store): Likewise.
19128 (rs6000_gen_lvx): Likewise.
19129 (replace_swapped_aligned_load): Likewise.
19130 (replace_swapped_load_constant): Capitalize argument name in
19131 comment describing this function.
19132 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19133 and stores that access quad-word aligned addresses and replace
19134 with stvx or lvx instructions when appropriate.
19135 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19136 New function prototype.
19137 (rs6000_quadword_masked_address_p): Likewise.
19138 (rs6000_gen_lvx): Likewise.
19139 (rs6000_gen_stvx): Likewise.
19140 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19141 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19142 when memory address is aligned.
19143 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19144 this split to select lvx instruction when memory address is aligned.
19145 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19146 instruction when memory address is aligned.
19147 (*vsx_le_perm_load_v16qi): Likewise.
19148 (four unnamed splitters): Modify to select the stvx instruction
19149 when memory is aligned.
19151 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19153 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19154 which appears in the queue multiple times.
19156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19157 Alan Hayward <alan.hayward@arm.com>
19158 David Sherwood <david.sherwood@arm.com>
19160 * tree-vectorizer.h (vec_lower_bound): New structure.
19161 (_loop_vec_info): Add check_nonzero and lower_bounds.
19162 (LOOP_VINFO_CHECK_NONZERO): New macro.
19163 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19164 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19165 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19166 fields. Make seg_len the distance travelled, not including the
19168 (dr_direction_indicator): Declare.
19169 (dr_zero_step_indicator): Likewise.
19170 (dr_known_forward_stride_p): Likewise.
19171 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19173 (runtime_alias_check_p): Allow runtime alias checks with
19175 (operator ==): Compare access_size and align.
19176 (prune_runtime_alias_test_list): Rework for new distinction between
19177 the access_size and seg_len.
19178 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19180 (get_segment_min_max): New function.
19181 (create_intersect_range_checks): Use it.
19182 (dr_step_indicator): New function.
19183 (dr_direction_indicator): Likewise.
19184 (dr_zero_step_indicator): Likewise.
19185 (dr_known_forward_stride_p): Likewise.
19186 * tree-loop-distribution.c (data_ref_segment_size): Return
19187 DR_STEP * (niters - 1).
19188 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19190 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19191 (vect_preserves_scalar_order_p): New function, split out from...
19192 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19193 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19194 (vect_vfa_access_size): New function.
19195 (vect_vfa_align): Likewise.
19196 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19197 (dump_lower_bound): New function.
19198 (vect_check_lower_bound): Likewise.
19199 (vect_small_gap_p): Likewise.
19200 (vectorizable_with_step_bound_p): Likewise.
19201 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19202 depencies if the vectorization factor is 1. Convert the checks
19203 for nonzero steps into checks on the bounds of DR_STEP. Try using
19204 a bunds check for variable steps if the minimum required step is
19205 relatively small. Update calls to the dr_with_seg_len
19206 constructor and to vect_compile_time_alias.
19207 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19209 (vect_loop_versioning): Call it.
19210 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19212 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19215 Alan Hayward <alan.hayward@arm.com>
19216 David Sherwood <david.sherwood@arm.com>
19218 * doc/sourcebuild.texi (vect_scatter_store): Document.
19219 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19221 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19223 * genopinit.c (main): Add supports_vec_scatter_store and
19224 supports_vec_scatter_store_cached to target_optabs.
19225 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19226 IFN_MASK_SCATTER_STORE.
19227 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19229 * internal-fn.h (internal_store_fn_p): Declare.
19230 (internal_fn_stored_value_index): Likewise.
19231 * internal-fn.c (scatter_store_direct): New macro.
19232 (expand_scatter_store_optab_fn): New function.
19233 (direct_scatter_store_optab_supported_p): New macro.
19234 (internal_store_fn_p): New function.
19235 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19236 IFN_MASK_SCATTER_STORE.
19237 (internal_fn_mask_index): Likewise.
19238 (internal_fn_stored_value_index): New function.
19239 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19240 for scatter stores.
19241 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19242 * optabs-query.c (supports_vec_scatter_store_p): New function.
19243 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19244 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19245 true for scatter stores.
19246 (vect_gather_scatter_fn_p): Handle scatter stores too.
19247 (vect_check_gather_scatter): Consider using scatter stores if
19248 supports_vec_scatter_store_p.
19249 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19250 scatter stores too.
19251 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19252 internal_fn_stored_value_index.
19253 (check_load_store_masking): Handle scatter stores too.
19254 (vect_get_store_rhs): Make public.
19255 (vectorizable_call): Use internal_store_fn_p.
19256 (vectorizable_store): Handle scatter store internal functions.
19257 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19258 when deciding whether the end of the group has been reached.
19259 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19260 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19261 (mask_scatter_store<mode>): New insns.
19263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19264 Alan Hayward <alan.hayward@arm.com>
19265 David Sherwood <david.sherwood@arm.com>
19267 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19268 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19269 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19271 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19272 Use vect_truncate_gather_scatter_offset if we can't treat the
19273 operation as a normal gather load or scatter store.
19274 (get_group_load_store_type): Take the gather_scatter_info
19275 as argument. Try using a gather load or scatter store for
19276 single-element groups.
19277 (get_load_store_type): Update calls to get_group_load_store_type
19278 and vect_use_strided_gather_scatters_p.
19280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19281 Alan Hayward <alan.hayward@arm.com>
19282 David Sherwood <david.sherwood@arm.com>
19284 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19285 optional tree argument.
19286 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19288 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19289 but continue to use the current value as a fallback.
19290 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19291 to compare the updates.
19292 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19293 (get_load_store_type): Use it when handling a strided access.
19294 (vect_get_strided_load_store_ops): New function.
19295 (vect_get_data_ptr_increment): Likewise.
19296 (vectorizable_load): Handle strided gather loads. Always pass
19297 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19300 Alan Hayward <alan.hayward@arm.com>
19301 David Sherwood <david.sherwood@arm.com>
19303 * doc/md.texi (gather_load@var{m}): Document.
19304 (mask_gather_load@var{m}): Likewise.
19305 * genopinit.c (main): Add supports_vec_gather_load and
19306 supports_vec_gather_load_cached to target_optabs.
19307 * optabs-tree.c (init_tree_optimization_optabs): Use
19308 ggc_cleared_alloc to allocate target_optabs.
19309 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19310 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19312 * internal-fn.h (internal_load_fn_p): Declare.
19313 (internal_gather_scatter_fn_p): Likewise.
19314 (internal_fn_mask_index): Likewise.
19315 (internal_gather_scatter_fn_supported_p): Likewise.
19316 * internal-fn.c (gather_load_direct): New macro.
19317 (expand_gather_load_optab_fn): New function.
19318 (direct_gather_load_optab_supported_p): New macro.
19319 (direct_internal_fn_optab): New function.
19320 (internal_load_fn_p): Likewise.
19321 (internal_gather_scatter_fn_p): Likewise.
19322 (internal_fn_mask_index): Likewise.
19323 (internal_gather_scatter_fn_supported_p): Likewise.
19324 * optabs-query.c (supports_at_least_one_mode_p): New function.
19325 (supports_vec_gather_load_p): Likewise.
19326 * optabs-query.h (supports_vec_gather_load_p): Declare.
19327 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19328 and memory_type field.
19329 (NUM_PATTERNS): Bump to 15.
19330 * tree-vect-data-refs.c: Include internal-fn.h.
19331 (vect_gather_scatter_fn_p): New function.
19332 (vect_describe_gather_scatter_call): Likewise.
19333 (vect_check_gather_scatter): Try using internal functions for
19334 gather loads. Recognize existing calls to a gather load function.
19335 (vect_analyze_data_refs): Consider using gather loads if
19336 supports_vec_gather_load_p.
19337 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19338 (vect_get_gather_scatter_offset_type): Likewise.
19339 (vect_convert_mask_for_vectype): Likewise.
19340 (vect_add_conversion_to_patterm): Likewise.
19341 (vect_try_gather_scatter_pattern): Likewise.
19342 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19343 (vect_vect_recog_func_ptrs): Add it.
19344 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19345 internal_fn_mask_index and internal_gather_scatter_fn_p.
19346 (check_load_store_masking): Take the gather_scatter_info as an
19347 argument and handle gather loads.
19348 (vect_get_gather_scatter_ops): New function.
19349 (vectorizable_call): Check internal_load_fn_p.
19350 (vectorizable_load): Likewise. Handle gather load internal
19352 (vectorizable_store): Update call to check_load_store_masking.
19353 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19354 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19355 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19356 (aarch64_gather_scale_operand_d): New predicates.
19357 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19358 (mask_gather_load<mode>): New insns.
19360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19361 Alan Hayward <alan.hayward@arm.com>
19362 David Sherwood <david.sherwood@arm.com>
19364 * optabs.def (fold_left_plus_optab): New optab.
19365 * doc/md.texi (fold_left_plus_@var{m}): Document.
19366 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19367 * internal-fn.c (fold_left_direct): Define.
19368 (expand_fold_left_optab_fn): Likewise.
19369 (direct_fold_left_optab_supported_p): Likewise.
19370 * fold-const-call.c (fold_const_fold_left): New function.
19371 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19372 * tree-parloops.c (valid_reduction_p): New function.
19373 (gather_scalar_reductions): Use it.
19374 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19375 (vect_finish_replace_stmt): Declare.
19376 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19377 (needs_fold_left_reduction_p): New function, split out from...
19378 (vect_is_simple_reduction): ...here. Accept reductions that
19379 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19380 (vect_force_simple_reduction): Also store the reduction type in
19381 the assignment's STMT_VINFO_REDUC_TYPE.
19382 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19383 (merge_with_identity): New function.
19384 (vect_expand_fold_left): Likewise.
19385 (vectorize_fold_left_reduction): Likewise.
19386 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19387 scalar phi in place for it. Check for target support and reject
19388 cases that would reassociate the operation. Defer the transform
19389 phase to vectorize_fold_left_reduction.
19390 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19391 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19392 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19396 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19397 call to ifc_temp_var.
19399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19400 Alan Hayward <alan.hayward@arm.com>
19401 David Sherwood <david.sherwood@arm.com>
19403 * target.def (legitimize_address_displacement): Take the original
19404 offset as a poly_int.
19405 * targhooks.h (default_legitimize_address_displacement): Update
19407 * targhooks.c (default_legitimize_address_displacement): Likewise.
19408 * doc/tm.texi: Regenerate.
19409 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19410 as an argument, moving assert of ad->disp == ad->disp_term to...
19411 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19412 Try calling targetm.legitimize_address_displacement before expanding
19413 the address rather than afterwards, and adjust for the new interface.
19414 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19415 Match the new hook interface. Handle SVE addresses.
19416 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19417 new hook interface.
19419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19421 * Makefile.in (OBJS): Add early-remat.o.
19422 * target.def (select_early_remat_modes): New hook.
19423 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19424 * doc/tm.texi: Regenerate.
19425 * targhooks.h (default_select_early_remat_modes): Declare.
19426 * targhooks.c (default_select_early_remat_modes): New function.
19427 * timevar.def (TV_EARLY_REMAT): New timevar.
19428 * passes.def (pass_early_remat): New pass.
19429 * tree-pass.h (make_pass_early_remat): Declare.
19430 * early-remat.c: New file.
19431 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19433 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19436 Alan Hayward <alan.hayward@arm.com>
19437 David Sherwood <david.sherwood@arm.com>
19439 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19440 vfm1 with a bound_epilog parameter.
19441 (vect_do_peeling): Update calls accordingly, and move the prologue
19442 call earlier in the function. Treat the base bound_epilog as 0 for
19443 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19444 this base when peeling for gaps.
19445 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19446 with fully-masked loops.
19447 (vect_estimate_min_profitable_iters): Handle the single peeled
19448 iteration in that case.
19450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19451 Alan Hayward <alan.hayward@arm.com>
19452 David Sherwood <david.sherwood@arm.com>
19454 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19455 single-element interleaving even if the size is not a power of 2.
19456 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19457 accesses for single-element interleaving if the group size is
19460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19461 Alan Hayward <alan.hayward@arm.com>
19462 David Sherwood <david.sherwood@arm.com>
19464 * doc/md.texi (fold_extract_last_@var{m}): Document.
19465 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19466 * optabs.def (fold_extract_last_optab): New optab.
19467 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19468 * internal-fn.c (fold_extract_direct): New macro.
19469 (expand_fold_extract_optab_fn): Likewise.
19470 (direct_fold_extract_optab_supported_p): Likewise.
19471 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19472 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19473 EXTRACT_LAST_REDUCTION.
19474 (get_initial_def_for_reduction): Do not create an initial vector
19475 for EXTRACT_LAST_REDUCTION reductions.
19476 (vectorizable_reduction): Leave the scalar phi in place for
19477 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19478 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19479 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19480 transform phase to vectorizable_condition.
19481 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19483 (vect_finish_stmt_generation): ...here.
19484 (vect_finish_replace_stmt): New function.
19485 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19486 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19488 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19491 Alan Hayward <alan.hayward@arm.com>
19492 David Sherwood <david.sherwood@arm.com>
19494 * doc/md.texi (extract_last_@var{m}): Document.
19495 * optabs.def (extract_last_optab): New optab.
19496 * internal-fn.def (EXTRACT_LAST): New internal function.
19497 * internal-fn.c (cond_unary_direct): New macro.
19498 (expand_cond_unary_optab_fn): Likewise.
19499 (direct_cond_unary_optab_supported_p): Likewise.
19500 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19501 loops using EXTRACT_LAST.
19502 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19503 (extract_last_<mode>): ...this optab.
19504 (vec_extract<mode><Vel>): Update accordingly.
19506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19507 Alan Hayward <alan.hayward@arm.com>
19508 David Sherwood <david.sherwood@arm.com>
19510 * target.def (empty_mask_is_expensive): New hook.
19511 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19512 * doc/tm.texi: Regenerate.
19513 * targhooks.h (default_empty_mask_is_expensive): Declare.
19514 * targhooks.c (default_empty_mask_is_expensive): New function.
19515 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19516 if the target says that empty masks are expensive.
19517 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19519 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19521 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19522 Alan Hayward <alan.hayward@arm.com>
19523 David Sherwood <david.sherwood@arm.com>
19525 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19526 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19527 (vect_use_loop_mask_for_alignment_p): New function.
19528 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19529 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19530 niters_skip argument. Make sure that the first niters_skip elements
19531 of the first iteration are inactive.
19532 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19533 Update call to vect_set_loop_masks_directly.
19534 (get_misalign_in_elems): New function, split out from...
19535 (vect_gen_prolog_loop_niters): ...here.
19536 (vect_update_init_of_dr): Take a code argument that specifies whether
19537 the adjustment should be added or subtracted.
19538 (vect_update_init_of_drs): Likewise.
19539 (vect_prepare_for_masked_peels): New function.
19540 (vect_do_peeling): Skip prologue peeling if we're using a mask
19541 instead. Update call to vect_update_inits_of_drs.
19542 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19544 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19545 alignment. Do not include the number of peeled iterations in
19546 the minimum threshold in that case.
19547 (vectorizable_induction): Adjust the start value down by
19548 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19549 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19550 Take the number of skipped iterations into account when calculating
19552 * tree-vect-stmts.c (vect_gen_while_not): New function.
19554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19555 Alan Hayward <alan.hayward@arm.com>
19556 David Sherwood <david.sherwood@arm.com>
19558 * doc/sourcebuild.texi (vect_fully_masked): Document.
19559 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19560 default value to 0.
19561 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19563 (vect_analyze_loop_2): ...here. Don't check the vectorization
19564 factor against the number of loop iterations if the loop is
19567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19568 Alan Hayward <alan.hayward@arm.com>
19569 David Sherwood <david.sherwood@arm.com>
19571 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19572 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19573 (dump_groups): Update accordingly.
19574 (iv_use::mem_type): New member variable.
19575 (address_p): New function.
19576 (record_use): Add a mem_type argument and initialize the new
19578 (record_group_use): Add a mem_type argument. Use address_p.
19579 Remove obsolete null checks of base_object. Update call to record_use.
19580 (find_interesting_uses_op): Update call to record_group_use.
19581 (find_interesting_uses_cond): Likewise.
19582 (find_interesting_uses_address): Likewise.
19583 (get_mem_type_for_internal_fn): New function.
19584 (find_address_like_use): Likewise.
19585 (find_interesting_uses_stmt): Try find_address_like_use before
19586 calling find_interesting_uses_op.
19587 (addr_offset_valid_p): Use the iv mem_type field as the type
19588 of the addressed memory.
19589 (add_autoinc_candidates): Likewise.
19590 (get_address_cost): Likewise.
19591 (split_small_address_groups_p): Use address_p.
19592 (split_address_groups): Likewise.
19593 (add_iv_candidate_for_use): Likewise.
19594 (autoinc_possible_for_pair): Likewise.
19595 (rewrite_groups): Likewise.
19596 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19597 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19598 (get_alias_ptr_type_for_ptr_address): New function.
19599 (rewrite_use_address): Rewrite address uses in calls that were
19600 identified by find_address_like_use.
19602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19603 Alan Hayward <alan.hayward@arm.com>
19604 David Sherwood <david.sherwood@arm.com>
19606 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19608 * gimple-expr.h (is_gimple_addressable: Likewise.
19609 * gimple-expr.c (is_gimple_address): Likewise.
19610 * internal-fn.c (expand_call_mem_ref): New function.
19611 (expand_mask_load_optab_fn): Use it.
19612 (expand_mask_store_optab_fn): Likewise.
19614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19615 Alan Hayward <alan.hayward@arm.com>
19616 David Sherwood <david.sherwood@arm.com>
19618 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19619 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19620 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19621 (cond_umax@var{mode}): Document.
19622 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19623 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19624 (cond_umin_optab, cond_umax_optab): New optabs.
19625 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19626 (COND_IOR, COND_XOR): New internal functions.
19627 * internal-fn.h (get_conditional_internal_fn): Declare.
19628 * internal-fn.c (cond_binary_direct): New macro.
19629 (expand_cond_binary_optab_fn): Likewise.
19630 (direct_cond_binary_optab_supported_p): Likewise.
19631 (get_conditional_internal_fn): New function.
19632 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19633 Cope with reduction statements that are vectorized as calls rather
19635 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19636 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19637 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19638 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19639 (UNSPEC_COND_EOR): New unspecs.
19640 (optab): Add mappings for them.
19641 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19642 (sve_int_op, sve_fp_op): New int attributes.
19644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19645 Alan Hayward <alan.hayward@arm.com>
19646 David Sherwood <david.sherwood@arm.com>
19648 * optabs.def (while_ult_optab): New optab.
19649 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19650 * internal-fn.def (WHILE_ULT): New internal function.
19651 * internal-fn.h (direct_internal_fn_supported_p): New override
19652 that takes two types as argument.
19653 * internal-fn.c (while_direct): New macro.
19654 (expand_while_optab_fn): New function.
19655 (convert_optab_supported_p): Likewise.
19656 (direct_while_optab_supported_p): New macro.
19657 * wide-int.h (wi::udiv_ceil): New function.
19658 * tree-vectorizer.h (rgroup_masks): New structure.
19659 (vec_loop_masks): New typedef.
19660 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19661 and fully_masked_p.
19662 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19663 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19664 (vect_max_vf): New function.
19665 (slpeel_make_loop_iterate_ntimes): Delete.
19666 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19667 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19668 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19669 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19670 internal-fn.h, stor-layout.h and optabs-query.h.
19671 (vect_set_loop_mask): New function.
19672 (add_preheader_seq): Likewise.
19673 (add_header_seq): Likewise.
19674 (interleave_supported_p): Likewise.
19675 (vect_maybe_permute_loop_masks): Likewise.
19676 (vect_set_loop_masks_directly): Likewise.
19677 (vect_set_loop_condition_masked): Likewise.
19678 (vect_set_loop_condition_unmasked): New function, split out from
19679 slpeel_make_loop_iterate_ntimes.
19680 (slpeel_make_loop_iterate_ntimes): Rename to..
19681 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19682 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19683 (vect_do_peeling): Update call accordingly.
19684 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19686 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19687 mask_compare_type, can_fully_mask_p and fully_masked_p.
19688 (release_vec_loop_masks): New function.
19689 (_loop_vec_info): Use it to free the loop masks.
19690 (can_produce_all_loop_masks_p): New function.
19691 (vect_get_max_nscalars_per_iter): Likewise.
19692 (vect_verify_full_masking): Likewise.
19693 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19694 retries, and free the mask rgroups before retrying. Check loop-wide
19695 reasons for disallowing fully-masked loops. Make the final decision
19696 about whether use a fully-masked loop or not.
19697 (vect_estimate_min_profitable_iters): Do not assume that peeling
19698 for the number of iterations will be needed for fully-masked loops.
19699 (vectorizable_reduction): Disable fully-masked loops.
19700 (vectorizable_live_operation): Likewise.
19701 (vect_halve_mask_nunits): New function.
19702 (vect_double_mask_nunits): Likewise.
19703 (vect_record_loop_mask): Likewise.
19704 (vect_get_loop_mask): Likewise.
19705 (vect_transform_loop): Handle the case in which the final loop
19706 iteration might handle a partial vector. Call vect_set_loop_condition
19707 instead of slpeel_make_loop_iterate_ntimes.
19708 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19709 (check_load_store_masking): New function.
19710 (prepare_load_store_mask): Likewise.
19711 (vectorizable_store): Handle fully-masked loops.
19712 (vectorizable_load): Likewise.
19713 (supportable_widening_operation): Use vect_halve_mask_nunits for
19715 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19716 (vect_gen_while): New function.
19717 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19718 (aarch64_uqdec<mode>): New insn.
19720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19721 Alan Hayward <alan.hayward@arm.com>
19722 David Sherwood <david.sherwood@arm.com>
19724 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19725 (reduc_xor_scal_optab): New optabs.
19726 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19727 (reduc_xor_scal_@var{m}): Document.
19728 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19729 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19730 internal functions.
19731 * fold-const-call.c (fold_const_call): Handle them.
19732 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19733 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19734 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19735 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19736 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19737 (UNSPEC_XORV): New unspecs.
19738 (optab): Add entries for them.
19739 (BITWISEV): New int iterator.
19740 (bit_reduc_op): New int attributes.
19742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19743 Alan Hayward <alan.hayward@arm.com>
19744 David Sherwood <david.sherwood@arm.com>
19746 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19747 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19748 * optabs.def (vec_shl_insert_optab): New optab.
19749 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19750 (duplicate_and_interleave): Likewise.
19751 * tree-vect-loop.c: Include internal-fn.h.
19752 (neutral_op_for_slp_reduction): New function, split out from
19753 get_initial_defs_for_reduction.
19754 (get_initial_def_for_reduction): Handle option 2 for variable-length
19755 vectors by loading the neutral value into a vector and then shifting
19756 the initial value into element 0.
19757 (get_initial_defs_for_reduction): Replace the code argument with
19758 the neutral value calculated by neutral_op_for_slp_reduction.
19759 Use gimple_build_vector for constant-length vectors.
19760 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19761 but the first group_size elements have a neutral value.
19762 Use duplicate_and_interleave otherwise.
19763 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19764 Update call to get_initial_defs_for_reduction. Handle SLP
19765 reductions for variable-length vectors by creating one vector
19766 result for each scalar result, with the elements associated
19767 with other scalar results stubbed out with the neutral value.
19768 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19769 Require IFN_VEC_SHL_INSERT for double reductions on
19770 variable-length vectors, or SLP reductions that have
19771 a neutral value. Require can_duplicate_and_interleave_p
19772 support for variable-length unchained SLP reductions if there
19773 is no neutral value, such as for MIN/MAX reductions. Also require
19774 the number of vector elements to be a multiple of the number of
19775 SLP statements when doing variable-length unchained SLP reductions.
19776 Update call to vect_create_epilog_for_reduction.
19777 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19778 and remove initial values.
19779 (duplicate_and_interleave): Make public.
19780 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19781 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19784 Alan Hayward <alan.hayward@arm.com>
19785 David Sherwood <david.sherwood@arm.com>
19787 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19788 (can_duplicate_and_interleave_p): New function.
19789 (vect_get_and_check_slp_defs): Take the vector of statements
19790 rather than just the current one. Remove excess parentheses.
19791 Restriction rejectinon of vect_constant_def and vect_external_def
19792 for variable-length vectors to boolean types, or types for which
19793 can_duplicate_and_interleave_p is false.
19794 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19795 (duplicate_and_interleave): New function.
19796 (vect_get_constant_vectors): Use gimple_build_vector for
19797 constant-length vectors and suitable variable-length constant
19798 vectors. Use duplicate_and_interleave for other variable-length
19799 vectors. Don't defer the update when inserting new statements.
19801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19802 Alan Hayward <alan.hayward@arm.com>
19803 David Sherwood <david.sherwood@arm.com>
19805 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19806 min_profitable_iters doesn't go negative.
19808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19809 Alan Hayward <alan.hayward@arm.com>
19810 David Sherwood <david.sherwood@arm.com>
19812 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19813 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19814 * optabs.def (vec_mask_load_lanes_optab): New optab.
19815 (vec_mask_store_lanes_optab): Likewise.
19816 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19817 (MASK_STORE_LANES): Likewise.
19818 * internal-fn.c (mask_load_lanes_direct): New macro.
19819 (mask_store_lanes_direct): Likewise.
19820 (expand_mask_load_optab_fn): Handle masked operations.
19821 (expand_mask_load_lanes_optab_fn): New macro.
19822 (expand_mask_store_optab_fn): Handle masked operations.
19823 (expand_mask_store_lanes_optab_fn): New macro.
19824 (direct_mask_load_lanes_optab_supported_p): Likewise.
19825 (direct_mask_store_lanes_optab_supported_p): Likewise.
19826 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19828 (vect_load_lanes_supported): Likewise.
19829 * tree-vect-data-refs.c (strip_conversion): New function.
19830 (can_group_stmts_p): Likewise.
19831 (vect_analyze_data_ref_accesses): Use it instead of checking
19832 for a pair of assignments.
19833 (vect_store_lanes_supported): Take a masked_p parameter.
19834 (vect_load_lanes_supported): Likewise.
19835 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19836 vect_store_lanes_supported and vect_load_lanes_supported.
19837 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19838 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19839 parameter. Don't allow gaps for masked accesses.
19840 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19841 and vect_load_lanes_supported.
19842 (get_load_store_type): Take a masked_p parameter and update
19843 call to get_group_load_store_type.
19844 (vectorizable_store): Update call to get_load_store_type.
19845 Handle IFN_MASK_STORE_LANES.
19846 (vectorizable_load): Update call to get_load_store_type.
19847 Handle IFN_MASK_LOAD_LANES.
19849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19850 Alan Hayward <alan.hayward@arm.com>
19851 David Sherwood <david.sherwood@arm.com>
19853 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19855 * config/aarch64/aarch64-protos.h
19856 (aarch64_sve_struct_memory_operand_p): Declare.
19857 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19858 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19859 (VPRED, vpred): Handle SVE structure modes.
19860 * config/aarch64/constraints.md (Utx): New constraint.
19861 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19862 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19863 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19864 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19865 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19866 structure modes. Split into pieces after RA.
19867 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19868 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19870 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19871 SVE structure modes.
19872 (aarch64_classify_address): Likewise.
19873 (sizetochar): Move earlier in file.
19874 (aarch64_print_operand): Handle SVE register lists.
19875 (aarch64_array_mode): New function.
19876 (aarch64_sve_struct_memory_operand_p): Likewise.
19877 (TARGET_ARRAY_MODE): Redefine.
19879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19880 Alan Hayward <alan.hayward@arm.com>
19881 David Sherwood <david.sherwood@arm.com>
19883 * target.def (array_mode): New target hook.
19884 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19885 * doc/tm.texi: Regenerate.
19886 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19887 * hooks.c (hook_optmode_mode_uhwi_none): New function.
19888 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19889 targetm.array_mode.
19890 * stor-layout.c (mode_for_array): Likewise. Support polynomial
19893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19894 Alan Hayward <alan.hayward@arm.com>
19895 David Sherwood <david.sherwood@arm.com>
19897 * fold-const.c (fold_binary_loc): Check the argument types
19898 rather than the result type when testing for a vector operation.
19900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19902 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19903 * doc/tm.texi: Regenerate.
19905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19906 Alan Hayward <alan.hayward@arm.com>
19907 David Sherwood <david.sherwood@arm.com>
19909 * doc/invoke.texi (-msve-vector-bits=): Document new option.
19910 (sve): Document new AArch64 extension.
19911 * doc/md.texi (w): Extend the description of the AArch64
19912 constraint to include SVE vectors.
19913 (Upl, Upa): Document new AArch64 predicate constraints.
19914 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19916 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19917 (msve-vector-bits=): New option.
19918 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19919 SVE when these are disabled.
19920 (sve): New extension.
19921 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19922 modes. Adjust their number of units based on aarch64_sve_vg.
19923 (MAX_BITSIZE_MODE_ANY_MODE): Define.
19924 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19925 aarch64_addr_query_type.
19926 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19927 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19928 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19929 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19930 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19931 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19932 (aarch64_simd_imm_zero_p): Delete.
19933 (aarch64_check_zero_based_sve_index_immediate): Declare.
19934 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19935 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19936 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19937 (aarch64_sve_float_mul_immediate_p): Likewise.
19938 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19939 rather than an rtx.
19940 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19941 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19942 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19943 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19944 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19945 (aarch64_regmode_natural_size): Likewise.
19946 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19947 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19949 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19950 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19951 for VG and the SVE predicate registers.
19952 (V_ALIASES): Add a "z"-prefixed alias.
19953 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19954 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19955 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19956 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19957 (REG_CLASS_NAMES): Add entries for them.
19958 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
19959 and the predicate registers.
19960 (aarch64_sve_vg): Declare.
19961 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19962 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19963 (REGMODE_NATURAL_SIZE): Define.
19964 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19966 * config/aarch64/aarch64.c: Include cfgrtl.h.
19967 (simd_immediate_info): Add a constructor for series vectors,
19968 and an associated step field.
19969 (aarch64_sve_vg): New variable.
19970 (aarch64_dbx_register_number): Handle VG and the predicate registers.
19971 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19972 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19973 (VEC_ANY_DATA, VEC_STRUCT): New constants.
19974 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19975 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19976 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19977 (aarch64_get_mask_mode): New functions.
19978 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19979 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19980 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
19981 predicate modes and predicate registers. Explicitly restrict
19982 GPRs to modes of 16 bytes or smaller. Only allow FP registers
19983 to store a vector mode if it is recognized by
19984 aarch64_classify_vector_mode.
19985 (aarch64_regmode_natural_size): New function.
19986 (aarch64_hard_regno_caller_save_mode): Return the original mode
19988 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19989 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19990 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19991 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19993 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
19994 does not overlap dest if the function is frame-related. Handle
19996 (aarch64_split_add_offset): New function.
19997 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19998 them aarch64_add_offset.
19999 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20000 and update call to aarch64_sub_sp.
20001 (aarch64_add_cfa_expression): New function.
20002 (aarch64_expand_prologue): Pass extra temporary registers to the
20003 functions above. Handle the case in which we need to emit new
20004 DW_CFA_expressions for registers that were originally saved
20005 relative to the stack pointer, but now have to be expressed
20006 relative to the frame pointer.
20007 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20009 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20010 IP0 and IP1 values for SVE frames.
20011 (aarch64_expand_vec_series): New function.
20012 (aarch64_expand_sve_widened_duplicate): Likewise.
20013 (aarch64_expand_sve_const_vector): Likewise.
20014 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20015 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20016 into the register, rather than emitting a SET directly.
20017 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20018 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20019 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20020 (offset_9bit_signed_scaled_p): New functions.
20021 (aarch64_replicate_bitmask_imm): New function.
20022 (aarch64_bitmask_imm): Use it.
20023 (aarch64_cannot_force_const_mem): Reject expressions involving
20024 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20025 (aarch64_classify_index): Handle SVE indices, by requiring
20026 a plain register index with a scale that matches the element size.
20027 (aarch64_classify_address): Handle SVE addresses. Assert that
20028 the mode of the address is VOIDmode or an integer mode.
20029 Update call to aarch64_classify_symbol.
20030 (aarch64_classify_symbolic_expression): Update call to
20031 aarch64_classify_symbol.
20032 (aarch64_const_vec_all_in_range_p): New function.
20033 (aarch64_print_vector_float_operand): Likewise.
20034 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20035 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20036 and the FP immediates 1.0 and 0.5.
20037 (aarch64_print_address_internal): Handle SVE addresses.
20038 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20039 (aarch64_regno_regclass): Handle predicate registers.
20040 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20042 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20043 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20044 (aarch64_convert_sve_vector_bits): New function.
20045 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20046 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20047 rather than an rtx.
20048 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20049 Handle SVE vector and predicate modes. Accept VL-based constants
20050 that need only one temporary register, and VL offsets that require
20051 no temporary registers.
20052 (aarch64_conditional_register_usage): Mark the predicate registers
20053 as fixed if SVE isn't available.
20054 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20055 Return true for SVE vector and predicate modes.
20056 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20057 rather than an unsigned int. Handle SVE modes.
20058 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20060 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20062 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20063 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20064 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20065 (aarch64_sve_float_mul_immediate_p): New functions.
20066 (aarch64_sve_valid_immediate): New function.
20067 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20068 Explicitly reject structure modes. Check for INDEX constants.
20069 Handle PTRUE and PFALSE constants.
20070 (aarch64_check_zero_based_sve_index_immediate): New function.
20071 (aarch64_simd_imm_zero_p): Delete.
20072 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20073 vector modes. Accept constants in the range of CNT[BHWD].
20074 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20075 ask for an Advanced SIMD mode.
20076 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20077 (aarch64_simd_vector_alignment): Handle SVE predicates.
20078 (aarch64_vectorize_preferred_vector_alignment): New function.
20079 (aarch64_simd_vector_alignment_reachable): Use it instead of
20081 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20082 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20084 (MAX_VECT_LEN): Delete.
20085 (expand_vec_perm_d): Add a vec_flags field.
20086 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20087 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20088 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20090 (aarch64_evpc_rev): Rename to...
20091 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20092 (aarch64_evpc_rev_global): New function.
20093 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20094 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20096 (aarch64_evpc_sve_tbl): New function.
20097 (aarch64_expand_vec_perm_const_1): Update after rename of
20098 aarch64_evpc_rev. Handle SVE permutes too, trying
20099 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20100 than aarch64_evpc_tbl.
20101 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20102 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20103 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20104 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20105 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20106 (aarch64_expand_sve_vcond): New functions.
20107 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20108 of aarch64_vector_mode_p.
20109 (aarch64_dwarf_poly_indeterminate_value): New function.
20110 (aarch64_compute_pressure_classes): Likewise.
20111 (aarch64_can_change_mode_class): Likewise.
20112 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20113 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20114 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20115 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20116 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20117 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20118 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20119 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20121 (Dn, Dl, Dr): Accept const as well as const_vector.
20122 (Dz): Likewise. Compare against CONST0_RTX.
20123 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20124 of "vector" where appropriate.
20125 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20126 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20127 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20128 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20129 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20130 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20131 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20132 (v_int_equiv): Extend to SVE modes.
20133 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20135 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20136 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20137 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20138 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20139 (SVE_COND_FP_CMP): New int iterators.
20140 (perm_hilo): Handle the new unpack unspecs.
20141 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20143 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20144 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20145 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20146 (aarch64_equality_operator, aarch64_constant_vector_operand)
20147 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20148 (aarch64_sve_nonimmediate_operand): Likewise.
20149 (aarch64_sve_general_operand): Likewise.
20150 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20151 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20152 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20153 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20154 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20155 (aarch64_sve_float_arith_immediate): Likewise.
20156 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20157 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20158 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20159 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20160 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20161 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20162 (aarch64_sve_float_arith_operand): Likewise.
20163 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20164 (aarch64_sve_float_mul_operand): Likewise.
20165 (aarch64_sve_vec_perm_operand): Likewise.
20166 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20167 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20168 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20169 as well as const_vector.
20170 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20171 in file. Use CONST0_RTX and CONSTM1_RTX.
20172 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20173 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20174 Use aarch64_simd_imm_zero.
20175 * config/aarch64/aarch64-sve.md: New file.
20176 * config/aarch64/aarch64.md: Include it.
20177 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20178 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20179 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20180 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20181 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20182 (sve): New attribute.
20183 (enabled): Disable instructions with the sve attribute unless
20185 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20186 aarch64_expand_mov_immediate.
20187 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20188 CNT[BHSD] immediates.
20189 (movti): Split CONST_POLY_INT moves into two halves.
20190 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20191 Split additions that need a temporary here if the destination
20192 is the stack pointer.
20193 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20194 (*add<mode>3_poly_1): New instruction.
20195 (set_clobber_cc): New expander.
20197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20199 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20200 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20201 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20202 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20203 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20204 Change innermode from fixed_mode_size to machine_mode.
20205 (simplify_subreg): Update call accordingly. Handle a constant-sized
20206 subreg of a variable-length CONST_VECTOR.
20208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20209 Alan Hayward <alan.hayward@arm.com>
20210 David Sherwood <david.sherwood@arm.com>
20212 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20213 (add_offset_to_base): New function, split out from...
20214 (create_mem_ref): ...here. When handling a scale other than 1,
20215 check first whether the address is valid without the offset.
20216 Add it into the base if so, leaving the index and scale as-is.
20218 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20221 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20222 fold_for_warn before checking if arg2 is INTEGER_CST.
20224 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20226 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20227 (store_multiple_operation): Delete.
20228 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20229 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20230 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20231 guarded by TARGET_STRING.
20232 (rs6000_output_load_multiple): Delete.
20233 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20234 OPTION_MASK_STRING / TARGET_STRING handling.
20235 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20236 (const rs6000_opt_masks) <"string">: Change mask to 0.
20237 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20238 (MASK_STRING): Delete.
20239 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20241 (load_multiple): Delete.
20248 (store_multiple): Delete.
20255 (movmemsi_8reg): Delete.
20256 (corresponding unnamed define_insn): Delete.
20257 (movmemsi_6reg): Delete.
20258 (corresponding unnamed define_insn): Delete.
20259 (movmemsi_4reg): Delete.
20260 (corresponding unnamed define_insn): Delete.
20261 (movmemsi_2reg): Delete.
20262 (corresponding unnamed define_insn): Delete.
20263 (movmemsi_1reg): Delete.
20264 (corresponding unnamed define_insn): Delete.
20265 * config/rs6000/rs6000.opt (mno-string): New.
20266 (mstring): Replace by deprecation warning stub.
20267 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20269 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20271 * regrename.c (regrename_do_replace): If replacing the same
20272 reg multiple times, try to reuse last created gen_raw_REG.
20275 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20276 main to workaround a bug in GDB.
20278 2018-01-12 Tom de Vries <tom@codesourcery.com>
20281 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20283 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20285 PR rtl-optimization/80481
20286 * ira-color.c (get_cap_member): New function.
20287 (allocnos_conflict_by_live_ranges_p): Use it.
20288 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20289 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20291 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20294 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20295 (*saddl_se_1): Ditto.
20296 (*ssubsi_1): Ditto.
20297 (*ssubl_se_1): Ditto.
20299 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20301 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20302 rather than wi::to_widest for DR_INITs.
20303 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20304 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20305 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20307 (vect_analyze_group_access_1): Note that here.
20309 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20311 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20312 polynomial type sizes.
20314 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20316 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20317 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20318 (gimple_add_tmp_var): Likewise.
20320 2018-01-12 Martin Liska <mliska@suse.cz>
20322 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20323 (gimple_alloc_sizes): Likewise.
20324 (dump_gimple_statistics): Use PRIu64 in printf format.
20325 * gimple.h: Change uint64_t to int.
20327 2018-01-12 Martin Liska <mliska@suse.cz>
20329 * tree-core.h: Use uint64_t instead of int.
20330 * tree.c (tree_node_counts): Likewise.
20331 (tree_node_sizes): Likewise.
20332 (dump_tree_statistics): Use PRIu64 in printf format.
20334 2018-01-12 Martin Liska <mliska@suse.cz>
20336 * Makefile.in: As qsort_chk is implemented in vec.c, add
20337 vec.o to linkage of gencfn-macros.
20338 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20339 passing the info to record_node_allocation_statistics.
20340 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20342 * ggc-common.c (struct ggc_usage): Add operator== and use
20343 it in operator< and compare function.
20344 * mem-stats.h (struct mem_usage): Likewise.
20345 * vec.c (struct vec_usage): Remove operator< and compare
20346 function. Can be simply inherited.
20348 2018-01-12 Martin Jambor <mjambor@suse.cz>
20351 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20352 * tree-ssa-math-opts.c: Include domwalk.h.
20353 (convert_mult_to_fma_1): New function.
20354 (fma_transformation_info): New type.
20355 (fma_deferring_state): Likewise.
20356 (cancel_fma_deferring): New function.
20357 (result_of_phi): Likewise.
20358 (last_fma_candidate_feeds_initial_phi): Likewise.
20359 (convert_mult_to_fma): Added deferring logic, split actual
20360 transformation to convert_mult_to_fma_1.
20361 (math_opts_dom_walker): New type.
20362 (math_opts_dom_walker::after_dom_children): New method, body moved
20363 here from pass_optimize_widening_mul::execute, added deferring logic
20365 (pass_optimize_widening_mul::execute): Moved most of code to
20366 math_opts_dom_walker::after_dom_children.
20367 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20368 * config/i386/i386.c (ix86_option_override_internal): Added
20369 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20371 2018-01-12 Richard Biener <rguenther@suse.de>
20374 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20375 inline instance vars.
20377 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20380 * config/rx/rx.c (rx_is_restricted_memory_address):
20381 Handle SUBREG case.
20383 2018-01-12 Richard Biener <rguenther@suse.de>
20385 PR tree-optimization/80846
20386 * target.def (split_reduction): New target hook.
20387 * targhooks.c (default_split_reduction): New function.
20388 * targhooks.h (default_split_reduction): Declare.
20389 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20390 target requests first reduce vectors by combining low and high
20392 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20393 (get_vectype_for_scalar_type_and_size): Export.
20394 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20395 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20396 * doc/tm.texi: Regenerate.
20397 * config/i386/i386.c (ix86_split_reduction): Implement
20398 TARGET_VECTORIZE_SPLIT_REDUCTION.
20400 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20403 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20404 in PIC mode except for TARGET_VXWORKS_RTP.
20405 * config/sparc/sparc.c: Include cfgrtl.h.
20406 (TARGET_INIT_PIC_REG): Define.
20407 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20408 (sparc_pic_register_p): New predicate.
20409 (sparc_legitimate_address_p): Use it.
20410 (sparc_legitimize_pic_address): Likewise.
20411 (sparc_delegitimize_address): Likewise.
20412 (sparc_mode_dependent_address_p): Likewise.
20413 (gen_load_pcrel_sym): Remove 4th parameter.
20414 (load_got_register): Adjust call to above. Remove obsolete stuff.
20415 (sparc_expand_prologue): Do not call load_got_register here.
20416 (sparc_flat_expand_prologue): Likewise.
20417 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20418 (sparc_use_pseudo_pic_reg): New function.
20419 (sparc_init_pic_reg): Likewise.
20420 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20421 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20423 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20425 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20426 Add item for branch_cost.
20428 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20430 PR rtl-optimization/83565
20431 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20432 not extend the result to a larger mode for rotate operations.
20433 (num_sign_bit_copies1): Likewise.
20435 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20438 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20440 Use values-Xc.o for -pedantic.
20441 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20443 2018-01-12 Martin Liska <mliska@suse.cz>
20446 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20448 (possible_polymorphic_call_targets): Use it.
20449 (ipa_devirt): Likewise.
20451 2018-01-12 Martin Liska <mliska@suse.cz>
20453 * profile-count.h (enum profile_quality): Use 0 as invalid
20454 enum value of profile_quality.
20456 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20458 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20459 -mext-string options.
20461 2018-01-12 Richard Biener <rguenther@suse.de>
20463 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20464 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20465 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20467 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20469 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20471 * configure.ac (--with-long-double-format): Add support for the
20472 configuration option to change the default long double format on
20474 * config.gcc (powerpc*-linux*-*): Likewise.
20475 * configure: Regenerate.
20476 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20477 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20478 used without modification.
20480 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20482 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20483 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20484 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20485 MISC_BUILTIN_SPEC_BARRIER.
20486 (rs6000_init_builtins): Likewise.
20487 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20489 (speculation_barrier): New define_insn.
20490 * doc/extend.texi: Document __builtin_speculation_barrier.
20492 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20495 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20496 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20497 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20499 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20500 integral modes instead of "ss" and "sd".
20501 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20502 vectors with 32-bit and 64-bit elements.
20503 (vecdupssescalarmodesuffix): New mode attribute.
20504 (vec_dup<mode>): Use it.
20506 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20509 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20510 frame if argument is passed on stack.
20512 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20515 * ree.c (combine_reaching_defs): Optimize also
20516 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20517 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20519 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20521 PR middle-end/83189
20522 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20524 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20526 PR middle-end/83718
20527 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20528 after they are computed.
20530 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20532 PR tree-optimization/83695
20533 * gimple-loop-linterchange.cc
20534 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20535 reset cached scev information after interchange.
20536 (pass_linterchange::execute): Remove call to scev_reset_htab.
20538 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20540 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20541 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20542 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20543 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20544 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20545 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20546 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20547 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20548 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20549 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20550 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20551 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20552 (V_lane_reg): Likewise.
20553 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20555 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20556 (vfmal_lane_low<mode>_intrinsic,
20557 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20558 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20559 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20560 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20561 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20562 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20564 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20566 * config/arm/arm-cpus.in (fp16fml): New feature.
20567 (ALL_SIMD): Add fp16fml.
20568 (armv8.2-a): Add fp16fml as an option.
20569 (armv8.3-a): Likewise.
20570 (armv8.4-a): Add fp16fml as part of fp16.
20571 * config/arm/arm.h (TARGET_FP16FML): Define.
20572 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20574 * config/arm/arm-modes.def (V2HF): Define.
20575 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20576 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20577 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20578 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20579 vfmsl_low, vfmsl_high): New set of builtins.
20580 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20581 (vfml_op): New code attribute.
20582 (VFMLHALVES): New int iterator.
20583 (VFML, VFMLSEL): New mode attributes.
20584 (V_reg): Define mapping for V2HF.
20585 (V_hi, V_lo): New mode attributes.
20586 (VF_constraint): Likewise.
20587 (vfml_half, vfml_half_selector): New int attributes.
20588 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20590 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20591 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20593 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20594 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20595 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20596 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20598 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20599 Document new effective target and option set.
20601 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20603 * config/arm/arm-cpus.in (armv8_4): New feature.
20604 (ARMv8_4a): New fgroup.
20605 (armv8.4-a): New arch.
20606 * config/arm/arm-tables.opt: Regenerate.
20607 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20608 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20609 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20610 Add matching rules for -march=armv8.4-a and extensions.
20611 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20613 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20616 * config/rx/rx.md (BW): New mode attribute.
20617 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20619 2018-01-11 Richard Biener <rguenther@suse.de>
20621 PR tree-optimization/83435
20622 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20623 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20624 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20626 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20627 Alan Hayward <alan.hayward@arm.com>
20628 David Sherwood <david.sherwood@arm.com>
20630 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20632 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20633 (aarch64_print_address_internal): Use it to check for a zero offset.
20635 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20636 Alan Hayward <alan.hayward@arm.com>
20637 David Sherwood <david.sherwood@arm.com>
20639 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20640 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20641 Return a poly_int64 rather than a HOST_WIDE_INT.
20642 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20643 rather than a HOST_WIDE_INT.
20644 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20645 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20646 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20647 final_offset from HOST_WIDE_INT to poly_int64.
20648 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20649 to_constant when getting the number of units in an Advanced SIMD
20651 (aarch64_builtin_vectorized_function): Check for a constant number
20653 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20655 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20656 attribute instead of GET_MODE_NUNITS.
20657 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20658 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20659 GET_MODE_SIZE for fixed-size registers.
20660 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20661 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20662 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20663 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20664 (aarch64_print_operand, aarch64_print_address_internal)
20665 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20666 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20667 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20668 Handle polynomial GET_MODE_SIZE.
20669 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20670 wider than SImode without modification.
20671 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20672 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20673 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20674 passing and returning SVE modes.
20675 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20676 rather than GEN_INT.
20677 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20678 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20679 (aarch64_allocate_and_probe_stack_space): Likewise.
20680 (aarch64_layout_frame): Cope with polynomial offsets.
20681 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20682 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20683 polynomial offsets.
20684 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20685 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20686 poly_int64 rather than a HOST_WIDE_INT.
20687 (aarch64_get_separate_components, aarch64_process_components)
20688 (aarch64_expand_prologue, aarch64_expand_epilogue)
20689 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20690 (aarch64_anchor_offset): New function, split out from...
20691 (aarch64_legitimize_address): ...here.
20692 (aarch64_builtin_vectorization_cost): Handle polynomial
20693 TYPE_VECTOR_SUBPARTS.
20694 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20696 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20697 number of elements from the PARALLEL rather than the mode.
20698 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20699 rather than GET_MODE_BITSIZE.
20700 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20701 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20702 (aarch64_expand_vec_perm_const_1): Handle polynomial
20703 d->perm.length () and d->perm elements.
20704 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20705 Apply to_constant to d->perm elements.
20706 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20707 polynomial CONST_VECTOR_NUNITS.
20708 (aarch64_move_pointer): Take amount as a poly_int64 rather
20710 (aarch64_progress_pointer): Avoid temporary variable.
20711 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20712 the mode attribute instead of GET_MODE.
20714 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20715 Alan Hayward <alan.hayward@arm.com>
20716 David Sherwood <david.sherwood@arm.com>
20718 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20719 x exists before using it.
20720 (aarch64_add_constant_internal): Rename to...
20721 (aarch64_add_offset_1): ...this. Replace regnum with separate
20722 src and dest rtxes. Handle the case in which they're different,
20723 including when the offset is zero. Replace scratchreg with an rtx.
20724 Use 2 additions if there is no spare register into which we can
20725 move a 16-bit constant.
20726 (aarch64_add_constant): Delete.
20727 (aarch64_add_offset): Replace reg with separate src and dest
20728 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20729 Use aarch64_add_offset_1.
20730 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20731 an rtx rather than an int. Take the delta as a poly_int64
20732 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20733 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20734 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20735 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20736 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20737 and aarch64_add_sp.
20738 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20739 aarch64_add_constant.
20741 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20743 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20744 Use scalar_float_mode.
20746 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20748 * config/aarch64/aarch64-simd.md
20749 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20750 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20751 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20752 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20753 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20754 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20755 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20756 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20757 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20758 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20760 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20763 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20764 targ_options->x_arm_arch_string is non NULL.
20766 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20768 * config/aarch64/aarch64.h
20769 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20771 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20774 * expmed.c (emit_store_flag_force): Swap if const op0
20775 and change VOIDmode to mode of op0.
20777 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20779 PR rtl-optimization/83761
20780 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20781 than bytes to mode_for_size.
20783 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20785 PR middle-end/83189
20786 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20787 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20790 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20792 PR middle-end/83575
20793 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20794 when in layout mode.
20795 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20796 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20799 2018-01-10 Michael Collison <michael.collison@arm.com>
20801 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20802 * config/aarch64/aarch64-option-extension.def: Add
20803 AARCH64_OPT_EXTENSION of 'fp16fml'.
20804 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20805 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20806 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20807 * config/aarch64/constraints.md (Ui7): New constraint.
20808 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20809 (VFMLA_SEL_W): Ditto.
20812 (VFMLA16_LOW): New int iterator.
20813 (VFMLA16_HIGH): Ditto.
20814 (UNSPEC_FMLAL): New unspec.
20815 (UNSPEC_FMLSL): Ditto.
20816 (UNSPEC_FMLAL2): Ditto.
20817 (UNSPEC_FMLSL2): Ditto.
20818 (f16mac): New code attribute.
20819 * config/aarch64/aarch64-simd-builtins.def
20820 (aarch64_fmlal_lowv2sf): Ditto.
20821 (aarch64_fmlsl_lowv2sf): Ditto.
20822 (aarch64_fmlalq_lowv4sf): Ditto.
20823 (aarch64_fmlslq_lowv4sf): Ditto.
20824 (aarch64_fmlal_highv2sf): Ditto.
20825 (aarch64_fmlsl_highv2sf): Ditto.
20826 (aarch64_fmlalq_highv4sf): Ditto.
20827 (aarch64_fmlslq_highv4sf): Ditto.
20828 (aarch64_fmlal_lane_lowv2sf): Ditto.
20829 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20830 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20831 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20832 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20833 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20834 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20835 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20836 (aarch64_fmlal_lane_highv2sf): Ditto.
20837 (aarch64_fmlsl_lane_highv2sf): Ditto.
20838 (aarch64_fmlal_laneq_highv2sf): Ditto.
20839 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20840 (aarch64_fmlalq_lane_highv4sf): Ditto.
20841 (aarch64_fmlsl_lane_highv4sf): Ditto.
20842 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20843 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20844 * config/aarch64/aarch64-simd.md:
20845 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20846 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20847 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20848 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20849 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20850 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20851 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20852 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20853 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20854 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20855 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20856 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20857 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20858 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20859 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20860 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20861 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20862 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20863 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20864 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20865 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20866 (vfmlsl_low_u32): Ditto.
20867 (vfmlalq_low_u32): Ditto.
20868 (vfmlslq_low_u32): Ditto.
20869 (vfmlal_high_u32): Ditto.
20870 (vfmlsl_high_u32): Ditto.
20871 (vfmlalq_high_u32): Ditto.
20872 (vfmlslq_high_u32): Ditto.
20873 (vfmlal_lane_low_u32): Ditto.
20874 (vfmlsl_lane_low_u32): Ditto.
20875 (vfmlal_laneq_low_u32): Ditto.
20876 (vfmlsl_laneq_low_u32): Ditto.
20877 (vfmlalq_lane_low_u32): Ditto.
20878 (vfmlslq_lane_low_u32): Ditto.
20879 (vfmlalq_laneq_low_u32): Ditto.
20880 (vfmlslq_laneq_low_u32): Ditto.
20881 (vfmlal_lane_high_u32): Ditto.
20882 (vfmlsl_lane_high_u32): Ditto.
20883 (vfmlal_laneq_high_u32): Ditto.
20884 (vfmlsl_laneq_high_u32): Ditto.
20885 (vfmlalq_lane_high_u32): Ditto.
20886 (vfmlslq_lane_high_u32): Ditto.
20887 (vfmlalq_laneq_high_u32): Ditto.
20888 (vfmlslq_laneq_high_u32): Ditto.
20889 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20890 (AARCH64_FL_FOR_ARCH8_4): New.
20891 (AARCH64_ISA_F16FML): New ISA flag.
20892 (TARGET_F16FML): New feature flag for fp16fml.
20893 (doc/invoke.texi): Document new fp16fml option.
20895 2018-01-10 Michael Collison <michael.collison@arm.com>
20897 * config/aarch64/aarch64-builtins.c:
20898 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20899 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20900 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20901 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20902 (AARCH64_ISA_SHA3): New ISA flag.
20903 (TARGET_SHA3): New feature flag for sha3.
20904 * config/aarch64/iterators.md (sha512_op): New int attribute.
20905 (CRYPTO_SHA512): New int iterator.
20906 (UNSPEC_SHA512H): New unspec.
20907 (UNSPEC_SHA512H2): Ditto.
20908 (UNSPEC_SHA512SU0): Ditto.
20909 (UNSPEC_SHA512SU1): Ditto.
20910 * config/aarch64/aarch64-simd-builtins.def
20911 (aarch64_crypto_sha512hqv2di): New builtin.
20912 (aarch64_crypto_sha512h2qv2di): Ditto.
20913 (aarch64_crypto_sha512su0qv2di): Ditto.
20914 (aarch64_crypto_sha512su1qv2di): Ditto.
20915 (aarch64_eor3qv8hi): Ditto.
20916 (aarch64_rax1qv2di): Ditto.
20917 (aarch64_xarqv2di): Ditto.
20918 (aarch64_bcaxqv8hi): Ditto.
20919 * config/aarch64/aarch64-simd.md:
20920 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20921 (aarch64_crypto_sha512su0qv2di): Ditto.
20922 (aarch64_crypto_sha512su1qv2di): Ditto.
20923 (aarch64_eor3qv8hi): Ditto.
20924 (aarch64_rax1qv2di): Ditto.
20925 (aarch64_xarqv2di): Ditto.
20926 (aarch64_bcaxqv8hi): Ditto.
20927 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20928 (vsha512h2q_u64): Ditto.
20929 (vsha512su0q_u64): Ditto.
20930 (vsha512su1q_u64): Ditto.
20931 (veor3q_u16): Ditto.
20932 (vrax1q_u64): Ditto.
20933 (vxarq_u64): Ditto.
20934 (vbcaxq_u16): Ditto.
20935 * config/arm/types.md (crypto_sha512): New type attribute.
20936 (crypto_sha3): Ditto.
20937 (doc/invoke.texi): Document new sha3 option.
20939 2018-01-10 Michael Collison <michael.collison@arm.com>
20941 * config/aarch64/aarch64-builtins.c:
20942 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20943 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20944 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20945 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20946 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20947 (AARCH64_ISA_SM4): New ISA flag.
20948 (TARGET_SM4): New feature flag for sm4.
20949 * config/aarch64/aarch64-simd-builtins.def
20950 (aarch64_sm3ss1qv4si): Ditto.
20951 (aarch64_sm3tt1aq4si): Ditto.
20952 (aarch64_sm3tt1bq4si): Ditto.
20953 (aarch64_sm3tt2aq4si): Ditto.
20954 (aarch64_sm3tt2bq4si): Ditto.
20955 (aarch64_sm3partw1qv4si): Ditto.
20956 (aarch64_sm3partw2qv4si): Ditto.
20957 (aarch64_sm4eqv4si): Ditto.
20958 (aarch64_sm4ekeyqv4si): Ditto.
20959 * config/aarch64/aarch64-simd.md:
20960 (aarch64_sm3ss1qv4si): Ditto.
20961 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20962 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20963 (aarch64_sm4eqv4si): Ditto.
20964 (aarch64_sm4ekeyqv4si): Ditto.
20965 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20966 (sm3part_op): Ditto.
20967 (CRYPTO_SM3TT): Ditto.
20968 (CRYPTO_SM3PART): Ditto.
20969 (UNSPEC_SM3SS1): New unspec.
20970 (UNSPEC_SM3TT1A): Ditto.
20971 (UNSPEC_SM3TT1B): Ditto.
20972 (UNSPEC_SM3TT2A): Ditto.
20973 (UNSPEC_SM3TT2B): Ditto.
20974 (UNSPEC_SM3PARTW1): Ditto.
20975 (UNSPEC_SM3PARTW2): Ditto.
20976 (UNSPEC_SM4E): Ditto.
20977 (UNSPEC_SM4EKEY): Ditto.
20978 * config/aarch64/constraints.md (Ui2): New constraint.
20979 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20980 * config/arm/types.md (crypto_sm3): New type attribute.
20981 (crypto_sm4): Ditto.
20982 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20983 (vsm3tt1aq_u32): Ditto.
20984 (vsm3tt1bq_u32): Ditto.
20985 (vsm3tt2aq_u32): Ditto.
20986 (vsm3tt2bq_u32): Ditto.
20987 (vsm3partw1q_u32): Ditto.
20988 (vsm3partw2q_u32): Ditto.
20989 (vsm4eq_u32): Ditto.
20990 (vsm4ekeyq_u32): Ditto.
20991 (doc/invoke.texi): Document new sm4 option.
20993 2018-01-10 Michael Collison <michael.collison@arm.com>
20995 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20996 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20997 (AARCH64_FL_FOR_ARCH8_4): New.
20998 (AARCH64_FL_V8_4): New flag.
20999 (doc/invoke.texi): Document new armv8.4-a option.
21001 2018-01-10 Michael Collison <michael.collison@arm.com>
21003 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21004 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21005 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21006 * config/aarch64/aarch64-option-extension.def: Add
21007 AARCH64_OPT_EXTENSION of 'sha2'.
21008 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21009 (crypto): Disable sha2 and aes if crypto disabled.
21010 (crypto): Enable aes and sha2 if enabled.
21011 (simd): Disable sha2 and aes if simd disabled.
21012 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21014 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21015 (TARGET_SHA2): New feature flag for sha2.
21016 (TARGET_AES): New feature flag for aes.
21017 * config/aarch64/aarch64-simd.md:
21018 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21019 conditional on TARGET_AES.
21020 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21021 (aarch64_crypto_sha1hsi): Make pattern conditional
21023 (aarch64_crypto_sha1hv4si): Ditto.
21024 (aarch64_be_crypto_sha1hv4si): Ditto.
21025 (aarch64_crypto_sha1su1v4si): Ditto.
21026 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21027 (aarch64_crypto_sha1su0v4si): Ditto.
21028 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21029 (aarch64_crypto_sha256su0v4si): Ditto.
21030 (aarch64_crypto_sha256su1v4si): Ditto.
21031 (doc/invoke.texi): Document new aes and sha2 options.
21033 2018-01-10 Martin Sebor <msebor@redhat.com>
21035 PR tree-optimization/83781
21036 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21039 2018-01-11 Martin Sebor <msebor@gmail.com>
21040 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21042 PR tree-optimization/83501
21043 PR tree-optimization/81703
21045 * tree-ssa-strlen.c (get_string_cst): Rename...
21046 (get_string_len): ...to this. Handle global constants.
21047 (handle_char_store): Adjust.
21049 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21050 Jim Wilson <jimw@sifive.com>
21052 * config/riscv/riscv-protos.h (riscv_output_return): New.
21053 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21054 (riscv_attribute_table, riscv_output_return),
21055 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21056 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21057 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21059 (riscv_expand_prologue): Add early return for naked function.
21060 (riscv_expand_epilogue): Likewise.
21061 (riscv_function_ok_for_sibcall): Return false for naked function.
21062 (riscv_set_current_function): New.
21063 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21064 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21065 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21066 * doc/extend.texi (RISC-V Function Attributes): New.
21068 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21070 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21071 check for 128-bit long double before checking TCmode.
21072 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21073 128-bit long doubles before checking TFmode or TCmode.
21074 (FLOAT128_IBM_P): Likewise.
21076 2018-01-10 Martin Sebor <msebor@redhat.com>
21078 PR tree-optimization/83671
21079 * builtins.c (c_strlen): Unconditionally return zero for the empty
21081 Use -Warray-bounds for warnings.
21082 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21083 for non-constant array indices with COMPONENT_REF, arrays of
21084 arrays, and pointers to arrays.
21085 (gimple_fold_builtin_strlen): Determine and set length range for
21086 non-constant character arrays.
21088 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21090 PR middle-end/81897
21091 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21094 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21096 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21098 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21101 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21102 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21103 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21104 indexed_or_indirect_operand predicate.
21105 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21106 (*vsx_le_perm_load_v8hi): Likewise.
21107 (*vsx_le_perm_load_v16qi): Likewise.
21108 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21109 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21110 (*vsx_le_perm_store_v8hi): Likewise.
21111 (*vsx_le_perm_store_v16qi): Likewise.
21112 (eight unnamed splitters): Likewise.
21114 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21116 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21117 * config/rs6000/emmintrin.h: Likewise.
21118 * config/rs6000/mmintrin.h: Likewise.
21119 * config/rs6000/xmmintrin.h: Likewise.
21121 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21124 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21126 * tree.c (tree_nop_conversion): Return true for location wrapper
21128 (maybe_wrap_with_location): New function.
21129 (selftest::check_strip_nops): New function.
21130 (selftest::test_location_wrappers): New function.
21131 (selftest::tree_c_tests): Call it.
21132 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21133 (maybe_wrap_with_location): New decl.
21134 (EXPR_LOCATION_WRAPPER_P): New macro.
21135 (location_wrapper_p): New inline function.
21136 (tree_strip_any_location_wrapper): New inline function.
21138 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21141 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21142 stack_realign_offset for the largest alignment of stack slot
21144 (ix86_find_max_used_stack_alignment): New function.
21145 (ix86_finalize_stack_frame_flags): Use it. Set
21146 max_used_stack_alignment if we don't realign stack.
21147 * config/i386/i386.h (machine_function): Add
21148 max_used_stack_alignment.
21150 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21152 * config/arm/arm.opt (-mbranch-cost): New option.
21153 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21156 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21159 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21160 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21162 2018-01-10 Richard Biener <rguenther@suse.de>
21165 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21166 early out so it also covers the case where we have a non-NULL
21169 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21171 PR tree-optimization/83753
21172 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21173 for non-strided grouped accesses if the number of elements is 1.
21175 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21178 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21179 * i386.h (TARGET_USE_GATHER): Define.
21180 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21182 2018-01-10 Martin Liska <mliska@suse.cz>
21185 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21186 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21188 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21189 CLEANUP_NO_PARTITIONING is not set.
21191 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21193 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21194 for vectors, as a partial revert of r254296.
21195 * rtl.h (const_vec_p): Delete.
21196 (const_vec_duplicate_p): Don't test for vector CONSTs.
21197 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21198 * expmed.c (make_tree): Likewise.
21201 * common.md (E, F): Use CONSTANT_P instead of checking for
21203 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21204 checking for CONST_VECTOR.
21206 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21208 PR middle-end/83575
21209 * predict.c (force_edge_cold): Handle in more sane way edges
21210 with no prediction.
21212 2018-01-09 Carl Love <cel@us.ibm.com>
21214 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21216 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21217 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21218 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21219 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21220 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21221 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21222 * config/rs6000/rs6000-protos.h: Add extern defition for
21223 rs6000_generate_float2_double_code.
21224 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21226 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21227 (float2_v2df): Add define_expand.
21229 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21232 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21233 op_mode in the force_to_mode call.
21235 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21237 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21238 instead of checking each element individually.
21239 (aarch64_evpc_uzp): Likewise.
21240 (aarch64_evpc_zip): Likewise.
21241 (aarch64_evpc_ext): Likewise.
21242 (aarch64_evpc_rev): Likewise.
21243 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21244 instead of checking each element individually. Return true without
21246 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21247 whether all selected elements come from the same input, instead of
21248 checking each element individually. Remove calls to gen_rtx_REG,
21249 start_sequence and end_sequence and instead assert that no rtl is
21252 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21254 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21255 order of HIGH and CONST checks.
21257 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21259 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21260 if the destination isn't an SSA_NAME.
21262 2018-01-09 Richard Biener <rguenther@suse.de>
21264 PR tree-optimization/83668
21265 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21267 (canonicalize_loop_form): ... here, renamed from ...
21268 (canonicalize_loop_closed_ssa_form): ... this and amended to
21269 swap successor edges for loop exit blocks to make us use
21270 the RPO order we need for initial schedule generation.
21272 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21274 PR tree-optimization/64811
21275 * match.pd: When optimizing comparisons with Inf, avoid
21276 introducing or losing exceptions from comparisons with NaN.
21278 2018-01-09 Martin Liska <mliska@suse.cz>
21281 * asan.c (shadow_mem_size): Add gcc_assert.
21283 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21285 Don't save registers in main().
21288 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21289 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21290 * config/avr/avr.c (avr_set_current_function): Don't error if
21291 naked, OS_task or OS_main are specified at the same time.
21292 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21294 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21296 * common/config/avr/avr-common.c (avr_option_optimization_table):
21297 Switch on -mmain-is-OS_task for optimizing compilations.
21299 2018-01-09 Richard Biener <rguenther@suse.de>
21301 PR tree-optimization/83572
21302 * graphite.c: Include cfganal.h.
21303 (graphite_transform_loops): Connect infinite loops to exit
21304 and remove fake edges at the end.
21306 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21308 * ipa-inline.c (edge_badness): Revert accidental checkin.
21310 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21313 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21314 symbols; not inline clones.
21316 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21319 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21320 hard registers. Formatting fixes.
21322 PR preprocessor/83722
21323 * gcc.c (try_generate_repro): Pass
21324 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21325 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21328 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21329 Kito Cheng <kito.cheng@gmail.com>
21331 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21332 (riscv_leaf_function_p): Delete.
21333 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21335 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21337 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21339 (do_ifelse): New function.
21340 (do_isel): New function.
21341 (do_sub3): New function.
21342 (do_add3): New function.
21343 (do_load_mask_compare): New function.
21344 (do_overlap_load_compare): New function.
21345 (expand_compare_loop): New function.
21346 (expand_block_compare): Call expand_compare_loop() when appropriate.
21347 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21348 option description.
21349 (-mblock-compare-inline-loop-limit): New option.
21351 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21354 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21355 Reverse order of second and third operands in first alternative.
21356 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21357 of first and second elements in UNSPEC_VPERMR vector.
21358 (altivec_expand_vec_perm_le): Likewise.
21360 2018-01-08 Jeff Law <law@redhat.com>
21362 PR rtl-optimizatin/81308
21363 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21364 (process_switch): If group_case_labels makes a change, then set
21366 (pass_convert_switch::execute): If a switch is converted, then
21367 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21369 PR rtl-optimization/81308
21370 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21373 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21375 PR target/83663 - Revert r255946
21376 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21377 generation for cases where splatting a value is not useful.
21378 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21379 across a vec_duplicate and a paradoxical subreg forming a vector
21380 mode to a vec_concat.
21382 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21384 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21385 -march=armv8.3-a variants.
21386 * config/arm/t-multilib: Likewise.
21387 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21389 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21391 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21393 (cceq_ior_compare_complement): Give it a name so I can use it, and
21394 change boolean_or_operator predicate to boolean_operator so it can
21395 be used to generate a crand.
21396 (eqne): New code iterator.
21397 (bd/bd_neg): New code_attrs.
21398 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21399 a single define_insn.
21400 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21401 decrement (bdnzt/bdnzf/bdzt/bdzf).
21402 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21403 with the new names of the branch decrement patterns, and added the
21404 names of the branch decrement conditional patterns.
21406 2018-01-08 Richard Biener <rguenther@suse.de>
21408 PR tree-optimization/83563
21409 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21412 2018-01-08 Richard Biener <rguenther@suse.de>
21414 PR middle-end/83713
21415 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21417 2018-01-08 Richard Biener <rguenther@suse.de>
21419 PR tree-optimization/83685
21420 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21421 references to abnormals.
21423 2018-01-08 Richard Biener <rguenther@suse.de>
21426 * dwarf2out.c (output_indirect_strings): Handle empty
21427 skeleton_debug_str_hash.
21428 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21430 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21432 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21433 (emit_store_direct): Likewise.
21434 (arc_trampoline_adjust_address): Likewise.
21435 (arc_asm_trampoline_template): New function.
21436 (arc_initialize_trampoline): Use asm_trampoline_template.
21437 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21438 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21439 * config/arc/arc.md (flush_icache): Delete pattern.
21441 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21443 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21444 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21447 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21450 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21451 by not USED_FOR_TARGET.
21452 (make_pass_resolve_sw_modes): Likewise.
21454 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21456 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21459 2018-01-08 Richard Biener <rguenther@suse.de>
21461 PR middle-end/83580
21462 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21464 2018-01-08 Richard Biener <rguenther@suse.de>
21466 PR middle-end/83517
21467 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21469 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21471 PR middle-end/81897
21472 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21473 basic blocks with a small number of successors.
21474 (convert_control_dep_chain_into_preds): Improve handling of
21476 (dump_predicates): Split apart into...
21477 (dump_pred_chain): ...here...
21478 (dump_pred_info): ...and here.
21479 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21480 (can_chain_union_be_invalidated_p): Improve check for invalidation
21482 (uninit_uses_cannot_happen): Avoid unnecessary if
21483 convert_control_dep_chain_into_preds yielded nothing.
21485 2018-01-06 Martin Sebor <msebor@redhat.com>
21487 PR tree-optimization/83640
21488 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21489 subtracting negative offset from size.
21490 (builtin_access::overlap): Adjust offset bounds of the access to fall
21491 within the size of the object if possible.
21493 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21495 PR rtl-optimization/83699
21496 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21497 extract_bit_field_as_subreg to cases in which the extracted
21498 value is also a vector.
21500 * lra-constraints.c (process_alt_operands): Test for the equivalence
21501 substitutions when detecting a possible reload cycle.
21503 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21506 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21507 by default if flag_selective_schedling{,2}. Formatting fixes.
21509 PR rtl-optimization/83682
21510 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21511 if it has non-VECTOR_MODE element mode.
21512 (vec_duplicate_p): Likewise.
21514 PR middle-end/83694
21515 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21516 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21518 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21521 * config/i386/i386-builtin.def
21522 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21523 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21524 Require also OPTION_MASK_ISA_AVX512F in addition to
21525 OPTION_MASK_ISA_GFNI.
21526 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21527 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21528 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21529 to OPTION_MASK_ISA_GFNI.
21530 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21531 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21532 OPTION_MASK_ISA_AVX512BW.
21533 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21534 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21535 addition to OPTION_MASK_ISA_GFNI.
21536 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21537 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21538 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21539 to OPTION_MASK_ISA_GFNI.
21540 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21541 a requirement for all ISAs rather than any of them with a few
21543 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21545 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21546 bitmasks to be enabled with 3 exceptions, instead of requiring any
21547 enabled ISA with lots of exceptions.
21548 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21549 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21550 Change avx512bw in isa attribute to avx512f.
21551 * config/i386/sgxintrin.h: Add license boilerplate.
21552 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21553 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21554 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21555 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21557 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21558 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21559 temporarily sse2 rather than sse if not enabled already.
21562 * config/i386/sse.md (VI248_VLBW): Rename to ...
21563 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21564 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21565 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21566 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21567 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21568 mode iterator instead of VI248_VLBW.
21570 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21572 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21573 (record_modified): Skip clobbers; add debug output.
21574 (param_change_prob): Use sreal frequencies.
21576 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21578 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21579 punt for user-aligned variables.
21581 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21583 * tree-chrec.c (chrec_contains_symbols): Return true for
21586 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21589 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21590 of (x|y) == x for BICS pattern.
21592 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21594 PR tree-optimization/83605
21595 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21596 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21599 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21601 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21602 * config/epiphany/rtems.h: New file.
21604 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21605 Uros Bizjak <ubizjak@gmail.com>
21608 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21609 QIreg_operand instead of register_operand predicate.
21610 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21611 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21612 comments instead of -fmitigate[-_]rop.
21614 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21617 * cgraphunit.c (symbol_table::compile): Switch to text_section
21618 before calling assembly_start debug hook.
21619 * run-rtl-passes.c (run_rtl_passes): Likewise.
21622 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21624 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21625 range_int_cst_p rather than !symbolic_range_p before calling
21626 extract_range_from_multiplicative_op_1.
21628 2018-01-04 Jeff Law <law@redhat.com>
21630 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21631 redundant test in assertion.
21633 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21635 * doc/rtl.texi: Document machine_mode wrapper classes.
21637 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21639 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21640 using tree_to_uhwi.
21642 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21644 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21645 the VEC_PERM_EXPR fold to fail.
21647 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21650 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21651 to switched_sections.
21653 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21656 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21657 test for d.testing.
21659 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21662 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21663 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21665 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21668 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21669 is BLKmode and bitpos not zero or mode change is needed.
21671 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21674 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21677 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21680 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21681 instead of MULT rtx. Update all corresponding splitters.
21682 (*saddl_se): Ditto.
21683 (*ssub<modesuffix>): Ditto.
21684 (*ssubl_se): Ditto.
21685 (*cmp_sadd_di): Update split patterns.
21686 (*cmp_sadd_si): Ditto.
21687 (*cmp_sadd_sidi): Ditto.
21688 (*cmp_ssub_di): Ditto.
21689 (*cmp_ssub_si): Ditto.
21690 (*cmp_ssub_sidi): Ditto.
21691 * config/alpha/predicates.md (const23_operand): New predicate.
21692 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21693 Look for ASHIFT, not MULT inner operand.
21694 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21696 2018-01-04 Martin Liska <mliska@suse.cz>
21698 PR gcov-profile/83669
21699 * gcov.c (output_intermediate_file): Add version to intermediate
21701 * doc/gcov.texi: Document new field 'version' in intermediate
21702 file format. Fix location of '-k' option of gcov command.
21704 2018-01-04 Martin Liska <mliska@suse.cz>
21707 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21709 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21711 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21713 2018-01-03 Martin Sebor <msebor@redhat.com>
21715 PR tree-optimization/83655
21716 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21717 checking calls with invalid arguments.
21719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21721 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21722 (vectorizable_mask_load_store): Delete.
21723 (vectorizable_call): Return false for masked loads and stores.
21724 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21725 instead of gimple_assign_rhs1.
21726 (vectorizable_load): Handle IFN_MASK_LOAD.
21727 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21731 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21733 (vectorizable_mask_load_store): ...here.
21734 (vectorizable_load): ...and here.
21736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21738 * tree-vect-stmts.c (vect_build_all_ones_mask)
21739 (vect_build_zero_merge_argument): New functions, split out from...
21740 (vectorizable_load): ...here.
21742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21744 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21746 (vectorizable_mask_load_store): ...here.
21747 (vectorizable_store): ...and here.
21749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21751 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21753 (vectorizable_mask_load_store): ...here.
21755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21757 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21758 (vect_model_store_cost): Take a vec_load_store_type instead of a
21760 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21761 (vect_model_store_cost): Take a vec_load_store_type instead of a
21763 (vectorizable_mask_load_store): Update accordingly.
21764 (vectorizable_store): Likewise.
21765 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21769 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21770 IFN_MASK_LOAD calls here rather than...
21771 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21774 Alan Hayward <alan.hayward@arm.com>
21775 David Sherwood <david.sherwood@arm.com>
21777 * expmed.c (extract_bit_field_1): For vector extracts,
21778 fall back to extract_bit_field_as_subreg if vec_extract
21781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21782 Alan Hayward <alan.hayward@arm.com>
21783 David Sherwood <david.sherwood@arm.com>
21785 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21786 they are variable or constant sized.
21787 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21788 slots for constant-sized data.
21790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21791 Alan Hayward <alan.hayward@arm.com>
21792 David Sherwood <david.sherwood@arm.com>
21794 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21795 handling COND_EXPRs with boolean comparisons, try to find a better
21796 basis for the mask type than the boolean itself.
21798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21800 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21801 is calculated and how it can be overridden.
21802 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21803 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21805 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21809 Alan Hayward <alan.hayward@arm.com>
21810 David Sherwood <david.sherwood@arm.com>
21812 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21813 Remove the mode argument.
21814 (aarch64_simd_valid_immediate): Remove the mode and inverse
21816 * config/aarch64/iterators.md (bitsize): New iterator.
21817 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21818 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21819 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21820 aarch64_simd_valid_immediate.
21821 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21822 (aarch64_reg_or_bic_imm): Likewise.
21823 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21824 with an insn_type enum and msl with a modifier_type enum.
21825 Replace element_width with a scalar_mode. Change the shift
21826 to unsigned int. Add constructors for scalar_float_mode and
21827 scalar_int_mode elements.
21828 (aarch64_vect_float_const_representable_p): Delete.
21829 (aarch64_can_const_movi_rtx_p)
21830 (aarch64_simd_scalar_immediate_valid_for_move)
21831 (aarch64_simd_make_constant): Update call to
21832 aarch64_simd_valid_immediate.
21833 (aarch64_advsimd_valid_immediate_hs): New function.
21834 (aarch64_advsimd_valid_immediate): Likewise.
21835 (aarch64_simd_valid_immediate): Remove mode and inverse
21836 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21837 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21838 and aarch64_float_const_representable_p on the result.
21839 (aarch64_output_simd_mov_immediate): Remove mode argument.
21840 Update call to aarch64_simd_valid_immediate and use of
21841 simd_immediate_info.
21842 (aarch64_output_scalar_simd_mov_immediate): Update call
21845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21846 Alan Hayward <alan.hayward@arm.com>
21847 David Sherwood <david.sherwood@arm.com>
21849 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21850 (mode_nunits): Likewise CONST_MODE_NUNITS.
21851 * machmode.def (ADJUST_NUNITS): Document.
21852 * genmodes.c (mode_data::need_nunits_adj): New field.
21853 (blank_mode): Update accordingly.
21854 (adj_nunits): New variable.
21855 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21857 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21858 listed in adj_nunits.
21859 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21860 listed in adj_nunits. Don't emit case statements for such modes.
21861 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21862 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21863 nothing if adj_nunits is nonnull.
21864 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21865 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21866 (emit_mode_fbit): Update use of print_maybe_const_decl.
21867 (emit_move_size): Likewise. Treat the array as non-const
21869 (emit_mode_adjustments): Handle adj_nunits.
21871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21873 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21874 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21875 (VECTOR_MODES): Use it.
21876 (make_vector_modes): Take the prefix as an argument.
21878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21879 Alan Hayward <alan.hayward@arm.com>
21880 David Sherwood <david.sherwood@arm.com>
21882 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21883 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21884 for MODE_VECTOR_BOOL.
21885 * machmode.def (VECTOR_BOOL_MODE): Document.
21886 * genmodes.c (VECTOR_BOOL_MODE): New macro.
21887 (make_vector_bool_mode): New function.
21888 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21890 * lto-streamer-in.c (lto_input_mode_table): Likewise.
21891 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21893 * stor-layout.c (int_mode_for_mode): Likewise.
21894 * tree.c (build_vector_type_for_mode): Likewise.
21895 * varasm.c (output_constant_pool_2): Likewise.
21896 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21897 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
21898 for MODE_VECTOR_BOOL.
21899 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21900 of mode class checks.
21901 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21902 instead of a list of mode class checks.
21903 (expand_vector_scalar_condition): Likewise.
21904 (type_for_widest_vector_mode): Handle BImode as an inner mode.
21906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21907 Alan Hayward <alan.hayward@arm.com>
21908 David Sherwood <david.sherwood@arm.com>
21910 * machmode.h (mode_size): Change from unsigned short to
21912 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21913 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21914 or if measurement_type is not polynomial.
21915 (fixed_size_mode::includes_p): Check for constant-sized modes.
21916 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21917 return a poly_uint16 rather than an unsigned short.
21918 (emit_mode_size): Change the type of mode_size from unsigned short
21919 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
21920 (emit_mode_adjustments): Cope with polynomial vector sizes.
21921 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21923 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21925 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21926 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21927 * caller-save.c (setup_save_areas): Likewise.
21928 (replace_reg_with_saved_mem): Likewise.
21929 * calls.c (emit_library_call_value_1): Likewise.
21930 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21931 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21932 (gen_lowpart_for_combine): Likewise.
21933 * convert.c (convert_to_integer_1): Likewise.
21934 * cse.c (equiv_constant, cse_insn): Likewise.
21935 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21936 (cselib_subst_to_values): Likewise.
21937 * dce.c (word_dce_process_block): Likewise.
21938 * df-problems.c (df_word_lr_mark_ref): Likewise.
21939 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21940 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21941 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21942 (rtl_for_decl_location): Likewise.
21943 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21944 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21945 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21946 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21947 (expand_expr_real_1): Likewise.
21948 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21949 (pad_below): Likewise.
21950 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21951 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21952 * ira.c (get_subreg_tracking_sizes): Likewise.
21953 * ira-build.c (ira_create_allocno_objects): Likewise.
21954 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21955 (ira_sort_regnos_for_alter_reg): Likewise.
21956 * ira-costs.c (record_operand_costs): Likewise.
21957 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21958 (resolve_simple_move): Likewise.
21959 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21960 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21961 (lra_constraints): Likewise.
21962 (CONST_POOL_OK_P): Reject variable-sized modes.
21963 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21964 (add_pseudo_to_slot, lra_spill): Likewise.
21965 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21966 * optabs-query.c (get_best_extraction_insn): Likewise.
21967 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21968 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21969 (expand_mult_highpart, valid_multiword_target_p): Likewise.
21970 * recog.c (offsettable_address_addr_space_p): Likewise.
21971 * regcprop.c (maybe_mode_change): Likewise.
21972 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21973 * regrename.c (build_def_use): Likewise.
21974 * regstat.c (dump_reg_info): Likewise.
21975 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21976 (find_reloads, find_reloads_subreg_address): Likewise.
21977 * reload1.c (eliminate_regs_1): Likewise.
21978 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21979 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21980 (simplify_binary_operation_1, simplify_subreg): Likewise.
21981 * targhooks.c (default_function_arg_padding): Likewise.
21982 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21983 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21984 (verify_gimple_assign_ternary): Likewise.
21985 * tree-inline.c (estimate_move_cost): Likewise.
21986 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21987 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21988 (get_address_cost_ainc): Likewise.
21989 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21990 (vect_supportable_dr_alignment): Likewise.
21991 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21992 (vectorizable_reduction): Likewise.
21993 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21994 (vectorizable_operation, vectorizable_load): Likewise.
21995 * tree.c (build_same_sized_truth_vector_type): Likewise.
21996 * valtrack.c (cleanup_auto_inc_dec): Likewise.
21997 * var-tracking.c (emit_note_insn_var_location): Likewise.
21998 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21999 (ADDR_VEC_ALIGN): Likewise.
22001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22002 Alan Hayward <alan.hayward@arm.com>
22003 David Sherwood <david.sherwood@arm.com>
22005 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22007 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22008 or if measurement_type is polynomial.
22009 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22010 * combine.c (make_extraction): Likewise.
22011 * dse.c (find_shift_sequence): Likewise.
22012 * dwarf2out.c (mem_loc_descriptor): Likewise.
22013 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22014 (extract_bit_field, extract_low_bits): Likewise.
22015 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22016 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22017 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22018 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22019 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22020 * reload.c (find_reloads): Likewise.
22021 * reload1.c (alter_reg): Likewise.
22022 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22023 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22024 * tree-if-conv.c (predicate_mem_writes): Likewise.
22025 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22026 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22027 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22028 * valtrack.c (dead_debug_insert_temp): Likewise.
22029 * varasm.c (mergeable_constant_section): Likewise.
22030 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22033 Alan Hayward <alan.hayward@arm.com>
22034 David Sherwood <david.sherwood@arm.com>
22036 * expr.c (expand_assignment): Cope with polynomial mode sizes
22037 when assigning to a CONCAT.
22039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22040 Alan Hayward <alan.hayward@arm.com>
22041 David Sherwood <david.sherwood@arm.com>
22043 * machmode.h (mode_precision): Change from unsigned short to
22045 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22047 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22048 or if measurement_type is not polynomial.
22049 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22050 in which the mode is already known to be a scalar_int_mode.
22051 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22052 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22054 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22055 for GET_MODE_PRECISION.
22056 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22057 for GET_MODE_PRECISION.
22058 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22060 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22061 (expand_field_assignment, make_extraction): Likewise.
22062 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22063 (get_last_value): Likewise.
22064 * convert.c (convert_to_integer_1): Likewise.
22065 * cse.c (cse_insn): Likewise.
22066 * expr.c (expand_expr_real_1): Likewise.
22067 * lra-constraints.c (simplify_operand_subreg): Likewise.
22068 * optabs-query.c (can_atomic_load_p): Likewise.
22069 * optabs.c (expand_atomic_load): Likewise.
22070 (expand_atomic_store): Likewise.
22071 * ree.c (combine_reaching_defs): Likewise.
22072 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22073 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22074 * tree.h (type_has_mode_precision_p): Likewise.
22075 * ubsan.c (instrument_si_overflow): Likewise.
22077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22078 Alan Hayward <alan.hayward@arm.com>
22079 David Sherwood <david.sherwood@arm.com>
22081 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22082 polynomial numbers of units.
22083 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22084 (valid_vector_subparts_p): New function.
22085 (build_vector_type): Remove temporary shim and take the number
22086 of units as a poly_uint64 rather than an int.
22087 (build_opaque_vector_type): Take the number of units as a
22088 poly_uint64 rather than an int.
22089 * tree.c (build_vector_from_ctor): Handle polynomial
22090 TYPE_VECTOR_SUBPARTS.
22091 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22092 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22093 (build_vector_from_val): If the number of units is variable,
22094 use build_vec_duplicate_cst for constant operands and
22095 VEC_DUPLICATE_EXPR otherwise.
22096 (make_vector_type): Remove temporary is_constant ().
22097 (build_vector_type, build_opaque_vector_type): Take the number of
22098 units as a poly_uint64 rather than an int.
22099 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22101 * cfgexpand.c (expand_debug_expr): Likewise.
22102 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22103 (store_constructor, expand_expr_real_1): Likewise.
22104 (const_scalar_mask_from_tree): Likewise.
22105 * fold-const-call.c (fold_const_reduction): Likewise.
22106 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22107 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22108 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22109 (fold_relational_const): Likewise.
22110 (native_interpret_vector): Likewise. Change the size from an
22111 int to an unsigned int.
22112 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22113 TYPE_VECTOR_SUBPARTS.
22114 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22115 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22116 duplicating a non-constant operand into a variable-length vector.
22117 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22118 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22119 * ipa-icf.c (sem_variable::equals): Likewise.
22120 * match.pd: Likewise.
22121 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22122 * print-tree.c (print_node): Likewise.
22123 * stor-layout.c (layout_type): Likewise.
22124 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22125 * tree-cfg.c (verify_gimple_comparison): Likewise.
22126 (verify_gimple_assign_binary): Likewise.
22127 (verify_gimple_assign_ternary): Likewise.
22128 (verify_gimple_assign_single): Likewise.
22129 * tree-pretty-print.c (dump_generic_node): Likewise.
22130 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22131 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22132 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22133 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22134 (vect_shift_permute_load_chain): Likewise.
22135 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22136 (expand_vector_condition, optimize_vector_constructor): Likewise.
22137 (lower_vec_perm, get_compute_type): Likewise.
22138 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22139 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22140 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22141 (vect_recog_mask_conversion_pattern): Likewise.
22142 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22143 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22144 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22145 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22146 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22147 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22148 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22149 (supportable_widening_operation): Likewise.
22150 (supportable_narrowing_operation): Likewise.
22151 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22153 * varasm.c (output_constant): Likewise.
22155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22156 Alan Hayward <alan.hayward@arm.com>
22157 David Sherwood <david.sherwood@arm.com>
22159 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22160 so that both the length == 3 and length != 3 cases set up their
22161 own permute vectors. Add comments explaining why we know the
22162 number of elements is constant.
22163 (vect_permute_load_chain): Likewise.
22165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22166 Alan Hayward <alan.hayward@arm.com>
22167 David Sherwood <david.sherwood@arm.com>
22169 * machmode.h (mode_nunits): Change from unsigned char to
22171 (ONLY_FIXED_SIZE_MODES): New macro.
22172 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22173 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22174 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22176 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22177 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22178 or if measurement_type is not polynomial.
22179 * genmodes.c (ZERO_COEFFS): New macro.
22180 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22182 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22183 Use ZERO_COEFFS when emitting initializers.
22184 * data-streamer.h (bp_pack_poly_value): New function.
22185 (bp_unpack_poly_value): Likewise.
22186 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22187 for GET_MODE_NUNITS.
22188 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22189 for GET_MODE_NUNITS.
22190 * tree.c (make_vector_type): Remove temporary shim and make
22191 the real function take the number of units as a poly_uint64
22192 rather than an int.
22193 (build_vector_type_for_mode): Handle polynomial nunits.
22194 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22195 * emit-rtl.c (const_vec_series_p_1): Likewise.
22196 (gen_rtx_CONST_VECTOR): Likewise.
22197 * fold-const.c (test_vec_duplicate_folding): Likewise.
22198 * genrecog.c (validate_pattern): Likewise.
22199 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22200 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22201 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22202 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22203 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22204 * rtlanal.c (subreg_get_info): Likewise.
22205 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22206 (vect_grouped_load_supported): Likewise.
22207 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22208 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22209 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22210 (simplify_const_unary_operation, simplify_binary_operation_1)
22211 (simplify_const_binary_operation, simplify_ternary_operation)
22212 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22213 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22214 instead of CONST_VECTOR_NUNITS.
22215 * varasm.c (output_constant_pool_2): Likewise.
22216 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22217 explicit-encoded elements in the XVEC for variable-length vectors.
22219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22221 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22224 Alan Hayward <alan.hayward@arm.com>
22225 David Sherwood <david.sherwood@arm.com>
22227 * coretypes.h (fixed_size_mode): Declare.
22228 (fixed_size_mode_pod): New typedef.
22229 * builtins.h (target_builtins::x_apply_args_mode)
22230 (target_builtins::x_apply_result_mode): Change type to
22231 fixed_size_mode_pod.
22232 * builtins.c (apply_args_size, apply_result_size, result_vector)
22233 (expand_builtin_apply_args_1, expand_builtin_apply)
22234 (expand_builtin_return): Update accordingly.
22236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22238 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22239 * cselib.c (cselib_hash_rtx): Likewise.
22240 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22241 CONST_VECTOR encoding.
22243 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22244 Jeff Law <law@redhat.com>
22247 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22248 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22249 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22250 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22253 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22254 explicitly probe *sp in a noreturn function if there were any callee
22255 register saves or frame pointer is needed.
22257 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22260 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22261 BLKmode for ternary, binary or unary expressions.
22264 * var-tracking.c (delete_vta_debug_insn): New inline function.
22265 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22266 insns from get_insns () to NULL instead of each bb separately.
22267 Use delete_vta_debug_insn. No longer static.
22268 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22269 delete_vta_debug_insns callers.
22270 * rtl.h (delete_vta_debug_insns): Declare.
22271 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22272 instead of variable_tracking_main.
22274 2018-01-03 Martin Sebor <msebor@redhat.com>
22276 PR tree-optimization/83603
22277 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22278 arguments past the endof the argument list in functions declared
22279 without a prototype.
22280 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22281 Avoid checking when arguments are null.
22283 2018-01-03 Martin Sebor <msebor@redhat.com>
22286 * doc/extend.texi (attribute const): Fix a typo.
22287 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22288 issuing -Wsuggest-attribute for void functions.
22290 2018-01-03 Martin Sebor <msebor@redhat.com>
22292 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22293 offset_int::from instead of wide_int::to_shwi.
22294 (maybe_diag_overlap): Remove assertion.
22295 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22296 * gimple-ssa-sprintf.c (format_directive): Same.
22297 (parse_directive): Same.
22298 (sprintf_dom_walker::compute_format_length): Same.
22299 (try_substitute_return_value): Same.
22301 2018-01-03 Jeff Law <law@redhat.com>
22303 PR middle-end/83654
22304 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22305 non-constant residual for zero at runtime and avoid probing in
22306 that case. Reorganize code for trailing problem to mirror handling
22309 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22311 PR tree-optimization/83501
22312 * tree-ssa-strlen.c (get_string_cst): New.
22313 (handle_char_store): Call get_string_cst.
22315 2018-01-03 Martin Liska <mliska@suse.cz>
22317 PR tree-optimization/83593
22318 * tree-ssa-strlen.c: Include tree-cfg.h.
22319 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22320 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22321 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22323 (strlen_dom_walker::before_dom_children): Call
22324 gimple_purge_dead_eh_edges. Dump tranformation with details
22326 (strlen_dom_walker::before_dom_children): Update call by adding
22327 new argument cleanup_eh.
22328 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22330 2018-01-03 Martin Liska <mliska@suse.cz>
22333 * cif-code.def (VARIADIC_THUNK): New enum value.
22334 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22337 2018-01-03 Jan Beulich <jbeulich@suse.com>
22339 * sse.md (mov<mode>_internal): Tighten condition for when to use
22340 vmovdqu<ssescalarsize> for TI and OI modes.
22342 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22344 Update copyright years.
22346 2018-01-03 Martin Liska <mliska@suse.cz>
22349 * ipa-visibility.c (function_and_variable_visibility): Skip
22350 functions with noipa attribure.
22352 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22354 * gcc.c (process_command): Update copyright notice dates.
22355 * gcov-dump.c (print_version): Ditto.
22356 * gcov.c (print_version): Ditto.
22357 * gcov-tool.c (print_version): Ditto.
22358 * gengtype.c (create_file): Ditto.
22359 * doc/cpp.texi: Bump @copying's copyright year.
22360 * doc/cppinternals.texi: Ditto.
22361 * doc/gcc.texi: Ditto.
22362 * doc/gccint.texi: Ditto.
22363 * doc/gcov.texi: Ditto.
22364 * doc/install.texi: Ditto.
22365 * doc/invoke.texi: Ditto.
22367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22369 * vector-builder.h (vector_builder::m_full_nelts): Change from
22370 unsigned int to poly_uint64.
22371 (vector_builder::full_nelts): Update prototype accordingly.
22372 (vector_builder::new_vector): Likewise.
22373 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22374 (vector_builder::operator ==): Likewise.
22375 (vector_builder::finalize): Likewise.
22376 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22377 Take the number of elements as a poly_uint64 rather than an
22379 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22380 from unsigned int to poly_uint64.
22381 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22382 (vec_perm_indices::new_vector): Likewise.
22383 (vec_perm_indices::length): Likewise.
22384 (vec_perm_indices::nelts_per_input): Likewise.
22385 (vec_perm_indices::input_nelts): Likewise.
22386 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22387 number of elements per input as a poly_uint64 rather than an
22388 unsigned int. Use the original encoding for variable-length
22389 vectors, rather than clamping each individual element.
22390 For the second and subsequent elements in each pattern,
22391 clamp the step and base before clamping their sum.
22392 (vec_perm_indices::series_p): Handle polynomial element counts.
22393 (vec_perm_indices::all_in_range_p): Likewise.
22394 (vec_perm_indices_to_tree): Likewise.
22395 (vec_perm_indices_to_rtx): Likewise.
22396 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22397 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22398 (tree_vector_builder::new_binary_operation): Handle polynomial
22399 element counts. Return false if we need to know the number
22400 of elements at compile time.
22401 * fold-const.c (fold_vec_perm): Punt if the number of elements
22402 isn't known at compile time.
22404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22406 * vec-perm-indices.h (vec_perm_builder): Change element type
22407 from HOST_WIDE_INT to poly_int64.
22408 (vec_perm_indices::element_type): Update accordingly.
22409 (vec_perm_indices::clamp): Handle polynomial element_types.
22410 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22411 (vec_perm_indices::all_in_range_p): Likewise.
22412 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22414 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22415 polynomial vec_perm_indices element types.
22416 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22417 * fold-const.c (fold_vec_perm): Likewise.
22418 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22419 * tree-vect-generic.c (lower_vec_perm): Likewise.
22420 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22421 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22422 element type to HOST_WIDE_INT.
22424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22425 Alan Hayward <alan.hayward@arm.com>
22426 David Sherwood <david.sherwood@arm.com>
22428 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22429 rather than an int. Use plus_constant.
22430 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22431 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22434 Alan Hayward <alan.hayward@arm.com>
22435 David Sherwood <david.sherwood@arm.com>
22437 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22438 a HOST_WIDE_INT to a poly_int64.
22440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22441 Alan Hayward <alan.hayward@arm.com>
22442 David Sherwood <david.sherwood@arm.com>
22444 * calls.c (load_register_parameters): Cope with polynomial
22445 mode sizes. Require a constant size for BLKmode parameters
22446 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22447 forces a parameter to be padded at the lsb end in order to
22448 fill a complete number of words, require the parameter size
22449 to be ordered wrt UNITS_PER_WORD.
22451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22452 Alan Hayward <alan.hayward@arm.com>
22453 David Sherwood <david.sherwood@arm.com>
22455 * reload1.c (spill_stack_slot_width): Change element type
22456 from unsigned int to poly_uint64_pod.
22457 (alter_reg): Treat mode sizes as polynomial.
22459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22460 Alan Hayward <alan.hayward@arm.com>
22461 David Sherwood <david.sherwood@arm.com>
22463 * reload.c (complex_word_subreg_p): New function.
22464 (reload_inner_reg_of_subreg, push_reload): Use it.
22466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22467 Alan Hayward <alan.hayward@arm.com>
22468 David Sherwood <david.sherwood@arm.com>
22470 * lra-constraints.c (process_alt_operands): Reject matched
22471 operands whose sizes aren't ordered.
22472 (match_reload): Refer to this check here.
22474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22475 Alan Hayward <alan.hayward@arm.com>
22476 David Sherwood <david.sherwood@arm.com>
22478 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22479 that the mode size is in the set {1, 2, 4, 8, 16}.
22481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22482 Alan Hayward <alan.hayward@arm.com>
22483 David Sherwood <david.sherwood@arm.com>
22485 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22486 Use plus_constant instead of gen_rtx_PLUS.
22488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22489 Alan Hayward <alan.hayward@arm.com>
22490 David Sherwood <david.sherwood@arm.com>
22492 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22493 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22494 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22495 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22496 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22497 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22498 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22499 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22500 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22501 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22503 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22504 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22505 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22506 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22507 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22508 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22509 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22510 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22511 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22512 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22514 * expr.c (emit_move_resolve_push): Treat the input and result
22515 of PUSH_ROUNDING as a poly_int64.
22516 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22517 (emit_push_insn): Likewise.
22518 * lra-eliminations.c (mark_not_eliminable): Likewise.
22519 * recog.c (push_operand): Likewise.
22520 * reload1.c (elimination_effects): Likewise.
22521 * rtlanal.c (nonzero_bits1): Likewise.
22522 * calls.c (store_one_arg): Likewise. Require the padding to be
22523 known at compile time.
22525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22526 Alan Hayward <alan.hayward@arm.com>
22527 David Sherwood <david.sherwood@arm.com>
22529 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22530 Use plus_constant instead of gen_rtx_PLUS.
22532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22533 Alan Hayward <alan.hayward@arm.com>
22534 David Sherwood <david.sherwood@arm.com>
22536 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22537 rather than an int.
22539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22540 Alan Hayward <alan.hayward@arm.com>
22541 David Sherwood <david.sherwood@arm.com>
22543 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22544 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22545 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
22552 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22553 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22554 as a poly_uint64 too.
22556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22557 Alan Hayward <alan.hayward@arm.com>
22558 David Sherwood <david.sherwood@arm.com>
22560 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22563 Alan Hayward <alan.hayward@arm.com>
22564 David Sherwood <david.sherwood@arm.com>
22566 * combine.c (can_change_dest_mode): Handle polynomial
22567 REGMODE_NATURAL_SIZE.
22568 * expmed.c (store_bit_field_1): Likewise.
22569 * expr.c (store_constructor): Likewise.
22570 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22571 and polynomial REGMODE_NATURAL_SIZE.
22572 (gen_lowpart_common): Likewise.
22573 * reginfo.c (record_subregs_of_mode): Likewise.
22574 * rtlanal.c (read_modify_subreg_p): Likewise.
22576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22577 Alan Hayward <alan.hayward@arm.com>
22578 David Sherwood <david.sherwood@arm.com>
22580 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22581 numbers of elements.
22583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22584 Alan Hayward <alan.hayward@arm.com>
22585 David Sherwood <david.sherwood@arm.com>
22587 * match.pd: Cope with polynomial numbers of vector elements.
22589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22590 Alan Hayward <alan.hayward@arm.com>
22591 David Sherwood <david.sherwood@arm.com>
22593 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22594 in a POINTER_PLUS_EXPR.
22596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22597 Alan Hayward <alan.hayward@arm.com>
22598 David Sherwood <david.sherwood@arm.com>
22600 * omp-simd-clone.c (simd_clone_subparts): New function.
22601 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22602 (ipa_simd_modify_function_body): Likewise.
22604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22605 Alan Hayward <alan.hayward@arm.com>
22606 David Sherwood <david.sherwood@arm.com>
22608 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22609 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22610 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22611 (expand_vector_condition, vector_element): Likewise.
22612 (subparts_gt): New function.
22613 (get_compute_type): Use subparts_gt.
22614 (count_type_subparts): Delete.
22615 (expand_vector_operations_1): Use subparts_gt instead of
22616 count_type_subparts.
22618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22619 Alan Hayward <alan.hayward@arm.com>
22620 David Sherwood <david.sherwood@arm.com>
22622 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22623 (vect_compile_time_alias): ...this new function. Do the calculation
22624 on poly_ints rather than trees.
22625 (vect_prune_runtime_alias_test_list): Update call accordingly.
22627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22628 Alan Hayward <alan.hayward@arm.com>
22629 David Sherwood <david.sherwood@arm.com>
22631 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22633 (vect_schedule_slp_instance): Likewise.
22635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22636 Alan Hayward <alan.hayward@arm.com>
22637 David Sherwood <david.sherwood@arm.com>
22639 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22640 constant and extern definitions for variable-length vectors.
22641 (vect_get_constant_vectors): Note that the number of units
22642 is known to be constant.
22644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22645 Alan Hayward <alan.hayward@arm.com>
22646 David Sherwood <david.sherwood@arm.com>
22648 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22649 of units as polynomial. Choose between WIDE and NARROW based
22652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22653 Alan Hayward <alan.hayward@arm.com>
22654 David Sherwood <david.sherwood@arm.com>
22656 * tree-vect-stmts.c (simd_clone_subparts): New function.
22657 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22660 Alan Hayward <alan.hayward@arm.com>
22661 David Sherwood <david.sherwood@arm.com>
22663 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22664 vectors as polynomial. Use build_index_vector for
22665 IFN_GOMP_SIMD_LANE.
22667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22668 Alan Hayward <alan.hayward@arm.com>
22669 David Sherwood <david.sherwood@arm.com>
22671 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22672 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22673 for variable-length vectors.
22674 (vectorizable_mask_load_store): Treat the number of units as
22675 polynomial, asserting that it is constant if the condition has
22676 already been enforced.
22677 (vectorizable_store, vectorizable_load): Likewise.
22679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22680 Alan Hayward <alan.hayward@arm.com>
22681 David Sherwood <david.sherwood@arm.com>
22683 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22684 of units as polynomial. Punt if we can't tell at compile time
22685 which vector contains the final result.
22687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22688 Alan Hayward <alan.hayward@arm.com>
22689 David Sherwood <david.sherwood@arm.com>
22691 * tree-vect-loop.c (vectorizable_induction): Treat the number
22692 of units as polynomial. Punt on SLP inductions. Use an integer
22693 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22694 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
22702 * tree.c (build_index_vector): New function.
22703 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22704 of units as polynomial, forcibly converting it to a constant if
22705 vectorizable_reduction has already enforced the condition.
22706 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22707 to create a {1,2,3,...} vector.
22708 (vectorizable_reduction): Treat the number of units as polynomial.
22709 Choose vectype_in based on the largest scalar element size rather
22710 than the smallest number of units. Enforce the restrictions
22713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22714 Alan Hayward <alan.hayward@arm.com>
22715 David Sherwood <david.sherwood@arm.com>
22717 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22718 number of units as polynomial.
22720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22721 Alan Hayward <alan.hayward@arm.com>
22722 David Sherwood <david.sherwood@arm.com>
22724 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22725 * target.def (autovectorize_vector_sizes): Return the vector sizes
22726 by pointer, using vector_sizes rather than a bitmask.
22727 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22728 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22729 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22731 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22732 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22733 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22734 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22735 * omp-general.c (omp_max_vf): Likewise.
22736 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22737 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22738 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22739 * tree-vect-slp.c (vect_slp_bb): Likewise.
22740 * doc/tm.texi: Regenerate.
22741 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22743 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22744 the vector size as a poly_uint64 rather than an unsigned int.
22745 (current_vector_size): Change from an unsigned int to a poly_uint64.
22746 (get_vectype_for_scalar_type): Update accordingly.
22747 * tree.h (build_truth_vector_type): Take the size and number of
22748 units as a poly_uint64 rather than an unsigned int.
22749 (build_vector_type): Add a temporary overload that takes
22750 the number of units as a poly_uint64 rather than an unsigned int.
22751 * tree.c (make_vector_type): Likewise.
22752 (build_truth_vector_type): Take the number of units as a poly_uint64
22753 rather than an unsigned int.
22755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22756 Alan Hayward <alan.hayward@arm.com>
22757 David Sherwood <david.sherwood@arm.com>
22759 * target.def (get_mask_mode): Take the number of units and length
22760 as poly_uint64s rather than unsigned ints.
22761 * targhooks.h (default_get_mask_mode): Update accordingly.
22762 * targhooks.c (default_get_mask_mode): Likewise.
22763 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22764 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22771 * omp-general.c (omp_max_vf): Likewise.
22772 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22773 (expand_omp_simd): Handle polynomial safelen.
22774 * omp-low.c (omplow_simd_context): Add a default constructor.
22775 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22776 (lower_rec_simd_input_clauses): Update accordingly.
22777 (lower_rec_input_clauses): Likewise.
22779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22780 Alan Hayward <alan.hayward@arm.com>
22781 David Sherwood <david.sherwood@arm.com>
22783 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22784 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22785 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22786 (vect_analyze_slp_cost): Likewise.
22787 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22788 (vect_model_load_cost): Likewise.
22790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22791 Alan Hayward <alan.hayward@arm.com>
22792 David Sherwood <david.sherwood@arm.com>
22794 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22795 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22796 from an unsigned int * to a poly_uint64_pod *.
22797 (calculate_unrolling_factor): New function.
22798 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22801 Alan Hayward <alan.hayward@arm.com>
22802 David Sherwood <david.sherwood@arm.com>
22804 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22805 from an unsigned int to a poly_uint64.
22806 (_loop_vec_info::slp_unrolling_factor): Likewise.
22807 (_loop_vec_info::vectorization_factor): Change from an int
22809 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22810 (vect_get_num_vectors): New function.
22811 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22812 (vect_get_num_copies): Use vect_get_num_vectors.
22813 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22814 to an unsigned int *.
22815 (vect_analyze_data_refs): Change min_vf from an int * to a
22817 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22818 than an unsigned HOST_WIDE_INT.
22819 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22820 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22821 to an unsigned int *.
22822 (vect_analyze_data_ref_dependences): Likewise.
22823 (vect_compute_data_ref_alignment): Handle polynomial vf.
22824 (vect_enhance_data_refs_alignment): Likewise.
22825 (vect_prune_runtime_alias_test_list): Likewise.
22826 (vect_shift_permute_load_chain): Likewise.
22827 (vect_supportable_dr_alignment): Likewise.
22828 (dependence_distance_ge_vf): Take the vectorization factor as a
22829 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22830 (vect_analyze_data_refs): Change min_vf from an int * to a
22832 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22833 vfm1 as a poly_uint64 rather than an int. Make the same change
22834 for the returned bound_scalar.
22835 (vect_gen_vector_loop_niters): Handle polynomial vf.
22836 (vect_do_peeling): Likewise. Update call to
22837 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22838 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22840 * tree-vect-loop.c (vect_determine_vectorization_factor)
22841 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22842 (vect_get_known_peeling_cost): Likewise.
22843 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22844 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22845 (vect_transform_loop): Likewise. Use the lowest possible VF when
22846 updating the upper bounds of the loop.
22847 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22848 rather than an int.
22849 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22850 polynomial unroll factors.
22851 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22852 (vect_make_slp_decision): Likewise.
22853 (vect_supported_load_permutation_p): Likewise, and polynomial
22855 (vect_analyze_slp_cost): Handle polynomial vf.
22856 (vect_slp_analyze_node_operations): Likewise.
22857 (vect_slp_analyze_bb_1): Likewise.
22858 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22859 than an unsigned HOST_WIDE_INT.
22860 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22861 (vectorizable_load): Handle polynomial vf.
22862 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22864 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22867 Alan Hayward <alan.hayward@arm.com>
22868 David Sherwood <david.sherwood@arm.com>
22870 * match.pd: Handle bit operations involving three constants
22871 and try to fold one pair.
22873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22875 * tree-vect-loop-manip.c: Include gimple-fold.h.
22876 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22877 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22878 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22879 Add a path that uses a step of VF instead of 1, but disable it
22881 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22882 and niters_no_overflow parameters. Update calls to
22883 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22884 Create a new SSA name if the latter choses to use a ste other
22885 than zero, and return it via niters_vector_mult_vf_var.
22886 * tree-vect-loop.c (vect_transform_loop): Update calls to
22887 vect_do_peeling, vect_gen_vector_loop_niters and
22888 slpeel_make_loop_iterate_ntimes.
22889 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22890 (vect_gen_vector_loop_niters): Update declarations after above changes.
22892 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
22894 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22895 128-bit round to integer instructions.
22896 (ceil<mode>2): Likewise.
22897 (btrunc<mode>2): Likewise.
22898 (round<mode>2): Likewise.
22900 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22902 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22903 unaligned VSX load/store on P8/P9.
22904 (expand_block_clear): Allow the use of unaligned VSX
22905 load/store on P8/P9.
22907 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22909 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22911 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22912 swap associated with both a load and a store.
22914 2018-01-02 Andrew Waterman <andrew@sifive.com>
22916 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22917 * config/riscv/riscv.md (clear_cache): Use it.
22919 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
22921 * web.c: Remove out-of-date comment.
22923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22925 * expr.c (fixup_args_size_notes): Check that any existing
22926 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22927 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22928 (emit_single_push_insn): ...here.
22930 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22932 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22933 (const_vector_encoded_nelts): New function.
22934 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22935 (const_vector_int_elt, const_vector_elt): Declare.
22936 * emit-rtl.c (const_vector_int_elt_1): New function.
22937 (const_vector_elt): Likewise.
22938 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22939 of CONST_VECTOR_ELT.
22941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22943 * expr.c: Include rtx-vector-builder.h.
22944 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22945 directly on the tree encoding.
22946 (const_vector_from_tree): Likewise.
22947 * optabs.c: Include rtx-vector-builder.h.
22948 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22949 sequence of "u" values.
22950 * vec-perm-indices.c: Include rtx-vector-builder.h.
22951 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22952 directly on the vec_perm_indices encoding.
22954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22956 * doc/rtl.texi (const_vector): Describe new encoding scheme.
22957 * Makefile.in (OBJS): Add rtx-vector-builder.o.
22958 * rtx-vector-builder.h: New file.
22959 * rtx-vector-builder.c: Likewise.
22960 * rtl.h (rtx_def::u2): Add a const_vector field.
22961 (CONST_VECTOR_NPATTERNS): New macro.
22962 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22963 (CONST_VECTOR_DUPLICATE_P): Likewise.
22964 (CONST_VECTOR_STEPPED_P): Likewise.
22965 (CONST_VECTOR_ENCODED_ELT): Likewise.
22966 (const_vec_duplicate_p): Check for a duplicated vector encoding.
22967 (unwrap_const_vec_duplicate): Likewise.
22968 (const_vec_series_p): Check for a non-duplicated vector encoding.
22969 Say that the function only returns true for integer vectors.
22970 * emit-rtl.c: Include rtx-vector-builder.h.
22971 (gen_const_vec_duplicate_1): Delete.
22972 (gen_const_vector): Call gen_const_vec_duplicate instead of
22973 gen_const_vec_duplicate_1.
22974 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22975 (gen_const_vec_duplicate): Use rtx_vector_builder.
22976 (gen_const_vec_series): Likewise.
22977 (gen_rtx_CONST_VECTOR): Likewise.
22978 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22979 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22980 Build a new vector rather than modifying a CONST_VECTOR in-place.
22981 (handle_special_swappables): Update call accordingly.
22982 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22983 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22984 Build a new vector rather than modifying a CONST_VECTOR in-place.
22985 (handle_special_swappables): Update call accordingly.
22987 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22989 * simplify-rtx.c (simplify_const_binary_operation): Use
22990 CONST_VECTOR_ELT instead of XVECEXP.
22992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22994 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22995 the selector elements to be different from the data elements
22996 if the selector is a VECTOR_CST.
22997 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22998 ssizetype for the selector.
23000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23002 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23003 before testing each element individually.
23004 * tree-vect-generic.c (lower_vec_perm): Likewise.
23006 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23008 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23009 * selftest-run-tests.c (selftest::run_tests): Call it.
23010 * vector-builder.h (vector_builder::operator ==): New function.
23011 (vector_builder::operator !=): Likewise.
23012 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23013 (vec_perm_indices::all_from_input_p): New function.
23014 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23015 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23016 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23017 instead of reading the VECTOR_CST directly. Detect whether both
23018 vector inputs are the same before constructing the vec_perm_indices,
23019 and update the number of inputs argument accordingly. Use the
23020 utility functions added above. Only construct sel2 if we need to.
23022 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23024 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23025 the broadcast of the low byte.
23026 (expand_mult_highpart): Use an explicit encoding for the permutes.
23027 * optabs-query.c (can_mult_highpart_p): Likewise.
23028 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23029 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23030 (vectorizable_bswap): Likewise.
23031 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23032 explicit encoding for the power-of-2 permutes.
23033 (vect_permute_store_chain): Likewise.
23034 (vect_grouped_load_supported): Likewise.
23035 (vect_permute_load_chain): Likewise.
23037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23039 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23040 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23041 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23042 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23043 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23044 (vect_gen_perm_mask_any): Likewise.
23046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23048 * int-vector-builder.h: New file.
23049 * vec-perm-indices.h: Include int-vector-builder.h.
23050 (vec_perm_indices): Redefine as an int_vector_builder.
23051 (auto_vec_perm_indices): Delete.
23052 (vec_perm_builder): Redefine as a stand-alone class.
23053 (vec_perm_indices::vec_perm_indices): New function.
23054 (vec_perm_indices::clamp): Likewise.
23055 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23056 (vec_perm_indices::new_vector): New function.
23057 (vec_perm_indices::new_expanded_vector): Update for new
23058 vec_perm_indices class.
23059 (vec_perm_indices::rotate_inputs): New function.
23060 (vec_perm_indices::all_in_range_p): Operate directly on the
23061 encoded form, without computing elided elements.
23062 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23063 encoding. Update for new vec_perm_indices class.
23064 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23065 the given vec_perm_builder.
23066 (expand_vec_perm_var): Update vec_perm_builder constructor.
23067 (expand_mult_highpart): Use vec_perm_builder instead of
23068 auto_vec_perm_indices.
23069 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23070 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23071 or double series encoding as appropriate.
23072 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23073 vec_perm_indices instead of auto_vec_perm_indices.
23074 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23075 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23076 (vect_permute_store_chain): Likewise.
23077 (vect_grouped_load_supported): Likewise.
23078 (vect_permute_load_chain): Likewise.
23079 (vect_shift_permute_load_chain): Likewise.
23080 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23081 (vect_transform_slp_perm_load): Likewise.
23082 (vect_schedule_slp_instance): Likewise.
23083 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23084 (vectorizable_mask_load_store): Likewise.
23085 (vectorizable_bswap): Likewise.
23086 (vectorizable_store): Likewise.
23087 (vectorizable_load): Likewise.
23088 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23089 vec_perm_indices instead of auto_vec_perm_indices. Use
23090 tree_to_vec_perm_builder to read the vector from a tree.
23091 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23092 vec_perm_builder instead of a vec_perm_indices.
23093 (have_whole_vector_shift): Use vec_perm_builder and
23094 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23095 truncation to calc_vec_perm_mask_for_shift.
23096 (vect_create_epilog_for_reduction): Likewise.
23097 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23098 from auto_vec_perm_indices to vec_perm_indices.
23099 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23100 instead of changing individual elements.
23101 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23102 the vector in d.perm.
23103 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23104 from auto_vec_perm_indices to vec_perm_indices.
23105 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23106 instead of changing individual elements.
23107 (arm_vectorize_vec_perm_const): Use new_vector to install
23108 the vector in d.perm.
23109 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23110 Update vec_perm_builder constructor.
23111 (rs6000_expand_interleave): Likewise.
23112 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23113 (rs6000_expand_interleave): Likewise.
23115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23117 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23118 to qimode could truncate the indices.
23119 * optabs.c (expand_vec_perm_var): Likewise.
23121 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23123 * Makefile.in (OBJS): Add vec-perm-indices.o.
23124 * vec-perm-indices.h: New file.
23125 * vec-perm-indices.c: Likewise.
23126 * target.h (vec_perm_indices): Replace with a forward class
23128 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23129 * optabs.h: Include vec-perm-indices.h.
23130 (expand_vec_perm): Delete.
23131 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23132 (expand_vec_perm_const): Declare.
23133 * target.def (vec_perm_const_ok): Replace with...
23134 (vec_perm_const): ...this new hook.
23135 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23136 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23137 * doc/tm.texi: Regenerate.
23138 * optabs.def (vec_perm_const): Delete.
23139 * doc/md.texi (vec_perm_const): Likewise.
23140 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23141 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23142 expand_vec_perm for constant permutation vectors. Assert that
23143 the mode of variable permutation vectors is the integer equivalent
23144 of the mode that is being permuted.
23145 * optabs-query.h (selector_fits_mode_p): Declare.
23146 * optabs-query.c: Include vec-perm-indices.h.
23147 (selector_fits_mode_p): New function.
23148 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23149 is defined, instead of checking whether the vec_perm_const_optab
23150 exists. Use targetm.vectorize.vec_perm_const instead of
23151 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23152 fit in the vector mode before using a variable permute.
23153 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23154 vec_perm_indices instead of an rtx.
23155 (expand_vec_perm): Replace with...
23156 (expand_vec_perm_const): ...this new function. Take the selector
23157 as a vec_perm_indices rather than an rtx. Also take the mode of
23158 the selector. Update call to shift_amt_for_vec_perm_mask.
23159 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23160 Use vec_perm_indices::new_expanded_vector to expand the original
23161 selector into bytes. Check whether the indices fit in the vector
23162 mode before using a variable permute.
23163 (expand_vec_perm_var): Make global.
23164 (expand_mult_highpart): Use expand_vec_perm_const.
23165 * fold-const.c: Includes vec-perm-indices.h.
23166 * tree-ssa-forwprop.c: Likewise.
23167 * tree-vect-data-refs.c: Likewise.
23168 * tree-vect-generic.c: Likewise.
23169 * tree-vect-loop.c: Likewise.
23170 * tree-vect-slp.c: Likewise.
23171 * tree-vect-stmts.c: Likewise.
23172 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23174 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23175 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23176 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23177 (aarch64_vectorize_vec_perm_const): ...this new function.
23178 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23179 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23180 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23181 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23182 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23183 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23184 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23186 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23187 check for NEON modes.
23188 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23189 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23190 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23191 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23193 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23194 the old VEC_PERM_CONST conditions.
23195 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23196 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23197 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23198 (ia64_vectorize_vec_perm_const_ok): Merge into...
23199 (ia64_vectorize_vec_perm_const): ...this new function.
23200 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23201 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23202 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23203 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23204 * config/mips/mips.c (mips_expand_vec_perm_const)
23205 (mips_vectorize_vec_perm_const_ok): Merge into...
23206 (mips_vectorize_vec_perm_const): ...this new function.
23207 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23208 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23209 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23210 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23211 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23212 (rs6000_expand_vec_perm_const): Delete.
23213 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23215 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23216 (altivec_expand_vec_perm_const_le): Take each operand individually.
23217 Operate on constant selectors rather than rtxes.
23218 (altivec_expand_vec_perm_const): Likewise. Update call to
23219 altivec_expand_vec_perm_const_le.
23220 (rs6000_expand_vec_perm_const): Delete.
23221 (rs6000_vectorize_vec_perm_const_ok): Delete.
23222 (rs6000_vectorize_vec_perm_const): New function.
23223 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23224 an element count and rtx array.
23225 (rs6000_expand_extract_even): Update call accordingly.
23226 (rs6000_expand_interleave): Likewise.
23227 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23228 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23229 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23230 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23231 (rs6000_expand_vec_perm_const): Delete.
23232 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23233 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23234 (altivec_expand_vec_perm_const_le): Take each operand individually.
23235 Operate on constant selectors rather than rtxes.
23236 (altivec_expand_vec_perm_const): Likewise. Update call to
23237 altivec_expand_vec_perm_const_le.
23238 (rs6000_expand_vec_perm_const): Delete.
23239 (rs6000_vectorize_vec_perm_const_ok): Delete.
23240 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23241 reference to the SPE evmerge intructions.
23242 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23243 an element count and rtx array.
23244 (rs6000_expand_extract_even): Update call accordingly.
23245 (rs6000_expand_interleave): Likewise.
23246 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23247 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23249 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23251 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23253 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23254 vector mode and that that mode matches the mode of the data
23256 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23257 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23258 directly using expand_vec_perm_1 when forcing selectors into
23260 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23264 * optabs-query.h (can_vec_perm_p): Delete.
23265 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23266 * optabs-query.c (can_vec_perm_p): Split into...
23267 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23268 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23269 particular selector is valid.
23270 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23271 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23272 (vect_grouped_load_supported): Likewise.
23273 (vect_shift_permute_load_chain): Likewise.
23274 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23275 (vect_transform_slp_perm_load): Likewise.
23276 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23277 (vectorizable_bswap): Likewise.
23278 (vect_gen_perm_mask_checked): Likewise.
23279 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23280 implementations of variable permutation vectors into account
23281 when deciding which selector to use.
23282 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23283 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23284 with a false third argument.
23285 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23286 to test whether the constant selector is valid and can_vec_perm_var_p
23287 to test whether a variable selector is valid.
23289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23291 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23292 * optabs-query.c (can_vec_perm_p): Likewise.
23293 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23294 instead of vec_perm_indices.
23295 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23296 (vect_gen_perm_mask_checked): Likewise,
23297 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23298 (vect_gen_perm_mask_checked): Likewise,
23300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23302 * optabs-query.h (qimode_for_vec_perm): Declare.
23303 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23304 (qimode_for_vec_perm): ...this new function.
23305 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23307 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23309 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23310 does not have a conditional at the top.
23312 2018-01-02 Richard Biener <rguenther@suse.de>
23314 * ipa-inline.c (big_speedup_p): Fix expression.
23316 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23319 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23322 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23326 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23327 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23328 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23329 cond_taken_branch_cost 3->4.
23331 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23333 PR tree-optimization/83581
23334 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23335 TODO_cleanup_cfg if any changes have been made.
23337 PR middle-end/83608
23338 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23339 convert_modes if target mode has the right side, but different mode
23342 PR middle-end/83609
23343 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23344 last argument when extracting from CONCAT. If either from_real or
23345 from_imag is NULL, use expansion through memory. If result is not
23346 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23347 the parts directly to inner mode, if even that fails, use expansion
23350 PR middle-end/83623
23351 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23352 check for bswap in mode rather than HImode and use that in expand_unop
23355 Copyright (C) 2018 Free Software Foundation, Inc.
23357 Copying and distribution of this file, with or without modification,
23358 are permitted in any medium without royalty provided the copyright
23359 notice and this notice are preserved.