1 2018-08-16 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/86853
4 * gimple-ssa-sprintf.c (struct format_result): Rename member.
5 (struct fmtresult): Add member and initialize it in ctors.
6 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
7 (format_string): Handle %S the same as %ls. Set MAYFAIL.
8 (format_directive): Set POSUNDER4K when MAYFAIL is set.
9 (parse_directive): Handle %C same as %c.
10 (sprintf_dom_walker::compute_format_length): Adjust.
11 (is_call_safe): Adjust.
13 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
15 * builtins.c (c_strlen): Add new parameter eltsize. Use it
16 for determining how to count the elements.
17 * builtins.h (c_strlen): Adjust prototype.
18 * expr.c (string_constant): Add new parameter mem_size.
19 Set *mem_size appropriately.
20 * expr.h (string_constant): Adjust protoype.
21 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
22 * gimple-fold.h (get_range_strlen): Adjust prototype.
23 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
24 (format_string): Call get_string_length with eltsize.
26 2018-08-16 David Malcolm <dmalcolm@redhat.com>
28 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
29 to emit the span, rather than setting it as the prefix.
31 2018-08-16 David Malcolm <dmalcolm@redhat.com>
33 * diagnostic-show-locus.c (layout::start_annotation_line): Add
34 "margin_char" parameter, defaulting to space. Use it in place
35 of pp_space for the initial part of the margin.
36 (layout::print_leading_fixits): Use '+' when filling the margin
37 of line-insertion fix-it hints.
39 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
41 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
44 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
46 * config/rs6000/altivec.md: Don't set length attribute to the default
48 * config/rs6000/darwin.md: Ditto.
49 * config/rs6000/dfp.md: Ditto.
50 * config/rs6000/htm.md: Ditto.
51 * config/rs6000/rs6000.md: Ditto.
52 * config/rs6000/sync.md: Ditto.
53 * config/rs6000/vsx.md: Ditto.
55 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
57 * config/rs6000/altivec.md: Don't set length attribute to the default
58 value, for branch instructions.
59 * config/rs6000/darwin.md: Ditto.
60 * config/rs6000/rs6000.md: Ditto.
62 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
64 * config/rs6000/rs6000.md (length): Always define as const_int 4.
65 (unnamed conditional branch define_insn): Set length to 4 or 8
67 (<bd>_<mode>): Similar, for alternative 0.
68 (<bd>tf_<mode>): Ditto.
70 2018-08-16 Tamar Christina <tamar.christina@arm.com>
72 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
74 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
76 * doc/rtl.texi: Replace old RTX class names with new names.
79 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
81 * expmed.h (canonicalize_comparison): New declaration.
82 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
83 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
84 * optabs.c (prepare_cmp_insn): Likewise.
85 * rtl.h (unsigned_condition_p): New function which checks if a
86 comparison operator is unsigned.
88 2018-08-16 Nathan Sidwell <nathan@acm.org>
90 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
91 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
93 2018-08-16 Tamar Christina <tamar.christina@arm.com>
96 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
97 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
98 (mov<mov>): ..this and enable unconditionally.
100 2018-08-16 Tamar Christina <tamar.christina@arm.com>
102 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
104 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
106 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
107 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
108 "Common" with "Target".
110 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
112 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
113 * doc/invoke.texi (mmitigate-rop): Remove.
114 * config/i386/i386.c: Do not include "regrename.h".
115 (ix86_rop_should_change_byte_p, reg_encoded_number)
116 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
118 (ix86_reorg): Remove call to ix86_mitigate_rop.
119 * config/i386/i386.md (attr "modrm_class"): Remove.
120 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
121 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
122 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
124 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
126 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
127 allow folding of mergeh() and mergel() for the float and double types.
128 (fold_mergehl_helper): Rework to handle building a permute tree
131 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
133 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
136 2018-08-15 David Malcolm <dmalcolm@redhat.com>
138 * common.opt (fdiagnostics-show-labels): New option.
139 * diagnostic-show-locus.c (class layout_range): Add field
141 (class layout): Add field "m_show_labels_p".
142 (layout_range::layout_range): Add param "label" and use it to
144 (make_range): Pass in NULL for new "label" param of layout_range's
146 (layout::layout): Initialize m_show_labels_p.
147 (layout::maybe_add_location_range): Pass in loc_range->m_label
148 when constructing layout_range instances.
149 (struct line_label): New struct.
150 (layout::print_any_labels): New member function.
151 (layout::print_line): Call it if label-printing is enabled.
152 (selftest::test_one_liner_labels): New test.
153 (selftest::test_diagnostic_show_locus_one_liner): Call it.
154 * diagnostic.c (diagnostic_initialize): Initialize
155 context->show_labels_p.
156 * diagnostic.h (struct diagnostic_context): Add field
158 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
159 -fno-diagnostics-show-labels.
160 * dwarf2out.c (gen_producer_string): Add
161 OPT_fdiagnostics_show_labels to the ignored options.
162 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
164 (gcc_rich_location::maybe_add_expr): Likewise.
165 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
166 label" param, defaulting to NULL.
167 (gcc_rich_location::add_expr): Add "label" param.
168 (gcc_rich_location::maybe_add_expr): Likewise.
169 (class text_range_label): New class.
170 (class range_label_for_type_mismatch): New class.
171 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
172 of format_warning_va.
173 (fmtwarn_n): Likewise for new params of format_warning_n_va.
174 * lto-wrapper.c (merge_and_complain): Add
175 OPT_fdiagnostics_show_labels to the "pick one setting" options.
176 (append_compiler_options): Likewise to the dropped options.
177 (append_diag_options): Likewise to the passed-on options.
178 * opts.c (common_handle_option): Handle the new option.
179 * selftest-diagnostic.c
180 (test_diagnostic_context::test_diagnostic_context): Enable
182 * substring-locations.c: Include "gcc-rich-location.h".
183 (format_warning_n_va): Add "fmt_label" and "param_label" params
184 and use them as appropriate.
185 (format_warning_va): Add "fmt_label" and "param_label" params,
186 passing them on to format_warning_n_va.
187 (format_warning_at_substring): Likewise.
188 (format_warning_at_substring_n): Likewise.
189 * substring-locations.h (format_warning_va): Add "fmt_label" and
190 "param_label" params.
191 (format_warning_n_va): Likewise.
192 (format_warning_at_substring): Likewise.
193 (format_warning_at_substring_n): Likewise.
194 * toplev.c (general_init): Initialize global_dc->show_labels_p.
196 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
199 * builtins.c (expand_builtin_memcmp): Do not expand the call
200 when overflow is detected.
202 2018-08-15 Martin Sebor <msebor@redhat.com>
204 PR tree-optimization/71625
205 * config/aarch64/aarch64-builtins.c
206 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
208 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
210 * config/s390/s390.c (s390_reorg): Remove loop.
212 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
215 (darwin_function_switched_text_sections): Delete.
216 * gcc/config/darwin.h
217 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
219 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
222 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
223 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
224 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
226 2018-08-15 Martin Liska <mliska@suse.cz>
228 PR tree-optimization/86925
229 * predict.c (expr_expected_value_1): When taking
230 later predictor, assign also probability.
231 Use fold_build2_initializer_loc in order to fold
232 the expression in -frounding-math.
234 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
236 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
238 (expand_vec_perm_1): Try the new method.
240 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
243 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
244 Check whether lra_live_max_point is 0 before dividing.
246 2018-08-14 Martin Sebor <msebor@redhat.com>
248 PR tree-optimization/86650
249 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
250 (vrp_prop::check_mem_ref): Same.
252 2018-08-13 Liu Hao <lh_mouse@126.com>
254 * pretty-print.c (eat_esc_sequence): Swap the foreground and
255 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
256 and clear it thereafter, as it only works for DBCS.
258 2018-08-13 Liu Hao <lh_mouse@126.com>
260 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
261 handle returned by _get_osf_handle().
263 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
265 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
266 for folding vec_perm.
268 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
270 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
271 Add support for gimple-folding of vec_pack() and vec_unpack()
274 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
276 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
277 vec_xst variants to the list.
278 (rs6000_gimple_fold_builtin): Add support for folding unaligned
279 vector loads and stores.
281 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
283 * config.gcc (rs6000-ibm-aix4.x): Delete.
284 (rs6000-ibm-aix5.1): Delete.
285 (rs6000-ibm-aix5.2): Delete.
286 (rs6000-ibm-aix5.3): Delete.
287 * config/rs6000/aix43.h: Delete.
288 * config/rs6000/aix51.h: Delete.
289 * config/rs6000/aix52.h: Delete.
290 * config/rs6000/t-aix43: Delete.
292 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
294 * config/s390/s390.c (s390_decompose_constant_pool_ref):
296 (s390_decompose_address): Factor out constant pool ref
299 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
301 * config/nds32/nds32-predicates.c
302 (nds32_can_use_bclr_p): Change return type as bool.
303 (nds32_can_use_bset_p): Ditto.
304 (nds32_can_use_btgl_p): Ditto.
305 (nds32_can_use_bitci_p): Ditto.
306 * config/nds32/nds32-protos.h
307 (nds32_can_use_bclr_p): Change declaration.
308 (nds32_can_use_bset_p): Ditto.
309 (nds32_can_use_btgl_p): Ditto.
310 (nds32_can_use_bitci_p): Ditto.
312 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
315 Support -msched-prolog-epilog option.
316 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
318 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
320 * common/config/nds32/nds32-common.c
321 (nds32_option_optimization_table): Enalbe -malways-align.
323 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
325 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
327 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
328 OPT_misr_secure_ case.
329 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
330 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
331 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
333 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
334 (nds32_isr_info): New field security_level.
335 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
336 * config/nds32/nds32.md (return_internal): Consider critical attribute.
337 * config/nds32/nds32.opt (misr-secure): New option.
338 * config/nds32/nds32_init.inc: New file.
339 * config/nds32/nds32_isr.h: New file.
341 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
343 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
344 Update comment for atomic instructions.
345 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
346 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
348 (atomic_loaddi): Revise fence expansion to only emit fence prior to
349 load for __ATOMIC_SEQ_CST model.
350 (atomic_loaddi_1): Remove float register target.
351 (atomic_storedi): Handle CONST_INT values.
352 (atomic_storedi_1): Remove float register source. Add special case
354 (memory_barrier): New expander and insn.
356 2018-08-11 Jakub Jelinek <jakub@redhat.com>
358 PR tree-optimization/86835
359 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
360 new_stmt after def_gsi, make sure to insert new_square_stmt after
361 that stmt, not 2 stmts before it.
363 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
366 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
369 2018-08-10 Martin Liska <mliska@suse.cz>
372 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
374 * builtins.c (expand_builtin_expect_with_probability):
376 (expand_builtin_expect_with_probability): New function.
377 (build_builtin_expect_predicate): Add new argumnet probability
378 for BUILT_IN_EXPECT_WITH_PROBABILITY.
379 (fold_builtin_expect):
382 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
383 * builtins.h (fold_builtin_expect): Set new argument.
384 * doc/extend.texi: Document __builtin_expect_with_probability.
385 * doc/invoke.texi: Likewise.
386 * gimple-fold.c (gimple_fold_call): Pass new argument.
387 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
388 also BUILT_IN_EXPECT_WITH_PROBABILITY.
389 * predict.c (get_predictor_value): New function.
390 (expr_expected_value): Add new argument probability. Assume
391 that predictor and probability are always non-null.
392 (expr_expected_value_1): Likewise. For __builtin_expect and
393 __builtin_expect_with_probability set probability. Handle
394 combination in binary expressions.
395 (tree_predict_by_opcode): Simplify code by simply calling
397 (pass_strip_predict_hints::execute): Add handling of
398 BUILT_IN_EXPECT_WITH_PROBABILITY.
399 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
401 * tree.h (DECL_BUILT_IN_P): New function.
403 2018-08-10 Martin Liska <mliska@suse.cz>
405 PR tree-optimization/85799
406 * passes.def: Add argument for pass_strip_predict_hints.
407 * predict.c (class pass_strip_predict_hints): Add new argument
409 (strip_predictor_early): New function.
410 (pass_strip_predict_hints::execute): Call the function to
412 (strip_predict_hints): New function.
413 * predict.def: Fix comment.
415 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
417 * Makefile.in: Clarify which tm.texi to copy over to assert the
418 right to grant a GFDL license for all.
420 2018-08-09 Jeff Law <law@redhat.com>
422 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
425 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
427 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
430 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
432 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
433 reductions for variable-length vectors.
435 2018-08-09 David Malcolm <dmalcolm@redhat.com>
438 * common.opt (fdiagnostics-show-line-numbers): New option.
439 * diagnostic-show-locus.c (class layout): Add fields
440 "m_show_line_numbers_p" and "m_linenum_width";
441 (num_digits): New function.
442 (test_num_digits): New function.
443 (layout::layout): Initialize new fields. Update m_x_offset
444 logic to handle any left margin.
445 (layout::print_source_line): Print line number when requested.
446 (layout::start_annotation_line): New member function.
447 (layout::print_annotation_line): Call it.
448 (layout::print_leading_fixits): Likewise.
449 (layout::print_trailing_fixits): Likewise. Update calls to
450 move_to_column for new parameter.
451 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
452 it to potentially call start_annotation_line.
453 (layout::show_ruler): Call start_annotation_line.
454 (selftest::test_line_numbers_multiline_range): New selftest.
455 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
456 and selftest::test_line_numbers_multiline_range.
457 * diagnostic.c (diagnostic_initialize): Initialize
459 * diagnostic.h (struct diagnostic_context): Add field
460 "show_line_numbers_p".
461 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
462 -fno-diagnostics-show-line-numbers.
463 * dwarf2out.c (gen_producer_string): Add
464 OPT_fdiagnostics_show_line_numbers to the ignored options.
465 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
466 one setting" options.
467 (append_compiler_options): Likewise to the dropped options.
468 (append_diag_options): Likewise to the passed-on options.
469 * opts.c (common_handle_option): Handle the new option.
470 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
472 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
474 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
475 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
476 third argument of type "const signed char" to descriptions of
477 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
478 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
479 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
480 __builtin_bcdsub_ov functions.
482 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
484 PR tree-optimization/86858
485 * tree-vect-loop.c (vect_is_simple_reduction): Restore
486 flow_bb_inside_loop_p calls.
488 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
490 PR tree-optimization/86871
491 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
492 instead of gimple_assign_lhs.
494 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
497 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
498 register constraint to operand 0.
499 (add<mode>3_carryinC): Likewise.
500 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
502 2018-08-09 Martin Liska <mliska@suse.cz>
505 * common.opt: Remove extra line.
507 2018-08-09 Martin Liska <mliska@suse.cz>
509 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
510 at the end of a line, make first letter capital and end up
511 a sentence with a dot.
512 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
513 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
514 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
515 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
516 (PARAM_MAX_ISL_OPERATIONS): Likewise.
517 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
518 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
519 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
520 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
521 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
522 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
523 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
524 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
525 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
526 (PARAM_TREE_REASSOC_WIDTH): Likewise.
527 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
528 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
529 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
531 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
534 * config/s390/s390.c (s390_option_override_internal): Reduce the
535 stack-clash-protection-probe-interval param if it would be too big
538 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
541 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
542 * config/m68k/m68k.c (handle_move_double): Don't call
543 m68k_final_prescan_insn.
544 (m68k_adjust_decorated_operand): Renamed from
545 m68k_final_prescan_insn, remove first and third operand and
547 (print_operand): Call it.
548 (print_operand_address): Call it.
550 2018-08-08 Nathan Sidwell <nathan@acm.org>
552 * diagnostic.c (diagnostic_report_current_module): Use
553 linemap_included_from & linemap_included_from_linemap.
555 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
557 * config/aarch64/aarch64-cores.def: Add phecda core.
558 * config/aarch64/aarch64-tune.md: Regenerate.
559 * doc/invoke.texi: Add phecda core.
561 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
564 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
566 * config/s390/s390.md ("movti"): Add more alternatives for
567 constant to GPR copies.
569 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
571 * config/s390/s390.c: Fix whitespace damage throughout the file.
572 * config/s390/s390.h: Likewise.
573 * config/s390/tpf.h: Likewise.
575 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
577 * config/s390/s390.c (s390_loadrelative_operand_p):
578 Remove TARGET_CPU_ZARCH usages.
579 (s390_rtx_costs): Likewise.
580 (s390_legitimate_constant_p): Likewise.
581 (s390_cannot_force_const_mem): Likewise.
582 (legitimate_reload_constant_p): Likewise.
583 (s390_preferred_reload_class): Likewise.
584 (legitimize_pic_address): Likewise.
585 (legitimize_tls_address): Likewise.
586 (s390_split_branches): Removed.
587 (s390_add_execute): Removed.
588 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
589 (s390_mainpool_start): Likewise.
590 (s390_mainpool_finish): Likewise.
591 (s390_mainpool_cancel): Removed.
592 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
593 (s390_chunkify_cancel): Likewise.
594 (s390_return_addr_rtx): Likewise.
595 (s390_register_info): Remove split_branches_pending_p uages.
596 (s390_optimize_register_info): Likewise.
597 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
598 split_branches_pending_p usages.
599 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
600 (s390_load_got): Likewise.
601 (s390_expand_split_stack_prologue): Likewise.
602 (output_asm_nops): Likewise.
603 (s390_function_profiler): Likewise.
604 (s390_emit_call): Likewise.
605 (s390_conditional_register_usage): Likewise.
606 (s390_optimize_prologue): Likewise.
607 (s390_reorg): Remove TARGET_CPU_ZARCH and
608 split_branches_pending_p usages.
609 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
611 (s390_output_indirect_thunk_function): Likewise.
612 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
613 (TARGET_CPU_ZARCH_P): Removed.
614 (struct machine_function): Remove split_branches_pending_p.
615 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
617 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
619 * common/config/s390/s390-common.c (processor_flags_table):
621 * config.gcc: Remove with_arch/with_tune support.
622 * config/s390/2064.md: Remove cpu attribute comparisons.
623 * config/s390/driver-native.c (s390_host_detect_local_cpu):
625 * config/s390/linux.h (ASM_SPEC):
626 Remove -march support.
627 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
628 Use a table to get an arch level.
629 * config/s390/s390-opts.h (enum processor_type):
632 (processor_table): Remove entries, add arch_level values.
633 (s390_issue_rate): Remove cases.
634 (s390_option_override): Adjust
635 s390_option_override_internal() call.
636 (s390_option_override_internal): Remove deprecation warning.
637 (s390_valid_target_attribute_tree): Adjust
638 s390_option_override_internal() call.
639 * config/s390/s390.h (struct s390_processor):
640 Share with s390-c.c, add arch_level field.
641 * config/s390/s390.md:
642 Remove occurrences in cpu attribute.
643 * config/s390/s390.opt: Remove -march/-mtune support.
644 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
645 * doc/invoke.texi: Remove deprecation warning.
647 2018-08-08 Luis Machado <luis.machado@linaro.org>
649 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
651 (qdf24xx_tunings): Set vector cost structure to
654 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
655 <register_sextend>: Set to 3.
657 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
660 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
661 * config/aarch64/aarch64-simd.md
662 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
663 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
664 (aarch64_frecpx<mode>): ...this new pattern.
665 * config/aarch64/aarch64-simd-builtins.def: Remove comment
666 about aarch64_frecp<FRECP:frecp_suffix><mode>.
668 2018-08-07 Martin Liska <mliska@suse.cz>
671 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
672 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
673 * predict.def (PRED_MALLOC_NONNULL): New predictor.
674 * doc/extend.texi: Document that malloc attribute adds
677 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
680 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
681 Define to speculation_safe_value_not_needed.
683 2018-08-06 Jeff Law <law@redhat.com>
685 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
686 the vr_values instance to cprop_into_stmt.
687 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
688 (cprop_operand): Also query EVRP to determine if OP is a constant.
690 2018-08-06 Nathan Sidwell <nathan@acm.org>
692 * diagnostic.c (diagnostic_report_current_module): Reroll
693 included-at loop. Translate text.
695 2018-08-06 David Malcolm <dmalcolm@redhat.com>
697 * function-tests.c (selftest::test_expansion_to_rtl): Call
698 free_after_compilation.
700 2018-08-06 Alan Hayward <alan.hayward@arm.com>
702 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
704 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
706 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
707 loops with memory block operations from getting unrolled.
709 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
712 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
713 Define to speculation_safe_value_not_needed.
715 2018-08-06 Jeff Law <law@redhat.com>
717 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
720 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
723 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
724 with all enabled __intN types.
726 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
728 2018-08-06 Alan Hayward <alan.hayward@arm.com>
730 * alias.c (record_set): Check for clobber high.
731 * cfgexpand.c (expand_gimple_stmt): Likewise.
732 * combine-stack-adj.c (single_set_for_csa): Likewise.
733 * combine.c (find_single_use_1): Likewise.
734 (set_nonzero_bits_and_sign_copies): Likewise.
735 (get_combine_src_dest): Likewise.
736 (is_parallel_of_n_reg_sets): Likewise.
737 (try_combine): Likewise.
738 (record_dead_and_set_regs_1): Likewise.
739 (reg_dead_at_p_1): Likewise.
740 (reg_dead_at_p): Likewise.
741 * dce.c (deletable_insn_p): Likewise.
742 (mark_nonreg_stores_1): Likewise.
743 (mark_nonreg_stores_2): Likewise.
744 * df-scan.c (df_find_hard_reg_defs): Likewise.
745 (df_uses_record): Likewise.
746 (df_get_call_refs): Likewise.
747 * dwarf2out.c (mem_loc_descriptor): Likewise.
748 * haifa-sched.c (haifa_classify_rtx): Likewise.
749 * ira-build.c (create_insn_allocnos): Likewise.
750 * ira-costs.c (scan_one_insn): Likewise.
751 * ira.c (equiv_init_movable_p): Likewise.
752 (rtx_moveable_p): Likewise.
753 (interesting_dest_for_shprep): Likewise.
754 * jump.c (mark_jump_label_1): Likewise.
755 * postreload-gcse.c (record_opr_changes): Likewise.
756 * postreload.c (reload_cse_simplify): Likewise.
757 (struct reg_use): Add source expr.
758 (reload_combine): Check for clobber high.
759 (reload_combine_note_use): Likewise.
760 (reload_cse_move2add): Likewise.
761 (move2add_note_store): Likewise.
762 * print-rtl.c (print_pattern): Likewise.
763 * recog.c (decode_asm_operands): Likewise.
764 (store_data_bypass_p): Likewise.
765 (if_test_bypass_p): Likewise.
766 * regcprop.c (kill_clobbered_value): Likewise.
767 (kill_set_value): Likewise.
768 * reginfo.c (reg_scan_mark_refs): Likewise.
769 * reload1.c (maybe_fix_stack_asms): Likewise.
770 (eliminate_regs_1): Likewise.
771 (elimination_effects): Likewise.
772 (mark_not_eliminable): Likewise.
773 (scan_paradoxical_subregs): Likewise.
774 (forget_old_reloads_1): Likewise.
775 * reorg.c (find_end_label): Likewise.
776 (try_merge_delay_insns): Likewise.
777 (redundant_insn): Likewise.
778 (own_thread_p): Likewise.
779 (fill_simple_delay_slots): Likewise.
780 (fill_slots_from_thread): Likewise.
781 (dbr_schedule): Likewise.
782 * resource.c (update_live_status): Likewise.
783 (mark_referenced_resources): Likewise.
784 (mark_set_resources): Likewise.
785 * rtl.c (copy_rtx): Likewise.
786 * rtlanal.c (reg_referenced_p): Likewise.
787 (single_set_2): Likewise.
788 (noop_move_p): Likewise.
789 (note_stores): Likewise.
790 * sched-deps.c (sched_analyze_reg): Likewise.
791 (sched_analyze_insn): Likewise.
793 2018-08-06 Alan Hayward <alan.hayward@arm.com>
795 * cse.c (invalidate_reg): New function extracted from...
796 (invalidate): ...here.
797 (canonicalize_insn): Check for clobber high.
798 (invalidate_from_clobbers): invalidate clobber highs.
799 (invalidate_from_sets_and_clobbers): Likewise.
800 (count_reg_usage): Check for clobber high.
801 (insn_live_p): Likewise.
802 * cselib.c (cselib_expand_value_rtx_1):Likewise.
803 (cselib_invalidate_regno): Check for clobber in setter.
804 (cselib_invalidate_rtx): Pass through setter.
805 (cselib_invalidate_rtx_note_stores):
806 (cselib_process_insn): Check for clobber high.
807 * cselib.h (cselib_invalidate_rtx): Add operand.
809 2018-08-06 Alan Hayward <alan.hayward@arm.com>
811 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
812 (mark_not_eliminable): Likewise.
813 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
814 * lra-lives.c (process_bb_lives): Check for clobber high.
815 * lra.c (new_insn_reg): Remember clobber highs.
816 (collect_non_operand_hard_regs): Check for clobber high.
817 (lra_set_insn_recog_data): Likewise.
818 (add_regs_to_insn_regno_info): Likewise.
819 (lra_update_insn_regno_info): Likewise.
821 2018-08-06 Alan Hayward <alan.hayward@arm.com>
823 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
824 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
826 2018-08-06 Alan Hayward <alan.hayward@arm.com>
828 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
829 (copy_insn_1): Likewise.
830 (gen_hard_reg_clobber_high): New gen function.
831 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
832 * genemit.c (gen_exp): Likewise.
833 (gen_emit_seq): Pass through info.
834 (gen_insn): Check for CLOBBER_HIGH.
835 (gen_expand): Pass through info.
836 (gen_split): Likewise.
837 (output_add_clobbers): Likewise.
838 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
839 (remove_clobbers): Likewise.
840 * rtl.h (gen_hard_reg_clobber_high): New declaration.
842 2018-08-06 Alan Hayward <alan.hayward@arm.com>
844 * doc/rtl.texi (clobber_high): Add.
845 (parallel): Add in clobber high
846 * rtl.c (rtl_check_failed_code3): Add function.
847 * rtl.def (CLOBBER_HIGH): Add expression.
848 * rtl.h (RTL_CHECKC3): Add macro.
849 (rtl_check_failed_code3): Add declaration.
852 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
855 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
856 cfun->machine->max_used_stack_alignment if needed.
858 2018-08-04 Martin Sebor <msebor@redhat.com>
860 PR tree-optimization/86571
861 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
864 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
867 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
870 2018-08-03 Jeff Law <law@redhat.com>
873 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
874 Define to speculation_safe_value_not_needed.
876 2018-08-03 David Malcolm <dmalcolm@redhat.com>
878 * doc/gcov.texi (-x): Remove duplicate "to".
879 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
880 (-Wif-not-aligned): Remove duplicate "is".
881 (-flto): Remove duplicate "the".
882 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
884 (MSP430 Options): Remove duplicate "and" from the description
885 of "-mgprel-sec=regexp".
886 (x86 Options): Remove duplicate copies of "vmldLog102" and
887 vmlsLog104 from description of "-mveclibabi=type".
889 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
891 * internal-fn.h (first_commutative_argument): Declare.
892 * internal-fn.c (first_commutative_argument): New function.
893 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
894 restrictions for pattern statements. Use first_commutative_argument
895 to look for commutative operands in calls to internal functions.
897 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
899 * Makefile.in (wide-int-range.o): New.
900 * tree-vrp.c: Move all the wide_int_* functions to...
901 * wide-int-range.cc: ...here.
902 * tree-vrp.h: Move all the wide_int_* prototypes to...
903 * wide-int-range.h: ...here.
905 2018-08-03 Tom de Vries <tdevries@suse.de>
907 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
909 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
910 * except.c (output_function_exception_table): Do early exit if
911 targetm_common.except_unwind_info (&global_options) == UI_NONE.
913 2018-08-03 Martin Liska <mliska@suse.cz>
915 * predict.c (dump_prediction): Change to 2 digits
918 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
920 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
922 (falkor_am_1_gtov_gtov): ... a new insn reservation.
924 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
926 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
927 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
928 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
929 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
930 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
932 2018-08-02 David Malcolm <dmalcolm@redhat.com>
934 * diagnostic-show-locus.c (diagnostic_show_locus): Use
935 pp_take_prefix when saving the existing prefix.
936 * diagnostic.c (diagnostic_append_note): Likewise.
937 * langhooks.c (lhd_print_error_function): Likewise.
938 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
939 param's type. Free the existing prefix.
940 (pp_take_prefix): New function.
941 (pretty_printer::pretty_printer): Drop the prefix parameter.
942 Rename the length parameter to match the comment.
943 (pretty_printer::~pretty_printer): Free the prefix.
944 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
946 (struct pretty_printer): Drop the "const" from "prefix" field's
947 type and clarify memory management.
948 (pp_set_prefix): Drop the "const" from the 2nd param.
949 (pp_take_prefix): New decl.
951 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
953 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
954 for word_mode_ok here instead of passing as argument.
955 (expand_block_compare): Change select_block_compare_mode() call.
956 (expand_strncmp_gpr_sequence): New function.
957 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
959 2018-08-02 Jeff Law <law@redhat.com>
962 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
963 Define to speculation_safe_value_not_needed.
966 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
967 Define to speculation_safe_value_not_needed.
969 2018-08-02 Tom de Vries <tdevries@suse.de>
972 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
973 function. Return UI_TARGET unconditionally.
974 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
975 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
977 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
979 * genemit.c (print_overload_test): Fix typo.
981 2018-08-02 Richard Biener <rguenther@suse.de>
983 PR tree-optimization/86816
984 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
985 which checks for value availability before querying it.
986 (gvn_uses_equal): Use it.
987 (same_succ_hash): Likewise.
988 (gimple_equal_p): Likewise.
990 2018-08-02 Nick Clifton <nickc@redhat.com>
993 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
994 Define to speculation_safe_value_not_needed.
997 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
998 Define to speculation_safe_value_not_needed.
1001 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1002 Define to speculation_safe_value_not_needed.
1005 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1006 Define to speculation_safe_value_not_needed.
1009 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1010 Define to speculation_safe_value_not_needed.
1013 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1014 Define to speculation_safe_value_not_needed.
1017 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1018 Define to speculation_safe_value_not_needed.
1021 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1022 Define to speculation_safe_value_not_needed.
1025 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1026 speculation_safe_value_not_needed.
1029 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1030 to speculation_safe_value_not_needed.
1032 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1034 * doc/md.texi: Expand the documentation of instruction names
1035 to mention port-local uses. Document '@' in pattern names.
1036 * read-md.h (overloaded_instance, overloaded_name): New structs.
1038 (md_reader::handle_overloaded_name): New member function.
1039 (md_reader::get_overloads): Likewise.
1040 (md_reader::m_first_overload): New member variable.
1041 (md_reader::m_next_overload_ptr): Likewise.
1042 (md_reader::m_overloads_htab): Likewise.
1043 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1044 m_next_overload_ptr and m_overloads_htab.
1045 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1046 (get_mode_token, get_code_token, get_int_token): New functions.
1047 (map_attr_string): Add an optional argument that passes back
1048 the associated iterator.
1049 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1050 (md_reader::handle_overloaded_name, add_overload_instance): New
1052 (apply_iterators): Handle '@' names. Report an error if '@'
1053 is used without iterators.
1054 (initialize_iterators): Initialize the new iterator_group fields.
1055 * genopinit.c (handle_overloaded_code_for)
1056 (handle_overloaded_gen): New functions.
1057 (main): Use them to print declarations of maybe_code_for_* and
1058 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1059 * genemit.c (print_overload_arguments, print_overload_test)
1060 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1061 (main): Use it to print definitions of maybe_code_for_* and
1062 maybe_gen_* functions.
1063 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1064 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1065 instead of explicit mode checks.
1066 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1067 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1068 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1069 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1070 (aarch64_expand_compare_and_swap): Likewise
1071 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1072 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1073 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1074 (aarch64_constant_pool_reload_icode): Delete.
1075 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1076 instead of aarch64_constant_pool_reload_icode. Use
1077 code_for_aarch64_reload_mov instead of explicit mode checks.
1078 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1079 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1080 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1081 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1082 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1083 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1084 (aarch64_atomic_load_op_code): Delete.
1085 (aarch64_emit_atomic_load_op): Likewise.
1086 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1087 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1088 instead of aarch64_emit_atomic_load_op.
1089 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1090 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1091 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1092 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1093 character before the pattern name.
1094 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1095 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1096 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1097 (aarch64_frecps<mode>): Likewise.
1098 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1099 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1100 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1101 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1102 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1104 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1106 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1107 Allow HFmode constants if TARGET_FP_F16INST.
1109 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1112 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1113 No longer check last store for clobber of address register.
1115 2018-08-02 Martin Liska <mliska@suse.cz>
1117 PR gcov-profile/86817
1118 * gcov.c (process_all_functions): New function.
1120 (process_file): Move functions processing to
1121 process_all_functions.
1123 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1125 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1126 "const" to the "gimple *" and "rtx_insn *" parameters.
1127 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1129 (dump_location_t::dump_location_t): Likewise.
1131 2018-08-01 Martin Sebor <msebor@redhat.com>
1133 PR tree-optimization/86650
1134 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1135 rather than a "gcall *". Directly pass the data of interest
1136 to percent_K_format, rather than building a temporary CALL_EXPR
1138 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1139 (gimple_fold_builtin_strncat): Adjust.
1140 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1141 gcall* argument with gimple*.
1142 * gimple-ssa-warn-restrict.c (check_call): Same.
1143 (wrestrict_dom_walker::before_dom_children): Same.
1144 (builtin_access::builtin_access): Same.
1145 (check_bounds_or_overlap): Same
1146 (maybe_diag_overlap): Same.
1147 (maybe_diag_offset_bounds): Same.
1148 * tree-diagnostic.c (default_tree_printer): Move usage of
1149 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1151 * tree-pretty-print.c (percent_K_format): Add argument.
1152 * tree-pretty-print.h: Add argument.
1153 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1154 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1155 (maybe_diag_stxncpy_trunc): Same.
1156 (handle_builtin_stxncpy): Same.
1157 (handle_builtin_strcat): Same.
1159 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1161 * match.pd: Optimise pointer range checks.
1163 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1165 PR tree-optimization/86758
1166 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1167 to remove pattern statements.
1169 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1171 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1172 result of dfs_enumerate_from when constructing stmt_vec_infos,
1173 instead of additionally calling get_loop_body.
1175 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1177 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1179 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1180 When creating an iv, assert that the step is not known to be zero.
1181 (vect_setup_realignment): Update call accordingly.
1182 * tree-vect-stmts.c (vectorizable_store): Likewise.
1183 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1185 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1187 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1188 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1189 (vectorizable_reduction): Likewise.
1190 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1191 (vect_detect_hybrid_slp_stmts): Likewise.
1192 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1194 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1196 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1197 (wide_int_set_zero_nonzero_bits): ...this.
1198 (zero_nonzero_bits_from_vr): Rename to...
1199 (vrp_set_zero_nonzero_bits): ...this.
1200 (extract_range_from_multiplicative_op_1): Abstract wide int
1202 (wide_int_range_multiplicative_op): ...here.
1203 (extract_range_from_binary_expr_1): Extract wide int binary
1204 operations into their own functions.
1205 (wide_int_range_lshift): New.
1206 (wide_int_range_can_optimize_bit_op): New.
1207 (wide_int_range_shift_undefined_p): New.
1208 (wide_int_range_bit_xor): New.
1209 (wide_int_range_bit_ior): New.
1210 (wide_int_range_bit_and): New.
1211 (wide_int_range_trunc_mod): New.
1212 (extract_range_into_wide_ints): New.
1213 (vrp_shift_undefined_p): New.
1214 (extract_range_from_multiplicative_op): New.
1215 (vrp_can_optimize_bit_op): New.
1216 * tree-vrp.h (value_range::dump): New.
1217 (wide_int_range_multiplicative_op): New.
1218 (wide_int_range_lshift):New.
1219 (wide_int_range_shift_undefined_p): New.
1220 (wide_int_range_bit_xor): New.
1221 (wide_int_range_bit_ior): New.
1222 (wide_int_range_bit_and): New.
1223 (wide_int_range_trunc_mod): New.
1224 (zero_nonzero_bits_from_bounds): Rename to...
1225 (wide_int_set_zero_nonzero_bits): ...this.
1226 (zero_nonzero_bits_from_vr): Rename to...
1227 (vrp_set_zero_nonzero_bits): ...this.
1228 (range_easy_mask_min_max): Rename to...
1229 (wide_int_range_can_optimize_bit_op): this.
1230 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1231 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1233 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1235 * tree-vectorizer.h (vect_orig_stmt): New function.
1236 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1237 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1238 (vect_create_epilog_for_reduction): Likewise.
1239 (vectorizable_live_operation): Likewise.
1240 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1241 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1242 * tree-vect-stmts.c (vectorizable_call): Likewise.
1243 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1245 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1247 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1249 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1250 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1251 (vect_transform_loop): Likewise.
1252 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1254 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1256 * tree-vectorizer.h (vect_schedule_slp): Return void.
1257 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1258 (vect_schedule_slp): Likewise.
1260 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1262 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1264 (vect_transform_loop): Update calls accordingly. Schedule SLP
1265 instances before the main loop, if any exist.
1267 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1269 PR tree-optimization/86749
1270 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1271 If the lhs is used in a COND_EXPR, check that it is being used
1272 as the "then" or "else" value.
1274 2018-08-01 Tom de Vries <tdevries@suse.de>
1277 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1278 speculation_safe_value_not_needed.
1280 2018-08-01 Richard Biener <rguenther@suse.de>
1282 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1285 2018-08-01 Martin Liska <mliska@suse.cz>
1287 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1288 format how successful transformation is dumped.
1289 (gimple_mod_pow2_value_transform): Likewise.
1290 (gimple_mod_subtract_transform): Likewise.
1291 (gimple_stringops_transform): Likewise.
1293 2018-08-01 Martin Liska <mliska@suse.cz>
1296 * value-prof.c (interesting_stringop_to_profile_p):
1297 Simplify the code and add BUILT_IN_MEMMOVE.
1298 (gimple_stringops_transform): Likewise.
1300 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1302 * config/aarch64/aarch64-simd.md
1303 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1304 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1305 use GPI iterator instead of SI mode.
1307 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1309 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1310 rs6000_speculation_barrier.
1311 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1312 new barrier pattern name.
1314 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1316 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1317 (speculation_barrier): New insn.
1319 2018-08-01 Richard Biener <rguenther@suse.de>
1322 * graphite.h: Include isl/id.h and isl/space.h to allow build
1325 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1328 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1329 mode for COFF targets.
1330 * defaults.h (TARGET_COFF): Define.
1331 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1332 TARGET_COFF): Define.
1333 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1334 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1335 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1337 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1339 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1340 * gimple-streamer-out.c (output_bb): Save it.
1341 * lto-streamer-in.c (input_struct_function_base): Restore
1342 instance discriminator if available. Create map on demand.
1343 * lto-streamer-out.c (output_struct_function_base): Save it if
1345 * final.c (decl_to_instance_map): Document LTO strategy.
1347 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1348 Olivier Hainque <hainque@adacore.com>
1350 * debug.h (decl_to_instance_map_t): New type.
1351 (decl_to_instance_map): Declare.
1352 (maybe_create_decl_to_instance_map): New inline function.
1353 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1354 to track basic block discriminators.
1355 (final_start_function_1): Initialize them.
1356 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1358 (decl_to_instance_map): New variable.
1359 (map_decl_to_instance, maybe_set_discriminator): New functions.
1360 (notice_source_line): Set discriminator.
1362 2018-07-31 Ian Lance Taylor <iant@golang.org>
1364 * targhooks.c (default_have_speculation_safe_value): Add
1367 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1369 * dump-context.h: Include "pretty-print.h".
1370 (dump_context::refresh_dumps_are_enabled): New decl.
1371 (dump_context::emit_item): New decl.
1372 (class dump_context): Add fields "m_test_pp" and
1374 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1375 (temp_dump_context::get_dumped_text): New decl.
1376 (class temp_dump_context): Add field "m_pp".
1377 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1378 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1380 (set_dump_file): Update for above change.
1381 (set_alt_dump_file): Likewise.
1382 (dump_loc): New overload, taking a pretty_printer *.
1383 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1384 to any test pretty-printer.
1385 (make_item_for_dump_gimple_stmt): New function, adapted from
1386 optinfo::add_gimple_stmt.
1387 (dump_context::dump_gimple_stmt): Call it, and use the result,
1388 eliminating the direct usage of dump_file and alt_dump_file in
1389 favor of indirectly using them via emit_item.
1390 (make_item_for_dump_gimple_expr): New function, adapted from
1391 optinfo::add_gimple_expr.
1392 (dump_context::dump_gimple_expr): Call it, and use the result,
1393 eliminating the direct usage of dump_file and alt_dump_file in
1394 favor of indirectly using them via emit_item.
1395 (make_item_for_dump_generic_expr): New function, adapted from
1397 (dump_context::dump_generic_expr): Call it, and use the result,
1398 eliminating the direct usage of dump_file and alt_dump_file in
1399 favor of indirectly using them via emit_item.
1400 (make_item_for_dump_printf_va): New function, adapted from
1401 optinfo::add_printf_va.
1402 (make_item_for_dump_printf): New function.
1403 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1404 and use the result, eliminating the direct usage of dump_file and
1405 alt_dump_file in favor of indirectly using them via emit_item.
1406 (make_item_for_dump_dec): New function.
1407 (dump_context::dump_dec): Call it, and use the result,
1408 eliminating the direct usage of dump_file and alt_dump_file in
1409 favor of indirectly using them via emit_item.
1410 (make_item_for_dump_symtab_node): New function, adapted from
1411 optinfo::add_symtab_node.
1412 (dump_context::dump_symtab_node): Call it, and use the result,
1413 eliminating the direct usage of dump_file and alt_dump_file in
1414 favor of indirectly using them via emit_item.
1415 (dump_context::begin_scope): Reimplement, avoiding direct usage
1416 of dump_file and alt_dump_file in favor of indirectly using them
1418 (dump_context::emit_item): New member function.
1419 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1420 Set up test pretty-printer on the underlying context. Call
1421 refresh_dumps_are_enabled.
1422 (temp_dump_context::~temp_dump_context): Call
1423 refresh_dumps_are_enabled.
1424 (temp_dump_context::get_dumped_text): New member function.
1425 (selftest::verify_dumped_text): New function.
1426 (ASSERT_DUMPED_TEXT_EQ): New macro.
1427 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1428 and then without optinfo enabled. Add uses of
1429 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1430 * dumpfile.h: Update comment for the dump_* API.
1431 * optinfo-emit-json.cc
1432 (selftest::test_building_json_from_dump_calls): Update for new
1433 param for temp_dump_context ctor.
1434 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1435 and "m_owned" field.
1436 (optinfo_item::~optinfo_item): Likewise.
1437 (optinfo::add_item): New member function.
1438 (optinfo::emit): Update comment.
1439 (optinfo::add_string): Delete.
1440 (optinfo::add_printf): Delete.
1441 (optinfo::add_printf_va): Delete.
1442 (optinfo::add_gimple_stmt): Delete.
1443 (optinfo::add_gimple_expr): Delete.
1444 (optinfo::add_tree): Delete.
1445 (optinfo::add_symtab_node): Delete.
1446 (optinfo::add_dec): Delete.
1447 * optinfo.h (class dump_context): New forward decl.
1448 (optinfo::add_item): New decl.
1449 (optinfo::add_string): Delete.
1450 (optinfo::add_printf): Delete.
1451 (optinfo::add_printf_va): Delete.
1452 (optinfo::add_gimple_stmt): Delete.
1453 (optinfo::add_gimple_expr): Delete.
1454 (optinfo::add_tree): Delete.
1455 (optinfo::add_symtab_node): Delete.
1456 (optinfo::add_dec): Delete.
1457 (optinfo::add_poly_int): Delete.
1458 (optinfo_item::optinfo_item): Remove "owned" param.
1459 (class optinfo_item): Remove field "m_owned".
1461 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1464 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1465 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1466 requested variable alignment.
1467 (expand_one_ssa_partition): Likewise.
1468 (expand_one_var): Likewise.
1470 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1472 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1473 to speculation_safe_value_not_needed.
1475 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1477 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1478 * targhooks.c (speculation_safe_value_not_needed): New function.
1479 * target.def (have_speculation_safe_value): Update documentation.
1480 * doc/tm.texi: Regenerated.
1482 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1484 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1485 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1487 (despeculate_copy<ALLI:mode>_insn): New insn.
1488 (despeculate_copyti_insn): New insn.
1489 (despeculate_simple<ALLI:mode>): New insn
1490 (despeculate_simpleti): New insn.
1491 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1493 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1494 aarch64_speculation_safe_value.
1495 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1497 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1499 * config/aarch64/aarch64-speculation.cc: New file.
1500 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1501 before pass_reorder_blocks.
1502 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1504 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1505 X14 and X15 when tracking speculation.
1506 * config/aarch64/aarch64.md (register name constants): Add
1507 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1508 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1509 (speculation_barrier): New insn attribute.
1510 (cmp<mode>): Allow SP in comparisons.
1511 (speculation_tracker): New insn.
1512 (speculation_barrier): Add speculation_barrier attribute.
1513 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1514 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1515 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1517 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1519 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1520 aarch64_track_speculation is true.
1521 (tb<optab><mode>1): Likewise.
1522 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1523 generate CB[N]Z when tracking speculation.
1524 (aarch64_split_compare_and_swap): Likewise.
1525 (aarch64_split_atomic_op): Likewise.
1527 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1529 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1531 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1533 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1534 (speculation_barrier): New insn.
1536 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1538 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1539 * config/arm/arm.md (speculation_barrier): New expand.
1540 (speculation_barrier_insn): New pattern.
1542 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1544 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1545 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1546 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1547 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1549 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1550 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1551 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1552 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1553 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1554 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1555 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1556 * builtins.c (expand_speculation_safe_value): New function.
1557 (expand_builtin): Call it.
1558 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1559 * doc/extend.texi: Document __builtin_speculation_safe_value.
1560 * doc/md.texi: Document "speculation_barrier" pattern.
1561 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1562 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1563 * doc/tm.texi: Regenerated.
1564 * target.def (have_speculation_safe_value, speculation_safe_value): New
1566 * targhooks.c (default_have_speculation_safe_value): New function.
1567 (default_speculation_safe_value): New function.
1568 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1569 (default_speculation_safe_value): Add prototype.
1571 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1573 * dump-context.h (dump_context::dump_loc): New decl.
1574 * dumpfile.c (dump_context::dump_loc): New member function.
1575 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1576 and dump_gimple_stmt.
1577 (dump_context::dump_gimple_expr_loc): Likewise, using
1579 (dump_context::dump_generic_expr_loc): Likewise, using
1581 (dump_context::dump_printf_loc_va): Likewise, using
1583 (dump_context::begin_scope): Explicitly using the global function
1584 "dump_loc", rather than the member function.
1586 2018-07-31 Martin Sebor <msebor@redhat.com>
1588 PR tree-optimization/86741
1589 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1591 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1593 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1594 depend on whether prefetch instructions will be emitted or not.
1595 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1596 will be emitted or not.
1597 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1598 (TARGET_SETMEM_PFD): New macros.
1600 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1602 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1603 (NULL_STMT_VEC_INFO): Delete.
1604 (stmt_vec_info::operator*): Likewise.
1605 (stmt_vec_info::operator gimple *): Likewise.
1606 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1607 of NULL_STMT_VEC_INFO.
1608 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1609 (vect_reassociating_reduction_p): Likewise.
1610 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1611 (vectorizable_store): Likewise.
1612 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1613 (vec_info::free_stmt_vec_infos): Likewise.
1615 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1617 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1618 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1620 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1622 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1623 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1624 (vec_info::free_stmt_vec_info): New private member functions.
1625 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1626 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1627 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1628 set_stmt_vec_info_vec and free_stmt_vec_infos.
1629 * tree-vect-loop.c (_loop_vec_info): Remove call to
1630 set_stmt_vec_info_vec.
1631 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1632 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1633 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1634 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1635 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1636 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1637 with the clearing in the xcalloc.
1638 (stmt_vec_info_vec): Delete.
1639 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1640 (vectorize_loops): Likewise.
1641 (vec_info::~vec_info): Remove argument from call to
1642 free_stmt_vec_infos.
1643 (vec_info::add_stmt): Remove vinfo argument from call to
1646 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1648 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1649 rather than a gimple stmt.
1650 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1651 information for pattern statements when passed the original
1652 statement; instead wait to be passed the pattern statement itself.
1653 Don't call set_vinfo_for_stmt here.
1654 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1655 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1656 stmt_vec_infos here.
1657 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1658 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1659 stmt_vec_infos entry.
1661 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1663 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1664 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1665 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1666 * tree-vect-stmts.c (vectorizable_call): Likewise.
1667 (vectorizable_simd_clone_call): Likewise.
1669 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1671 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1672 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1673 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1674 * tree-vect-loop.c (vect_transform_loop): Likewise.
1675 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1676 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1678 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1680 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1681 (vect_dr_stmt): Delete.
1682 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1683 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1685 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1686 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1687 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1688 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1689 (vect_analyze_data_refs): Likewise.
1690 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1692 (vect_find_same_alignment_drs): Likewise.
1693 (vect_slp_analyze_node_dependences): Update calls accordingly.
1694 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1695 instead of DR_VECT_AUX.
1696 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1697 of a vector data references. Use vec_info::lookup_dr instead of
1699 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1700 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1701 instead of DR_VECT_AUX.
1703 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1705 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1707 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1709 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1710 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1711 (vect_gen_prolog_loop_niters): Likewise.
1713 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1715 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1716 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1717 (vect_known_alignment_in_bytes, vect_dr_behavior)
1718 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1719 instead of data_references. Update calls to other routines for
1720 which the same change has been made.
1721 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1722 dr_vec_infos instead of stmt_vec_infos.
1723 (vect_analyze_data_ref_dependence): Update call accordingly.
1724 (vect_slp_analyze_data_ref_dependence)
1725 (vect_record_base_alignments): Use DR_VECT_AUX.
1726 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1727 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1728 (vector_alignment_reachable_p, vect_get_data_access_cost)
1729 (vect_peeling_supportable, vect_analyze_group_access_1)
1730 (vect_analyze_group_access, vect_analyze_data_ref_access)
1731 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1732 (vect_compile_time_alias, vect_small_gap_p)
1733 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1734 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1735 instead of data_references. Update calls to other routines for
1736 which the same change has been made.
1737 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1738 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1739 (vect_slp_analyze_and_verify_node_alignment)
1740 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1741 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1742 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1744 (_vect_peel_info::dr): Replace with...
1745 (_vect_peel_info::dr_info): ...this new field.
1746 (vect_peeling_hash_get_most_frequent)
1747 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1748 (vect_peeling_hash_get_lowest_cost):
1749 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1750 routines for which the same change has been made.
1751 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1753 * tree-vect-loop-manip.c (get_misalign_in_elems)
1754 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1756 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1757 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1758 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1759 (get_group_load_store_type, get_negative_load_store_type)
1760 (vect_get_data_ptr_increment, vectorizable_store)
1761 (vectorizable_load): Likewise.
1762 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1763 Update calls to other routines for which the same change has been made.
1765 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1767 * tree-vectorizer.h (vec_info::move_dr): New member function.
1768 (dataref_aux): Rename to...
1769 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1770 (_stmt_vec_info::dr_aux): Update accordingly.
1771 (_stmt_vec_info::data_ref_info): Delete.
1772 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1773 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1774 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1775 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1777 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1778 (STMT_VINFO_DR_INFO): New macro.
1779 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1780 (set_dr_misalignment): Update after rename of dataref_aux.
1781 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1782 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1783 initialization of STMT_VINFO_DATA_REF.
1784 * tree-vectorizer.c (vec_info::move_dr): New function.
1785 * tree-vect-patterns.c (vect_recog_bool_pattern)
1786 (vect_recog_mask_conversion_pattern)
1787 (vect_recog_gather_scatter_pattern): Use it.
1788 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1789 the "dr" and "stmt" fields of dr_vec_info instead of
1790 STMT_VINFO_DATA_REF.
1792 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1794 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1795 (is_pattern_stmt_p): Use it.
1796 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1797 on pattern statements.
1799 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1801 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1802 original stmt as a stmt_vec_info rather than a gimple stmt.
1803 (vect_pattern_recog_1): Take the statement directly as a
1804 stmt_vec_info, rather than via a gimple_stmt_iterator.
1805 Update call to vect_mark_pattern_stmts.
1806 (vect_pattern_recog): Update calls accordingly.
1808 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1810 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1811 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1812 a vect_def_type for the first argument.
1813 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1814 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1815 operand if it isn't defined by a vectorized statement.
1816 (vect_build_gather_load_calls): Remove the mask_dt argument and
1817 update calls to vect_get_vec_def_for_stmt_copy.
1818 (vectorizable_bswap): Likewise the dt argument.
1819 (vectorizable_call): Update calls to vectorizable_bswap and
1820 vect_get_vec_def_for_stmt_copy.
1821 (vectorizable_simd_clone_call, vectorizable_assignment)
1822 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1823 (vectorizable_comparison): Update calls to
1824 vect_get_vec_def_for_stmt_copy.
1825 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1827 (vect_get_loop_based_defs): Remove dt argument and update call
1828 to vect_get_vec_def_for_stmt_copy.
1829 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1830 and vect_get_vec_def_for_stmt_copy.
1831 (vectorizable_load): Update calls to vect_build_gather_load_calls
1832 and vect_get_vec_def_for_stmt_copy.
1833 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1834 (vectorizable_reduction, vectorizable_live_operation): Update calls
1835 to vect_get_vec_def_for_stmt_copy.
1837 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1839 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1840 and gimple stmt arguments with a stmt_vec_info.
1841 (vect_record_base_alignments): Update calls accordingly.
1842 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1843 and gimple stmt arguments with a stmt_vec_info.
1844 (vect_build_slp_tree_1): Remove vinfo argument and update call
1845 to vect_record_max_nunits.
1846 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1847 and vect_record_max_nunits.
1849 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1851 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1852 file and take a stmt_vec_info instead of a gimple stmt.
1853 (supportable_widening_operation, vect_finish_replace_stmt)
1854 (vect_finish_stmt_generation, vect_get_store_rhs)
1855 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1856 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1857 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1858 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1859 (vect_create_data_ref_ptr, bump_vector_ptr)
1860 (vect_permute_store_chain, vect_setup_realignment)
1861 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1862 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1863 (vectorizable_reduction, vectorizable_induction)
1864 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1865 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1866 than gimple stmts as arguments.
1867 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1868 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1869 (can_group_stmts_p, vect_check_gather_scatter)
1870 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1871 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1872 (vect_permute_load_chain, vect_shift_permute_load_chain)
1873 (vect_transform_grouped_load)
1874 (vect_record_grouped_load_vectors): Likewise.
1875 * tree-vect-loop.c (vect_fixup_reduc_chain)
1876 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1877 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1878 (vectorizable_reduction, vectorizable_induction)
1879 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1880 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1881 (vect_get_load_store_mask): Likewise.
1882 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1883 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1884 * tree-vect-stmts.c (vect_mark_relevant)
1885 (is_simple_and_all_uses_invariant)
1886 (exist_non_indexing_operands_for_use_p, process_use)
1887 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1888 (vect_get_vec_def_for_operand, vect_get_vec_defs)
1889 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1890 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1891 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1892 (get_negative_load_store_type, get_load_store_type)
1893 (vect_check_load_store_mask, vect_check_store_rhs)
1894 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1895 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1896 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1897 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1898 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1899 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1900 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1901 (supportable_widening_operation): Likewise.
1903 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1905 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1906 a stmt_vec_info instead of a gcall.
1907 (vect_check_gather_scatter): Update call accordingly.
1908 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1910 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1911 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1912 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1913 instead of a gimple stmt.
1914 (vect_transform_loop): Update calls accordingly.
1915 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1916 stmt_vec_infos instead of gimple stmts.
1917 (vect_analyze_slp_instance): Update use accordingly.
1918 * tree-vect-stmts.c (read_vector_array, write_vector_array)
1919 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1920 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1921 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1922 (vect_gen_widened_results_half, vect_get_loop_based_defs)
1923 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1924 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1925 down to subroutines.
1927 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1929 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1930 of the worklist from a vector of gimple stmts to a vector of
1932 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1933 (vect_mark_stmts_to_be_vectorized): Likewise
1935 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1937 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1938 statement before passing it to vect_analyze_stmt.
1939 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1940 the chain of phi vector definitions. Track the exit phi via its
1942 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1943 STMT_VINFO_REDUC_DEF.
1944 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1945 stmt_vec_infos to handle the statement chains.
1946 (vect_get_slp_defs): Record the first statement in the node
1947 using a stmt_vec_info.
1948 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1949 statements here and pass their stmt_vec_info down to subroutines.
1950 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1951 down to vect_finish_stmt_generation.
1952 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1953 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1954 stmt_vec_infos to subroutines.
1955 (vect_remove_stores): Use stmt_vec_infos to handle the statement
1958 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1960 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1961 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1962 (vect_permute_store_chain, vect_setup_realignment)
1963 (vect_permute_load_chain, vect_shift_permute_load_chain)
1964 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1965 stmts internally, and when passing values to other vectorizer routines.
1966 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1967 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
1968 (vect_analyze_loop_operations, get_initial_def_for_reduction)
1969 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
1970 (vectorizable_reduction, vectorizable_induction)
1971 (vectorizable_live_operation, vect_transform_loop_stmt)
1972 (vect_transform_loop): Likewise.
1973 * tree-vect-patterns.c (vect_reassociating_reduction_p)
1974 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
1975 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
1976 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1977 (vect_slp_analyze_node_operations_1): Likewise.
1978 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1979 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
1980 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
1981 (vect_finish_stmt_generation_1, get_group_load_store_type)
1982 (get_load_store_type, vect_build_gather_load_calls)
1983 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1984 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1985 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1986 (vectorizable_store, vectorizable_load, vectorizable_condition)
1987 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1988 (supportable_widening_operation): Likewise.
1989 (vect_get_vector_types_for_stmt): Likewise.
1990 * tree-vectorizer.h (vect_dr_behavior): Likewise.
1992 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1994 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1995 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1996 (vect_permute_store_chain, vect_permute_load_chain)
1997 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1998 repeated stmt_vec_info lookups.
1999 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2000 (vect_update_ivs_after_vectorizer): Likewise.
2001 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2002 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2003 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2004 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2005 (vect_bb_slp_scalar_cost): Likewise.
2006 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2008 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2010 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2011 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2012 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2013 to get gassigns and gcalls, rather than operating on generc gimple
2015 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2016 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2017 (vectorizable_load, vect_analyze_stmt): Likewise.
2018 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2020 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2022 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2023 return stmt_vec_infos rather than gimple stmts. Do not accept
2025 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2027 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2028 Update use of get_later_stmt.
2029 (vect_get_constant_vectors): Update call accordingly.
2030 (vect_schedule_slp_instance): Likewise
2031 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2032 (vect_slp_analyze_instance_dependence): Likewise.
2033 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2035 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2037 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2038 (stmt_info_for_cost::stmt_info): ...this new field.
2039 (add_stmt_costs): Update accordingly.
2040 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2041 (vect_get_known_peeling_cost): Likewise.
2042 (vect_estimate_min_profitable_iters): Likewise.
2043 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2045 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2047 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2048 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2049 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2051 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2053 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2055 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2056 a gimple stmt to a stmt_vec_info.
2057 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2059 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2061 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2062 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2063 (_loop_vec_info::reduction_chains): Likewise.
2064 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2066 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2068 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2070 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2071 a gimple stmt to a stmt_vec_info.
2072 (_stmt_vec_info::next_element): Likewise.
2073 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2074 (vect_slp_analyze_and_verify_node_alignment)
2075 (vect_analyze_group_access_1, vect_analyze_group_access)
2076 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2077 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2078 (vect_supportable_dr_alignment): Update accordingly.
2079 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2080 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2081 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2082 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2083 * tree-vect-slp.c (vect_build_slp_tree_1)
2084 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2085 (vect_split_slp_store_group, vect_analyze_slp_instance)
2086 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2087 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2088 (get_group_load_store_type, get_load_store_type)
2089 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2090 (vect_transform_stmt, vect_remove_stores): Likewise.
2092 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2094 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2096 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2097 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2098 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2099 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2100 (vector_alignment_reachable_p, vect_get_data_access_cost)
2101 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2102 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2103 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2104 (vect_analyze_group_access_1, vect_analyze_group_access)
2105 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2106 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2107 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2108 result of vect_dr_stmt and use the stmt_vec_info instead of
2109 the associated gimple stmt.
2110 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2111 (vect_gen_prolog_loop_niters): Likewise.
2112 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2114 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2116 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2117 to a vec<stmt_vec_info>.
2118 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2119 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2121 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2122 to a vec<stmt_vec_info>.
2123 (bst_traits::value_type, bst_traits::value_type): Likewise.
2124 (bst_traits::hash): Update accordingly.
2125 (vect_get_and_check_slp_defs): Change the stmts parameter from
2126 a vec<gimple *> to a vec<stmt_vec_info>.
2127 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2128 (vect_build_slp_tree): Likewise.
2129 (vect_build_slp_tree_2): Likewise. Update uses of
2130 SLP_TREE_SCALAR_STMTS.
2131 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2132 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2133 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2134 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2135 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2136 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2137 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2138 (vect_get_constant_vectors, vect_get_slp_defs)
2139 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2140 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2141 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2142 instead of gimple stmts.
2143 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2144 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2145 (vect_slp_analyze_instance_dependence): Update uses of
2146 SLP_TREE_SCALAR_STMTS.
2147 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2148 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2149 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2150 (get_initial_defs_for_reduction): Likewise.
2151 (vect_create_epilog_for_reduction): Likewise.
2152 (vectorize_fold_left_reduction): Likewise.
2153 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2154 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2155 (can_vectorize_live_stmts): Likewise.
2157 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2159 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2160 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2161 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2163 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2165 (gather_scalar_reductions): Update after above interface changes.
2166 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2167 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2169 (vect_force_simple_reduction): Likewise.
2170 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2171 LOOP_VINFO_REDUCTIONS.
2172 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2174 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2176 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2177 a gimple stmt to a stmt_vec_info.
2178 * tree-vect-loop.c (vect_active_double_reduction_p)
2179 (vect_force_simple_reduction, vectorizable_reduction): Update
2182 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2184 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2185 vec<gimple *> to a vec<stmt_vec_info>.
2186 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2187 the reduction_phis argument from a vec<gimple *> to a
2189 (vectorizable_reduction): Likewise the phis local variable that
2190 is passed to vect_create_epilog_for_reduction. Update for new type
2191 of SLP_TREE_VEC_STMTS.
2192 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2193 (vectorizable_live_operation): Likewise.
2194 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2195 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2197 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2199 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2200 a gimple stmt to a stmt_vec_info.
2201 (vectorizable_condition, vectorizable_live_operation)
2202 (vectorizable_reduction, vectorizable_induction): Pass back the
2203 vectorized statement as a stmt_vec_info.
2204 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2205 use of STMT_VINFO_VEC_STMT.
2206 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2207 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2208 as stmt_vec_infos rather than gimple stmts.
2209 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2211 (vectorizable_live_operation): Likewise.
2212 (vectorizable_reduction, vectorizable_induction): Likewise,
2213 updating use of STMT_VINFO_VEC_STMT.
2214 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2215 of STMT_VINFO_VEC_STMT.
2216 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2217 (vectorizable_simd_clone_call, vectorizable_conversion)
2218 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2219 (vectorizable_store, vectorizable_load, vectorizable_condition)
2220 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2221 from a gimple stmt to a stmt_vec_info.
2222 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2223 pointer to a stmt_vec_info to the vectorizable_* routines.
2225 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2227 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2228 a gimple stmt to a stmt_vec_info.
2229 (is_pattern_stmt_p): Update accordingly.
2230 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2231 (vect_record_grouped_load_vectors): Likewise.
2232 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2233 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2234 (vect_model_reduction_cost): Likewise.
2235 (vect_create_epilog_for_reduction): Likewise.
2236 (vectorizable_reduction, vectorizable_induction): Likewise.
2237 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2238 Return the stmt_vec_info for the pattern statement.
2239 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2240 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2241 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2242 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2243 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2244 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2245 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2246 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2248 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2250 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2251 (vect_finish_stmt_generation): Likewise.
2252 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2253 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2254 (vect_build_gather_load_calls): Use the return value of the above
2255 functions instead of a separate call to vinfo_for_stmt. Use narrow
2256 scopes for the input gimple stmt and wider scopes for the associated
2257 stmt_vec_info. Use vec_info::lookup_def when setting these
2258 stmt_vec_infos from an SSA_NAME definition.
2259 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2260 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2261 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2262 (vectorizable_store, vectorizable_load, vectorizable_condition)
2263 (vectorizable_comparison): Likewise.
2264 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2265 (vectorizable_reduction): Likewise.
2267 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2269 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2270 stmt_vec_info * parameter before the optional gimple **.
2271 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2272 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2273 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2274 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2275 (vectorizable_live_operation): Likewise.
2276 * tree-vect-patterns.c (type_conversion_p): Likewise.
2277 (vect_look_through_possible_promotion): Likewise.
2278 (vect_recog_rotate_pattern): Likewise.
2279 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2281 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2283 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2284 a typedef to a wrapper class.
2285 (NULL_STMT_VEC_INFO): New macro.
2286 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2287 (stmt_vec_info::operator*): New function.
2288 (stmt_vec_info::operator gimple *): Likewise.
2289 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2290 (add_stmt_costs): Likewise.
2291 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2292 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2293 (vect_get_known_peeling_cost): Likewise.
2294 (vect_estimate_min_profitable_iters): Likewise.
2295 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2296 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2297 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2298 (vectorizable_store, free_stmt_vec_infos): Likewise.
2299 (new_stmt_vec_info): Change return type of xcalloc to
2302 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2304 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2305 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2306 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2307 a single_imm_use-based sequence.
2308 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2310 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2312 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2313 * tree-vectorizer.c (vec_info::lookup_def): New function.
2314 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2315 (vect_widened_op_tree): Likewise.
2316 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2317 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2318 (vectorizable_reduction): Likewise.
2319 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2321 (vect_is_slp_reduction): Update calls accordingly. Use
2322 vec_info::lookup_def.
2323 (vect_is_simple_reduction): Likewise
2324 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2326 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2328 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2329 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2330 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2332 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2333 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2334 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2335 (vect_is_slp_reduction, vectorizable_induction)
2336 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2337 * tree-vect-patterns.c (vect_init_pattern_stmt):
2338 (vect_determine_min_output_precision_1, vect_determine_precisions)
2339 (vect_pattern_recog): Likewise.
2340 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2341 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2342 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2343 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2344 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2345 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2346 info field from a loop to a loop_vec_info.
2348 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2350 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2351 (vec_info::add_stmt): Declare.
2352 * tree-vectorizer.c (vec_info::add_stmt): New function.
2353 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2354 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2355 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2356 (vectorizable_induction): Likewise.
2357 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2358 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2359 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2360 (vectorizable_load): Likewise.
2361 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2362 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2363 (vect_recog_gather_scatter_pattern): Likewise.
2364 (append_pattern_def_seq): Likewise. Remove a check that is
2365 performed by add_stmt itself.
2367 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2369 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2370 which make_ssa_name was called with new_stmt before new_stmt
2373 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2375 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2377 (vect_is_slp_reduction): ...here...
2378 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2379 that are already known to be false.
2381 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2383 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2384 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2385 STMT_VINFO_NUM_SLP_USES when it's true.
2386 (vect_free_slp_instance): Add a final_p parameter and pass it to
2388 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2389 (vect_analyze_slp_instance): Likewise.
2390 (vect_slp_analyze_operations): Likewise.
2391 (vect_slp_analyze_bb_1): Likewise.
2392 * tree-vectorizer.c (vec_info): Likewise.
2393 * tree-vect-loop.c (vect_transform_loop): Likewise.
2395 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2397 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2398 function is not called for second and subsequent members of
2401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2403 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2404 cases for nested loops from here to ...
2405 (vect_create_epilog_for_reduction): ...here. Only call
2406 vect_is_simple_use for inner-loop reductions.
2408 2018-07-31 Martin Liska <mliska@suse.cz>
2410 PR gcov-profile/85338
2411 PR gcov-profile/85350
2412 PR gcov-profile/85372
2413 * profile.c (struct location_triplet): New.
2414 (struct location_triplet_hash): Likewise.
2415 (output_location): Do not output a BB that
2416 is already recorded for a line.
2417 (branch_prob): Use streamed_locations.
2419 2018-07-31 Martin Liska <mliska@suse.cz>
2421 PR gcov-profile/85370
2422 * coverage.c (coverage_begin_function): Do not mark target
2423 clones as artificial functions.
2425 2018-07-31 Martin Liska <mliska@suse.cz>
2427 PR gcov-profile/83813
2428 PR gcov-profile/84758
2429 PR gcov-profile/85217
2430 PR gcov-profile/85332
2431 * profile.c (branch_prob): Do not record GOTO expressions
2432 for GIMPLE statements which locations are already streamed.
2434 2018-07-31 Olivier Hainque <hainque@adacore.com>
2436 * gcc.c (handle_spec_function): Accept a soft_matched_part
2437 argument, as do_spec_1. Pass it down to ...
2438 (eval_spec_function): Accept a soft_matched_part argument,
2439 and pass it down to ...
2440 (do_spec_2): Accept a soft_matched_part argument, and pass
2441 it down to do_spec_1.
2442 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2443 (handle_braces): Update call to handle_spec_function.
2444 (driver::set_up_specs): Update calls to do_spec_2.
2445 (compare_debug_dump_opt_spec_function): Likewise.
2446 (compare_debug_self_opt_spec_function): Likewise.
2448 2018-07-31 Olivier Hainque <hainque@adacore.com>
2450 * common.opt (nolibc): New option.
2451 * doc/invoke.texi (Link Options): Document it.
2452 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2453 * config/alpha/linux.h: Likewise.
2454 * config/arc/elf.h: Likewise.
2455 * config/arm/uclinux-elf.h: Likewise.
2456 * config/arm/unknown-elf.h: Likewise.
2457 * config/avr/avrlibc.h: Likewise.
2458 * config/bfin/bfin.h: Likewise.
2459 * config/bfin/linux.h: Likewise.
2460 * config/bfin/uclinux.h: Likewise.
2461 * config/darwin.h: Likewise.
2462 * config/darwin10.h: Likewise.
2463 * config/darwin12.h: Likewise.
2464 * config/gnu-user.h: Likewise.
2465 * config/lm32/uclinux-elf.h: Likewise.
2466 * config/pa/pa-hpux11.h: Likewise.
2467 * config/pa/pa64-hpux.h: Likewise.
2468 * config/sparc/sparc.h: Likewise.
2470 2018-07-31 Olivier Hainque <hainque@adacore.com>
2472 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2473 undefined variables.
2475 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2478 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2481 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2483 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2484 terminated string literal.
2486 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2488 PR rtl-optimization/85160
2489 * combine.c (is_just_move): New function.
2490 (try_combine): Allow combining two instructions into two if neither of
2491 the original instructions was a move.
2493 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2496 * doc/extend.texi (Global Register Variables): Discourage use of type
2498 (Local Register Variables): Likewise.
2500 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2502 PR tree-optimization/86506
2503 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2505 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2508 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2509 hard_regno, make sure no insn between `from` and `to` clobbers it.
2511 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2512 Tom de Vries <tdevries@suse.de>
2514 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2515 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2516 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2517 PTX_DEFAULT_RUNTIME_DIM.
2518 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2520 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2522 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2523 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2524 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2526 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2528 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2530 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2531 predicate to reflect toc fusion has been deleted.
2532 (toc_fusion_mem_raw): Delete, no longer used.
2533 (toc_fusion_mem_wrapped): Likewise.
2534 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2536 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2537 Delete, no longer used.
2538 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2539 meant to be used for toc fusion.
2540 (rs6000_debug_print_mode): Delete toc fusion debugging.
2541 (rs6000_debug_reg_global): Likewise.
2542 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2543 fusion and secondary reload support that were never used.
2544 (rs6000_option_override_internal): Delete TOC fusion, that was only
2545 partially defined, and it did not work unless you also used the
2547 (rs6000_legitimate_address_p): Delete TOC fusion support.
2548 (rs6000_opt_masks): Likewise.
2549 (fusion_wrap_memory_address): Delete function, no longer used.
2550 (fusion_split_address); Delete TOC fusion support.
2551 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2552 longer used with toc fusion being deleted.
2553 (TARGET_TOC_FUSION_FP): Likewise.
2554 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2556 (toc fusion spliter): Delete TOC fusion support.
2557 (toc_fusionload_<mode>): Likewise.
2558 (toc_fusionload_di): Likewise.
2559 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2560 longer needs to be named. Rename predicate to delete TOC fusion.
2561 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2562 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2563 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2564 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2565 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2567 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2569 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2570 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2571 __int128 in built-in function prototypes.
2572 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2573 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2575 2018-07-27 Martin Sebor <msebor@redhat.com>
2577 PR tree-optimization/86696
2578 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2579 types, including enums.
2580 (handle_char_store): Be prepared for the above function to fail.
2582 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2584 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2585 when optimization level is lower than 2 or optimize for size.
2587 2018-07-26 Martin Sebor <msebor@redhat.com>
2589 PR tree-optimization/86043
2590 PR tree-optimization/86042
2591 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2592 (get_string_cst_length): Rename...
2593 (get_min_string_length): ...to this. Add argument.
2594 (handle_char_store): Extend to handle multi-character stores by
2596 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2597 * tree.h (initializer_zerop): Add argument.
2599 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2602 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2603 declare target to variables if they have always,{to,from,tofrom} map
2606 2018-07-26 Martin Liska <mliska@suse.cz>
2609 * lto-wrapper.c: Add linker_output as prefix
2610 for ltrans_output_file.
2612 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2614 PR rtl-optimization/85805
2615 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2616 value for hard registers if that was written in the same mode.
2618 2018-07-26 Martin Liska <mliska@suse.cz>
2620 PR gcov-profile/86536
2621 * gcov.c (format_gcov): Use printf format %.*f directly
2622 and do not handle special values.
2624 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2626 * common/config/arc/arc-common.c (arc_option_optimization_table):
2627 Update default optimizations for size.
2629 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2631 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2632 * config/arc/constraints.md (CfZ): New constraint.
2633 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2634 (subsf3_fpu): Likewise.
2635 (cmpsf_fpu): Likewise.
2636 (cmpsf_fpu_uneq): Likewise.
2638 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2640 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2642 (arc_is_uncached_mem_p): uncached applies to both the variable and
2645 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2647 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2650 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2652 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2653 field "m_scopes" from vec to auto_vec.
2655 2018-07-25 Martin Liska <mliska@suse.cz>
2657 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2660 2018-07-25 Richard Biener <rguenther@suse.de>
2663 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2664 special wrt context_die late.
2665 (gen_subprogram_die): Re-use DIEs in local scope.
2667 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2669 PR tree-optimization/86644
2670 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2672 2018-07-25 Martin Liska <mliska@suse.cz>
2675 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2676 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2678 2018-07-25 Martin Liska <mliska@suse.cz>
2681 * params.def: Explain ASan abbreviation and provide
2682 a documentation link.
2684 2018-07-24 Martin Sebor <msebor@redhat.com>
2686 PR tree-optimization/86622
2687 PR tree-optimization/86532
2688 * builtins.h (string_length): Declare.
2689 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2690 (check_access): Be prepared for non-constant length ranges.
2691 (string_length): Make extern.
2692 * expr.c (string_constant): Only handle the minor non-constant
2693 array index. Use string_constant to compute the length of
2694 a generic string constant.
2696 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2698 PR tree-optimization/86618
2699 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2700 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2702 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2704 PR tree-optimization/86636
2705 * json.cc (json::object::set): Fix comment. Add assertions.
2706 (json::array::append): Move here from json.h. Add comment and an
2708 (json::string::string): Likewise.
2709 * json.h (json::array::append): Move to json.cc.
2710 (json::string::string): Likewise.
2711 * optinfo-emit-json.cc
2712 (optrecord_json_writer::impl_location_to_json): Assert that we
2713 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2714 wrapper around it. Expand the location once, rather than three
2716 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2717 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2719 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2720 places. Fix some overlong lines.
2722 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2724 * config/aarch64/aarch64-simd.md
2725 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2726 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2727 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2728 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2729 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2730 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2731 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2732 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2733 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2735 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2738 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2739 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2740 and abs_d == d, do the power of two handling if profitable.
2742 2018-07-24 Richard Biener <rguenther@suse.de>
2744 * match.pd: Add BIT_FIELD_REF canonicalizations.
2746 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2749 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2751 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2753 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2754 terminated STRING_CST object.
2756 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2758 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2760 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2762 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2763 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2764 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2765 the elements into a register.
2766 (rs6000_split_v4si_init_di_reg): Delete.
2767 (rs6000_split_v4si_init): Delete.
2768 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2769 (vsx_init_v4si): Rewrite as a define_expand.
2771 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2773 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2774 zero_extend argument from memory): New.
2776 2018-07-22 Martin Sebor <msebor@redhat.com>
2779 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2780 diagnosing calls with unknown arguments unless -Walloca-larger-than
2781 is restricted to less than PTRDIFF_MAX bytes.
2783 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2785 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2787 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2789 * pretty-print.c (text_info::set_location): Remove redundant
2790 "line_table" parameter from call to rich_location::set_range.
2792 2018-07-20 Martin Sebor <msebor@redhat.com>
2795 * builtins.c (expand_builtin_alloca): Adjust.
2796 * calls.c (alloc_max_size): Simplify.
2797 * cgraphunit.c (cgraph_node::expand): Adjust.
2798 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2800 (frame_larger_than_size): Same.
2801 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2802 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2804 * doc/invoke.texi (GCC Command Options): Document option arguments.
2805 Explain byte-size arguments and suffixes.
2806 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2807 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2808 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2809 * doc/options.texi (UInteger): Expand.
2810 (Host_Wide_Int, ByteSize): Document new properties.
2811 * final.c (final_start_function_1): Include sizes in an error message.
2812 * function.c (frame_offset_overflow): Same.
2813 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2814 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2815 Diagnose unbounded alloca calls only for limits of less than
2817 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2818 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2820 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2821 only for limits of less than PTRDIFF_MAX.
2822 * langhooks-def.h (lhd_handle_option): Change function argument
2824 * langhooks.c (lhd_handle_option): Same.
2825 * langhooks.h (handle_option): Same.
2826 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2828 (var_type, var_type_struct): Same.
2829 (var_set): Handle ByteSize flag.
2830 * optc-gen.awk: Add comments to output to ease debugging. Make
2831 use of HOST_WIDE_INT where appropriate.
2832 * opts-gen-save.awk: Use %lx to format unsigned long.
2833 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2834 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2835 arguments. Parse bytes-size suffixes.
2836 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2837 (enum_value_to_arg): Same.
2838 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2839 (handle_option): Adjust.
2840 (generate_option): Change function argument to HOST_WIDE_INT.
2841 (cmdline_handle_error): Adjust.
2842 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2843 (set_option): Change function argument to HOST_WIDE_INT.
2844 (option_enabled): Handle cl_host_wide_int.
2845 (get_option_state): Handle CLVC_SIZE.
2846 (control_warning_option): Same.
2847 * opts.c (common_handle_option): Change function argument to
2848 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2849 OPT_Wvla_larger_than_.
2850 * opts.h (enum cl_var_type): Add an enumerator.
2851 * stor-layout.c (layout_decl): Print a more meaningful warning.
2852 * toplev.c (output_stack_usage): Adjust.
2854 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
2856 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2857 call to inline_expand_builtin_string_cmp.
2858 (expand_builtin_strcmp): Likewise.
2859 (expand_builtin_strncmp): Likewise.
2860 (inline_string_cmp): Delete the last parameter, change char_type_node
2861 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2863 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2864 the inlining expansion on target where the type of the call has same or
2865 narrower precision than unsigned char.
2867 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2869 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2870 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2871 * common.opt (fsave-optimization-record): New option.
2872 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2873 * doc/invoke.texi (-fsave-optimization-record): New option.
2874 * dumpfile.c: Include "optinfo-emit-json.h".
2875 (struct kv_pair): Move to coretypes.h.
2876 (optgroup_options): Make non-static.
2877 (dump_context::end_scope): Call
2878 optimization_records_maybe_pop_dump_scope.
2879 * dumpfile.h (optgroup_options): New decl.
2880 * json.cc: New file.
2882 * optinfo-emit-json.cc: New file.
2883 * optinfo-emit-json.h: New file.
2884 * optinfo.cc: Include "optinfo-emit-json.h".
2885 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2886 (optinfo_enabled_p): Check optimization_records_enabled_p.
2887 (optinfo_wants_inlining_info_p): Likewise.
2888 * optinfo.h: Update comment.
2889 * profile-count.c (profile_quality_as_string): New function.
2890 * profile-count.h (profile_quality_as_string): New decl.
2891 (profile_count::quality): New accessor.
2892 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2893 and optinfo_emit_json_cc_tests.
2894 * selftest.h (selftest::json_cc_tests): New decl.
2895 (selftest::optinfo_emit_json_cc_tests): New decl.
2896 * toplev.c: Include "optinfo-emit-json.h".
2897 (compile_file): Call optimization_records_finish.
2898 (do_compile): Call optimization_records_start.
2899 * tree-ssa-live.c: Include optinfo.h.
2900 (remove_unused_scope_block_p): Retain inlining information if
2901 optinfo_wants_inlining_info_p returns true.
2903 2018-07-20 Richard Biener <rguenther@suse.de>
2906 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2907 to cover -flto-partition=none.
2909 2018-07-20 Martin Liska <mliska@suse.cz>
2911 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2912 (get_decl_source_range): Remove unused function.
2914 2018-07-20 Richard Biener <rguenther@suse.de>
2916 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2917 (struct vn_phi_s): Likewise.
2918 (struct vn_reference_s): Likewise.
2919 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2920 for searching the slot of an entry known to be in the hash itself.
2921 (vn_phi_hasher::equal): Likewise.
2922 (vn_reference_hasher::equal): Likewise.
2923 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2925 (optimistic_info, current_info): Remove, keeping only valid_info.
2926 (vn_reference_lookup_1): Remove fallback lookup.
2927 (vn_reference_lookup_2): Likewise.
2928 (vn_nary_op_lookup_1): Likewise.
2929 (vn_phi_lookup): Likewise.
2930 (vn_nary_build_or_lookup_1): Make sure to not chain the built
2932 (vn_reference_insert): Adjust, chain the inserted hash element
2933 at last_inserted_ref.
2934 (vn_reference_insert_pieces): Likewise.
2935 (visit_reference_op_call): Likewise.
2936 (vn_nary_op_insert_into): Chain the inserted hash element at
2938 (vn_nary_op_insert_pieces): Adjust.
2939 (vn_nary_op_insert): Likewise.
2940 (vn_nary_op_insert_stmt): Likewise.
2941 (vn_phi_insert): Adjust, chain the inserted hash element at
2943 (process_scc): Remove clearing and copying the optimistic
2944 table. Instead remove elements inserted during an optimistic
2945 iteration from the single table we maintain.
2946 (init_scc_vn): Adjust.
2947 (free_scc_vn): Likewise.
2948 (sccvn_dom_walker::record_cond): Likewise.
2949 (sccvn_dom_walker::after_dom_children): Likewise.
2951 2018-07-19 Martin Sebor <msebor@redhat.com>
2953 PR tree-optimization/84047
2954 PR tree-optimization/83776
2955 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2956 (check_array_bounds): Call it.
2958 2018-07-19 Martin Sebor <msebor@redhat.com>
2960 * align.h (align_flags): Use member initialization.
2962 2018-07-19 David Malcolm <dmalcolm@redhat.com>
2964 * Makefile.in (OBJS): Add optinfo.o.
2965 * coretypes.h (class symtab_node): New forward decl.
2966 (struct cgraph_node): New forward decl.
2967 (class varpool_node): New forward decl.
2968 * dump-context.h: New file.
2969 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
2971 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
2972 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
2973 (set_alt_dump_file): Likewise.
2974 (dump_context::~dump_context): New dtor.
2975 (dump_gimple_stmt): Move implementation to...
2976 (dump_context::dump_gimple_stmt): ...this new member function.
2977 Add the stmt to any pending optinfo, creating one if need be.
2978 (dump_gimple_stmt_loc): Move implementation to...
2979 (dump_context::dump_gimple_stmt_loc): ...this new member function.
2980 Start a new optinfo and add the stmt to it.
2981 (dump_gimple_expr): Move implementation to...
2982 (dump_context::dump_gimple_expr): ...this new member function.
2983 Add the stmt to any pending optinfo, creating one if need be.
2984 (dump_gimple_expr_loc): Move implementation to...
2985 (dump_context::dump_gimple_expr_loc): ...this new member function.
2986 Start a new optinfo and add the stmt to it.
2987 (dump_generic_expr): Move implementation to...
2988 (dump_context::dump_generic_expr): ...this new member function.
2989 Add the tree to any pending optinfo, creating one if need be.
2990 (dump_generic_expr_loc): Move implementation to...
2991 (dump_context::dump_generic_expr_loc): ...this new member
2992 function. Add the tree to any pending optinfo, creating one if
2994 (dump_printf): Move implementation to...
2995 (dump_context::dump_printf_va): ...this new member function. Add
2996 the text to any pending optinfo, creating one if need be.
2997 (dump_printf_loc): Move implementation to...
2998 (dump_context::dump_printf_loc_va): ...this new member function.
2999 Start a new optinfo and add the stmt to it.
3000 (dump_dec): Move implementation to...
3001 (dump_context::dump_dec): ...this new member function. Add the
3002 value to any pending optinfo, creating one if need be.
3003 (dump_context::dump_symtab_node): New member function.
3004 (dump_context::get_scope_depth): New member function.
3005 (dump_context::begin_scope): New member function.
3006 (dump_context::end_scope): New member function.
3007 (dump_context::ensure_pending_optinfo): New member function.
3008 (dump_context::begin_next_optinfo): New member function.
3009 (dump_context::end_any_optinfo): New member function.
3010 (dump_context::s_current): New global.
3011 (dump_context::s_default): New global.
3012 (dump_scope_depth): Delete global.
3013 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3014 (dump_symtab_node): New function.
3015 (get_dump_scope_depth): Reimplement in terms of dump_context.
3016 (dump_begin_scope): Likewise.
3017 (dump_end_scope): Likewise.
3018 (selftest::temp_dump_context::temp_dump_context): New ctor.
3019 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3020 (selftest::verify_item): New function.
3021 (ASSERT_IS_TEXT): New macro.
3022 (ASSERT_IS_TREE): New macro.
3023 (ASSERT_IS_GIMPLE): New macro.
3024 (selftest::test_capture_of_dump_calls): New test.
3025 (selftest::dumpfile_c_tests): Call it.
3026 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3027 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3028 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3029 descriptive comment.
3030 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3031 (dump_node, dump_bb): Move these unrelated decls.
3032 (class dump_manager): Add leading comment.
3033 * optinfo.cc: New file.
3034 * optinfo.h: New file.
3036 2018-07-19 Michael Collison <michael.collison@arm.com>
3037 Richard Henderson <rth@redhat.com>
3039 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3040 (subti): Handle op1 zero.
3041 (subvti4, usub4ti4): New.
3042 (*sub<GPI>3_compare1_imm): New.
3043 (sub<GPI>3_carryinCV): New.
3044 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3045 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3047 2018-07-19 Michael Collison <michael.collison@arm.com>
3048 Richard Henderson <rth@redhat.com>
3050 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3051 (addti3): Create simpler code if low part is already known to be 0.
3052 (addvti4, uaddvti4): New.
3053 (*add<GPI>3_compareC_cconly_imm): New.
3054 (*add<GPI>3_compareC_cconly): New.
3055 (*add<GPI>3_compareC_imm): New.
3056 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3057 handle constants within this pattern..
3058 (*add<GPI>3_compareV_cconly_imm): New.
3059 (*add<GPI>3_compareV_cconly): New.
3060 (*add<GPI>3_compareV_imm): New.
3061 (add<GPI>3_compareV): New.
3062 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3063 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3064 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3065 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3066 with 'comparison' operator.
3067 (*add<GPI>3_compareV_cconly_imm): Ditto.
3068 (*add<GPI>3_compareV_cconly): Ditto.
3069 (*add<GPI>3_compareV_imm): Ditto.
3070 (add<GPI>3_compareV): Ditto.
3071 (add<mode>3_carryinC): Ditto.
3072 (*add<mode>3_carryinC_zero): Ditto.
3073 (*add<mode>3_carryinC): Ditto.
3074 (add<mode>3_carryinV): Ditto.
3075 (*add<mode>3_carryinV_zero): Ditto.
3076 (*add<mode>3_carryinV): Ditto.
3078 2018-07-19 Michael Collison <michael.collison@arm.com>
3079 Richard Henderson <rth@redhat.com>
3081 * config/aarch64/aarch64-modes.def (CC_V): New.
3082 * config/aarch64/aarch64-protos.h
3083 (aarch64_addti_scratch_regs): Declare
3084 (aarch64_subvti_scratch_regs): Declare.
3085 (aarch64_expand_subvti): Declare.
3086 (aarch64_gen_unlikely_cbranch): Declare
3087 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3088 for signed overflow using CC_Vmode.
3089 (aarch64_get_condition_code_1): Handle CC_Vmode.
3090 (aarch64_gen_unlikely_cbranch): New function.
3091 (aarch64_addti_scratch_regs): New function.
3092 (aarch64_subvti_scratch_regs): New function.
3093 (aarch64_expand_subvti): New function.
3095 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3097 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3099 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3100 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3103 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3106 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3107 address check not strict.
3109 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3111 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3113 (store_pair_lanes<mode>): Likewise.
3114 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3115 enum value 'ADDR_QUERY_LDP_STP_N'.
3116 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3117 (aarch64_print_address_internal): Add declaration.
3118 (aarch64_print_ldpstp_address): Remove.
3119 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3120 (aarch64_print_operand): Change printing of 'y'.
3121 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3122 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3123 'true' rather than '1'.
3124 * config/aarch64/constraints.md (Uml): Likewise.
3125 (Uml): Rename to Umn.
3128 2018-07-19 Richard Biener <rguenther@suse.de>
3130 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3132 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3133 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3134 (vn_reference_hasher): Likewise.
3135 (struct vn_tables_s): Remove obstack and alloc-pool members.
3136 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3137 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3138 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3139 (vn_reference_insert_pieces): Likewise.
3140 (alloc_vn_nary_op_noinit): Adjust.
3141 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3142 (vn_phi_eq): Adjust.
3143 (shared_lookup_phiargs): Remove.
3144 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3145 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3146 (visit_reference_op_call): Likewise.
3147 (copy_nary, copy_phi, copy_reference): Remove.
3148 (process_scc): Rewind the obstack when iterating. Do not
3149 copy the elements to valid_info but just move them from one
3150 hashtable to the other.
3151 (allocate_vn_table): Adjust.
3152 (free_vn_table): Likewise.
3153 (init_scc_vn): Likewise.
3154 (free_scc_vn): Likewise.
3156 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3159 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3160 indirect_return as function type attribute.
3161 (ix86_attribute_table): Change indirect_return to function
3163 * doc/extend.texi: Update indirect_return attribute.
3165 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3167 * wide-int.h (widest2_int): New.
3168 * gimple-fold.c (arith_overflowed_p): Use it.
3169 * tree.h (widest2_int_cst): New.
3170 * tree-vrp.c (wide_int_binop_overflow): Rename from
3171 vrp_int_const_binop.
3172 Rewrite to work on trees.
3173 (extract_range_from_multiplicative_op_1): Abstract code to...
3174 (wide_int_range_min_max): ...here.
3175 (wide_int_range_cross_product): ...and here.
3176 (extract_range_from_binary_expr_1): Abstract overflow code to...
3177 (wide_int_range_mult_wrapping): ...here.
3178 * tree-vrp.h (wide_int_range_cross_product): New.
3179 (wide_int_range_mult_wrapping): New.
3181 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3182 Julia Koval <julia.koval@intel.com>
3184 * config/i386/x86-tune-costs.h (skylake_memcpy,
3185 skylake_memset): Replace rep_prefix with unrolling for size 512.
3187 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3190 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3191 comparision with EQ_EXPR in last stmt.
3193 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3195 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3196 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3197 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3198 previously known as "PowerPC AltiVec Built-in Functions". Move
3199 some material to new subsubsections "PowerPC AltiVec Built-in
3200 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3202 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3203 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3204 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3205 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3207 2018-07-18 Richard Biener <rguenther@suse.de>
3209 PR tree-optimization/86557
3210 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3213 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3215 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3217 2018-07-17 Jeff Law <law@redhat.com>
3219 * config/arm/arm.c (get_label_padding): Update for recent
3220 changes to label_to_alignment.
3222 PR tree-optimization/86010
3223 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3225 * config/mips/mips.c (vr4130_align_insns): Update for recent
3226 changes to label_to_alignment.
3228 * config/frv/frv.c (frv_label_align): Update for recent changes
3229 to label_to_alignment.
3231 * config/nios2/nios2.c (nios2_label_align): Update for recent
3232 changes which dropped ALIGN_LABELS_LOG.
3234 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3236 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3237 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3239 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3241 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3242 deprecated align_labels_log.
3244 2018-07-17 Richard Biener <rguenther@suse.de>
3247 * dwarf2out.c (init_sections_and_labels): Always generate
3248 a debug_line_str_section for early LTO debug.
3249 (dwarf2out_finish): Reset debug_line_str_hash output early.
3250 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3252 (dwarf2out_early_finish): Output debug_line_str.
3254 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3256 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3257 index register on z196 or later.
3259 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3261 * config/s390/s390.c (s390_default_align): Set default function
3263 (s390_override_options_after_change): Call s390_default align.
3264 (s390_option_override_internal): Call s390_default_align.
3265 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3267 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3270 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3271 to _looptemp_ clauses, other than the first two.
3273 2018-07-17 Martin Liska <mliska@suse.cz>
3275 * opts.c: Do not enable OPT_falign_* for -Os.
3277 2018-07-17 Martin Liska <mliska@suse.cz>
3279 * align.h (MAX_CODE_ALIGN): New.
3280 (MAX_CODE_ALIGN_VALUE): New.
3281 * common/config/i386/i386-common.c (ix86_handle_option):
3282 (MAX_CODE_ALIGN): Moved to align.h.
3283 * final.c (MAX_CODE_ALIGN): Likewise.
3284 * opts.c (parse_and_check_align_values):
3285 (MAX_CODE_ALIGN): Likewise.
3286 (MAX_CODE_ALIGN_VALUE): Likewise.
3288 2018-07-17 Martin Liska <mliska@suse.cz>
3290 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3291 in order to fulfil coding style.
3292 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3293 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3294 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3295 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3296 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3297 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3298 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3299 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3300 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3302 2018-07-17 Martin Liska <mliska@suse.cz>
3304 * align.h: New file.
3305 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3307 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3308 align_flags of label_to_alignment.
3309 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3311 * config/m68k/m68k.c: Do not use removed align_labels_value and
3313 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3314 (LOOP_ALIGN): Likewise.
3315 (LABEL_ALIGN): Likewise.
3316 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3317 Remove not used macro.
3318 (rs6000_loop_align): Change return type to align_flags.
3319 (rs6000_loop_align_max_skip): Remove.
3320 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3321 Change return type to align_flags.
3322 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3323 Remove not used macro.
3324 (rs6000_loop_align): Change return type to align_flags.
3325 (rs6000_loop_align_max_skip): Remove.
3326 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3327 * config/rx/rx-protos.h (rx_align_for_label): Make it
3329 * config/rx/rx.c (rx_align_for_label): Change return type
3331 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3333 into align_flags class.
3334 (LABEL_ALIGN): Likewise.
3335 (LOOP_ALIGN): Likewise.
3336 * config/s390/s390.c (s390_label_align): Use align_flags
3338 (s390_asm_output_function_label): Likewise.
3339 * config/sh/sh.c (sh_override_options_after_change):
3340 Use align_flags class directly without macros.
3341 (find_barrier): Likewise.
3342 (barrier_align): Likewise.
3343 (sh_loop_align): Likewise.
3344 * config/spu/spu.c (spu_option_override):
3345 Use align_flags_tuple::get_value instead of removed macros.
3346 (spu_sched_init): Likewise.
3347 * config/spu/spu.h (GTY): Likewise.
3348 * config/visium/visium.c (visium_option_override):
3349 Set "8" as default secondary alignment.
3350 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3351 in order to guarantee secondary alignment of 8.
3352 * coretypes.h: Include align.h header file.
3353 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3354 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3355 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3356 * doc/tm.texi.in: Likewise.
3357 * final.c (struct label_alignment): Remove not used structure.
3358 (LABEL_ALIGN): Change type to align_flags.
3359 (LOOP_ALIGN): Likewise.
3360 (JUMP_ALIGN): Likewise.
3361 (default_loop_align_max_skip): Remove.
3362 (default_label_align_max_skip): Likewise.
3363 (default_jump_align_max_skip): Likewise.
3364 (default_label_align_after_barrier_max_skip):
3365 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3366 (LABEL_TO_MAX_SKIP): Remove.
3367 (label_to_alignment): Return align_flags type instead of integer.
3368 (label_to_max_skip): Remove.
3369 (align_fuzz): Use align_flags type.
3370 (compute_alignments): Use align_flags type and use align_flags::max
3371 to combine multiple alignments.
3372 (grow_label_align): Grow vec instead of C array.
3373 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3374 (shorten_branches): Use align_flags type and use align_flags::max
3375 to combine multiple alignments.
3376 (final_scan_insn_1): Remove usage of secondary alignment that comes
3377 from label alignment, but instead use proper secondary alignment
3378 which is computed in grow_label_align.
3379 * flags.h (struct align_flags_tuple): Move to align.h.
3380 (struct align_flags): Likewise.
3381 (state_align_loops): Rename to align_loops.
3382 (state_align_jumps): Rename to align_jumps.
3383 (state_align_labels): Rename to align_labels.
3384 (state_align_functions): Rename to align_functions.
3385 (align_loops_log): Remove.
3386 (align_jumps_log): Remove.
3387 (align_labels_log): Remove.
3388 (align_functions_log): Remove.
3389 (align_loops_max_skip): Remove.
3390 (align_jumps_max_skip): Remove.
3391 (align_labels_max_skip): Remove.
3392 (align_functions_max_skip): Remove.
3393 (align_loops_value): Remove.
3394 (align_jumps_value): Remove.
3395 (align_labels_value): Remove.
3396 (align_functions_value): Remove.
3397 * output.h (label_to_alignment): Change return type to align_flags.
3398 (label_to_max_skip): Remove.
3399 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3400 jump_align_max_skip macros.
3401 * targhooks.h (default_loop_align_max_skip): Remove.
3402 (default_label_align_max_skip): Likewise.
3403 (default_jump_align_max_skip): Likewise.
3404 (default_label_align_after_barrier_max_skip): Remove.
3405 * toplev.c (read_log_maxskip): Use ::normalize function.
3406 (parse_N_M): Remove not used argument and also call ::normalize.
3407 (parse_alignment_opts): Do not pass unused arguments.
3408 * varasm.c (assemble_start_function): Use directly align_functions
3409 instead of removed macros.
3410 * system.h: Do not poison removed macros.
3412 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3415 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3416 and cond temporaries don't have reference type if iterator has
3417 pointer type. For init use &for_pre_body instead of pre_p if
3418 for_pre_body is non-empty.
3420 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3422 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3423 double-double modes to SFmode directly directly.
3424 (trunc<mode>sf2_fprs): Delete.
3426 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3428 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3429 for conversions between IFmode and the decimal floating point modes.
3430 (init_float128_ieee): Use the correct names for conversions between
3431 KFmode and the decimal floating point modes.
3433 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3435 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3436 for the conversions between TDmode and IFmode.
3437 (init_float128_ieee): Use more correct names for the conversions
3438 between TDmode and KFmode.
3440 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3442 PR tree-optimization/86526
3443 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3444 (inline_expand_builtin_string_cmp): Likewise.
3445 (inline_string_cmp): Likewise. Use c_readstr instead of
3446 builtin_memcpy_read_str. Add unit_mode temporary.
3448 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3451 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3452 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3454 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3456 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3457 Alphabetize prototypes of built-in functions, separating out
3458 built-in functions that are listed in this section but should be
3459 described elsewhere.
3461 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3464 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3465 conditional move when trapping comparison was split to a
3466 non-trapping one (and vice versa).
3468 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3470 * config/s390/s390.c (s390_function_profiler): Generate nops
3471 instead of profiler call sequences.
3472 * config/s390/s390.opt: Add the new option.
3474 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3476 * config/s390/s390.c (s390_function_profiler): Generate
3477 __mcount_loc section.
3478 * config/s390/s390.opt: Add the new option.
3480 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3482 * common.opt: Add the new warning.
3483 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3484 %r0,__fentry__" when -mfentry is specified.
3485 (s390_option_override_internal): Disallow -mfentry for 31-bit
3487 * config/s390/s390.opt: Add the new option.
3489 2018-07-16 Richard Biener <rguenther@suse.de>
3492 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3493 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3495 2018-07-16 Martin Liska <mliska@suse.cz>
3498 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3501 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3503 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3505 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3507 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3509 * config/arc/arc-cpus.def (hs4x): New cpu.
3511 * config/arc/arc-tables.opt: Regenerate.
3512 * config/arc/arc.c (arc_sched_issue_rate): New function.
3513 (TARGET_SCHED_ISSUE_RATE): Define.
3514 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3515 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3517 (attr tune): Add ARCHS4x tune values.
3518 (attr tune_dspmpy): Define.
3519 (*tst): Correct instruction type.
3520 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3521 * config/arc/arcHS4x.md: New file.
3522 * config/arc/fpu.md: Update instruction type attributes.
3523 * config/arc/t-multilib: Regenerate.
3525 2018-07-16 Tom de Vries <tdevries@suse.de>
3528 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3530 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3532 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3535 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3537 PR tree-optimization/86514
3538 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3539 conversion to a boolean type from a type with greater precision.
3541 2018-07-16 Tom de Vries <tdevries@suse.de>
3543 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3544 dump_flags request TDF_SLIM.
3546 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3548 * fold-const.c (int_const_binop_1): Abstract...
3549 (wide_int_binop): ...wide int code here.
3550 (poly_int_binop): ...poly int code here.
3551 Abstract the rest of int_const_binop_1 into int_const_binop.
3552 * fold-const.h (wide_int_binop): New.
3553 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3554 Remove useless PLUS/MINUS_EXPR case.
3555 (zero_nonzero_bits_from_vr): Move wide int code...
3556 (zero_nonzero_bits_from_bounds): ...here.
3557 (extract_range_from_binary_expr_1): Move mask optimization code...
3558 (range_easy_mask_min_max): ...here.
3559 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3560 (range_easy_mask_min_max): New.
3562 2018-07-15 Jeff Law <law@redhat.com>
3565 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3568 2018-07-14 Jim Wilson <jimw@sifive.com>
3570 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3572 2018-07-14 Paul Koning <ni1d@arrl.net>
3574 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3576 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3578 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3581 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3582 Steve Munroe <munroesj52@gmail.com>
3584 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3585 (_mm_andnot_si128): Likewise.
3586 (_mm_or_si128): Likewise.
3587 (_mm_xor_si128): Likewise.
3589 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3592 * builtins.c (expand_builtin_memcmp): Inline the calls first
3593 when result_eq is false.
3594 (expand_builtin_strcmp): Inline the calls first.
3595 (expand_builtin_strncmp): Likewise.
3596 (inline_string_cmp): New routine. Expand a string compare
3597 call by using a sequence of char comparison.
3598 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3599 a call to str(n)cmp/memcmp.
3600 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3602 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3604 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3606 * config/arm/driver-arm.c: Include arm-native.h.
3607 (host_detect_local_cpu): Use auto-generated data tables.
3608 (vendors, arm_cpu_table): Delete. Move part information to ...
3609 * config/arm/arm-cpus.in: ... here.
3610 * config/arm/parsecpu.awk (gen_native): New function.
3611 (vendor, part): New CPU fields.
3612 (END): Add support for building the native CPU detection tables.
3613 * config/arm/t-arm (arm-native.h): Add build rule.
3614 (driver-arm.o): Add dependency on arm-native.h.
3616 2018-07-13 Richard Biener <rguenther@suse.de>
3619 * match.pd (addr1 - addr2): Allow either of the operand to
3622 2018-07-13 Tom de Vries <tdevries@suse.de>
3624 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3627 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3629 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3630 arrays instead of numbered variables.
3632 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3634 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3635 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3637 2018-07-13 Richard Biener <rguenther@suse.de>
3640 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3641 instead of get_context_die.
3643 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3644 Richard Biener <rguenther@suse.de>
3647 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3648 that the loop latch destination where phi is defined.
3650 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3652 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3653 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3654 (riscv_set_current_function): Extract getting interrupt type to new
3656 (riscv_get_interrupt_type): New function.
3657 (riscv_merge_decl_attributes): New function, checking interrupt type is
3659 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3661 2018-07-12 Paul Koning <ni1d@arrl.net>
3663 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3666 2018-07-12 Paul Koning <ni1d@arrl.net>
3668 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3669 zero reference, add doloop_end instead.
3670 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3671 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3672 detail for doloop_end.
3674 2018-07-12 Martin Sebor <msebor@redhat.com>
3677 * attribs.c (decl_attributes): Reject conflicting attributes before
3678 calling attribute handlers.
3680 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3682 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3684 (gcc::dump_manager::get_dump_file_name): likewise.
3685 (dump_begin): Likewise.
3686 * dumpfile.h (dump_begin): Update prototype.
3687 (gcc::dump_manager::get_dump_file_name,
3688 gcc::dump_manager::get_dump_file_name): Update prototype.
3690 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3692 * internal-fn.h (vectorizable_internal_fn_p): New function.
3693 * tree-vect-slp.c (compatible_calls_p): Likewise.
3694 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3695 to internal functions.
3696 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3698 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3700 * fold-const.h (inverse_conditions_p): Declare.
3701 * fold-const.c (inverse_conditions_p): New function.
3702 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3703 that test the inverse condition of a conditional internal function.
3704 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3705 * internal-fn.c (internal_fn_mask_index): Handle conditional
3707 (vectorized_internal_fn_supported_p): New function.
3708 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3709 (any_pred_load_store): Replace with...
3710 (need_to_predicate): ...this new variable.
3711 (redundant_ssa_names): New variable.
3712 (ifcvt_can_use_mask_load_store): Move initial checks to...
3713 (ifcvt_can_predicate): ...this new function. Handle tree codes
3714 for which a conditional internal function exists.
3715 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3716 instead of ifcvt_can_use_mask_load_store. Update after variable
3718 (predicate_load_or_store): New function, split out from
3719 predicate_mem_writes.
3720 (check_redundant_cond_expr): New function.
3721 (value_available_p): Likewise.
3722 (predicate_rhs_code): Likewise.
3723 (predicate_mem_writes): Rename to...
3724 (predicate_statements): ...this. Use predicate_load_or_store
3725 and predicate_rhs_code.
3726 (combine_blocks, tree_if_conversion): Update after above name changes.
3727 (ifcvt_local_dce): Handle redundant_ssa_names.
3728 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3729 general conditional functions.
3730 * tree-vect-stmts.c (vectorizable_call): Likewise.
3732 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3733 Alan Hayward <alan.hayward@arm.com>
3734 David Sherwood <david.sherwood@arm.com>
3736 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3737 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3738 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3739 plus and minus and convert them into IFN_COND_FMA-based sequences.
3740 (convert_mult_to_fma): Handle conditional plus and minus.
3742 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3744 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3745 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3746 (cond_fnms_optab): New optabs.
3747 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3749 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3750 * internal-fn.h (get_conditional_internal_fn): Declare.
3751 (get_unconditional_internal_fn): Likewise.
3752 * internal-fn.c (cond_ternary_direct): New macro.
3753 (expand_cond_ternary_optab_fn): Likewise.
3754 (direct_cond_ternary_optab_supported_p): Likewise.
3755 (FOR_EACH_COND_FN_PAIR): Likewise.
3756 (get_conditional_internal_fn): New function.
3757 (get_unconditional_internal_fn): Likewise.
3758 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3759 (gimple_match_op::gimple_match_op): Add a new overload for 5
3761 (gimple_match_op::set_op): Likewise.
3762 (gimple_resimplify5): Declare.
3763 * genmatch.c (decision_tree::gen): Generate simplifications for
3765 * gimple-match-head.c (gimple_simplify): Define an overload for
3766 5 operands. Handle calls with 5 arguments in the top-level overload.
3767 (convert_conditional_op): Handle conversions from unconditional
3768 internal functions to conditional ones.
3769 (gimple_resimplify5): New function.
3770 (build_call_internal): Pass a fifth operand.
3771 (maybe_push_res_to_seq): Likewise.
3772 (try_conditional_simplification): Try converting conditional
3773 internal functions to unconditional internal functions.
3774 Handle 3-operand unconditional forms.
3775 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3776 Define ternary equivalents of the current rules for binary conditional
3778 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3780 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3781 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3782 (optab): Handle them.
3783 (SVE_COND_FP_TERNARY): New int iterator.
3784 (sve_fmla_op, sve_fmad_op): New int attributes.
3785 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3786 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3787 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3789 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3791 * target.def (preferred_else_value): New target hook.
3792 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3793 * doc/tm.texi: Regenerate.
3794 * targhooks.h (default_preferred_else_value): Declare.
3795 * targhooks.c (default_preferred_else_value): New function.
3796 * internal-fn.h (conditional_internal_fn_code): Declare.
3797 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3798 (get_conditional_internal_fn): Use it.
3799 (conditional_internal_fn_code): New function.
3800 * gimple-match.h (gimple_match_cond): New struct.
3801 (gimple_match_op): Add a cond member function.
3802 (gimple_match_op::gimple_match_op): Update all forms to take a
3804 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3805 for the suboperation, but don't specify a particular else_value.
3806 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3807 (visit_nary_op, visit_reference_op_load): Pass
3808 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3809 * gimple-match-head.c: Include tree-eh.h
3810 (convert_conditional_op): New function.
3811 (maybe_resimplify_conditional_op): Likewise.
3812 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3813 (gimple_resimplify2): Likewise.
3814 (gimple_resimplify3): Likewise.
3815 (gimple_resimplify4): Likewise.
3816 (maybe_push_res_to_seq): Return null for conditional operations.
3817 (try_conditional_simplification): New function.
3818 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3820 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3822 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3824 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3826 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3828 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3830 (hash_tree): Do not hash DECL_FCONTEXT
3831 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3832 Do not stream DECL_FCONTEXT.
3833 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3834 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3836 2018-07-12 Richard Biener <rguenther@suse.de>
3839 * dwarf2out.c (gen_block_die): Only output blocks when they have
3840 at least one !DECL_IGNORED_P variable.
3842 2018-07-12 Richard Biener <rguenther@suse.de>
3845 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3846 Remove -mieee-fp handling.
3848 2018-07-12 Richard Biener <rguenther@suse.de>
3850 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3851 left-over from last patch.
3853 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3855 PR tree-optimization/86492
3856 * gimple-ssa-store-merging.c
3857 (imm_store_chain_info::coalesce_immediate_stores): Call
3858 check_no_overlap even for the merge_overlapping case. Formatting fix.
3860 2018-07-12 Richard Biener <rguenther@suse.de>
3863 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3864 move possibly trapping operations into the conditional.
3866 2018-07-12 Richard Biener <rguenther@suse.de>
3868 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3869 (vn_lookup_simplify_result): Remove recursion limit applied
3871 (vn_nary_build_or_lookup_1): Adjust.
3872 (try_to_simplify): Likewise.
3873 * gimple-match-head.c (gimple_resimplify1): Instead apply one
3875 (gimple_resimplify2): Likewise.
3876 (gimple_resimplify3): Likewise.
3877 (gimple_resimplify4): Likewise.
3879 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3881 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3882 Use __mmask64 type instead of __mmask8 for __M argument.
3883 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3884 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3886 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3887 __mmask16 for __M argument.
3888 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3889 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3890 to __mmask16 instead of __mmask8.
3891 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3892 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3893 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3894 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3895 instead of __mmask16 for __U argument.
3896 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3897 __mmask16 instead of __mmask8 for __U argument.
3898 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3900 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3902 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3904 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3906 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3908 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3910 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3912 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3914 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3916 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3918 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3919 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3920 return type as well as __M argument type and all casts from __mmask8
3922 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3923 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3924 return type as well as __M argument type and all casts from __mmask8
3926 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3927 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3928 return type as well as __M argument type and all casts from __mmask8
3930 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3931 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3932 return type as well as __M argument type and all casts from __mmask8
3934 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3935 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3938 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3940 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3941 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3944 2018-07-11 Paul Koning <ni1d@arrl.net>
3946 * doc/md.texi (define_subst): Document how multiple occurrences of
3947 the same argument in the replacement pattern are handled.
3949 2018-07-11 Paul Koning <ni1d@arrl.net>
3951 * doc/extend.texi (Common Variable Attributes): Move "mode" into
3953 (Common Type Attributes): Add "mode" attribute.
3955 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
3957 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3958 stream DECL_ORIGINAL_TYPE.
3959 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3960 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3961 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3962 Do not walk original type.
3963 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3965 (write_ts_decl_non_common_tree_pointers): Do not stream
3967 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
3968 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
3970 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3972 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
3973 thread twice from the same starting edge.
3975 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3977 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
3978 * gimple.c (gimple_call_nonnull_result_p): ...here...
3979 (gimple_call_nonnull_arg): ...and here.
3980 * gimple.h (gimple_call_nonnull_result_p): New.
3981 (gimple_call_nonnull_arg): New.
3983 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
3985 * config/arm/arm-cpus.in: Move information from fpu field of each
3986 cpu definition to the isa field.
3987 * config/arm/parsecpu.awk (fpu): Delete match rule.
3988 (gen_comm_data): Don't add bits from the CPU's FPU entry.
3990 2018-07-11 Richard Biener <rguenther@suse.de>
3993 * dwarf2out.c (init_sections_and_labels): Use
3994 output_asm_line_debug_info consistently.
3995 (dwarf2out_early_finish): Likewise.
3996 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3999 2018-07-11 Richard Biener <rguenther@suse.de>
4001 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4002 Rework father_bb setting in a way to avoid propagating constants
4003 multiple times on a loop body.
4005 2018-07-10 Mark Wielaard <mark@klomp.org>
4008 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4010 2018-07-10 Richard Biener <rguenther@suse.de>
4012 * hash-map.h (hash_map::iterator::operator*): Return
4013 references to key and value.
4015 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4018 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4019 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4020 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4021 (gimplify_omp_for): For composite loops, move outer
4022 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4023 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4024 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4025 TREE_LIST for both the original class iterator and the "last" helper
4026 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4027 loop, remember has_decl_expr from outer composite loops for the
4028 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4030 2018-07-09 Martin Sebor <msebor@redhat.com>
4034 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4035 accessing implicitly initialized array elements.
4036 * expr.c (string_constant): Handle string initializers of
4037 character arrays within aggregates.
4038 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4039 Store element offset. As a special case, handle zero size.
4040 (fold_nonarray_ctor_reference): Same.
4041 (fold_ctor_reference): Add argument. Store subobject offset.
4042 * gimple-fold.h (fold_ctor_reference): Add argument.
4044 2018-07-09 Paul Koning <ni1d@arrl.net>
4046 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4047 (pdp11_insn_cost): New function.
4048 (pdp11_md_asm_adjust): New function.
4049 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4050 (pdp11_rtx_costs): Update to match machine better.
4051 (output_addr_const_pdp11): Correct format mismatch warnings.
4052 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4053 * config/pdp11/pdp11.md: General change to add base_cost and/or
4054 length attributes for use by new pdp11_insn_cost function.
4055 (MIN_BRANCH): Correct definition.
4057 (doloop_end): Use standard pattern name for looping pattern.
4058 (doloop_end_nocc): New.
4059 (movsf): Add another constraint alternative.
4060 (zero_extendqihi2): Add constraint alternatives for not in place
4062 (zero_extendhisi2): Remove.
4063 (shift patterns): Add CC handling variants.
4067 (define_peephole2): New peephole to recognize mov that sets CC for
4070 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4072 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4073 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4076 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4079 * real.c (real_nextafter): Return true if result is denormal.
4081 2018-07-09 Martin Liska <mliska@suse.cz>
4083 * common.opt: Add back wrongly removed attribute.
4085 2018-07-09 Richard Biener <rguenther@suse.de>
4088 * dwarf2out.c (gen_block_die): For an early generated DIE
4089 always output high/low PC attributes.
4091 2018-07-09 Tom de Vries <tdevries@suse.de>
4093 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4094 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4095 onto VAR_DECL with abstract origin.
4097 2018-07-07 Jim Wilson <jimw@sifive.com>
4099 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4101 2018-07-07 Tom de Vries <tdevries@suse.de>
4103 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4105 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4107 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4109 (combine_bound): Use wide-int overflow calculation instead of
4111 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4113 * fold-const.c (int_const_binop_2): Same.
4114 (extract_muldiv_1): Same.
4115 (fold_div_compare): Same.
4116 (fold_abs_const): Same.
4118 * poly-int.h (add): Same.
4122 * predict.c (predict_iv_comparison): Same.
4123 * profile-count.c (slow_safe_scale_64bit): Same.
4124 * simplify-rtx.c (simplify_const_binary_operation): Same.
4125 * tree-chrec.c (tree_fold_binomial): Same.
4126 * tree-data-ref.c (split_constant_offset_1): Same.
4127 * tree-if-conv.c (idx_within_array_bound): Same.
4128 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4129 * tree-ssa-phiopt.c (minmax_replacement): Same.
4130 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4131 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4132 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4133 * wide-int.cc (wi::add_large): Same.
4134 (wi::mul_internal): Same.
4135 (wi::sub_large): Same.
4136 (wi::divmod_internal): Same.
4137 * wide-int.h: Change overflow type to overflow_type for neg, add,
4138 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4139 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4140 mul_internal, divmod_internal.
4141 (overflow_type): New enum.
4142 (accumulate_overflow): New.
4144 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4146 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4147 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4149 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4151 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4152 argument is checked for zero before entering loop, avoid checking again.
4154 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4156 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4157 * gimplify.h (generic_expr_could_trap_p): Likwise.
4158 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4160 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4162 PR tree-optimization/86401
4163 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4164 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4165 (fold_bit_and_mask): ... here. New helper function for match.pd.
4166 * fold-const.h (fold_bit_and_mask): Declare.
4167 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4169 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4172 * target.def (translate_mode_attribute): New hook.
4173 * targhooks.h (default_translate_mode_attribute): Declare.
4174 * targhooks.c (default_translate_mode_attribute): New function.
4175 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4176 * doc/tm.texi: Regenerate.
4177 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4178 (rs6000_translate_mode_attribute): New function.
4180 2018-07-06 Paul Koning <ni1d@arrl.net>
4182 * doc/md.texi (define_split): Document DONE and FAIL.
4183 (define_peephole2): Ditto.
4185 2018-07-05 Jeff Law <law@redhat.com>
4187 PR tree-optimization/86010
4188 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4189 both the head and tail of mem* and str* calls.
4191 2018-07-05 Jim Wilson <jimw@sifive.com>
4193 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4195 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4197 * config/aarch64/aarch64-simd.md: correct flags text for
4198 MIN_EXPR replacement.
4200 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4202 * configure: Regenerated.
4204 2018-07-05 Carl Love <cel@us.ibm.com>
4206 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4207 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4208 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4209 VSX_BUILTIN_DOUBLEL_V4SF.
4211 2018-07-05 Martin Sebor <msebor@redhat.com>
4214 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4215 than its domain to compute its the upper bound of a char array.
4217 2018-07-05 Nathan Sidwell <nathan@acm.org>
4219 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4220 * doc/cpp.texi: Update comment.
4221 * doc/tm.texi: Rebuilt.
4222 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4223 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4224 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4225 headers affected by extern "C".
4226 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4227 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4228 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4229 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4230 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4231 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4232 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4233 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4234 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4235 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4236 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4237 config/sparc/sp64-elf.h, config/spu/spu.h,
4238 config/stormy16/stormy16.h, config/v850/v850.h,
4239 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4240 define NO_IMPLICIT_EXTERN_C.
4241 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4243 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4246 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4247 instead of GET_MODE_SIZE when comparing Units.
4249 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4252 * rtlanal.c (set_noop_p): Constrain on mode change,
4253 include hard-reg-set.h
4255 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4257 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4259 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4262 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4265 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4267 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4269 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4272 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4274 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4276 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4279 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4281 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4283 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4284 neg_*_op* variables.
4286 2018-07-04 Martin Liska <mliska@suse.cz>
4288 * tree-switch-conversion.c: Define
4289 max_ratio_for_speed and max_ratio_for_size constants.
4291 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4292 Martin Liska <mliska@suse.cz>
4297 * common.opt: Rename align options with 'str_' prefix.
4298 * common/config/i386/i386-common.c (set_malign_value): New
4300 (ix86_handle_option): Use it to set -falign-* options/
4301 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4302 type from int to string.
4303 * config/aarch64/aarch64.c: Update default values from int
4305 * config/alpha/alpha.c (alpha_override_options_after_change):
4307 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4308 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4309 max skip conditionally.
4310 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4311 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4312 max skip conditionally.
4313 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4314 max skip conditionally.
4315 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4316 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4317 max skip conditionally.
4318 * config/i386/i386.c (struct ptt): Change type from int to
4320 (ix86_default_align): Set default values.
4321 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4322 max skip conditionally.
4323 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4324 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4325 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4326 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4327 max skip conditionally.
4328 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4329 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4330 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4331 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4332 max skip conditionally.
4333 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4334 * config/ia64/ia64.c (ia64_option_override): Set default values
4335 for alignment options.
4336 * config/m68k/m68k.c: Handle new str_align_* options.
4337 * config/mips/mips.c (mips_set_compression_mode): Change
4339 (mips_option_override): Set default values for options.
4340 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4342 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4344 * config/rx/rx.c (rx_option_override): Likewise.
4345 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4346 (LABEL_ALIGN): Use align_labels_log.
4347 (LOOP_ALIGN): Use align_loops_align.
4348 * config/s390/s390.c (s390_asm_output_function_label): Use new
4350 * config/sh/sh.c (sh_override_options_after_change):
4351 Change type of constants.
4352 * config/spu/spu.c (spu_sched_init): Likewise.
4353 * config/sparc/sparc.c (sparc_option_override): Set default
4355 * config/visium/visium.c (visium_option_override): Likewise.
4356 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4357 emit p2align format with last argument if it's not needed.
4358 * doc/invoke.texi: Document extended format of -falign-*.
4359 * final.c: Use align_labels alignment.
4360 * flags.h (struct target_flag_state): Change type to use
4362 (struct align_flags_tuple): New.
4363 (struct align_flags): Likewise.
4364 (align_loops_log): Redefine macro to use new types.
4365 (align_loops_max_skip): Redefine macro to use new types.
4366 (align_jumps_log): Redefine macro to use new types.
4367 (align_jumps_max_skip): Redefine macro to use new types.
4368 (align_labels_log): Redefine macro to use new types.
4369 (align_labels_max_skip): Redefine macro to use new types.
4370 (align_functions_log): Redefine macro to use new types.
4371 (align_loops): Redefine macro to use new types.
4372 (align_jumps): Redefine macro to use new types.
4373 (align_labels): Redefine macro to use new types.
4374 (align_functions): Redefine macro to use new types.
4375 (align_functions_max_skip): Redefine macro to use new types.
4376 (align_loops_value): New macro.
4377 (align_jumps_value): New macro.
4378 (align_labels_value): New macro.
4379 (align_functions_value): New macro.
4380 * function.c (invoke_set_current_function_hook): Propagate
4381 alignment values from flags to global variables default in
4383 * ipa-icf.c (sem_function::equals_wpa): Use
4384 cl_optimization_option_eq instead of memcmp.
4385 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4387 (cl_optimization_stream_in): Likewise.
4388 * optc-save-gen.awk: Support strings in cl_optimization.
4389 * opth-gen.awk: Likewise.
4390 * opts.c (finish_options): Remove error checking of invalid
4392 (MAX_CODE_ALIGN): Remove.
4393 (MAX_CODE_ALIGN_VALUE): Likewise.
4394 (parse_and_check_align_values): New function.
4395 (check_alignment_argument): Likewise.
4396 (common_handle_option): Use check_alignment_argument.
4397 * opts.h (parse_and_check_align_values): Declare.
4398 * toplev.c (init_alignments): Remove.
4399 (read_log_maxskip): New.
4400 (parse_N_M): Likewise.
4401 (parse_alignment_opts): Likewise.
4402 (backend_init_target): Remove usage of init_alignments.
4403 * toplev.h (parse_alignment_opts): Declare.
4404 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4406 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4407 * tree.c (cl_option_hasher::equal): New.
4408 * varasm.c: Use new global macros.
4410 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4412 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4413 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4414 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4415 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4416 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4417 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4418 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4419 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4420 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4421 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4422 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4423 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4424 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4426 2018-07-04 Martin Liska <mliska@suse.cz>
4427 Jonathan Wakely <jwakely@redhat.com>
4429 * coverage.c: Use correct type.
4430 * doc/invoke.texi: Language correction.
4432 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4435 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4436 ENDBRANCH for non-tail call which may return via indirect branch.
4437 * doc/extend.texi: Document indirect_return attribute.
4439 2018-07-03 Martin Sebor <msebor@redhat.com>
4441 PR tree-optimization/86274
4442 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4444 (format_floating): Correct handling of infinities and NaNs.
4446 2018-07-03 Martin Sebor <msebor@redhat.com>
4448 * print-tree.c (print_real_cst): New function.
4449 (print_node_brief): Call it.
4450 (print_node): Ditto.
4452 2018-07-03 Jeff Law <law@redhat.com>
4454 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4455 into a single pattern.
4457 * config/h8300/h8300.md (ors code_iterator): New.
4458 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4459 a single pattern and single splitter.
4460 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4461 (iorqi3_1, xorqi3_1): Likewise.
4462 (iorqi3, xorqi3 expanders): Similarly.
4464 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4465 (movmd_internal) into a single pattern using the P mode iterator.
4466 (movmd splitters): Similarly.
4467 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4468 (movsd splitters): Similarly.
4470 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4471 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4472 right table based on the size of the operand.
4473 * config/h8300/h8300.md (length_table): Corresponding changes. All
4474 references to "addb", "addw" and "addl" changed to "add".
4475 (btst patterns): Merge two variants into a single pattern.
4476 (tstqi, tsthi): Likewise.
4477 (addhi3_incdec, addsi3_incdec): Likewise.
4478 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4479 (mulhi3, mulsi3): Likewise.
4480 (udivhi3, udivsi3): Likewise.
4481 (divhi3, divsi3): Likewise.
4482 (andorqi3, andorhi3, andorsi3): Likewise.
4484 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4487 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4488 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4490 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4492 PR tree-optimization/85694
4493 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4494 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4496 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4497 (<u>avg<mode>3_ceil): New patterns.
4499 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4501 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4502 scan-tree-dump directive.
4503 * gcc.dg/vect/slp-perm-2.c: Likewise.
4504 * gcc.dg/vect/slp-perm-3.c: Likewise.
4505 * gcc.dg/vect/slp-perm-5.c: Likewise.
4506 * gcc.dg/vect/slp-perm-6.c: Likewise.
4507 * gcc.dg/vect/slp-perm-7.c: Likewise.
4508 * gcc.dg/vect/slp-perm-8.c: Likewise.
4510 2018-07-03 Marek Polacek <polacek@redhat.com>
4513 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4515 2018-07-03 Richard Biener <rguenther@suse.de>
4518 * tree-ssa-structalias.c (find_func_clobbers): Properly
4519 handle indirect calls.
4521 2018-07-03 Jeff Law <law@redhat.com>
4523 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4524 (shifts): New code iterator.
4525 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4526 expander. Fix HImode handling on H8/SX.
4527 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4528 (subqi3, subhi3, subsi3 expanders): Likewise.
4529 (andqi3, andhi3, andsi3 expanders): Likewise.
4530 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4531 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4532 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4533 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4534 (zero_extendqihi2, zero_extendqisi2): Likewise.
4535 (extendqihi2, extendqisi2): Likewise.
4536 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4537 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4538 (rotlqi3_1, rotlhi3_1): Likewise.
4539 (logicalhi3_sn, logicalsi3_sn): Likewise.
4540 (logicalhi3, logicalsi3): Likewise.
4542 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4544 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4545 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4546 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4547 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4548 type to append_pattern_def_seq instead of creating a stmt_vec_info
4550 (build_mask_conversion): Likewise. Remove vinfo argument.
4551 (vect_add_conversion_to_patterm): Likewise, renaming to...
4552 (vect_add_conversion_to_pattern): ...this.
4553 (vect_recog_mask_conversion_pattern): Update call to
4554 build_mask_conversion. Pass the vector type to
4555 append_pattern_def_seq here too.
4556 (vect_recog_gather_scatter_pattern): Update call to
4557 vect_add_conversion_to_pattern.
4559 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4561 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4562 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4563 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4564 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4565 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4566 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4567 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4568 append_pattern_def_seq instead of new_pattern_def_seq.
4569 (vect_recog_divmod_pattern): Do both of the above.
4570 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4573 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4575 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4576 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4577 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4578 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4579 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4580 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4581 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4582 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4583 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4584 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4585 parameter with a single stmt_vec_info.
4586 (vect_recog_func_ptr): Likewise.
4587 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4588 (vect_try_gather_scatter_pattern): ...this.
4589 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4590 the stmt_vec_info of the statement to be matched. Don't clear
4591 STMT_VINFO_RELATED_STMT.
4592 (vect_pattern_recog): Update call accordingly.
4594 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4596 PR tree-optimization/85694
4597 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4598 (uavgM3_ceil): Document new optabs.
4599 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4600 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4602 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4603 (savg_ceil_optab): New optabs.
4604 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4605 (vect_vect_recog_func_ptrs): Add it.
4606 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4607 constant directly from the associated lhs.
4609 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4611 * tree-vect-patterns.c (vect_split_statement): New function.
4612 (vect_convert_input): Use it to try to split an existing cast.
4614 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4616 * poly-int.h (print_hex): New function.
4617 * dumpfile.h (dump_dec, dump_hex): Declare.
4618 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4619 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4620 min_input_precision, operation_precision and operation_sign.
4621 * tree-vect-patterns.c (vect_get_range_info): New function.
4622 (vect_same_loop_or_bb_p, vect_single_imm_use)
4623 (vect_operation_fits_smaller_type): Delete.
4624 (vect_look_through_possible_promotion): Add an optional
4625 single_use_p parameter.
4626 (vect_recog_over_widening_pattern): Rewrite to use new
4627 stmt_vec_info infomration. Handle one operation at a time.
4628 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4629 (vect_truncatable_operation_p, vect_set_operation_type)
4630 (vect_set_min_input_precision): New functions.
4631 (vect_determine_min_output_precision_1): Likewise.
4632 (vect_determine_min_output_precision): Likewise.
4633 (vect_determine_precisions_from_range): Likewise.
4634 (vect_determine_precisions_from_users): Likewise.
4635 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4636 (vect_vect_recog_func_ptrs): Put over_widening first.
4638 (vect_pattern_recog): Call vect_determine_precisions.
4640 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4642 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4643 statements that have been replaced by further pattern statements.
4644 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4646 2018-07-03 Richard Biener <rguenther@suse.de>
4648 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4649 always set *dt. Dump vectype in vectype overload.
4650 * dumpfile.h (dump_gimple_expr): New function.
4651 (dump_gimple_expr_loc): Likewise.
4652 * dumpfile.c (dump_gimple_expr): New function.
4653 (dump_gimple_expr_loc): Likewise.
4655 2018-07-02 Jeff Law <law@redhat.com>
4657 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4658 the H8/300, H8/300H and H8/S variants into a single pattern.
4659 (movhi_h8300, movqi_h8300hs): Similarly.
4660 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4661 (QHI mode iterator): New.
4663 * config/h8300/h8300.md: Remove trailing whitespace.
4665 2018-07-02 Jim Wilson <jimw@sifive.com>
4667 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4668 instead of emit_insn for interrupt returns.
4669 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4670 (riscv_sret, riscv_uret): Likewise.
4672 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4674 * pretty-print.c (selftest::test_pp_format): Move save and restore
4675 of quotes to class auto_fix_quotes, and add an instance.
4676 * selftest.c: Include "intl.h".
4677 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4678 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4679 * selftest.h (selftest::auto_fix_quotes): New class.
4681 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4683 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4684 (aarch64_sve_prepare_conditional_op): Remove.
4685 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4686 Allow aarch64_simd_reg_or_zero as select operand; remove
4687 the aarch64_sve_prepare_conditional_op call.
4688 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4689 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4690 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4691 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4692 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4693 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4694 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4695 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4696 and a splitters to match all of the *_any patterns.
4697 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4699 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4700 (SVE_COND_FP_BINARY_REV): Remove.
4701 (sve_int_op_rev, sve_fp_op_rev): New.
4702 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4703 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4704 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4705 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4706 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4707 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4708 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4709 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4710 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4712 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4713 Remove match_dup 1 from the inner unspec.
4714 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4716 * config/aarch64/aarch64.md (movprfx): New attr.
4717 (length): Default movprfx to 8.
4718 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4719 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4720 (*<su>mul<SVE_I>3_highpart): Likewise.
4721 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4722 (*v<ASHIFT><SVE_I>3): Likewise.
4723 (*<su><MAXMIN><SVE_I>3): Likewise.
4724 (*<su><MAXMIN><SVE_F>3): Likewise.
4725 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4726 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4727 (*div<SVE_F>4): Likewise.
4729 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4731 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4734 2018-07-02 Richard Biener <rguenther@suse.de>
4736 PR tree-optimization/86363
4737 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4738 memset argument refers to a non-variable address.
4740 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4742 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4743 {PLUS,MINUS}_EXPR code to...
4744 (adjust_symbolic_bound): ...here,
4745 (combine_bound): ...here,
4746 (set_value_range_with_overflow): ...and here.
4748 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4750 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4752 (extract_range_from_abs_expr): ...here.
4754 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4756 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4757 -fno-omit-frame-pointer when not optimizing.
4759 2018-07-02 Martin Liska <mliska@suse.cz>
4762 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4763 (propagate_nothrow): Likewise.
4765 2018-07-02 Martin Liska <mliska@suse.cz>
4768 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4770 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4772 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4773 function in r262149, changing "loc" param from source_location to
4774 const dump_location_t &.
4775 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4776 declaration, as above.
4778 2018-07-01 Paul Koning <ni1d@arrl.net>
4780 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4781 -munit-asm, -mgnu-asm, -mdec-asm.
4782 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4783 (pdp11_output_labelref): New.
4784 (pdp11_output_def): New.
4785 (pdp11_output_addr_vec_elt): New.
4786 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4787 %# and %@ format codes.
4788 (pdp11_option_override): New.
4789 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4790 (pdp11_output_ident): New.
4791 (pdp11_asm_named_section): New.
4792 (pdp11_asm_init_sections): New.
4793 (pdp11_file_start): New.
4794 (pdp11_file_end): New.
4795 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4796 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4797 %o, like %c but octal.
4798 (pdp11_option_override): New.
4799 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4801 (DATA_SECTION_ASM_OP): Ditto.
4802 (READONLY_DATA_SECTION_ASM_OP): New.
4803 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4804 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4805 (ASM_OUTPUT_LABELREF): Ditto.
4806 (ASM_OUTPUT_DEF): Ditto.
4807 (ASM_OUTPUT_EXTERNAL): New.
4808 (ASM_OUTPUT_SOURCE_FILENAME): New.
4809 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4810 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4811 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4812 %# and %@ format codes.
4813 * config/pdp11/pdp11.opt (mgnu-asm): New.
4814 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4815 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4816 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4818 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4820 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4821 dereferencing path[] beyond its length.
4823 (debug_all_paths): New.
4824 (rewire_first_differing_edge): New.
4825 (adjust_paths_after_duplication): New.
4826 (duplicate_thread_path): Call adjust_paths_after_duplication.
4828 (thread_through_all_blocks): Add new argument to
4829 duplicate_thread_path.
4831 2018-06-30 Jim Wilson <jimw@sifive.com>
4833 * config/riscv/predicates.md (p2m1_shift_operand): New.
4834 (high_mask_shift_operand): New.
4835 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4836 pattern using p2m1_shift_operand.
4837 (lshsi3_zero_extend_3+2): New combiner pattern using
4838 high_mask_shift_operand.
4840 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4842 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4844 (vect_recog_rotate_pattern): ...here.
4845 (vect_convert_input): Try to insert casts of invariants in the
4847 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4848 preheader to be empty.
4850 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4852 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4853 vector type. If given, install it in the new statement's
4855 (vect_element_precision): New function.
4856 (vect_unpromoted_value): New struct.
4857 (vect_unpromoted_value::vect_unpromoted_value): New function.
4858 (vect_unpromoted_value::set_op): Likewise.
4859 (vect_look_through_possible_promotion): Likewise.
4860 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4861 (vect_widened_op_tree, vect_convert_input): Likewise.
4862 (vect_convert_inputs, vect_convert_output): Likewise.
4863 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4864 to handle the optional cast of the multiplication result and
4865 vect_widened_op_tree to detect the widened multiplication itself.
4866 Do not require the input and output of promotion casts to have
4867 the same sign, but base the signedness of the operation on the
4868 input rather than the result. If the pattern includes two
4869 promotions, check that those promotions have the same sign.
4870 Do not restrict the MULT_EXPR handling to a double-width result;
4871 handle quadruple-width results and wider. Use vect_convert_inputs
4872 to convert the inputs to the common type.
4873 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
4874 to handle the optional cast of the ABS result. Also allow a sign
4875 change or a sign extension between the ABS and MINUS.
4876 Use vect_widened_op_tree to detect the widened subtraction and use
4877 vect_convert_inputs to convert the inputs to the common type.
4878 (vect_handle_widen_op_by_const): Delete.
4879 (vect_recog_widen_op_pattern): New function.
4880 (vect_recog_widen_mult_pattern): Use it.
4881 (vect_recog_widen_shift_pattern): Likewise.
4882 (vect_recog_widen_sum_pattern): Use
4883 vect_look_through_possible_promotion to handle the promoted
4886 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4888 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4889 the containing gimple_seq *.
4890 * gimple-iterator.h (gsi_for_stmt): Declare it.
4891 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4892 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4893 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4894 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4895 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4897 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4899 (vect_mark_pattern_stmts): ...here. Handle cases in which the
4900 statement being replaced is part of an existing pattern
4901 definition sequence, inserting the new pattern statements before
4903 (vect_pattern_recog_1): Don't return a bool. If the statement
4904 is already part of a pattern, instead apply pattern matching
4905 to the pattern definition statements. Don't clear the
4906 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4907 (vect_pattern_recog): Don't break after the first match;
4908 continue processing the pattern definition statements instead.
4909 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4911 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4913 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4914 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4915 (vect_recog_widen_sum_pattern): Use it.
4917 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4919 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4920 phi is not a pattern statement and has not been replaced by
4921 a pattern statement.
4922 * tree-vect-patterns.c (type_conversion_p): Don't check
4923 STMT_VINFO_IN_PATTERN_P.
4924 (vect_recog_vector_vector_shift_pattern): Likewise.
4925 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4926 the pattern statement rather than the original statement; check
4927 directly for a WIDEN_MULT_EXPR here.
4928 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4929 vect_is_simple_use to return the pattern statement rather
4930 than the original statement; use is_pattern_stmt_p to check
4931 for such a pattern statement.
4932 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4933 to return the pattern statement rather than the original statement;
4934 don't do the same transformation here.
4935 (vect_is_simple_use): If the defining statement has been replaced
4936 by a pattern statement, return the pattern statement instead.
4937 Remove the corresponding (local) transformation from the vectype
4940 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4942 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4943 end and default to null.
4944 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4945 (vectorizable_reduction): Update calls accordingly, dropping the
4946 gimple ** argument if the passed-back statement isn't needed.
4947 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4948 (vect_recog_rotate_pattern): Likewise.
4949 (vect_recog_mask_conversion_pattern): Likewise.
4950 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4951 (vect_mask_constant_operand_p): Likewise.
4952 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4953 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4954 (get_group_load_store_type, get_load_store_type): Likewise.
4955 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4956 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4957 (vectorizable_conversion, vectorizable_assignment): Likewise.
4958 (vectorizable_shift, vectorizable_operation): Likewise.
4959 (vectorizable_store, vect_is_simple_cond): Likewise.
4960 (vectorizable_condition, vectorizable_comparison): Likewise.
4961 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4962 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4963 and move it to the end. Cope with null def_stmt_outs.
4965 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4967 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
4969 2018-06-29 Jeff Law <law@redhat.com>
4971 * config/v850/v850.c (v850_legitimate_address_p): Handle large
4972 displacements for TARGET_V850E2V3 and newer.
4973 (TARGET_LRA_P): Remove. Defaults to LRA now.
4974 * config/v850/v850.md (sign23byte_load): Remove.
4975 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
4976 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
4978 2018-06-29 Martin Liska <mliska@suse.cz>
4981 * coverage.c (coverage_init): Mangle full path name.
4982 * doc/invoke.texi: Document the change.
4983 * gcov-io.c (mangle_path): New.
4984 * gcov-io.h (mangle_path): Likewise.
4985 * gcov.c (mangle_name): Use mangle_path for path mangling.
4987 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4989 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4990 if starting source register is not even.
4992 2018-06-29 Martin Liska <mliska@suse.cz>
4994 PR tree-optimization/86263
4995 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4996 Make edge redirection.
4998 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5000 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5001 (dump_scope_depth): New variable.
5002 (get_dump_scope_depth): New function.
5003 (dump_begin_scope): New function.
5004 (dump_end_scope): New function.
5005 * dumpfile.h (get_dump_scope_depth): New declaration.
5006 (dump_begin_scope): New declaration.
5007 (dump_end_scope): New declaration.
5008 (class auto_dump_scope): New class.
5009 (AUTO_DUMP_SCOPE): New macro.
5010 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5013 2018-06-29 Richard Biener <rguenther@suse.de>
5015 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5016 compute_all_dependences succeeds.
5017 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5018 exceed --param loop-max-datarefs-for-datadeps.
5020 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5022 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5024 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5027 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5028 alternative 0 in preferred_for_speed attribute.
5030 2018-06-28 Paul Koning <ni1d@arrl.net>
5032 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5033 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5034 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5035 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5036 for shift insn patterns.
5038 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5040 * cgraph.c (cgraph_node::get_body): Replace assignments to
5041 "dump_file" with calls to set_dump_file.
5042 * dumpfile.c (alt_dump_file): Make static, and group with...
5043 (alt_flags): ...this definition.
5044 (dumps_are_enabled): New variable.
5045 (refresh_dumps_are_enabled): New function.
5046 (set_dump_file): New function.
5047 (set_alt_dump_file): New function.
5048 (gcc::dump_manager::dump_start): Replace assignments to
5049 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5051 (gcc::dump_manager::dump_finish): Likewise.
5052 * dumpfile.h (alt_dump_file): Delete decl.
5053 (dumps_are_enabled): New variable decl.
5054 (set_dump_file): New function decl.
5055 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5057 * tree-nested.c (lower_nested_functions): Replace assignments to
5058 "dump_file" with calls to set_dump_file.
5060 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5062 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5063 goto_locus of each outgoing edge of each basic block.
5065 2018-06-28 Richard Biener <rguenther@suse.de>
5067 * dwarf2out.c (decl_scope_table): Remove.
5068 (push_decl_scope): Likewise.
5069 (pop_decl_scope): Likewise.
5070 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5071 (gen_struct_or_union_type_die): Likewise.
5072 (gen_tagged_type_die): Likewise.
5073 (dwarf2out_init): Do not initialize decl_scope_table.
5074 (dwarf2out_c_finalize): Do not free it.
5076 2018-06-28 Richard Biener <rguenther@suse.de>
5078 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5079 deciding whether to not re-use a DIE.
5081 2018-06-28 Richard Biener <rguenther@suse.de>
5083 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5084 DW_AT_abstract_origin attribute.
5086 2018-06-28 Martin Liska <mliska@suse.cz>
5088 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5089 Use newly introduced constants.
5090 * tree-switch-conversion.h (struct jump_table_cluster):
5091 Define max_ratio_for_size and max_ratio_for_speed.
5093 2018-06-28 Martin Liska <mliska@suse.cz>
5095 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5096 Add new checking assert to catch invalid state.
5097 (jump_table_cluster::can_be_handled): Handle single case
5099 (jump_table_cluster::is_beneficial): Bail out for such case.
5100 (bit_test_cluster::find_bit_tests):
5101 Add new checking assert to catch invalid state.
5102 (bit_test_cluster::can_be_handled): Handle single case
5104 (bit_test_cluster::is_beneficial): Bail out for such case.
5105 (switch_decision_tree::analyze_switch_statement):
5108 2018-06-28 Martin Liska <mliska@suse.cz>
5110 * common.opt: Introduce -completion option.
5111 * gcc.c (driver_handle_option): Handle it.
5112 (driver::main): Print completions if completion
5114 * opt-suggestions.c (option_proposer::get_completions):
5116 (option_proposer::suggest_completion): Likewise.
5117 (option_proposer::find_param_completions): Likewise.
5118 (verify_autocompletions): Likewise.
5119 (test_completion_valid_options): Likewise.
5120 (test_completion_valid_params): Likewise.
5121 (in_completion_p): Likewise.
5122 (empty_completion_p): Likewise.
5123 (test_completion_partial_match): Likewise.
5124 (test_completion_garbage): Likewise.
5125 (opt_proposer_c_tests): Likewise.
5126 * opt-suggestions.h: Declare new functions.
5127 * opts.c (common_handle_option): Handle OPT__completion_.
5128 * selftest-run-tests.c (selftest::run_tests): Add
5129 opt_proposer_c_tests.
5130 * selftest.c (assert_str_startswith): New.
5131 * selftest.h (assert_str_startswith): Likewise.
5132 (opt_proposer_c_tests): New.
5133 (ASSERT_STR_STARTSWITH): Likewise.
5135 2018-06-28 Martin Liska <mliska@suse.cz>
5137 * Makefile.in: Add opt-suggestions.o.
5138 * gcc-main.c: Include opt-suggestions.h.
5139 * gcc.c (driver::driver): Likewise.
5140 (driver::~driver): Remove m_option_suggestions.
5141 (driver::build_option_suggestions): Moved to option_proposer.
5142 (driver::suggest_option): Likewise.
5143 (driver::handle_unrecognized_options): Use option_proposer.
5144 * gcc.h (class driver): Add new memver m_option_proposer.
5145 * opt-suggestions.c: New file.
5146 * opt-suggestions.h: New file.
5148 2018-06-28 Martin Liska <mliska@suse.cz>
5150 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5151 (auto_string_vec::~auto_string_vec): Likewise.
5153 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5155 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5156 prevent_decl_creation_for_types fields up and add reset_location field.
5157 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5158 statement if id->reset_location is true.
5159 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5160 id->reset_location is true.
5161 (copy_phis_for_bb): Force input_location on the arguments if
5162 id->reset_location is true.
5163 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5164 is set on the function to be inlined.
5166 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5168 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5170 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5172 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5173 registers for Pmode.
5174 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5175 hard registers for the clobbered pseudo.
5177 2018-06-27 Paul Koning <ni1d@arrl.net>
5179 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5180 mutually exclusive options.
5181 * config/pdp11/constraints.md (h): New constraint.
5182 (O): Update definition to match shift code generation.
5183 (D): New constraint.
5184 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5186 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5188 (output_jump): Change arguments.
5189 (pdp11_fixed_cc_regs): New function.
5190 (pdp11_cc_mode): Ditto.
5191 (pdp11_expand_shift): Ditto.
5192 (pdp11_assemble_shift): Ditto.
5193 (pdp11_small_shift): Ditto.
5194 (pdp11_branch_cost): Remove.
5195 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5197 (pdp11_register_move_cost): Update for CC registers.
5198 (pdp11_rtx_costs): Add case for LSHIFTRT.
5199 (pdp11_output_jump): Add CCNZ mode conditional branches.
5200 (notice_update_cc_on_set): Remove.
5201 (pdp11_cc_mode): New function.
5202 (simple_memory_operand): Correct pre/post decrement case.
5203 (no_side_effect_operand): New function.
5204 (pdp11_regno_reg_class): Add CC_REGS class.
5205 (pdp11_fixed_cc_regs): New function.
5206 (pdp11_small_shift): New function.
5207 (pdp11_expand_shift): New function to expand shift insns.
5208 (pdp11_assemble_shift): New function to output shifts.
5209 (pdp11_branch_cost): Remove.
5210 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5211 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5212 (WCHAR_TYPE): Ditto.
5213 (PTRDIFF_TYPE): Ditto.
5214 (ADJUST_INSN_LENGTH): New macro.
5215 (FIXED_REGISTERS): Add CC registers.
5216 (CALL_USED_REGISTERS): Ditto.
5218 (REG_CLASS_NAMES): Ditto.
5219 (REG_CLASS_CONTENTS): Ditto.
5220 (SELECT_CC_MODE): Use new function.
5221 (TARGET_FLAGS_REGNUM): New macro.
5222 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5223 (cc0_reg_rtx): Remove.
5224 (CC_STATUS_MDEP): Remove.
5225 (CC_STATUS_MDEFP_INIT): Remove.
5226 (CC_IN_FPU): Remove.
5227 (NOTICE_UPDATE_CC): Remove.
5228 (REGISTER_NAMES): Add CC registers.
5229 (BRANCH_COST): Change to constant 1.
5230 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5232 * config/pdp11/pdp11.opt (mbcopy): Remove.
5233 (mbcopy-builtin): Remove.
5234 (mbranch-cheap): Remove.
5235 (mbranch-expensive): Remove.
5236 * config/pdp11/predicates.md (expand_shift_operand): Update to
5237 match shift code generation.
5238 (ccnz_operator): New predicate.
5239 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5240 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5241 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5243 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5244 description of O constraint.
5246 2018-06-27 Jeff Law <law@redhat.com>
5247 Austin Law <austinklaw@gmail.com>
5249 * config/v850/v850.md (addsi3_set_flags): New pattern.
5250 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5251 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5252 (zero_extendhisi2_v850_set_flags): Likewise.
5253 (zero_extendqisi2_v850_set_flags): Likewise.
5254 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5255 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5256 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5258 * config/v850/v850-protos.h (notice_update_cc): Remove.
5259 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5260 (v850_print_operand): Handle 'D' and "d".
5261 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5262 Add handling of arithmetic/logical operations compared against zero.
5263 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5264 Do not look at v850_compare_op, instead get mode from last argument.
5265 (v850_gen_compare): Remove
5266 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5267 after reload for prologue insns.
5268 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5270 (construct_save_jarl): Likewise.
5271 (TARGET_FLAGS_REGNUM): Define.
5272 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5273 (NOTICE_UPDATE_CC): Remove.
5274 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5275 than cc0. Conditionalize on reload_completed.
5276 (cmpsi_insn, setfcc_insn): Likewise.
5277 (tst1 splitter): Turn into define_and_split which sets the flags
5279 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5280 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5281 (cstoresf4, cstoredf4): Clobber the flags.
5282 (cmpsi, cmpsf, cmpdf): Remove expanders.
5283 (setf_insn): Remove pattern.
5284 (addsi3): Turn into define_and_split which clobbers the flags after
5285 reload and a suitable pattern (addsi3_clobber_flags) for use after
5287 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5288 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5289 (ashrsi3, ashrsi3_v850e2): Likewise.
5290 (bins): Clobber the flags.
5291 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5292 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5293 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5294 (call_value_internal_short, call_value_internal_long): Likewise.
5295 (callt_save_interrupt, callt_return_interrupt): Likewise.
5296 (save_interrupt, return_interrupt): Likewise.
5297 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5298 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5299 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5300 (All FP comparisons): Only allow after reload has completed.
5302 (divh, divhu): Tweak output template.
5303 (branch_z_normal, branch_z_invert): Remove
5304 (branch_nz_normal, branch_nz_invert): Likewise.
5305 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5307 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5308 * config/v850/v850.c (notice_update_cc): Remove.
5309 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5310 (CC_NO_CARRY): Likewise.
5311 (NOTICE_UPDATE_CC): Define to nothing.
5312 * config/v850/v850.md: Remove block comment on cc0 handling
5313 Remove "cc" attribute from all patterns. Remove cc_status handling
5314 from all patterns. Minor formatting fixes.
5316 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5318 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5319 (cortex-a76.cortex-a55): Likewise.
5320 * config/aarch64/aarch64-tune.md: Regenerate.
5321 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5322 cortex-a76.cortex-a55.
5324 2018-06-27 Jeff Law <law@redhat.com>
5326 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5327 (MULTILIB_DIRNAMES): Similarly.
5329 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5331 * gimple.h (gimple_return_retbnd): Delete.
5332 (gimple_return_set_retbnd): Likewise.
5333 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5334 gimple_return_set_retbnd.
5335 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5336 gimple_return_retbnd and adjust.
5337 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5338 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5339 Explicitly return NULL in a couple more cases. Move assertion
5340 on debug statements and remove unreachable code.
5341 (reset_debug_binding): Do not test id->retbnd.
5342 (expand_call_inline): Do not set it.
5344 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5346 * configure.ac: Add --disable-gcov option.
5347 * configure: Regenerate.
5348 * Makefile.in: Honour @enable_gcov@.
5349 * doc/install.texi: Document --disable-gcov.
5351 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5353 * config/arm/arm-cpus.in (cortex-a76): New entry.
5354 (cortex-a76.cortex-a55): Likewise.
5355 * config/arm/arm-tables.opt: Regenerate.
5356 * config/arm/arm-tune.md: Likewise.
5357 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5358 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5359 cortex-a76.cortex-a55.
5361 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5364 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5366 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5368 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5370 (EPILOGUE_USES): Likewise.
5372 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5374 * tree-inline.c (remap_location): New function extracted from...
5375 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5376 (copy_phis_for_bb): ...here. Call remap_location.
5377 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5379 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5381 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5382 unaligned vsx for 16B memset.
5384 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5387 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5388 ieee128_float_type_node to long_double_type_node unless
5389 TARGET_LONG_DOUBLE_128 is set.
5391 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5393 * cfgloop.c (get_loop_location): Convert return type from
5394 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5395 by implicit construction from rtx_insn *, and using
5396 dump_user_location_t::from_function_decl for the fallback case.
5397 * cfgloop.h (get_loop_location): Convert return type from
5398 location_t to dump_user_location_t.
5399 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5400 dump_printf_loc to pass in a dump_location_t rather than a
5401 location_t, via the gimple stmt.
5402 * coverage.c (get_coverage_counts): Update calls to
5403 dump_printf_loc to pass in dump_location_t rather than a
5405 * doc/optinfo.texi (Dump types): Convert example of
5406 dump_printf_loc from taking "locus" to taking "insn". Update
5407 description of the "_loc" calls to cover dump_location_t.
5408 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5410 (dump_user_location_t::dump_user_location_t): New constructors,
5411 from gimple *stmt and rtx_insn *.
5412 (dump_user_location_t::from_function_decl): New function.
5413 (dump_loc): Make static.
5414 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5415 const dump_location_t &.
5416 (dump_generic_expr_loc): Delete.
5417 (dump_printf_loc): Convert param "loc" from location_t to
5418 const dump_location_t &.
5419 (selftest::test_impl_location): New function.
5420 (selftest::dumpfile_c_tests): New function.
5421 * dumpfile.h: Include "profile-count.h".
5422 (class dump_user_location_t): New class.
5423 (struct dump_impl_location_t): New struct.
5424 (class dump_location_t): New class.
5425 (dump_printf_loc): Convert 2nd param from source_location to
5426 const dump_location_t &.
5427 (dump_generic_expr_loc): Delete.
5428 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5429 const dump_location_t &.
5430 * gimple-fold.c (fold_gimple_assign): Update call to
5431 dump_printf_loc to pass in a dump_location_t rather than a
5432 location_t, via the gimple stmt.
5433 (gimple_fold_call): Likewise.
5434 * gimple-loop-interchange.cc
5435 (loop_cand::analyze_iloop_reduction_var): Update for change to
5436 check_reduction_path.
5437 (tree_loop_interchange::interchange): Update for change to
5439 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5440 change in return-type of find_loop_location.
5441 (graphite_regenerate_ast_isl): Likewise.
5442 * graphite-optimize-isl.c (optimize_isl): Likewise.
5443 * graphite.c (graphite_transform_loops): Likewise.
5444 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5445 pass in a dump_location_t rather than a location_t, via the
5447 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5448 * ipa.c (walk_polymorphic_call_targets): Likewise.
5449 * loop-unroll.c (report_unroll): Convert "locus" param from
5450 location_t to dump_location_t.
5451 (decide_unrolling): Update for change to get_loop_location's
5453 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5454 location_t to dump_user_location_t.
5455 (grid_find_single_omp_among_assignments_1): Updates calls to
5456 dump_printf_loc to pass in a dump_location_t rather than a
5457 location_t, via the gimple stmt.
5458 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5459 location_t to dump_location_t. Updates calls to dump_printf_loc
5460 to pass in a dump_location_t rather than a location_t, via the
5462 (grid_inner_loop_gridifiable_p): Likewise.
5463 (grid_dist_follows_simple_pattern): Likewise.
5464 (grid_gfor_follows_tiling_pattern): Likewise.
5465 (grid_target_follows_gridifiable_pattern): Likewise.
5466 (grid_attempt_target_gridification): Convert initialization
5467 of local "grid" from memset to zero-initialization; FIXME: does
5468 this require C++11? Update call to dump_printf_loc to pass in a
5469 optinfo_location rather than a location_t, via the gimple stmt.
5470 * profile.c (read_profile_edge_counts): Updates call to
5471 dump_printf_loc to pass in a dump_location_t rather than a
5473 (compute_branch_probabilities): Likewise.
5474 * selftest-run-tests.c (selftest::run_tests): Call
5476 * selftest.h (dumpfile_c_tests): New decl.
5477 * tree-loop-distribution.c (pass_loop_distribution::execute):
5478 Update for change in return type of find_loop_location.
5479 * tree-parloops.c (parallelize_loops): Likewise.
5480 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5481 "locus" from location_t to dump_user_location_t.
5482 (canonicalize_loop_induction_variables): Likewise.
5483 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5484 for change in return type of find_loop_location.
5485 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5486 to dump_printf_loc to pass in a dump_location_t rather than a
5487 location_t, via the stmt.
5488 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5490 * tree-vect-loop-manip.c (find_loop_location): Convert return
5491 type from source_location to dump_user_location_t.
5492 (vect_do_peeling): Update for above change.
5493 (vect_loop_versioning): Update for change in type of
5495 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5496 from location_t to dump_user_location_t.
5497 (vect_estimate_min_profitable_iters): Update for change in type
5499 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5500 location_t to dump_location_t.
5501 (vect_slp_bb): Update for change in type of vect_location.
5502 * tree-vectorizer.c (vect_location): Convert from source_location
5503 to dump_user_location_t.
5504 (try_vectorize_loop_1): Update for change in vect_location's type.
5505 (vectorize_loops): Likewise.
5506 (increase_alignment): Likewise.
5507 * tree-vectorizer.h (vect_location): Convert from source_location
5508 to dump_user_location_t.
5509 (find_loop_location): Convert return type from source_location to
5510 dump_user_location_t.
5511 (check_reduction_path): Convert 1st param from location_t to
5512 dump_user_location_t.
5513 * value-prof.c (check_counter): Update call to dump_printf_loc to
5514 pass in a dump_user_location_t rather than a location_t; update
5515 call to error_at for change in type of "locus".
5516 (check_ic_target): Update call to dump_printf_loc to
5517 pass in a dump_user_location_t rather than a location_t, via the
5520 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5522 * config/s390/s390.h (enum processor_flags): Do not use
5523 default tune parameter when -march was specified.
5525 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5528 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5529 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5532 2018-06-26 Richard Biener <rguenther@suse.de>
5534 PR tree-optimization/86287
5536 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5537 (vect_analyze_loop): Initialize n_stmts.
5539 2018-06-26 Richard Biener <rguenther@suse.de>
5542 * fold-const.c (fold_convertible_p): Pointer extension
5545 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5548 * dwarf2out.c (loc_list_has_views): Adjust comments.
5549 (dw_loc_list): Split single cross-partition range with
5552 2018-06-25 Jeff Law <law@redhat.com>
5554 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5555 on -mbig-switch by default.
5557 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5559 (const_float_0_operand): Remove unused predicate.
5560 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5561 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5562 (recipsf2): New expander. Original pattern now called
5564 (recipdf2, recipdf2_insn): Similarly.
5565 (rsqrtsf2, rsqrtsf2_insn): Similarly
5566 (rsqrtdf2, rsqrtdf2_insn): Similarly
5568 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5570 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5571 Simplify logic for FreeBSD (twice).
5573 2018-06-25 Martin Sebor <msebor@redhat.com>
5575 PR tree-optimization/86204
5576 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5577 a strnlen result if it's less than the length of the string.
5579 2018-06-25 Martin Sebor <msebor@redhat.com>
5581 PR tree-optimization/85700
5582 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5583 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5584 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5586 2018-06-25 Martin Sebor <msebor@redhat.com>
5588 * doc/extend.texi (Zero-length arrays): Update and clarify.
5590 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5592 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5593 added IEEE/IBM long double multilib support on PowerPC little
5594 endian Linux systems.
5595 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5596 (MULTILIB_DEFAULTS): Likewise.
5597 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5599 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5600 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5601 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5603 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5606 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5607 (REORDER_45): Likewise.
5609 2018-06-25 Jeff Law <law@redhat.com>
5611 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5612 dividend to 32 bits. Adjust length.
5613 (udivmodhi4): Cleanup output template. Fix length.
5615 2018-06-25 Carl Love <cel@us.ibm.com>
5617 * config/rs6000/vsx.md: Change word selector to prefered location.
5619 2018-06-25 Richard Biener <rguenther@suse.de>
5621 PR tree-optimization/86304
5622 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5623 epilogue-if-converted loops as well.
5625 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5627 * lto-section-out.c (lto_begin_section): Do not print section
5628 name for noaddr and unnumbered dumps.
5630 2018-06-25 Richard Biener <rguenther@suse.de>
5632 * tree-vectorizer.h (struct vec_info_shared): New structure
5633 with parts split out from struct vec_info and loop_nest from
5634 struct _loop_vec_info.
5635 (struct vec_info): Adjust accordingly.
5636 (struct _loop_vec_info): Likewise.
5637 (LOOP_VINFO_LOOP_NEST): Adjust.
5638 (LOOP_VINFO_DATAREFS): Likewise.
5639 (LOOP_VINFO_DDRS): Likewise.
5640 (struct _bb_vec_info): Likewise.
5641 (BB_VINFO_DATAREFS): Likewise.
5642 (BB_VINFO_DDRS): Likewise.
5643 (struct _stmt_vec_info): Add dr_aux member.
5644 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5645 (DR_MISALIGNMENT_UNINITIALIZED): New.
5646 (set_dr_misalignment): Adjust.
5647 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5648 (vect_analyze_loop): Adjust prototype.
5649 (vect_analyze_loop_form): Likewise.
5650 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5651 Compute dependences lazily.
5652 (vect_record_base_alignments): Use shared datarefs/ddrs.
5653 (vect_verify_datarefs_alignment): Likewise.
5654 (vect_analyze_data_refs_alignment): Likewise.
5655 (vect_analyze_data_ref_accesses): Likewise.
5656 (vect_analyze_data_refs): Likewise.
5657 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5658 constructor parameter for shared part.
5659 (vect_analyze_loop_form): Pass in shared part and adjust.
5660 (vect_analyze_loop_2): Pass in storage for the number of
5661 stmts. Move loop nest finding to the caller. Compute
5663 (vect_analyze_loop): Pass in shared part.
5664 (vect_transform_loop): Verify shared datarefs are unchanged.
5665 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5666 constructor parameter for shared part.
5667 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5668 (vect_slp_bb): Verify shared datarefs are unchanged before
5670 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5672 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5673 DR_MISALIGNMENT_UNINITIALIZED.
5674 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5675 parameter for shared part.
5676 (vec_info::~vec_info): Adjust.
5677 (vec_info_shared::vec_info_shared): New.
5678 (vec_info_shared::~vec_info_shared): Likewise.
5679 (vec_info_shared::save_datarefs): Likewise.
5680 (vec_info_shared::check_datarefs): Likewise.
5681 (try_vectorize_loop_1): Construct shared part live for analyses
5682 of a single loop for multiple vector sizes.
5683 * tree-parloops.c (gather_scalar_reductions): Adjust.
5685 2018-06-25 Richard Biener <rguenther@suse.de>
5687 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5688 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5689 (vect_analyze_data_refs): Remove similar code from here and
5690 simplify accordingly.
5692 2018-06-25 Richard Biener <rguenther@suse.de>
5694 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5695 for reverse storage order accesses rather than asserting
5696 they cannot happen here.
5698 2018-06-25 Tom de Vries <tdevries@suse.de>
5701 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5702 Use data16 instead of .byte for insn prefix.
5704 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5707 * parser.c (make_char_string_pack): Pass this literal chars
5708 through cpp_interpret_string.
5709 (cp_parser_userdef_numeric_literal): Check the result of
5710 make_char_string_pack.
5712 2018-06-24 Maya Rashish <coypu@sdf.org>
5714 * ginclude/stddef.h: Simplify conditions around avoiding
5715 re-definition of __size_t.
5717 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5719 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5720 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5722 2018-06-22 Maya Rashish <coypu@sdf.org>
5724 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5725 made to OpenBSD and NetBSD through the years.
5727 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5729 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5730 behavior of vec_pack (vector double, vector double) to match
5731 behavior of vec_float2 (vector double, vector double).
5733 2018-06-22 Olivier Hainque <hainque@adacore.com>
5735 * gimplify.c (gimplify_function_tree): Prevent creation
5736 of a trampoline for the address of the current function
5737 passed to entry/exit instrumentation hooks.
5739 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5742 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5745 2018-06-22 Martin Liska <mliska@suse.cz>
5747 PR tree-optimization/86263
5748 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5749 Bail out if is_enabled is false.
5750 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5752 (jump_table_cluster::is_enabled): New function.
5754 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5756 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5757 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5758 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5759 (lto_input_ts_binfo_tree_pointers): Likewise.
5760 * tree-streamer-out.c (streamer_write_tree_bitfields,
5761 write_ts_binfo_tree_pointers): Likewise.
5762 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5764 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5766 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5768 2018-06-22 Martin Liska <mliska@suse.cz>
5770 * symbol-summary.h (get): Make it pure and inline move
5771 functionality from ::get function.
5772 (get): Remove and inline into ::get and ::get_create.
5773 (get_create): Move code from ::get function.
5775 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5778 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5779 -x assembler-with-cpp.
5781 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5783 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5784 _FILE_OFFSET_BITS=64 for C++.
5786 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5788 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5789 conversion insn that shows up when pr85657-3.c is compiled using
5790 IEEE 128-bit long double.
5791 (neg<mode>2_internal): Use the correct mode to check whether the
5792 mode is IBM extended.
5793 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5794 multiply and divide external functions from being created more
5797 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5799 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5801 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5802 the edge can be forwarded.
5803 (cfg_layout_merge_blocks): Likewise.
5805 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5807 * except.c (finish_eh_generation): Commit edge insertions only after
5808 the EH edges have been redirected from post-landing to landing pads.
5810 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5812 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5813 create_tmp_var_for to create the FRAME decl.
5814 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5816 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5818 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5819 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5820 debug statement when resetting its value.
5821 (expand_call_inline): Copy the locus of the call onto the assignment
5822 of the return value, if any. Use local variable in more cases.
5824 2018-06-21 Martin Liska <mliska@suse.cz>
5826 * ipa-pure-const.c (propagate_nothrow): Use
5827 funct_state_summaries->get.
5828 (dump_malloc_lattice): Likewise.
5829 (propagate_malloc): Likewise.
5831 2018-06-21 Richard Biener <rguenther@suse.de>
5833 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5834 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5835 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5837 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5838 BLOCK_ABSTRACT_ORIGIN unconditionally.
5840 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5842 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5844 * ipa-reference.c (ipa_reference_c_finalize): Delete
5845 ipa_ref_opt_sum_summaries and set it to NULL.
5847 2018-06-21 Tom de Vries <tdevries@suse.de>
5849 PR tree-optimization/85859
5850 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5851 test with comment from bb_no_side_effects_p.
5853 2018-06-21 Richard Biener <rguenther@suse.de>
5855 PR tree-optimization/86232
5856 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5857 max for constant niter.
5859 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5861 * config/aarch64/aarch64-simd.md
5862 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5864 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5866 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5867 Make opernads of the unspec commutative.
5869 2018-06-21 Richard Biener <rguenther@suse.de>
5871 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5872 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5873 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5874 (vect_analyze_data_ref_dependence): Re-order checks to deal with
5876 (vect_record_base_alignments): Do not record base alignment
5877 for gathers or scatters.
5878 (vect_compute_data_ref_alignment): Drop return value that is always
5879 true. Bail out early for gathers or scatters.
5880 (vect_enhance_data_refs_alignment): Bail out early for gathers
5882 (vect_find_same_alignment_drs): Likewise.
5883 (vect_analyze_data_refs_alignment): Remove dead code.
5884 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5885 (vect_analyze_data_refs): For possible gathers or scatters do
5886 not create an alternate DR, just check their possible validity
5887 and mark them. Adjust DECL_NONALIASED handling to not rely
5889 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5890 update inits of gathers or scatters.
5891 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5892 Also copy gather/scatter flag to pattern vinfo.
5894 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
5896 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5897 behavior of vec_packsu (vector unsigned long long, vector unsigned
5898 long long) to match behavior of vec_packs with same signature.
5900 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
5901 Thomas Schwinge <thomas@codesourcery.com>
5902 Cesar Philippidis <cesar@codesourcery.com>
5904 * gimplify.c (gimplify_scan_omp_clauses): Add support for
5905 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5906 (gimplify_adjust_omp_clauses): Likewise.
5907 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5908 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5909 (gimplify_omp_target_update): Update handling of acc update and
5911 * omp-low.c (install_var_field): Remove unused parameter
5912 base_pointers_restrict.
5913 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5914 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5916 (omp_target_base_pointers_restrict_p): Delete.
5917 (scan_omp_target): Update call to scan_sharing_clauses.
5918 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5920 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5921 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5922 (convert_local_omp_clauses): Likewise.
5923 * tree-pretty-print.c (dump_omp_clause): Likewise.
5924 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
5926 (omp_clause_code_name): Likewise.
5928 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5931 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5934 PR tree-optimization/86231
5935 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5936 anti-range don't overwrite *vr0min before using it to compute *vr0max.
5938 2018-06-20 Tom de Vries <tdevries@suse.de>
5940 PR tree-optimization/86097
5941 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5942 iv type if signedness of iv type is not the same as that of *nit.
5944 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5946 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
5947 EDGE_EH edges, verify they are all EDGE_EH.
5949 2018-06-20 Maya Rashish <coypu@sdf.org>
5951 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5953 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5955 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5956 * config/aarch64/aarch64.c (xgene1_tunings): Add
5957 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5958 (aarch64_mode_valid_for_sched_fusion_p):
5959 Allow 16-byte modes.
5960 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5961 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5963 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5965 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
5966 * config/aarch64/iterators.md (VQ2): New mode iterator.
5968 2018-06-20 Martin Liska <mliska@suse.cz>
5970 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5971 Change default ratio from 10 to 8.
5973 2018-06-20 Martin Liska <mliska@suse.cz>
5975 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5977 (bit_test_cluster::find_bit_tests): Likewise.
5978 (switch_decision_tree::analyze_switch_statement): Find clusters.
5979 * tree-switch-conversion.h (struct jump_table_cluster): Document
5982 2018-06-20 Martin Liska <mliska@suse.cz>
5984 * tree-switch-conversion.c (switch_conversion::collect):
5985 Record m_uniq property.
5986 (switch_conversion::expand): Bail out for special conditions.
5987 (group_cluster::~group_cluster): New.
5988 (group_cluster::group_cluster): Likewise.
5989 (group_cluster::dump): Likewise.
5990 (jump_table_cluster::emit): New.
5991 (switch_decision_tree::fix_phi_operands_for_edges): New.
5992 (struct case_node): Remove struct.
5993 (jump_table_cluster::can_be_handled): New.
5994 (case_values_threshold): Moved to header.
5995 (reset_out_edges_aux): Likewise.
5996 (jump_table_cluster::is_beneficial): New.
5997 (bit_test_cluster::can_be_handled): Likewise.
5998 (add_case_node): Remove.
5999 (bit_test_cluster::is_beneficial): New.
6000 (case_bit_test::cmp): New.
6001 (bit_test_cluster::emit): New.
6002 (expand_switch_as_decision_tree_p): Remove.
6003 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6004 (fix_phi_operands_for_edge): Likewise.
6005 (switch_decision_tree::analyze_switch_statement): New.
6006 (compute_cases_per_edge): Move ...
6007 (switch_decision_tree::compute_cases_per_edge): ... here.
6008 (try_switch_expansion): Likewise.
6009 (switch_decision_tree::try_switch_expansion): Likewise.
6010 (record_phi_operand_mapping): Likewise.
6011 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6012 (emit_case_decision_tree): Likewise.
6013 (switch_decision_tree::emit): Likewise.
6014 (balance_case_nodes): Likewise.
6015 (switch_decision_tree::balance_case_nodes): Likewise.
6016 (dump_case_nodes): Likewise.
6017 (switch_decision_tree::dump_case_nodes): Likewise.
6018 (emit_jump): Likewise.
6019 (switch_decision_tree::emit_jump): Likewise.
6020 (emit_cmp_and_jump_insns): Likewise.
6021 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6022 (emit_case_nodes): Likewise.
6023 (switch_decision_tree::emit_case_nodes): Likewise.
6024 (conditional_probability): Remove.
6025 * tree-switch-conversion.h (enum cluster_type): New.
6027 (struct cluster): Likewise.
6028 (cluster::cluster): Likewise.
6029 (struct simple_cluster): Likewise.
6030 (simple_cluster::simple_cluster): Likewise.
6031 (struct group_cluster): Likewise.
6032 (struct jump_table_cluster): Likewise.
6033 (struct bit_test_cluster): Likewise.
6034 (struct min_cluster_item): Likewise.
6035 (struct case_tree_node): Likewise.
6036 (case_tree_node::case_tree_node): Likewise.
6037 (jump_table_cluster::case_values_threshold): Likewise.
6038 (struct case_bit_test): Likewise.
6039 (struct switch_decision_tree): Likewise.
6040 (struct switch_conversion): Likewise.
6041 (switch_decision_tree::reset_out_edges_aux): Likewise.
6043 2018-06-20 Martin Liska <mliska@suse.cz>
6045 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6046 (hoist_edge_and_branch_if_true): Likewise.
6047 (expand_switch_using_bit_tests_p): Likewise.
6048 (struct case_bit_test): Likewise.
6049 (case_bit_test_cmp): Likewise.
6050 (emit_case_bit_tests): Likewise.
6051 (switch_conversion::switch_conversion): New class.
6052 (struct switch_conv_info): Remove old struct.
6053 (collect_switch_conv_info): More to ...
6054 (switch_conversion::collect): ... this.
6055 (check_range): Likewise.
6056 (switch_conversion::check_range): Likewise.
6057 (check_all_empty_except_final): Likewise.
6058 (switch_conversion::check_all_empty_except_final): Likewise.
6059 (check_final_bb): Likewise.
6060 (switch_conversion::check_final_bb): Likewise.
6061 (create_temp_arrays): Likewise.
6062 (switch_conversion::create_temp_arrays): Likewise.
6063 (free_temp_arrays): Likewise.
6064 (gather_default_values): Likewise.
6065 (switch_conversion::gather_default_values): Likewise.
6066 (build_constructors): Likewise.
6067 (switch_conversion::build_constructors): Likewise.
6068 (constructor_contains_same_values_p): Likewise.
6069 (switch_conversion::contains_same_values_p): Likewise.
6070 (array_value_type): Likewise.
6071 (switch_conversion::array_value_type): Likewise.
6072 (build_one_array): Likewise.
6073 (switch_conversion::build_one_array): Likewise.
6074 (build_arrays): Likewise.
6075 (switch_conversion::build_arrays): Likewise.
6076 (gen_def_assigns): Likewise.
6077 (switch_conversion::gen_def_assigns): Likewise.
6078 (prune_bbs): Likewise.
6079 (switch_conversion::prune_bbs): Likewise.
6080 (fix_phi_nodes): Likewise.
6081 (switch_conversion::fix_phi_nodes): Likewise.
6082 (gen_inbound_check): Likewise.
6083 (switch_conversion::gen_inbound_check): Likewise.
6084 (process_switch): Use the newly created class.
6085 (switch_conversion::expand): New.
6086 (switch_conversion::~switch_conversion): New.
6087 * tree-switch-conversion.h: New file.
6089 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6091 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6092 tree-vect-patterns.c.
6093 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6094 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6095 (vect_recog_sad_pattern): Likewise.
6096 (vect_recog_widen_sum_pattern): Likewise.
6097 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6098 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6099 (vect_recog_rotate_pattern): Likewise.
6100 (vect_recog_mult_pattern): Likewise.
6101 (vect_recog_vector_vector_shift_pattern): Likewise.
6102 (vect_recog_divmod_pattern): Likewise.
6103 (vect_recog_mixed_size_cond_pattern): Likewise.
6104 (vect_recog_bool_pattern): Likewise.
6105 (vect_recog_mask_conversion_pattern): Likewise.
6106 (vect_try_gather_scatter_pattern): Likewise.
6107 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6108 (vect_recog_over_widening_pattern): Likewise.
6109 (vect_recog_gather_scatter_pattern): Likewise.
6110 (vect_recog_func_ptr): Move from tree-vectorizer.h
6111 (vect_vect_recog_func_ptrs): Move further down the file.
6112 (vect_recog_func): Likewise. Remove the third argument.
6113 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6114 (vect_pattern_recog_1): Expect the pattern function to do any
6115 necessary target tests. Also expect it to provide a vector type.
6116 Remove the type_in handling.
6118 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6120 * tree-vect-patterns.c (vect_pattern_detected): New function.
6121 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6122 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6123 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6124 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6125 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6126 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6127 (vect_recog_mask_conversion_pattern)
6128 (vect_try_gather_scatter_pattern): Likewise.
6130 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6132 * tree-vect-patterns.c (vect_get_internal_def): New function.
6133 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6134 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6135 (search_type_for_mask_1): Use it.
6137 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6139 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6140 redundant WIDEN_SUM_EXPR handling.
6141 (vect_recog_sad_pattern): Likewise.
6143 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6145 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6146 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6147 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6148 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6149 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6151 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6153 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6154 use the stmt_vec_info of the original bb statement for the
6155 new zero assignment, even if the call is part of a pattern.
6157 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6159 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6160 that the sequence is attached to the original statement rather
6161 than the pattern statement.
6162 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6163 PATTERN_DEF_SEQ from the original statement rather than
6164 the main pattern statement.
6165 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6166 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6167 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6169 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6171 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6172 definition statements before the early exit for statements that aren't
6174 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6176 (vect_transform_loop): ...here. Process pattern definition
6177 statements without first checking whether the main pattern
6178 statement is live or relevant.
6180 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6182 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6183 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6185 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6187 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6188 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6189 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6190 (expand_block_compare): Change select_block_compare_mode call.
6191 (expand_strncmp_align_check): Use new functions, fix comment.
6192 (emit_final_str_compare_gpr): New function.
6193 (expand_strn_compare): Refactor and clean up code.
6194 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6196 2018-06-19 Tony Reix <tony.reix@atos.com>
6197 Damien Bergamini <damien.bergamini@atos.com>
6198 David Edelsohn <dje.gcc@gmail.com>
6200 * collect2.c (static_obj): New variable.
6201 (static_libs): New variable.
6202 (is_in_list): Uncomment declaration.
6203 (main): Track AIX libraries linked statically.
6204 (is_in_list): Uncomment definition.
6205 (scan_prog_file): Don't add AIX shared libraries initializer
6206 to constructor list if linking statically.
6208 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6210 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6212 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6214 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6216 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6219 2018-06-19 Martin Liska <mliska@suse.cz>
6221 * config/i386/i386.c (ix86_can_inline_p): Do not use
6222 ipa_fn_summaries::get_create.
6223 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6225 (devirtualization_time_bonus): Likewise.
6226 (ipcp_propagate_stage): Likewise.
6227 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6228 (edge_set_predicate): Likewise.
6229 (evaluate_conditions_for_known_args): Likewise.
6230 (evaluate_properties_for_edge): Likewise.
6231 (ipa_call_summary::reset): Tranform to ...
6232 (ipa_call_summary::~ipa_call_summary): ... this.
6233 (ipa_fn_summary::reset): Transform to ...
6234 (ipa_fn_summary::~ipa_fn_summary): ... this.
6235 (ipa_fn_summary_t::remove): Rename to ...
6236 (ipa_fn_summary_t::remove_callees): ... this.
6237 (ipa_fn_summary_t::duplicate): Use placement new
6238 instead of memory copy.
6239 (ipa_call_summary_t::duplicate): Likewise.
6240 (ipa_call_summary_t::remove): Remove.
6241 (dump_ipa_call_summary): Change get_create to get.
6242 (ipa_dump_fn_summary): Dump only when summary exists.
6243 (analyze_function_body): Use symbol_summary::get instead
6245 (compute_fn_summary): Likewise.
6246 (estimate_edge_devirt_benefit): Likewise.
6247 (estimate_edge_size_and_time): Likewise.
6248 (inline_update_callee_summaries): Likewise.
6249 (remap_edge_change_prob): Likewise.
6250 (remap_edge_summaries): Likewise.
6251 (ipa_merge_fn_summary_after_inlining): Likewise.
6252 (write_ipa_call_summary): Likewise.
6253 (ipa_fn_summary_write): Likewise.
6254 (ipa_free_fn_summary): Likewise.
6255 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6256 (struct ipa_call_summary): Likewise.
6257 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6259 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6260 (estimate_size_after_inlining): Likewise.
6261 (estimate_growth): Likewise.
6262 (growth_likely_positive): Likewise.
6263 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6264 (inline_call): Likewise.
6265 * ipa-inline.c (caller_growth_limits): Likewise.
6266 (can_inline_edge_p): Likewise.
6267 (can_inline_edge_by_limits_p): Likewise.
6268 (compute_uninlined_call_time): Likewise.
6269 (compute_inlined_call_time): Likewise.
6270 (want_inline_small_function_p): Likewise.
6271 (edge_badness): Likewise.
6272 (update_caller_keys): Likewise.
6273 (update_callee_keys): Likewise.
6274 (inline_small_functions): Likewise.
6275 (inline_to_all_callers_1): Likewise.
6276 (dump_overall_stats): Likewise.
6277 (early_inline_small_functions): Likewise.
6278 (early_inliner): Likewise.
6279 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6280 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6281 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6282 * ipa-split.c (execute_split_functions): Likewise.
6283 * symbol-summary.h: Likewise.
6284 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6286 2018-06-19 Richard Biener <rguenther@suse.de>
6288 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6289 (vectorize_loops): ... here. Fix dbgcnt handling.
6290 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6292 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6295 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6296 ieee128 argument takes up only one (vector) register, not two (floating
6299 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6301 * gimplify.c (gimplify_init_constructor): Really never clear for an
6302 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6304 2018-06-19 Richard Biener <rguenther@suse.de>
6306 PR tree-optimization/86179
6307 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6308 after failed recognition.
6310 2018-06-18 Martin Sebor <msebor@redhat.com>
6313 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6314 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6315 Handle integer subtraction.
6316 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6317 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6319 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6321 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6322 param from rtx to rtx_insn *.
6323 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6325 (frv_ifcvt_modify_insn): Likwise.
6326 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6327 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6328 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6329 as_a <rtx_insn *> cast to local "unprotected_region" once
6330 it's been established that it's not NULL or pc_rtx.
6331 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6332 param "sethi" from rtx to rtx_insn *.
6333 (nds32_group_float_insns): Likewise for param "insn".
6334 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6336 (vax_output_int_subtract): Likewise.
6337 * config/vax/vax.c (vax_output_int_add): Likewise for param
6339 (vax_output_int_subtract): Likewise.
6340 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6341 (emit_pattern_after): Likewise for param "after".
6342 (emit_insn_after): Likewise.
6343 (emit_jump_insn_after): Likewise.
6344 (emit_call_insn_after): Likewise.
6345 (emit_debug_insn_after): Likewise.
6346 (emit_pattern_before): Likewise for param "before".
6347 (emit_insn_before): Likewise.
6348 (emit_jump_insn_before): Likewise.
6349 * final.c (get_insn_template): Likewise for param "insn", removing
6351 * output.h (get_insn_template): Likewise for 2nd param.
6352 * rtl.h (emit_insn_before): Likewise.
6353 (emit_jump_insn_before): Likewise.
6354 (emit_debug_insn_before_noloc): Likewise.
6355 (emit_insn_after): Likewise.
6356 (emit_jump_insn_after): Likewise.
6357 (emit_call_insn_after): Likewise.
6358 (emit_debug_insn_after): Likewise.
6359 (set_insn_deleted): Likewise for param.
6361 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6364 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6365 floating point modes, so that IFmode is numerically greater than
6366 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6367 to declare the ordering. This prevents IFmode from being
6368 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6369 machine. Include rs6000-modes.h to share the fractional values
6370 between genmodes* and the rest of the compiler.
6374 * config/rs6000/rs6000-modes.h: New file.
6375 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6376 meaning of rs6000_long_double_size so that 126..128 selects an
6377 appropriate 128-bit floating point type.
6378 (rs6000_option_override_internal): Likewise.
6379 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6380 (TARGET_LONG_DOUBLE_128): Change the meaning of
6381 rs6000_long_double_size so that 126..128 selects an appropriate
6382 128-bit floating point type.
6383 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6384 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6385 source and destination to match the standard usage.
6386 (truncifkf2): Likewise.
6387 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6388 ISA 2.07 to use an explicit clobber, instead of passing in a
6390 (copysign<mode>3_soft): Likewise.
6392 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6394 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6395 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6396 (vect_slp_analyze_instance_dependence): Likewise.
6397 (vect_enhance_data_refs_alignment): Likewise.
6398 (vect_analyze_data_refs_alignment): Likewise.
6399 (vect_slp_analyze_and_verify_instance_alignment
6400 (vect_analyze_data_ref_accesses): Likewise.
6401 (vect_prune_runtime_alias_test_list): Likewise.
6402 (vect_analyze_data_refs): Likewise.
6403 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6404 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6405 (vect_analyze_scalar_cycles_1): Likewise.
6406 (vect_get_loop_niters): Likewise.
6407 (vect_analyze_loop_form_1): Likewise.
6408 (vect_update_vf_for_slp): Likewise.
6409 (vect_analyze_loop_operations): Likewise.
6410 (vect_analyze_loop): Likewise.
6411 (vectorizable_induction): Likewise.
6412 (vect_transform_loop): Likewise.
6413 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6414 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6415 (vect_make_slp_decision): Likewise.
6416 (vect_detect_hybrid_slp): Likewise.
6417 (vect_slp_analyze_operations): Likewise.
6418 (vect_slp_bb): Likewise.
6419 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6420 (vectorizable_bswap): Likewise.
6421 (vectorizable_call): Likewise.
6422 (vectorizable_simd_clone_call): Likewise.
6423 (vectorizable_conversion): Likewise.
6424 (vectorizable_assignment): Likewise.
6425 (vectorizable_shift): Likewise.
6426 (vectorizable_operation): Likewise.
6427 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6429 2018-06-18 Martin Sebor <msebor@redhat.com>
6431 PR tree-optimization/81384
6432 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6433 * builtins.c (expand_builtin_strnlen): New function.
6434 (expand_builtin): Call it.
6435 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6436 * builtins.def (BUILT_IN_STRNLEN): New.
6437 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6438 Warn for bounds in excess of maximum object size.
6439 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6440 single-value ranges. Handle strnlen.
6441 (handle_builtin_strlen): Handle strnlen.
6442 (strlen_check_and_optimize_stmt): Same.
6443 * doc/extend.texi (Other Builtins): Document strnlen.
6445 2018-06-18 Maya Rashish <coypu@sdf.org>
6447 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6448 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6449 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6451 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6453 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6455 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6457 * tree.c (escaped_string::escape): Replace cast to char * by
6458 const_cast<char *> (unescaped).
6460 2018-06-18 Nick Clifton <nickc@redhat.com>
6463 * tree.c (escaped_string): New class. Converts an unescaped
6464 string into its escaped equivalent.
6465 (warn_deprecated_use): Use the new class to convert the
6466 deprecation message, if present.
6467 (test_escaped_strings): New self test.
6468 (test_c_tests): Add test_escaped_strings.
6469 * doc/extend.texi (deprecated): Add a note that the
6470 deprecation message is affected by the -fmessage-length
6471 option, and that control characters will be escaped.
6472 (#pragma GCC error): Document this pragma.
6473 (#pragma GCC warning): Likewise.
6474 * doc/invoke.texi (-fmessage-length): Document this option's
6475 effect on the #warning and #error preprocessor directives and
6476 the deprecated attribute.
6478 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6480 * tree.c (decl_value_expr_lookup): Revert latest change.
6481 (decl_value_expr_insert): Likewise.
6483 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6485 * gimplify.c (nonlocal_vlas): Delete.
6486 (nonlocal_vla_vars): Likewise.
6487 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6489 (gimplify_body): Do not create and destroy nonlocal_vlas.
6490 * tree-nested.c: Include diagnostic.h.
6491 (use_pointer_in_frame): Tweak.
6492 (lookup_field_for_decl): Add assertion and declare the transformation.
6493 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6494 internal error when the reference is in a wrong context. Do not
6495 create a debug decl by default.
6496 (note_nonlocal_block_vlas): Delete.
6497 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6498 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6499 create a debug decl by default.
6500 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6501 call is in a wrong context.
6502 (fixup_vla_decls): New function.
6503 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6504 debug variables were created.
6505 * tree.c (decl_value_expr_lookup): Add checking assertion.
6506 (decl_value_expr_insert): Likewise.
6508 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6511 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6512 * tree-scalar-evolution.c (interpret_expr): Likewise.
6513 (expression_expensive_p): Likewise.
6514 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6515 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6516 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6517 (ssa_defined_by_minus_one_stmt_p): New.
6519 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6522 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6523 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6524 * dojump.c (do_jump): Likewise.
6525 * expr.c (expand_expr_real_2): Check operand type's sign.
6526 * fold-const.c (const_unop): Handle ABSU_EXPR.
6527 (fold_abs_const): Likewise.
6528 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6529 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6530 (strip_sign_op_1): Likesise.
6531 * match.pd: Add new pattern to generate ABSU_EXPR.
6532 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6533 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6534 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6535 * tree-inline.c (estimate_operator_cost): Likewise.
6536 * tree-pretty-print.c (dump_generic_node): Likewise.
6537 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6538 * tree.def (ABSU_EXPR): New.
6540 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6543 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6544 documented as preserved for backward compatibility only.
6545 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6547 PR rtl-optimization/86108
6548 * bb-reorder.c (create_forwarder_block): Renamed to ...
6549 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6550 jump from new landing pad to the second part.
6551 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6554 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6557 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6558 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6559 Only call store_expr for halves if the mode is the same.
6562 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6563 Fix up comment formatting.
6565 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6567 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6568 (typed_splay_tree::closure,
6569 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6570 (typed_splay_tree::typed_splay_tree,
6571 typed_splay_tree::operator =): Declared private.
6572 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6573 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6574 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6575 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6576 typed_splay_tree::splay_tree_splay,
6577 typed_splay_tree::splay_tree_foreach_helper,
6578 typed_splay_tree::splay_tree_insert,
6579 typed_splay_tree::splay_tree_remove,
6580 typed_splay_tree::splay_tree_lookup,
6581 typed_splay_tree::splay_tree_predecessor,
6582 typed_splay_tree::splay_tree_successor,
6583 typed_splay_tree::splay_tree_min,
6584 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6585 (typed_splay_tree::root, typed_splay_tree::comp,
6586 typed_splay_tree::delete_key,
6587 typed_splay_tree::delete_value): New data members.
6588 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6589 typed_splay_tree::remove.
6591 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6593 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6594 -mginv and -mno-ginv to the assembler.
6595 * config/mips/mips.opt (-mcrc): New option.
6597 * doc/invoke.text (-mcrc): Document.
6600 2018-06-15 Nick Clifton <nickc@redhat.com>
6603 * tree.c (escaped_string): New class. Converts an unescaped
6604 string into its escaped equivalent.
6605 (warn_deprecated_use): Use the new class to convert the
6606 deprecation message, if present.
6607 (test_escaped_strings): New self test.
6608 (test_c_tests): Add test_escaped_strings.
6609 * doc/extend.texi (deprecated): Add a note that the
6610 deprecation message is affected by the -fmessage-length
6611 option, and that control characters will be escaped.
6612 (#pragma GCC error): Document this pragma.
6613 (#pragma GCC warning): Likewise.
6614 * doc/invoke.texi (-fmessage-length): Document this option's
6615 effect on the #warning and #error preprocessor directives and
6616 the deprecated attribute.
6618 2018-06-15 Richard Biener <rguenther@suse.de>
6620 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6621 here, also noting vector size used.
6622 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6623 size used in MSG_OPTIMIZED_LOCATIONS dump.
6624 (pass_slp_vectorize::execute): Adjust.
6626 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6629 * config/arc/arc.c (arc_return_address_register): Fix
6632 2018-06-15 Richard Biener <rguenther@suse.de>
6635 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6636 leave useless conversion stripping to force_gimple_operand_gsi.
6637 (gimplify_build2): Likewise.
6638 (gimplify_build1): Likewise.
6640 2018-06-15 Richard Biener <rguenther@suse.de>
6643 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6644 before adjusting their block.
6646 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6648 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6649 multilibs for *-*-rtems*.
6650 * config/riscv/t-rtems: New file.
6652 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6655 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6656 unsigned_type_for returns NULL.
6659 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6660 subregs of multi-word pseudos unless the float mode has word size.
6662 2018-06-14 Richard Biener <rguenther@suse.de>
6665 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6666 duplicate and harmful type_hash_canon.
6667 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6669 2018-06-14 Richard Biener <rguenther@suse.de>
6672 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6675 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6677 * config/rtems.h (STDINT_LONG32): Define.
6679 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6680 Prachi Godbole <prachi.godbole@imgtec.com>
6682 * config/mips/mips-cpus.def: Define P6600.
6683 * config/mips/mips-tables.opt: Regenerate.
6684 * config/mips/mips.c (mips_ucbranch_type): New enum.
6685 (mips_rtx_cost_data): Add support for P6600.
6686 (mips_issue_rate): Likewise.
6687 (mips_multipass_dfa_lookahead): Likewise.
6688 (mips_avoid_hazard): Likewise.
6689 (mips_reorg_process_insns): Likewise.
6690 (mips_classify_branch_p6600): New function.
6691 * config/mips/mips.h (TUNE_P6600): New define.
6692 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6693 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6694 * config/mips/mips.md: Include p6600.md.
6695 (processor): Add p6600.
6696 * config/mips/p6600.md: New file.
6697 * doc/invoke.texi: Add p6600 to supported architectures.
6699 2018-06-13 Martin Sebor <msebor@redhat.com>
6701 PR tree-optimization/86114
6702 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6704 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6706 2018-06-13 Richard Biener <rguenther@suse.de>
6708 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6709 Properly set vector type of the intermediate stmt.
6710 * tree-vect-stmts.c (vectorizable_operation): The destination
6711 var always has vectype_out type.
6713 2018-06-13 Jeff Law <law@redhat.com>
6715 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6716 integer 0 for argument to print_rtl_with_bb.
6717 (rl78_reorg): Likewise.
6719 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6721 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6722 from rtx to rtx_insn *.
6723 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6725 (add_sched_insns_for_speculation): Likewise for local "target",
6726 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6727 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6728 from rtx_insn ** to rtx_code_label **.
6729 (reorg_emit_nops): Likewise.
6730 (c6x_reorg): Likewise for local "call_labels".
6731 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6733 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6734 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6735 the loops over LABEL_REFS.
6736 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6738 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6739 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6740 (split_branches): Strengthen local "olabel" from rtx to
6741 rtx_insn *, adding a safe_as_a cast.
6742 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6744 (add_insn_after): Likewise for first two params.
6745 (add_insn_before): Likewise.
6746 (remove_insn): Likewise for param.
6747 (emit_pattern_before_noloc): Likewise for second and third params.
6748 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6749 (emit_call_insn_before_noloc): Likewise.
6750 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6752 (emit_barrier_before): Likewise.
6753 (emit_label_before): Strengthen "label" param from "rtx" to
6754 "rtx_code_label *". Strengthen "before" param from "rtx" to
6756 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6758 (emit_pattern_after_noloc): Likewise.
6759 (emit_insn_after_noloc): Likewise.
6760 (emit_jump_insn_after_noloc): Likewise.
6761 (emit_call_insn_after_noloc): Likewise.
6762 (emit_debug_insn_after_noloc): Likewise.
6763 (emit_barrier_after): Likewise.
6764 (emit_label_after): Likewise for both params.
6765 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6766 "loc" param from "int" to "location_t".
6767 (emit_insn_after_setloc): Likewise.
6768 (emit_jump_insn_after_setloc): Likewise.
6769 (emit_call_insn_after_setloc): Likewise.
6770 (emit_debug_insn_after_setloc): Likewise.
6771 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6772 "loc" param from "int" to "location_t".
6773 (emit_pattern_before): Convert NULL_RTX to NULL.
6774 (emit_insn_before_setloc): Convert "loc" param from "int" to
6776 (emit_jump_insn_before_setloc): Likewise.
6777 (emit_call_insn_before_setloc): Likewise.
6778 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6779 rtx_insn *. Convert "loc" param from "int" to "location_t".
6780 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6781 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6782 Convert 3rd param from "int" to "location_t".
6783 (emit_barrier_before, emit_barrier_after, next_real_insn):
6784 Strengthen param from rtx to rtx_insn *.
6785 (emit_label_before): Strengthen 1st param from "rtx" to
6786 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6788 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6789 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6790 Strengthen 2nd param from "rtx" to "rtx_insn *".
6791 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6792 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6793 Likewise. Convert 3rd param from "int" to "location_t".
6794 (emit_label_after): Strengthen 1st param from "rtx" to
6796 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6798 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6799 from "rtx" to "rtx_insn *".
6801 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6803 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6804 bodies streamed in with -Q.
6805 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6806 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6807 * gimple-streamer-out.c: Include gimple-pretty-print.h
6808 (output_bb): Dump stmts streamed.
6809 * lto-section-out.c: Include print-tree.h
6810 (lto_begin_section): Dump sections created.
6811 (lto_output_decl_index): Dump decl encoded.
6812 * lto-streamer-out.c: Include print-tree.h
6813 (create_output_block): Dump output block created.
6814 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6815 (output_function): Dump function output.
6816 (output_constructor): Dump constructor streamed.
6817 (write_global_stream): Output indexes encoded.
6818 (produce_asm_for_decls): Dump streams encoded.
6819 * lto-streamer.c (streamer_dump_file): New global var.
6820 * lto-streamer.h (streamer_dump_file): Declare.
6821 * passes.c (ipa_write_summaries): Initialize streamer dump.
6822 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6825 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6828 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6829 offsets for register save directives. Emit a second batch of save
6830 directives, if need be, when the function accesses prior frames.
6832 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6834 * config/arc/fpu.md (fmasf4): Force operand to register.
6835 (fnmasf4): Likewise.
6837 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6839 * config/arc/arc-protos.h (arc_pad_return): Remove.
6840 * config/arc/arc.c (machine_function): Remove force_short_suffix
6842 (arc_print_operand): Adjust printing of '&'.
6843 (arc_verify_short): Remove conditional printing of short suffix.
6844 (arc_final_prescan_insn): Remove reference to size_reason.
6845 (pad_return): New function.
6846 (arc_reorg): Call pad_return.
6847 (arc_pad_return): Remove.
6848 (arc_init_machine_status): Remove reference to force_short_suffix.
6849 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6850 (attr length): When attribute iscompact is true force to 2
6851 regardless; in the case of maybe check if we want to force the
6852 instruction to have 4 bytes length.
6853 (nopv): Change it to generate 4 byte long nop as well.
6854 (blockage): New pattern.
6855 (simple_return): Remove call to arc_pad_return.
6856 (p_return_i): Likewise.
6858 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6860 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6862 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6864 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6867 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6869 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6870 for ARC700 and ARCv2.
6872 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
6875 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6876 operands[2] instead of operands[1].
6879 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6881 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6882 case, check whether the outer register overlaps an unallocatable
6883 register, not just whether it fits the required class.
6885 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6887 * poly-int.h (can_div_trunc_p): Add new overload in which all values
6889 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6890 (memrefs_conflict_p): Likewise.
6891 (init_alias_analysis): Likewise.
6892 * cfgexpand.c (expand_debug_expr): Likewise.
6893 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6894 * cse.c (fold_rtx): Likewise.
6895 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6896 * expr.c (emit_block_move_hints): Likewise.
6897 (clear_storage_hints, push_block, emit_push_insn): Likewise.
6898 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6899 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6900 (emit_group_store): Likewise.
6901 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
6902 to read the PRE/POST_MODIFY increment.
6903 * calls.c (store_one_arg): Use strip_offset.
6904 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6906 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6908 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6909 (simplify_binary_operation_1): Extend CONST_INT handling to
6911 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6912 than a HOST_WIDE_INT.
6913 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6915 (adjust_mems, add_stores): Update accodingly.
6916 (vt_canonicalize_addr): Track polynomial offsets.
6917 (emit_note_insn_var_location): Likewise.
6918 (vt_add_function_parameter): Likewise.
6919 (vt_initialize): Likewise.
6921 2018-06-12 Jeff Law <law@redhat.com>
6923 * config.gcc (alpha*-*-freebsd*): Remove.
6924 * config/alpha/freebsd.h: Remove.
6926 2018-06-12 David Malcolm <dmalcolm@redhat.com>
6929 * spellcheck-tree.c (levenshtein_distance): Rename to...
6930 (get_edit_distance): ...this, and update for underlying renaming.
6931 * spellcheck-tree.h (levenshtein_distance): Rename to...
6932 (get_edit_distance): ...this.
6933 * spellcheck.c (levenshtein_distance): Rename to...
6934 (get_edit_distance): ...this. Convert from Levenshtein distance
6935 to Damerau-Levenshtein distance by supporting transpositions of
6936 adjacent characters. Rename "v1" to "v_next" and "v0" to
6938 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6939 (selftest::test_edit_distance_unit_test_oneway): ...this, and
6940 update for underlying renaming.
6941 (selftest::levenshtein_distance_unit_test): Rename to...
6942 (selftest::test_get_edit_distance_unit): ...this, and update for
6943 underlying renaming.
6944 (selftest::test_find_closest_string): Add example from PR 69968
6945 where transposition helps
6946 (selftest::test_metric_conditions): Update for renaming.
6947 (selftest::test_metric_conditions): Likewise.
6948 (selftest::spellcheck_c_tests): Likewise.
6949 * spellcheck.h (levenshtein_distance): Rename both overloads to...
6950 (get_edit_distance): ...this.
6951 (best_match::consider): Update for renaming.
6953 2018-06-12 Martin Sebor <msebor@redhat.com>
6955 PR tree-optimization/85259
6956 * builtins.c (compute_objsize): Handle constant offsets.
6957 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6958 true iff a warning has been issued.
6959 * gimple.h (gimple_nonartificial_location): New function.
6960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6961 gimple_nonartificial_location and handle -Wno-system-headers.
6962 (handle_builtin_stxncpy): Same.
6964 2018-06-12 Martin Sebor <msebor@redhat.com>
6967 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
6969 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
6971 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
6972 BUILTIN_VEC_XST entries for pointer to double and long long.
6974 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
6977 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6979 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6982 2018-06-12 Martin Liska <mliska@suse.cz>
6984 * doc/options.texi: Document IntegerRange.
6986 2018-06-12 Martin Liska <mliska@suse.cz>
6988 * config/i386/i386.opt: Make MPX-related options as Deprecated.
6989 * opt-functions.awk: Handle Deprecated flag.
6990 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6992 (read_cmdline_option): Report warning for a deprecated option.
6993 * opts.h (struct cl_option): Add new field cl_deprecated.
6994 (CL_ERR_DEPRECATED): New.
6996 2018-06-12 Martin Liska <mliska@suse.cz>
6998 * doc/options.texi: Document Deprecated option flag.
7000 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7002 * config/arc/arc-arch.h (arc_extras): New enum.
7003 (arc_cpu_t):Add field extra.
7004 (arc_cpu_types): Consider the extras.
7005 * config/arc/arc-cpus.def: Add extras info.
7006 * config/arc/arc-opts.h (processor_type): Consider extra field.
7007 * config/arc/arc.c (arc_override_options): Handle extra field.
7009 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7011 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7012 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7014 * config/arc/arc.c (arc_init): Update pic warning.
7015 (irq_range): Update irq range parsing warnings.
7016 (arc_override_options): Update various warning messages.
7017 (arc_handle_aux_attribute): Likewise.
7019 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7021 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7023 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7025 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7026 compiler messages directives.
7028 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7030 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7031 * config/mips/mips-tables.opt: Regenerate.
7032 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7034 * doc/invoke.texi: Document -march=i6500.
7036 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7038 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7039 (i6400_gpmul): Add cpu_unit.
7040 (i6400_gpdiv): Likewise.
7041 (i6400_msa_add_d): Update reservations.
7042 (i6400_msa_int_add) Likewise.
7043 (i6400_msa_short_logic3) Likewise.
7044 (i6400_msa_short_logic2) Likewise.
7045 (i6400_msa_short_logic) Likewise.
7046 (i6400_msa_move) Likewise.
7047 (i6400_msa_cmp) Likewise.
7048 (i6400_msa_short_float2) Likewise.
7049 (i6400_msa_div_d) Likewise.
7050 (i6400_msa_long_logic1) Likewise.
7051 (i6400_msa_long_logic2) Likewise.
7052 (i6400_msa_mult) Likewise.
7053 (i6400_msa_long_float2) Likewise.
7054 (i6400_msa_long_float4) Likewise.
7055 (i6400_msa_long_float5) Likewise.
7056 (i6400_msa_long_float8) Likewise.
7057 (i6400_fpu_fadd): Include frint type.
7058 (i6400_fpu_store): New define_insn_reservation.
7059 (i6400_fpu_load): Likewise.
7060 (i6400_fpu_move): Likewise.
7061 (i6400_fpu_fcmp): Likewise.
7062 (i6400_fpu_fmadd): Likewise.
7063 (i6400_int_mult): Include imul3nc type and update reservation.
7064 (i6400_int_div): Include idiv3 type and update reservation.
7065 (i6400_int_load): Update to check type not move_type.
7066 (i6400_int_store): Likewise.
7067 (i6400_int_prefetch): Set zero latency.
7069 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7071 * gcc.c: Document new %@{...} sequence.
7072 (LINK_COMMAND_SPEC): Use it for the -L switches.
7073 (cpp_unique_options): Use it for the -I switches.
7074 (at_file_argbuf): New global variable.
7075 (in_at_file): Likewise.
7076 (alloc_args): Create at_file_argbuf.
7077 (clear_args): Truncate at_file_argbuf.
7078 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7079 (open_at_file): New function.
7080 (close_at_file): Likewise.
7081 (create_at_file): Delete.
7082 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7085 (validate_switches_from_spec): Deal with %@{...} sequence.
7086 (validate_switches): Likewise.
7087 (driver::finalize): Call clear_args.
7089 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7091 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7093 2018-06-11 Martin Sebor <msebor@redhat.com>
7095 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7096 (Wno-class-memaccess): Add @opindex.
7097 (Wno-templates, Wno-multiple-inheritance): Same.
7098 (Wno-virtual-inheritance, Wno-namespaces): Same.
7099 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7100 (Wno-format-overflow, Wno-format-truncation): Same.
7101 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7102 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7103 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7104 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7105 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7106 (Wno-misspelled-isr): Same.
7108 2018-06-11 Martin Sebor <msebor@redhat.com>
7110 * PR tree-optimization/86083
7111 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7113 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7115 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7117 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7120 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7121 on the correct operand.
7122 (*movdi_internal64): Ditto.
7124 2018-06-11 Martin Liska <mliska@suse.cz>
7126 PR tree-optimization/86089
7127 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7129 2018-06-11 Julia Koval <julia.koval@intel.com>
7131 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7132 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7133 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7135 2018-06-11 Olivier Hainque <hainque@adacore.com>
7137 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7138 for Ada with strict dwarf2.
7140 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7143 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7146 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7148 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7150 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7152 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7155 2018-06-08 Martin Liska <mliska@suse.cz>
7157 * tree-cfg.h (debug_function): Fix argument type to match
7160 2018-06-08 Martin Liska <mliska@suse.cz>
7162 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7163 Remove usage of MPX-related (and removed) fields.
7164 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7166 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7168 * cfg.c (debug): Use TDF_NONE rather than 0.
7169 * cfghooks.c (debug): Likewise.
7170 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7171 (struct dump_option_value_info): Convert to...
7172 (struct kv_pair): ...this template type.
7173 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7175 (optinfo_verbosity_options): Likewise.
7176 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7178 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7179 than int for "optgroup_flags" param.
7180 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7182 (dump_dec): Likewise.
7183 (dump_finish): Use TDF_NONE rather than 0.
7184 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7185 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7186 than 0. Update for change to option_ptr.
7187 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7188 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7189 0. Update for changes to optinfo_verbosity_options and
7191 (opt_info_switch_p): Convert optgroup_flags from int to
7193 (dump_basic_block): Use dump_flags_t rather than int
7194 for "dump_kind" param.
7195 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7196 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7197 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7198 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7199 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7200 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7201 TDF_NONE): Convert from macros to...
7202 (enum dump_flag): ...this new enum.
7203 (dump_flags_t): Update to use enum.
7204 (operator|, operator&, operator~, operator|=, operator&=):
7205 Implement for dump_flags_t.
7206 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7207 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7208 Convert from macros to...
7209 (enum optgroup_flag): ...this new enum.
7210 (optgroup_flags_t): New typedef.
7211 (operator|, operator|=): Implement for optgroup_flags_t.
7212 (struct dump_file_info): Convert field "alt_flags" to
7213 dump_flags_t. Convert field "optgroup_flags" to
7215 (dump_basic_block): Use dump_flags_t rather than int for param.
7216 (dump_generic_expr_loc): Likewise.
7217 (dump_dec): Likewise.
7218 (dump_register): Convert param "optgroup_flags" to
7220 (opt_info_enable_passes): Likewise.
7221 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7223 * gimple-pretty-print.c (debug): Likewise.
7224 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7225 (merged_store_group::apply_stores): Likewise.
7226 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7227 * gimple.c (verify_gimple_pp): Likewise.
7228 * graphite-poly.c (print_pbb_body): Likewise.
7229 * passes.c (pass_manager::register_one_dump_file): Convert
7230 local "optgroup_flags" to optgroup_flags_t.
7231 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7233 (debug_body): Likewise.
7234 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7235 to optgroup_flags_t.
7236 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7238 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7239 (convert_mult_to_fma): Likewise.
7240 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7241 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7242 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7243 "dump_kind" to dump_flags_t.
7245 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7247 * config/rs6000/rs6000.c (min, max): Delete.
7249 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7251 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7252 -mabi=spe and -mabi=no-spe.
7254 2018-06-08 Martin Liska <mliska@suse.cz>
7256 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7257 where we expect an existing summary.
7259 2018-06-08 Martin Liska <mliska@suse.cz>
7261 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7262 * ipa-inline.h (estimate_edge_growth): Likewise.
7264 2018-06-08 Martin Liska <mliska@suse.cz>
7266 * cgraph.c (function_version_hasher::hash): Use
7267 cgraph_node::get_uid ().
7268 (function_version_hasher::equal):
7269 * cgraph.h (cgraph_node::get_uid): New method.
7270 * ipa-inline.c (update_caller_keys): Use
7271 cgraph_node::get_uid ().
7272 (update_callee_keys): Likewise.
7273 * ipa-utils.c (searchc): Likewise.
7274 (ipa_reduced_postorder): Likewise.
7275 * lto-cgraph.c (input_node): Likewise.
7276 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7277 * symbol-summary.h (symtab_insertion): Likewise.
7278 (symtab_removal): Likewise.
7279 (symtab_duplication): Likewise.
7280 * tree-pretty-print.c (dump_function_header): Likewise.
7281 * tree-sra.c (convert_callers_for_node): Likewise.
7283 2018-06-08 Martin Liska <mliska@suse.cz>
7285 * cgraph.c (symbol_table::create_edge): Always assign a new
7287 (symbol_table::free_edge): Do not recycle numbers.
7288 * cgraph.h (cgraph_edge::get): New method.
7289 * symbol-summary.h (symtab_removal): Use it.
7290 (symtab_duplication): Likewise.
7291 (call_summary::hashable_uid): Remove.
7293 2018-06-08 Martin Liska <mliska@suse.cz>
7295 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7296 (initialize_growth_caches): Remove.
7297 (free_growth_caches): Likewise.
7298 (do_estimate_edge_time): Use edge_growth_cache.
7299 (do_estimate_edge_size): Likewise.
7300 (do_estimate_edge_hints): Likewise.
7301 * ipa-inline.c (reset_edge_caches): Likewise.
7302 (recursive_inlining): Likewise.
7303 (inline_small_functions): Likewise.
7304 * ipa-inline.h (initialize_growth_caches): Remove.
7305 (estimate_edge_size): Likewise.
7306 (estimate_edge_time): Likewise.
7307 (estimate_edge_hints): Likewise.
7308 (reset_edge_growth_cache): Likewise.
7309 * symbol-summary.h (call_summary::remove): New method.
7311 2018-06-08 Martin Liska <mliska@suse.cz>
7313 * ipa-cp.c (class edge_clone_summary): New summary.
7314 (grow_edge_clone_vectors): Remove.
7315 (ipcp_edge_duplication_hook): Remove.
7316 (class edge_clone_summary_t): New call_summary class.
7317 (ipcp_edge_removal_hook): Remove.
7318 (edge_clone_summary_t::duplicate): New function.
7319 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7320 (create_specialized_node): Likewise.
7321 (ipcp_driver): Initialize edge_clone_summaries and do not
7324 2018-06-08 Martin Liska <mliska@suse.cz>
7326 * symbol-summary.h (get): New function.
7327 (call_summary::m_initialize_when_cloning): New class member.
7329 2018-06-08 Martin Liska <mliska@suse.cz>
7331 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7332 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7333 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7334 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7335 (remove_cgraph_node_from_order): Use the removed_nodes set.
7336 (do_per_function_toporder): Likwise.
7337 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7338 instead of summary_uid.
7339 (symtab_removal): Likewise.
7340 (symtab_duplication): Likewise.
7342 2018-06-08 Martin Liska <mliska@suse.cz>
7344 * ipa-cp.c (ipcp_store_bits_results): Use
7345 ipcp_transformation_sum.
7346 (ipcp_store_vr_results): Likewise.
7347 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7349 (ipcp_transformation_initialize): ... this.
7350 (ipa_set_node_agg_value_chain):
7351 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7352 (write_ipcp_transformation_info): Likewise.
7353 (read_ipcp_transformation_info): Likewise.
7354 (ipcp_update_bits): Likewise.
7355 (ipcp_update_vr): Likewise.
7356 (ipcp_transform_function): Likewise.
7357 * ipa-prop.h: Rename ipcp_transformation_summary to
7358 ipcp_transformation.
7359 (class ipcp_transformation_t): New function summary.
7360 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7361 (ipa_get_agg_replacements_for_node): Likewise.
7363 2018-06-08 Martin Liska <mliska@suse.cz>
7365 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7367 (class funct_state_summary_t): New function_summary class.
7368 (has_function_state): Remove.
7369 (get_function_state): Likewise.
7370 (set_function_state): Likewise.
7371 (add_new_function): Likewise.
7372 (funct_state_summary_t::insert): New function.
7373 (duplicate_node_data): Remove.
7374 (remove_node_data): Remove.
7375 (funct_state_summary_t::duplicate): New function.
7376 (register_hooks): Create new funct_state_summaries.
7377 (pure_const_generate_summary): Use it.
7378 (pure_const_write_summary): Likewise.
7379 (pure_const_read_summary): Likewise.
7380 (propagate_pure_const): Likewise.
7381 (propagate_nothrow): Likewise.
7382 (dump_malloc_lattice): Likewise.
7383 (propagate_malloc): Likewise.
7384 (execute): Do not register hooks, just remove summary
7386 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7389 2018-06-08 Martin Liska <mliska@suse.cz>
7391 * ipa-reference.c (remove_node_data): Remove.
7392 (duplicate_node_data): Likewise.
7393 (class ipa_ref_var_info_summary_t): New class.
7394 (class ipa_ref_opt_summary_t): Likewise.
7395 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7396 (get_reference_optimization_summary): Use
7397 ipa_ref_opt_sum_summaries.
7398 (set_reference_vars_info): Remove.
7399 (set_reference_optimization_summary): Likewise.
7400 (ipa_init): Create summaries.
7401 (init_function_info): Use function summary.
7402 (ipa_ref_opt_summary_t::duplicate): New function.
7403 (ipa_ref_opt_summary_t::remove): New function.
7404 (get_read_write_all_from_node): Fix GNU coding style.
7405 (propagate): Use function summary.
7406 (write_node_summary_p): Fix GNU coding style.
7407 (stream_out_bitmap): Likewise.
7408 (ipa_reference_read_optimization_summary): Use function summary.
7409 (ipa_reference_c_finalize): Do not release hooks.
7411 2018-06-08 Martin Liska <mliska@suse.cz>
7413 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7414 (analyze_function_body): Extract multiple calls of get_create.
7415 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7416 * ipa-inline.c (recursive_inlining): Use ::get method.
7417 * ipa-inline.h (estimate_edge_growth): Likewise.
7419 2018-06-08 Martin Liska <mliska@suse.cz>
7421 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7423 (hsa_function_summary::hsa_function_summary): Use the new enum
7425 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7426 * hsa-gen.c (hsa_get_host_function): Likewise.
7427 (get_brig_function_name): Likewise.
7428 * ipa-hsa.c (process_hsa_functions): Likewise.
7429 (ipa_hsa_write_summary): Likewise.
7430 * symbol-summary.h (symtab_duplication): Use ::get function/
7431 (get): New function.
7433 2018-06-08 Martin Liska <mliska@suse.cz>
7435 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7437 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7438 (hsa_register_kernel): Likewise.
7439 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7440 * hsa-gen.c (hsa_get_host_function): Likewise.
7441 (get_brig_function_name): Likewise.
7442 (generate_hsa): Likewise.
7443 (pass_gen_hsail::execute): Likewise.
7444 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7445 (devirtualization_time_bonus): Likewise.
7446 (ipcp_propagate_stage): Likewise.
7447 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7448 (edge_set_predicate): Likewise.
7449 (evaluate_conditions_for_known_args): Likewise.
7450 (evaluate_properties_for_edge): Likewise.
7451 (ipa_fn_summary::reset): Likewise.
7452 (ipa_fn_summary_t::duplicate): Likewise.
7453 (dump_ipa_call_summary): Likewise.
7454 (ipa_dump_fn_summary): Likewise.
7455 (analyze_function_body): Likewise.
7456 (compute_fn_summary): Likewise.
7457 (estimate_edge_devirt_benefit): Likewise.
7458 (estimate_edge_size_and_time): Likewise.
7459 (estimate_calls_size_and_time): Likewise.
7460 (estimate_node_size_and_time): Likewise.
7461 (inline_update_callee_summaries): Likewise.
7462 (remap_edge_change_prob): Likewise.
7463 (remap_edge_summaries): Likewise.
7464 (ipa_merge_fn_summary_after_inlining): Likewise.
7465 (ipa_update_overall_fn_summary): Likewise.
7466 (read_ipa_call_summary): Likewise.
7467 (inline_read_section): Likewise.
7468 (write_ipa_call_summary): Likewise.
7469 (ipa_fn_summary_write): Likewise.
7470 (ipa_free_fn_summary): Likewise.
7471 * ipa-hsa.c (process_hsa_functions): Likewise.
7472 (ipa_hsa_write_summary): Likewise.
7473 (ipa_hsa_read_section): Likewise.
7474 * ipa-icf.c (sem_function::merge): Likewise.
7475 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7476 (do_estimate_edge_time): Likewise.
7477 (estimate_size_after_inlining): Likewise.
7478 (estimate_growth): Likewise.
7479 (growth_likely_positive): Likewise.
7480 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7481 (inline_call): Likewise.
7482 * ipa-inline.c (caller_growth_limits): Likewise.
7483 (can_inline_edge_p): Likewise.
7484 (can_inline_edge_by_limits_p): Likewise.
7485 (compute_uninlined_call_time): Likewise.
7486 (compute_inlined_call_time): Likewise.
7487 (want_inline_small_function_p): Likewise.
7488 (edge_badness): Likewise.
7489 (update_caller_keys): Likewise.
7490 (update_callee_keys): Likewise.
7491 (recursive_inlining): Likewise.
7492 (inline_small_functions): Likewise.
7493 (inline_to_all_callers_1): Likewise.
7494 (dump_overall_stats): Likewise.
7495 (early_inline_small_functions): Likewise.
7496 (early_inliner): Likewise.
7497 * ipa-inline.h (estimate_edge_growth): Likewise.
7498 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7499 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7500 * ipa-prop.h (IPA_NODE_REF): Likewise.
7501 (IPA_EDGE_REF): Likewise.
7502 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7503 (propagate_malloc): Likewise.
7504 * ipa-split.c (execute_split_functions): Likewise.
7505 * symbol-summary.h: Rename get to get_create.
7507 (get_create): Likewise.
7508 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7510 2018-06-08 Martin Liska <mliska@suse.cz>
7512 * symbol-summary.h (release): Move definition out of class
7514 (symtab_removal): Likewise.
7515 (symtab_duplication): Likewise.
7517 2018-06-08 Martin Liska <mliska@suse.cz>
7519 * symbol-summary.h (function_summary): Move constructor
7520 implementation out of class declaration.
7521 (release): Likewise.
7522 (symtab_insertion): Likewise.
7523 (symtab_removal): Likewise.
7524 (symtab_duplication): Likewise.
7527 2018-06-08 Martin Liska <mliska@suse.cz>
7529 * Makefile.in: Remove support for MPX (macros, related functions,
7530 fields in cgraph_node, ...).
7531 * builtin-types.def (BT_BND): Likewise.
7532 (BT_FN_BND_CONST_PTR): Likewise.
7533 (BT_FN_CONST_PTR_BND): Likewise.
7534 (BT_FN_VOID_PTR_BND): Likewise.
7535 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7536 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7537 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7538 (expand_builtin_mempcpy_with_bounds): Likewise.
7539 (expand_builtin_memset_with_bounds): Likewise.
7540 (expand_builtin_memset_args): Likewise.
7541 (std_expand_builtin_va_start): Likewise.
7542 (expand_builtin): Likewise.
7543 (expand_builtin_with_bounds): Likewise.
7544 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7545 (DEF_LIB_BUILTIN_CHKP): Likewise.
7546 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7547 (DEF_CHKP_BUILTIN): Likewise.
7548 (BUILT_IN_MEMCPY): Likewise.
7549 (BUILT_IN_MEMMOVE): Likewise.
7550 (BUILT_IN_MEMPCPY): Likewise.
7551 (BUILT_IN_MEMSET): Likewise.
7552 (BUILT_IN_STPCPY): Likewise.
7553 (BUILT_IN_STRCAT): Likewise.
7554 (BUILT_IN_STRCHR): Likewise.
7555 (BUILT_IN_STRCPY): Likewise.
7556 (BUILT_IN_STRLEN): Likewise.
7557 (BUILT_IN_MEMCPY_CHK): Likewise.
7558 (BUILT_IN_MEMMOVE_CHK): Likewise.
7559 (BUILT_IN_MEMPCPY_CHK): Likewise.
7560 (BUILT_IN_MEMSET_CHK): Likewise.
7561 (BUILT_IN_STPCPY_CHK): Likewise.
7562 (BUILT_IN_STRCAT_CHK): Likewise.
7563 (BUILT_IN_STRCPY_CHK): Likewise.
7564 * calls.c (store_bounds): Likewise.
7565 (emit_call_1): Likewise.
7566 (special_function_p): Likewise.
7567 (maybe_warn_nonstring_arg): Likewise.
7568 (initialize_argument_information): Likewise.
7569 (finalize_must_preallocate): Likewise.
7570 (compute_argument_addresses): Likewise.
7571 (expand_call): Likewise.
7572 * cfgexpand.c (expand_call_stmt): Likewise.
7573 (expand_return): Likewise.
7574 (expand_gimple_stmt_1): Likewise.
7575 (pass_expand::execute): Likewise.
7576 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7577 (cgraph_node::remove): Likewise.
7578 (cgraph_node::dump): Likewise.
7579 (cgraph_node::verify_node): Likewise.
7580 * cgraph.h (chkp_function_instrumented_p): Likewise.
7581 (symtab_node::get_alias_target): Likewise.
7582 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7583 (cgraph_local_p): Likewise.
7584 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7585 (cgraph_edge::rebuild_references): Likewise.
7586 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7587 (walk_polymorphic_call_targets): Likewise.
7588 (cgraph_node::expand_thunk): Likewise.
7589 (symbol_table::output_weakrefs): Likewise.
7590 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7591 (ix86_handle_option): Likewise.
7592 * config/i386/constraints.md: Likewise.
7593 * config/i386/i386-builtin-types.def (BND): Likewise.
7597 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7598 (BDESC_FIRST): Likewise.
7600 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7601 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7602 * config/i386/i386.c (enum reg_class): Likewise.
7603 (ix86_target_string): Likewise.
7604 (ix86_option_override_internal): Likewise.
7605 (ix86_conditional_register_usage): Likewise.
7606 (ix86_valid_target_attribute_inner_p): Likewise.
7607 (ix86_set_indirect_branch_type): Likewise.
7608 (ix86_set_current_function): Likewise.
7609 (ix86_function_arg_regno_p): Likewise.
7610 (init_cumulative_args): Likewise.
7611 (ix86_function_arg_advance): Likewise.
7612 (ix86_function_arg): Likewise.
7613 (ix86_pass_by_reference): Likewise.
7614 (ix86_function_value_regno_p): Likewise.
7615 (ix86_function_value_1): Likewise.
7616 (ix86_function_value_bounds): Likewise.
7617 (ix86_return_in_memory): Likewise.
7618 (ix86_setup_incoming_vararg_bounds): Likewise.
7619 (ix86_va_start): Likewise.
7620 (indirect_thunk_need_prefix): Likewise.
7621 (print_reg): Likewise.
7622 (ix86_print_operand): Likewise.
7623 (ix86_expand_call): Likewise.
7624 (ix86_output_function_return): Likewise.
7625 (reg_encoded_number): Likewise.
7626 (BDESC_VERIFYS): Likewise.
7627 (ix86_init_mpx_builtins): Likewise.
7628 (ix86_init_builtins): Likewise.
7629 (ix86_emit_cmove): Likewise.
7630 (ix86_emit_move_max): Likewise.
7631 (ix86_expand_builtin): Likewise.
7632 (ix86_builtin_mpx_function): Likewise.
7633 (ix86_get_arg_address_for_bt): Likewise.
7634 (ix86_load_bounds): Likewise.
7635 (ix86_store_bounds): Likewise.
7636 (ix86_load_returned_bounds): Likewise.
7637 (ix86_store_returned_bounds): Likewise.
7638 (ix86_class_likely_spilled_p): Likewise.
7639 (ix86_hard_regno_mode_ok): Likewise.
7640 (x86_order_regs_for_local_alloc): Likewise.
7641 (ix86_mitigate_rop): Likewise.
7642 (ix86_bnd_prefixed_insn_p): Likewise.
7643 (ix86_mpx_bound_mode): Likewise.
7644 (ix86_make_bounds_constant): Likewise.
7645 (ix86_initialize_bounds): Likewise.
7646 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7647 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7648 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7649 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7650 (TARGET_CHKP_BOUND_MODE): Likewise.
7651 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7652 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7653 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7654 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7655 * config/i386/i386.h (TARGET_MPX): Likewise.
7656 (TARGET_MPX_P): Likewise.
7657 (VALID_BND_REG_MODE): Likewise.
7658 (FIRST_BND_REG): Likewise.
7659 (LAST_BND_REG): Likewise.
7660 (enum reg_class): Likewise.
7661 (BND_REG_P): Likewise.
7662 (BND_REGNO_P): Likewise.
7663 (BNDmode): Likewise.
7664 (ADJUST_INSN_LENGTH): Likewise.
7665 * config/i386/i386.md: Likewise.
7666 * config/i386/i386.opt: Likewise.
7667 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7668 (defined): Likewise.
7669 (LINK_MPX): Likewise.
7670 (MPX_SPEC): Likewise.
7671 (LIBMPX_SPEC): Likewise.
7672 (LIBMPXWRAPPERS_SPEC): Likewise.
7673 (CHKP_SPEC): Likewise.
7674 * config/i386/predicates.md: Likewise.
7675 * dbxout.c (dbxout_type): Likewise.
7676 * doc/extend.texi: Likewise.
7677 * doc/invoke.texi: Likewise.
7678 * doc/md.texi: Likewise.
7679 * doc/tm.texi: Likewise.
7680 * doc/tm.texi.in: Likewise.
7681 * dwarf2out.c (is_base_type): Likewise.
7682 (gen_formal_types_die): Likewise.
7683 (gen_subprogram_die): Likewise.
7684 (gen_type_die_with_usage): Likewise.
7685 (gen_decl_die): Likewise.
7686 (dwarf2out_late_global_decl): Likewise.
7687 * expr.c (expand_assignment): Likewise.
7688 (emit_storent_insn): Likewise.
7689 (store_expr_with_bounds): Likewise.
7690 (store_expr): Likewise.
7691 (expand_expr_real_1): Likewise.
7692 * expr.h (store_expr_with_bounds): Likewise.
7693 * function.c (use_register_for_decl): Likewise.
7694 (struct bounds_parm_data): Likewise.
7695 (assign_parms_augmented_arg_list): Likewise.
7696 (assign_parm_find_entry_rtl): Likewise.
7697 (assign_parm_is_stack_parm): Likewise.
7698 (assign_parm_load_bounds): Likewise.
7699 (assign_bounds): Likewise.
7700 (assign_parms): Likewise.
7701 (expand_function_start): Likewise.
7702 * gcc.c (CHKP_SPEC): Likewise.
7703 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7704 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7705 (wrestrict_dom_walker::check_call): Likewise.
7706 * gimple.c (gimple_build_call_from_tree): Likewise.
7707 * gimple.h (enum gf_mask): Likewise.
7708 (gimple_call_with_bounds_p): Likewise.
7709 (gimple_call_set_with_bounds): Likewise.
7710 * gimplify.c (gimplify_init_constructor): Likewise.
7711 * ipa-cp.c (initialize_node_lattices): Likewise.
7712 (propagate_constants_across_call): Likewise.
7713 (find_more_scalar_values_for_callers_subset): Likewise.
7714 * ipa-hsa.c (process_hsa_functions): Likewise.
7715 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7716 * ipa-icf.c (sem_function::merge): Likewise.
7717 * ipa-inline.c (early_inliner): Likewise.
7718 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7719 (warn_function_cold): Likewise.
7720 (propagate_pure_const): Likewise.
7721 * ipa-ref.h (enum GTY): Likewise.
7722 * ipa-split.c (find_retbnd): Likewise.
7723 (consider_split): Likewise.
7724 (split_function): Likewise.
7725 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7726 * ipa.c (walk_polymorphic_call_targets): Likewise.
7727 (symbol_table::remove_unreachable_nodes): Likewise.
7728 (process_references): Likewise.
7729 (cgraph_build_static_cdtor_1): Likewise.
7730 * lto-cgraph.c (lto_output_node): Likewise.
7731 (output_refs): Likewise.
7732 (compute_ltrans_boundary): Likewise.
7733 (input_overwrite_node): Likewise.
7734 (input_node): Likewise.
7735 (input_cgraph_1): Likewise.
7736 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7737 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7738 (class pass_chkp_instrumentation_passes): Likewise.
7739 (make_pass_chkp_instrumentation_passes): Likewise.
7740 * passes.def: Likewise.
7741 * rtl.h (struct GTY): Likewise.
7742 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7743 * stor-layout.c (layout_type): Likewise.
7744 * symtab.c: Likewise.
7745 * target.def: Likewise.
7746 * targhooks.c (default_chkp_bound_type): Likewise.
7747 (default_chkp_bound_mode): Likewise.
7748 (default_builtin_chkp_function): Likewise.
7749 (default_chkp_function_value_bounds): Likewise.
7750 (default_chkp_make_bounds_constant): Likewise.
7751 (default_chkp_initialize_bounds): Likewise.
7752 * targhooks.h (default_chkp_bound_type): Likewise.
7753 (default_chkp_bound_mode): Likewise.
7754 (default_builtin_chkp_function): Likewise.
7755 (default_chkp_function_value_bounds): Likewise.
7756 (default_chkp_make_bounds_constant): Likewise.
7757 (default_chkp_initialize_bounds): Likewise.
7758 * toplev.c (compile_file): Likewise.
7759 (process_options): Likewise.
7760 * tree-core.h (DEF_BUILTIN): Likewise.
7761 (DEF_BUILTIN_CHKP): Likewise.
7762 * tree-inline.c (declare_return_variable): Likewise.
7763 (remap_gimple_stmt): Likewise.
7764 (copy_bb): Likewise.
7765 (initialize_inlined_parameters): Likewise.
7766 (expand_call_inline): Likewise.
7767 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7768 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7769 (make_pass_ipa_chkp_produce_thunks): Likewise.
7770 (make_pass_chkp): Likewise.
7771 (make_pass_chkp_opt): Likewise.
7772 (make_pass_chkp_instrumentation_passes): Likewise.
7773 * tree-pretty-print.c (dump_generic_node): Likewise.
7774 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7775 * tree-ssa-dce.c (propagate_necessity): Likewise.
7776 (eliminate_unnecessary_stmts): Likewise.
7777 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7778 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7779 * tree-ssa-sccvn.h: Likewise.
7780 * tree-ssa-strlen.c (get_string_length): Likewise.
7781 (valid_builtin_call): Likewise.
7782 (adjust_last_stmt): Likewise.
7783 (handle_builtin_strchr): Likewise.
7784 (handle_builtin_strcpy): Likewise.
7785 (handle_builtin_stxncpy): Likewise.
7786 (handle_builtin_memcpy): Likewise.
7787 (handle_builtin_strcat): Likewise.
7788 (strlen_check_and_optimize_stmt): Likewise.
7789 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7790 * tree-streamer-in.c: Likewise.
7791 * tree-streamer.c (record_common_node): Likewise.
7792 * tree.c (tree_code_size): Likewise.
7793 (wide_int_to_tree_1): Likewise.
7794 (type_contains_placeholder_1): Likewise.
7795 (build_common_tree_nodes): Likewise.
7796 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7797 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7798 (POINTER_BOUNDS_P): Likewise.
7799 (BOUNDED_TYPE_P): Likewise.
7800 (BOUNDED_P): Likewise.
7801 (CALL_WITH_BOUNDS_P): Likewise.
7802 (pointer_bounds_type_node): Likewise.
7803 * value-prof.c (gimple_ic): Likewise.
7804 * var-tracking.c (vt_add_function_parameters): Likewise.
7805 * varasm.c (make_decl_rtl): Likewise.
7806 (assemble_start_function): Likewise.
7807 (output_constant): Likewise.
7808 (maybe_assemble_visibility): Likewise.
7809 * varpool.c (ctor_for_folding): Likewise.
7810 * chkp-builtins.def: Remove.
7811 * ipa-chkp.c: Remove.
7812 * ipa-chkp.h: Remove.
7813 * rtl-chkp.c: Remove.
7814 * rtl-chkp.h: Remove.
7815 * tree-chkp-opt.c: Remove.
7816 * tree-chkp.c: Remove.
7817 * tree-chkp.h: Remove.
7819 2018-06-07 Carl Love <cel@us.ibm.com>
7821 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7822 vextract_fp_from_shortl): Add BE support.
7824 2018-06-07 Paul Koning <ni1d@arrl.net>
7826 * compare-elim.c (try_merge_compare): Don't merge compare if
7827 address contains a side effect.
7828 (try_eliminate_compare): Likewise.
7830 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7832 * config.gcc: Support "tremont".
7833 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7834 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7836 * config/i386/i386.c (m_TREMONT): Define.
7837 (processor_target_table): Add "tremont".
7838 (PTA_TREMONT): Define.
7839 (ix86_lea_outperforms): Add TARGET_TREMONT.
7840 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7841 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7842 and M_INTEL_GOLDMONT_PLUS.
7843 (fold_builtin_cpu): Add "tremont".
7844 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7845 (ix86_option_override_internal): Add "tremont".
7846 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7847 (processor_type): Add PROCESSOR_TREMONT.
7848 * config/i386/x86-tune.def: Add m_TREMONT.
7849 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7851 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7853 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7854 symbol defined for msp430i* devices to be lower case.
7856 2018-06-07 Richard Biener <rguenther@suse.de>
7858 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7859 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
7860 Properly wrap signed arithmetic if overflow wraps.
7862 2018-06-07 Jakub Jelinek <jakub@redhat.com>
7864 PR tree-optimization/69615
7865 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7866 of a cast from a same precision integral SSA_NAME in a bb dominated
7867 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7868 cast to utype if rhs2 has already a compatible type.
7870 2018-06-07 Richard Biener <rguenther@suse.de>
7872 PR tree-optimization/85935
7873 * graphite-scop-detection.c (find_params_in_bb): Analyze
7874 condition operands with respect to the correct loop. Assert
7875 the analysis doesn't fail.
7877 2018-06-04 Carl Love <cel@us.ibm.com>
7879 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7880 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
7881 as it is slightly cheaper.
7882 (first_match_or_eos_index_<mode>):
7883 Calculate index using natural element order.
7884 (first_match_index_<mode>):
7885 Calculate index using natural element order.
7886 (first_match_or_eos_index_<mode>):
7887 Calculate index using natural order.
7888 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7889 for BE and LE modes.
7890 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7891 P9V_BUILTIN_VCLZLSBB_V16QI.
7892 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7895 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7897 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7898 indentation and line wrap for many prototypes. Add missing
7899 @smallexample directives around block of prototypes for vec_xl and
7902 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
7904 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7905 track if we pass or return IEEE 128-bit floating point.
7906 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7907 C++ mangling that is compatible with GCC 8.1.
7908 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7909 (init_cumulative_args): Note if we pass or return IEEE 128-bit
7910 floating point types.
7911 (rs6000_function_arg_advance_1): Likewise.
7912 (rs6000_mangle_type): Optionally generate mangled names that match
7913 what GCC 8.1 generated for IEEE 128-bit floating point types.
7914 (rs6000_globalize_decl_name): If we have an external function that
7915 passes or returns IEEE 128-bit types, generate a weak reference
7916 from the mangled name used in GCC 8.1 to the current mangled
7918 (rs6000_init_builtins): Make __ibm128 use the long double type if
7919 long double is IBM extended double. Make __float128 use the long
7920 double type if long double is IEEE 128-bit.
7923 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7924 macro for __ibm128 built-in functions.
7925 (PACK_IF): Add __ibm128 pack/unpack functions.
7926 (UNPACK_IF): Likewise.
7927 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7928 enable long double built-in functions if long double is IEEE
7929 128-bit floating point.
7930 (rs6000_invalid_builtin): Update long double built-in function
7932 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7933 functions, adjust the built-in function to use the long double
7934 built-in function if __ibm128 and long double are the same type.
7935 * doc/extend.texi (PowerPC builtins): Update documention for
7936 __builtin_{,un}pack_longdouble. Add documentation for
7937 __builtin_{,un}pack_ibm128.
7939 2018-06-06 Jim Wilson <jimw@sifive.com>
7941 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7942 (struct machine_function): New field interrupt_mode.
7943 (riscv_handle_type_attribute): New function. Add forward declaration.
7944 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7945 (riscv_expand_epilogue): Check interrupt_mode field.
7946 (riscv_set_current_function): Check interrupt attribute args and
7947 set interrupt_mode field.
7948 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7949 (riscv_sret, riscv_uret): New.
7950 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7951 new arguments to interrupt attribute.
7953 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
7956 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7957 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7959 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7961 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7962 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7963 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7964 several redundant entries.
7966 2018-06-06 David Malcolm <dmalcolm@redhat.com>
7968 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
7969 type from "rtx" to "rtx_insn *".
7970 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
7971 for local "call_insn", removing cast.
7972 (ix86_expand_call): Likewise, introducing a "call_insn" local.
7974 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
7976 PR tree-optimization/86066
7977 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
7978 for BIT_INSERT_EXPR stores.
7980 2018-06-06 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/86062
7983 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7984 component refs ontop
7985 of to be offsetted base.
7987 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7989 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7990 to be static and remove check on interrupt attribute name.
7992 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
7994 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7995 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7997 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8000 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8002 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8003 Remove second argument, change how error is called.
8004 (aarch64_layout_arg): Remove second argument from
8005 aarch64_err_no_fpadvsimd call.
8006 (aarch64_init_cumulative_args): Ditto.
8007 (aarch64_gimplify_va_arg_expr): Ditto.
8008 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8010 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8012 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8013 (*simple_return_indirect_internal<mode>): Rename from
8014 simple_return_indirect_internal. Use W mode iterator.
8015 (rstorssp): New expander.
8016 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8017 (clrssbsy): New expander.
8018 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8020 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8022 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8024 (cmse_check_pointed_object): Likewise.
8026 2018-06-05 Martin Liska <mliska@suse.cz>
8028 PR gcov-profile/47618
8029 * doc/invoke.texi: Document how -fprofile-dir format
8032 2018-06-05 Richard Biener <rguenther@suse.de>
8034 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8035 removal pretend DOM info isn't available so we do not update
8036 it and only remove edges, not dominated blocks. Actually free
8037 DOM info in case we removed something. Remove unreachable blocks.
8038 (mfb_keep_latches): Work with either DOM info or marked backedges.
8039 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8040 first. Mark backedges if DOM info isn't available.
8041 (Re-)compute DOM info after cleanup_control_flow_pre.
8043 2018-06-05 Richard Biener <rguenther@suse.de>
8045 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8046 (locus_discrim_hasher::hash): Adjust.
8047 (locus_discrim_hasher::equal): Likewise.
8048 (next_discriminator_for_locus): Work on line directly.
8049 (same_line_p): Pass in expanded locus1 as well.
8050 (assign_discriminators): Avoid redundant location expansions.
8052 2018-06-05 Richard Biener <rguenther@suse.de>
8054 PR tree-optimization/86046
8055 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8056 if required after clearing TREE_ADDRESSABLE.
8058 2018-06-05 Richard Biener <rguenther@suse.de>
8060 PR tree-optimization/86047
8061 * tree-ssa-loop.c (for_each_index): Glob handling of all
8062 decls and constants and really handle all of them.
8064 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8067 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8068 qualifier_void_pointer and qualifier_const_void_pointer.
8069 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8070 (arm_init_builtins): Handle the above.
8071 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8072 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8073 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8076 2018-06-05 Martin Liska <mliska@suse.cz>
8078 * auto-profile.c (read_autofdo_file): Do not use
8079 gcov_ctr_summary struct.
8080 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8081 * coverage.c (struct counts_entry): Likewise.
8082 (read_counts_file): Read just single summary entry.
8083 (get_coverage_counts): Use gcov_summary struct.
8084 * coverage.h (get_coverage_counts): Likewise.
8085 * gcov-dump.c (dump_working_sets): Likewise.
8086 (tag_summary): Dump just single summary.
8087 * gcov-io.c (gcov_write_summary): Write just histogram
8089 (gcov_read_summary): Read just single summary.
8090 (compute_working_sets): Use gcov_summary struct.
8091 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8092 of GCOV_COUNTERS_SUMMABLE.
8093 (GCOV_COUNTERS_SUMMABLE): Remove.
8094 (GCOV_FIRST_VALUE_COUNTER): Replace with
8095 GCOV_COUNTER_V_INTERVAL.
8096 (struct gcov_ctr_summary): Remove.
8097 (struct gcov_summary): Directly use fields of former
8099 (compute_working_sets): Use gcov_summary struct.
8100 * gcov.c (read_count_file): Do not use ctrs fields.
8101 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8103 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8105 * profile.c: Likewise.
8106 * profile.h: Likewise.
8108 2018-06-05 Martin Liska <mliska@suse.cz>
8110 PR gcov-profile/84846
8111 * gcov.c (output_lines): Print working directory only
8112 in intermediate format.
8114 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8116 * config/s390/s390-builtin-types.def: Add void function type.
8117 * config/s390/s390-builtins.def: Use the function type for the
8120 2018-06-04 Jim Wilson <jimw@sifive.com>
8122 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8124 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8125 and maybe_eh_return. Change regno to unsigned int. Use new args to
8126 handle EH_RETURN_DATA_REGNO registers properly.
8127 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8128 (riscv_expand_epilogue): Update comment. Change argument name and
8129 type. Update code to use new name and type. Pass new args to
8130 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8132 * config/riscv/riscv.md (NORMAL_RETURN): New.
8133 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8134 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8135 (eh_return): Call gen_eh_return_internal and emit barrier.
8136 (eh_return_internal): Call riscv_expand_epilogue.
8138 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8140 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8141 bit_insertion field and declare can_be_merged_into method.
8142 (merged_store_group::can_be_merged_into): New method.
8143 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8144 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8145 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8147 2018-06-04 Richard Biener <rguenther@suse.de>
8149 PR tree-optimization/85955
8150 * builtins.c (fold_builtin_sincos): Convert pointers to
8151 destination to appropriate type before dereferencing.
8153 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8155 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8157 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8159 * expr.c (expand_expr_real_1): Force the operand into memory if
8160 its TYPE_MODE is BLKmode and if there is no integer mode for
8161 the number of bits being extracted.
8163 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8167 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8168 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8170 2018-06-04 Richard Biener <rguenther@suse.de>
8172 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8173 (cleanup_tree_cfg_noloop): ... single caller. Do
8174 start_recording_case_labels later.
8176 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8178 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8179 to _IMMINTRIN_H_INCLUDED.
8180 * config/i386/pconfigintrin.h: Ditto.
8181 * config/i386/waitpkgintrin.h: Ditto.
8182 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8183 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8184 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8185 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8186 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8187 waitpkgintrin.h and cldemoteintrin.h.
8189 2018-06-04 Richard Biener <rguenther@suse.de>
8191 PR tree-optimization/86038
8192 * tracer.c (find_best_successor): Check probability for
8193 being initialized, bail out if not.
8195 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8198 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8199 of bits to ignore when comparing architectures.
8201 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8203 PR tree-optimization/69615
8204 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8205 maximum or minimum of the type, try to merge it also as if
8206 range1 is + [-, x - 1] or + [x + 1, -].
8209 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8211 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8213 PR tree-optimization/86034
8214 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8215 the unsigned bitfield type in a bit insertion sequence if it does not
8216 have a larger precision than the bitfield size.
8217 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8219 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8221 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8223 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8225 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8226 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8227 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8228 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8230 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8232 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8233 Disable -fdelete-null-pointer-checks for ELF toolchain.
8235 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8236 Kito Cheng <kito.cheng@gmail.com>
8238 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8239 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8240 (nds32*-*-*): Add glibc and uclibc conditions.
8241 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8242 (TARGET_EXCEPT_UNWIND_INFO): Define.
8243 * config/nds32/elf.h: New file.
8244 * config/nds32/linux.h: New file.
8245 * config/nds32/nds32-elf.opt: New file.
8246 * config/nds32/nds32-linux.opt: New file.
8247 * config/nds32/nds32-fp-as-gp.c
8248 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8249 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8251 (nds32_asm_file_end): Ditto.
8252 (nds32_print_operand): Ditto.
8253 (nds32_insert_attributes): Ditto.
8254 (nds32_init_libfuncs): New function.
8255 (TARGET_HAVE_TLS): Define.
8256 (TARGET_INIT_LIBFUNCS): Define.
8257 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8259 (TARGET_ELF): Apply different mcmodel setting.
8260 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8261 been migrated into elf.h and linux.h files.
8262 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8263 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8264 (mcmodel): The content has been migrated into nds32-elf.opt and
8265 nds32-linux.opt files.
8266 * config/nds32/t-elf: New file.
8267 * config/nds32/t-linux: New file.
8269 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8270 Shiva Chen <shiva0217@gmail.com>
8272 * config/nds32/constants.md (unspec_volatile_element): Add
8273 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8274 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8276 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8277 (make_pass_nds32_fp_as_gp): Declare.
8278 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8280 (nds32_asm_function_end_prologue): Remove unused asm output.
8281 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8282 (nds32_asm_file_start): Output necessary fp_as_gp information.
8283 (nds32_option_override): Adjust register usage.
8284 (nds32_expand_prologue): Consider fp_as_gp situation.
8285 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8286 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8289 (simple_return): Ditto.
8290 (omit_fp_begin): Output special directive for fp_as_gp.
8291 (omit_fp_end): Output special directive for fp_as_gp.
8292 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8293 mforbid-fp-as-gp): New options.
8295 2018-06-01 Mark Wielaard <mark@klomp.org>
8297 * dwarf2out.c (dwarf2out_finish): Remove generation of
8298 DW_AT_loclists_base.
8300 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8302 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8303 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8304 (struct merged_store_group): Add bit_insertion field.
8305 (dump_char_array): Use standard hexadecimal format.
8306 (merged_store_group::merged_store_group): Set bit_insertion to false.
8307 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8308 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8309 also print the mask in the dump file.
8310 (pass_store_merging::gate): Minor tweak.
8311 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8312 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8313 stores with INTEGER_CST stores.
8314 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8315 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8316 and use it throughout. Generate bit insertion sequences if need be.
8317 (pass_store_merging::process_store): Remove redundant condition.
8318 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8320 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8322 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8323 the 128-bit floating point types. Fix function comment.
8325 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8327 * config/aarch64/aarch64-simd.md
8328 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8330 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8333 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8335 PR tree-optimization/85989
8336 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8338 (backprop::intersect_uses): Check it when deciding whether this
8339 is a backedge reference.
8340 (backprop::process_block): Add each phi to m_visited_phis
8341 after visiting it, then clear it at the end.
8343 2018-06-01 Richard Biener <rguenther@suse.de>
8345 * tree-vectorizer.h (vect_dr_stmt): New function.
8346 (vect_get_load_cost): Adjust.
8347 (vect_get_store_cost): Likewise.
8348 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8349 Use vect_dr_stmt instead of DR_SMTT.
8350 (vect_record_base_alignments): Likewise.
8351 (vect_calculate_target_alignment): Likewise.
8352 (vect_compute_data_ref_alignment): Likewise and make static.
8353 (vect_update_misalignment_for_peel): Likewise.
8354 (vect_verify_datarefs_alignment): Likewise.
8355 (vector_alignment_reachable_p): Likewise.
8356 (vect_get_data_access_cost): Likewise. Pass down
8357 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8358 (vect_get_peeling_costs_all_drs): Likewise.
8359 (vect_peeling_hash_get_lowest_cost): Likewise.
8360 (vect_enhance_data_refs_alignment): Likewise.
8361 (vect_find_same_alignment_drs): Likewise.
8362 (vect_analyze_data_refs_alignment): Likewise.
8363 (vect_analyze_group_access_1): Likewise.
8364 (vect_analyze_group_access): Likewise.
8365 (vect_analyze_data_ref_access): Likewise.
8366 (vect_analyze_data_ref_accesses): Likewise.
8367 (vect_vfa_segment_size): Likewise.
8368 (vect_small_gap_p): Likewise.
8369 (vectorizable_with_step_bound_p): Likewise.
8370 (vect_prune_runtime_alias_test_list): Likewise.
8371 (vect_analyze_data_refs): Likewise.
8372 (vect_supportable_dr_alignment): Likewise.
8373 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8374 (vect_gen_prolog_loop_niters): Likewise.
8375 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8376 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8378 (vect_recog_mask_conversion_pattern): Likewise.
8379 (vect_try_gather_scatter_pattern): Likewise.
8380 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8381 to vect_get_store_cost.
8382 (vect_get_store_cost): Get stmt_info instead of DR.
8383 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8384 (vect_get_load_cost): Get stmt_info instead of DR.
8386 2018-06-01 Richard Biener <rguenther@suse.de>
8389 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8390 vars wrapped in MEM_REFs.
8392 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8394 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8395 Fix subreg tests so that we only return a choice between
8396 GENERAL_REGS and FP_REGS if the original classes included both.
8398 2018-06-01 Richard Biener <rguenther@suse.de>
8401 * tree-ssa-structalias.c (get_function_part_constraint):
8402 Handle NULL fi->decl.
8403 (find_func_aliases_for_call): Properly handle indirect
8404 fi from direct call.
8405 (find_func_clobbers): Likewise.
8406 (ipa_pta_execute): Likewise.
8407 (create_variable_info_for): For functions that are ifunc_resolver
8408 resolve to a varinfo that contains the result of the resolver call.
8409 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8412 2018-05-31 Michael Collison <michael.collison@arm.com>
8414 * config/aarch64/aarch64.md:
8415 (*fix_to_zero_extenddfdi2): New pattern.
8416 * gcc.target/aarch64/fix_extend1.c: New testcase.
8418 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8422 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8423 and BUILT_IN_STRNCMP_EQ.
8424 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8425 BUILT_IN_STRNCMP_EQ.
8426 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8427 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8428 (gimple_fold_builtin): Likewise.
8429 * tree-ssa-strlen.c (compute_string_length): New function.
8430 (determine_min_obsize): New function.
8431 (handle_builtin_string_cmp): New function to handle calls to
8432 string compare functions.
8433 (strlen_optimize_stmt): Add handling to builtin string compare
8435 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8436 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8437 * tree.c (build_common_builtin_nodes): Add new defines of
8438 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8440 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8443 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8444 functions with naked attribute.
8446 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8448 * config/i386/sse.md (avx_vec_concat<mode>):
8449 Substitute concat_tg_mode mode attribute with xtg_mode.
8450 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8451 (concat_tg_mode): Remove mode attribute.
8453 2018-05-31 Martin Sebor <msebor@redhat.com>
8456 * calls.c (alloc_max_size): Correct a logic error/typo.
8457 Treat excessive arguments as infinite. Warn for invalid arguments.
8458 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8460 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8463 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8464 and movx for Haswell.
8466 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8467 Cesar Philippidis <cesar@codesourcery.com>
8470 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8471 when emitting error on private/firstprivate reductions.
8472 * omp-low.c (lower_omp_target): Avoid reference-type processing
8473 on pointers for firstprivate clause.
8475 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8477 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8480 * config/aarch64/aarch64-simd.md
8481 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8482 (aarch64_ld1_x3_<mode>): Likewise
8483 (aarch64_st1x2<VALLDIF:mode>): Likewise
8484 (aarch64_st1_x2_<mode>): Likewise
8485 (aarch64_st1x3<VALLDIF:mode>): Likewise
8486 (aarch64_st1_x3_<mode>): Likewise
8487 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8488 (vld1_s8_x3): Likewise.
8489 (vld1_u16_x3): Likewise.
8490 (vld1_s16_x3): Likewise.
8491 (vld1_u32_x3): Likewise.
8492 (vld1_s32_x3): Likewise.
8493 (vld1_u64_x3): Likewise.
8494 (vld1_s64_x3): Likewise.
8495 (vld1_f16_x3): Likewise.
8496 (vld1_f32_x3): Likewise.
8497 (vld1_f64_x3): Likewise.
8498 (vld1_p8_x3): Likewise.
8499 (vld1_p16_x3): Likewise.
8500 (vld1_p64_x3): Likewise.
8501 (vld1q_u8_x3): Likewise.
8502 (vld1q_s8_x3): Likewise.
8503 (vld1q_u16_x3): Likewise.
8504 (vld1q_s16_x3): Likewise.
8505 (vld1q_u32_x3): Likewise.
8506 (vld1q_s32_x3): Likewise.
8507 (vld1q_u64_x3): Likewise.
8508 (vld1q_s64_x3): Likewise.
8509 (vld1q_f16_x3): Likewise.
8510 (vld1q_f32_x3): Likewise.
8511 (vld1q_f64_x3): Likewise.
8512 (vld1q_p8_x3): Likewise.
8513 (vld1q_p16_x3): Likewise.
8514 (vld1q_p64_x3): Likewise.
8515 (vst1_s64_x2): Likewise.
8516 (vst1_u64_x2): Likewise.
8517 (vst1_f64_x2): Likewise.
8518 (vst1_s8_x2): Likewise.
8519 (vst1_p8_x2): Likewise.
8520 (vst1_s16_x2): Likewise.
8521 (vst1_p16_x2): Likewise.
8522 (vst1_s32_x2): Likewise.
8523 (vst1_u8_x2): Likewise.
8524 (vst1_u16_x2): Likewise.
8525 (vst1_u32_x2): Likewise.
8526 (vst1_f16_x2): Likewise.
8527 (vst1_f32_x2): Likewise.
8528 (vst1_p64_x2): Likewise.
8529 (vst1q_s8_x2): Likewise.
8530 (vst1q_p8_x2): Likewise.
8531 (vst1q_s16_x2): Likewise.
8532 (vst1q_p16_x2): Likewise.
8533 (vst1q_s32_x2): Likewise.
8534 (vst1q_s64_x2): Likewise.
8535 (vst1q_u8_x2): Likewise.
8536 (vst1q_u16_x2): Likewise.
8537 (vst1q_u32_x2): Likewise.
8538 (vst1q_u64_x2): Likewise.
8539 (vst1q_f16_x2): Likewise.
8540 (vst1q_f32_x2): Likewise.
8541 (vst1q_f64_x2): Likewise.
8542 (vst1q_p64_x2): Likewise.
8543 (vst1_s64_x3): Likewise.
8544 (vst1_u64_x3): Likewise.
8545 (vst1_f64_x3): Likewise.
8546 (vst1_s8_x3): Likewise.
8547 (vst1_p8_x3): Likewise.
8548 (vst1_s16_x3): Likewise.
8549 (vst1_p16_x3): Likewise.
8550 (vst1_s32_x3): Likewise.
8551 (vst1_u8_x3): Likewise.
8552 (vst1_u16_x3): Likewise.
8553 (vst1_u32_x3): Likewise.
8554 (vst1_f16_x3): Likewise.
8555 (vst1_f32_x3): Likewise.
8556 (vst1_p64_x3): Likewise.
8557 (vst1q_s8_x3): Likewise.
8558 (vst1q_p8_x3): Likewise.
8559 (vst1q_s16_x3): Likewise.
8560 (vst1q_p16_x3): Likewise.
8561 (vst1q_s32_x3): Likewise.
8562 (vst1q_s64_x3): Likewise.
8563 (vst1q_u8_x3): Likewise.
8564 (vst1q_u16_x3): Likewise.
8565 (vst1q_u32_x3): Likewise.
8566 (vst1q_u64_x3): Likewise.
8567 (vst1q_f16_x3): Likewise.
8568 (vst1q_f32_x3): Likewise.
8569 (vst1q_f64_x3): Likewise.
8570 (vst1q_p64_x3): Likewise.
8572 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8574 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8575 user_label_prefix to name.
8577 * tree-core.h: Update comment about the format of NAME string
8578 passed to handler in attribute_spec.
8580 * config/msp430/msp430.md: Remove erroneous subreg expression from
8581 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8582 zero_extend{q,h}isi2.
8584 2018-05-30 Borislav Petkov <bp@suse.de>
8586 * doc/extend.texi: Document some architecture specific
8587 constraints and sort entries.
8589 2018-05-30 Martin Sebor <msebor@redhat.com>
8592 * builtins.c (expand_builtin_stpcpy_1): New function.
8593 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8594 only if the former succeeds.
8596 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8598 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8601 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8603 * doc/invoke.texi (-flinker-output): Document
8605 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8607 * passes.c (ipa_write_summaries): Only modify statements if body
8609 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8610 incrementally linking.
8611 (ipa_passes): Likewise.
8612 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8613 pass down resolution info.
8614 * common.opt (flag_incremental_link): Update info.
8615 * gcc.c (plugin specs): Turn flinker-output=* to
8616 -plugin-opt=-linker-output-known
8617 * toplev.c (compile_file): Also cut compilation when doing incremental
8619 * flag-types. (enum lto_partition_model): Add
8620 LTO_LINKER_OUTPUT_NOLTOREL.
8621 (invoke.texi): Add -flinker-output docs.
8622 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8623 link same way as WPA; do not stream in dead initializers.
8625 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8626 darf2out_register_external_decl): Support incremental link.
8628 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8630 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8632 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8634 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8635 it down to simple_object_copy_lto_debug_sections.
8636 (run_gcc): Determine incremental LTO link time and configure
8637 lto1 into non-wpa mode, disable renaming of debug sections.
8639 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8641 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8642 descriptions of various incorrectly documented functions.
8644 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8647 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8648 address check not strict.
8650 2018-05-30 Richard Biener <rguenther@suse.de>
8652 PR tree-optimization/85964
8653 * tracer.c (better_p): Drop initialized count check, we only
8654 call the function with initialized counts now.
8655 (find_best_successor): Do find a best edge if one
8656 has uninitialized count.
8657 (find_best_predecessor): Likewise. Do BB frequency check only
8658 if count is initialized.
8660 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8662 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8663 (aarch64_ldrstr_offset_compare): New.
8664 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8665 load/store orderings.
8666 (aarch64_gen_adjusted_ldpstp): Likewise.
8668 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8670 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8671 Check for subset of GENERAL_REGS and FP_REGS.
8672 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8675 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8677 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8678 and wi::to_poly_offset. Add the current offset and then check
8679 whether the sum fits, rather than using an unchecked addition of
8680 a checked term. Check for a shwi rather than a uhwi.
8681 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8682 (store_constructor): Use poly_int_tree_p.
8683 (expand_expr_real_1): Likewise.
8684 * function.c (assign_temp): Likewise.
8685 * fold-const.c (const_binop): Use poly_int_tree_p and
8687 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8689 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8690 to_poly_offset for MEM offsets.
8691 * ipa-icf.c (sem_variable::equals): Likewise.
8692 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8693 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8694 wi::to_poly_offset for BIT_FIELD_REF offsets.
8695 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8697 * var-tracking.c (emit_note_insn_var_location): Use
8698 tree_to_poly_uint64.
8700 2018-05-29 Jim Wilson <jimw@sifive.com>
8702 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8704 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8707 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8708 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8710 (sse4_1_round<mode>2): Use nonimmediate_operand
8711 for operand 1 predicate.
8713 2018-05-29 Martin Sebor <msebor@redhat.com>
8714 Richard Biener <rguenther@suse.de>
8717 * calls.c (get_size_range): Call determine_value_range instead
8718 of get_value_range..
8719 * tree-vrp.h (determine_value_range): Declared new function.
8720 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8722 2018-05-29 Richard Biener <rguenther@suse.de>
8724 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8725 sure to use non-pattern stmts for get_earlier_stmt arguments.
8726 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8727 called on pattern stmts.
8728 (get_later_stmt): Likewise.
8730 2018-05-29 Martin Liska <mliska@suse.cz>
8732 PR gcov-profile/85759
8733 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8736 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8738 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8740 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8744 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8745 VEC_PACK_FLOAT_EXPR): New tree codes.
8746 * tree-pretty-print.c (op_code_prio): Handle
8747 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8748 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8749 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8750 * tree-inline.c (estimate_operator_cost): Likewise.
8751 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8752 * fold-const.c (const_binop): Likewise.
8753 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8754 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8755 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8756 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8757 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8758 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8759 * expr.c (expand_expr_real_2): Likewise.
8760 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8761 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8762 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8764 * optabs.c (expand_widen_pattern_expr): For
8765 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8766 sign from result type rather than operand's type.
8767 (expand_binop_directly): For vec_packu_float_optab and
8768 vec_packs_float_optab allow result type to be different from operand's
8770 * optabs-tree.c (optab_for_tree_code): Handle
8771 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8772 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8773 * tree-vect-generic.c (expand_vector_operations_1): Handle
8774 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8775 VEC_PACK_FLOAT_EXPR.
8776 * tree-vect-stmts.c (supportable_widening_operation): Handle
8778 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8779 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8780 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8781 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8782 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8784 (vec_pack<floatprefix>_float_<mode>): New expander.
8785 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8787 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8788 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8789 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8790 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8791 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8793 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8794 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8795 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8796 VEC_PACK_FLOAT_EXPR): Document.
8798 2018-05-29 Richard Biener <rguenther@suse.de>
8800 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8802 (stmt_vec_info_vec): Make pointer.
8803 (init_stmt_vec_info_vec): Remove.
8804 (free_stmt_vec_info_vec): Likewise.
8805 (set_stmt_vec_info_vec): New function.
8806 (free_stmt_vec_infos): Likewise.
8807 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8808 (set_vinfo_for_stmt): Likewise.
8809 (get_earlier_stmt): Likewise.
8810 (get_later_stmt): Likewise.
8811 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8812 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8813 (vec_info::~vec_info): Free stmt_vec_infos.
8814 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8815 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8816 (pass_slp_vectorize::execute): Likewise.
8817 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8818 (free_stmt_vec_info_vec): Likewise.
8819 (set_stmt_vec_info_vec): New function.
8820 (free_stmt_vec_infos): Likewise.
8821 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8822 the global stmt_vec_info_vec.
8823 * tree-parloops.c (gather_scalar_reductions): Use
8824 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8827 2018-05-29 Richard Biener <rguenther@suse.de>
8829 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8831 2018-05-29 Martin Liska <mliska@suse.cz>
8832 David Malcolm <dmalcolm@redhat.com>
8834 * vec.c (test_reverse): New.
8835 (vec_c_tests): Add new test.
8836 * vec.h (vl_ptr>::reverse): New function.
8838 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8840 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8842 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8845 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8847 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8849 2018-05-28 Richard Biener <rguenther@suse.de>
8851 PR tree-optimization/85933
8852 * tree-vect-data-refs.c (vect_record_base_alignments): Only
8853 look at stmts marked as vectorizable.
8855 2018-05-28 Richard Biener <rguenther@suse.de>
8857 PR tree-optimization/85934
8858 * tree-vect-generic.c (expand_vector_operations_1): Hoist
8859 vector boolean check before scalar optimization.
8861 2018-05-28 Jakub Jelinek <jakub@redhat.com>
8863 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8866 2018-05-28 Mark Wielaard <mark@klomp.org>
8868 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8869 if it is an expression containing a minus sign.
8871 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
8873 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8875 2018-05-27 Paul Koning <ni1d@arrl.net>
8877 * config/pdp11/pdp11.md (truncsihi2): Remove.
8879 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
8880 Chung-Ju Wu <jasonwucj@gmail.com>
8882 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8884 (unaligned_store_dw): Ditto.
8885 * config/nds32/nds32-memory-manipulation.c
8886 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8887 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8888 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8889 (emit_setmem_word_loop): Rename to ...
8890 (emit_setmem_doubleword_loop): ... this.
8891 (nds32_gen_dup_4_byte_to_word_value): New function.
8892 (nds32_gen_dup_8_byte_to_double_word_value): New function.
8893 (nds32_expand_setmem_loop): Refine implementation.
8894 (nds32_expand_setmem_loop_v3m): Ditto.
8895 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8898 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8900 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8902 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8904 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8905 (nds32_init_machine_status): Initialize machine->attr_naked_p and
8906 machine->attr_no_prologue_p.
8907 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8908 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8909 (nds32_expand_epilogue): Consider attr_naked_p.
8910 (nds32_expand_epilogue_v3pop): Likewise.
8911 (nds32_can_use_return_insn): Likewise.
8912 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8913 attr_no_prologue_p fields.
8914 * config/nds32/nds32.opt (mret-in-naked-func): New option.
8916 2018-05-27 Jakub Jelinek <jakub@redhat.com>
8919 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8921 * config/i386/sse.md
8922 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8924 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8926 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8928 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8930 (*<floatsuffix>floatv2div2sf2): Rename to ...
8931 (*float<floatunssuffix>v2div2sf2): ... this.
8932 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8933 (float<floatunssuffix>v2div2sf2_mask): ... this.
8934 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8935 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8936 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8938 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8940 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8942 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8944 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8946 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8948 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8949 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8950 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8951 gen_ufix_truncv8dfv8si2.
8952 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8953 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8954 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8955 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8956 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8957 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8958 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8959 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8961 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8965 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
8967 2018-05-25 Jim Wilson <jimw@sifive.com>
8969 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
8970 * config/riscv/riscv.c (struct machine_function): Add
8971 interrupt_handler_p and attribute_checked_p fields.
8972 (riscv_attribute_table): Add interrupt.
8973 (riscv_interrupt_type_p): New.
8974 (riscv_save_reg_p): Save extra regs for interrupt handler.
8975 (riscv_use_save_libcall): Return false for interrupt handler.
8976 (riscv_first_stack_step): Add forward declaration.
8977 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
8978 for interrupt handler with large frame. Use it for saved reg list.
8979 (riscv_expand_prologue): Move flag_stack_usage_info support to
8980 eliminate duplication.
8981 (riscv_expand_epilogue): Generate mret for interrupt handler.
8982 (riscv_epilogue_uses): New.
8983 (riscv_can_use_return_insn): Return false for interrupt handler.
8984 (riscv_function_ok_for_sibcall): Likewise.
8985 (riscv_set_current_function): Add interrupt handler support.
8986 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8987 * config/riscv/riscv.md (UNSPECV_MRET): New.
8989 (riscv_frflags, riscv_fsflags): Use tab after opcode.
8991 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8993 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
8995 PR tree-optimization/85712
8996 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8997 this candidate has already been replaced in-situ by a copy.
8999 2018-05-25 Jason Merrill <jason@redhat.com>
9001 PR c++/80485 - inline function non-zero address.
9002 * symtab.c (nonzero_address): Check DECL_COMDAT.
9004 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9007 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9008 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9009 extension of SImode operation. Use const123_operand predicate.
9010 (*saddsi_1): Remove.
9011 (*saddl_se_1): Ditto.
9013 (*ssubl_se_1): Ditto.
9014 * config/alpha/predicates.md (const123_operand): New predicate.
9015 * config/alpha/constraints.md (P): Use IN_RANGE.
9017 2018-05-25 Richard Biener <rguenther@suse.de>
9019 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9021 (ref_maybe_used_by_stmt_p): Likewise.
9022 (stmt_may_clobber_ref_p): Likewise.
9023 (stmt_may_clobber_ref_p_1): Likewise.
9024 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9026 (ref_maybe_used_by_stmt_p): Likewise.
9027 (stmt_may_clobber_ref_p): Likewise.
9028 (stmt_may_clobber_ref_p_1): Likewise.
9029 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9030 the alias oracle to disambiguate DRs with stmts DR analysis
9032 (vect_analyze_data_refs): Do not give up on not analyzable
9033 DRs for BB vectorization. Remove code truncating the dataref
9036 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9039 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9040 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9041 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9043 2018-05-25 Richard Biener <rguenther@suse.de>
9045 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9046 function, combining stmt data ref gathering and fatal analysis
9048 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9049 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9050 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9051 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9052 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9054 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9056 PR tree-optimization/85720
9057 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9058 SCC if all partitions are builtins.
9059 (version_loop_by_alias_check): New parameter. Generate cancelable
9060 runtime alias check if all partitions are builtins.
9061 (distribute_loop): Update call to above function.
9063 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9065 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9066 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9067 (parm_default_def_partition_arg): Ditto.
9068 (set_parm_default_def_partition): Ditto.
9069 (get_parm_default_def_partitions): Ditto and make it static.
9070 (get_undefined_value_partitions): Ditto and make it static.
9071 (remove_ssa_form): Refactor call to init_var_map here.
9072 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9073 computation for loop region.
9074 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9075 (register_default_def): Delete.
9076 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9077 (parm_default_def_partition_arg): Ditto.
9078 (set_parm_default_def_partition): Ditto.
9079 (get_parm_default_def_partitions): Ditto and make it static.
9080 (get_undefined_value_partitions): Ditto and make it static.
9081 (coalesce_with_default, coalesce_with_default): Update comment.
9082 (create_coalesce_list_for_region): New func factored out from
9083 create_outofssa_var_map.
9084 (populate_coalesce_list_for_outofssa): New func factored out from
9085 create_outofssa_var_map and coalesce_ssa_name.
9086 (create_outofssa_var_map): Delete.
9087 (coalesce_ssa_name): Refactor to support live range computation.
9088 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9089 (get_parm_default_def_partitions): Delete.
9090 (get_undefined_value_partitions): Ditto.
9091 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9092 computation for loop region.
9093 (new_tree_live_info, loe_visit_block): Ditto.
9094 (live_worklist, set_var_live_on_entry): Ditto.
9095 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9096 * tree-ssa-live.h (struct _var_map): New fields.
9097 (init_var_map): Change decl.
9098 (region_contains_p): New.
9100 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9102 * tree-ssa-live.h (live_merge_and_clear): Delete.
9104 2018-05-25 Richard Biener <rguenther@suse.de>
9107 * tree-dump.c (dequeue_and_dump): Remove access to removed
9108 operand 2 of a SWITCH_EXPR.
9110 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9112 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9113 multiplication and division.
9114 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9115 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9116 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9117 (cond_udiv_optab, cond_umod_optab): New optabs.
9118 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9119 (IFN_COND_RDIV): New internal functions.
9120 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9121 TRUNC_MOD_EXPR and RDIV_EXPR.
9122 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9123 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9125 (SVE_INT_BINARY): Include mult.
9126 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9127 (optab, sve_int_op): Handle mult.
9128 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9130 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9131 for SVE_INT_BINARY_SD.
9133 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9135 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9136 (optab, sve_int_op): Handle div and udiv.
9137 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9138 for SVE_INT_BINARY_SD.
9139 (*<optab><mode>3): New insn for the same.
9141 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9143 * tree-vect-patterns.c: Include predict.h.
9144 (vect_recog_divmod_pattern): Restrict check for division support
9145 to when optimizing for size.
9147 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9149 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9150 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9151 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9152 (gimple_match_op::set_op): Likewise.
9153 (gimple_resimplify4): Declare.
9154 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9155 (expr::gen_transform): Likewise.
9156 (decision_tree::gen): Generate a simplification routine for 4 operands.
9157 * gimple-match-head.c (gimple_simplify): Add an overload for
9158 4 operands. In the top-level function, handle up to 4 call
9159 arguments and call gimple_resimplify4.
9160 (gimple_resimplify4): New function.
9161 (build_call_internal): Pass a fourth operand.
9162 (maybe_push_to_seq): Likewise.
9163 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9164 Fold VEC_COND_EXPRs of an operation and a default value into
9165 an IFN_COND_* function if possible.
9166 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9168 (SVE_COND_FP_BINARY): Include them.
9169 (optab, sve_fp_op): Handle them.
9170 (SVE_INT_BINARY_REV): New code iterator.
9171 (SVE_COND_FP_BINARY_REV): New int iterator.
9172 (commutative): New int attribute.
9173 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9175 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9177 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9178 (*cond_<optab><mode>): New patterns for reversed operands.
9180 2018-05-25 Richard Biener <rguenther@suse.de>
9182 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9183 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9184 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9185 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9186 * tree-vect-data-refs.c (everywhere): Adjust users.
9187 * tree-vect-loop.c (everywhere): Likewise.
9188 * tree-vect-slp.c (everywhere): Likewise.
9189 * tree-vect-stmts.c (everywhere): Likewise.
9190 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9192 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9194 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9196 (gcc_cv_as_section_exclude): ... this.
9197 Try Solaris as #exclude syntax.
9198 * configure: Regenerate.
9199 * config.in: Regenerate.
9200 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9202 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9203 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9205 * varasm.c (default_elf_asm_named_section): Don't check if
9206 HAVE_GAS_SECTION_EXCLUDE is defined.
9208 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9210 * doc/md.texi: Update the documentation of the cond_* optabs
9211 to mention the new final operand. Fix GET_MODE_NUNITS call.
9212 Describe the scalar case too.
9213 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9214 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9216 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9217 (get_conditional_internal_fn): Update comment.
9218 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9219 accumulator value as a final argument to conditional functions.
9220 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9221 a define_expand and add an "else" operand. Assert for now that
9222 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9223 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9224 (*cond_<optab><mode>): New patterns.
9225 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9226 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9227 (UNSPEC_COND_EOR): Delete.
9228 (optab): Remove associated mappings.
9229 (SVE_INT_BINARY): New code iterator.
9230 (sve_int_op): Remove int attribute and add "minus" to the code
9232 (SVE_COND_INT_OP): Delete.
9233 (SVE_COND_FP_OP): Rename to...
9234 (SVE_COND_FP_BINARY): ...this.
9236 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9238 * optabs.c (can_reuse_operands_p): New function.
9239 (maybe_legitimize_operands): Try to reuse the results for
9242 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9244 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9245 Add {q} suffix to insn mnemonic.
9247 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9249 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9250 (msp430_warn_func_return): New.
9252 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9254 * fold-const.c (tree_nonzero_bits): New function.
9255 * fold-const.h (tree_nonzero_bits): Likewise.
9256 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9257 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9259 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9263 * varasm.c (assemble_alias): Check ifunc_resolver only on
9266 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9269 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9270 when memory input operand is handled.
9272 2018-05-24 Luis Machado <luis.machado@linaro.org>
9274 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9276 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9278 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9280 * match.pd: Delay FMA folds until after vectorization.
9282 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9285 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9286 address check not strict.
9288 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9290 * gimple-match.h (gimple_match_op): New class.
9291 (mprts_hook): Replace parameters with a gimple_match_op *.
9292 (maybe_build_generic_op): Likewise.
9293 (gimple_simplified_result_is_gimple_val): Replace parameters with
9294 a const gimple_match_op *.
9295 (gimple_simplify): Replace code_helper * and tree * parameters with
9296 a gimple_match_op * parameter.
9297 (gimple_resimplify1): Replace code_helper *, tree and tree *
9298 parameters with a gimple_match_op * parameter.
9299 (gimple_resimplify2): Likewise.
9300 (gimple_resimplify3): Likewise.
9301 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9302 parameters with a gimple_match_op * parameter.
9303 * gimple-match-head.c (gimple_simplify): Change prototypes of
9304 auto-generated functions to take a gimple_match_op * instead of
9305 separate code_helper * and tree * parameters. Make the same
9306 change in the top-level overload and update calls to the
9307 gimple_resimplify routines. Update calls to the auto-generated
9308 functions and to maybe_push_res_to_seq in the publicly-facing
9309 operation-specific gimple_simplify overloads.
9310 (gimple_match_op::MAX_NUM_OPS): Define.
9311 (gimple_resimplify1): Replace rcode and ops with a single res_op
9312 parameter. Update call to gimple_simplify.
9313 (gimple_resimplify2): Likewise.
9314 (gimple_resimplify3): Likewise.
9315 (mprts_hook): Replace parameters with a gimple_match_op *.
9316 (maybe_build_generic_op): Likewise.
9317 (build_call_internal): Replace type, nargs and ops with
9318 a gimple_match_op *.
9319 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9320 with a single gimple_match_op *. Update calls to mprts_hook,
9321 build_call_internal and gimple_simplified_result_is_gimple_val.
9322 Factor out code that is common to the tree_code and combined_fn cases.
9323 * genmatch.c (expr::gen_transform): Replace tem_code and
9324 tem_ops with a gimple_match_op called tem_op. Update calls
9325 to the gimple_resimplify functions and maybe_push_res_to_seq.
9326 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9327 res_ops. Update call to the gimple_resimplify functions.
9328 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9329 (decision_tree::gen): Make the functions take a gimple_match_op *
9330 called res_op instead of separate res_code and res_ops parameters.
9331 Update call accordingly.
9332 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9333 and ops with a single res_op parameter. Update calls to
9334 maybe_build_generic_op and maybe_push_res_to_seq.
9335 (fold_stmt_1): Update calls to gimple_simplify and
9336 replace_stmt_with_simplification.
9337 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9338 and gimple_simplified_result_is_gimple_val.
9339 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9341 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9342 with a gimple_match_op *.
9343 (vn_nary_build_or_lookup): Likewise. Update call to
9344 vn_nary_build_or_lookup_1.
9345 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9346 gimple_match_op *. Update calls to the gimple_resimplify routines
9347 and to gimple_simplified_result_is_gimple_val.
9348 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9349 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9350 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9351 (visit_nary_op): Likewise.
9352 (visit_reference_op_load): Likewise.
9354 2018-05-23 Luis Machado <luis.machado@linaro.org>
9356 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9357 modifier for printing the step amount.
9359 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9362 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9365 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9367 * doc/sourcebuild.texi (Endianness): New subsubsection.
9369 2018-05-23 Luis Machado <luis.machado@linaro.org>
9371 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9372 <prefetch_dynamic_strides>: New const bool field.
9373 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9374 prefetch_dynamic_strides.
9375 (exynosm1_prefetch_tune): Likewise.
9376 (thunderxt88_prefetch_tune): Likewise.
9377 (thunderx_prefetch_tune): Likewise.
9378 (thunderx2t99_prefetch_tune): Likewise.
9379 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9381 (aarch64_override_options_internal): Update to set
9382 PARAM_PREFETCH_DYNAMIC_STRIDES.
9383 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9384 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9385 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9386 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9387 prefetch-dynamic-strides setting.
9389 2018-05-23 Luis Machado <luis.machado@linaro.org>
9391 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9392 <minimum_stride>: New const int field.
9393 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9394 minimum_stride field defaulting to -1.
9395 (exynosm1_prefetch_tune): Likewise.
9396 (thunderxt88_prefetch_tune): Likewise.
9397 (thunderx_prefetch_tune): Likewise.
9398 (thunderx2t99_prefetch_tune): Likewise.
9399 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9400 <default_opt_level>: Set to 3.
9401 (aarch64_override_options_internal): Update to set
9402 PARAM_PREFETCH_MINIMUM_STRIDE.
9403 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9404 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9405 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9406 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9407 stride is constant and is below the minimum stride threshold.
9409 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9411 * config/arm/arm-cpus.in (mode26): Delete.
9412 (armv4): Delete mode26 reference.
9413 * config/arm/arm.c (arm_configure_build_target): Delete use of
9416 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9418 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9420 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9421 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9423 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9424 (floatunsdidf2): Ditto.
9426 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9428 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9429 (fixuns_trunc<mode>si2_avx512f): Ditto.
9430 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9431 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9432 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9434 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9436 PR rtl-optimization/79985
9437 * df-scan.c (df_insn_refs_collect): Remove special case for
9438 global registers and asm statements.
9440 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9442 * extend.texi (Global Register Variables): Rewrite the bullet list.
9443 Note that the register is available for allocation. Note that access
9444 via inline asm must use constraints. Add note about async-signal
9445 handlers. Remove paragraph about automagic register selection.
9447 2018-05-23 Richard Biener <rguenther@suse.de>
9449 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9450 of fixed offset from memset VN.
9452 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9454 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9456 (alloc_cand_and_find_basis): Initialize first_interp field.
9457 (slsr_process_mul): Modify first_interp field.
9458 (slsr_process_add): Likewise.
9459 (slsr_process_cast): Modify first_interp field for each new
9461 (slsr_process_copy): Likewise.
9462 (dump_candidate): Dump first_interp field.
9463 (replace_mult_candidate): Process all interpretations, not just
9465 (replace_rhs_if_not_dup): Likewise.
9466 (replace_one_candidate): Likewise.
9468 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9470 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9472 (aarch64_needs_frame_chain): New function.
9473 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9475 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9478 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9479 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9480 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9481 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9482 as true for strict-align.
9483 (aarch64_can_inline_p): Perform checks even when callee has no
9484 attributes to check for strict alignment.
9485 * doc/extend.texi (AArch64 Function Attributes): Document
9487 * doc/invoke.texi: (AArch64 Options): Likewise.
9489 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9491 PR tree-optimization/85853
9492 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9493 the handling of the root of the node to...
9494 (vect_slp_analyze_node_operations_1): ...this new function,
9495 and run the whole thing with the child nodes' def types
9496 set according to their SLP node's def type.
9498 2018-05-23 Richard Biener <rguenther@suse.de>
9501 * tree-data-ref.c (create_runtime_alias_checks): Defer
9502 and ignore overflow warnings.
9504 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9506 PR tree-optimization/85822
9507 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9510 2018-05-23 Richard Biener <rguenther@suse.de>
9512 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9513 memset constants via native_interpret_expr.
9515 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9518 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9520 (cgraph_node::create_alias): Likewise.
9521 (cgraph_node::get_availability): Check ifunc_resolver instead
9522 of looking up ifunc attribute.
9523 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9524 * varasm.c (do_assemble_alias): Likewise.
9525 (assemble_alias): Likewise.
9526 (default_binds_local_p_3): Likewise.
9527 * cgraph.h (cgraph_node): Add ifunc_resolver.
9528 (cgraph_node::only_called_directly_or_aliased_p): Return false
9530 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9532 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9533 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9534 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9535 instead of looking up ifunc attribute.
9537 2018-05-22 Luis Machado <luis.machado@linaro.org>
9539 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9541 2018-05-22 Martin Sebor <msebor@redhat.com>
9544 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9545 only when expasion succeeds.
9546 (expand_builtin_strcmp): Same.
9547 (expand_builtin_strncmp): Same.
9549 2018-05-22 Martin Sebor <msebor@redhat.com>
9551 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9553 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9554 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9556 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9557 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9559 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9560 Add check for legitimate_address.
9561 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9562 (aarch64_swap_ldrstr_operands): New.
9563 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9566 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9567 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9569 * config/aarch64/aarch64.md: New patterns to generate stp
9571 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9572 single words and double words.
9573 (load_pair_sw, load_pair_dw): Likewise.
9574 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9576 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9578 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9579 for different mode ldpstp and add peephole for merged zero stores.
9581 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9583 (aarch64_gen_store_pair): Rename calls to match new patterns.
9584 (aarch64_gen_load_pair): Rename calls to match new patterns.
9585 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9586 (load_pair<DREG:mode><DREG2:mode>): ... This.
9587 (store_pair<mode>): Rename to...
9588 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9589 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9591 (V_INT_EQUIV): Handle SImode.
9592 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9595 2018-05-22 Martin Sebor <msebor@redhat.com>
9598 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9599 or ajust the presumed bound on an operation to avoid unnecessary
9602 2018-05-22 Martin Sebor <msebor@redhat.com>
9604 PR tree-optimization/85826
9605 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9606 assuming that a DECL necesarily has a constant size.
9608 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9611 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9613 2018-05-22 Richard Biener <rguenther@suse.de>
9615 PR tree-optimization/85834
9616 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9617 non-constant and non-zero memset arguments.
9619 2018-05-22 Martin Liska <mliska@suse.cz>
9622 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9624 2018-05-22 Richard Biener <rguenther@suse.de>
9626 PR tree-optimization/85863
9627 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9628 comparisons when vectype is specified.
9629 (vectorizable_condition): Do not specify vectype for
9630 vect_is_simple_cond when SLP vectorizing.
9632 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9635 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9636 define __ibm128 as long double.
9637 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9638 as a distinct type when IEEE 128-bit support is enabled.
9639 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9640 128-bit types to use the correct functions.
9641 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9642 convert between 128-bit floating point types that have different
9643 modes but the same representation, instead of using gen_lowpart to
9645 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9647 (IFKF_reg): New attributes to give the register constraints for
9649 (extend<mode>tf2_internal): New insns to mark an explicit
9650 conversion between 128-bit floating point types that have a
9651 different mode but share the same representation.
9653 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9655 PR tree-optimization/85814
9656 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9657 a null return from get_strinfo when unsharing the next
9658 strinfo in the chain.
9660 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9663 * varasm.c (weak_finish): Clean up weak_decls.
9665 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9667 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9668 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9669 UNSPEC_UADALP values.
9670 * config/aarch64/iterators.md (ABAL): New int iterator.
9673 (sur): Add mappings for the above.
9674 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9676 (aarch64_<sur>abal<mode>_4): Likewise.
9677 (aarch64_<sur>adalp<mode>_3): Likewise.
9678 (<sur>sadv16qi): New define_expand.
9680 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9682 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9683 (*movdf_internal): Ditto.
9684 (*rcpsf2_sse): Ditto.
9685 (*rsqrtsf2_sse): Ditto.
9686 (*sqrt<mode>2_sse): Ditto.
9688 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9690 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9692 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9693 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9694 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9695 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9697 * config/aarch64/arm_neon.h: Likewise.
9698 * config/aarch64/iterators.md (VQ_I): New.
9700 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9702 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9704 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9707 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9709 * config/nds32/constraints.md (S): New constraint.
9710 * config/nds32/nds32.md (call_internal): Use constraint S.
9711 (call_value_internal): Likewise.
9712 (sibcall_internal): Likewise.
9713 (sibcall_value_internal): Likewise.
9715 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9716 Chung-Ju Wu <jasonwucj@gmail.com>
9718 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9721 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9722 Chung-Ju Wu <jasonwucj@gmail.com>
9724 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9725 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9726 (nds32_rtx_costs_impl): Simplify.
9727 (nds32_address_cost_impl): Simplify.
9728 (nds32_init_rtx_costs): New function.
9729 (nds32_rtx_costs_speed_prefer): Likewise.
9730 (nds32_rtx_costs_size_prefer): Likewise.
9731 (nds32_address_cost_speed_prefer): Likewise.
9732 (nds32_address_cost_speed_fwprop): Likewise.
9733 (nds32_address_cost_size_prefer): Likewise.
9734 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9735 * config/nds32/nds32.c (nds32_option_override): Use
9736 nds32_init_rtx_costs function.
9738 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9740 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9741 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9742 (TARGET_PIPELINE_N8): Likewise.
9743 (TARGET_PIPELINE_N10): Likewise.
9744 (TARGET_PIPELINE_N13): Likewise.
9745 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9747 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9749 * config/nds32/nds32-fpu.md: Update copyright year.
9751 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9753 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9755 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9757 * config/nds32/nds32.c
9758 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9759 * config/nds32/nds32.opt (minline-asm-r15): New option.
9761 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9763 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9765 * config/nds32/nds32.md (abssi2): New pattern.
9767 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9769 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9770 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9771 Merge insn pattern from sse_cvtsi2ss<round_name> and
9772 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9773 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9774 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9775 using SWI48 mode iterator.
9776 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9777 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9778 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9779 pattern from sse_cvttss2si<round_saeonly_name>
9780 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9781 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9782 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9783 using SWI48 mode iterator.
9784 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9785 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9786 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9787 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9788 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9789 using SWI48 mode iterator.
9790 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9791 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9792 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9793 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9794 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9795 SWI48 mode iterator.
9796 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9797 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9798 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9799 pattern from sse_cvttsd2si<round_saeonly_name>
9800 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9802 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9804 * config/nds32/nds32-md-auxiliary.c
9805 (nds32_valid_smw_lwm_base_p): Refine.
9806 (nds32_output_smw_single_word): Refine.
9807 (nds32_output_smw_double_word): New.
9808 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9810 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9812 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9813 (nds32_output_stack_pop): Refine.
9814 (nds32_expand_unaligned_load): Refine.
9815 (nds32_expand_unaligned_store): Refine.
9817 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9818 Chung-Ju Wu <jasonwucj@gmail.com>
9820 * config/nds32/constants.md: Add TP_REGNUM constant.
9821 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9822 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9824 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9825 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9826 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9827 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9828 and PIC code generation.
9829 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9831 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9833 * config/nds32/nds32.md: Support TLS and PIC.
9834 * config/nds32/nds32.c: Support TLS and PIC.
9835 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9836 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9839 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9841 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9842 mode with E_ prefix.
9844 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9845 Chung-Ju Wu <jasonwucj@gmail.com>
9847 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9848 * config/nds32/nds32-md-auxiliary.c
9849 (symbolic_reference_mentioned_p): New.
9850 (nds32_legitimize_ict_address): New.
9851 (nds32_expand_ict_move): New.
9852 (nds32_indirect_call_referenced_p): New.
9853 (nds32_symbol_binds_local_p): Delete.
9854 (nds32_long_call_p): Modify.
9855 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9856 * config/nds32/nds32-protos.h
9857 (symbolic_reference_mentioned_p): Declare.
9858 (nds32_legitimize_ict_address): Declare.
9859 (nds32_expand_ict_move): Declare.
9860 (nds32_indirect_call_referenced_p): Declare.
9861 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9862 (nds32_relax_group): Use nds32_ict_const_p as condition.
9863 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9864 (nds32_asm_file_start): Output ict_model directive in asm code.
9865 (nds32_legitimate_address_p): Consider indirect call.
9866 (nds32_print_operand): Consider indirect call.
9867 (nds32_print_operand_address): Consider indirect call.
9868 (nds32_insert_attributes): Handle "indirect_call" attribute.
9869 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9870 (TARGET_LEGITIMATE_CONSTANT_P): Define.
9871 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9872 (TARGET_DELEGITIMIZE_ADDRESS): Define.
9873 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9874 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9875 (TARGET_ICT_MODEL_SMALL): Define.
9876 (TARGET_ICT_MODEL_LARGE): Define.
9877 * config/nds32/nds32.md (movsi): Consider ict model.
9878 (call, call_value): Consider ict model.
9879 (sibcall, sibcall_value): Consider ict model.
9880 * config/nds32/nds32.opt (mict-model): New option.
9881 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9884 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
9885 Monk Chiang <sh.chiang04@gmail.com>
9886 Jim Wilson <jimw@sifive.com>
9888 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9889 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
9890 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9891 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9892 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
9893 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9894 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9895 compute save_libcall_adjustment properly.
9896 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9897 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9898 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9899 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9900 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9901 (ABI_SPEC): Handle mabi=ilp32e.
9902 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9903 (RVE): Add RVE mask.
9904 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9905 <-march>: Add rv32e as an example.
9907 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
9910 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9911 (intra_create_variable_infos): Handle C++ constructors.
9913 2018-05-18 Martin Liska <mliska@suse.cz>
9915 * passes.def: Remove a redundant pass.
9917 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
9920 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9922 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9924 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9926 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9928 (armv2, armv2a, armv3, armv3m): Delete architectures.
9929 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9930 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9931 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9933 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9934 (*mulsidi3adddi): Likewise.
9935 (mulsidi3): Likewise.
9936 (*mulsidi3_nov6): Likewise.
9937 (umulsidi3): Likewise.
9938 (umulsidi3_nov6): Likewise.
9939 (umaddsidi4): Likewise.
9940 (*umulsidi3adddi): Likewise.
9941 (smulsi3_highpart): Likewise.
9942 (*smulsi3_highpart_nov6): Likewise.
9943 (umulsi3_highpart): Likewise.
9944 (*umulsi3_highpart_nov6): Likewise.
9945 * config/arm/arm.h (arm_arch3m): Delete.
9946 * config/arm/arm.c (arm_arch3m): Delete.
9947 (arm_option_override_internal): Update armv3-related comment.
9948 (arm_configure_build_target): Delete use of isa_bit_mode32.
9949 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9950 (arm_rtx_costs_internal): Delete check of arm_arch3m.
9951 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9953 (mulusa3): Likewise.
9954 * config/arm/arm-protos.h (arm_arch3m): Delete.
9955 * config/arm/arm-tables.opt: Regenerate.
9956 * config/arm/arm-tune.md: Likewise.
9957 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9958 deleted architectures.
9960 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9962 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9963 (armv5t, armv5te): New features.
9964 (ARMv5, ARMv5e): Delete fgroups.
9965 (ARMv5t, ARMv5te): Adjust for above changes.
9967 (armv5, armv5e): Delete arches.
9968 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
9970 (*call_reg_arm): Likewise.
9971 (*call_value_reg_armv5): Likewise.
9972 (*call_value_reg_arm): Likewise.
9973 (*call_symbol): Likewise.
9974 (*call_value_symbol): Likewise.
9975 (*sibcall_insn): Likewise.
9976 (*sibcall_value_insn): Likewise.
9978 (prefetch): Likewise.
9979 (define_split and define_peephole2 dependent on arm_arch5):
9981 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
9983 (TARGET_ARM_QBIT): Likewise.
9984 (TARGET_DSP_MULTIPLY): Likewise.
9985 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
9986 (arm_arch5, arm_arch5e): Delete.
9987 (arm_arch5t, arm_arch5te): Declare.
9988 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9989 (arm_arch5t): Declare.
9990 (arm_option_reconfigure_globals): Update for the above.
9991 (arm_options_perform_arch_sanity_checks): Update comment, replace
9992 use of arm_arch5 with arm_arch5t.
9993 (use_return_insn): Likewise.
9994 (arm_emit_call_insn): Likewise.
9995 (output_return_instruction): Likewise.
9996 (arm_final_prescan_insn): Likewise.
9997 (arm_coproc_builtin_available): Likewise.
9998 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9999 arm_arch5e with arm_arch5t and arm_arch5te.
10000 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10001 (arm_arch5t, arm_arch5te): Declare.
10002 * config/arm/arm-tables.opt: Regenerate.
10003 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10004 * config/arm/t-multilib: Likewise.
10005 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10006 instead of arm_arch5.
10007 (*call_reg_thumb1): Likewise.
10008 (*call_value_reg_thumb1_v5): Likewise.
10009 (*call_value_reg_thumb1): Likewise.
10010 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10012 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10014 2018-05-18 Martin Liska <mliska@suse.cz>
10016 PR gcov-profile/84846
10017 * doc/gcov.texi: Document -t option of gcov tool.
10019 2018-05-18 Martin Liska <mliska@suse.cz>
10021 PR gcov-profile/84846
10022 * gcov.c (print_usage): Add new -t option.
10023 (process_args): Handle the option.
10024 (generate_results): Use stdout as output when requested by
10027 2018-05-18 Martin Liska <mliska@suse.cz>
10029 PR gcov-profile/84846
10030 * coverage.c (coverage_init): Write PWD to .gcno file.
10031 * doc/gcov.texi: Document how working directory is printed.
10032 * gcov-dump.c (dump_gcov_file): Print PWD.
10033 * gcov.c (output_intermediate_file): Likewise.
10034 (read_graph_file): Read PWD string.
10035 (output_lines): Print PWD.
10037 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10039 PR middle-end/85817
10040 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10041 for retval and return false if all args to phi are zero.
10043 2018-05-18 Richard Biener <rguenther@suse.de>
10045 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10047 (evrp_dom_walker::before_dom_children): Call it.
10049 2018-05-18 Richard Biener <rguenther@suse.de>
10051 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10052 results when processing array refs with variable index.
10054 2018-05-18 Toon Moene <toon@moene.org>
10056 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10057 directly after that of -floop-interchange. Indicate that both
10058 options are enabled by default when specifying -O3.
10060 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10062 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10063 iterator. Delete separate integer-mode vec_set<mode> expander.
10064 (aarch64_simd_vec_setv2di): Delete.
10065 (vec_setv2di): Delete.
10066 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10067 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10068 the "w, r" alternative.
10070 2018-05-18 Martin Liska <mliska@suse.cz>
10072 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10073 * tree-pass.h (make_pass_lower_switch_O0): New function.
10074 * tree-switch-conversion.c (node_has_low_bound): Remove.
10075 (node_has_high_bound): Likewise.
10076 (node_is_bounded): Likewise.
10077 (class pass_lower_switch): Make it a template type and create
10079 (pass_lower_switch::execute): Add template argument.
10080 (make_pass_lower_switch): New function.
10081 (make_pass_lower_switch_O0): New function.
10082 (do_jump_if_equal): Remove.
10083 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10084 all the hard work to tree optimization passes.
10086 2018-05-18 Martin Liska <mliska@suse.cz>
10088 * dbgcnt.c (limit_low): Renamed from limit.
10089 (limit_high): New variable.
10090 (dbg_cnt_is_enabled): Check for upper limit.
10091 (dbg_cnt): Adjust dumping.
10092 (dbg_cnt_set_limit_by_index): Add new argument for high
10094 (dbg_cnt_set_limit_by_name): Likewise.
10095 (dbg_cnt_process_single_pair): Parse new format.
10096 (dbg_cnt_process_opt): Use strtok.
10097 (dbg_cnt_list_all_counters): Remove 'value' and add
10099 * doc/invoke.texi: Document changes.
10101 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10103 * doc/sourcebuild.texi (scalar_all_fma): Document.
10104 * tree.def (FMA_EXPR): Delete.
10105 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10106 * internal-fn.c (ternary_direct): New macro.
10107 (expand_ternary_optab_fn): Likewise.
10108 (direct_ternary_optab_supported_p): Likewise.
10109 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10110 * builtins.c (fold_builtin_fma): Delete.
10111 (fold_builtin_3): Don't call it.
10112 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10113 * expr.c (expand_expr_real_2): Likewise.
10114 * fold-const.c (operand_equal_p): Likewise.
10115 (fold_ternary_loc): Likewise.
10116 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10117 * gimple.c (DEFTREECODE): Likewise.
10118 * gimplify.c (gimplify_expr): Likewise.
10119 * optabs-tree.c (optab_for_tree_code): Likewise.
10120 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10121 * tree-eh.c (operation_could_trap_p): Likewise.
10122 (stmt_could_throw_1_p): Likewise.
10123 * tree-inline.c (estimate_operator_cost): Likewise.
10124 * tree-pretty-print.c (dump_generic_node): Likewise.
10125 (op_code_prio): Likewise.
10126 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10127 * tree-ssa-operands.c (get_expr_operands): Likewise.
10128 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10129 * fold-const-call.h (fold_fma): Delete.
10130 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10131 CFN_FNMA and CFN_FNMS.
10132 (fold_fma): Delete.
10133 * genmatch.c (combined_fn): New enum.
10134 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10135 (commutative_op): New function.
10136 (commutate): Use it. Handle more than 2 operands.
10137 (dt_operand::gen_gimple_expr): Use commutative_op.
10138 (parser::parse_expr): Allow :c to be used with non-binary
10139 operators if the commutative operand is known.
10140 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10141 CFN_FMS, CFN_FNMA and CFN_FNMS.
10142 (backprop::process_assign_use): Remove FMA_EXPR handling.
10143 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10144 (gen_hsa_fma): New function.
10145 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10146 IFN_FNMA and IFN_FNMS.
10147 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10148 * gimple-fold.h (follow_all_ssa_edges): Declare.
10149 * gimple-fold.c (follow_all_ssa_edges): New function.
10150 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10151 gimple_build interface and use follow_all_ssa_edges to fold the result.
10152 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10153 instead of checking for optabs directly.
10154 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10155 rather than FMA_EXPRs.
10156 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10157 call to IFN_FMA instead of an FMA_EXPR.
10159 2018-05-17 Jim Wilson <jimw@sifive.com>
10161 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10162 sign extended promoted subreg, and the range does not have the sign bit
10163 set, then do a sign extend.
10165 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10166 test, check for sign extended subreg and/or constant operands, and
10167 do a sign extend in that case.
10169 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10171 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10172 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10174 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10175 Change logics_shift_reg to logics_shift_imm.
10176 (thunderx2t99_fp_loadpair_basic): Delete.
10177 (thunderx2t99_fp_storepair_basic): Delete.
10178 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10179 (thunderx2t99_asimd_polynomial): Delete.
10180 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10181 and neon_fp_mul_d_scalar_q.
10182 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10183 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10184 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10185 (thunderx2t99_asimd_lut): Add missing tbl types.
10186 (thunderx2t99_asimd_ext): Delete.
10187 (thunderx2t99_asimd_load1_1_mult): Delete.
10188 (thunderx2t99_asimd_load1_2_mult): Delete.
10189 (thunderx2t99_asimd_load1_ldp): New.
10190 (thunderx2t99_asimd_load1): New.
10191 (thunderx2t99_asimd_load2): Add missing *load2* types.
10192 (thunderx2t99_asimd_load3): New.
10193 (thunderx2t99_asimd_load4): New.
10194 (thunderx2t99_asimd_store1_1_mult): Delete.
10195 (thunderx2t99_asimd_store1_2_mult): Delete.
10196 (thunderx2t99_asimd_store2_mult): Delete.
10197 (thunderx2t99_asimd_store2_onelane): Delete.
10198 (thunderx2t99_asimd_store_stp): New.
10199 (thunderx2t99_asimd_store1): New.
10200 (thunderx2t99_asimd_store2): New.
10201 (thunderx2t99_asimd_store3): New.
10202 (thunderx2t99_asimd_store4): New.
10204 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10206 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10207 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10209 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10210 Segher Boessenkool <segher@kernel.crashing.org>
10213 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10216 2018-05-17 Richard Biener <rguenther@suse.de>
10218 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10219 for pruning loop and prune defs feeding only already visited PHIs.
10221 2018-05-17 Richard Biener <rguenther@suse.de>
10223 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10225 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10226 Richard Biener <rguenther@suse.de>
10228 PR tree-optimization/85793
10229 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10230 for VMAT_ELEMENTWISE.
10232 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10234 * internal-fn.h (lookup_internal_fn): Declare
10235 * internal-fn.c (lookup_internal_fn): New function.
10236 * gimple.c (gimple_build_call_from_tree): Handle calls to
10237 internal functions.
10238 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10239 internal function names.
10240 * tree-pretty-print.c (dump_generic_node): Likewise.
10241 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10243 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10245 * gimple-fold.h (gimple_build): Make the function forms take
10246 combined_fn rather than built_in_function.
10247 (gimple_simplify): Likewise.
10248 * gimple-match-head.c (gimple_simplify): Likewise.
10249 * gimple-fold.c (gimple_build): Likewise.
10250 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10251 rather than gimple_build_call_internal.
10252 (get_initial_defs_for_reduction): Likewise.
10253 (vect_create_epilog_for_reduction): Likewise.
10254 (vectorizable_live_operation): Likewise.
10256 2018-05-17 Martin Liska <mliska@suse.cz>
10258 * gimple-ssa-sprintf.c (format_directive): Do not use
10259 space in between 'G_' and '('.
10261 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10264 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10265 even if the mask is not all ones.
10268 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10270 (ix86_gimple_fold_builtin): Likewise.
10273 * config/i386/i386.c: Include tree-vector-builder.h.
10274 (ix86_vector_shift_count): New function.
10275 (ix86_fold_builtin): Fold shift builtins by scalar count.
10276 (ix86_gimple_fold_builtin): Likewise.
10278 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10279 _mm512_setzero): New intrinsics.
10281 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10282 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10284 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10285 code generation for cases where splatting a value is not useful.
10286 * simplify-rtx.c (simplify_ternary_operation): Simplify
10287 vec_merge across a vec_duplicate and a paradoxical subreg forming
10288 a vector mode to a vec_concat.
10290 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10292 * config.gcc: Support "goldmont-plus".
10293 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10295 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10296 PROCESSOR_GOLDMONT_PLUS.
10297 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10298 (processor_target_table): Add "goldmont-plus".
10299 (PTA_GOLDMONT_PLUS): Define.
10300 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10301 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10302 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10303 (fold_builtin_cpu): Add "goldmont-plus".
10304 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10305 (ix86_option_override_internal): Add "goldmont-plus".
10306 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10307 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10308 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10309 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10311 2018-05-17 Richard Biener <rguenther@suse.de>
10313 PR tree-optimization/85757
10314 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10315 remove defs that only feed that PHI from further processing.
10317 2018-05-16 Jim Wilson <jimw@sifive.com>
10319 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10321 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10323 2018-05-16 Mark Wielaard <mark@klomp.org>
10325 * dwarf2out.c (count_index_strings): New function.
10326 (output_indirect_strings): Call count_index_strings and generate
10327 header for dwarf_version >= 5.
10329 2018-05-16 Mark Wielaard <mark@klomp.org>
10331 * dwarf2out.c (dwarf_FORM): New function.
10332 (set_indirect_string): Use dwarf_FORM.
10333 (reset_indirect_string): Likewise.
10334 (size_of_die): Likewise.
10335 (value_format): Likewise.
10336 (output_die): Likewise.
10337 (add_skeleton_AT_string): Likewise.
10338 (output_macinfo_op): Likewise.
10339 (index_string): Likewise.
10340 (output_index_string_offset): Likewise.
10341 (output_index_string): Likewise.
10342 (count_index_strings): Likewise.
10344 2018-05-16 Carl Love <cel@us.ibm.com>
10346 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10347 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10349 2018-05-16 Martin Jambor <mjambor@suse.cz>
10351 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10352 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10354 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10356 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10357 (fnma<mode>4): Likewise.
10358 (fms<mode>4): Likewise.
10359 (fnms<mode>4): Likewise.
10360 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10361 (aarch64_fnma<mode>4): Likewise.
10362 (aarch64_fms<mode>4): Likewise.
10363 (aarch64_fnms<mode>4): Likewise.
10364 (aarch64_fnmadd<mode>4): Likewise.
10366 2018-05-16 Jason Merrill <jason@redhat.com>
10368 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10370 2018-05-16 Richard Biener <rguenther@suse.de>
10372 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10373 (dump_stmt_cost): Declare.
10374 (add_stmt_cost): Dump cost we add.
10375 (add_stmt_costs): New function.
10376 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10377 No longer exported.
10378 (vect_analyze_stmt): Adjust prototype.
10379 (vectorizable_condition): Likewise.
10380 (vectorizable_live_operation): Likewise.
10381 (vectorizable_reduction): Likewise.
10382 (vectorizable_induction): Likewise.
10383 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10384 cost vector to pass to vectorizable_ and record afterwards.
10385 (vect_model_reduction_cost): Take cost vector argument and adjust.
10386 (vect_model_induction_cost): Likewise.
10387 (vectorizable_reduction): Likewise.
10388 (vectorizable_induction): Likewise.
10389 (vectorizable_live_operation): Likewise.
10390 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10391 SLP_TREE_NUMBER_OF_VEC_STMTS.
10392 (vect_analyze_slp_cost_1): Remove.
10393 (vect_analyze_slp_cost): Likewise.
10394 (vect_slp_analyze_node_operations): Take visited args and
10395 a target cost vector. Avoid processing already visited stmt sets.
10396 (vect_slp_analyze_operations): Use a local cost vector to gather
10397 costs and register those of non-discarded instances.
10398 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10399 (vect_schedule_slp_instance): Remove copying of
10400 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10402 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10403 adding cost. Record cost entry location.
10404 (vect_prologue_cost_for_slp_op): Function to compute cost of
10405 a constant or invariant generated for SLP vect in the prologue,
10406 split out from vect_analyze_slp_cost_1.
10407 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10408 (vect_model_promotion_demotion_cost): Likewise.
10409 (vect_model_store_cost): Likewise, make static.
10410 (vect_model_load_cost): Likewise.
10411 (vectorizable_bswap): Add cost vector arg and adjust.
10412 (vectorizable_call): Likewise.
10413 (vectorizable_simd_clone_call): Likewise.
10414 (vectorizable_conversion): Likewise.
10415 (vectorizable_assignment): Likewise.
10416 (vectorizable_shift): Likewise.
10417 (vectorizable_operation): Likewise.
10418 (vectorizable_store): Likewise.
10419 (vectorizable_load): Likewise.
10420 (vectorizable_condition): Likewise.
10421 (vectorizable_comparison): Likewise.
10422 (can_vectorize_live_stmts): Likewise.
10423 (vect_analyze_stmt): Likewise.
10424 (vect_transform_stmt): Adjust calls to vectorizable_*.
10425 * tree-vectorizer.c: Include gimple-pretty-print.h.
10426 (dump_stmt_cost): New function.
10428 2018-05-16 Richard Biener <rguenther@suse.de>
10430 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10431 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10432 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10433 (check_name): New callback.
10434 (dse_classify_store): Track cycles via a visited bitmap of PHI
10435 defs and simplify handling of in-loop and across loop dead stores
10436 and properly fail for loop-variant refs. Handle byte-tracking with
10437 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10440 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10442 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10443 (vect_get_mask_type_for_stmt): Likewise.
10444 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10446 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10447 to determine the statement's vector type and the vector type that
10448 should be used for calculating nunits. Deal with cases in which
10449 the type has to be deferred.
10450 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10451 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10452 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10453 (vect_determine_vf_for_stmt): New functions, split out from...
10454 (vect_determine_vectorization_factor): ...here.
10455 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10456 (vect_get_mask_type_for_stmt): New functions, split out from
10457 vect_determine_vectorization_factor.
10459 2018-05-16 Richard Biener <rguenther@suse.de>
10461 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10462 verify the [VEC_]COND_EXPR embedded comparison.
10464 2018-05-15 Martin Sebor <msebor@redhat.com>
10466 PR tree-optimization/85753
10467 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10468 RECORD_TYPE in addition to ARRAY_TYPE.
10470 2018-05-15 Martin Sebor <msebor@redhat.com>
10472 PR middle-end/85643
10473 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10475 2018-05-15 Richard Biener <rguenther@suse.de>
10477 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10478 add by_clobber_p one. Change algorithm to collect all defs
10479 representing uses we need to walk and try reducing them to
10480 a single one before failing.
10481 (dse_dom_walker::dse_optimize_stmt): Adjust.
10483 2018-05-13 Mark Wielaard <mark@klomp.org>
10485 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10486 (size_of_loc_descr): Likewise.
10487 (output_loc_operands): Likewise.
10488 (output_loc_operands_raw): Likewise.
10489 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10490 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10491 (hash_loc_operands): Likewise.
10492 (compare_loc_operands): Likewise.
10494 2018-05-14 Mark Wielaard <mark@klomp.org>
10496 * dwarf2out.c (count_index_addrs): New function.
10497 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10499 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10501 PR tree-optimization/83648
10502 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10503 return value as malloc candidate.
10505 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10508 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10509 param as true in call to suggest_attribute.
10511 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10513 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10514 -mreadonly-in-sdata.
10516 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10518 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10520 (aarch64_crypto_aesd_fused): Likewise.
10522 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10524 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10525 (movsi_aarch64): Likewise.
10526 (load_pairsi): Likewise.
10527 (load_pairdi): Likewise.
10528 (store_pairsi): Likewise.
10529 (store_pairdi): Likewise.
10530 (load_pairsf): Likewise.
10531 (load_pairdf): Likewise.
10532 (store_pairsf): Likewise.
10533 (store_pairdf): Likewise.
10534 (zero_extend): Likewise.
10535 (trunc): Swap alternatives.
10536 (fcvt_target): Add '?' to prefer w over r.
10538 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10541 * config/i386/i386.md: Disallow non-commutative arithmetics in
10542 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10543 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10544 in the peephole2 before it.
10546 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10548 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10549 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10550 (ix86_handle_option): Handle -mcldemote.
10551 * config.gcc: New header.
10552 * config/i386/cldemoteintrin.h: New file.
10553 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10554 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10556 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10557 OPTION_MASK_ISA_CLDEMOTE.
10558 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10559 (ix86_valid_target_attribute_inner_p): Ditto.
10560 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10561 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10562 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10563 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10564 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10566 * config/i386/i386.opt: Add -mcldemote.
10567 * config/i386/x86intrin.h: New header.
10568 * doc/invoke.texi: Add -mcldemote.
10570 2018-05-14 Richard Biener <rguenther@suse.de>
10572 * doc/match-and-simplify.texi: Adjust :s documentation.
10574 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10576 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10577 intended memcpy size.
10578 (REORDER_45): Likewise.
10580 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10582 * sort.cc: New file.
10583 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10584 * vec.c (qsort_chk): Use gcc_qsort.
10585 * Makefile.in (OBJS-libcommon): Add sort.o.
10586 (build/sort.o): New target. Use it...
10587 (BUILD_RTL): ... here, and...
10588 (build/gencfn-macros): ... here, and...
10589 (build/genmatch): ... here.
10591 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10592 Chung-Ju Wu <jasonwucj@gmail.com>
10594 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10595 * config/nds32/nds32-graywolf.md: New file.
10596 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10597 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10599 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10600 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10601 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10602 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10603 * config/nds32/pipelines.md: Include n15 settings.
10605 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10606 Chung-Ju Wu <jasonwucj@gmail.com>
10608 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10609 * config/nds32/nds32-n13.md: New file.
10610 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10611 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10613 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10614 * config/nds32/nds32.md (pipeline_model): Add n13.
10615 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10616 * config/nds32/pipelines.md: Include n13 settings.
10618 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10619 Chung-Ju Wu <jasonwucj@gmail.com>
10621 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10622 * config/nds32/nds32-n10.md: New file.
10623 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10624 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10626 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10627 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10628 * config/nds32/nds32.md (pipeline_model): Add n10.
10629 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10630 * config/nds32/pipelines.md: Include n10 settings.
10632 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10633 Kito Cheng <kito.cheng@gmail.com>
10634 Chung-Ju Wu <jasonwucj@gmail.com>
10636 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10637 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10638 Add enum values for DSP extension instructions.
10639 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10641 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10642 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10643 New code iterators.
10644 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10645 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10646 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10647 * config/nds32/nds32-intrinsic.md: Likewise.
10648 * config/nds32/nds32_intrinsic.h: Likewise.
10649 * config/nds32/nds32-md-auxiliary.c: Likewise.
10650 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10651 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10652 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10653 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10654 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10655 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10656 TYPE_DMAC in switch statement.
10657 * config/nds32/nds32.c: New checking and implementation for DSP
10658 extension instructions.
10659 * config/nds32/nds32.h: Likewise.
10660 * config/nds32/nds32.md: Likewise.
10661 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10662 * config/nds32/predicates.md: Implement new predicates for DSP
10665 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10667 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10668 Reformat alternatives and attributes so it is easier to identify
10669 which constraints/attributes go with which instruction.
10670 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10671 (mov<mode>_softfloat32, FMOVE64): Likewise.
10672 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10673 (mov<mode>_softfloat64, FMOVE64): Likewise.
10675 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10677 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10679 (Basic PowerPC Built-in Functions): The new name of the
10680 subsection previously known as "PowerPC Built-in Functions".
10681 (Basic PowerPC Built-in Functions Available on all Configurations):
10683 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10684 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10685 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10686 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10688 2018-05-11 Martin Jambor <mjambor@suse.cz>
10691 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10694 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10697 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10699 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10701 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10702 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10703 (ix86_handle_option): Handle -mwaitpkg.
10704 * config.gcc: New header.
10705 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10706 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10707 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10709 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10710 OPTION_MASK_ISA_WAITPKG.
10711 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10712 (ix86_option_override_internal): Add PTA_WAITPKG.
10713 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10714 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10715 IX86_BUILTIN_TPAUSE.
10716 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10717 __builtin_ia32_umwait and __builtin_ia32_tpause.
10718 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10719 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10720 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10721 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10722 UNSPECV_TPAUSE): New.
10723 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10724 * config/i386/i386.opt: Add -mwaitpkg.
10725 * config/i386/waitpkgintrin.h: New file.
10726 * config/i386/x86intrin.h: New header.
10727 * doc/invoke.texi: Add -mwaitpkg.
10729 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10732 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10734 (cortex-m0): Use armv6s-m isa.
10735 (cortex-m0plus): Likewise.
10736 (cortex-m1): Likewise.
10737 (cortex-m0.small-multiply): Likewise.
10738 (cortex-m0plus.small-multiply): Likewise.
10739 (cortex-m1.small-multiply): Likewise.
10741 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10742 Jakub Jelinek <jakub@redhat.com>
10744 PR tree-optimization/85692
10745 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10746 source permute as well.
10748 2018-05-11 Martin Liska <mliska@suse.cz>
10751 * doc/extend.texi: Document LLVM style format for no_sanitize
10754 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10756 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10757 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10758 (mode_supports_vsx_dform_quad): Likewise.
10759 (mode_supports_vmx_dform): Move these functions to be next to the
10760 other mode_supports functions.
10761 (mode_supports_dq_form): Likewise.
10762 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10763 mode_supports_dq_form.
10764 (reg_offset_addressing_ok_p): Likewise.
10765 (offsettable_ok_by_alignment): Likewise.
10766 (rs6000_legitimate_offset_address_p): Likewise.
10767 (legitimate_lo_sum_address_p): Likewise.
10768 (rs6000_legitimize_address): Likewise.
10769 (rs6000_legitimize_reload_address): Likewise.
10770 (rs6000_secondary_reload_inner): Likewise.
10771 (rs6000_preferred_reload_class): Likewise.
10772 (rs6000_output_move_128bit): Likewise.
10774 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10776 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10777 Generate SImode target register for null target.
10778 <case IX86_BUILTIN_XGETBV>: Ditto.
10779 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10780 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10782 2018-05-10 Carl Love <cel@us.ibm.com>
10784 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10785 dcbtt and dcbtstt if operands[2] is 0.
10787 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10790 * config/i386/sse.md (usadv64qi): New expander.
10792 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10794 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10795 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10796 -maltivec=be support.
10797 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10798 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10799 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10800 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10801 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10802 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10803 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10804 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10805 altivec_vsumsws): Adjust.
10806 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10807 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10808 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10810 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10811 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10812 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10813 (altivec_lve<VI_char>x): Delete expand.
10814 (*altivec_lve<VI_char>x_internal): Rename to...
10815 (altivec_lve<VI_char>x): ... this.
10816 (altivec_lvxl_<mode>): Delete expand.
10817 (*altivec_lvxl_<mode>_internal): Rename to ...
10818 (altivec_lvxl_<mode>): ... this.
10819 (altivec_stvxl_<mode>): Delete expand.
10820 (*altivec_stvxl_<mode>_internal): Rename to ...
10821 (altivec_stvxl_<mode>): ... this.
10822 (altivec_stve<VI_char>x): Delete expand.
10823 (*altivec_stve<VI_char>x_internal): Rename to ...
10824 (altivec_stve<VI_char>x): ... this.
10825 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10826 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10827 reduc_plus_scal_<mode>): Adjust.
10828 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10830 (rs6000_cpu_cpp_builtins): Adjust.
10831 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10832 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10833 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10834 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10835 -maltivec=be support.
10836 (rs6000_split_vec_extract_var): Adjust.
10837 (rs6000_split_v4si_init): Adjust.
10838 (swap_selector_for_mode): Delete.
10839 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10840 altivec_expand_stvex_be): Delete.
10841 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10842 -maltivec=be support.
10843 (rs6000_gimple_fold_builtin): Ditto.
10844 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10846 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10847 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10848 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10849 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10850 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10851 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10852 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10853 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10854 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10855 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10856 anonymous split): Adjust.
10857 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10858 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10860 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
10862 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10863 when --with-gxx-include-dir is also specified.
10864 * configure: Regenerate.
10866 2018-05-09 Jim Wilson <jimw@sifive.com>
10869 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10870 * config/riscv/t-withmultilib: New.
10871 * config/riscv/withmultilib.h: New.
10872 * doc/install.texi: Document RISC-V --with-multilib-list support.
10874 2018-05-09 Richard Biener <rguenther@suse.de>
10876 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10878 (vect_bb_vectorization_profitable_p): Adjust. Compute
10879 actual scalar cost using the cost vector and the add_stmt_cost
10882 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10884 PR rtl-optimization/85645
10885 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10886 in the REG_CFA_REGISTER note for LR, don't leave it empty.
10888 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10890 PR rtl-optimization/85645
10891 * shrink-wrap.c (spread_components): Return a boolean saying if
10892 anything was changed.
10893 (try_shrink_wrapping_separate): Iterate spread_components until
10894 nothing changes anymore.
10896 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10898 PR rtl-optimization/85645
10899 * regrename.c (build_def_use): Also kill the chains that include the
10900 destination of a REG_CFA_REGISTER note.
10902 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10904 PR rtl-optimization/85645
10905 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10906 insn that has a REG_CFA_REGISTER note.
10908 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
10910 * cfgexpand.c (expand_clobber): New function.
10911 (expand_gimple_stmt_1): Use it.
10912 * tree-vect-stmts.c (vect_clobber_variable): New function,
10914 (vectorizable_simd_clone_call): ...here.
10915 (vectorizable_store): Emit a clobber either side of an
10916 IFN_STORE_LANES sequence.
10917 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10919 2018-05-09 Tom de Vries <tom@codesourcery.com>
10922 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10923 (define_insn "trap_if_false"): Add exit after trap.
10925 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
10927 PR rtl-optimization/85638
10928 * bb-reorder.c: Include common/common-target.h.
10929 (create_forwarder_block): New function extracted from...
10930 (fix_up_crossing_landing_pad): ...here. Rename into...
10931 (dw2_fix_up_crossing_landing_pad): ...this.
10932 (sjlj_fix_up_crossing_landing_pad): New function.
10933 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10934 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10935 from both partitions and exit the loop after one iteration.
10937 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10940 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10942 (Basic PowerPC Built-in Functions): The new name of the
10943 subsection previously known as "PowerPC Built-in Functions".
10944 (Basic PowerPC Built-in Functions Available on all Configurations):
10946 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10948 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10950 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10952 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10955 2018-05-08 Jim Wilson <jimw@sifive.com>
10957 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10958 (LD_EMUL_SUFFIX): New.
10959 (LINK_SPEC): Use it.
10961 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10963 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10965 (Basic PowerPC Built-in Functions): The new name of the
10966 subsection previously known as "PowerPC Built-in Functions".
10967 (Basic PowerPC Built-in Functions Available on all Configurations):
10969 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10971 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10973 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10975 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10978 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10981 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
10982 after cmpelim optimization.
10984 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
10986 * config.gcc: Support "goldmont".
10987 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10988 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10989 PROCESSOR_GOLDMONT.
10990 * config/i386/i386.c (m_GOLDMONT): Define.
10991 (processor_target_table): Add "goldmont".
10992 (PTA_GOLDMONT): Define.
10993 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10994 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10995 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10996 (fold_builtin_cpu): Add "goldmont".
10997 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10998 (ix86_option_override_internal): Add "goldmont".
10999 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11000 (processor_type): Add PROCESSOR_GOLDMONT.
11001 * config/i386/i386.md: Add CPU "glm".
11002 * config/i386/glm.md: New file.
11003 * config/i386/x86-tune.def: Add m_GOLDMONT.
11004 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11006 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11009 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11011 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11012 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11013 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11016 * config/i386/i386.c (ix86_fold_builtin): Handle
11017 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11020 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11021 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11023 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11026 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11027 (check_arch): Likewise.
11028 (check_fpu): Return the result rather than printing it.
11029 (end arch): Fix operator precedence.
11030 (end cpu): Likewise.
11031 (END): Print the result from check_fpu.
11033 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11034 Alan Hayward <alan.hayward@arm.com>
11035 David Sherwood <david.sherwood@arm.com>
11037 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11038 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11039 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11040 (*fcmuo<mode>_and): New patterns.
11042 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11044 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11045 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11046 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11047 (cmp_op, sve_imm_con): New code attributes.
11048 (SVE_COND_INT_CMP, imm_con): Delete.
11049 (cmp_op): Remove above unspecs from int attribute.
11050 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11052 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11053 comparison-specific unspecs.
11054 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11055 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11056 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11057 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11058 (*vec_fcm<cmp_op><mode>): Rename to...
11059 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11060 (*vec_fcmuo<mode>): Rename to...
11061 (*fcmuo<mode>): ...this and adjust likewise.
11062 (*pred_fcm<cmp_op><mode>): New pattern.
11063 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11064 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11066 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11068 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11069 (aarch64_emit_sve_predicated_cond): New function.
11070 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11071 (aarch64_emit_unspec_cond_or): Replace with...
11072 (aarch64_emit_sve_or_conds): ...this new function. Use
11073 aarch64_emit_sve_ptrue_op for the individual comparisons and
11074 aarch64_emit_binop to OR them together.
11075 (aarch64_emit_inverted_unspec_cond): Replace with...
11076 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11077 aarch64_emit_sve_ptrue_op for the comparison and
11078 aarch64_emit_unop to invert the result.
11079 (aarch64_expand_sve_vec_cmp_float): Update after the above
11080 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11082 2018-05-07 Nathan Sidwell <nathan@acm.org>
11084 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11085 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11086 (Backwards Compatibility): Likewise.
11088 2018-05-07 Luis Machado <luis.machado@linaro.org>
11092 2018-05-07 Luis Machado <luis.machado@linaro.org>
11094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11095 <prefetch_dynamic_strides>: New const bool field.
11096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11097 prefetch_dynamic_strides.
11098 (exynosm1_prefetch_tune): Likewise.
11099 (thunderxt88_prefetch_tune): Likewise.
11100 (thunderx_prefetch_tune): Likewise.
11101 (thunderx2t99_prefetch_tune): Likewise.
11102 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11104 (aarch64_override_options_internal): Update to set
11105 PARAM_PREFETCH_DYNAMIC_STRIDES.
11106 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11107 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11108 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11109 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11110 prefetch-dynamic-strides setting.
11112 2018-05-07 Luis Machado <luis.machado@linaro.org>
11114 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11115 <minimum_stride>: New const int field.
11116 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11117 minimum_stride field.
11118 (exynosm1_prefetch_tune): Likewise.
11119 (thunderxt88_prefetch_tune): Likewise.
11120 (thunderx_prefetch_tune): Likewise.
11121 (thunderx2t99_prefetch_tune): Likewise.
11122 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11123 (aarch64_override_options_internal): Update to set
11124 PARAM_PREFETCH_MINIMUM_STRIDE.
11125 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11126 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11127 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11128 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11129 stride is constant and is below the minimum stride threshold.
11131 2018-05-07 Luis Machado <luis.machado@linaro.org>
11133 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11136 2018-05-07 Luis Machado <luis.machado@linaro.org>
11138 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11139 <prefetch_dynamic_strides>: New const bool field.
11140 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11141 prefetch_dynamic_strides.
11142 (exynosm1_prefetch_tune): Likewise.
11143 (thunderxt88_prefetch_tune): Likewise.
11144 (thunderx_prefetch_tune): Likewise.
11145 (thunderx2t99_prefetch_tune): Likewise.
11146 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11148 (aarch64_override_options_internal): Update to set
11149 PARAM_PREFETCH_DYNAMIC_STRIDES.
11150 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11151 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11152 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11153 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11154 prefetch-dynamic-strides setting.
11156 2018-05-07 Luis Machado <luis.machado@linaro.org>
11158 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11159 <minimum_stride>: New const int field.
11160 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11161 minimum_stride field.
11162 (exynosm1_prefetch_tune): Likewise.
11163 (thunderxt88_prefetch_tune): Likewise.
11164 (thunderx_prefetch_tune): Likewise.
11165 (thunderx2t99_prefetch_tune): Likewise.
11166 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11167 (aarch64_override_options_internal): Update to set
11168 PARAM_PREFETCH_MINIMUM_STRIDE.
11169 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11170 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11171 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11172 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11173 stride is constant and is below the minimum stride threshold.
11175 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11178 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11179 the type is addressable. Don't force op into register if it has
11182 2018-05-05 Roland McGrath <mcgrathr@google.com>
11185 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11186 any section for which we don't know a specific type it should have,
11187 regardless of name. Previously this was done only for the exact
11188 names ".init_array", ".fini_array", and ".preinit_array".
11189 (default_elf_asm_named_section): Add comment about
11190 relationship with default_section_type_flags and SECTION_NOTYPE.
11191 (get_section): Don't consider it a type conflict if one side has
11192 SECTION_NOTYPE and the other doesn't, as long as neither has the
11193 SECTION_BSS et al used in the default_section_type_flags logic.
11195 2018-05-05 Tom de Vries <tom@codesourcery.com>
11198 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11199 (workaround_barsyncs): New function.
11200 (nvptx_reorg): Use workaround_barsyncs.
11201 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11202 (define_expand "nvptx_membar_cta"): New define_expand.
11203 (define_insn "*nvptx_membar_cta"): New insn.
11205 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11207 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11208 To improve optimization opportunities.
11209 * builtin-types.def: The new needed builtin types for the above.
11211 2018-05-04 Richard Biener <rguenther@suse.de>
11213 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11214 * gimple-ssa-store-merging.c
11215 (imm_store_chain_info::output_merged_store): Remove redundant create,
11216 release split_store vector contents on failure.
11217 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11218 scalar stmt vector on cache hit.
11220 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11222 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11224 * config.gcc (powerpc-xilinx-eabi*): Remove.
11225 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11227 (fusion_addis_mem_combo_load): Ditto.
11228 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11230 (rs6000_cpu_cpp_builtins): Ditto.
11231 * config/rs6000/rs6000-linux.c
11232 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11233 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11234 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11236 (rs6000_setup_reg_addr_masks): Ditto.
11237 (rs6000_init_hard_regno_mode_ok): Ditto.
11238 (rs6000_option_override_internal): Ditto.
11239 (legitimate_lo_sum_address_p): Ditto.
11240 (rs6000_legitimize_address): Ditto.
11241 (rs6000_legitimize_reload_address): Ditto.
11242 (rs6000_legitimate_address_p): Ditto.
11243 (abi_v4_pass_in_fpr): Ditto.
11244 (setup_incoming_varargs): Ditto.
11245 (rs6000_gimplify_va_arg): Ditto.
11246 (rs6000_split_multireg_move): Ditto.
11247 (rs6000_savres_strategy): Ditto.
11248 (rs6000_emit_prologue_components): Ditto.
11249 (rs6000_emit_epilogue_components): Ditto.
11250 (rs6000_emit_prologue): Ditto.
11251 (rs6000_emit_epilogue): Ditto.
11252 (rs6000_elf_file_end): Ditto.
11253 (rs6000_function_value): Ditto.
11254 (rs6000_libcall_value): Ditto.
11255 * config/rs6000/rs6000.h: Ditto.
11256 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11257 (TARGET_MINMAX): ... this. New.
11258 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11259 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11260 (*movsi_internal1_single): Delete.
11261 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11262 mfpu=, mxilinx-fpu): Delete.
11263 * config/rs6000/singlefp.h: Delete.
11264 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11265 * config/rs6000/t-rs6000: Ditto.
11266 * config/rs6000/t-xilinx: Delete.
11267 * config/rs6000/titan.md: Adjust for fp_type removal.
11268 * config/rs6000/vsx.md: Remove Xilinx FP support.
11269 (VStype_simple): Delete.
11270 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11271 * config/rs6000/xfpu.h: Delete.
11272 * config/rs6000/xfpu.md: Delete.
11273 * config/rs6000/xilinx.h: Delete.
11274 * config/rs6000/xilinx.opt: Delete.
11275 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11276 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11278 2018-05-04 Tom de Vries <tom@codesourcery.com>
11281 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11284 2018-05-04 Richard Biener <rguenther@suse.de>
11286 PR middle-end/85627
11287 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11288 (expand_complex_div_wide): Likewise.
11289 (expand_complex_operations_1): Likewise.
11290 (expand_complex_libcall): Preserve EH info of the original stmt.
11291 (tree_lower_complex): Handle removed blocks.
11292 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11293 on complex multiplication and division libcall builtins.
11295 2018-05-04 Richard Biener <rguenther@suse.de>
11297 PR middle-end/85574
11298 * fold-const.c (negate_expr_p): Restrict negation of operand
11299 zero of a division to when we know that can happen without
11301 (fold_negate_expr_1): Likewise.
11303 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11306 * real.h (real_nextafter): Declare.
11307 * real.c (real_nextafter): New function.
11308 * fold-const-call.c (fold_const_nextafter): New function.
11309 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11310 CASE_CFN_NEXTTOWARD.
11311 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11312 even when arg1_mode is different from arg0_mode.
11314 2018-05-03 Nathan Sidwell <nathan@acm.org>
11316 * doc/extend.texi (Deprecated Features): Remove
11317 -ffriend-injection.
11318 (Backwards Compatibility): Likewise.
11319 * doc/invoke.texi (C++ Language Options): Likewise.
11320 (C++ Dialect Options): Likewise.
11322 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11325 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11326 _mm512_mask_mullox_epi64): New intrinsics.
11328 2018-05-03 Tom de Vries <tom@codesourcery.com>
11331 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11332 dump files): Add offload-tree.
11334 2018-05-03 Richard Biener <rguenther@suse.de>
11336 PR tree-optimization/85615
11337 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11338 to loops not nested in BBs loop father to avoid creating multi-entry
11341 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11343 PR tree-optimization/70291
11344 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11345 arguments. Change return type to tree. Emit libcall as a new
11346 statement rather than replacing existing one when inplace_p is true.
11347 (expand_complex_multiplication_components): New function.
11348 (expand_complex_multiplication): Expand floating-point complex
11349 multiplication using the above.
11350 (expand_complex_division): Rename inner_type parameter to type.
11351 Update expand_complex_libcall call-site.
11352 (expand_complex_operations_1): Update expand_complex_multiplication
11353 and expand_complex_division call-sites.
11355 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11358 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11359 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11360 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11361 the highest significant bit of the shift count mask is clear. In
11362 check whether and[sq]i3 is needed verify that all significant bits
11363 of the shift count other than the highest are set.
11365 2018-05-02 Tom de Vries <tom@codesourcery.com>
11368 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11369 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11370 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11371 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11372 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11373 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11374 __builtin_goacc_parlevel_size.
11376 2018-05-02 Richard Biener <rguenther@suse.de>
11378 PR tree-optimization/85597
11379 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11380 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11383 2018-05-02 Tom de Vries <tom@codesourcery.com>
11386 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11387 dump files): Add ltrans-tree.
11389 2018-05-02 Tom de Vries <tom@codesourcery.com>
11392 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11393 dump files): Add wpa-ipa.
11395 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11397 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11398 powerpc*-*-linux*paired* target.
11399 * config/rs6000/750cl.h: Delete.
11400 * config/rs6000/paired.h: Delete.
11401 * config/rs6000/paired.md: Delete.
11402 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11404 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11405 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11406 comment. Remove paired float support.
11407 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11408 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11410 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11411 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11413 * config/rs6000/rs6000.c: Remove paired float support.
11414 (paired_expand_vector_init, paired_expand_vector_move,
11415 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11416 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11417 paired_expand_builtin, paired_expand_predicate_builtin,
11418 paired_init_builtins): Delete.
11419 * config/rs6000/rs6000.h: Remove paired float support.
11420 * config/rs6000/rs6000.md: Remove paired float support.
11421 (move_from_CR_ov_bit): Delete.
11422 * config/rs6000/rs6000.opt (mpaired): Delete.
11423 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11424 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11426 2018-05-02 Richard Biener <rguenther@suse.de>
11428 PR middle-end/85567
11429 * gimplify.c (gimplify_save_expr): When in SSA form allow
11430 SAVE_EXPRs to compute to SSA vars.
11432 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11435 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11436 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11437 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11438 clobber operands[2], instead use a new pseudo. Formatting fixes.
11440 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11442 PR tree-optimization/85586
11443 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11444 exit early for statements in the same group if the accesses are
11447 2018-05-02 Tom de Vries <tom@codesourcery.com>
11450 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11453 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11455 PR tree-optimization/85143
11456 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11458 2018-05-01 Tom de Vries <tom@codesourcery.com>
11461 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11462 not found" error message.
11464 2018-05-01 Tom de Vries <tom@codesourcery.com>
11467 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11468 * vec.c (test_ordered_remove_if): New function.
11469 (vec_c_tests): Call test_ordered_remove_if.
11470 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11471 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11472 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11473 VEC_ORDERED_REMOVE_IF.
11475 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11477 PR tree-optimization/82665
11478 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11479 pointer subtraction where arguments come from a memchr call.
11481 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11483 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11484 --push-state --as-needed and --pop-state instead of --as-needed and
11485 --no-as-needed if ld supports it.
11486 * configure: Regenerated.
11489 * doc/install.texi2html: Replace _002d with - and _002a with * in
11490 generated html files using sed.
11492 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11495 * gcc-rich-location.c (blank_line_before_p): New function.
11496 (use_new_line): New function.
11497 (gcc_rich_location::add_fixit_insert_formatted): New function.
11498 * gcc-rich-location.h
11499 (gcc_rich_location::add_fixit_insert_formatted): New function.
11501 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11503 * selftest.c (assert_streq): Rename "expected" and "actual" to
11504 "val1" and "val2". Extend NULL-handling to cover both inputs
11505 symmetrically, while still requiring both to be non-NULL for a pass.
11506 * selftest.h (assert_streq): Rename "expected" and "actual" to
11508 (ASSERT_EQ): Likewise.
11509 (ASSERT_EQ_AT): Likewise.
11510 (ASSERT_KNOWN_EQ): Likewise.
11511 (ASSERT_KNOWN_EQ_AT): Likewise.
11512 (ASSERT_NE): Likewise.
11513 (ASSERT_MAYBE_NE): Likewise.
11514 (ASSERT_MAYBE_NE_AT): Likewise.
11515 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11516 the assertion to pass.
11517 (ASSERT_STREQ_AT): Likewise.
11519 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11521 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11522 interaction with -pie.
11524 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11526 * selftest.h: Fix alphabetization of per-source-file selftest
11529 2018-04-30 Jason Merrill <jason@redhat.com>
11531 PR c++/61982 - dead stores to destroyed objects.
11532 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11535 2018-04-30 Jason Merrill <jason@redhat.com>
11537 * tree.c (build_clobber): New.
11538 * tree.h: Declare it.
11539 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11541 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11543 * diagnostic-show-locus.c (layout::layout): Update for
11544 location_get_source_line returning a char_span.
11545 (struct char_span): Move to input.h.
11546 (struct correction): Update for fields in char_span becoming
11548 (struct source_line): Update for location_get_source_line
11549 returning a char_span.
11550 (layout::print_line): Likewise.
11551 * edit-context.c (edited_file::print_content): Likewise.
11552 (edited_file::print_diff_hunk): Likewise.
11553 (edited_file::print_run_of_changed_lines): Likewise.
11554 (edited_file::get_num_lines): Likewise.
11555 (edited_line::edited_line): Likewise.
11556 * final.c (asm_show_source): Likewise.
11557 * input.c (location_get_source_line): Convert return type
11558 from const char * to char_span, losing the final "line_len"
11560 (dump_location_info): Update for the above.
11561 (get_substring_ranges_for_loc): Likewise. Use a char_span
11562 when handling the literal within the line.
11563 (test_reading_source_line): Update for location_get_source_line
11564 returning a char_span.
11565 * input.h (class char_span): Move here from
11566 diagnostic-show-locus.c, converting from a struct to a class.
11567 Make data members private.
11568 (char_span::operator bool): New.
11569 (char_span::length): New.
11570 (char_span::get_buffer): New.
11571 (char_span::operator[]): New.
11572 (char_span::subspan): Make const.
11573 (char_span::xstrdup): New.
11574 (location_get_source_line): Convert return type from const char *
11575 to char_span, losing the final "line_size" param.
11577 2018-04-30 Jan Hubicka <jh@suse.cz>
11579 * lto-wrapper.c (ltrans_priorities): New static var.
11580 (cmp_priority): New.
11581 (run_gcc): Read priorities and if doing parallel build order
11582 the Makefile by them.
11584 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11586 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11588 2018-04-30 Richard Biener <rguenther@suse.de>
11590 * tree-cfg.c (verify_address): Remove base argument, add
11591 flag whether to check TREE_ADDRESSABLE and do that.
11592 (verify_expr): Remove.
11593 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11594 (verify_gimple_assign_single): Likewise.
11595 (verify_gimple_switch): Likewise.
11596 (verify_expr_location_1): Dereference tp once. Add (disabled)
11597 piece from verify_expr.
11598 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11600 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11602 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11604 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11606 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11607 (small_data_pattern): Likewise.
11608 (arc_rewrite_small_data): Likewise.
11609 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11610 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11611 (get_symbol_alignment): New function.
11612 (legitimate_small_data_address_p): Likewise.
11613 (legitimate_scaled_address): Update, call
11614 legitimate_small_data_address_p.
11615 (output_sdata): New static variable.
11616 (arc_print_operand): Update how we handle small data operands.
11617 (arc_print_operand_address): Likewise.
11618 (arc_legitimate_address_p): Update, use
11619 legitimate_small_data_address_p.
11620 (arc_rewrite_small_data_p): Remove.
11621 (arc_rewrite_small_data_1): Likewise.
11622 (arc_rewrite_small_data): Likewise.
11623 (small_data_pattern): Likewise.
11624 (compact_sda_memory_operand): Update to use
11625 legitimate_small_data_address_p and get_symbol_alignment.
11626 (prepare_move_operands): Don't rewite sdata pattern.
11627 (prepare_extend_operands): Remove.
11628 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11630 (zero_extendqisi2): Likewise.
11631 (zero_extendhisi2): Likewise.
11632 (extendqihi2): Likewise.
11633 (extendqisi2): Likewise.
11634 (extendhisi2): Likewise.
11635 (addsi3): Likewise.
11636 (subsi3): Likewise.
11637 (andsi3): Likewise.
11638 * config/arc/constraints.md (Usd): Change it to memory constraint.
11640 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11642 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11643 as source of std instructions.
11644 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11645 allow 6-bit constants as source for store instructions.
11646 (movdi_insn): Update instruction pattern to allow 6-bit constants
11647 as source for store instructions.
11649 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11651 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11653 2018-04-30 Nathan Sidwell <nathan@acm.org>
11654 Sandra Loosemore <sandra@codesourcery.com>
11656 * dumpfile.c (dump_open): Allow '-' for stdout.
11657 * doc/invoke.texi (Developer Options): Document dump filename
11658 determination early. Document stdin/stdout selection.
11660 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11662 Microblaze Target: PIC data text relative
11664 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11665 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11667 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11668 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11669 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11670 New addressing mode for data-text relative position indepenedent code.
11671 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11672 'ADDRESS_SYMBOLIC_TXT_REL'.
11673 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11674 (microblaze_legitimate_pic_operand): Exclude function calls from
11675 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11676 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11678 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11679 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11680 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11681 for 'address + offset'.
11682 (microblaze_expand_prologue): Add new function prologue call for
11684 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11685 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11686 table in case of TARGET_PIC_DATA_TEXT_REL.
11687 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11688 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11689 Add new macros 'UNSPEC_TEXT',
11690 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11691 + exclude function calls from 'UNSPEC_PLT' in case of data text
11693 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11694 new target hook for generating address diff vector tables in case of
11696 * doc/tm.texi : Regenerate.
11697 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11698 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11699 of addr diff vector generation.
11700 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11701 target hook definition.
11702 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11703 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11704 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11705 option for data text relative.
11707 2018-04-30 Richard Biener <rguenther@suse.de>
11709 * tree-chrec.h (evolution_function_is_constant_p): Remove
11711 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11713 2018-04-30 Richard Biener <rguenther@suse.de>
11716 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11718 2018-04-30 Richard Biener <rguenther@suse.de>
11720 PR tree-optimization/28364
11721 PR tree-optimization/85275
11722 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11723 copying first exit test.
11725 2018-04-28 Mark Wielaard <mark@klomp.org>
11727 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11728 dwarf_version >= 5.
11729 (dwarf_AT): Handle DW_AT_addr_base.
11730 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11732 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11735 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11736 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11737 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11738 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11740 2018-04-28 Richard Biener <rguenther@suse.de>
11742 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11743 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11744 to reflect use. Only add interesting stmts.
11746 2018-04-27 Martin Jambor <mjambor@suse.cz>
11749 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11750 the jump function allows for passing through aggregate values.
11752 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11754 * input.h (in_system_header_at): Convert from macro to inline
11756 (from_macro_expansion_at): Likewise.
11757 (from_macro_definition_at): Likewise.
11759 2018-04-27 Jeff Law <law@redhat.com>
11761 * config.gcc: Mark tile* targets as deprecated/obsolete.
11763 2018-04-27 Richard Biener <rguenther@suse.de>
11765 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11768 2018-04-27 Richard Biener <rguenther@suse.de>
11770 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11772 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11774 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11775 with Yd constraint. Set "preferred_for_speed" attribute from
11776 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11777 with Yd constraint.
11778 (*movdi_internal): Ditto.
11779 (movti_interunit splitters): Remove
11780 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11781 (movdi_interunit splitters): Ditto.
11782 * config/i386/constraints.md (Ye): Remove.
11783 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11785 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11788 * config/aarch64/constraints.md (Usg): Limit to 31.
11789 (Usj): Limit to 63.
11791 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11793 PR tree-optimization/85529
11794 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11795 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11796 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11797 zero extension or masking of the MSB bit.
11798 (optimize_range_tests): Add FIRST_BB argument, pass it through
11799 to optimize_range_tests_var_bound.
11800 (maybe_optimize_range_tests, reassociate_bb): Adjust
11801 optimize_range_tests callers.
11803 2018-04-26 Richard Biener <rguenther@suse.de>
11804 Jakub Jelinek <jakub@redhat.com>
11806 * cgraph.h (symbol_table): Just declare debug method here.
11807 * symtab.c (symbol_table::debug): Define.
11809 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11811 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11813 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11815 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11816 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11817 (*movdi_internal): Substitute Yi and Yj constraint with x
11818 and Ym and Yn constraint with y constraint. Update "isa"
11819 attribute and set "preferred_for_speed" attribute from
11820 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11821 (*movsi_internal): Ditto.
11822 (*movdf_internal): Ditto.
11823 (*movsf_internal): Ditto.
11824 (*zero_extendsidi2): Ditto.
11825 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11826 (sse2_loadld): Ditto.
11827 (*vec_extract<ssevecmodelower>_0): Ditto.
11828 (*vec_extractv4si_0_zext_sse4): Ditto.
11829 (vec_concatv2di): Ditto.
11830 (*vec_dup<mode>): Ditto.
11831 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11832 * config/i386/constraints.md (Yi): Remove.
11837 2018-04-26 Nathan Sidwell <nathan@acm.org>
11839 * dumpfile.c (dump_open): New.
11840 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11841 (dump_finish): Detect stdio/stderr by value not name.
11843 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11845 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11847 2018-04-26 Tom de Vries <tom@codesourcery.com>
11850 * config/nvptx/nvptx.c (verify_neutering_jumps)
11851 (verify_neutering_labels): New function
11852 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11854 2018-04-26 Tom de Vries <tom@codesourcery.com>
11857 * config/nvptx/nvptx.c (needs_neutering_p): New function.
11858 (nvptx_single): Use needs_neutering_p to skip over insns that do not
11861 2018-04-26 Richard Biener <rguenther@suse.de>
11862 Tom de Vries <tom@codesourcery.com>
11865 * lto-streamer-out.c (output_function): Fixup loops if required to match
11866 discovery done in the reader.
11868 2018-04-26 Richard Biener <rguenther@suse.de>
11870 PR tree-optimization/85116
11871 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11872 have a loop exit from the single latch predecessor. Remove
11873 case of header with just condition.
11874 (ch_base::copy_headers): Exclude infinite loops from any
11876 (pass_ch::execute): Record exits.
11878 2018-04-26 Richard Biener <rguenther@suse.de>
11880 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11881 prologue cost vector and pass it to vect_get_load_cost.
11882 (vect_get_peeling_costs_all_drs): Likewise.
11883 (vect_peeling_hash_get_lowest_cost): Likewise.
11884 (vect_enhance_data_refs_alignment): Likewise.
11886 2018-04-26 Richard Biener <rguenther@suse.de>
11888 PR middle-end/85450
11889 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11890 checking of integer<->pointer conversions.
11891 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11892 sign-/zero-extending pointer types.
11893 (expand_omp_for_static_chunk): Likewise.
11895 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
11896 Jean Lee <xiaoyur347@gmail.com>
11898 * config/mips/mips.c (mips_asan_shadow_offset): New function.
11899 (TARGET_ASAN_SHADOW_OFFSET): Define.
11900 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11901 true for -fsanitize=address.
11903 2018-04-25 Mark Wielaard <mark@klomp.org>
11905 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11908 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11910 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11911 than "alu", remove explicit "memory" and "imm_disp" attributes.
11912 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11914 PR middle-end/85414
11915 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11916 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11917 gen_lowpart_no_emit.
11919 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
11922 * config/i386/i386.c (ix86_expand_builtin): Change memory
11923 operand to XI, extend p0 to Pmode.
11924 * config/i386/i386.md: Change unspec volatile and operand
11925 1 mode to XI, change operand 0 mode to P.
11927 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11929 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11930 GET_MODE_MASK before any checking.
11931 (nds32_can_use_bset_p): Likewise.
11932 (nds32_can_use_btgl_p): Likewise.
11934 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11936 * config/nds32/nds32-doubleword.md: New define_split pattern for
11937 illegal register number.
11939 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11941 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11943 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11945 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11947 2018-04-25 Richard Biener <rguenther@suse.de>
11949 * lto-streamer.h (LTO_major_version): Bump to 8.
11951 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11953 * BASE-VER: Set to 9.0.0.
11955 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
11957 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11958 in __abskf2 and __powikf2.
11960 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11963 * config/aarch64/constraints.md (Usg, Usj): New constraints.
11964 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11965 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
11966 Use the above on operand 2. Reindent.
11967 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
11969 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11972 * common/config/i386/i386-common.c (ix86_handle_option): Don't
11974 * config/i386/i386.opt (mcet): Removed.
11975 * doc/install.texi: Remove -mcet documentation.
11976 * doc/invoke.texi: Likewise.
11978 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11981 * doc/install.texi: Remove -mcet from bootstrap-cet.
11983 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11986 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11987 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11991 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11992 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11993 containing a CONST_VECTOR.
11995 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
11997 * doc/install.texi: Update newlib dependency for nvptx.
11999 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12002 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12003 instead of INTVAL when shifting x left.
12005 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12007 PR tree-optimization/85478
12008 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12009 vect_grouped_store_supported for single element vectors.
12011 2018-04-24 Richard Biener <rguenther@suse.de>
12014 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12015 load cost increase to the case of non-constant step.
12017 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12020 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12021 destination if any_malformed_asm.
12023 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12025 PR middle-end/85496
12026 * expr.c (store_field): In the bitfield case, if the value comes from
12027 a function call and is returned in registers by means of a PARALLEL,
12028 do not change the mode of the temporary unless BLKmode and VOIDmode.
12030 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12032 PR rtl-optimization/85423
12033 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12034 dependencies to debug insns when the previous insn is non-debug.
12036 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12038 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12039 enums into a single definition.
12040 (fls): Fix predicates and printing.
12043 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12045 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12046 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12047 and short u6 immediate.
12048 (check_if_valid_sleep_operand): Remove.
12049 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12051 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12053 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12054 flag_always_save_lp condition.
12055 * config/nds32/nds32.opt (malways-save-lp): New option.
12057 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12059 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12060 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12061 * config/nds32/nds32.h
12062 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12063 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12065 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12067 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12068 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12070 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12071 Chung-Ju Wu <jasonwucj@gmail.com>
12073 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12075 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12076 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12078 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12080 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12082 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12084 * config/nds32/nds32-protos.h (nds32_data_alignment,
12085 nds32_local_alignment): Declare.
12086 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12087 nds32_local_alignment): New functions.
12088 (TARGET_CONSTANT_ALIGNMENT): Define.
12089 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12091 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12093 * config/nds32/nds32.c
12094 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12095 (TARGET_MODES_TIEABLE_P): Likewise.
12097 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12099 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12100 level Ofast and Og.
12102 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12103 Chung-Ju Wu <jasonwucj@gmail.com>
12105 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12106 for unaligned access.
12107 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12109 * config/nds32/nds32-intrinsic.md: Likewise.
12110 * config/nds32/nds32_intrinsic.h: Likewise.
12111 * config/nds32/nds32.h (nds32_builtins): Likewise.
12112 * config/nds32/nds32.opt (munaligned-access): New option.
12113 * config/nds32/nds32.c (nds32_asm_file_start): Display
12114 flag_unaligned_access status.
12116 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12118 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12119 -mno-relax is present.
12120 * config/riscv/linux.h (LINK_SPEC): Ditto.
12122 2018-04-20 Martin Sebor <msebor@redhat.com>
12125 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12127 (gimple_fold_builtin_stxcpy_chk): Same.
12128 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12130 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12133 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12134 __powikf2 when long double is IEEE 128-bit.
12136 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12138 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12139 step to make sure stack always aligned.
12141 2018-04-20 Carl Love <cel@us.ibm.com>
12144 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12145 size check for arg0.
12147 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12148 Tom de Vries <tom@codesourcery.com>
12151 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12152 Emit insns for calls too.
12153 (nvptx_find_par): Always look for worker-level predecessor insn.
12154 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12156 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12157 (nvptx_process_pars): Propagate frames for calls.
12159 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12162 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12164 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12165 (ix86_handle_option): Don't handle OPT_mibt.
12166 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12168 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12170 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12171 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12172 (ix86_target_macros): Define __CET__ with flag_cf_protection
12173 for -fcf-protection.
12174 * config/i386/i386.c (isa2_opts): Remove -mibt.
12175 * config/i386/i386.h (TARGET_IBT): Removed.
12176 (TARGET_IBT_P): Likewise.
12177 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12178 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12179 * config/i386/i386.opt (mcet): Update help message.
12180 (mshstk): Likewise.
12182 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12183 -mcet as an alias for -mshstk.
12185 2018-04-20 Richard Biener <rguenther@suse.de>
12187 PR middle-end/85475
12188 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12189 complexity by forcing a single use of the multiply operand.
12191 2018-04-20 Martin Jambor <mjambor@suse.cz>
12194 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12195 recursion dependency to only apply to non-clones.
12197 2018-04-20 Martin Jambor <mjambor@suse.cz>
12200 * ipa-cp.c (create_specialized_node): Check that clones of
12201 self-recursive edges exist during IPA-CP.
12203 2018-04-19 Toon Moene <toon@moene.org>
12205 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12208 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12210 PR tree-optimization/85467
12211 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12212 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12213 VECTOR_CST element to type.
12215 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12218 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12219 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12220 (builtin_longjmp): Likewise.
12221 (save_stack_nonlocal): New pattern.
12222 (restore_stack_nonlocal): Likewise.
12224 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12227 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12228 Replace ASM_OUTPUT_LABEL with fprintf.
12230 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12233 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12234 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12235 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12236 define __IBT__ and __SHSTK__ for -fcf-protection.
12237 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12239 (ix86_trampoline_init): Likewise.
12240 (x86_output_mi_thunk): Likewise.
12241 (ix86_notrack_prefixed_insn_p): Likewise.
12242 (ix86_option_override_internal): Don't disallow -fcf-protection.
12243 * config/i386/i386.md (rdssp<mode>): Also enable for
12245 (incssp<mode>): Likewise.
12246 (nop_endbr): Likewise.
12247 * config/i386/i386.opt (mcet): Change help message to built-in
12250 (mshstk): Likewise.
12251 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12252 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12253 enable CET built-in functions.
12255 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12257 * common/config/i386/i386-common.c
12258 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12259 OPTION_MASK_ISA_MOVDIRI_UNSET,
12260 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12261 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12262 * config.gcc (movdirintrin.h): New header.
12263 * config/i386/cpuid.h (bit_MOVDIRI,
12264 bit_MOVDIR64B): New bits.
12265 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12267 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12268 (VOID, PVOID, PCVOID)): New function types.
12269 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12270 __builtin_ia32_directstoreu_u64,
12271 __builtin_ia32_movdir64b): New builtins.
12272 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12273 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12275 (ix86_valid_target_attribute_inner_p): Ditto.
12276 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12277 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12278 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12279 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12280 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12281 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12282 (movdiri<mode>, movdir64b_<mode>): New.
12283 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12284 * config/i386/immintrin.h: Include movdirintrin.h.
12285 * config/i386/movdirintrin.h: New file.
12286 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12288 2018-04-19 Richard Biener <rguenther@suse.de>
12290 PR middle-end/85455
12291 * cfg.c (clear_bb_flags): When loop state says we have
12292 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12294 2018-04-19 Richard Biener <rguenther@suse.de>
12296 PR tree-optimization/84737
12297 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12298 copying restrict info.
12299 (vect_setup_realignment): Use it.
12300 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12301 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12302 the first DR to all generated stores.
12303 (vectorizable_load): Likewise for loads.
12305 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12307 PR tree-optimization/85446
12308 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12309 the integral and pointer types to have the same precision.
12311 * doc/install.texi: Document --disable-cet being the default and
12314 2018-04-18 Martin Liska <mliska@suse.cz>
12316 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12319 2018-04-18 Martin Liska <mliska@suse.cz>
12322 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12325 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12326 arguments if they are comparable.
12328 2018-04-18 Martin Liska <mliska@suse.cz>
12331 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12334 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12337 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12340 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12341 ENDBR after calling __morestack.
12343 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12346 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12347 by using gcc_base_ver to generate a gcc_driver_version, and use
12348 it when generating GCC_DRIVER_NAME.
12349 * configure: Regenerate.
12351 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12354 * config.gcc: Obsolete powerpc*-*-*spe*.
12356 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12359 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12360 (stabstr_D): Change type of unum from unsigned int to
12361 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12364 2018-04-17 Jim Wilson <jimw@sifive.com>
12367 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12368 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12369 Set arg_pointer_offset after using pretend_args_size.
12371 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12373 PR rtl-optimization/85431
12374 * dse.c (record_store): Ignore zero width stores.
12377 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12378 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12379 __builtin_stack_restore rather than after it.
12380 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12381 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12382 argument instead of virtual_dynamic_stack_rtx.
12384 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12386 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12388 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12389 Add note to error message to explain internal mapping of overloaded
12390 built-in function name to non-overloaded built-in function name.
12391 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12394 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12397 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12398 where the inputs overlap with the output.
12400 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12402 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12403 (=v, v) alternative and explicit "memory" attribute.
12404 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12405 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12407 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12408 "sselog1" type instead of "sselog".
12409 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12410 "sselog". Remove explicit "memory" attribute.
12411 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12412 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12414 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12415 "sselog1" type instead of "sselog". Remove explicit "memory"
12417 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12418 use "sselog1" type instead of "sselog". Remove explicit "memory"
12420 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12421 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12423 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12424 "sselog1" type instead of "sselog". Remove explicit "memory"
12426 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12427 use "sselog1" type instead of "sselog". Remove explicit "memory"
12431 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12433 PR middle-end/85414
12434 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12437 2018-04-17 Martin Jambor <mjambor@suse.cz>
12440 * ipa-cp.c (create_specialized_node): Call
12441 expand_all_artificial_thunks if necessary.
12443 2018-04-17 Martin Liska <mliska@suse.cz>
12446 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12447 in message, remote space in between '_G' and '('.
12449 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12452 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12453 avx512f_vmcmp<mode>3<round_saeonly_name>,
12454 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12455 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12456 avx512f_rndscale<mode><round_saeonly_name>,
12457 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12458 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12459 Use %<iptr>2 instead of %2 for -masm=intel.
12460 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12461 avx512f_vcvttss2usi<round_saeonly_name>,
12462 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12464 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12465 avx512f_vcvttsd2usi<round_saeonly_name>,
12466 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12467 Use %q1 instead of %1 for -masm=intel.
12468 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12469 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12470 of %3 for -masm=intel.
12471 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12473 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12475 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12477 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12478 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12480 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12482 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12483 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12484 %g1 and one with %0 and %1.
12485 (avx512er_vmrcp28<mode><round_saeonly_name>,
12486 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12487 %1 for -masm=intel.
12488 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12489 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12490 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12491 of %0 and %{%4%} for -masm=intel.
12492 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12493 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12494 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12495 order of %0 and %{%5%}%{z%} for -masm=intel.
12497 2018-04-17 Jan Hubicka <jh@suse.cz>
12500 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12502 2018-04-17 Martin Liska <mliska@suse.cz>
12505 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12506 for target_clone error message. Make default implementation
12507 clone to be a local declaration.
12508 (separate_attrs): Add new argument and check for an empty
12510 (expand_target_clones): Handle it.
12511 (ipa_target_clone): Make redirection just for target_clones
12514 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12515 Tom de Vries <tom@codesourcery.com>
12517 PR middle-end/84955
12518 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12519 tiled basic blocks without omp continue statements.
12521 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12524 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12525 vec_extract expression as having side effects to make sure it gets
12528 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12531 * config/i386/i386.c (get_builtin_code_for_version): Check
12534 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12537 * config.gcc: Support "skylake".
12538 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12540 * config/i386/i386.c (m_SKYLAKE): Define.
12541 (processor_target_table): Add "skylake".
12542 (ix86_option_override_internal): Add "skylake".
12543 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12544 PROCESSOR_CANNONLAKE.
12545 (get_builtin_code_for_version): Fix priority for
12546 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12547 PROCESSOR_SKYLAKE-AVX512.
12548 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12549 (processor_type): Add PROCESSOR_SKYLAKE.
12551 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12552 Jason Merrill <jason@redhat.com>
12555 * convert.c (convert_to_integer_1): Use direct recursion for
12556 enumeral types and types with a precision less than the number
12557 of bits in their mode.
12559 2018-04-16 Julia Koval <julia.koval@intel.com>
12562 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12563 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12565 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12568 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12569 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12570 and -mno-direct-move.
12572 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12575 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12576 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12577 Ensure negative shifts result in {0}.
12579 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12581 PR rtl-optimization/79916
12582 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12583 regs (if any) to define how to gnerate SD moves when LRA is in
12586 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12588 PR rtl-optimization/85393
12589 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12590 * except.c (expand_dw2_landing_pad_for_region): Make static.
12591 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12592 a label and unconditional jump to old_bb, rather than
12593 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12596 PR rtl-optimization/85376
12597 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12598 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12599 instead of a specific value.
12601 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12602 Bin Cheng <bin.cheng@arm.com>
12604 PR tree-optimization/82965
12605 PR tree-optimization/83991
12606 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12607 by_profile_only parameter.
12608 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12609 information if the loop was predicted to iterate too many times.
12610 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12612 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12615 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12618 2018-04-13 Martin Liska <mliska@suse.cz>
12619 Jakub Jelinek <jakub@redhat.com>
12621 PR middle-end/81657
12622 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12623 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12624 * builtins.c (expand_builtin_memory_copy_args): Use
12625 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12626 handle dest_addr == pc_rtx.
12628 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12631 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12632 asked to not generate direct moves.
12633 (fix_trunc<mode>si2_stfiwx): Similar.
12634 (fix_trunc<mode>si2_internal): Similar.
12636 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12639 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12640 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12641 lookup if dest in some wider mode is known to be const0_rtx and
12642 if so, record permanent equivalence for it to be ZERO_EXTEND of
12643 the narrower mode destination.
12645 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12647 * lto-streamer-out.c (output_function): Revert 259346.
12648 * omp-expand.c (expand_oacc_for): Likewise.
12650 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12652 PR rtl-optimization/85354
12653 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12654 * sel-sched.c (sel_global_init): ... here.
12656 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12659 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12660 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12661 mode for PE-COFF targets.
12662 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12663 (i386_pe_asm_lto_end): Likewise.
12664 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12665 (TARGET_ASM_LTO_END): Likewise.
12666 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12667 (i386_pe_asm_lto_start): New function.
12668 (i386_pe_asm_lto_end): Likewise.
12670 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12671 Richard Biener <rguenther@suse.de>
12673 PR middle-end/84955
12674 * lto-streamer-out.c (output_function): Fix CFG loop state before
12676 * omp-expand.c (expand_oacc_for): Handle calls to internal
12677 functions like regular functions.
12679 2018-04-12 Richard Biener <rguenther@suse.de>
12682 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12683 for the early LTO debug to properly generate references to it
12684 during DIE emission. Do not re-use that for the skeleton for
12686 (dwarf2out_early_finish): Likewise.
12688 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12691 * config/i386/sse.md
12692 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12693 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12694 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12695 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12696 and output is a reg, avoid creating invalid lowpart subreg, but
12697 instead split into a 512-bit move. Don't split if not AVX512VL,
12698 input is xmm16+ reg and output is a mem.
12699 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12700 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12701 xmm16+ reg and output is a mem.
12703 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12705 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12706 also for flag_dwarf2_cfi_asm.
12708 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12710 PR rtl-optimization/85342
12711 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12712 a bool scalar var inside of the loop instead. Don't try to update
12713 recog_data.operand after failed apply_change_group.
12715 2018-04-12 Tom de Vries <tom@codesourcery.com>
12718 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12719 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12720 array with flexible array member as array without given dimension.
12721 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12722 argument for undefined param to true.
12724 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12727 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12728 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12729 from PowerPC section.
12730 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12731 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12732 help text that is too long.
12733 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12734 help text that is too long.
12735 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12736 help text that is too long.
12738 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12740 * config/alpha/alpha.md (stack_probe_internal): Rename
12741 from "probe_stack". Update all callers.
12743 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12745 PR rtl-optimization/84566
12746 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12747 sched_macro_fuse_insns.
12749 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12752 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12753 (compute_block_dependences): ... from here.
12755 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12757 PR tree-optimization/85331
12758 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12759 from int to HOST_WIDE_INT.
12761 2018-04-11 Martin Jambor <mjambor@suse.cz>
12764 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12765 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12766 not the same as the source val.
12767 (cgraph_edge_brings_value_p): New parameter.
12768 (gather_edges_for_value): Pass destination value to
12769 cgraph_edge_brings_value_p.
12770 (perhaps_add_new_callers): Likewise.
12771 (get_info_about_necessary_edges): Likewise and exclude values brought
12772 only by self-recursive edges.
12773 (create_specialized_node): Redirect only clones of self-calling edges.
12774 (+self_recursive_pass_through_p): New function.
12775 (find_more_scalar_values_for_callers_subset): Use it.
12776 (find_aggregate_values_for_callers_subset): Likewise.
12777 (known_aggs_to_agg_replacement_list): Removed.
12778 (decide_whether_version_node): Re-calculate known constants for all
12779 remaining context clones.
12781 2018-04-11 Richard Biener <rguenther@suse.de>
12784 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12785 from early DWARF output.
12786 (dwarf2out_early_finish): Output line info unconditionally into
12787 early DWARF and add reference to it.
12789 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12792 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12793 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12794 other than V2DFmode using iptr mode attribute.
12795 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12797 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12799 PR rtl-optimization/84659
12800 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12802 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12805 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12807 (output_loc_list): Pass address of a dummy size variable even in the
12808 locview handling loop.
12809 (index_location_lists): Add comment on why skip_loc_list_entry can't
12812 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12815 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12818 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12821 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12822 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12823 and -mstring-compare-inline-limit.
12825 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12828 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12829 for stack clash protection in a register whenever we need it to be in
12832 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12834 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12835 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12837 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12840 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12842 (mlong-double-): Ditto.
12843 * config/rs6000/sysv4.opt (msdata=): Ditto.
12844 (mtls-size=): Ditto.
12846 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12848 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12849 erroneous entries for
12850 "vector int vec_ldl (int, long int *)", and
12851 "vector unsigned int vec_ldl (int, unsigned long int *)".
12852 Add comments and entries for
12853 "vector bool char vec_ldl (int, bool char *)",
12854 "vector bool short vec_ldl (int, bool short *)",
12855 "vector bool int vec_ldl (int, bool int *)",
12856 "vector bool long long vec_ldl (int, bool long long *)",
12857 "vector pixel vec_ldl (int, pixel *)",
12858 "vector long long vec_ldl (int, long long *)",
12859 "vector unsigned long long vec_ldl (int, unsigned long long *)".
12860 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12861 type tree bool_long_long_type_node and correct definition of
12862 bool_V2DI_type_node to make reference to this new type tree.
12863 (rs6000_mangle_type): Replace erroneous reference to
12864 bool_long_type_node with bool_long_long_type_node.
12865 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12866 comments to emphasize sign distinctions for char and int types and
12867 replace RS6000_BTI_bool_long constant with
12868 RS6000_BTI_bool_long_long constant. Also add comment to restrict
12869 use of RS6000_BTI_pixel.
12870 (bool_long_type_node): Remove this macro definition.
12871 (bool_long_long_type_node): New macro definition
12873 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12875 PR rtl-optimization/85300
12876 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12877 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12878 simplify_unary_operation fails.
12880 2018-04-10 Martin Liska <mliska@suse.cz>
12882 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12883 cgraph_edge and ipa_ref.
12885 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12889 * config/i386/sse.md
12890 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12891 computation of the VEC_MERGE selector from mask.
12892 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12893 Fix decoding of the VEC_MERGE selector into mask.
12895 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12897 PR tree-optimization/85286
12898 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12900 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12902 * final.c (final_1): Set insn_last_address as well as
12903 insn_current_address.
12905 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12908 * explow.c (emit_stack_probe): Call validize_mem on memory location
12909 before passing it to gen_probe_stack. Create address operand and
12910 legitimize it for the probe_stack_address case.
12912 2018-04-09 Jan Hubicka <jh@suse.cz>
12915 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12916 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12917 * tree.c (free_lang_data_in_type): Fix handling of binfos;
12919 (free_lang_data): Rebuild type inheritance graph.
12921 2018-04-09 Martin Sebor <msebor@redhat.com>
12923 * invoke.texi (-finline-small-functions): Mention other optimization
12925 (-findirect-inlining, -fpartial-inlining): Same.
12926 (-finline-functions-called-once): Same.
12927 (-freorder-blocks-and-partition): Same.
12929 2018-04-09 Jan Hubicka <jh@suse.cz>
12932 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12933 jumps; choose last target that matches the criteria (i.e.
12934 no partition changes for non-crossing jumps).
12935 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12936 support for redirecting crossing jumps to non-crossing.
12938 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
12940 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12941 also for naked functions.
12943 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
12945 * config/arc/arc.md (add_shift): New pattern.
12946 (add_shift2): Likewise.
12947 (sub_shift): Likewise.
12948 (sub_shift_cmp0_noout): Likewise.
12949 (compare_si_ashiftsi): Likewise.
12950 (xbfu_cmp0_noout): New combine pattern.
12951 (xbfu_cmp0"): Likewise.
12952 (movsi_set_cc_insn): Place the predicable variant first.
12953 (commutative_binary_cmp0_noout): Remove clobber.
12954 (commutative_binary_cmp0): New pattern.
12955 (noncommutative_binary_cmp0): Likewise.
12956 (noncommutative_binary_cmp0_noout): Likewise.
12957 (noncommutative_binary_comparison_result_used): Removed.
12958 (rsub_cmp0): New pattern.
12959 (rsub_cmp0_noout): Likewise.
12960 (extzvsi): Changed, keep only meaningful variants.
12961 (SQH, SEZ): New iterators.
12962 (SQH_postfix): New mode attribute.
12963 (SEZ_prefix): New code attribute.
12964 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12965 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
12966 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
12967 of numerical value.
12968 (noncommutative_operator): Check the availability of barrel
12971 2018-04-09 Richard Biener <rguenther@suse.de>
12973 PR tree-optimization/85284
12974 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
12975 Only use the niter constraining form of simple_iv when the exit
12976 is always executed.
12978 2018-04-09 Tom de Vries <tom@codesourcery.com>
12981 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12982 (define_expand "*memory_barrier"): New define_expand.
12983 (define_insn "memory_barrier"): New insn.
12985 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12987 PR rtl-optimization/80463
12988 PR rtl-optimization/83972
12989 PR rtl-optimization/83480
12991 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12992 correct producer for the insn.
12993 (tidy_control_flow): Fixup seqnos in case of debug insns.
12995 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12997 PR rtl-optimization/83913
12999 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13000 different sched-times when merging exprs.
13002 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13004 PR rtl-optimization/83962
13006 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13007 tidy_fallthru_edge and tidy_control_flow.
13009 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13011 PR rtl-optimization/83530
13013 * sel-sched.c (force_next_insn): New global variable.
13014 (remove_insn_for_debug): When force_next_insn is true, also leave only
13015 next insn in the ready list.
13016 (sel_sched_region): When the region wasn't scheduled, make another pass
13017 over it with force_next_insn set to 1.
13019 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13021 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13023 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13024 for interrupt control.
13025 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13026 functions for interrupt control.
13027 * config/nds32/nds32-intrinsic.md: Likewise.
13028 * config/nds32/nds32_intrinsic.h: Likewise.
13029 * config/nds32/nds32.h (nds32_builtins): Likewise.
13031 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13033 * config/nds32/nds32.c (nds32_init_machine_status,
13034 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13035 strict_aligned_p field.
13036 (nds32_expand_to_rtl_hook): New function.
13037 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13038 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13040 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13041 Chung-Ju Wu <jasonwucj@gmail.com>
13043 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13044 * config/nds32/nds32-n7.md: New file.
13045 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13046 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13048 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13049 * config/nds32/nds32.md (pipeline_model): Add n7.
13050 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13051 * config/nds32/pipelines.md: Include n7 settings.
13053 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13054 Chung-Ju Wu <jasonwucj@gmail.com>
13056 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13057 * config/nds32/nds32-e8.md: New file.
13058 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13059 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13061 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13062 * config/nds32/nds32.md (pipeline_model): Add e8.
13063 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13064 * config/nds32/pipelines.md: Include e8 settings.
13066 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13067 Chung-Ju Wu <jasonwucj@gmail.com>
13069 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13070 * config/nds32/nds32-n8.md: New file.
13071 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13072 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13074 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13075 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13076 * config/nds32/nds32.md (pipeline_model): Add n8.
13077 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13078 * config/nds32/pipelines.md: Include n8 settings.
13080 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13081 Chung-Ju Wu <jasonwucj@gmail.com>
13083 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13084 * config/nds32/nds32-n9-2r1w.md: New file.
13085 * config/nds32/nds32-n9-3r2w.md: New file.
13086 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13087 nds32_register_ports): New or modify for cpu n9.
13088 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13090 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13091 * config/nds32/nds32-utils.c: New file.
13092 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13093 TARGET_MUL_SLOW): Define.
13094 * config/nds32/nds32.md (pipeline_model): New attribute.
13095 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13096 New options that support cpu n9.
13097 * config/nds32/pipelines.md: Include n9 settings.
13098 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13100 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13102 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13103 information if necessary.
13104 (output_cond_branch_compare_zero): Likewise.
13105 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13106 (nds32_target_alignment): Refine for alignment.
13107 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13108 (FUNCTION_BOUNDARY): Modify.
13109 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13111 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13113 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13115 * config/nds32/constants.md (unspec_volatile_element): Add values for
13116 TLB operation and data prefetch.
13117 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13118 functions for TLB operation and data prefetch.
13119 * config/nds32/nds32-intrinsic.md: Likewise.
13120 * config/nds32/nds32_intrinsic.h: Likewise.
13121 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13122 (nds32_print_operand): Likewise.
13123 * config/nds32/nds32.h (nds32_builtins): Likewise.
13125 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13126 Andrew Pinski <pinsika@gcc.gnu.org>
13128 PR middle-end/82976
13129 * match.pd: Use constant_boolean_node of correct type instead of
13130 boolean_true_node or boolean_false_node for simplifying
13131 pointer comparisons to zero.
13133 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13135 PR tree-optimization/80021
13136 * tree.c (verify_type_variant): Make error call in verify_variant_match
13137 translatable and remove final full stop.
13139 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13141 * config/nds32/constants.md (unspec_volatile_element): Add
13142 UNSPEC_VOLATILE_EH_RETURN.
13143 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13144 nds32_output_stack_pop): Support dwarf exception handling process.
13145 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13146 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13147 exception handling process.
13148 (nds32_compute_stack_frame): Likewise.
13149 (nds32_return_addr_rtx): Likewise.
13150 (nds32_initial_elimination_offset): Likewise.
13151 (nds32_expand_prologue): Likewise.
13152 (nds32_expand_epilogue): Likewise.
13153 (nds32_dynamic_chain_address): New function.
13154 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13155 exception handling.
13156 (DYNAMIC_CHAIN_ADDRESS): Define.
13157 (EH_RETURN_DATA_REGNO): Define.
13158 (EH_RETURN_STACKADJ_RTX): Define.
13159 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13160 patterns for dwarf exception handling.
13162 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13164 * config/nds32/nds32.h: Clean up obsolete macros.
13166 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13168 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13169 Add enum values for particular instructions.
13170 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13171 particular intrinsic functions.
13172 * config/nds32/nds32-intrinsic.md: Likewise.
13173 * config/nds32/nds32_intrinsic.h: Likewise.
13174 * config/nds32/nds32.h (nds32_builtins): Likewise.
13175 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13176 (btst, ave): New patterns for particular instructions.
13178 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13180 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13181 Add enum values for atomic load/store and memory sync.
13182 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13184 * config/nds32/nds32-intrinsic.md: Likewise.
13185 * config/nds32/nds32_intrinsic.h: Likewise.
13186 * config/nds32/nds32.h (nds32_builtins): Likewise.
13188 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13190 PR tree-optimization/85257
13191 * fold-const.c (native_encode_vector): If not all elts could fit
13192 and off is -1, return 0 rather than offset.
13193 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13194 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13195 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13196 adjust buffer in native_interpret_expr call.
13198 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13200 * config/nds32/constants.md (unspec_volatile_element): Add cache
13201 control enum values.
13202 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13203 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13204 * config/nds32/nds32.c (nds32_cctl_names): New.
13205 (nds32_print_operand): Handle cache control register names.
13206 * config/nds32/nds32.h (nds32_builtins): New enum values.
13207 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13209 * config/nds32/nds32.md (type): Add mmu.
13210 * config/nds32/pipelines.md (simple_insn): Add mmu.
13212 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13214 * config/nds32/nds32.md (type): Remove call.
13215 * config/nds32/pipelines.md (simple_insn): Likewise.
13217 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13219 * config/nds32/constants.md (unspec_volatile_element): Add
13220 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13221 UNSPEC_VOLATILE_FMFCFG.
13222 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13223 description for fmfcfg and fmfcsr.
13224 (bdesc_1arg): Add fmtcsr.
13225 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13226 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13227 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13228 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13229 unspec_fmfcfg): New patterns.
13230 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13231 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13232 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13233 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13234 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13235 __nds32__fmfcfg): Define.
13237 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13239 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13240 intrinsic register names.
13241 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13242 intrinsic register enum values and macros.
13244 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13246 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13247 for load/store addressing form.
13248 (nds32_print_operand_address): Likewise.
13250 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13253 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13254 based on LABEL_REF. Remove useless assertion.
13255 (pic_address_needs_scratch): Fix formatting.
13256 (sparc_legitimize_pic_address): Minor tweaks.
13257 (sparc_delegitimize_address): Adjust assertion accordingly.
13258 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13259 into symbolic_operand.
13260 (movsi_high_pic_label_ref): Likewise.
13261 (movsi_lo_sum_pic_label_ref): Likewise.
13262 (movdi_pic_label_ref): Likewise.
13263 (movdi_high_pic_label_ref): Likewise.
13264 (movdi_lo_sum_pic_label_ref): Likewise.
13266 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13268 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13269 custom LIB_SPEC setup.
13271 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13272 Kito Cheng <kito.cheng@gmail.com>
13274 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13275 * config/riscv/freebsd.h: New.
13277 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13279 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13280 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13283 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13284 Kito Cheng <kito.cheng@gmail.com>
13286 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13287 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13288 * config/nds32/nds32-protos.h (nds32_output_call,
13289 nds32_output_return): Declare.
13290 * config/nds32/nds32.md: Refine all the call and return patterns.
13292 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13295 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13296 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13298 PR rtl-optimization/84872
13299 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13300 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13301 EDGE_CROSSING edge.
13303 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13305 * expr.c (copy_blkmode_to_reg): Revert 254862.
13306 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13308 2018-04-06 Richard Biener <rguenther@suse.de>
13310 PR middle-end/85244
13311 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13312 after seeing a component reference with an adjacent field. Treat
13313 refs to arrays at struct end of external decls similar to
13314 refs to unconstrained commons.
13316 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13319 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13320 look through SAVE_EXPRs with non-side-effects argument. Adjust
13322 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13325 2018-04-06 Richard Biener <rguenther@suse.de>
13327 PR middle-end/85180
13328 * alias.c (find_base_term): New wrapper around find_base_term
13329 unwinding CSELIB_VAL_PTR changes.
13330 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13333 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13335 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13337 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13338 constant definitions.
13339 ("nop"): lr 0,0 -> nopr r0
13340 ("nop_lr0", "nop_lr1"): New insn definitions.
13342 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13344 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13345 NDS32_V3PUSH_AVAILABLE_P macro.
13347 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13348 Chung-Ju Wu <jasonwucj@gmail.com>
13350 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13351 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13352 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13353 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13354 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13355 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13356 * config/nds32/constraints.md: New constraints and checking for hard
13357 float configuration.
13358 * config/nds32/iterators.md: New mode iterator and attribute for hard
13359 float configuration.
13360 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13362 * config/nds32/nds32-fpu.md: New file.
13363 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13364 deal with hard float code generation.
13365 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13367 (abi_type, float_reg_number): New enum type.
13368 * config/nds32/nds32-predicates.c: New predicates for hard float.
13369 * config/nds32/nds32-protos.h: Declare functions for hard float.
13370 * config/nds32/nds32.c: Implementation for hard float configuration.
13371 * config/nds32/nds32.h: Definitions for hard float configuration.
13372 * config/nds32/nds32.md: Include hard float machine description and
13373 modify patterns for hard float configuration.
13374 * config/nds32/nds32.opt: New options for hard float configuration.
13375 * config/nds32/predicates.md: New predicates for hard float
13378 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13380 * common/config/nds32/nds32-common.c
13381 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13383 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13385 PR middle-end/85195
13386 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13387 CONSTRUCTOR_ELT (ctor, ...)->value.
13389 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13392 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13394 2018-04-05 Tom de Vries <tom@codesourcery.com>
13397 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13400 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13401 Kito Cheng <kito.cheng@gmail.com>
13403 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13404 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13405 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13406 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13408 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13409 Kito Cheng <kito.cheng@gmail.com>
13411 * config/nds32/constraints.md (Ufe): New memory constraint.
13412 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13413 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13414 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13416 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13417 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13419 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13421 * config/nds32/nds32.md: Use optimize_size in the condition for
13422 alu-shift instructions.
13424 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13426 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13428 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13430 * config/nds32/nds32.md (negsi2): Refine pattern.
13432 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13433 Chung-Ju Wu <jasonwucj@gmail.com>
13435 * config/nds32/iterators.md (shift_rotate): New code iterator.
13436 (shift): New code attribute.
13437 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13438 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13439 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13440 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13441 bit-wise operations.
13442 (andsi3, *andsi3): Ditto.
13443 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13444 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13445 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13446 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13447 nds32_ior_operand, nds32_xor_operand): New predicates.
13449 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13451 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13452 (addsi3, subsi3): ... this.
13454 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13456 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13458 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13460 * config/nds32/nds32.md: Adjust indention.
13462 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13464 * config/nds32/nds32.md (feature): New attribute.
13466 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13468 * config/nds32/nds32.md (subtype): New attribute.
13470 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13473 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13474 expansion to perform a bitwise AND of the argument followed by a
13475 boolean negation of the result.
13477 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13479 PR rtl-optimization/84878
13480 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13481 the basic block. Assert the use reference is not artificial and that
13482 it has an associated insn.
13484 2018-04-04 Michael Matz <matz@suse.de>
13486 * builtins.c (compute_objsize): Pass correct operand
13487 to array_at_struct_end_p.
13489 2018-04-04 Richard Biener <rguenther@suse.de>
13492 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13493 from contexts for DINFO_LEVEL_TERSE and below.
13495 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13497 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13498 resiter_operand condition.
13499 * config/nds32/nds32.md (*move<mode>): Ditto.
13501 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13502 Monk Chiang <sh.chiang04@gmail.com>
13504 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13506 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13508 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13510 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13511 Kito Cheng <kito.cheng@gmail.com>
13513 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13514 nds32_cond_code_str, output_cond_branch,
13515 output_cond_branch_compare_zero, nds32_expand_cbranch,
13516 nds32_expand_cstore, nds32_expand_movcc,
13517 nds32_output_cbranchsi4_equality_zero,
13518 nds32_output_cbranchsi4_equality_reg,
13519 nds32_output_cbranchsi4_equality_reg_or_const_int,
13520 nds32_output_cbranchsi4_greater_less_zero: New functions.
13521 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13522 nds32_expand_cstore, nds32_expand_movcc,
13523 nds32_output_cbranchsi4_equality_zero,
13524 nds32_output_cbranchsi4_equality_reg,
13525 nds32_output_cbranchsi4_equality_reg_or_const_int,
13526 nds32_output_cbranchsi4_greater_less_zero): Declare.
13527 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13528 nds32_rimm11s_operand): New predicates.
13529 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13530 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13533 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13535 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13536 * config/nds32/nds32.md: Ditto.
13537 * config/nds32/pipelines.md: Ditto.
13539 2018-04-04 Richard Biener <rguenther@suse.de>
13541 PR tree-optimization/85168
13542 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13543 propagating abnormals.
13545 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13547 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13549 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13550 Kito Cheng <kito.cheng@gmail.com>
13552 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13553 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13554 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13555 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13556 * config/nds32/nds32.md (sibcall_internal): New.
13557 (sibcall_register): Remove.
13558 (sibcall_immediate): Remove.
13559 (sibcall_value_internal): New.
13560 (sibcall_value_register): Remove.
13561 (sibcall_value_immediate): Remove.
13562 * config/nds32/predicates.md (nds32_general_register_operand): New.
13563 (nds32_call_address_operand): New.
13565 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13567 PR rtl-optimization/85167
13568 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13569 bb_defs if *split_p, instead preinitialize it to NULL.
13571 PR tree-optimization/85156
13572 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13573 evaluating the argument multiple times.
13575 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13577 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13579 (_mm_cvtpd_ps): Likewise.
13580 (_mm_cvttpd_epi32): Likewise.
13581 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13582 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13583 vector, pixel, and bool following altivec.h include.
13585 2018-04-03 Martin Sebor <msebor@redhat.com>
13587 * doc/extend.texi (Common Function Attributes): Clarify.
13588 (const attribute): Likewise.
13589 (pure attribute): Likewise.
13591 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13594 * config/i386/i386.c (ix86_expand_vector_set): Use
13595 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13597 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13599 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13600 instructions when changing rounding bits to preserve precision bits
13601 in the x87 control word.
13603 2018-04-03 Martin Liska <mliska@suse.cz>
13605 PR tree-optimization/82491
13606 * rtl.h (strip_offset_and_add): Replace += suboffset with
13607 poly_uint64 () + suboffset.
13609 2018-03-29 Martin Liska <mliska@suse.cz>
13610 Martin Jambor <mjambor@suse.cz>
13613 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13614 param_type is not an integral or pointer type.
13616 2018-04-03 Richard Biener <rguenther@suse.de>
13618 * sese.h (recompute_all_dominators): Remove.
13620 2018-04-02 Martin Sebor <msebor@redhat.com>
13622 * doc/invoke.texi (-Wrestrict): Fix typos.
13624 2018-04-02 Jim Wilson <jimw@sifive.com>
13626 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13627 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13628 (<optab>di3, <optab>si3_extend): Likewise.
13629 (<optab>si3_mask, <optab>si3_mask_1): New.
13630 (<optab>di3_mask, <optab>di3_mask_1): New.
13631 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13632 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13633 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13635 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13637 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13640 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13642 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13643 (nds32_canonicalize_comparison): New function.
13645 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13646 Kito Cheng <kito.cheng@gmail.com>
13647 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13649 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13650 * config/nds32/constants.md (unspec_volatile_element): Add
13651 UNSPEC_VOLATILE_RELAX_GROUP.
13652 * config/nds32/nds32-relax-opt.c: New file.
13653 * config/nds32/nds32-predicates.c
13654 (nds32_symbol_load_store_p): New function.
13655 * config/nds32/nds32-protos.h
13656 (nds32_symbol_load_store_p): Declare function.
13657 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13658 * config/nds32/nds32.c
13659 (nds32_register_pass): New function to register pass.
13660 (nds32_register_passes): New function to register passes.
13661 * config/nds32/nds32.md (relax_group): New pattern.
13662 * config/nds32/nds32.opt (mrelax-hint): New option.
13663 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13665 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13667 * config/nds32/t-nds32: Modify files dependency.
13669 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13671 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13672 (PROFILE_HOOK): Define its implementation.
13674 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13676 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13677 type and 32-bit size.
13679 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13681 PR middle-end/85090
13682 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13683 (V_128_256): New mode iterator.
13684 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13685 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13686 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13688 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13689 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13691 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13694 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13695 NaN inputs correctly.
13697 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13700 * config/rs6000/vsx.md (??r): New mode attribute.
13701 (*vsx_mov<mode>_64bit): Use it.
13702 (*vsx_mov<mode>_32bit): Likewise.
13704 2018-03-30 Martin Sebor <msebor@redhat.com>
13706 PR tree-optimization/84818
13707 * builtins.c (check_access): Use warning_n.
13709 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13712 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13714 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13717 2018-03-30 Julia Koval <julia.koval@intel.com>
13720 * x86-tune.def (movx, partial_reg_dependency): Enable for
13723 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13725 PR inline-asm/84985
13726 * lra-constraints.c (process_alt_operands): Move setting
13727 this_alternative_matches below.
13729 2018-03-29 Martin Liska <mliska@suse.cz>
13732 * doc/invoke.texi: Document how LTO works with debug info.
13733 Describe auto-load support of binutils. Mention 'x86-64'
13734 as valid option value of -march option.
13736 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13738 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13741 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13742 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13743 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13746 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13749 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13750 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13751 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13752 for RS6000_BTM_POWERPC64.
13753 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13754 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13755 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13758 (DIVDEU): Likewise.
13760 2018-03-28 Carl Love <cel@us.ibm.com>
13763 2017-09-27 Carl Love <cel@us.ibm.com>
13765 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13766 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13767 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13770 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13772 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13773 instead of __vector bool.
13774 (_mm_max_pu8): Likewise.
13775 (_mm_min_pi16): Likewise.
13777 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13780 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13781 (DIVWEUO): Likewise.
13782 (DIVDEO): Likewise.
13783 (DIVDEUO): Likewise.
13784 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13785 DIVWEUO and DIVDEUO.
13786 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13787 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13788 (div_extend): Likewise.
13789 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13791 (__builtin_divweuo): Likewise.
13792 (__builtin_divdeo): Likewise.
13793 (__builtin_divdeuo): Likewise.
13795 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13798 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13799 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13801 PR tree-optimization/82004
13802 * gimple-match-head.c (optimize_pow_to_exp): New function.
13803 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13804 Don't fold to exp if optimize_pow_to_exp is false.
13806 2018-03-28 Martin Liska <mliska@suse.cz>
13809 * calls.c (initialize_argument_information): Fix trailing space.
13810 * common.opt: Fix typo and provide better explanation for
13811 -fsanitize-coverage option.
13812 * config/i386/i386.opt: Fix typo.
13814 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13815 Martin Liska <mliska@suse.cz>
13818 * gimplify.c (asan_poison_variable): Don't do the check for
13819 gimplify_omp_ctxp here.
13820 (gimplify_decl_expr): Do it here.
13821 (gimplify_target_expr): Likewise.
13823 2018-03-28 Martin Liska <mliska@suse.cz>
13826 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13827 chkp_type_bounds_count if MPX is not enabled.
13829 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13831 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13833 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13836 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13837 function to create the function decl for complex long double
13838 multiply and divide for -mabi=ieeelongdouble.
13839 (init_float128_ieee): Call it.
13841 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13844 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13845 -fcf-protection=branch -mibt.
13846 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13848 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13851 * config/arm/arm.c (arm_valid_symbolic_address): Handle
13852 arm_word_relocations.
13854 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
13857 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13858 extern array declarations.
13860 2018-03-27 Richard Biener <rguenther@suse.de>
13862 PR middle-end/84067
13863 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13864 explicit single_use checks.
13866 2018-03-27 Richard Biener <rguenther@suse.de>
13868 PR tree-optimization/85082
13869 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13872 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13874 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13875 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13876 Turn on fasynchronous-unwind-tables and funwind-tables.
13878 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
13881 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13882 (*bmi_blsr_<mode>_ccz): Ditto.
13884 2018-03-26 Tom de Vries <tom@codesourcery.com>
13886 PR tree-optimization/85063
13887 * omp-general.c (offloading_function_p): New function. Factor out
13889 * omp-offload.c (pass_omp_target_link::gate): ... here.
13890 * omp-general.h (offloading_function_p): Declare.
13891 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13892 with attribute omp declare target for offloading functions.
13894 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
13896 PR tree-optimization/84005
13897 * tree-data-ref.h (get_base_for_alignment): Declare.
13898 * tree-data-ref.c (get_base_for_alignment_1): New function.
13899 (get_base_for_alignment): Likewise.
13900 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13901 get_base_for_alignment to find a suitable base object, instead
13902 of always using drb->base_address.
13904 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13906 PR inline-asm/85022
13907 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13908 known size by default.
13910 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
13912 PR inline-asm/85030
13913 * lra-constraints.c (process_alt_operands): Don't match BLKmode
13914 and non BLKmode operands.
13916 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13919 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13920 Clean up attributes.
13922 2018-03-23 Richard Biener <rguenther@suse.de>
13925 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13926 we are going to emit early debug for LTO.
13928 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13930 PR inline-asm/85034
13931 * function.c (match_asm_constraints_1): Don't optimize if input
13932 doesn't satisfy general_operand predicate for output's mode.
13934 PR inline-asm/85022
13935 * alias.c (write_dependence_p): Don't require for x_canonicalized
13936 non-VOIDmode if x has VOIDmode.
13939 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13940 just don't try to optimize it rather than assert it never happens.
13942 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13944 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13945 macro expansions for definition of ST_INTERNAL_<mode> and
13946 LD_INTERNAL_<mode> builtins.
13947 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13949 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13951 (altivec_expand_st_builtin): Likewise.
13952 (altivec_expand_builtin): Remove calls to deleted functions.
13953 (rs6000_address_for_altivec): Delete this function.
13954 * config/rs6000/vector.md: Remove expands for
13955 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13957 2018-03-22 Sudakshina Das <sudi.das@arm.com>
13960 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13961 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13962 re-computing once computed.
13963 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13964 (arm_init_machine_status): Initialize
13965 machine->static_chain_stack_bytes.
13967 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13970 * doc/extend.texi: Add four new prototypes for vec_ld.
13971 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
13972 definitions for more logical presentation.
13973 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
13974 entries for V1TI variants of __builtin_altivec_ld builtin.
13975 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
13976 handling of V1TI variant of LVX icode pattern.
13977 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
13978 (rs6000_gimple_fold_builtin): Likewise.
13979 (altivec_init_builtins): Add code to define
13980 __builtin_altivec_lvx_v1ti function.
13982 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13984 PR inline-asm/84941
13985 * function.c (match_asm_constraints_1): Don't do the optimization
13986 if input isn't a REG, SUBREG, MEM or constant.
13988 2018-03-22 Tom de Vries <tom@codesourcery.com>
13990 PR tree-optimization/84956
13991 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13992 bb_has_abnormal_pred.
13994 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13997 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13998 DECL_INITIAL (decl) to decl at the end.
13999 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14000 adjust the comment.
14002 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14004 * doc/extend.texi (__builtin_tgmath): Document when complex
14005 integer types are treated as _Complex _Float64.
14007 2018-03-21 Tom de Vries <tom@codesourcery.com>
14009 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14011 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14013 PR tree-optimization/84960
14014 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14015 if it is ENTRY block, move them into single succ of ENTRY in that case.
14017 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14019 PR tree-optimization/84811
14020 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14021 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14022 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14023 (known_subrange_p): Update accordingly. Cast each value involved
14024 in the size comparison, rather than casting the result of the
14027 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14029 PR tree-optimization/84982
14030 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14031 by flipping the least significant bit rather than all bits from
14032 bitpos to bitpos + bitsize - 1.
14034 2018-03-21 Nathan Sidwell <nathan@acm.org>
14036 * doc/extend.texi (Deprecated Features): Remove mention of
14037 long-deleted deprecations.
14039 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14042 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14043 * configure: Regenerate.
14045 2018-03-21 Tom de Vries <tom@codesourcery.com>
14047 PR tree-optimization/83126
14048 * tree-parloops.c (num_phis): New function.
14049 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14051 2018-03-21 Nathan Sidwell <nathan@acm.org>
14053 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14054 mention anon-struct/union members and trailing attributes.
14056 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14058 PR tree-optimization/84969
14059 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14060 builtin memset partitions if they set different rhs values.
14062 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14064 PR rtl-optimization/84989
14065 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14066 VEC_DUPLICATE with scalar result mode.
14068 2018-03-21 Martin Liska <mliska@suse.cz>
14071 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14072 not intended return statement.
14074 2018-03-21 Martin Liska <mliska@suse.cz>
14077 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14078 (chkp_find_bound_slots_1): Limit number of iterations.
14080 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14083 * Minor grammar fixes for x86 options.
14085 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14088 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14089 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14091 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14094 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14095 (altivec_lvx_<mode>_1op): Likewise.
14096 (altivec_stvx_<mode>_2op): Likewise.
14097 (altivec_stvx_<mode>_1op): Likewise.
14098 (altivec_lvx_<VM2:mode>): New define_expand.
14099 (altivec_stvx_<VM2:mode>): Likewise.
14100 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14101 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14102 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14103 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14104 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14105 (rs6000_gen_lvx): Likewise.
14106 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14107 (altivec_expand_stv_builtin): Likewise.
14108 (altivec_expand_builtin): Likewise.
14109 * config/rs6000/vector.md: Likewise.
14111 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14114 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14117 2018-03-20 Richard Biener <rguenther@suse.de>
14120 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14121 sign-conversions as zero, fall back to standard scalar_stmt
14124 2018-03-20 Martin Liska <mliska@suse.cz>
14127 * predict.c (rebuild_frequencies): Handle case when we have
14128 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14130 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14133 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14134 flag_section_anchors.
14135 * varasm.c (use_blocks_for_decl_p): Remove hack for
14136 dw2_force_const_mem.
14139 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14141 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14142 be created, use lowpart_subreg of operands[0] rather than operands[0]
14144 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14145 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14146 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14147 and n constraint instead of aarch64_shift_imm_di and Usd.
14148 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14149 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14151 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14154 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14155 to favor GPR over NEON registers.
14156 (<shift>di3_neon): Likewise.
14158 2018-03-20 Tom de Vries <tom@codesourcery.com>
14161 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14162 (nvptx_process_pars): Emit bar.sync asap and alap.
14164 2018-03-20 Tom de Vries <tom@codesourcery.com>
14167 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14168 seen_label if seen_label is already set.
14170 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14173 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14174 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14175 Use 1U instead of 1. Formatting fixes.
14178 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14179 instead of TREE_TYPE (s1) for the return value.
14181 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14183 PR tree-optimization/84946
14184 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14185 bitsize + bitsize in poly_uint64 rather than poly_int64.
14188 * dwarf2asm.c: Include fold-const.c.
14189 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14190 of decl rather than decl itself.
14192 PR rtl-optimization/84643
14193 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14195 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14198 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14199 calling assemble_variable.
14201 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14204 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14205 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14207 2018-03-19 Jim Wilson <jimw@sifive.com>
14210 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14211 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14212 (riscv_first_stack_step): Likewise.
14213 (riscv_option_override): Use STACK_BOUNDARY instead of
14214 MIN_STACK_BOUNDARY.
14215 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14216 MIN_STACK_BOUNDARY.
14217 (BIGGEST_ALIGNMENT): Set to 128.
14218 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14219 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14222 2018-03-19 Richard Biener <rguenther@suse.de>
14224 PR tree-optimization/84933
14225 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14226 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14228 2018-03-19 Richard Biener <rguenther@suse.de>
14230 PR tree-optimization/84859
14231 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14232 (cond_if_else_store_replacement): Perform sinking operation on
14233 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14234 Generalize what a BB with a single eligible store is.
14236 2018-03-19 Richard Biener <rguenther@suse.de>
14238 PR tree-optimization/84929
14239 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14240 chrec_is_positive against non-chrec arg.
14242 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14245 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14247 2018-03-18 Martin Liska <mliska@suse.cz>
14249 PR rtl-optimization/84635
14250 * regrename.c (build_def_use): Use matches_mode only when
14253 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14255 PR tree-optimization/84913
14256 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14257 vectorize chains of COND_EXPRs.
14259 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14261 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14263 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14265 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14267 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14269 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14271 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14272 Kito Cheng <kito.cheng@gmail.com>
14274 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14275 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14276 (nds32_adjust_reg_alloc_order): New function.
14277 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14279 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14281 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14282 nds32_print_operand, nds32_print_operand_address): Use
14283 HOST_WIDE_INT_PRINT_DEC instead.
14285 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14287 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14289 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14292 * config/i386/i386.c (initial_ix86_tune_features,
14293 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14294 unsigned long long.
14295 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14296 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14297 rather than 1u << ix86_tune. Formatting fix.
14298 (ix86_option_override_internal): Change ix86_arch_mask from
14299 unsigned int to unsigned HOST_WIDE_INT, initialize to
14300 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14301 (ix86_function_specific_restore): Likewise.
14303 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14306 * postreload.c (reload_combine_recognize_pattern): Perform
14307 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14308 truncate_int_for_mode the result for the destination's mode.
14311 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14312 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14314 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14317 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14320 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14321 regno_allocno_class_array and sorted_pseudos.
14322 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14323 insns where regno is used.
14325 2018-03-16 Martin Liska <mliska@suse.cz>
14328 * multiple_target.c (create_dispatcher_calls): Redirect
14329 reference in the symbol table.
14331 2018-03-16 Martin Liska <mliska@suse.cz>
14334 * multiple_target.c (create_dispatcher_calls): Redirect also
14337 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14341 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14342 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14343 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14345 2018-03-16 Julia Koval <julia.koval@intel.com>
14347 * doc/invoke.texi (Skylake Server): Add CLWB.
14348 Cannonlake): Remove CLWB.
14350 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14352 PR tree-optimization/84841
14353 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14355 (FLOAT_ONE_CONST_TYPE): Define.
14356 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14357 (sort_by_operand_rank): Put entries with higher constant_type last
14358 rather than first to match comments.
14360 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14362 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14365 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14368 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14369 check and use address of target always.
14371 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14374 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14375 (indirect_thunk_bnd_needed): Likewise.
14376 (indirect_thunks_used): Likewise.
14377 (indirect_thunks_bnd_used): Likewise.
14378 (indirect_return_needed): New.
14379 (indirect_return_bnd_needed): Likewise.
14380 (output_indirect_thunk_function): Add a bool argument for
14382 (output_indirect_thunk_function): Don't generate alias for
14383 function return thunk.
14384 (ix86_code_end): Call output_indirect_thunk_function to generate
14385 function return thunks.
14386 (ix86_output_function_return): Set indirect_return_bnd_needed
14387 and indirect_return_needed instead of indirect_thunk_bnd_needed
14388 and indirect_thunk_needed.
14390 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14392 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14393 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14394 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14396 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14397 Paul Hua <paul.hua.gm@gmail.com>
14400 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14402 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14404 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14405 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14406 resp. SFmode cases.
14408 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14411 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14412 instead of GET_MODE_SIZE when comparing Units.
14414 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14417 * varasm.c (hash_section): Return an unchangeble hash value
14418 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14419 Return !aarch64_can_use_per_function_literal_pools_p ().
14421 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14424 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14425 rather than address of cmode as last argument to prepare_cmp_insn.
14427 2018-03-15 Julia Koval <julia.koval@intel.com>
14429 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14430 F_AVX512VNNI, F_AVX512BITALG): New.
14432 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14435 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14436 insn for floating-point loads and stores.
14438 2018-03-14 Carl Love <cel@us.ibm.com>
14440 * config/rs6000/rs6000-c.c: Add macro definitions for
14441 ALTIVEC_BUILTIN_VEC_PERMXOR.
14442 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14443 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14444 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14445 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14447 * config/doc/extend.texi: Add prototypes for vec_permxor.
14449 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14452 * diagnostic-show-locus.c (class layout_point): Convert m_line
14453 from int to linenum_type.
14454 (line_span::comparator): Use linenum "compare" function when
14455 comparing line numbers.
14456 (test_line_span): New function.
14457 (layout_range::contains_point): Convert param "row" from int to
14459 (layout_range::intersects_line_p): Likewise.
14460 (layout::will_show_line_p): Likewise.
14461 (layout::print_source_line): Likewise.
14462 (layout::should_print_annotation_line_p): Likewise.
14463 (layout::print_annotation_line): Likewise.
14464 (layout::print_leading_fixits): Likewise.
14465 (layout::annotation_line_showed_range_p): Likewise.
14466 (struct line_corrections): Likewise for field m_row.
14467 (line_corrections::line_corrections): Likewise for param "row".
14468 (layout::print_trailing_fixits): Likewise.
14469 (layout::get_state_at_point): Likewise.
14470 (layout::get_x_bound_for_row): Likewise.
14471 (layout::print_line): Likewise.
14472 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14473 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14474 * input.c (selftest::test_linenum_comparisons): New function.
14475 (selftest::input_c_tests): Call it.
14476 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14477 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14478 * selftest.h (ASSERT_GT): New macro.
14479 (ASSERT_GT_AT): New macro.
14480 (ASSERT_LT): New macro.
14481 (ASSERT_LT_AT): New macro.
14483 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14485 PR rtl-optimization/84780
14486 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14488 2018-03-14 Martin Liska <mliska@suse.cz>
14490 * tree.c (record_node_allocation_statistics): Use
14491 get_stats_node_kind.
14492 (get_stats_node_kind): New function extracted from
14493 record_node_allocation_statistics.
14494 (free_node): Use get_stats_node_kind.
14496 2018-03-14 Richard Biener <rguenther@suse.de>
14498 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14499 that the value-set of ANTIC_IN doesn't grow.
14502 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14504 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14505 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14507 2018-03-14 Julia Koval <julia.koval@intel.com>
14509 * config.gcc (icelake-client, icelake-server): New.
14511 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14512 (initial_ix86_arch_features): Ditto.
14513 (PTA_SKYLAKE): Add SGX.
14514 (PTA_ICELAKE): Remove.
14515 (PTA_ICELAKE_CLIENT): New.
14516 (PTA_ICELAKE_SERVER): New.
14517 (ix86_option_override_internal): Split up icelake on icelake client and
14519 (get_builtin_code_for_version): Ditto.
14520 (fold_builtin_cpu): Ditto.
14521 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14522 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14523 * config/i386/i386.h (processor_type): Ditto.
14524 * doc/invoke.texi: Ditto.
14526 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14529 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14530 INTEGER_CST offset, add it together with bitpos / 8 and
14531 sign extend based on POINTER_SIZE.
14535 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14538 * config/i386/constraints.md (Yc): New register constraint.
14539 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14540 Use Yc constraint for alternative 2 of operand 0. Remove
14541 preferred_for_speed attribute.
14543 2018-03-14 Richard Biener <rguenther@suse.de>
14545 PR tree-optimization/84830
14546 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14547 with the old one to avoid oscillations.
14549 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14552 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14554 (assign_by_spills): Return a flag of reload assignment failure.
14555 Do not process the reload assignment failures. Do not spill other
14556 reload pseudos if they has the same reg class. Update n if
14558 (lra_assign): Add a return arg. Set up from the result of
14559 assign_by_spills call.
14560 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14561 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14562 usage_insns if it is not NULL.
14563 (spill_hard_reg_in_range): New function.
14564 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14565 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14566 function prototypes.
14567 (lra_assign): Change prototype.
14568 * lra.c (lra): Add code to deal with fails by splitting hard reg
14571 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14573 * config/riscv/riscv.opt (mrelax): New option.
14574 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14575 "norelax" when riscv_mrelax is disabled.
14576 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14578 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14581 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14582 reassociation for int modes.
14584 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14586 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14587 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14589 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14590 * config/aarch64/aarch64-sve.md
14591 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14592 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14593 (*extend<mode><Vwide>2): Rename to...
14594 (aarch64_sve_extend<mode><Vwide>2): ...this.
14595 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14596 renaming the old pattern to...
14597 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14599 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14600 define_expand, renaming the old pattern to...
14601 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14602 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14603 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14604 account when deciding which SVE instruction the optab should use.
14605 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14607 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14609 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14610 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14611 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14612 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14613 (tlsdesc_small_<mode>): Turn a define_expand and use
14614 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14615 (tlsdesc_small_advsimd_<mode>): ...this.
14616 (tlsdesc_small_sve_<mode>): New pattern.
14618 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14620 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14621 (UNSPEC_UMUL_HIGHPART): New constants.
14622 (MUL_HIGHPART): New int iteraor.
14623 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14624 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14626 (*<su>mul<mode>3_highpart): New define_insn.
14628 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14631 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14634 2018-03-13 Martin Liska <mliska@suse.cz>
14637 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14639 (sem_item_optimizer::~sem_item_optimizer): Release it.
14640 (sem_item_optimizer::merge_classes): Register variable aliases.
14641 (sem_item_optimizer::fixup_pt_set): New function.
14642 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14643 * ipa-icf.h: Declare new variables and functions.
14645 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14647 PR middle-end/84834
14648 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14649 integer_pow2p@2 and test integer_pow2p in condition.
14650 (A < 0 ? C : 0): Similarly for @1.
14652 PR middle-end/84831
14653 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14654 characters starting at p contain '\0' character, don't look beyond
14658 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14659 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14662 * reg-stack.c (change_stack): Change update_end var from int to
14663 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14664 also call set_block_for_insn on the newly added insns and rescan.
14667 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14668 on the last operand.
14671 * tree.c (stabilize_reference_1): Return save_expr (e) for
14672 STATEMENT_LIST even if it doesn't have side-effects.
14674 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14676 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14678 2018-03-12 Renlin Li <renlin.li@arm.com>
14680 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14681 aarch64_output_scalar_simd_mov_immediate.
14683 2018-03-12 Martin Sebor <msebor@redhat.com>
14685 PR tree-optimization/83456
14686 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14687 for perfectly overlapping calls to memcpy.
14688 (gimple_fold_builtin_memory_chk): Same.
14689 (gimple_fold_builtin_strcpy): Handle no-warning.
14690 (gimple_fold_builtin_stxcpy_chk): Same.
14691 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14693 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14695 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14696 parameter. Use it for SFmode.
14697 (rs6000_function_arg_advance_1): Adjust.
14698 (rs6000_function_arg): Adjust.
14699 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14701 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14703 PR rtl-optimization/84169
14704 PR rtl-optimization/84780
14705 * combine.c (can_combine_p): Check for a 2-insn combination whether
14706 the destination register is used between the two insns, too.
14708 2018-03-12 Richard Biener <rguenther@suse.de>
14710 PR tree-optimization/84803
14711 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14712 for refs DR analysis didn't process.
14714 2018-03-12 Richard Biener <rguenther@suse.de>
14716 PR tree-optimization/84777
14717 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14718 force-vectorize loops ignore whether we are optimizing for size.
14720 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14722 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14723 (TARGET_MD_ASM_ADJUST): Define.
14725 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14726 Kito Cheng <kito.cheng@gmail.com>
14727 Chung-Ju Wu <jasonwucj@gmail.com>
14729 * config/nds32/nds32.c (nds32_compute_stack_frame,
14730 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14731 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14732 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14733 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14734 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14735 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14736 * config/nds32/nds32.md (prologue, epilogue): Use macro
14737 NDS32_V3PUSH_AVAILABLE_P to do checking.
14739 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14742 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14743 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14744 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14745 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14746 addition of most attributes on !orig_type_die or the attribute not
14747 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14749 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14750 Chung-Ju Wu <jasonwucj@gmail.com>
14752 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14753 __NDS32_VH__ macro.
14754 * config/nds32/nds32.opt (mvh): New option.
14756 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14757 Chung-Ju Wu <jasonwucj@gmail.com>
14759 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14761 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14762 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14765 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14766 Chung-Ju Wu <jasonwucj@gmail.com>
14768 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14770 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14771 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14773 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14774 Kito Cheng <kito.cheng@gmail.com>
14775 Chung-Ju Wu <jasonwucj@gmail.com>
14777 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14778 UNSPEC_FFMISM and UNSPEC_FLMISM.
14779 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14780 for ffb, ffmism and flmism.
14781 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14782 (unspec_ffmism): Ditto.
14783 (unspec_flmism): Ditto.
14784 (nds32_expand_builtin_impl): Check if string extension is available.
14785 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14786 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14788 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14791 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14794 * lra-assigns.c (assign_by_spills): Return a flag of reload
14795 assignment failure. Do not process the reload assignment
14796 failures. Do not spill other reload pseudos if they has the same
14798 (lra_assign): Add a return arg. Set up from the result of
14799 assign_by_spills call.
14800 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14801 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14802 usage_insns if it is not NULL.
14803 (spill_hard_reg_in_range): New function.
14804 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14805 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14806 function prototypes.
14807 (lra_assign): Change prototype.
14808 * lra.c (lra): Add code to deal with fails by splitting hard reg
14811 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14814 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14816 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14819 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14820 (dw_val_node): Add val_symbolic_view.
14821 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14822 (symview_upper_bound): New.
14823 (new_line_info_table): Initialize symviews_since_reset.
14824 (dwarf2out_source_line): Count symviews_since_reset and set
14825 symview_upper_bound.
14826 (dw_val_equal_p): Handle symview.
14827 (add_AT_symview): New.
14828 (print_dw_val): Handle symview.
14829 (attr_checksum, attr_checksum_ordered): Likewise.
14830 (same_dw_val_p, size_of_die): Likewise.
14831 (value_format, output_die): Likewise.
14832 (add_high_low_attributes): Use add_AT_symview for entry_view.
14833 (dwarf2out_finish): Reset symview_upper_bound, clear
14836 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14839 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14840 Add strict argument and use it.
14841 (rs6000_split_multireg_move): Update for new strict argument.
14842 (mem_operand_gpr): Disallow all non-offsettable addresses.
14843 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14845 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14848 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14849 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14850 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14853 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14854 decl, use remap_type if we want to use the type.
14856 2018-03-09 Martin Sebor <msebor@redhat.com>
14858 PR tree-optimization/84526
14859 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14861 (builtin_access::generic_overlap): Be prepared to handle non-array
14864 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
14866 PR rtl-optimization/84682
14867 * lra-constraints.c (process_address_1): Check is_address flag
14868 for address constraints.
14869 (process_alt_operands): Likewise.
14870 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14871 preprocess_constraints.
14872 * recog.h (preprocess_constraints): Add oploc parameter.
14874 * recog.c (preprocess_constraints): Test address_operand for
14875 CT_ADDRESS constraints.
14877 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14880 * lra-assigns.c (assign_by_spills): Return a flag of reload
14881 assignment failure. Do not process the reload assignment
14882 failures. Do not spill other reload pseudos if they has the same
14884 (lra_assign): Add a return arg. Set up from the result of
14885 assign_by_spills call.
14886 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14887 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14888 usage_insns if it is not NULL.
14889 (spill_hard_reg_in_range): New function.
14890 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14891 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14892 function prototypes.
14893 (lra_assign): Change prototype.
14894 * lra.c (lra): Add code to deal with fails by splitting hard reg
14897 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14900 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14901 Accept complain bool parameter. Only emit errors if it is true.
14902 (arm_parse_cpu_option_name): Likewise.
14903 (arm_target_thumb_only): Adjust callers of the above.
14904 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14905 prototype to take a default true bool parameter.
14906 (arm_parse_arch_option_name): Likewise.
14908 2018-03-09 David Malcolm <dmalcolm@redhat.com>
14909 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
14913 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14914 * configure: Regenerate.
14915 * configure.ac ("linker --version-script option"): New.
14916 ("linker soname option"): New.
14918 2018-03-09 Richard Biener <rguenther@suse.de>
14920 PR tree-optimization/84775
14921 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14922 immediate uses of predicate stmts and mark them modified.
14925 PR tree-optimization/84178
14926 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14928 (version_loop_for_if_conversion): Delay update_ssa call.
14929 (tree_if_conversion): Delay update_ssa until after predicate
14932 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
14935 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14936 when the function accesses prior frames.
14938 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14941 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14942 gen_llsym, otherwise call maybe_gen_llsym.
14944 PR inline-asm/84742
14945 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14946 has ',' character inside of it.
14948 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14951 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14952 as clobbering CC_REGNUM.
14954 2018-03-08 Richard Biener <rguenther@suse.de>
14956 PR middle-end/84552
14957 * tree-scalar-evolution.c: Include tree-into-ssa.h.
14958 (follow_copies_to_constant): Do not follow SSA names registered
14961 2018-03-08 Richard Biener <rguenther@suse.de>
14963 PR tree-optimization/84178
14964 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14966 (version_loop_for_if_conversion): Delay update_ssa call.
14967 (tree_if_conversion): Delay update_ssa until after predicate
14970 2018-03-08 David Malcolm <dmalcolm@redhat.com>
14972 PR tree-optimization/84178
14973 * tree-if-conv.c (release_bb_predicate): Remove the
14974 the assertion that the stmts have NULL use_ops.
14975 Discard the statements, asserting that they haven't
14976 yet been added to a BB.
14978 2018-03-08 Richard Biener <rguenther@suse.de>
14980 PR tree-optimization/84746
14981 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
14982 (phi_translate): Pass in destination ANTIC_OUT set.
14983 (phi_translate_1): Likewise. For a simplified result lookup
14984 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
14985 (phi_translate_set): Adjust.
14986 (do_pre_regular_insertion): Likewise.
14987 (do_pre_partial_partial_insertion): Likewise.
14989 2018-03-08 Martin Liska <mliska@suse.cz>
14991 PR gcov-profile/84735
14992 * doc/gcov.texi: Document usage of profile files.
14993 * gcov-io.h: Document changes in the format.
14995 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
14999 * dwarf2out.c (struct dw_line_info_table): Update comments for
15001 (FORCE_RESET_NEXT_VIEW): New.
15002 (FORCE_RESETTING_VIEW_P): New.
15003 (RESETTING_VIEW_P): Check for -1 too.
15004 (ZERO_VIEW_P): Likewise.
15005 (new_line_info_table): Force-reset next view.
15006 (dwarf2out_begin_function): Likewise.
15007 (dwarf2out_source_line): Simplify zero_view_p initialization.
15008 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15009 view directly. Omit view when omitting .loc at line 0.
15011 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15013 PR tree-optimization/84740
15014 * tree-switch-conversion.c (process_switch): Call build_constructors
15015 only if info.phi_count is non-zero.
15017 PR tree-optimization/84739
15018 * tree-tailcall.c (find_tail_calls): Check call arguments against
15019 DECL_ARGUMENTS (current_function_decl) rather than
15020 DECL_ARGUMENTS (func) when checking for tail recursion.
15022 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15024 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15025 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15026 Volker Reichelt's entry and add entries for people that perform
15027 GCC fuzzy testing and report numerous bugs.
15029 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15032 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15033 readonly data in sdata, if that is disabled.
15034 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15035 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15036 -mreadonly-in-sdata option.
15038 2018-03-07 Martin Sebor <msebor@redhat.com>
15040 PR tree-optimization/84468
15041 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15042 basic block when looking for nul assignment.
15044 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15047 * except.h (output_function_exception_table): Adjust prototype.
15048 * except.c (output_function_exception_table): Remove FNNAME parameter
15049 and add SECTION parameter. Ouput one part of the table at a time.
15050 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15051 the first part of the exception table and emit unwind directives.
15052 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15053 (i386_pe_seh_cold_init): Likewise.
15054 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15055 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15056 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15057 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15058 * config/i386/winnt.c: Include except.h.
15059 (struct seh_frame_state): Add reg_offset, after_prologue and
15060 in_cold_section fields.
15061 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15062 (i386_pe_seh_cold_init): New function.
15063 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15064 to seh->in_cold_section.
15065 (seh_emit_push): Record the offset of the push.
15066 (seh_emit_save): Record the offet of the save.
15067 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15068 Test seh->after_prologue to disregard the epilogue.
15069 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15070 (i386_pe_end_cold_function): New function.
15072 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15075 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15076 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15079 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15081 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15082 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15084 PR middle-end/84723
15085 * multiple_target.c: Include tree-inline.h and intl.h.
15086 (expand_target_clones): Diagnose and fail if node->definition and
15087 !tree_versionable_function_p (node->decl).
15089 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15091 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15093 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15094 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15095 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15097 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15100 * combine.c (try_combine): Use reg_or_subregno instead of handling
15101 just paradoxical SUBREGs and REGs.
15103 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15105 * config/arc/arc.c (arc_finalize_pic): Remove function.
15106 (arc_must_save_register): We use single base PIC register, remove
15107 checks to save/restore the PIC register.
15108 (arc_expand_prologue): Likewise.
15109 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15111 (arc_verify_short): Likewise.
15112 (arc_attr_type): Likewise.
15113 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15114 (walk_stores): Likewise.
15115 (arc_address_cost): Make it static.
15116 (arc_verify_short): Likewise.
15117 (branch_dest): Likewise.
15118 (arc_attr_type): Likewise.
15119 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15120 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15121 (arc_final_prescan_insn): Remove inserting the nops due to
15122 hardware hazards. It is done in reorg step.
15123 (insn_length_variant_t): Remove.
15124 (insn_length_parameters_t): Likewise.
15125 (arc_insn_length_parameters): Likewise.
15126 (arc_get_insn_variants): Likewise.
15127 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15129 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15131 PR inline-asm/84683
15132 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15135 PR tree-optimization/84687
15136 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15138 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15140 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15142 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15143 Rename to ppc_speculation_barrier.
15144 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15145 __builtin_ppc_speculation_barrier.
15147 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15150 * combine.c (combine_simplify_rtx): Don't try to simplify if
15151 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15154 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15156 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15157 to 32 bytes when compiling for POWER9.
15159 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15162 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15163 regparm >= 3 with no arg reg available also for calls with
15164 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15167 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15169 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15171 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15174 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15176 2018-03-05 Richard Biener <rguenther@suse.de>
15178 PR tree-optimization/84486
15179 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15180 When inserting a __builtin_assume_aligned call set the LHS
15181 SSA name alignment info accordingly.
15183 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15185 PR tree-optimization/84114
15186 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15187 Avoid reassociation of FLOAT_MODE addition.
15189 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15191 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15192 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15193 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15194 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15195 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15196 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15197 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15199 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15200 __builtin_ia32_wbinvd): New builtins.
15201 (SPECIAL_ARGS2): New.
15202 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15203 (SPECIAL_ARGS2): New.
15204 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15205 (ix86_valid_target_attribute_inner_p): Ditto.
15206 (ix86_init_mmx_sse_builtins): Add special_args2.
15207 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15208 TARGET_WBNOINVD_P): New.
15209 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15210 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15211 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15212 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15213 * config/i386/pconfigintrin.h: New file.
15214 * config/i386/wbnoinvdintrin.h: Ditto.
15215 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15217 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15219 2018-03-05 Richard Biener <rguenther@suse.de>
15221 PR tree-optimization/84670
15222 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15224 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15225 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15226 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15227 doesn't grow if all successors (recursively) were visited at least
15230 2018-03-05 Richard Biener <rguenther@suse.de>
15232 PR tree-optimization/84650
15233 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15234 if executed in the loop pipeline.
15236 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15238 * doc/configfiles.texi (Configuration Files): Move info about
15239 conditionalizing $target-protos.h to...
15240 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15241 differs from $target-protos.h.
15243 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15244 Chung-Ju Wu <jasonwucj@gmail.com>
15246 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15247 * config/nds32/nds32-multiple.md (setmemsi): Define.
15248 * config/nds32/nds32-memory-manipulation.c
15249 (nds32_gen_dup_4_byte_to_word_value): New.
15250 (emit_setmem_word_loop): New.
15251 (emit_setmem_byte_loop): New.
15252 (nds32_expand_setmem_loop): New.
15253 (nds32_expand_setmem_loop_v3m): New.
15254 (nds32_expand_setmem_unroll): New.
15255 (nds32_expand_setmem): New.
15257 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15258 Chung-Ju Wu <jasonwucj@gmail.com>
15260 * config/nds32/nds32-memory-manipulation.c
15261 (nds32_emit_load_store): New.
15262 (nds32_emit_post_inc_load_store): New.
15263 (nds32_emit_mem_move): New.
15264 (nds32_emit_mem_move_block): New.
15265 (nds32_expand_movmemsi_loop_unknown_size): New.
15266 (nds32_expand_movmemsi_loop_known_size): New.
15267 (nds32_expand_movmemsi_loop): New.
15268 (nds32_expand_movmemsi_unroll): New.
15269 (nds32_expand_movmemqi): Rename ...
15270 (nds32_expand_movmemsi): ... to this.
15271 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15272 (movmemsi): ... to this.
15273 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15274 (nds32_expand_movmemsi): ... to this.
15276 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15277 Monk Chiang <sh.chiang04@gmail.com>
15278 Chung-Ju Wu <jasonwucj@gmail.com>
15280 * config/nds32/nds32-protos.h
15281 (nds32_expand_load_multiple): New arguments.
15282 (nds32_expand_store_multiple): Ditto.
15283 (nds32_valid_multiple_load_store): Rename ...
15284 (nds32_valid_multiple_load_store_p): ... to this.
15285 * config/nds32/nds32-memory-manipulation.c
15286 (nds32_expand_load_multiple): Refine implementation.
15287 (nds32_expand_store_multiple): Ditto.
15288 * config/nds32/nds32-multiple.md
15289 (load_multiple): Update nds32_expand_load_multiple interface.
15290 (store_multiple): Update nds32_expand_store_multiple interface.
15291 * config/nds32/nds32-predicates.c
15292 (nds32_valid_multiple_load_store): Rename ...
15293 (nds32_valid_multiple_load_store_p): ... to this and refine
15295 * config/nds32/predicates.md
15296 (nds32_load_multiple_and_update_address_operation): New predicate.
15297 (nds32_store_multiple_and_update_address_operation): New predicate.
15299 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15300 Chung-Ju Wu <jasonwucj@gmail.com>
15302 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15303 (combo): New attribute.
15304 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15306 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15308 * config/nds32/nds32.opt: Change -mcmodel= default value.
15310 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15311 Monk Chiang <sh.chiang04@gmail.com>
15312 Chung-Ju Wu <jasonwucj@gmail.com>
15314 * config/nds32/constants.md (unspec_element): New enum.
15315 * config/nds32/constraints.md (Umw): New constraint.
15316 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15317 * config/nds32/nds32-intrinsic.md: Likewise.
15318 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15319 (nds32_valid_smw_lwm_base_p): New.
15320 (nds32_output_smw_single_word): New.
15321 (nds32_output_lmw_single_word): New.
15322 (nds32_expand_unaligned_load): New.
15323 (nds32_expand_unaligned_store): New.
15324 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15325 (nds32_output_smw_single_word): Declare.
15326 (nds32_output_lmw_single_word): Declare.
15327 (nds32_expand_unaligned_load): Declare.
15328 (nds32_expand_unaligned_store): Declare.
15329 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15330 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15331 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15332 NDS32_BUILTIN_UASTORE_DW.
15333 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15336 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15337 Kito Cheng <kito.cheng@gmail.com>
15338 Chung-Ju Wu <jasonwucj@gmail.com>
15340 * config/nds32/nds32-intrinsic.c
15341 (nds32_expand_builtin_null_ftype_reg): Delete.
15342 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15343 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15344 (nds32_read_argument): New.
15345 (nds32_legitimize_target): Ditto.
15346 (nds32_legitimize_argument): Ditto.
15347 (nds32_check_constant_argument): Ditto.
15348 (nds32_expand_unop_builtin): Ditto.
15349 (nds32_expand_unopimm_builtin): Ditto.
15350 (nds32_expand_binop_builtin): Ditto.
15351 (nds32_builtin_decl_impl): Ditto.
15352 (builtin_description): Ditto.
15353 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15354 (nds32_init_builtins_impl): Ditto.
15355 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15356 (nds32_builtin_decl): New.
15357 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15358 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15360 2018-03-02 Jeff Law <law@redhat.com>
15362 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15363 (redundant_insn, fill_simple_delay_slots): Likewise.
15364 (fill_slots_from_thread): Likewise.
15365 * resource.c (mark_referenced_resources): Likewise.
15366 (mark_set_resources, find_dead_or_set_registers): Likewise.
15368 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15370 * substring-locations.h (format_warning_va): Formatting fix for
15371 ATTRIBUTE_GCC_DIAG.
15372 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15374 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15375 * substring-locations.c: Include intl.h.
15376 (format_warning_va): Turned into small wrapper around
15377 format_warning_n_va, renamed to ...
15378 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15379 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15381 (format_warning_at_substring_n): New function.
15382 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15383 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15384 format_warning_at_substring with just a shorter name instead of
15385 const function pointer.
15386 (fmtwarn_n): New function.
15387 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15388 appropriate, get rid of all the fmtstr temporaries, move conditionals
15389 with G_() wrapped string literals directly into fmtwarn arguments,
15390 cast dir.len to (int), formatting fixes.
15392 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15394 * doc/invoke.texi: Remove "Cilk Plus" references.
15396 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15397 Richard Biener <rguenther@suse.de>
15400 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15401 for error or warning attributes if CALL_FROM_THUNK_P is set.
15404 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15407 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15408 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15411 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15412 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15414 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15415 instead of -1U in last predictors element's probability member.
15417 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15420 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15421 arguments if they are comparable.
15423 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15425 PR tree-optimization/84634
15426 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15427 masks and masked_loop_p with a single loop_masks, making sure it's
15428 null for bb vectorization.
15430 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15432 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15433 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15434 loop->force_vectorize to check whether there is no alias.
15436 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15439 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15441 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15443 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15444 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15445 instead of a loop around prev_real_insn.
15446 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15449 PR inline-asm/84625
15450 * config/i386/i386.c (ix86_print_operand): Use conditional
15451 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15454 2018-03-02 Richard Biener <rguenther@suse.de>
15456 PR tree-optimization/84427
15457 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15458 (bitmap_set_subtract_values): Rewrite to handle multiple
15461 (prune_clobbered_mems): Likewise.
15462 (phi_translate): Take edge instead of pred/phiblock.
15463 (phi_translate_1): Likewise.
15464 (phi_translate_set): Likewise. Insert all translated
15465 exprs for a value into the set, keeping possibly multiple
15466 expressions per value.
15467 (compute_antic_aux): Adjust for phi_translate changes.
15468 When intersecting union the expressions and prune those
15469 not in the final value set, keeping possibly multiple
15470 expressions per value. Do not use value-insertion
15471 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15472 all expressions. Add verification that the value-sets
15473 only shrink during iteration.
15474 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15475 (do_pre_regular_insertion): Likewise.
15476 (do_pre_partial_partial_insertion): Likewise.
15478 2018-03-02 Richard Biener <rguenther@suse.de>
15481 * config/darwin.c (saved_debug_info_level): New static global.
15482 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15483 (darwin_asm_lto_end): Restore debug info generation settings.
15485 2018-03-01 Martin Liska <mliska@suse.cz>
15488 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15489 volatile arguments.
15491 2018-03-01 Richard Biener <rguenther@suse.de>
15494 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15495 type annotation with LTO when debug was disabled at compile-time.
15497 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15499 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15501 (mips_final_postscan_insn): Likewise.
15503 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15505 PR rtl-optimization/84528
15506 * alias.c (init_alias_target): Add commentary.
15507 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15508 a unique base value if the frame pointer is not eliminated
15509 to the stack pointer.
15511 2018-03-01 Tom de Vries <tom@codesourcery.com>
15513 PR rtl-optimization/83327
15514 * lra-int.h (hard_regs_spilled_into): Declare.
15515 * lra.c (hard_regs_spilled_into): Define.
15516 (init_reg_info): Init hard_regs_spilled_into.
15517 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15518 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15519 (process_bb_lives): Handle hard_regs_spilled_into.
15520 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15521 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15523 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15525 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15526 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15527 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15528 * config/rs6000/aix72.h: New file.
15530 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15532 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15533 instead of warning_at with conditional singular and plural messages
15537 * stor-layout.c (update_alignment_for_field): For
15538 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15539 && !DECL_PACKED (field), do the alignment update, just use
15540 only desired_align instead of MAX (type_align, desired_align)
15542 (place_field): Don't do known_align < desired_align handling
15543 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15544 is non-NULL, instead do it after rli->prev_field handling and
15545 only if not within a bitfield word. For DECL_PACKED (field)
15546 use type_align of BITS_PER_UNIT.
15548 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15550 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15551 superfluous parentheses and trailing spaces.
15553 2018-02-28 Richard Biener <rguenther@suse.de>
15555 PR tree-optimization/84584
15556 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15557 SCoPs with fake exit edge.
15559 2018-02-28 Martin Liska <mliska@suse.cz>
15562 * timevar.c (timer::print): Fix format to properly print 100%
15565 2018-02-28 Richard Biener <rguenther@suse.de>
15567 PR middle-end/84607
15568 * genmatch.c (capture_info::walk_match): Do not mark
15569 captured expressions without operands as expr_p given
15570 they act more like predicates and should be subject to
15571 "lost tail" side-effect preserving.
15573 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15575 PR rtl-optimization/81611
15576 * auto-inc-dec.c (attempt_change): Move dead note from
15577 mem_insn if it's the next use of regno
15578 (find_address): Take address use of reg holding
15579 non-incremented value. Add parm to limit search to the named
15581 (merge_in_block): Attempt to use a mem insn that is the next
15582 use of the original regno.
15584 2018-02-27 Martin Sebor <msebor@redhat.com>
15587 * doc/invoke.texi (-Wmissing-attributes): New option.
15588 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15590 2018-02-27 Martin Sebor <msebor@redhat.com>
15592 PR translation/84207
15593 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15594 n argument to unsigned HOST_WIDE_INT.
15595 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15596 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15597 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15598 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15600 2018-02-27 Richard Biener <rguenther@suse.de>
15602 PR tree-optimization/84512
15603 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15604 Do not use the estimate returned from record_stmt_cost for
15605 the scalar iteration cost but sum properly using add_stmt_cost.
15607 2018-02-27 Richard Biener <rguenther@suse.de>
15609 PR tree-optimization/84466
15610 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15611 Adjust last change to less strictly validate use operands.
15613 2018-02-27 Martin Liska <mliska@suse.cz>
15615 PR gcov-profile/84548
15616 * gcov.c (process_file): Allow partial overlap and consider it
15617 also as group functions.
15618 (output_lines): Properly calculate range of lines for a group.
15620 2018-02-27 Martin Liska <mliska@suse.cz>
15622 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15623 'ggc' suffixes. Change first column width.
15624 (timer::print): Fix formatting of the column.
15626 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15628 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15629 preserve inline entry blocks for the sake of debug inline
15630 entry point markers alone.
15631 (remove_unused_locals): Suggest in comments a better place to
15632 force the preservation of inline entry blocks that are
15633 otherwise unused, but do not preserve them.
15635 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15637 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15639 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15642 * config/i386/constraints.md (Bs): Replace
15643 ix86_indirect_branch_register with
15644 TARGET_INDIRECT_BRANCH_REGISTER.
15646 * config/i386/i386.md (indirect_jump): Likewise.
15647 (tablejump): Likewise.
15648 (*sibcall_memory): Likewise.
15649 (*sibcall_value_memory): Likewise.
15650 Peepholes of indirect call and jump via memory: Likewise.
15651 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15652 (*sibcall_value_GOT_32): Likewise.
15653 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15654 (GOT_memory_operand): Likewise.
15655 (call_insn_operand): Likewise.
15656 (sibcall_insn_operand): Likewise.
15657 (GOT32_symbol_operand): Likewise.
15658 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15660 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15662 PR rtl-optimization/83496
15663 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15664 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15665 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15666 redundant insn, if any.
15667 (relax_delay_slots): Likewise.
15668 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15670 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15672 PR tree-optimization/83965
15673 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15674 that grouped statements are part of a reduction chain. Return
15675 true if the statement is not marked as a reduction itself but
15676 is part of a group.
15677 (vect_recog_dot_prod_pattern): Don't check whether the statement
15678 is part of a group here.
15679 (vect_recog_sad_pattern): Likewise.
15680 (vect_recog_widen_sum_pattern): Likewise.
15682 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15685 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15687 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15690 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15692 (ix86_output_indirect_function_return): New prototype.
15693 (ix86_split_simple_return_pop_internal): Likewise.
15694 * config/i386/i386.c (indirect_return_via_cx): New.
15695 (indirect_return_via_cx_bnd): Likewise.
15696 (indirect_thunk_name): Handle return va CX_REG.
15697 (output_indirect_thunk_function): Create alias for
15698 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15699 (ix86_output_indirect_jmp): Remove the bool argument.
15700 (ix86_output_indirect_function_return): New function.
15701 (ix86_split_simple_return_pop_internal): Likewise.
15702 * config/i386/i386.md (*indirect_jump): Don't pass false
15703 to ix86_output_indirect_jmp.
15704 (*tablejump_1): Likewise.
15705 (simple_return_pop_internal): Change it to define_insn_and_split.
15706 Call ix86_split_simple_return_pop_internal to split it for
15707 -mfunction-return=.
15708 (simple_return_indirect_internal): Call
15709 ix86_output_indirect_function_return instead of
15710 ix86_output_indirect_jmp.
15712 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15715 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15716 memset and value initialization afterwards.
15718 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15720 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15722 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15725 * common/config/aarch64/aarch64-common.c
15726 (aarch_option_optimization_table[]): Switch
15727 off fomit-frame-pointer
15729 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15730 Chung-Ju Wu <jasonwucj@gmail.com>
15732 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15734 (store_multiple): Ditto.
15736 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15738 * config.gcc: Add --with-cpu support for nds32 target.
15739 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15740 * config/nds32/nds32.opt: Add -mcpu= option.
15742 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15744 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15745 isel=yes): Warn for these deprecated options.
15747 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15749 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15750 ISA_2_5_MASKS_EMBEDDED.
15752 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15754 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15755 p->max as pointers rather than using iterative_hash_expr.
15757 2018-02-23 Carl Love <cel@us.ibm.com>
15759 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15760 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15762 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15763 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15764 P8V_BUILTIN_VEC_VUNSIGNED2.
15766 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15769 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15770 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15772 * lra-constraints.c (curr_insn_transform): Set up
15773 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15775 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15776 * lra-eliminations.c (spill_pseudos): Ditto.
15777 (process_insn_for_elimination): Ditto.
15778 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15779 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15780 LRA_NON_CLOBBERED_ALT.
15782 2018-02-22 Martin Sebor <msebor@redhat.com>
15784 PR tree-optimization/84480
15785 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15786 to maybe_diag_stxncpy_trunc. Call it.
15787 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15788 from gimple_fold_builtin_strcpy. Print inlining stack.
15789 (handle_builtin_stxncpy): Print inlining stack.
15790 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15792 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15795 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15796 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15797 and -fcheck-pointer-bounds are used together.
15798 (indirect_thunk_prefix): New enum.
15799 (indirect_thunk_need_prefix): New function.
15800 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15801 "_nt" instead of "_bnd" for NOTRACK prefix.
15802 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15803 (output_indirect_thunk_function): Likewise.
15805 (ix86_code_end): Update output_indirect_thunk_function calls.
15806 (ix86_output_indirect_branch_via_reg): Replace
15807 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15808 (ix86_output_indirect_branch_via_push): Likewise.
15809 (ix86_output_function_return): Likewise.
15810 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15811 incompatible with -fcf-protection=branch and
15812 -fcheck-pointer-bounds.
15814 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15817 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15818 Change gcc_assert call to output_operand_lossage.
15820 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15822 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15824 2018-02-22 DJ Delorie <dj@redhat.com>
15825 Sebastian Perta <sebastian.perta@renesas.com>
15826 Oleg Endo <olegendo@gcc.gnu.org>
15828 * config/rx/rx.c (rx_rtx_costs): New function.
15829 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15831 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15833 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15835 2018-02-22 Martin Liska <mliska@suse.cz>
15838 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15839 Add "native" as a possible value.
15841 2018-02-22 Martin Liska <mliska@suse.cz>
15844 * config/i386/i386.c (ix86_option_override_internal):
15845 Add "native" as a possible value for -march and -mtune.
15847 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15850 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15851 to all type variants.
15853 PR tree-optimization/84503
15854 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15855 width as info->bitpos + info->bitsize - start.
15856 (merged_store_group::merge_overlapping): Simplify width computation.
15857 (check_no_overlap): New function.
15858 (imm_store_chain_info::try_coalesce_bswap): Compute expected
15859 start + width and last_order of the group, fail if check_no_overlap
15861 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15862 to group if check_no_overlap fails.
15864 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15866 * config/rs6000/altivec.md: Delete contraint arguments to
15867 define_expand, define_split, and define_peephole2, and in
15868 define_insn_and_split if always unused.
15869 * config/rs6000/darwin.md: Ditto.
15870 * config/rs6000/dfp.md: Ditto.
15871 * config/rs6000/rs6000.md: Ditto.
15872 * config/rs6000/sync.md: Ditto.
15873 * config/rs6000/vector.md: Ditto.
15874 * config/rs6000/vsx.md: Ditto.
15876 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15878 * config/rs6000/altivec.md: Write output control strings as braced
15879 blocks instead of double-quoted strings.
15880 * config/rs6000/darwin.md: Ditto.
15881 * config/rs6000/rs6000.md: Ditto.
15882 * config/rs6000/vector.md: Ditto.
15883 * config/rs6000/vsx.md: Ditto.
15885 2018-02-21 Jason Merrill <jason@redhat.com>
15887 PR c++/84314 - ICE with templates and fastcall attribute.
15888 * attribs.c (build_type_attribute_qual_variant): Remove assert.
15890 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15892 * ipa-cp.c (determine_versionability): Fix comment typos.
15894 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15897 * ipa-cp.c (determine_versionability): Do not version functions caling
15900 2018-02-21 Martin Liska <mliska@suse.cz>
15903 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15904 Add "native" as a possible value.
15905 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
15906 the macro when native cpu detection is available.
15908 2018-02-21 Martin Liska <mliska@suse.cz>
15911 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15912 Add "native" as a possible value.
15913 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15914 when native cpu detection is available.
15916 2018-02-21 Jakub Jelinek <jakub@redhat.com>
15917 Martin Sebor <msebor@redhat.com>
15919 PR tree-optimization/84478
15920 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15922 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15923 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15924 support which is conservatively correct, for 2 only stay conservative
15925 for maxlen. Formatting and comment capitalization fixes. Add STRICT
15926 argument to the 2 argument get_range_strlen, adjust 6 arg
15927 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15929 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15930 (gimple_fold_builtin_strlen): Pass true as last argument to
15933 2018-02-20 Martin Sebor <msebor@redhat.com>
15935 PR middle-end/84095
15936 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15937 (builtin_memref::set_base_and_offset): Same. Handle inner references.
15938 (builtin_memref::builtin_memref): Factor out parts into
15939 set_base_and_offset and call it.
15941 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15943 PR middle-end/84406
15944 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15945 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15946 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
15947 search at the associated MODE_INT.
15949 2018-02-20 Jeff Law <law@redhat.com>
15951 PR middle-end/82123
15952 PR tree-optimization/81592
15953 PR middle-end/79257
15954 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15955 for range data rather than using global data.
15956 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15957 range data rather than using global data.
15958 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15959 pass it to children as needed.
15960 (struct directive::fmtresult): Similarly.
15961 (struct directive::set_width): Similarly.
15962 (struct directive::set_precision): Similarly.
15963 (format_integer, format_directive, parse_directive): Similarly.
15964 (format_none): Accept unnamed vr_values parameter.
15965 (format_percent, format_floating, format_character): Similarly.
15966 (format_string, format_plain): Similarly.
15967 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
15968 the EVRP range analyzer for range data rather than using global data.
15969 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
15970 gimple-ssa-evrp-analyze.h
15971 (class sprintf_dom_walker): Add after_dom_children member function.
15972 Add evrp_range_analyzer member.
15973 (sprintf_dom_walker::before_dom_children): Call into the EVRP
15974 range analyzer as needed.
15975 (sprintf_dom_walker::after_dom_children): New member function.
15976 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
15978 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
15979 (evrp_range_analyzer::pop_to_marker): Likewise.
15981 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15983 PR tree-optimization/84419
15984 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
15985 with the required type if its current type is compatible but
15988 2018-02-20 Jakub Jelinek <jakub@redhat.com>
15990 PR middle-end/82004
15991 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15992 after vectorization.
15994 2018-02-20 Martin Liska <mliska@suse.cz>
15997 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15998 possible values if we don't have a hint.
16000 2018-02-20 Martin Liska <mliska@suse.cz>
16004 * final.c (shorten_branches): Build align_tab array with one
16006 * opts.c (finish_options): Add alignment option limit check.
16007 (MAX_CODE_ALIGN): Likewise.
16008 (MAX_CODE_ALIGN_VALUE): Likewise.
16009 * doc/invoke.texi: Document maximum allowed option value for
16010 all -falign-* options.
16012 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16015 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16016 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16017 * var-tracking.c (emit_note_insn_var_location): Remove all references
16018 to NOTE_INSN_CALL_ARG_LOCATION.
16019 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16020 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16021 Use copy_rtx_if_shared.
16022 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16023 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16024 (dwarf2out_var_location): Remove handling of
16025 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16027 * final.c (final_scan_insn): Remove all references to
16028 NOTE_INSN_CALL_ARG_LOCATION.
16029 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16030 before dumping final insns.
16031 * except.c (emit_note_eh_region_end): Remove all references to
16032 NOTE_INSN_CALL_ARG_LOCATION.
16033 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16034 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16035 * config/arc/arc.c (hwloop_optimize): Likewise.
16036 * config/arm/arm.c (create_fix_barrier): Likewise.
16037 * config/s390/s390.c (s390_chunkify_start): Likewise.
16038 * config/sh/sh.c (find_barrier): Likewise.
16039 * config/i386/i386.c (rest_of_insert_endbranch,
16040 ix86_seh_fixup_eh_fallthru): Likewise.
16041 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16042 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16043 * config/frv/frv.c (frv_function_prologue): Likewise.
16044 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16046 (note_outside_basic_block_p): Remove all references to
16047 NOTE_INSN_CALL_ARG_LOCATION.
16048 * gengtype.c (adjust_field_rtx_def): Likewise.
16049 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16051 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16052 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16055 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16058 PR tree-optimization/84452
16059 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16060 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16063 2018-02-19 Martin Liska <mliska@suse.cz>
16066 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16068 2018-02-19 Martin Liska <mliska@suse.cz>
16069 Richard Sandiford <richard.sandiford@linaro.org>
16071 PR tree-optimization/82491
16072 * gimple-fold.c (get_base_constructor): Make earlier bail out
16075 2018-02-19 Carl Love <cel@us.ibm.com>
16077 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16078 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16080 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16081 P8V_BUILTIN_VEC_NEG.
16083 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16085 * config/rl78/rl78.md (movdf): New define expand.
16087 2018-02-19 Martin Liska <mliska@suse.cz>
16090 * doc/invoke.texi: Fix typo.
16091 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16093 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16095 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16096 handle rs6000_single_float and rs6000_double_float specially for
16099 2018-02-16 Jeff Law <law@redhat.com>
16101 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16102 (pop_regs): Use it.
16104 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16107 * ipa-inline.c (inline_small_functions): Fix a typo.
16109 2018-02-16 Nathan Sidwell <nathan@acm.org>
16111 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16113 2018-02-16 Carl Love <cel@us.ibm.com>
16115 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16116 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16117 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16118 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16119 expansion to P8V_BUILTIN_VEC_FLOAT2.
16121 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16123 PR rtl-optimization/70023
16124 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16125 src_regno into account.
16127 2018-02-16 Carl Love <cel@us.ibm.com>
16129 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16130 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16131 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16132 * config/rs6000/rs6000.c: Remove case statements for
16133 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16134 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16135 and P9V_BUILTIN_VEC_VINSERT4B.
16136 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16137 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16138 * config/rs6000/vsx.md:
16139 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16142 2018-02-16 Carl Love <cel@us.ibm.com>
16144 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16146 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16148 * config/rs6000/rs6000-c.c: Add the definitions for
16149 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16150 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16151 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16152 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16153 definition for insert4b and define insn *insert3b_internal.
16154 * doc/extend.texi: Add documentation for vec_extract4b.
16156 2018-02-16 Nathan Sidwell <nathan@acm.org>
16158 * doc/extend.texi (Backwards Compatibility): Mention friend
16159 injection. Note for-scope is deprecated.
16160 * doc/invoke.texi (-ffriend-injection): Deprecate.
16162 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16164 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16165 that moved to I2, also allow destinations that are a paradoxical
16166 subreg (instead of a normal reg).
16168 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16171 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16174 2018-02-16 Richard Biener <rguenther@suse.de>
16176 PR tree-optimization/84037
16177 PR tree-optimization/84016
16179 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16180 Adjust vec_construct for the fact we need additional higher latency
16181 128bit inserts for AVX256 and AVX512 vector builds.
16182 (ix86_add_stmt_cost): Scale vector construction cost for
16185 2018-02-16 Richard Biener <rguenther@suse.de>
16187 PR tree-optimization/84417
16188 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16189 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16190 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16192 2018-02-16 Martin Liska <mliska@suse.cz>
16195 * internal-fn.def (ASAN_CHECK): Set proper flags.
16196 (ASAN_MARK): Likewise.
16198 2018-02-16 Julia Koval <julia.koval@intel.com>
16200 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16201 from PTA_CANNONLAKE.
16203 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16206 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16207 Use ++iter rather than iter++ for std::list iterators.
16208 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16209 defer deleting them until all nodes in the forest are processed. Do
16210 free even leaf nodes. Change to_process into auto_vec.
16213 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16214 * vec.h (vec_default_construct): Use memset instead of placement new
16215 if BROKEN_VALUE_INITIALIZATION is defined.
16216 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16217 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16220 PR rtl-optimization/83723
16221 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16222 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16223 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16225 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16227 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16229 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16231 PR rtl-optimization/81443
16232 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16233 from inner REGs to paradoxical SUBREGs.
16235 2018-02-16 Richard Biener <rguenther@suse.de>
16237 PR tree-optimization/84399
16238 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16239 For operands we can analyze at their definition make sure we can
16240 analyze them at each use as well.
16242 2018-02-16 Richard Biener <rguenther@suse.de>
16244 PR tree-optimization/84190
16245 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16246 volatile accesses if the decl isn't volatile.
16248 2018-02-15 Jason Merrill <jason@redhat.com>
16250 PR c++/84314 - ICE with templates and fastcall attribute.
16251 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16252 TYPE_CANONICAL on an existing type.
16254 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16256 PR tree-optimization/84383
16257 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16258 dstoff nor call operand_equal_p if dstbase is NULL.
16260 PR tree-optimization/84334
16261 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16262 also a CONSTANT_CLASS_P, punt.
16264 2018-02-14 Jim Wilson <jimw@sifive.com>
16266 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16267 first SMALL_OPERAND check. New local min_second_step. Move assert
16268 to where locals are set. Add TARGET_RVC support.
16269 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16271 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16273 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16275 2018-02-14 Martin Sebor <msebor@redhat.com>
16277 PR tree-optimization/83698
16278 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16279 arrays constrain the offset range to their bounds.
16280 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16281 (builtin_access::overlap): Avoid setting the size of overlap if it's
16283 (maybe_diag_overlap): Also consider arrays when deciding what values
16284 of offsets to include in diagnostics.
16286 2018-02-14 Martin Sebor <msebor@redhat.com>
16289 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16290 that correspond to the kind of a declaration.
16292 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16295 * config/pa/pa.md: Load address of PIC label using the linkage table
16296 if the label is nonlocal.
16298 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16300 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16301 warning message if user requests -maltivec=be.
16302 * doc/invoke.texi: Document deprecation of -maltivec=be.
16304 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16307 * config/rs6000/rs6000-c.c: Update definitions for
16308 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16309 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16311 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16314 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16315 add _get_ssp intrinsics. Remove argument from
16316 __builtin_ia32_rdssp[d|q].
16317 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16318 * config/i386/i386-builtin.def: Remove argument from
16319 __builtin_ia32_rdssp[d|q].
16320 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16321 ix86_expand_special_args_builtin for _rdssp[d|q].
16322 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16323 Clear register before usage.
16324 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16325 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16327 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16329 PR tree-optimization/84357
16330 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16331 operand 1 of an ARRAY_REF too.
16333 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16336 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16337 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16339 (set_of_reg): New struct.
16340 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16341 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16342 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16344 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16345 Split into bitclr, bitset, bitinvert patterns if appropriate.
16346 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16347 use rx_fuse_in_memory_bitop.
16348 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16349 to named insn, correct maximum insn length.
16351 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16354 * machmode.def: Define a complex mode for PARTIAL_INT.
16355 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16357 * doc/rtl.texi: Document CSPImode.
16358 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16360 (msp430_hard_regno_nregs_with_padding): Likewise.
16362 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16365 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16367 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16369 PR rtl-optimization/84169
16370 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16371 we generated a parallel as new i3 and we split that to new i2 and i3
16372 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16373 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16374 those to i2, not i1. Partially rewrite this scan code.
16376 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16379 * stor-layout.c (place_field): For variable length fields, adjust
16380 offset_align afterwards not just based on the field's alignment,
16381 but also on the size.
16383 PR middle-end/84309
16384 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16385 of exps and logs in the use_exp2 case.
16387 2018-02-13 Jeff Law <law@redhat.com>
16389 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16390 entry for "vector".
16392 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16395 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16399 * common.opt (gas-loc-support, gas-locview-support): New.
16400 (ginline-points, ginternal-reset-location-views): New.
16401 * doc/invoke.texi: Document them. Use @itemx where intended.
16402 (gvariable-location-views): Adjust.
16403 * target.def (reset_location_view): New.
16404 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16405 (TARGET_RESET_LOCATION_VIEW): New.
16406 * doc/tm.texi: Rebuilt.
16407 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16408 (dwarf2out_default_as_locview_support): New.
16409 (output_asm_line_debug_info): Use option variables.
16410 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16411 (output_loc_list): Likewise.
16412 (add_high_low_attributes): Check option variables.
16413 Don't output entry view attribute in strict mode.
16414 (gen_inlined_subroutine_die): Check option variables.
16415 (dwarf2out_inline_entry): Likewise.
16416 (init_sections_and_labels): Likewise.
16417 (dwarf2out_early_finish): Likewise.
16418 (maybe_reset_location_view): New, from...
16419 (dwarf2out_var_location): ... here. Call it.
16420 * debug.h (dwarf2out_default_as_loc_support): Declare.
16421 (dwarf2out_default_as_locview_support): Declare.
16422 * hooks.c (hook_int_rtx_insn_0): New.
16423 * hooks.h (hook_int_rtx_insn_0): Declare.
16424 * toplev.c (process_options): Take -gas-loc-support and
16425 -gas-locview-support from dwarf2out. Enable
16426 -gvariable-location-views by default only with locview
16427 assembler support. Enable -ginternal-reset-location-views by
16428 default only if the target defines the corresponding hook.
16429 Enable -ginline-points by default if location views are
16430 enabled; force it disabled if statement frontiers are
16432 * tree-inline.c (expand_call_inline): Check option variables.
16433 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16435 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16437 PR tree-optimization/84321
16438 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16439 handling. Also check whether the anti-range contains any values
16440 that satisfy the mask; switch to a VR_RANGE if not.
16442 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16445 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16447 2018-02-13 Martin Jambor <mjambor@suse.cz>
16450 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16451 of call statements, also set location of a load to a temporary.
16453 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16455 * config/rl78/rl78.c (add_vector_labels): New function.
16456 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16457 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16458 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16459 which checks that no arguments are passed.
16460 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16461 * doc/extend.texi: Documentation for the new attribute.
16463 2018-02-13 Andreas Schwab <schwab@suse.de>
16465 * config/riscv/linux.h (CPP_SPEC): Define.
16467 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16470 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16471 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16472 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16473 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16474 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16475 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16476 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16477 temporarily for AES and PCLMUL builtins.
16479 PR tree-optimization/84339
16480 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16481 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16484 PR middle-end/84309
16485 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16486 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16487 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16489 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16491 * omp-simd-clone.h: New file.
16492 * omp-simd-clone.c: Include omp-simd-clone.h.
16493 (expand_simd_clones): No longer static.
16494 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16495 cgraph.h and omp-simd-clone.h.
16496 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16497 (vect_recog_widen_shift_pattern): Formatting fix.
16498 (vect_pattern_recog_1): Don't check optab for calls.
16501 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16502 operands[2] into a REG before using gen_lowpart on it.
16504 2018-02-12 Jeff Law <law@redhat.com>
16507 * config/sh/sh.c (find_barrier): Consider a sibling call
16510 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16511 successfully back substituting a reg.
16513 2018-02-12 Richard Biener <rguenther@suse.de>
16515 PR tree-optimization/84037
16516 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16517 parameter, move visited init to caller.
16518 (vect_slp_analyze_operations): Separate cost from validity
16519 check, initialize visited once for all instances.
16520 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16522 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16524 (vect_model_promotion_demotion_cost): Likewise.
16525 (vectorizable_bswap): Guard cost modeling with !slp_node
16526 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16528 (vectorizable_call): Likewise.
16529 (vectorizable_conversion): Likewise.
16530 (vectorizable_assignment): Likewise.
16531 (vectorizable_shift): Likewise.
16532 (vectorizable_operation): Likewise.
16533 (vectorizable_store): Likewise.
16534 (vectorizable_load): Likewise.
16535 (vectorizable_condition): Likewise.
16536 (vectorizable_comparison): Likewise.
16538 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16541 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16542 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16543 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16545 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16547 PR middle-end/83665
16548 * params.def (inline-min-speedup): Increase from 8 to 15.
16549 (max-inline-insns-auto): Decrease from 40 to 30.
16550 * ipa-split.c (consider_split): Add some buffer for function to
16551 be considered inlining candidate.
16552 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16555 2018-02-12 Richard Biener <rguenther@suse.de>
16557 PR tree-optimization/84037
16558 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16559 matched stmts if we cannot swap the non-matched ones.
16561 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16563 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16564 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16565 _mm_maskz_scalef_round_ss): New intrinsics.
16566 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16567 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16568 __builtin_ia32_scalefss_round): Remove.
16569 (__builtin_ia32_scalefsd_mask_round,
16570 __builtin_ia32_scalefss_mask_round): New intrinsics.
16571 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16572 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16573 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16574 "<round_constraint>")): Changed to ...
16575 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16576 "<round_scalar_constraint>")): ... this.
16577 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16578 %0, %1, %2<round_op3>}"): Changed to ...
16579 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16580 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16581 %2<round_scalar_mask_op3>}"): ... this.
16582 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16584 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16586 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16587 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16588 (_mm_maskz_sqrt_round_ss): New intrinsics.
16589 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16590 (__builtin_ia32_sqrtsd_mask_round)
16591 (__builtin_ia32_sqrtss_mask_round): New builtins.
16592 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16593 (__builtin_ia32_sqrtss_round): Remove.
16594 (__builtin_ia32_sqrtsd_mask_round)
16595 (__builtin_ia32_sqrtss_mask_round): New builtins.
16596 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16597 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16598 ((match_operand:VF_128 1 "vector_operand"
16599 "xBm,<round_constraint>")): Changed to ...
16600 ((match_operand:VF_128 1 "vector_operand"
16601 "xBm,<round_scalar_constraint>")): ... this.
16602 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16603 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16604 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16605 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16606 %<iptr>1<round_scalar_mask_op3>}): ... this.
16607 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16608 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16610 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16613 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16614 Cast vec_cmpeq result to correct type.
16615 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16616 Cast vec_cmpgt result to correct type.
16618 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16620 * final.c (final_scan_insn_1): Renamed from...
16621 (final_scan_insn): ... this. New wrapper, to recover
16622 seen from the outermost call in recursive ones.
16623 * config/sparc/sparc.c (output_return): Drop seen from call.
16624 (output_sibcall): Likewise.
16625 * config/visium/visium.c (output_branch): Likewise.
16627 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16629 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16632 2018-02-10 Alan Modra <amodra@gmail.com>
16635 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16636 Specify LR as an input.
16638 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16641 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16642 remove_member_access_dummy_vars): New functions.
16643 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16644 lower_omp_1, execute_lower_omp): Use them.
16646 PR rtl-optimization/84308
16647 * shrink-wrap.c (spread_components): Release todo vector.
16649 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16651 PR rtl-optimization/57193
16652 * ira-color.c (struct allocno_color_data): Add member
16653 conflict_allocno_hard_prefs.
16654 (update_conflict_allocno_hard_prefs): New.
16655 (bucket_allocno_compare_func): Add a preference based on
16656 conflict_allocno_hard_prefs.
16657 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16658 (color_allocnos): Remove a dead code. Initiate
16659 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16661 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16664 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16665 constraint from =wa to wa. Avoid a subreg on the output operand,
16666 instead use a pseudo and subreg it in a move.
16667 (p9_xxbrd_<mode>): Changed to ...
16668 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16669 (p9_xxbrd_v2df): New expander.
16670 (p9_xxbrw_<mode>): Changed to ...
16671 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16672 (p9_xxbrw_v4sf): New expander.
16674 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16676 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16678 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16681 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16682 multiply in 32-bit mode.
16683 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16684 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16687 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16689 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16690 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16691 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16692 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16694 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16697 * dwarf2out.c (is_trivial_indirect_ref): New function.
16698 (dwarf2out_late_global_decl): Do not generate a location
16699 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16700 and that are not defined in the current unit.
16702 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16704 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16705 instead of a libcall for UNORDERED.
16707 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16710 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16711 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16713 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16716 * config/s390/s390.c (s390_set_current_function): Invoke
16717 s390_indirect_branch_settings also if fndecl didn't change.
16719 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16721 * config/rs6000/rs6000.md (blockage): Set length to zero.
16723 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16725 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16727 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16730 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16731 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16735 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16736 PARALLEL incoming that failed vt_get_decl_and_offset check.
16738 PR middle-end/84237
16739 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16740 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16742 (get_variable_section): For decls in named .bss* sections pass true as
16743 second argument to bss_initializer_p.
16745 2018-02-09 Marek Polacek <polacek@redhat.com>
16746 Jakub Jelinek <jakub@redhat.com>
16749 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16750 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16751 Sync some changes from cxx_fold_indirect_ref.
16753 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16755 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16757 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16758 (BLOCK_INLINE_ENTRY_LABEL): New.
16759 (dwarf2out_var_location): Disregard inline entry markers.
16760 (inline_entry_data): New struct.
16761 (inline_entry_data_hasher): New hashtable type.
16762 (inline_entry_data_hasher::hash): New.
16763 (inline_entry_data_hasher::equal): New.
16764 (inline_entry_data_table): New variable.
16765 (add_high_low_attributes): Add DW_AT_entry_pc and
16766 DW_AT_GNU_entry_view attributes if a pending entry is found
16767 in inline_entry_data_table. Add old entry_pc attribute only
16768 if debug nonbinding markers are disabled.
16769 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16770 markers are enabled.
16771 (block_within_block_p, dwarf2out_inline_entry): New.
16772 (dwarf2out_finish): Check that no entries remained in
16773 inline_entry_data_table.
16774 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16775 (final_scan_insn, notice_source_line): Likewise.
16776 (rest_of_clean_state): Skip inline entry markers.
16777 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16779 * gimple.c (gimple_build_debug_inline_entry): New.
16780 * gimple.h (enum gimple_debug_subcode): Add
16781 GIMPLE_DEBUG_INLINE_ENTRY.
16782 (gimple_build_debug_inline_entry): Declare.
16783 (gimple_debug_inline_entry_p): New.
16784 (gimple_debug_nonbind_marker_p): Adjust.
16785 * insn-notes.def (INLINE_ENTRY): New.
16786 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16787 inline entry marker notes.
16788 (print_insn): Likewise.
16789 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16790 (INSN_DEBUG_MARKER_KIND): Likewise.
16791 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16792 * tree-inline.c (expand_call_inline): Build and insert
16793 debug_inline_entry stmt.
16794 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16795 inline entry blocks early, if nonbind markers are enabled.
16796 (dump_scope_block): Dump fragment info.
16797 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16798 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16799 (gimple_build_debug_inline_entry): New.
16800 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16801 Enable/disable inline entry points too.
16802 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16803 (DEBUG_INSN): Describe inline entry markers.
16805 * common.opt (gvariable-location-views): New.
16806 (gvariable-location-views=incompat5): New.
16807 * config.in: Rebuilt.
16808 * configure: Rebuilt.
16809 * configure.ac: Test assembler for view support.
16810 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16811 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16812 * dwarf2out.c (var_loc_view): New typedef.
16813 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16814 (dwarf2out_locviews_in_attribute): New.
16815 (dwarf2out_locviews_in_loclist): New.
16816 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16817 (enum dw_line_info_opcode): Add LI_adv_address.
16818 (struct dw_line_info_table): Add view.
16819 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16820 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16821 (zero_view_p): New variable.
16822 (ZERO_VIEW_P): New macro.
16823 (output_asm_line_debug_info): New.
16824 (struct var_loc_node): Add view.
16825 (add_AT_view_list, AT_loc_list): New.
16826 (add_var_loc_to_decl): Add view param. Test it against last.
16827 (new_loc_list): Add view params. Record them.
16828 (AT_loc_list_ptr): Handle loc and view lists.
16829 (view_list_to_loc_list_val_node): New.
16830 (print_dw_val): Handle dw_val_class_view_list.
16831 (size_of_die): Likewise.
16832 (value_format): Likewise.
16833 (loc_list_has_views): New.
16834 (gen_llsym): Set vl_symbol too.
16835 (maybe_gen_llsym, skip_loc_list_entry): New.
16836 (dwarf2out_maybe_output_loclist_view_pair): New.
16837 (output_loc_list): Output view list or entries too.
16838 (output_view_list_offset): New.
16839 (output_die): Handle dw_val_class_view_list.
16840 (output_dwarf_version): New.
16841 (output_compilation_unit_header): Use it.
16842 (output_skeleton_debug_sections): Likewise.
16843 (output_rnglists, output_line_info): Likewise.
16844 (output_pubnames, output_aranges): Update version comments.
16845 (output_one_line_info_table): Output view numbers in asm comments.
16846 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16847 Call maybe_gen_llsym.
16848 (loc_list_from_tree_1): Adjust.
16849 (add_AT_location_description): Create view list attribute if
16850 needed, check it's absent otherwise.
16851 (convert_cfa_to_fb_loc_list): Adjust.
16852 (maybe_emit_file): Call output_asm_line_debug_info for test.
16853 (dwarf2out_var_location): Reset views as needed. Precompute
16854 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
16855 attribute. Set view.
16856 (new_line_info_table): Reset next view.
16857 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16858 (dwarf2out_source_line): Likewise. Output view resets and labels to
16859 the assembler, or select appropriate line info opcodes.
16860 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16861 (optimize_string_length): Catch it. Adjust.
16862 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
16863 dw_val_class_view_list, and remove it if no longer needed.
16864 (hash_loc_list): Hash view numbers.
16865 (loc_list_hasher::equal): Compare them.
16866 (optimize_location_lists): Check whether a view list symbol is
16867 needed, and whether the locview attribute is present, and
16868 whether they match. Remove the locview attribute if no longer
16870 (index_location_lists): Call skip_loc_list_entry for test.
16871 (dwarf2out_finish): Call output_asm_line_debug_info for test.
16872 Use output_dwarf_version.
16873 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16874 (struct dw_val_node): Add val_view_list.
16875 * final.c (SEEN_NEXT_VIEW): New.
16876 (set_next_view_needed): New.
16877 (clear_next_view_needed): New.
16878 (maybe_output_next_view): New.
16879 (final_start_function): Rename to...
16880 (final_start_function_1): ... this. Take pointer to FIRST,
16881 add SEEN parameter. Emit param bindings in the initial view.
16882 (final_start_function): Reintroduce SEEN-less interface.
16883 (final): Rename to...
16884 (final_1): ... this. Take SEEN parameter. Output final pending
16885 next view at the end.
16886 (final): Reintroduce seen-less interface.
16887 (final_scan_insn): Output pending next view before switching
16888 sections or ending a block. Mark the next view as needed when
16889 outputting variable locations. Notify debug backend of section
16890 changes, and of location view changes.
16891 (rest_of_handle_final): Adjust.
16892 * toplev.c (process_options): Autodetect value for debug variable
16893 location views option. Warn on incompat5 without -gdwarf-5.
16894 * doc/invoke.texi (gvariable-location-views): New.
16895 (gvariable-location-views=incompat5): New.
16896 (gno-variable-location-views): New.
16898 2018-02-08 David Malcolm <dmalcolm@redhat.com>
16900 PR tree-optimization/84136
16901 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16902 that the result of find_edge is non-NULL.
16904 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16907 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16908 storing integer register in SImode. Fix cost of 256 and 512
16909 byte aligned SSE register store.
16911 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16913 * config/i386/i386.c (ix86_multiplication_cost): Fix
16914 multiplication cost for TARGET_AVX512DQ.
16916 2018-02-08 Marek Polacek <polacek@redhat.com>
16918 PR tree-optimization/84238
16919 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16922 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16924 PR tree-optimization/84265
16925 * tree-vect-stmts.c (vectorizable_store): Don't treat
16926 VMAT_CONTIGUOUS accesses as grouped.
16927 (vectorizable_load): Likewise.
16929 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16931 PR tree-optimization/81635
16932 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16933 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16934 (test_round_for_mask): New functions.
16935 (wide_int_cc_tests): Call test_round_for_mask.
16936 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16937 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16938 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16939 range returned by get_range_info.
16941 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16944 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16945 * symtab.c: Include builtins.h
16946 (symtab_node::output_to_lto_symbol_table_p): Move here
16947 from lto-streamer-out.c:output_symbol_p.
16948 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16949 (output_symbol_p): Move all logic to symtab.c
16950 (produce_symtab): Update.
16952 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16954 * config/s390/s390-opts.h (enum indirect_branch): Define.
16955 * config/s390/s390-protos.h (s390_return_addr_from_memory)
16956 (s390_indirect_branch_via_thunk)
16957 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16958 (enum s390_indirect_branch_type): Define.
16959 * config/s390/s390.c (struct s390_frame_layout, struct
16960 machine_function): Remove.
16961 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16962 (indirect_branch_table_label_no, indirect_branch_table_name):
16964 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16965 (enum s390_indirect_branch_option): Define.
16966 (s390_return_addr_from_memory): New function.
16967 (s390_handle_string_attribute): New function.
16968 (s390_attribute_table): Add new attribute handler.
16969 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
16970 (s390_indirect_branch_via_thunk): New function.
16971 (s390_indirect_branch_via_inline_thunk): New function.
16972 (s390_function_ok_for_sibcall): When jumping via thunk disallow
16973 sibling call optimization for non z10 compiles.
16974 (s390_emit_call): Force indirect branch target to be a single
16975 register. Add r1 clobber for non-z10 compiles.
16976 (s390_emit_epilogue): Emit return jump via return_use expander.
16977 (s390_reorg): Handle JUMP_INSNs as execute targets.
16978 (s390_option_override_internal): Perform validity checks for the
16979 new command line options.
16980 (s390_indirect_branch_attrvalue): New function.
16981 (s390_indirect_branch_settings): New function.
16982 (s390_set_current_function): Invoke s390_indirect_branch_settings.
16983 (s390_output_indirect_thunk_function): New function.
16984 (s390_code_end): Implement target hook.
16985 (s390_case_values_threshold): Implement target hook.
16986 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16988 * config/s390/s390.h (struct s390_frame_layout)
16989 (struct machine_function): Move here from s390.c.
16990 (TARGET_INDIRECT_BRANCH_NOBP_RET)
16991 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16992 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16993 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16994 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16995 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16996 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16997 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16998 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16999 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17000 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17001 (mnemonic attribute): Add values which aren't recognized
17003 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17004 pattern for branch conversion. Fix mnemonic attribute.
17005 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17006 indirect branch via thunk if requested.
17007 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17008 ("*indirect_jump"): Disable for branch conversion using out of
17010 ("indirect_jump_via_thunk<mode>_z10")
17011 ("indirect_jump_via_thunk<mode>")
17012 ("indirect_jump_via_inlinethunk<mode>_z10")
17013 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17014 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17015 ("casesi_jump_via_inlinethunk<mode>_z10")
17016 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17017 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17018 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17019 ("*indirect2_jump"): Disable for branch conversion.
17020 ("casesi_jump"): Turn into expander and expand patterns for branch
17022 ("return_use"): New expander.
17023 ("*return"): Emit return via thunk and rename it to ...
17024 ("*return<mode>"): ... this one.
17025 * config/s390/s390.opt: Add new options and and enum for the
17028 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17030 * lra-constraints.c (match_reload): Unconditionally use
17031 gen_lowpart_SUBREG, rather than selecting between that
17032 and equivalent gen_rtx_SUBREG code.
17034 2018-02-08 Richard Biener <rguenther@suse.de>
17036 PR tree-optimization/84233
17037 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17038 changed flag instead of boguously re-using phi_inserted.
17040 2018-02-08 Martin Jambor <mjambor@suse.cz>
17042 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17043 static local variables.
17045 2018-02-08 Richard Biener <rguenther@suse.de>
17047 PR tree-optimization/84278
17048 * tree-vect-stmts.c (vectorizable_store): When looking for
17049 smaller vector types to perform grouped strided loads/stores
17050 make sure the mode is supported by the target.
17051 (vectorizable_load): Likewise.
17053 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17055 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17056 Increase LDP/STP opportunities by adding adjacent callee-saves.
17058 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17060 PR rtl-optimization/84068
17061 PR rtl-optimization/83459
17062 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17064 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17066 PR tree-optimization/84224
17067 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17068 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17069 non-zero arguments.
17071 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17074 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17075 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17076 count, remove one USE.
17078 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17080 * doc/install.texi (Configuration): Document the
17081 --with-long-double-format={ibm,ieee} PowerPC configuration
17085 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17086 Convert from define_expand to be define_insn_and_split. Rework
17087 float/double/_Float128 conversions to QI/HI/SImode to work with
17088 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17089 conversions to QI/HImode types did a store and then a load to
17090 truncate the value. For conversions to VSX registers, don't split
17091 the insn, instead emit the code directly. Use the code iterator
17092 any_fix to combine signed and unsigned conversions.
17093 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17094 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17095 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17096 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17097 (fix_<mode>di2_hw): Likewise.
17098 (fixuns_<mode>di2_hw): Likewise.
17099 (fix_<mode>si2_hw): Likewise.
17100 (fixuns_<mode>si2_hw): Likewise.
17101 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17102 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17103 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17104 fix<uns>_trunc<SFDF:mode>si2_p8.
17105 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17106 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17107 (fix<uns>_<mode>_mem): Likewise.
17108 (fctiw<u>z_<mode>_mem): Likewise.
17109 (fix<uns>_<mode>_mem): Likewise.
17110 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17111 the register allocator from doing a direct move to the GPRs to do
17112 a store, and instead use the ISA 3.0 store byte/half-word from
17113 vector register instruction. For IEEE 128-bit floating point,
17114 also optimize stores of 32-bit ints.
17115 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17117 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17119 * genextract.c (push_pathstr_operand): New function to support
17121 (walk_rtx): Call push_pathstr_operand.
17122 (print_path): Support [a-zA-Z].
17124 2018-02-07 Richard Biener <rguenther@suse.de>
17126 PR tree-optimization/84037
17127 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17128 (cse_and_gimplify_to_preheader): Declare.
17129 (vect_get_place_in_interleaving_chain): Likewise.
17130 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17132 (_loop_vec_info::~_loop_vec_info): Delete it.
17133 (cse_and_gimplify_to_preheader): New function.
17134 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17135 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17136 (vectorizable_load): Likewise. For grouped stores always base
17137 the IV on the first element.
17138 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17139 condition before gimplifying.
17141 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17143 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17144 *DIV_EXPR and *MOD_EXPR.
17146 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17149 * config/i386/i386.c (ix86_option_override_internal): Mask out
17150 the CF_SET bit when checking -fcf-protection.
17152 2018-02-07 Tom de Vries <tom@codesourcery.com>
17155 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17158 2018-02-07 Richard Biener <rguenther@suse.de>
17160 PR tree-optimization/84204
17161 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17164 PR tree-optimization/84205
17165 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17166 special-case isl_ast_op_zdiv_r.
17168 PR tree-optimization/84223
17169 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17170 Only add conditions from within the region.
17171 (gather_bbs::after_dom_children): Adjust.
17173 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17176 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17177 * config/avr/avr.md: Only post-reload split REG-REG moves if
17178 either register is GENERAL_REG_P.
17180 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17182 PR tree-optimization/84235
17183 * tree-ssa-scopedtables.c
17184 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17185 if the subtraction is performed in floating point type where NaNs are
17186 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17187 build 1. Formatting fix.
17189 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17192 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17193 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17194 and skip it regardless of bb boundaries. Use CALL_P macro,
17195 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17196 unnecessarily, formatting fix.
17198 2018-02-06 Michael Collison <michael.collison@arm.com>
17200 * config/arm/thumb2.md:
17201 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17202 (*thumb_mov_notscc): Ditto.
17204 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17207 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17208 unsigned_fix, not "s".
17210 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17212 * configure.ac (gcc_fn_eh_frame_ro): New function.
17213 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17214 correct .eh_frame permissions.
17215 * configure: Regenerate.
17217 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17219 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17220 irrelevant options.
17222 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17224 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17225 Display warning message for -mno-speculate-indirect-jumps.
17227 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17229 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17231 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17233 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17235 PR tree-optimization/84225
17236 * tree-eh.c (find_trapping_overflow): Only call
17237 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17239 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17242 * config/i386/i386.c: Reimplement the check of possible options
17243 -mibt/-mshstk conbination. Change error messages.
17244 * doc/invoke.texi: Fix a typo: remove extra '='.
17246 2018-02-06 Marek Polacek <polacek@redhat.com>
17248 PR tree-optimization/84228
17249 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17251 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17254 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17255 emitted arch directives.
17256 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17257 __ARM_FEATURE_COPROC before changing architectures.
17259 2018-02-06 Richard Biener <rguenther@suse.de>
17261 * config/i386/i386.c (print_reg): Fix typo.
17262 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17264 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17266 * configure: Regenerate.
17268 2018-02-05 Martin Sebor <msebor@redhat.com>
17270 PR tree-optimization/83369
17271 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17274 2018-02-05 Martin Liska <mliska@suse.cz>
17276 * doc/invoke.texi: Cherry-pick upstream r323995.
17278 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17280 * ira.c (ira_init_register_move_cost): Adjust comment.
17282 2018-02-05 Martin Liska <mliska@suse.cz>
17284 PR gcov-profile/84137
17285 * doc/gcov.texi: Fix typo in documentation.
17287 2018-02-05 Martin Liska <mliska@suse.cz>
17289 PR gcov-profile/83879
17290 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17291 using dlopen functionality.
17293 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17295 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17296 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17297 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17298 _mm_maskz_range_round_ss): New intrinsics.
17299 (__builtin_ia32_rangesd128_round)
17300 (__builtin_ia32_rangess128_round): Remove.
17301 (__builtin_ia32_rangesd128_mask_round,
17302 __builtin_ia32_rangess128_mask_round): New builtins.
17303 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17304 __builtin_ia32_rangess128_round): Remove.
17305 (__builtin_ia32_rangesd128_mask_round,
17306 __builtin_ia32_rangess128_mask_round): New builtins.
17307 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17308 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17309 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17310 "<round_saeonly_constraint>")): Changed to ...
17311 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17312 "<round_saeonly_scalar_constraint>")): ... this.
17313 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17314 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17315 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17316 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17317 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17319 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17321 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17323 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17324 Remove all values except native, 8540 and 8548.
17326 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17328 * config/i386/i386.c (ix86_output_function_return): Pass
17329 INVALID_REGNUM, instead of -1, as invalid register number to
17330 indirect_thunk_name and output_indirect_thunk.
17332 2018-02-02 Julia Koval <julia.koval@intel.com>
17334 * config.gcc: Add -march=icelake.
17335 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17336 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17337 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17338 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17339 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17340 (processor_target_table): Add icelake.
17341 (ix86_option_override_internal): Handle new PTAs.
17342 (get_builtin_code_for_version): Handle icelake.
17343 (M_INTEL_COREI7_ICELAKE): New.
17344 (fold_builtin_cpu): Handle icelake.
17345 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17346 * doc/invoke.texi: Add -march=icelake.
17348 2018-02-02 Julia Koval <julia.koval@intel.com>
17350 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17351 to wide_int_bitmask.
17352 * wide-int-bitmask.h: New.
17354 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17357 * config/i386/i386.md: Replace Pmode with word_mode in
17358 builtin_setjmp_setup and builtin_longjmp to support x32.
17360 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17364 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17366 (rs6000_supported_cpu_names): New static variable.
17367 (linux_cpu_translation_table): Likewise.
17368 (elf_platform) <cpu>: Define new static variable and use it.
17369 Translate kernel AT_PLATFORM name to canonical name if needed.
17370 Error if platform name is unknown.
17372 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17375 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17377 2018-02-01 Jeff Law <law@redhat.com>
17380 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17381 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17382 the scratch if RELEASE_VIA_POP is false.
17383 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17384 If we have to save a temporary register, decrement SIZE appropriately.
17385 Pass new arguments to release_scratch_register_on_entry.
17386 (ix86_adjust_stack_and_probe): Likewise.
17387 (ix86_emit_probe_stack_range): Pass new arguments to
17388 release_scratch_register_on_entry.
17390 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17392 PR rtl-optimization/84157
17393 * combine.c (change_zero_ext): Use REG_P predicate in
17394 front of HARD_REGISTER_P predicate.
17396 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17398 * config/avr/avr.c (avr_option_override): Move disabling of
17399 -fdelete-null-pointer-checks to...
17400 * common/config/avr/avr-common.c (avr_option_optimization_table):
17403 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17405 PR tree-optimization/81635
17406 * tree-data-ref.c (split_constant_offset_1): For types that
17407 wrap on overflow, try to use range info to prove that wrapping
17410 2018-02-01 Renlin Li <renlin.li@arm.com>
17413 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17414 TAILCALL_ADDR_REGS.
17415 (aarch64_register_move_cost): Likewise.
17416 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17417 TAILCALL_ADDR_REGS.
17418 (REG_CLASS_NAMES): Likewise.
17419 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17420 TAILCALL_ADDR_REGS. Remove IP registers.
17421 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17423 2018-02-01 Richard Biener <rguenther@suse.de>
17425 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17426 for specifying RPO order and allow NULL for that.
17427 * domwalk.c (dom_walker::dom_walker): Likewise.
17428 (dom_walker::walk): Handle NULL RPO order.
17429 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17431 (rewrite_update_dom_walker): Likewise.
17432 (mark_def_dom_walker): Likewise.
17434 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17436 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17437 (aarch64_maybe_expand_sve_subreg_move): Declare.
17438 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17439 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17441 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17442 that are semantically a reverse operation.
17443 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17444 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17445 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17447 (aarch64_can_change_mode_class): For big-endian, forbid changes
17448 between two SVE modes if they have different element sizes.
17450 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17452 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17453 the TImode handling for big-endian targets.
17455 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17457 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17458 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17460 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17461 Remove BSWAP handing for big-endian targets and use the form of
17462 LD1RQ appropariate for the mode.
17464 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17466 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17467 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17468 duplicated element.
17470 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17473 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17474 check for operands that need to go through aarch64_sve_reload_be.
17476 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17478 PR tree-optimization/81661
17479 PR tree-optimization/84117
17480 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17481 * tree-eh.c: Include gimplify.h.
17482 (find_trapping_overflow, replace_trapping_overflow,
17483 rewrite_to_non_trapping_overflow): New functions.
17484 * tree-vect-loop.c: Include tree-eh.h.
17485 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17486 * tree-data-ref.c: Include tree-eh.h.
17487 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17489 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17491 PR rtl-optimization/84123
17492 * combine.c (change_zero_ext): Check if hard register satisfies
17493 can_change_dest_mode before calling gen_lowpart_SUBREG.
17495 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17498 * ira.c (ira_init_register_move_cost): Remove assert.
17500 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17502 PR rtl-optimization/84071
17503 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17504 * doc/tm.texi: Regenerate.
17506 2018-01-31 Richard Biener <rguenther@suse.de>
17508 PR tree-optimization/84132
17509 * tree-data-ref.c (analyze_miv_subscript): Properly
17510 check whether evolution_function_is_affine_multivariate_p
17511 before calling gcd_of_steps_may_divide_p.
17513 2018-01-31 Julia Koval <julia.koval@intel.com>
17516 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17517 * config/i386/i386.md (rdpid_rex64) New.
17518 (rdpid): Make 32bit only.
17520 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17523 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17524 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17526 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17529 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17531 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17533 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17535 PR rtl-optimization/84071
17536 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17537 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17539 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17541 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17542 (arc_attribute_table): Add 'aux' attribute.
17543 (arc_in_small_data_p): Consider aux like variables.
17544 (arc_is_aux_reg_p): New function.
17545 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17546 (arc_get_aux_arg): New function.
17547 (prepare_move_operands): Handle aux-register access.
17548 (arc_handle_aux_attribute): New function.
17549 * doc/extend.texi (ARC Variable attributes): Add subsection.
17551 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17553 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17554 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17555 (arc_attribute_table): Add 'uncached' attribute.
17556 (arc_print_operand): Print '.di' flag for uncached memory
17558 (arc_in_small_data_p): Do not consider for small data the uncached
17560 (arc_is_uncached_mem_p): New function.
17561 * config/arc/predicates.md (compact_store_memory_operand): Check
17562 for uncached memory accesses.
17563 (nonvol_nonimm_operand): Likewise.
17564 * doc/extend.texi (ARC Type Attribute): New subsection.
17566 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17569 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17570 falign-loops=): Add Optimization flag.
17572 2018-01-30 Jeff Law <law@redhat.com>
17575 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17576 INT_REGISTERS_SAVED. Check it prior to calling
17577 get_scratch_register_on_entry.
17578 (ix86_adjust_stack_and_probe): Similarly.
17579 (ix86_emit_probe_stack_range): Similarly.
17580 (ix86_expand_prologue): Corresponding changes.
17582 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17585 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17586 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17588 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17591 * lra-constraints.c (curr_insn_transform): Process AND in the
17594 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17596 PR rtl-optimization/83986
17597 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17598 dependence against last_pending_memory_flush in addition to
17599 pending_jump_insns.
17601 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17603 PR tree-optimization/81611
17604 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17607 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17610 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17613 2018-01-30 Richard Biener <rguenther@suse.de>
17614 Jakub Jelinek <jakub@redhat.com>
17616 PR tree-optimization/84111
17617 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17618 inner loops added during recursion, as they don't have up-to-date
17621 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17624 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17625 (can_inline_edge_by_limits_p): ... here.
17626 (can_early_inline_edge_p, check_callers,
17627 update_caller_keys, update_callee_keys, recursive_inlining,
17628 add_new_edges_to_heap, speculation_useful_p,
17629 inline_small_functions,
17630 inline_small_functions, flatten_function,
17631 inline_to_all_callers_1): Update.
17633 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17635 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17638 2018-01-30 Richard Biener <rguenther@suse.de>
17640 PR tree-optimization/83008
17641 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17642 invariant and constant vector uses in stmts when they need
17643 more than one stmt.
17645 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17648 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17649 * configure: Regenerate.
17651 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17653 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17655 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17656 Use gen_rtx_REG rather than gen_lowpart.
17658 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17660 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17661 rather than 0 when creating partial subregs.
17663 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17665 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17668 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17671 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17672 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17673 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17674 flags. This restores the settings used before the 2017-07-24.
17675 Turning off pre increment/decrement/modify allows IVOPTS to
17676 optimize DF/SF loops where the index is an int.
17678 2018-01-29 Richard Biener <rguenther@suse.de>
17679 Kelvin Nilsen <kelvin@gcc.gnu.org>
17682 * tree-vect-stmts.c (vectorizable_call): Don't call
17683 targetm.vectorize_builtin_md_vectorized_function if callee is
17686 2018-01-22 Carl Love <cel@us.ibm.com>
17688 * doc/extend.tex: Fix typo in second arg in
17689 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17691 2018-01-29 Richard Biener <rguenther@suse.de>
17693 PR tree-optimization/84086
17694 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17695 (flush_ssaname_freelist): When SSA names were released reset
17696 the SCEV hash table.
17698 2018-01-29 Richard Biener <rguenther@suse.de>
17700 PR tree-optimization/84057
17701 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17702 removed paths when removing edges.
17704 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17706 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17707 -mfunction-return=@var{choice}.
17709 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17711 PR diagnostic/84034
17712 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17713 Handle CR like TAB.
17714 (layout::print_source_line): Likewise.
17715 (test_get_line_width_without_trailing_whitespace): Add test cases.
17717 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17719 PR middle-end/84040
17720 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17723 2018-01-26 Jim Wilson <jimw@sifive.com>
17725 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17727 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17730 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17732 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17733 and CMP + SUB-immediate -> SUBS.
17735 2018-01-26 Martin Sebor <msebor@redhat.com>
17737 PR tree-optimization/83896
17738 * tree-ssa-strlen.c (get_string_len): Rename...
17739 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17740 Avoid assuming length is constant.
17741 (handle_char_store): Use HOST_WIDE_INT for string length.
17743 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17746 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17747 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17749 2018-01-26 Richard Biener <rguenther@suse.de>
17751 PR rtl-optimization/84003
17752 * dse.c (record_store): Only record redundant stores when
17753 the earlier store aliases at least all accesses the later one does.
17755 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17757 PR rtl-optimization/83985
17758 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17759 REG_CFA_RESTORE insns.
17760 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17761 REG_CFA_RESTORE insns here.
17764 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17765 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17767 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17769 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17770 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17771 (arc_init): Likewise.
17772 (arc_override_options): Likewise.
17773 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17775 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17777 * config/arc/arc.h (TARGET_DBNZ): Define.
17778 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17779 properly set the tune attribute.
17780 (dbnz): Use TARGET_DBNZ guard.
17781 * config/arc/arc.opt (mtune): Add core3 option.
17783 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17785 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17786 recognize new pic like addresses.
17787 (arc_delegitimize_address): Clean up.
17789 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17791 * config/arc/arc-arches.def: Option mrf16 valid for all
17793 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17794 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17795 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17796 * config/arc/arc-tables.opt: Regenerate.
17797 * config/arc/arc.c (arc_conditional_register_usage): Handle
17798 reduced register file case.
17799 (arc_file_start): Set must have build attributes.
17800 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17801 mrf16 option value.
17802 * config/arc/arc.opt (mrf16): Add new option.
17803 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17804 * config/arc/genmultilib.awk: Handle new mrf16 option.
17805 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17806 * config/arc/t-multilib: Regenerate.
17807 * doc/invoke.texi (ARC Options): Document mrf16 option.
17809 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17811 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17812 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17813 (arc_attribute_table): Add 'secure_call' attribute.
17814 (arc_print_operand): Print secure call operand.
17815 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17817 (arc_is_secure_call_p): New function. * config/arc/arc.md
17818 (call_i): Add support for sjli instruction.
17819 (call_value_i): Likewise.
17820 * config/arc/constraints.md (Csc): New constraint.
17822 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17823 John Eric Martin <John.Martin@emmicro-us.com>
17825 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17826 * config/arc/arc.c (_arc_jli_section): New struct.
17827 (arc_jli_section): New type.
17828 (rc_jli_sections): New static variable.
17829 (arc_handle_jli_attribute): New function.
17830 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17831 (arc_file_end): New function.
17832 (TARGET_ASM_FILE_END): Define.
17833 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17834 (arc_add_jli_section): New function.
17835 (jli_call_scan): Likewise.
17836 (arc_reorg): Call jli_call_scan.
17837 (arc_output_addsi): Remove 'S' from printing asm operand.
17838 (arc_is_jli_call_p): New function.
17839 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17841 (movhi_insn): Likewise.
17842 (movsi_insn): Likewise.
17843 (movsi_set_cc_insn): Likewise.
17844 (loadqi_update): Likewise.
17845 (load_zeroextendqisi_update): Likewise.
17846 (load_signextendqisi_update): Likewise.
17847 (loadhi_update): Likewise.
17848 (load_zeroextendhisi_update): Likewise.
17849 (load_signextendhisi_update): Likewise.
17850 (loadsi_update): Likewise.
17851 (loadsf_update): Likewise.
17852 (movsicc_insn): Likewise.
17853 (bset_insn): Likewise.
17854 (bxor_insn): Likewise.
17855 (bclr_insn): Likewise.
17856 (bmsk_insn): Likewise.
17857 (bicsi3_insn): Likewise.
17858 (cmpsi_cc_c_insn): Likewise.
17859 (movsi_ne): Likewise.
17860 (movsi_cond_exec): Likewise.
17861 (clrsbsi2): Likewise.
17862 (norm_f): Likewise.
17872 (call_i): Remove 'S' asm letter, add jli instruction.
17873 (call_value_i): Likewise.
17874 * config/arc/arc.op (mjli-always): New option.
17875 * config/arc/constraints.md (Cji): New constraint.
17876 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17878 (subsf3_fpx): Likewise.
17879 (mulsf3_fpx): Likewise.
17880 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17882 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17883 function attrbutes.
17884 * doc/invoke.texi (ARC): Document mjli-always option.
17886 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
17888 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17889 avoid addition with 0 and use incw and decw where possible.
17891 2018-01-26 Richard Biener <rguenther@suse.de>
17893 PR tree-optimization/81082
17894 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17895 association if it requires casting to unsigned.
17896 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17897 from fold_plusminus_mult_expr to catch important cases late when
17898 range info is available.
17900 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17902 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17903 * configure.ac (hidden_linkonce): New test.
17904 * configure: Regenerate.
17905 * config.in: Regenerate.
17907 2018-01-26 Julia Koval <julia.koval@intel.com>
17909 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17910 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17911 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17912 _mm_mask_bitshuffle_epi64_mask): Fix type.
17913 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17914 USI_FTYPE_V4DI_V4DI_USI): Remove.
17915 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17916 __builtin_ia32_vpshufbitqmb256_mask,
17917 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17918 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17919 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17921 2018-01-26 Alan Modra <amodra@gmail.com>
17924 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17925 UNSPEC_VBPERMQ. Sort other unspecs.
17927 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
17929 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17931 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
17933 PR middle-end/83055
17934 * predict.c (drop_profile): Do not push/pop cfun; update also
17936 (handle_missing_profiles): Fix logic looking for zero profiles.
17938 2018-01-25 Jakub Jelinek <jakub@redhat.com>
17940 PR middle-end/83977
17941 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17942 on functions with #pragma omp declare simd or functions with simd
17944 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17945 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17946 Remove trailing \n from warning_at calls.
17948 2018-01-25 Tom de Vries <tom@codesourcery.com>
17951 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17952 for neutered workers.
17954 2018-01-24 Joseph Myers <joseph@codesourcery.com>
17957 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17958 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17960 2018-01-24 Jeff Law <law@redhat.com>
17963 * i386.c (get_probe_interval): Move to earlier point.
17964 (ix86_compute_frame_layout): If -fstack-clash-protection and
17965 the frame is larger than the probe interval, then use pushes
17966 to save registers rather than reg->mem moves.
17967 (ix86_expand_prologue): Remove conditional for int_registers_saved
17970 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
17973 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
17974 min/max for never referenced object.
17976 2018-01-24 Jakub Jelinek <jakub@redhat.com>
17978 PR middle-end/83977
17979 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
17981 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
17982 attributes from DECL_ATTRIBUTES (decl) without affecting
17983 DECL_ATTRIBUTES (current_function_decl).
17984 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
17985 functions with non-NULL DECL_ABSTRACT_ORIGIN.
17987 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
17989 PR tree-optimization/83979
17990 * fold-const.c (fold_comparison): Use constant_boolean_node
17991 instead of boolean_{true,false}_node.
17993 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
17995 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17998 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18000 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18001 Simplify the clause that sets the length attribute.
18002 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18003 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18004 clause that sets the length attribute.
18005 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18007 2018-01-24 Tom de Vries <tom@codesourcery.com>
18010 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18011 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18012 Add strict parameter.
18013 (prevent_branch_around_nothing): Insert dummy insn between branch to
18014 label and label with no ptx insn inbetween.
18015 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18017 2018-01-24 Tom de Vries <tom@codesourcery.com>
18020 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18021 for neutered threads in warp.
18022 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18024 2018-01-24 Richard Biener <rguenther@suse.de>
18026 PR tree-optimization/83176
18027 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18030 2018-01-24 Richard Biener <rguenther@suse.de>
18032 PR tree-optimization/82819
18033 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18034 code generating pluses that are no-ops in the target precision.
18036 2018-01-24 Richard Biener <rguenther@suse.de>
18038 PR middle-end/84000
18039 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18041 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18043 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18044 to merge probabilities.
18045 * predict.c (probably_never_executed): Also mark as cold functions
18046 with global 0 profile and guessed local profile.
18047 * profile-count.c (profile_probability::combine_with_count): New
18049 * profile-count.h (profile_probability::operator*,
18050 profile_probability::operator*=, profile_probability::operator/,
18051 profile_probability::operator/=): Reduce precision to adjusted
18052 and set value to guessed on contradictory divisions.
18053 (profile_probability::combine_with_freq): Remove.
18054 (profile_probability::combine_wiht_count): Declare.
18055 (profile_count::force_nonzero):: Set to adjusted.
18056 (profile_count::probability_in):: Set quality to adjusted.
18057 * tree-ssa-tail-merge.c (replace_block_by): Use
18058 combine_with_count.
18060 2018-01-23 Andrew Waterman <andrew@sifive.com>
18061 Jim Wilson <jimw@sifive.com>
18063 * config/riscv/riscv.c (riscv_stack_boundary): New.
18064 (riscv_option_override): Set riscv_stack_boundary. Handle
18065 riscv_preferred_stack_boundary_arg.
18066 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18067 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18068 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18069 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18070 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18071 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18073 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18076 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18077 of struct ix86_frame.
18078 (ix86_expand_epilogue): Likewise. Add a local variable for
18079 the reg_save_offset field in struct ix86_frame.
18081 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18083 PR tree-optimization/82604
18084 * tree-loop-distribution.c (enum partition_kind): New enum item
18085 PKIND_PARTIAL_MEMSET.
18086 (partition_builtin_p): Support above new enum item.
18087 (generate_code_for_partition): Ditto.
18088 (compute_access_range): Differentiate cases that equality can be
18089 proven at all loops, the innermost loops or no loops.
18090 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18091 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18092 (finalize_partitions, distribute_loop): Don't fuse partition of
18093 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18094 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18095 parloop is enabled.
18097 2018-01-23 Martin Liska <mliska@suse.cz>
18099 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18100 order to ignore the predictor.
18101 (PRED_POLYMORPHIC_CALL): Likewise.
18102 (PRED_RECURSIVE_CALL): Likewise.
18104 2018-01-23 Martin Liska <mliska@suse.cz>
18106 * tree-profile.c (tree_profiling): Print function header to
18107 aware reader which function we are working on.
18108 * value-prof.c (gimple_find_values_to_profile): Do not print
18109 not interesting value histograms.
18111 2018-01-23 Martin Liska <mliska@suse.cz>
18113 * profile-count.h (enum profile_quality): Add
18114 profile_uninitialized as the first value. Do not number values
18115 as they are zero based.
18116 (profile_count::verify): Update sanity check.
18117 (profile_probability::verify): Likewise.
18119 2018-01-23 Nathan Sidwell <nathan@acm.org>
18121 * doc/invoke.texi (ffor-scope): Deprecate.
18123 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18125 PR tree-optimization/83510
18126 * domwalk.c (set_all_edges_as_executable): New function.
18127 (dom_walker::dom_walker): Convert bool param
18128 "skip_unreachable_blocks" to enum reachability. Move setup of
18129 edge flags to set_all_edges_as_executable and only do it when
18130 reachability is REACHABLE_BLOCKS.
18131 * domwalk.h (enum dom_walker::reachability): New enum.
18132 (dom_walker::dom_walker): Convert bool param
18133 "skip_unreachable_blocks" to enum reachability.
18134 (set_all_edges_as_executable): New decl.
18135 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18136 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18138 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18139 but converting true to REACHABLE_BLOCKS.
18140 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18142 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18143 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18144 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18146 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18147 if check_all_array_refs will be called.
18149 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18151 * tree.c (selftest::test_location_wrappers): Add more test
18154 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18156 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18157 (selftest::test_bit_in_range): Likewise.
18159 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18162 * doc/sourcebuild.texi (vect_float): Say that the selector
18163 only describes the situation when -funsafe-math-optimizations is on.
18164 (vect_float_strict): Document.
18166 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18168 PR tree-optimization/83965
18169 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18170 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18171 instead of checking only for a reduction.
18172 (vect_recog_widen_sum_pattern): Likewise.
18174 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18176 * predict.c (probably_never_executed): Only use precise profile info.
18177 (compute_function_frequency): Skip after inlining hack since we now
18178 have quality checking.
18180 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18182 * profile-count.h (profile_probability::very_unlikely,
18183 profile_probability::unlikely, profile_probability::even): Set
18184 precision to guessed.
18186 2018-01-23 Richard Biener <rguenther@suse.de>
18188 PR tree-optimization/83963
18189 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18190 Properly terminate dominator walk when crossing the exit edge not
18191 when visiting its source block.
18193 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18196 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18197 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18199 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18201 PR tree-optimization/83957
18202 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18203 semicolon after for body surrounded by braces.
18205 PR tree-optimization/83081
18206 * profile-count.h (profile_probability::split): New method.
18207 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18208 Use profile_probability::split.
18209 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18210 when splitting a single conditional jump into 2.
18212 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18214 PR tree-optimization/69452
18215 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18218 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18220 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18221 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18222 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18224 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18226 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18228 * config/rl78/rl78.md (movdi): New define_expand.
18229 * config/rl78/rl78.c (rl78_split_movdi): New function.
18231 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18234 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18236 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18237 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18238 128-bit to produce an UNSPEC move to get the double word with the
18239 signbit and then a shift directly to do signbit.
18240 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18241 implementation with a new version that just does either a direct
18242 move or a regular move. Move memory interface to separate insns.
18243 Move insns so they are next to the expander.
18244 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18245 with signbit move. Split big and little endian case.
18246 (signbit<mode>2_dm_mem_le): Likewise.
18247 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18248 (signbit<mode>2_dm2): Likewise.
18250 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18252 * config/rl78/rl78.md (anddi3): New define_expand.
18254 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18256 * config/rl78/rl78.md (umindi3): New define_expand.
18258 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18260 * config/rl78/rl78.md (smindi3): New define_expand.
18262 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18264 * config/rl78/rl78.md (smaxdi3): New define_expand.
18266 2018-01-22 Carl Love <cel@us.ibm.com>
18268 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18269 LVX_V1TI): Add macro expansion.
18270 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18271 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18272 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18273 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18274 Change check to determine if the instruction is a byte reversing
18275 entry. Fix typo in comment.
18276 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18277 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18278 Add def_builtin calls for new builtins.
18279 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18280 Add define_insn expansion.
18282 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18284 * config/rl78/rl78.md (umaxdi3): New define_expand.
18286 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18288 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18289 for non-QImode registers.
18291 2018-01-22 Richard Biener <rguenther@suse.de>
18293 PR tree-optimization/83963
18294 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18295 including the loop exit block.
18296 (scop_detection::merge_sese): Likewise.
18297 (scop_detection::add_scop): Do it here instead.
18299 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18301 * doc/sourcebuild.texi (arm_softfloat): Document.
18303 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18306 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18307 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18308 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18310 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18311 David Edelsohn <dje.gcc@gmail.com>
18314 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18315 Change "crset eq" to "crset 2".
18316 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18317 (*call_indirect_aix<mode>_nospec): Likewise.
18318 (*call_value_indirect_aix<mode>_nospec): Likewise.
18319 (*call_indirect_elfv2<mode>_nospec): Likewise.
18320 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18321 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18322 change assembly output from . to $.
18323 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18324 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18325 (*tablejump<mode>_internal1_nospec): Likewise.
18327 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18330 * config/sh/sh_optimize_sett_clrt.cc:
18331 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18333 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18335 PR tree-optimization/83940
18336 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18337 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18338 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18339 use it to pass back the definition type.
18340 (vect_check_store_rhs): Likewise rhs_dt_out.
18341 (vect_build_gather_load_calls): Add a mask_dt argument and use
18342 it instead of a call to vect_is_simple_use.
18343 (vectorizable_store): Update calls to vect_check_load_store_mask
18344 and vect_check_store_rhs. Use the dt returned by the latter instead
18345 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18346 instead of calls to vect_is_simple_use. Pass the scalar rather
18347 than the vector operand to vect_is_simple_use when handling
18348 second and subsequent copies of an rhs value.
18349 (vectorizable_load): Update calls to vect_check_load_store_mask
18350 and vect_build_gather_load_calls. Use the cached mask_dt and
18351 gs_info.offset_dt instead of calls to vect_is_simple_use.
18353 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18355 PR middle-end/83945
18356 * tree-emutls.c: Include gimplify.h.
18357 (lower_emutls_2): New function.
18358 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18359 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18360 it before further processing.
18363 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18364 UINTVAL (trueop1) instead of INTVAL (op1).
18366 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18370 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18371 INCOMING_FRAME_SP_OFFSET if not defined.
18372 (scan_trace): Add ENTRY argument. If true and
18373 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18374 emit a note to adjust the CFA offset.
18375 (create_cfi_notes): Adjust scan_trace callers.
18376 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18377 INCOMING_FRAME_SP_OFFSET in the CIE.
18378 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18379 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18381 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18382 * doc/tm.texi: Regenerated.
18384 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18386 PR rtl-optimization/83147
18387 * lra-constraints.c (remove_inheritance_pseudos): Use
18388 lra_substitute_pseudo_within_insn.
18390 2018-01-19 Tom de Vries <tom@codesourcery.com>
18391 Cesar Philippidis <cesar@codesourcery.com>
18394 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18396 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18399 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18400 spaces for function labels.
18402 2018-01-19 Martin Liska <mliska@suse.cz>
18404 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18405 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18406 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18407 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18408 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18409 (PRED_CONST_RETURN): Change from 69 to 65.
18410 (PRED_NULL_RETURN): Change from 91 to 71.
18411 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18412 (PRED_LOOP_GUARD): Change from 66 to 73.
18414 2018-01-19 Martin Liska <mliska@suse.cz>
18416 * predict.c (predict_insn_def): Add new assert.
18417 (struct branch_predictor): Change type to signed integer.
18418 (test_prediction_value_range): Amend test to cover
18419 PROB_UNINITIALIZED.
18420 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18421 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18422 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18423 (PRED_LOOP_IV_COMPARE): Likewise.
18424 * predict.h (PROB_UNINITIALIZED): Define new constant.
18426 2018-01-19 Martin Liska <mliska@suse.cz>
18428 * predict.c (dump_prediction): Add new format for
18429 analyze_brprob.py script which is enabled with -details
18431 * profile-count.h (precise_p): New function.
18433 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18435 PR tree-optimization/83922
18436 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18437 there are no statements that need masking.
18438 (vect_active_double_reduction_p): New function.
18439 (vect_analyze_loop_operations): Use it when handling phis that
18440 are not in the loop header.
18442 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18444 PR tree-optimization/83914
18445 * tree-vect-loop.c (vectorizable_induction): Don't convert
18446 init_expr or apply the peeling adjustment for inductions
18447 that are nested within the vectorized loop.
18449 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18451 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18454 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18458 * function.h (gimplify_parameters): Add gimple_seq * argument.
18459 * function.c: Include gimple.h and options.h.
18460 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18461 for the added local temporaries if needed.
18462 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18463 if there are any parameter cleanups, wrap whole body into a
18464 try/finally with the cleanups.
18466 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18469 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18470 Use GET_MODE_CLASS for scalar floating point.
18472 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18476 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18477 Fix call of call_cgraph_insertion_hooks.
18479 2018-01-18 Martin Sebor <msebor@redhat.com>
18481 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18483 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18486 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18489 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18492 * common.opt: (-ffile-prefix-map): New option.
18493 * opts.c (common_handle_option): Defer it.
18494 * opts-global.c (handle_common_deferred_options): Handle it.
18495 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18496 * file-prefix-map.h: New file.
18497 (remap_debug_filename, add_debug_prefix_map): ...here.
18498 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18499 * final.c (debug_prefix_map, add_debug_prefix_map
18500 remap_debug_filename): Move to...
18501 * file-prefix-map.c: New file.
18502 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18503 generalize, get rid of alloca(), use strrchr() instead of strchr().
18504 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18505 Implement in terms of add_prefix_map().
18506 (remap_macro_filename, remap_debug_filename): Implement in term of
18508 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18509 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18510 * dbxout.c: Include file-prefix-map.h.
18511 * varasm.c: Likewise.
18512 * vmsdbgout.c: Likewise.
18513 * xcoffout.c: Likewise.
18514 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18515 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18516 * doc/invoke.texi (-ffile-prefix-map): Document.
18517 (-fdebug-prefix-map): Update description.
18519 2018-01-18 Martin Liska <mliska@suse.cz>
18521 * config/i386/i386.c (indirect_thunk_name): Document that also
18523 (output_indirect_thunk): Document why both instructions
18524 (pause and lfence) are generated.
18526 2018-01-18 Richard Biener <rguenther@suse.de>
18528 PR tree-optimization/83887
18529 * graphite-scop-detection.c
18530 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18531 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18532 (scop_detection::merge_sese): Re-implement with a flood-fill
18533 algorithm that properly finds a SESE region if it exists.
18535 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18538 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18539 pointer_diff optimizations use view_convert instead of convert.
18541 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18543 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18544 Generate different code for -mno-speculate-indirect-jumps.
18545 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18546 (*call_indirect_aix<mode>): Disable for
18547 -mno-speculate-indirect-jumps.
18548 (*call_indirect_aix<mode>_nospec): New define_insn.
18549 (*call_value_indirect_aix<mode>): Disable for
18550 -mno-speculate-indirect-jumps.
18551 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18552 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18553 -mno-speculate-indirect-jumps.
18554 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18556 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18558 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18559 long double type, set the flags for noting the default long double
18560 type, even if we don't pass or return a long double type.
18562 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18565 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18568 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18570 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18571 support for merge[hl].
18572 (fold_mergehl_helper): New helper function.
18573 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18574 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18575 (altivec_vmrglw_direct): Add xxmrglw insn.
18577 2018-01-17 Andrew Waterman <andrew@sifive.com>
18579 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18580 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18582 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18585 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18586 call the lto_location_cache before reading the
18587 DECL_SOURCE_LOCATION of the types.
18589 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18590 Richard Sandiford <richard.sandiford@linaro.org>
18592 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18593 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18594 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18595 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18596 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18598 * config/aarch64/constraints.md (aarch64_movti_operand):
18600 * config/aarch64/predicates.md (Uti): Add new constraint.
18602 2018-01-17 Carl Love <cel@us.ibm.com>
18604 * config/rs6000/vsx.md (define_expand xl_len_r,
18605 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18606 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18608 (define_expand, define_insn): Move the shift left from the
18609 define_insn to the define_expand for lxvl and stxvl instructions.
18610 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18611 and XL_LEN_R definitions to PURE.
18613 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18615 * config/i386/i386.c (indirect_thunk_name): Declare regno
18616 as unsigned int. Compare regno with INVALID_REGNUM.
18617 (output_indirect_thunk): Ditto.
18618 (output_indirect_thunk_function): Ditto.
18619 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18620 in the call to output_indirect_thunk_function.
18622 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18624 PR middle-end/83884
18625 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18626 rather than the size of inner_type to determine the stack slot size
18627 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18629 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18632 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18635 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18637 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18638 endian Linux systems to optionally enable multilibs for selecting
18639 the long double type if the user configured an explicit type.
18640 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18641 have no long double multilibs if not defined.
18642 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18643 warn if the user used -mabi={ieee,ibm}longdouble and we built
18644 multilibs for long double.
18645 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18646 appropriate multilib option.
18647 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18649 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18650 for building long double multilibs.
18651 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18653 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18655 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18658 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18660 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18663 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18666 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18669 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18671 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18672 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18674 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18676 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18677 different rtl trees depending on TARGET_64BIT.
18678 (rs6000_gen_lvx): Likewise.
18680 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18682 * config/visium/visium.md (nop): Tweak comment.
18683 (hazard_nop): Likewise.
18685 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18687 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18688 -mspeculate-indirect-jumps.
18689 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18690 for -mno-speculate-indirect-jumps.
18691 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18692 (*call_value_indirect_elfv2<mode>): Disable for
18693 -mno-speculate-indirect-jumps.
18694 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18695 (indirect_jump): Emit different RTL for
18696 -mno-speculate-indirect-jumps.
18697 (*indirect_jump<mode>): Disable for
18698 -mno-speculate-indirect-jumps.
18699 (*indirect_jump<mode>_nospec): New define_insn.
18700 (tablejump): Emit different RTL for
18701 -mno-speculate-indirect-jumps.
18702 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18703 (tablejumpsi_nospec): New define_expand.
18704 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18705 (tablejumpdi_nospec): New define_expand.
18706 (*tablejump<mode>_internal1): Disable for
18707 -mno-speculate-indirect-jumps.
18708 (*tablejump<mode>_internal1_nospec): New define_insn.
18709 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18712 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18714 * caller-save.c (insert_save): Drop unnecessary parameter. All
18717 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18718 Richard Biener <rguenth@suse.de>
18721 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18722 return early, inline manually is_gimple_sizepos. Make sure if we
18723 call gimplify_expr we don't end up with a gimple constant.
18724 * tree.c (variably_modified_type_p): Don't return true for
18725 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18726 * gimplify.h (is_gimple_sizepos): Remove.
18728 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18730 PR tree-optimization/83857
18731 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18732 vectorizable_live_operation for pure SLP statements.
18733 (vectorizable_live_operation): Handle PHIs.
18735 2018-01-16 Richard Biener <rguenther@suse.de>
18737 PR tree-optimization/83867
18738 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18739 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18741 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18744 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18745 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18746 If off is not INTEGER_CST, issue a may not be aligned warning
18747 rather than isn't aligned. Use isn%'t rather than isn't.
18748 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18750 <case MULT_EXPR>: Improve the case when bottom and one of the
18751 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18752 operand, in that case check if the other operand is multiple of
18753 bottom divided by the INTEGER_CST operand.
18755 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18758 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18759 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18760 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18761 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18762 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18763 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18764 (pa_function_arg_size): New function.
18766 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18768 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18769 in a separate statement.
18771 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18773 PR tree-optimization/83847
18774 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18775 group gathers and scatters.
18777 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18779 PR rtl-optimization/86620
18780 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18782 PR rtl-optimization/83213
18783 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18784 to last if both are JUMP_INSNs.
18786 PR tree-optimization/83843
18787 * gimple-ssa-store-merging.c
18788 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18789 store_immediate_info for bswap/nop orig_stores.
18791 2018-01-15 Andrew Waterman <andrew@sifive.com>
18793 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18795 <UDIV>: Increase cost if !TARGET_DIV.
18797 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18799 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18800 (define_attr "cr_logical_3op"): New.
18801 (cceq_ior_compare): Adjust.
18802 (cceq_ior_compare_complement): Adjust.
18803 (*cceq_rev_compare): Adjust.
18804 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18805 (is_cracked_insn): Adjust.
18806 (insn_must_be_first_in_group): Adjust.
18807 * config/rs6000/40x.md: Adjust.
18808 * config/rs6000/440.md: Adjust.
18809 * config/rs6000/476.md: Adjust.
18810 * config/rs6000/601.md: Adjust.
18811 * config/rs6000/603.md: Adjust.
18812 * config/rs6000/6xx.md: Adjust.
18813 * config/rs6000/7450.md: Adjust.
18814 * config/rs6000/7xx.md: Adjust.
18815 * config/rs6000/8540.md: Adjust.
18816 * config/rs6000/cell.md: Adjust.
18817 * config/rs6000/e300c2c3.md: Adjust.
18818 * config/rs6000/e500mc.md: Adjust.
18819 * config/rs6000/e500mc64.md: Adjust.
18820 * config/rs6000/e5500.md: Adjust.
18821 * config/rs6000/e6500.md: Adjust.
18822 * config/rs6000/mpc.md: Adjust.
18823 * config/rs6000/power4.md: Adjust.
18824 * config/rs6000/power5.md: Adjust.
18825 * config/rs6000/power6.md: Adjust.
18826 * config/rs6000/power7.md: Adjust.
18827 * config/rs6000/power8.md: Adjust.
18828 * config/rs6000/power9.md: Adjust.
18829 * config/rs6000/rs64.md: Adjust.
18830 * config/rs6000/titan.md: Adjust.
18832 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18834 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18835 ix86_indirect_branch_register logic.
18837 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18839 * config/i386/constraints.md (Bs): Update
18840 ix86_indirect_branch_register check. Don't check
18841 ix86_indirect_branch_register with GOT_memory_operand.
18843 * config/i386/predicates.md (GOT_memory_operand): Don't check
18844 ix86_indirect_branch_register here.
18845 (GOT32_symbol_operand): Likewise.
18847 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18849 * config/i386/predicates.md (constant_call_address_operand):
18850 Rewrite ix86_indirect_branch_register logic.
18851 (sibcall_insn_operand): Likewise.
18853 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18855 * config/i386/constraints.md (Bs): Replace
18856 ix86_indirect_branch_thunk_register with
18857 ix86_indirect_branch_register.
18859 * config/i386/i386.md (indirect_jump): Likewise.
18860 (tablejump): Likewise.
18861 (*sibcall_memory): Likewise.
18862 (*sibcall_value_memory): Likewise.
18863 Peepholes of indirect call and jump via memory: Likewise.
18864 * config/i386/i386.opt: Likewise.
18865 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18866 (GOT_memory_operand): Likewise.
18867 (call_insn_operand): Likewise.
18868 (sibcall_insn_operand): Likewise.
18869 (GOT32_symbol_operand): Likewise.
18871 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18873 PR middle-end/83837
18874 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18875 type rather than type addr's type points to.
18876 (expand_omp_atomic_mutex): Likewise.
18877 (expand_omp_atomic): Likewise.
18879 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18882 * config/i386/i386.c (output_indirect_thunk_function): Use
18883 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18884 for __x86_return_thunk.
18886 2018-01-15 Richard Biener <rguenther@suse.de>
18888 PR middle-end/83850
18889 * expmed.c (extract_bit_field_1): Fix typo.
18891 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18894 * config/arm/iterators.md (VF): New mode iterator.
18895 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18896 Remove integer-related logic from pattern.
18897 (neon_vabd<mode>_3): Likewise.
18899 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18901 PR middle-end/82694
18902 * common.opt (fstrict-overflow): No longer an alias.
18903 (fwrapv-pointer): New option.
18904 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18905 also for pointer types based on flag_wrapv_pointer.
18906 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18907 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18908 opts->x_flag_wrapv got set.
18909 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18910 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18911 POINTER_TYPE_OVERFLOW_UNDEFINED.
18912 * match.pd: Likewise in address comparison pattern.
18913 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18915 2018-01-15 Richard Biener <rguenther@suse.de>
18918 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18919 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
18920 Reset type names to their identifier if their TYPE_DECL doesn't
18921 have linkage (and thus is used for ODR and devirt).
18922 (save_debug_info_for_decl): Remove.
18923 (save_debug_info_for_type): Likewise.
18924 (add_tree_to_fld_list): Adjust.
18925 * tree-pretty-print.c (dump_generic_node): Make dumping of
18926 type names more robust.
18928 2018-01-15 Richard Biener <rguenther@suse.de>
18930 * BASE-VER: Bump to 8.0.1.
18932 2018-01-14 Martin Sebor <msebor@redhat.com>
18935 * builtins.c (check_access): Avoid warning when the no-warning bit
18938 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
18940 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18941 * ira-color (allocno_hard_regs_compare): Likewise.
18943 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
18946 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18947 Use .pushsection/.popsection.
18949 2018-01-14 Martin Sebor <msebor@redhat.com>
18952 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18954 2018-01-14 Jakub Jelinek <jakub@redhat.com>
18956 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18957 entry from extra_headers.
18958 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18959 extra_headers, make the list bitwise identical to the i?86-*-* one.
18961 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18963 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18964 -mcmodel=large with -mindirect-branch=thunk,
18965 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
18966 -mfunction-return=thunk-extern.
18967 * doc/invoke.texi: Document -mcmodel=large is incompatible with
18968 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
18969 -mfunction-return=thunk and -mfunction-return=thunk-extern.
18971 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18973 * config/i386/i386.c (print_reg): Print the name of the full
18974 integer register without '%'.
18975 (ix86_print_operand): Handle 'V'.
18976 * doc/extend.texi: Document 'V' modifier.
18978 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18980 * config/i386/constraints.md (Bs): Disallow memory operand for
18981 -mindirect-branch-register.
18983 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18984 (GOT_memory_operand): Likewise.
18985 (call_insn_operand): Likewise.
18986 (sibcall_insn_operand): Likewise.
18987 (GOT32_symbol_operand): Likewise.
18988 * config/i386/i386.md (indirect_jump): Call convert_memory_address
18989 for -mindirect-branch-register.
18990 (tablejump): Likewise.
18991 (*sibcall_memory): Likewise.
18992 (*sibcall_value_memory): Likewise.
18993 Disallow peepholes of indirect call and jump via memory for
18994 -mindirect-branch-register.
18995 (*call_pop): Replace m with Bw.
18996 (*call_value_pop): Likewise.
18997 (*sibcall_pop_memory): Replace m with Bs.
18998 * config/i386/i386.opt (mindirect-branch-register): New option.
18999 * doc/invoke.texi: Document -mindirect-branch-register option.
19001 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19003 * config/i386/i386-protos.h (ix86_output_function_return): New.
19004 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19005 set function_return_type.
19006 (indirect_thunk_name): Add ret_p to indicate thunk for function
19008 (output_indirect_thunk_function): Pass false to
19009 indirect_thunk_name.
19010 (ix86_output_indirect_branch_via_reg): Likewise.
19011 (ix86_output_indirect_branch_via_push): Likewise.
19012 (output_indirect_thunk_function): Create alias for function
19013 return thunk if regno < 0.
19014 (ix86_output_function_return): New function.
19015 (ix86_handle_fndecl_attribute): Handle function_return.
19016 (ix86_attribute_table): Add function_return.
19017 * config/i386/i386.h (machine_function): Add
19018 function_return_type.
19019 * config/i386/i386.md (simple_return_internal): Use
19020 ix86_output_function_return.
19021 (simple_return_internal_long): Likewise.
19022 * config/i386/i386.opt (mfunction-return=): New option.
19023 (indirect_branch): Mention -mfunction-return=.
19024 * doc/extend.texi: Document function_return function attribute.
19025 * doc/invoke.texi: Document -mfunction-return= option.
19027 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19029 * config/i386/i386-opts.h (indirect_branch): New.
19030 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19031 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19032 with local indirect jump when converting indirect call and jump.
19033 (ix86_set_indirect_branch_type): New.
19034 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19035 (indirectlabelno): New.
19036 (indirect_thunk_needed): Likewise.
19037 (indirect_thunk_bnd_needed): Likewise.
19038 (indirect_thunks_used): Likewise.
19039 (indirect_thunks_bnd_used): Likewise.
19040 (INDIRECT_LABEL): Likewise.
19041 (indirect_thunk_name): Likewise.
19042 (output_indirect_thunk): Likewise.
19043 (output_indirect_thunk_function): Likewise.
19044 (ix86_output_indirect_branch_via_reg): Likewise.
19045 (ix86_output_indirect_branch_via_push): Likewise.
19046 (ix86_output_indirect_branch): Likewise.
19047 (ix86_output_indirect_jmp): Likewise.
19048 (ix86_code_end): Call output_indirect_thunk_function if needed.
19049 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19051 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19052 (ix86_attribute_table): Add indirect_branch.
19053 * config/i386/i386.h (machine_function): Add indirect_branch_type
19054 and has_local_indirect_jump.
19055 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19057 (tablejump): Likewise.
19058 (*indirect_jump): Use ix86_output_indirect_jmp.
19059 (*tablejump_1): Likewise.
19060 (simple_return_indirect_internal): Likewise.
19061 * config/i386/i386.opt (mindirect-branch=): New option.
19062 (indirect_branch): New.
19065 (thunk-inline): Likewise.
19066 (thunk-extern): Likewise.
19067 * doc/extend.texi: Document indirect_branch function attribute.
19068 * doc/invoke.texi: Document -mindirect-branch= option.
19070 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19073 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19075 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19077 * ipa-inline.c (want_inline_small_function_p): Return false if
19078 inlining has already failed with CIF_FINAL_ERROR.
19079 (update_caller_keys): Call want_inline_small_function_p before
19081 (update_callee_keys): Likewise.
19083 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19085 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19087 (rs6000_quadword_masked_address_p): Likewise.
19088 (quad_aligned_load_p): Likewise.
19089 (quad_aligned_store_p): Likewise.
19090 (const_load_sequence_p): Add comment to describe the outer-most loop.
19091 (mimic_memory_attributes_and_flags): New function.
19092 (rs6000_gen_stvx): Likewise.
19093 (replace_swapped_aligned_store): Likewise.
19094 (rs6000_gen_lvx): Likewise.
19095 (replace_swapped_aligned_load): Likewise.
19096 (replace_swapped_load_constant): Capitalize argument name in
19097 comment describing this function.
19098 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19099 and stores that access quad-word aligned addresses and replace
19100 with stvx or lvx instructions when appropriate.
19101 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19102 New function prototype.
19103 (rs6000_quadword_masked_address_p): Likewise.
19104 (rs6000_gen_lvx): Likewise.
19105 (rs6000_gen_stvx): Likewise.
19106 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19107 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19108 when memory address is aligned.
19109 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19110 this split to select lvx instruction when memory address is aligned.
19111 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19112 instruction when memory address is aligned.
19113 (*vsx_le_perm_load_v16qi): Likewise.
19114 (four unnamed splitters): Modify to select the stvx instruction
19115 when memory is aligned.
19117 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19119 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19120 which appears in the queue multiple times.
19122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19123 Alan Hayward <alan.hayward@arm.com>
19124 David Sherwood <david.sherwood@arm.com>
19126 * tree-vectorizer.h (vec_lower_bound): New structure.
19127 (_loop_vec_info): Add check_nonzero and lower_bounds.
19128 (LOOP_VINFO_CHECK_NONZERO): New macro.
19129 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19130 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19131 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19132 fields. Make seg_len the distance travelled, not including the
19134 (dr_direction_indicator): Declare.
19135 (dr_zero_step_indicator): Likewise.
19136 (dr_known_forward_stride_p): Likewise.
19137 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19139 (runtime_alias_check_p): Allow runtime alias checks with
19141 (operator ==): Compare access_size and align.
19142 (prune_runtime_alias_test_list): Rework for new distinction between
19143 the access_size and seg_len.
19144 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19146 (get_segment_min_max): New function.
19147 (create_intersect_range_checks): Use it.
19148 (dr_step_indicator): New function.
19149 (dr_direction_indicator): Likewise.
19150 (dr_zero_step_indicator): Likewise.
19151 (dr_known_forward_stride_p): Likewise.
19152 * tree-loop-distribution.c (data_ref_segment_size): Return
19153 DR_STEP * (niters - 1).
19154 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19156 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19157 (vect_preserves_scalar_order_p): New function, split out from...
19158 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19159 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19160 (vect_vfa_access_size): New function.
19161 (vect_vfa_align): Likewise.
19162 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19163 (dump_lower_bound): New function.
19164 (vect_check_lower_bound): Likewise.
19165 (vect_small_gap_p): Likewise.
19166 (vectorizable_with_step_bound_p): Likewise.
19167 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19168 depencies if the vectorization factor is 1. Convert the checks
19169 for nonzero steps into checks on the bounds of DR_STEP. Try using
19170 a bunds check for variable steps if the minimum required step is
19171 relatively small. Update calls to the dr_with_seg_len
19172 constructor and to vect_compile_time_alias.
19173 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19175 (vect_loop_versioning): Call it.
19176 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19178 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19181 Alan Hayward <alan.hayward@arm.com>
19182 David Sherwood <david.sherwood@arm.com>
19184 * doc/sourcebuild.texi (vect_scatter_store): Document.
19185 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19187 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19189 * genopinit.c (main): Add supports_vec_scatter_store and
19190 supports_vec_scatter_store_cached to target_optabs.
19191 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19192 IFN_MASK_SCATTER_STORE.
19193 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19195 * internal-fn.h (internal_store_fn_p): Declare.
19196 (internal_fn_stored_value_index): Likewise.
19197 * internal-fn.c (scatter_store_direct): New macro.
19198 (expand_scatter_store_optab_fn): New function.
19199 (direct_scatter_store_optab_supported_p): New macro.
19200 (internal_store_fn_p): New function.
19201 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19202 IFN_MASK_SCATTER_STORE.
19203 (internal_fn_mask_index): Likewise.
19204 (internal_fn_stored_value_index): New function.
19205 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19206 for scatter stores.
19207 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19208 * optabs-query.c (supports_vec_scatter_store_p): New function.
19209 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19210 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19211 true for scatter stores.
19212 (vect_gather_scatter_fn_p): Handle scatter stores too.
19213 (vect_check_gather_scatter): Consider using scatter stores if
19214 supports_vec_scatter_store_p.
19215 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19216 scatter stores too.
19217 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19218 internal_fn_stored_value_index.
19219 (check_load_store_masking): Handle scatter stores too.
19220 (vect_get_store_rhs): Make public.
19221 (vectorizable_call): Use internal_store_fn_p.
19222 (vectorizable_store): Handle scatter store internal functions.
19223 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19224 when deciding whether the end of the group has been reached.
19225 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19226 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19227 (mask_scatter_store<mode>): New insns.
19229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19230 Alan Hayward <alan.hayward@arm.com>
19231 David Sherwood <david.sherwood@arm.com>
19233 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19234 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19235 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19237 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19238 Use vect_truncate_gather_scatter_offset if we can't treat the
19239 operation as a normal gather load or scatter store.
19240 (get_group_load_store_type): Take the gather_scatter_info
19241 as argument. Try using a gather load or scatter store for
19242 single-element groups.
19243 (get_load_store_type): Update calls to get_group_load_store_type
19244 and vect_use_strided_gather_scatters_p.
19246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19247 Alan Hayward <alan.hayward@arm.com>
19248 David Sherwood <david.sherwood@arm.com>
19250 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19251 optional tree argument.
19252 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19254 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19255 but continue to use the current value as a fallback.
19256 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19257 to compare the updates.
19258 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19259 (get_load_store_type): Use it when handling a strided access.
19260 (vect_get_strided_load_store_ops): New function.
19261 (vect_get_data_ptr_increment): Likewise.
19262 (vectorizable_load): Handle strided gather loads. Always pass
19263 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19266 Alan Hayward <alan.hayward@arm.com>
19267 David Sherwood <david.sherwood@arm.com>
19269 * doc/md.texi (gather_load@var{m}): Document.
19270 (mask_gather_load@var{m}): Likewise.
19271 * genopinit.c (main): Add supports_vec_gather_load and
19272 supports_vec_gather_load_cached to target_optabs.
19273 * optabs-tree.c (init_tree_optimization_optabs): Use
19274 ggc_cleared_alloc to allocate target_optabs.
19275 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19276 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19278 * internal-fn.h (internal_load_fn_p): Declare.
19279 (internal_gather_scatter_fn_p): Likewise.
19280 (internal_fn_mask_index): Likewise.
19281 (internal_gather_scatter_fn_supported_p): Likewise.
19282 * internal-fn.c (gather_load_direct): New macro.
19283 (expand_gather_load_optab_fn): New function.
19284 (direct_gather_load_optab_supported_p): New macro.
19285 (direct_internal_fn_optab): New function.
19286 (internal_load_fn_p): Likewise.
19287 (internal_gather_scatter_fn_p): Likewise.
19288 (internal_fn_mask_index): Likewise.
19289 (internal_gather_scatter_fn_supported_p): Likewise.
19290 * optabs-query.c (supports_at_least_one_mode_p): New function.
19291 (supports_vec_gather_load_p): Likewise.
19292 * optabs-query.h (supports_vec_gather_load_p): Declare.
19293 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19294 and memory_type field.
19295 (NUM_PATTERNS): Bump to 15.
19296 * tree-vect-data-refs.c: Include internal-fn.h.
19297 (vect_gather_scatter_fn_p): New function.
19298 (vect_describe_gather_scatter_call): Likewise.
19299 (vect_check_gather_scatter): Try using internal functions for
19300 gather loads. Recognize existing calls to a gather load function.
19301 (vect_analyze_data_refs): Consider using gather loads if
19302 supports_vec_gather_load_p.
19303 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19304 (vect_get_gather_scatter_offset_type): Likewise.
19305 (vect_convert_mask_for_vectype): Likewise.
19306 (vect_add_conversion_to_patterm): Likewise.
19307 (vect_try_gather_scatter_pattern): Likewise.
19308 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19309 (vect_vect_recog_func_ptrs): Add it.
19310 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19311 internal_fn_mask_index and internal_gather_scatter_fn_p.
19312 (check_load_store_masking): Take the gather_scatter_info as an
19313 argument and handle gather loads.
19314 (vect_get_gather_scatter_ops): New function.
19315 (vectorizable_call): Check internal_load_fn_p.
19316 (vectorizable_load): Likewise. Handle gather load internal
19318 (vectorizable_store): Update call to check_load_store_masking.
19319 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19320 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19321 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19322 (aarch64_gather_scale_operand_d): New predicates.
19323 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19324 (mask_gather_load<mode>): New insns.
19326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19327 Alan Hayward <alan.hayward@arm.com>
19328 David Sherwood <david.sherwood@arm.com>
19330 * optabs.def (fold_left_plus_optab): New optab.
19331 * doc/md.texi (fold_left_plus_@var{m}): Document.
19332 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19333 * internal-fn.c (fold_left_direct): Define.
19334 (expand_fold_left_optab_fn): Likewise.
19335 (direct_fold_left_optab_supported_p): Likewise.
19336 * fold-const-call.c (fold_const_fold_left): New function.
19337 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19338 * tree-parloops.c (valid_reduction_p): New function.
19339 (gather_scalar_reductions): Use it.
19340 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19341 (vect_finish_replace_stmt): Declare.
19342 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19343 (needs_fold_left_reduction_p): New function, split out from...
19344 (vect_is_simple_reduction): ...here. Accept reductions that
19345 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19346 (vect_force_simple_reduction): Also store the reduction type in
19347 the assignment's STMT_VINFO_REDUC_TYPE.
19348 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19349 (merge_with_identity): New function.
19350 (vect_expand_fold_left): Likewise.
19351 (vectorize_fold_left_reduction): Likewise.
19352 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19353 scalar phi in place for it. Check for target support and reject
19354 cases that would reassociate the operation. Defer the transform
19355 phase to vectorize_fold_left_reduction.
19356 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19357 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19358 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19362 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19363 call to ifc_temp_var.
19365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19366 Alan Hayward <alan.hayward@arm.com>
19367 David Sherwood <david.sherwood@arm.com>
19369 * target.def (legitimize_address_displacement): Take the original
19370 offset as a poly_int.
19371 * targhooks.h (default_legitimize_address_displacement): Update
19373 * targhooks.c (default_legitimize_address_displacement): Likewise.
19374 * doc/tm.texi: Regenerate.
19375 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19376 as an argument, moving assert of ad->disp == ad->disp_term to...
19377 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19378 Try calling targetm.legitimize_address_displacement before expanding
19379 the address rather than afterwards, and adjust for the new interface.
19380 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19381 Match the new hook interface. Handle SVE addresses.
19382 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19383 new hook interface.
19385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19387 * Makefile.in (OBJS): Add early-remat.o.
19388 * target.def (select_early_remat_modes): New hook.
19389 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19390 * doc/tm.texi: Regenerate.
19391 * targhooks.h (default_select_early_remat_modes): Declare.
19392 * targhooks.c (default_select_early_remat_modes): New function.
19393 * timevar.def (TV_EARLY_REMAT): New timevar.
19394 * passes.def (pass_early_remat): New pass.
19395 * tree-pass.h (make_pass_early_remat): Declare.
19396 * early-remat.c: New file.
19397 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19399 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19402 Alan Hayward <alan.hayward@arm.com>
19403 David Sherwood <david.sherwood@arm.com>
19405 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19406 vfm1 with a bound_epilog parameter.
19407 (vect_do_peeling): Update calls accordingly, and move the prologue
19408 call earlier in the function. Treat the base bound_epilog as 0 for
19409 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19410 this base when peeling for gaps.
19411 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19412 with fully-masked loops.
19413 (vect_estimate_min_profitable_iters): Handle the single peeled
19414 iteration in that case.
19416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19417 Alan Hayward <alan.hayward@arm.com>
19418 David Sherwood <david.sherwood@arm.com>
19420 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19421 single-element interleaving even if the size is not a power of 2.
19422 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19423 accesses for single-element interleaving if the group size is
19426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19427 Alan Hayward <alan.hayward@arm.com>
19428 David Sherwood <david.sherwood@arm.com>
19430 * doc/md.texi (fold_extract_last_@var{m}): Document.
19431 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19432 * optabs.def (fold_extract_last_optab): New optab.
19433 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19434 * internal-fn.c (fold_extract_direct): New macro.
19435 (expand_fold_extract_optab_fn): Likewise.
19436 (direct_fold_extract_optab_supported_p): Likewise.
19437 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19438 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19439 EXTRACT_LAST_REDUCTION.
19440 (get_initial_def_for_reduction): Do not create an initial vector
19441 for EXTRACT_LAST_REDUCTION reductions.
19442 (vectorizable_reduction): Leave the scalar phi in place for
19443 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19444 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19445 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19446 transform phase to vectorizable_condition.
19447 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19449 (vect_finish_stmt_generation): ...here.
19450 (vect_finish_replace_stmt): New function.
19451 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19452 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19454 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19457 Alan Hayward <alan.hayward@arm.com>
19458 David Sherwood <david.sherwood@arm.com>
19460 * doc/md.texi (extract_last_@var{m}): Document.
19461 * optabs.def (extract_last_optab): New optab.
19462 * internal-fn.def (EXTRACT_LAST): New internal function.
19463 * internal-fn.c (cond_unary_direct): New macro.
19464 (expand_cond_unary_optab_fn): Likewise.
19465 (direct_cond_unary_optab_supported_p): Likewise.
19466 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19467 loops using EXTRACT_LAST.
19468 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19469 (extract_last_<mode>): ...this optab.
19470 (vec_extract<mode><Vel>): Update accordingly.
19472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19473 Alan Hayward <alan.hayward@arm.com>
19474 David Sherwood <david.sherwood@arm.com>
19476 * target.def (empty_mask_is_expensive): New hook.
19477 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19478 * doc/tm.texi: Regenerate.
19479 * targhooks.h (default_empty_mask_is_expensive): Declare.
19480 * targhooks.c (default_empty_mask_is_expensive): New function.
19481 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19482 if the target says that empty masks are expensive.
19483 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19485 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19488 Alan Hayward <alan.hayward@arm.com>
19489 David Sherwood <david.sherwood@arm.com>
19491 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19492 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19493 (vect_use_loop_mask_for_alignment_p): New function.
19494 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19495 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19496 niters_skip argument. Make sure that the first niters_skip elements
19497 of the first iteration are inactive.
19498 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19499 Update call to vect_set_loop_masks_directly.
19500 (get_misalign_in_elems): New function, split out from...
19501 (vect_gen_prolog_loop_niters): ...here.
19502 (vect_update_init_of_dr): Take a code argument that specifies whether
19503 the adjustment should be added or subtracted.
19504 (vect_update_init_of_drs): Likewise.
19505 (vect_prepare_for_masked_peels): New function.
19506 (vect_do_peeling): Skip prologue peeling if we're using a mask
19507 instead. Update call to vect_update_inits_of_drs.
19508 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19510 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19511 alignment. Do not include the number of peeled iterations in
19512 the minimum threshold in that case.
19513 (vectorizable_induction): Adjust the start value down by
19514 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19515 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19516 Take the number of skipped iterations into account when calculating
19518 * tree-vect-stmts.c (vect_gen_while_not): New function.
19520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19521 Alan Hayward <alan.hayward@arm.com>
19522 David Sherwood <david.sherwood@arm.com>
19524 * doc/sourcebuild.texi (vect_fully_masked): Document.
19525 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19526 default value to 0.
19527 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19529 (vect_analyze_loop_2): ...here. Don't check the vectorization
19530 factor against the number of loop iterations if the loop is
19533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19534 Alan Hayward <alan.hayward@arm.com>
19535 David Sherwood <david.sherwood@arm.com>
19537 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19538 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19539 (dump_groups): Update accordingly.
19540 (iv_use::mem_type): New member variable.
19541 (address_p): New function.
19542 (record_use): Add a mem_type argument and initialize the new
19544 (record_group_use): Add a mem_type argument. Use address_p.
19545 Remove obsolete null checks of base_object. Update call to record_use.
19546 (find_interesting_uses_op): Update call to record_group_use.
19547 (find_interesting_uses_cond): Likewise.
19548 (find_interesting_uses_address): Likewise.
19549 (get_mem_type_for_internal_fn): New function.
19550 (find_address_like_use): Likewise.
19551 (find_interesting_uses_stmt): Try find_address_like_use before
19552 calling find_interesting_uses_op.
19553 (addr_offset_valid_p): Use the iv mem_type field as the type
19554 of the addressed memory.
19555 (add_autoinc_candidates): Likewise.
19556 (get_address_cost): Likewise.
19557 (split_small_address_groups_p): Use address_p.
19558 (split_address_groups): Likewise.
19559 (add_iv_candidate_for_use): Likewise.
19560 (autoinc_possible_for_pair): Likewise.
19561 (rewrite_groups): Likewise.
19562 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19563 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19564 (get_alias_ptr_type_for_ptr_address): New function.
19565 (rewrite_use_address): Rewrite address uses in calls that were
19566 identified by find_address_like_use.
19568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19569 Alan Hayward <alan.hayward@arm.com>
19570 David Sherwood <david.sherwood@arm.com>
19572 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19574 * gimple-expr.h (is_gimple_addressable: Likewise.
19575 * gimple-expr.c (is_gimple_address): Likewise.
19576 * internal-fn.c (expand_call_mem_ref): New function.
19577 (expand_mask_load_optab_fn): Use it.
19578 (expand_mask_store_optab_fn): Likewise.
19580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19581 Alan Hayward <alan.hayward@arm.com>
19582 David Sherwood <david.sherwood@arm.com>
19584 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19585 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19586 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19587 (cond_umax@var{mode}): Document.
19588 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19589 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19590 (cond_umin_optab, cond_umax_optab): New optabs.
19591 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19592 (COND_IOR, COND_XOR): New internal functions.
19593 * internal-fn.h (get_conditional_internal_fn): Declare.
19594 * internal-fn.c (cond_binary_direct): New macro.
19595 (expand_cond_binary_optab_fn): Likewise.
19596 (direct_cond_binary_optab_supported_p): Likewise.
19597 (get_conditional_internal_fn): New function.
19598 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19599 Cope with reduction statements that are vectorized as calls rather
19601 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19602 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19603 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19604 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19605 (UNSPEC_COND_EOR): New unspecs.
19606 (optab): Add mappings for them.
19607 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19608 (sve_int_op, sve_fp_op): New int attributes.
19610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19611 Alan Hayward <alan.hayward@arm.com>
19612 David Sherwood <david.sherwood@arm.com>
19614 * optabs.def (while_ult_optab): New optab.
19615 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19616 * internal-fn.def (WHILE_ULT): New internal function.
19617 * internal-fn.h (direct_internal_fn_supported_p): New override
19618 that takes two types as argument.
19619 * internal-fn.c (while_direct): New macro.
19620 (expand_while_optab_fn): New function.
19621 (convert_optab_supported_p): Likewise.
19622 (direct_while_optab_supported_p): New macro.
19623 * wide-int.h (wi::udiv_ceil): New function.
19624 * tree-vectorizer.h (rgroup_masks): New structure.
19625 (vec_loop_masks): New typedef.
19626 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19627 and fully_masked_p.
19628 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19629 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19630 (vect_max_vf): New function.
19631 (slpeel_make_loop_iterate_ntimes): Delete.
19632 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19633 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19634 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19635 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19636 internal-fn.h, stor-layout.h and optabs-query.h.
19637 (vect_set_loop_mask): New function.
19638 (add_preheader_seq): Likewise.
19639 (add_header_seq): Likewise.
19640 (interleave_supported_p): Likewise.
19641 (vect_maybe_permute_loop_masks): Likewise.
19642 (vect_set_loop_masks_directly): Likewise.
19643 (vect_set_loop_condition_masked): Likewise.
19644 (vect_set_loop_condition_unmasked): New function, split out from
19645 slpeel_make_loop_iterate_ntimes.
19646 (slpeel_make_loop_iterate_ntimes): Rename to..
19647 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19648 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19649 (vect_do_peeling): Update call accordingly.
19650 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19652 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19653 mask_compare_type, can_fully_mask_p and fully_masked_p.
19654 (release_vec_loop_masks): New function.
19655 (_loop_vec_info): Use it to free the loop masks.
19656 (can_produce_all_loop_masks_p): New function.
19657 (vect_get_max_nscalars_per_iter): Likewise.
19658 (vect_verify_full_masking): Likewise.
19659 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19660 retries, and free the mask rgroups before retrying. Check loop-wide
19661 reasons for disallowing fully-masked loops. Make the final decision
19662 about whether use a fully-masked loop or not.
19663 (vect_estimate_min_profitable_iters): Do not assume that peeling
19664 for the number of iterations will be needed for fully-masked loops.
19665 (vectorizable_reduction): Disable fully-masked loops.
19666 (vectorizable_live_operation): Likewise.
19667 (vect_halve_mask_nunits): New function.
19668 (vect_double_mask_nunits): Likewise.
19669 (vect_record_loop_mask): Likewise.
19670 (vect_get_loop_mask): Likewise.
19671 (vect_transform_loop): Handle the case in which the final loop
19672 iteration might handle a partial vector. Call vect_set_loop_condition
19673 instead of slpeel_make_loop_iterate_ntimes.
19674 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19675 (check_load_store_masking): New function.
19676 (prepare_load_store_mask): Likewise.
19677 (vectorizable_store): Handle fully-masked loops.
19678 (vectorizable_load): Likewise.
19679 (supportable_widening_operation): Use vect_halve_mask_nunits for
19681 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19682 (vect_gen_while): New function.
19683 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19684 (aarch64_uqdec<mode>): New insn.
19686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19687 Alan Hayward <alan.hayward@arm.com>
19688 David Sherwood <david.sherwood@arm.com>
19690 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19691 (reduc_xor_scal_optab): New optabs.
19692 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19693 (reduc_xor_scal_@var{m}): Document.
19694 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19695 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19696 internal functions.
19697 * fold-const-call.c (fold_const_call): Handle them.
19698 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19699 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19700 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19701 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19702 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19703 (UNSPEC_XORV): New unspecs.
19704 (optab): Add entries for them.
19705 (BITWISEV): New int iterator.
19706 (bit_reduc_op): New int attributes.
19708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19709 Alan Hayward <alan.hayward@arm.com>
19710 David Sherwood <david.sherwood@arm.com>
19712 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19713 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19714 * optabs.def (vec_shl_insert_optab): New optab.
19715 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19716 (duplicate_and_interleave): Likewise.
19717 * tree-vect-loop.c: Include internal-fn.h.
19718 (neutral_op_for_slp_reduction): New function, split out from
19719 get_initial_defs_for_reduction.
19720 (get_initial_def_for_reduction): Handle option 2 for variable-length
19721 vectors by loading the neutral value into a vector and then shifting
19722 the initial value into element 0.
19723 (get_initial_defs_for_reduction): Replace the code argument with
19724 the neutral value calculated by neutral_op_for_slp_reduction.
19725 Use gimple_build_vector for constant-length vectors.
19726 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19727 but the first group_size elements have a neutral value.
19728 Use duplicate_and_interleave otherwise.
19729 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19730 Update call to get_initial_defs_for_reduction. Handle SLP
19731 reductions for variable-length vectors by creating one vector
19732 result for each scalar result, with the elements associated
19733 with other scalar results stubbed out with the neutral value.
19734 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19735 Require IFN_VEC_SHL_INSERT for double reductions on
19736 variable-length vectors, or SLP reductions that have
19737 a neutral value. Require can_duplicate_and_interleave_p
19738 support for variable-length unchained SLP reductions if there
19739 is no neutral value, such as for MIN/MAX reductions. Also require
19740 the number of vector elements to be a multiple of the number of
19741 SLP statements when doing variable-length unchained SLP reductions.
19742 Update call to vect_create_epilog_for_reduction.
19743 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19744 and remove initial values.
19745 (duplicate_and_interleave): Make public.
19746 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19747 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19750 Alan Hayward <alan.hayward@arm.com>
19751 David Sherwood <david.sherwood@arm.com>
19753 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19754 (can_duplicate_and_interleave_p): New function.
19755 (vect_get_and_check_slp_defs): Take the vector of statements
19756 rather than just the current one. Remove excess parentheses.
19757 Restriction rejectinon of vect_constant_def and vect_external_def
19758 for variable-length vectors to boolean types, or types for which
19759 can_duplicate_and_interleave_p is false.
19760 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19761 (duplicate_and_interleave): New function.
19762 (vect_get_constant_vectors): Use gimple_build_vector for
19763 constant-length vectors and suitable variable-length constant
19764 vectors. Use duplicate_and_interleave for other variable-length
19765 vectors. Don't defer the update when inserting new statements.
19767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19768 Alan Hayward <alan.hayward@arm.com>
19769 David Sherwood <david.sherwood@arm.com>
19771 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19772 min_profitable_iters doesn't go negative.
19774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19775 Alan Hayward <alan.hayward@arm.com>
19776 David Sherwood <david.sherwood@arm.com>
19778 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19779 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19780 * optabs.def (vec_mask_load_lanes_optab): New optab.
19781 (vec_mask_store_lanes_optab): Likewise.
19782 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19783 (MASK_STORE_LANES): Likewise.
19784 * internal-fn.c (mask_load_lanes_direct): New macro.
19785 (mask_store_lanes_direct): Likewise.
19786 (expand_mask_load_optab_fn): Handle masked operations.
19787 (expand_mask_load_lanes_optab_fn): New macro.
19788 (expand_mask_store_optab_fn): Handle masked operations.
19789 (expand_mask_store_lanes_optab_fn): New macro.
19790 (direct_mask_load_lanes_optab_supported_p): Likewise.
19791 (direct_mask_store_lanes_optab_supported_p): Likewise.
19792 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19794 (vect_load_lanes_supported): Likewise.
19795 * tree-vect-data-refs.c (strip_conversion): New function.
19796 (can_group_stmts_p): Likewise.
19797 (vect_analyze_data_ref_accesses): Use it instead of checking
19798 for a pair of assignments.
19799 (vect_store_lanes_supported): Take a masked_p parameter.
19800 (vect_load_lanes_supported): Likewise.
19801 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19802 vect_store_lanes_supported and vect_load_lanes_supported.
19803 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19804 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19805 parameter. Don't allow gaps for masked accesses.
19806 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19807 and vect_load_lanes_supported.
19808 (get_load_store_type): Take a masked_p parameter and update
19809 call to get_group_load_store_type.
19810 (vectorizable_store): Update call to get_load_store_type.
19811 Handle IFN_MASK_STORE_LANES.
19812 (vectorizable_load): Update call to get_load_store_type.
19813 Handle IFN_MASK_LOAD_LANES.
19815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19816 Alan Hayward <alan.hayward@arm.com>
19817 David Sherwood <david.sherwood@arm.com>
19819 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19821 * config/aarch64/aarch64-protos.h
19822 (aarch64_sve_struct_memory_operand_p): Declare.
19823 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19824 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19825 (VPRED, vpred): Handle SVE structure modes.
19826 * config/aarch64/constraints.md (Utx): New constraint.
19827 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19828 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19829 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19830 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19831 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19832 structure modes. Split into pieces after RA.
19833 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19834 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19836 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19837 SVE structure modes.
19838 (aarch64_classify_address): Likewise.
19839 (sizetochar): Move earlier in file.
19840 (aarch64_print_operand): Handle SVE register lists.
19841 (aarch64_array_mode): New function.
19842 (aarch64_sve_struct_memory_operand_p): Likewise.
19843 (TARGET_ARRAY_MODE): Redefine.
19845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19846 Alan Hayward <alan.hayward@arm.com>
19847 David Sherwood <david.sherwood@arm.com>
19849 * target.def (array_mode): New target hook.
19850 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19851 * doc/tm.texi: Regenerate.
19852 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19853 * hooks.c (hook_optmode_mode_uhwi_none): New function.
19854 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19855 targetm.array_mode.
19856 * stor-layout.c (mode_for_array): Likewise. Support polynomial
19859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19860 Alan Hayward <alan.hayward@arm.com>
19861 David Sherwood <david.sherwood@arm.com>
19863 * fold-const.c (fold_binary_loc): Check the argument types
19864 rather than the result type when testing for a vector operation.
19866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19868 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19869 * doc/tm.texi: Regenerate.
19871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19872 Alan Hayward <alan.hayward@arm.com>
19873 David Sherwood <david.sherwood@arm.com>
19875 * doc/invoke.texi (-msve-vector-bits=): Document new option.
19876 (sve): Document new AArch64 extension.
19877 * doc/md.texi (w): Extend the description of the AArch64
19878 constraint to include SVE vectors.
19879 (Upl, Upa): Document new AArch64 predicate constraints.
19880 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19882 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19883 (msve-vector-bits=): New option.
19884 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19885 SVE when these are disabled.
19886 (sve): New extension.
19887 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19888 modes. Adjust their number of units based on aarch64_sve_vg.
19889 (MAX_BITSIZE_MODE_ANY_MODE): Define.
19890 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19891 aarch64_addr_query_type.
19892 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19893 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19894 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19895 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19896 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19897 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19898 (aarch64_simd_imm_zero_p): Delete.
19899 (aarch64_check_zero_based_sve_index_immediate): Declare.
19900 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19901 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19902 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19903 (aarch64_sve_float_mul_immediate_p): Likewise.
19904 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19905 rather than an rtx.
19906 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19907 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19908 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19909 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19910 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19911 (aarch64_regmode_natural_size): Likewise.
19912 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19913 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19915 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19916 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19917 for VG and the SVE predicate registers.
19918 (V_ALIASES): Add a "z"-prefixed alias.
19919 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19920 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19921 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19922 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19923 (REG_CLASS_NAMES): Add entries for them.
19924 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
19925 and the predicate registers.
19926 (aarch64_sve_vg): Declare.
19927 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19928 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19929 (REGMODE_NATURAL_SIZE): Define.
19930 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19932 * config/aarch64/aarch64.c: Include cfgrtl.h.
19933 (simd_immediate_info): Add a constructor for series vectors,
19934 and an associated step field.
19935 (aarch64_sve_vg): New variable.
19936 (aarch64_dbx_register_number): Handle VG and the predicate registers.
19937 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19938 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19939 (VEC_ANY_DATA, VEC_STRUCT): New constants.
19940 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19941 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19942 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19943 (aarch64_get_mask_mode): New functions.
19944 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19945 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19946 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
19947 predicate modes and predicate registers. Explicitly restrict
19948 GPRs to modes of 16 bytes or smaller. Only allow FP registers
19949 to store a vector mode if it is recognized by
19950 aarch64_classify_vector_mode.
19951 (aarch64_regmode_natural_size): New function.
19952 (aarch64_hard_regno_caller_save_mode): Return the original mode
19954 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19955 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19956 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19957 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19959 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
19960 does not overlap dest if the function is frame-related. Handle
19962 (aarch64_split_add_offset): New function.
19963 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19964 them aarch64_add_offset.
19965 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
19966 and update call to aarch64_sub_sp.
19967 (aarch64_add_cfa_expression): New function.
19968 (aarch64_expand_prologue): Pass extra temporary registers to the
19969 functions above. Handle the case in which we need to emit new
19970 DW_CFA_expressions for registers that were originally saved
19971 relative to the stack pointer, but now have to be expressed
19972 relative to the frame pointer.
19973 (aarch64_output_mi_thunk): Pass extra temporary registers to the
19975 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
19976 IP0 and IP1 values for SVE frames.
19977 (aarch64_expand_vec_series): New function.
19978 (aarch64_expand_sve_widened_duplicate): Likewise.
19979 (aarch64_expand_sve_const_vector): Likewise.
19980 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
19981 Handle SVE constants. Use emit_move_insn to move a force_const_mem
19982 into the register, rather than emitting a SET directly.
19983 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
19984 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
19985 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
19986 (offset_9bit_signed_scaled_p): New functions.
19987 (aarch64_replicate_bitmask_imm): New function.
19988 (aarch64_bitmask_imm): Use it.
19989 (aarch64_cannot_force_const_mem): Reject expressions involving
19990 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
19991 (aarch64_classify_index): Handle SVE indices, by requiring
19992 a plain register index with a scale that matches the element size.
19993 (aarch64_classify_address): Handle SVE addresses. Assert that
19994 the mode of the address is VOIDmode or an integer mode.
19995 Update call to aarch64_classify_symbol.
19996 (aarch64_classify_symbolic_expression): Update call to
19997 aarch64_classify_symbol.
19998 (aarch64_const_vec_all_in_range_p): New function.
19999 (aarch64_print_vector_float_operand): Likewise.
20000 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20001 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20002 and the FP immediates 1.0 and 0.5.
20003 (aarch64_print_address_internal): Handle SVE addresses.
20004 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20005 (aarch64_regno_regclass): Handle predicate registers.
20006 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20008 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20009 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20010 (aarch64_convert_sve_vector_bits): New function.
20011 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20012 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20013 rather than an rtx.
20014 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20015 Handle SVE vector and predicate modes. Accept VL-based constants
20016 that need only one temporary register, and VL offsets that require
20017 no temporary registers.
20018 (aarch64_conditional_register_usage): Mark the predicate registers
20019 as fixed if SVE isn't available.
20020 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20021 Return true for SVE vector and predicate modes.
20022 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20023 rather than an unsigned int. Handle SVE modes.
20024 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20026 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20028 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20029 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20030 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20031 (aarch64_sve_float_mul_immediate_p): New functions.
20032 (aarch64_sve_valid_immediate): New function.
20033 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20034 Explicitly reject structure modes. Check for INDEX constants.
20035 Handle PTRUE and PFALSE constants.
20036 (aarch64_check_zero_based_sve_index_immediate): New function.
20037 (aarch64_simd_imm_zero_p): Delete.
20038 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20039 vector modes. Accept constants in the range of CNT[BHWD].
20040 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20041 ask for an Advanced SIMD mode.
20042 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20043 (aarch64_simd_vector_alignment): Handle SVE predicates.
20044 (aarch64_vectorize_preferred_vector_alignment): New function.
20045 (aarch64_simd_vector_alignment_reachable): Use it instead of
20047 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20048 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20050 (MAX_VECT_LEN): Delete.
20051 (expand_vec_perm_d): Add a vec_flags field.
20052 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20053 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20054 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20056 (aarch64_evpc_rev): Rename to...
20057 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20058 (aarch64_evpc_rev_global): New function.
20059 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20060 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20062 (aarch64_evpc_sve_tbl): New function.
20063 (aarch64_expand_vec_perm_const_1): Update after rename of
20064 aarch64_evpc_rev. Handle SVE permutes too, trying
20065 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20066 than aarch64_evpc_tbl.
20067 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20068 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20069 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20070 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20071 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20072 (aarch64_expand_sve_vcond): New functions.
20073 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20074 of aarch64_vector_mode_p.
20075 (aarch64_dwarf_poly_indeterminate_value): New function.
20076 (aarch64_compute_pressure_classes): Likewise.
20077 (aarch64_can_change_mode_class): Likewise.
20078 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20079 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20080 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20081 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20082 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20083 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20084 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20085 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20087 (Dn, Dl, Dr): Accept const as well as const_vector.
20088 (Dz): Likewise. Compare against CONST0_RTX.
20089 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20090 of "vector" where appropriate.
20091 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20092 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20093 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20094 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20095 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20096 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20097 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20098 (v_int_equiv): Extend to SVE modes.
20099 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20101 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20102 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20103 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20104 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20105 (SVE_COND_FP_CMP): New int iterators.
20106 (perm_hilo): Handle the new unpack unspecs.
20107 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20109 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20110 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20111 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20112 (aarch64_equality_operator, aarch64_constant_vector_operand)
20113 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20114 (aarch64_sve_nonimmediate_operand): Likewise.
20115 (aarch64_sve_general_operand): Likewise.
20116 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20117 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20118 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20119 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20120 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20121 (aarch64_sve_float_arith_immediate): Likewise.
20122 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20123 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20124 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20125 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20126 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20127 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20128 (aarch64_sve_float_arith_operand): Likewise.
20129 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20130 (aarch64_sve_float_mul_operand): Likewise.
20131 (aarch64_sve_vec_perm_operand): Likewise.
20132 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20133 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20134 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20135 as well as const_vector.
20136 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20137 in file. Use CONST0_RTX and CONSTM1_RTX.
20138 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20139 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20140 Use aarch64_simd_imm_zero.
20141 * config/aarch64/aarch64-sve.md: New file.
20142 * config/aarch64/aarch64.md: Include it.
20143 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20144 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20145 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20146 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20147 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20148 (sve): New attribute.
20149 (enabled): Disable instructions with the sve attribute unless
20151 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20152 aarch64_expand_mov_immediate.
20153 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20154 CNT[BHSD] immediates.
20155 (movti): Split CONST_POLY_INT moves into two halves.
20156 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20157 Split additions that need a temporary here if the destination
20158 is the stack pointer.
20159 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20160 (*add<mode>3_poly_1): New instruction.
20161 (set_clobber_cc): New expander.
20163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20165 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20166 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20167 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20168 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20169 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20170 Change innermode from fixed_mode_size to machine_mode.
20171 (simplify_subreg): Update call accordingly. Handle a constant-sized
20172 subreg of a variable-length CONST_VECTOR.
20174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20175 Alan Hayward <alan.hayward@arm.com>
20176 David Sherwood <david.sherwood@arm.com>
20178 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20179 (add_offset_to_base): New function, split out from...
20180 (create_mem_ref): ...here. When handling a scale other than 1,
20181 check first whether the address is valid without the offset.
20182 Add it into the base if so, leaving the index and scale as-is.
20184 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20187 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20188 fold_for_warn before checking if arg2 is INTEGER_CST.
20190 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20192 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20193 (store_multiple_operation): Delete.
20194 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20195 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20196 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20197 guarded by TARGET_STRING.
20198 (rs6000_output_load_multiple): Delete.
20199 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20200 OPTION_MASK_STRING / TARGET_STRING handling.
20201 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20202 (const rs6000_opt_masks) <"string">: Change mask to 0.
20203 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20204 (MASK_STRING): Delete.
20205 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20207 (load_multiple): Delete.
20214 (store_multiple): Delete.
20221 (movmemsi_8reg): Delete.
20222 (corresponding unnamed define_insn): Delete.
20223 (movmemsi_6reg): Delete.
20224 (corresponding unnamed define_insn): Delete.
20225 (movmemsi_4reg): Delete.
20226 (corresponding unnamed define_insn): Delete.
20227 (movmemsi_2reg): Delete.
20228 (corresponding unnamed define_insn): Delete.
20229 (movmemsi_1reg): Delete.
20230 (corresponding unnamed define_insn): Delete.
20231 * config/rs6000/rs6000.opt (mno-string): New.
20232 (mstring): Replace by deprecation warning stub.
20233 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20235 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20237 * regrename.c (regrename_do_replace): If replacing the same
20238 reg multiple times, try to reuse last created gen_raw_REG.
20241 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20242 main to workaround a bug in GDB.
20244 2018-01-12 Tom de Vries <tom@codesourcery.com>
20247 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20249 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20251 PR rtl-optimization/80481
20252 * ira-color.c (get_cap_member): New function.
20253 (allocnos_conflict_by_live_ranges_p): Use it.
20254 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20255 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20257 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20260 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20261 (*saddl_se_1): Ditto.
20262 (*ssubsi_1): Ditto.
20263 (*ssubl_se_1): Ditto.
20265 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20267 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20268 rather than wi::to_widest for DR_INITs.
20269 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20270 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20271 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20273 (vect_analyze_group_access_1): Note that here.
20275 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20277 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20278 polynomial type sizes.
20280 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20282 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20283 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20284 (gimple_add_tmp_var): Likewise.
20286 2018-01-12 Martin Liska <mliska@suse.cz>
20288 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20289 (gimple_alloc_sizes): Likewise.
20290 (dump_gimple_statistics): Use PRIu64 in printf format.
20291 * gimple.h: Change uint64_t to int.
20293 2018-01-12 Martin Liska <mliska@suse.cz>
20295 * tree-core.h: Use uint64_t instead of int.
20296 * tree.c (tree_node_counts): Likewise.
20297 (tree_node_sizes): Likewise.
20298 (dump_tree_statistics): Use PRIu64 in printf format.
20300 2018-01-12 Martin Liska <mliska@suse.cz>
20302 * Makefile.in: As qsort_chk is implemented in vec.c, add
20303 vec.o to linkage of gencfn-macros.
20304 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20305 passing the info to record_node_allocation_statistics.
20306 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20308 * ggc-common.c (struct ggc_usage): Add operator== and use
20309 it in operator< and compare function.
20310 * mem-stats.h (struct mem_usage): Likewise.
20311 * vec.c (struct vec_usage): Remove operator< and compare
20312 function. Can be simply inherited.
20314 2018-01-12 Martin Jambor <mjambor@suse.cz>
20317 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20318 * tree-ssa-math-opts.c: Include domwalk.h.
20319 (convert_mult_to_fma_1): New function.
20320 (fma_transformation_info): New type.
20321 (fma_deferring_state): Likewise.
20322 (cancel_fma_deferring): New function.
20323 (result_of_phi): Likewise.
20324 (last_fma_candidate_feeds_initial_phi): Likewise.
20325 (convert_mult_to_fma): Added deferring logic, split actual
20326 transformation to convert_mult_to_fma_1.
20327 (math_opts_dom_walker): New type.
20328 (math_opts_dom_walker::after_dom_children): New method, body moved
20329 here from pass_optimize_widening_mul::execute, added deferring logic
20331 (pass_optimize_widening_mul::execute): Moved most of code to
20332 math_opts_dom_walker::after_dom_children.
20333 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20334 * config/i386/i386.c (ix86_option_override_internal): Added
20335 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20337 2018-01-12 Richard Biener <rguenther@suse.de>
20340 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20341 inline instance vars.
20343 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20346 * config/rx/rx.c (rx_is_restricted_memory_address):
20347 Handle SUBREG case.
20349 2018-01-12 Richard Biener <rguenther@suse.de>
20351 PR tree-optimization/80846
20352 * target.def (split_reduction): New target hook.
20353 * targhooks.c (default_split_reduction): New function.
20354 * targhooks.h (default_split_reduction): Declare.
20355 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20356 target requests first reduce vectors by combining low and high
20358 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20359 (get_vectype_for_scalar_type_and_size): Export.
20360 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20361 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20362 * doc/tm.texi: Regenerate.
20363 * config/i386/i386.c (ix86_split_reduction): Implement
20364 TARGET_VECTORIZE_SPLIT_REDUCTION.
20366 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20369 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20370 in PIC mode except for TARGET_VXWORKS_RTP.
20371 * config/sparc/sparc.c: Include cfgrtl.h.
20372 (TARGET_INIT_PIC_REG): Define.
20373 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20374 (sparc_pic_register_p): New predicate.
20375 (sparc_legitimate_address_p): Use it.
20376 (sparc_legitimize_pic_address): Likewise.
20377 (sparc_delegitimize_address): Likewise.
20378 (sparc_mode_dependent_address_p): Likewise.
20379 (gen_load_pcrel_sym): Remove 4th parameter.
20380 (load_got_register): Adjust call to above. Remove obsolete stuff.
20381 (sparc_expand_prologue): Do not call load_got_register here.
20382 (sparc_flat_expand_prologue): Likewise.
20383 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20384 (sparc_use_pseudo_pic_reg): New function.
20385 (sparc_init_pic_reg): Likewise.
20386 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20387 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20389 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20391 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20392 Add item for branch_cost.
20394 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20396 PR rtl-optimization/83565
20397 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20398 not extend the result to a larger mode for rotate operations.
20399 (num_sign_bit_copies1): Likewise.
20401 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20404 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20406 Use values-Xc.o for -pedantic.
20407 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20409 2018-01-12 Martin Liska <mliska@suse.cz>
20412 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20414 (possible_polymorphic_call_targets): Use it.
20415 (ipa_devirt): Likewise.
20417 2018-01-12 Martin Liska <mliska@suse.cz>
20419 * profile-count.h (enum profile_quality): Use 0 as invalid
20420 enum value of profile_quality.
20422 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20424 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20425 -mext-string options.
20427 2018-01-12 Richard Biener <rguenther@suse.de>
20429 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20430 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20431 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20433 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20435 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20437 * configure.ac (--with-long-double-format): Add support for the
20438 configuration option to change the default long double format on
20440 * config.gcc (powerpc*-linux*-*): Likewise.
20441 * configure: Regenerate.
20442 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20443 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20444 used without modification.
20446 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20448 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20449 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20450 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20451 MISC_BUILTIN_SPEC_BARRIER.
20452 (rs6000_init_builtins): Likewise.
20453 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20455 (speculation_barrier): New define_insn.
20456 * doc/extend.texi: Document __builtin_speculation_barrier.
20458 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20461 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20462 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20463 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20465 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20466 integral modes instead of "ss" and "sd".
20467 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20468 vectors with 32-bit and 64-bit elements.
20469 (vecdupssescalarmodesuffix): New mode attribute.
20470 (vec_dup<mode>): Use it.
20472 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20475 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20476 frame if argument is passed on stack.
20478 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20481 * ree.c (combine_reaching_defs): Optimize also
20482 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20483 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20485 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20487 PR middle-end/83189
20488 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20490 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20492 PR middle-end/83718
20493 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20494 after they are computed.
20496 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20498 PR tree-optimization/83695
20499 * gimple-loop-linterchange.cc
20500 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20501 reset cached scev information after interchange.
20502 (pass_linterchange::execute): Remove call to scev_reset_htab.
20504 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20506 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20507 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20508 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20509 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20510 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20511 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20512 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20513 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20514 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20515 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20516 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20517 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20518 (V_lane_reg): Likewise.
20519 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20521 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20522 (vfmal_lane_low<mode>_intrinsic,
20523 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20524 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20525 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20526 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20527 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20528 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20530 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20532 * config/arm/arm-cpus.in (fp16fml): New feature.
20533 (ALL_SIMD): Add fp16fml.
20534 (armv8.2-a): Add fp16fml as an option.
20535 (armv8.3-a): Likewise.
20536 (armv8.4-a): Add fp16fml as part of fp16.
20537 * config/arm/arm.h (TARGET_FP16FML): Define.
20538 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20540 * config/arm/arm-modes.def (V2HF): Define.
20541 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20542 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20543 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20544 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20545 vfmsl_low, vfmsl_high): New set of builtins.
20546 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20547 (vfml_op): New code attribute.
20548 (VFMLHALVES): New int iterator.
20549 (VFML, VFMLSEL): New mode attributes.
20550 (V_reg): Define mapping for V2HF.
20551 (V_hi, V_lo): New mode attributes.
20552 (VF_constraint): Likewise.
20553 (vfml_half, vfml_half_selector): New int attributes.
20554 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20556 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20557 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20559 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20560 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20561 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20562 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20564 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20565 Document new effective target and option set.
20567 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20569 * config/arm/arm-cpus.in (armv8_4): New feature.
20570 (ARMv8_4a): New fgroup.
20571 (armv8.4-a): New arch.
20572 * config/arm/arm-tables.opt: Regenerate.
20573 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20574 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20575 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20576 Add matching rules for -march=armv8.4-a and extensions.
20577 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20579 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20582 * config/rx/rx.md (BW): New mode attribute.
20583 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20585 2018-01-11 Richard Biener <rguenther@suse.de>
20587 PR tree-optimization/83435
20588 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20589 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20590 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20592 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20593 Alan Hayward <alan.hayward@arm.com>
20594 David Sherwood <david.sherwood@arm.com>
20596 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20598 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20599 (aarch64_print_address_internal): Use it to check for a zero offset.
20601 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20602 Alan Hayward <alan.hayward@arm.com>
20603 David Sherwood <david.sherwood@arm.com>
20605 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20606 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20607 Return a poly_int64 rather than a HOST_WIDE_INT.
20608 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20609 rather than a HOST_WIDE_INT.
20610 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20611 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20612 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20613 final_offset from HOST_WIDE_INT to poly_int64.
20614 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20615 to_constant when getting the number of units in an Advanced SIMD
20617 (aarch64_builtin_vectorized_function): Check for a constant number
20619 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20621 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20622 attribute instead of GET_MODE_NUNITS.
20623 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20624 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20625 GET_MODE_SIZE for fixed-size registers.
20626 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20627 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20628 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20629 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20630 (aarch64_print_operand, aarch64_print_address_internal)
20631 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20632 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20633 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20634 Handle polynomial GET_MODE_SIZE.
20635 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20636 wider than SImode without modification.
20637 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20638 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20639 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20640 passing and returning SVE modes.
20641 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20642 rather than GEN_INT.
20643 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20644 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20645 (aarch64_allocate_and_probe_stack_space): Likewise.
20646 (aarch64_layout_frame): Cope with polynomial offsets.
20647 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20648 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20649 polynomial offsets.
20650 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20651 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20652 poly_int64 rather than a HOST_WIDE_INT.
20653 (aarch64_get_separate_components, aarch64_process_components)
20654 (aarch64_expand_prologue, aarch64_expand_epilogue)
20655 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20656 (aarch64_anchor_offset): New function, split out from...
20657 (aarch64_legitimize_address): ...here.
20658 (aarch64_builtin_vectorization_cost): Handle polynomial
20659 TYPE_VECTOR_SUBPARTS.
20660 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20662 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20663 number of elements from the PARALLEL rather than the mode.
20664 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20665 rather than GET_MODE_BITSIZE.
20666 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20667 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20668 (aarch64_expand_vec_perm_const_1): Handle polynomial
20669 d->perm.length () and d->perm elements.
20670 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20671 Apply to_constant to d->perm elements.
20672 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20673 polynomial CONST_VECTOR_NUNITS.
20674 (aarch64_move_pointer): Take amount as a poly_int64 rather
20676 (aarch64_progress_pointer): Avoid temporary variable.
20677 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20678 the mode attribute instead of GET_MODE.
20680 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20681 Alan Hayward <alan.hayward@arm.com>
20682 David Sherwood <david.sherwood@arm.com>
20684 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20685 x exists before using it.
20686 (aarch64_add_constant_internal): Rename to...
20687 (aarch64_add_offset_1): ...this. Replace regnum with separate
20688 src and dest rtxes. Handle the case in which they're different,
20689 including when the offset is zero. Replace scratchreg with an rtx.
20690 Use 2 additions if there is no spare register into which we can
20691 move a 16-bit constant.
20692 (aarch64_add_constant): Delete.
20693 (aarch64_add_offset): Replace reg with separate src and dest
20694 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20695 Use aarch64_add_offset_1.
20696 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20697 an rtx rather than an int. Take the delta as a poly_int64
20698 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20699 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20700 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20701 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20702 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20703 and aarch64_add_sp.
20704 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20705 aarch64_add_constant.
20707 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20709 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20710 Use scalar_float_mode.
20712 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20714 * config/aarch64/aarch64-simd.md
20715 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20716 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20717 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20718 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20719 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20720 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20721 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20722 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20723 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20724 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20726 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20729 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20730 targ_options->x_arm_arch_string is non NULL.
20732 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20734 * config/aarch64/aarch64.h
20735 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20737 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20740 * expmed.c (emit_store_flag_force): Swap if const op0
20741 and change VOIDmode to mode of op0.
20743 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20745 PR rtl-optimization/83761
20746 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20747 than bytes to mode_for_size.
20749 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20751 PR middle-end/83189
20752 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20753 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20756 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20758 PR middle-end/83575
20759 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20760 when in layout mode.
20761 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20762 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20765 2018-01-10 Michael Collison <michael.collison@arm.com>
20767 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20768 * config/aarch64/aarch64-option-extension.def: Add
20769 AARCH64_OPT_EXTENSION of 'fp16fml'.
20770 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20771 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20772 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20773 * config/aarch64/constraints.md (Ui7): New constraint.
20774 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20775 (VFMLA_SEL_W): Ditto.
20778 (VFMLA16_LOW): New int iterator.
20779 (VFMLA16_HIGH): Ditto.
20780 (UNSPEC_FMLAL): New unspec.
20781 (UNSPEC_FMLSL): Ditto.
20782 (UNSPEC_FMLAL2): Ditto.
20783 (UNSPEC_FMLSL2): Ditto.
20784 (f16mac): New code attribute.
20785 * config/aarch64/aarch64-simd-builtins.def
20786 (aarch64_fmlal_lowv2sf): Ditto.
20787 (aarch64_fmlsl_lowv2sf): Ditto.
20788 (aarch64_fmlalq_lowv4sf): Ditto.
20789 (aarch64_fmlslq_lowv4sf): Ditto.
20790 (aarch64_fmlal_highv2sf): Ditto.
20791 (aarch64_fmlsl_highv2sf): Ditto.
20792 (aarch64_fmlalq_highv4sf): Ditto.
20793 (aarch64_fmlslq_highv4sf): Ditto.
20794 (aarch64_fmlal_lane_lowv2sf): Ditto.
20795 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20796 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20797 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20798 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20799 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20800 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20801 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20802 (aarch64_fmlal_lane_highv2sf): Ditto.
20803 (aarch64_fmlsl_lane_highv2sf): Ditto.
20804 (aarch64_fmlal_laneq_highv2sf): Ditto.
20805 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20806 (aarch64_fmlalq_lane_highv4sf): Ditto.
20807 (aarch64_fmlsl_lane_highv4sf): Ditto.
20808 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20809 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20810 * config/aarch64/aarch64-simd.md:
20811 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20812 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20813 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20814 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20815 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20816 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20817 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20818 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20819 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20820 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20821 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20822 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20823 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20824 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20825 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20826 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20827 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20828 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20829 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20830 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20831 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20832 (vfmlsl_low_u32): Ditto.
20833 (vfmlalq_low_u32): Ditto.
20834 (vfmlslq_low_u32): Ditto.
20835 (vfmlal_high_u32): Ditto.
20836 (vfmlsl_high_u32): Ditto.
20837 (vfmlalq_high_u32): Ditto.
20838 (vfmlslq_high_u32): Ditto.
20839 (vfmlal_lane_low_u32): Ditto.
20840 (vfmlsl_lane_low_u32): Ditto.
20841 (vfmlal_laneq_low_u32): Ditto.
20842 (vfmlsl_laneq_low_u32): Ditto.
20843 (vfmlalq_lane_low_u32): Ditto.
20844 (vfmlslq_lane_low_u32): Ditto.
20845 (vfmlalq_laneq_low_u32): Ditto.
20846 (vfmlslq_laneq_low_u32): Ditto.
20847 (vfmlal_lane_high_u32): Ditto.
20848 (vfmlsl_lane_high_u32): Ditto.
20849 (vfmlal_laneq_high_u32): Ditto.
20850 (vfmlsl_laneq_high_u32): Ditto.
20851 (vfmlalq_lane_high_u32): Ditto.
20852 (vfmlslq_lane_high_u32): Ditto.
20853 (vfmlalq_laneq_high_u32): Ditto.
20854 (vfmlslq_laneq_high_u32): Ditto.
20855 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20856 (AARCH64_FL_FOR_ARCH8_4): New.
20857 (AARCH64_ISA_F16FML): New ISA flag.
20858 (TARGET_F16FML): New feature flag for fp16fml.
20859 (doc/invoke.texi): Document new fp16fml option.
20861 2018-01-10 Michael Collison <michael.collison@arm.com>
20863 * config/aarch64/aarch64-builtins.c:
20864 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20865 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20866 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20867 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20868 (AARCH64_ISA_SHA3): New ISA flag.
20869 (TARGET_SHA3): New feature flag for sha3.
20870 * config/aarch64/iterators.md (sha512_op): New int attribute.
20871 (CRYPTO_SHA512): New int iterator.
20872 (UNSPEC_SHA512H): New unspec.
20873 (UNSPEC_SHA512H2): Ditto.
20874 (UNSPEC_SHA512SU0): Ditto.
20875 (UNSPEC_SHA512SU1): Ditto.
20876 * config/aarch64/aarch64-simd-builtins.def
20877 (aarch64_crypto_sha512hqv2di): New builtin.
20878 (aarch64_crypto_sha512h2qv2di): Ditto.
20879 (aarch64_crypto_sha512su0qv2di): Ditto.
20880 (aarch64_crypto_sha512su1qv2di): Ditto.
20881 (aarch64_eor3qv8hi): Ditto.
20882 (aarch64_rax1qv2di): Ditto.
20883 (aarch64_xarqv2di): Ditto.
20884 (aarch64_bcaxqv8hi): Ditto.
20885 * config/aarch64/aarch64-simd.md:
20886 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20887 (aarch64_crypto_sha512su0qv2di): Ditto.
20888 (aarch64_crypto_sha512su1qv2di): Ditto.
20889 (aarch64_eor3qv8hi): Ditto.
20890 (aarch64_rax1qv2di): Ditto.
20891 (aarch64_xarqv2di): Ditto.
20892 (aarch64_bcaxqv8hi): Ditto.
20893 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20894 (vsha512h2q_u64): Ditto.
20895 (vsha512su0q_u64): Ditto.
20896 (vsha512su1q_u64): Ditto.
20897 (veor3q_u16): Ditto.
20898 (vrax1q_u64): Ditto.
20899 (vxarq_u64): Ditto.
20900 (vbcaxq_u16): Ditto.
20901 * config/arm/types.md (crypto_sha512): New type attribute.
20902 (crypto_sha3): Ditto.
20903 (doc/invoke.texi): Document new sha3 option.
20905 2018-01-10 Michael Collison <michael.collison@arm.com>
20907 * config/aarch64/aarch64-builtins.c:
20908 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20909 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20910 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20911 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20912 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20913 (AARCH64_ISA_SM4): New ISA flag.
20914 (TARGET_SM4): New feature flag for sm4.
20915 * config/aarch64/aarch64-simd-builtins.def
20916 (aarch64_sm3ss1qv4si): Ditto.
20917 (aarch64_sm3tt1aq4si): Ditto.
20918 (aarch64_sm3tt1bq4si): Ditto.
20919 (aarch64_sm3tt2aq4si): Ditto.
20920 (aarch64_sm3tt2bq4si): Ditto.
20921 (aarch64_sm3partw1qv4si): Ditto.
20922 (aarch64_sm3partw2qv4si): Ditto.
20923 (aarch64_sm4eqv4si): Ditto.
20924 (aarch64_sm4ekeyqv4si): Ditto.
20925 * config/aarch64/aarch64-simd.md:
20926 (aarch64_sm3ss1qv4si): Ditto.
20927 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20928 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20929 (aarch64_sm4eqv4si): Ditto.
20930 (aarch64_sm4ekeyqv4si): Ditto.
20931 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20932 (sm3part_op): Ditto.
20933 (CRYPTO_SM3TT): Ditto.
20934 (CRYPTO_SM3PART): Ditto.
20935 (UNSPEC_SM3SS1): New unspec.
20936 (UNSPEC_SM3TT1A): Ditto.
20937 (UNSPEC_SM3TT1B): Ditto.
20938 (UNSPEC_SM3TT2A): Ditto.
20939 (UNSPEC_SM3TT2B): Ditto.
20940 (UNSPEC_SM3PARTW1): Ditto.
20941 (UNSPEC_SM3PARTW2): Ditto.
20942 (UNSPEC_SM4E): Ditto.
20943 (UNSPEC_SM4EKEY): Ditto.
20944 * config/aarch64/constraints.md (Ui2): New constraint.
20945 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20946 * config/arm/types.md (crypto_sm3): New type attribute.
20947 (crypto_sm4): Ditto.
20948 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20949 (vsm3tt1aq_u32): Ditto.
20950 (vsm3tt1bq_u32): Ditto.
20951 (vsm3tt2aq_u32): Ditto.
20952 (vsm3tt2bq_u32): Ditto.
20953 (vsm3partw1q_u32): Ditto.
20954 (vsm3partw2q_u32): Ditto.
20955 (vsm4eq_u32): Ditto.
20956 (vsm4ekeyq_u32): Ditto.
20957 (doc/invoke.texi): Document new sm4 option.
20959 2018-01-10 Michael Collison <michael.collison@arm.com>
20961 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20962 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20963 (AARCH64_FL_FOR_ARCH8_4): New.
20964 (AARCH64_FL_V8_4): New flag.
20965 (doc/invoke.texi): Document new armv8.4-a option.
20967 2018-01-10 Michael Collison <michael.collison@arm.com>
20969 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20970 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
20971 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
20972 * config/aarch64/aarch64-option-extension.def: Add
20973 AARCH64_OPT_EXTENSION of 'sha2'.
20974 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
20975 (crypto): Disable sha2 and aes if crypto disabled.
20976 (crypto): Enable aes and sha2 if enabled.
20977 (simd): Disable sha2 and aes if simd disabled.
20978 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
20980 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
20981 (TARGET_SHA2): New feature flag for sha2.
20982 (TARGET_AES): New feature flag for aes.
20983 * config/aarch64/aarch64-simd.md:
20984 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
20985 conditional on TARGET_AES.
20986 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20987 (aarch64_crypto_sha1hsi): Make pattern conditional
20989 (aarch64_crypto_sha1hv4si): Ditto.
20990 (aarch64_be_crypto_sha1hv4si): Ditto.
20991 (aarch64_crypto_sha1su1v4si): Ditto.
20992 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20993 (aarch64_crypto_sha1su0v4si): Ditto.
20994 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20995 (aarch64_crypto_sha256su0v4si): Ditto.
20996 (aarch64_crypto_sha256su1v4si): Ditto.
20997 (doc/invoke.texi): Document new aes and sha2 options.
20999 2018-01-10 Martin Sebor <msebor@redhat.com>
21001 PR tree-optimization/83781
21002 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21005 2018-01-11 Martin Sebor <msebor@gmail.com>
21006 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21008 PR tree-optimization/83501
21009 PR tree-optimization/81703
21011 * tree-ssa-strlen.c (get_string_cst): Rename...
21012 (get_string_len): ...to this. Handle global constants.
21013 (handle_char_store): Adjust.
21015 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21016 Jim Wilson <jimw@sifive.com>
21018 * config/riscv/riscv-protos.h (riscv_output_return): New.
21019 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21020 (riscv_attribute_table, riscv_output_return),
21021 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21022 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21023 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21025 (riscv_expand_prologue): Add early return for naked function.
21026 (riscv_expand_epilogue): Likewise.
21027 (riscv_function_ok_for_sibcall): Return false for naked function.
21028 (riscv_set_current_function): New.
21029 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21030 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21031 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21032 * doc/extend.texi (RISC-V Function Attributes): New.
21034 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21036 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21037 check for 128-bit long double before checking TCmode.
21038 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21039 128-bit long doubles before checking TFmode or TCmode.
21040 (FLOAT128_IBM_P): Likewise.
21042 2018-01-10 Martin Sebor <msebor@redhat.com>
21044 PR tree-optimization/83671
21045 * builtins.c (c_strlen): Unconditionally return zero for the empty
21047 Use -Warray-bounds for warnings.
21048 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21049 for non-constant array indices with COMPONENT_REF, arrays of
21050 arrays, and pointers to arrays.
21051 (gimple_fold_builtin_strlen): Determine and set length range for
21052 non-constant character arrays.
21054 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21056 PR middle-end/81897
21057 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21060 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21062 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21064 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21067 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21068 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21069 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21070 indexed_or_indirect_operand predicate.
21071 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21072 (*vsx_le_perm_load_v8hi): Likewise.
21073 (*vsx_le_perm_load_v16qi): Likewise.
21074 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21075 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21076 (*vsx_le_perm_store_v8hi): Likewise.
21077 (*vsx_le_perm_store_v16qi): Likewise.
21078 (eight unnamed splitters): Likewise.
21080 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21082 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21083 * config/rs6000/emmintrin.h: Likewise.
21084 * config/rs6000/mmintrin.h: Likewise.
21085 * config/rs6000/xmmintrin.h: Likewise.
21087 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21090 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21092 * tree.c (tree_nop_conversion): Return true for location wrapper
21094 (maybe_wrap_with_location): New function.
21095 (selftest::check_strip_nops): New function.
21096 (selftest::test_location_wrappers): New function.
21097 (selftest::tree_c_tests): Call it.
21098 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21099 (maybe_wrap_with_location): New decl.
21100 (EXPR_LOCATION_WRAPPER_P): New macro.
21101 (location_wrapper_p): New inline function.
21102 (tree_strip_any_location_wrapper): New inline function.
21104 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21107 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21108 stack_realign_offset for the largest alignment of stack slot
21110 (ix86_find_max_used_stack_alignment): New function.
21111 (ix86_finalize_stack_frame_flags): Use it. Set
21112 max_used_stack_alignment if we don't realign stack.
21113 * config/i386/i386.h (machine_function): Add
21114 max_used_stack_alignment.
21116 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21118 * config/arm/arm.opt (-mbranch-cost): New option.
21119 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21122 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21125 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21126 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21128 2018-01-10 Richard Biener <rguenther@suse.de>
21131 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21132 early out so it also covers the case where we have a non-NULL
21135 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21137 PR tree-optimization/83753
21138 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21139 for non-strided grouped accesses if the number of elements is 1.
21141 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21144 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21145 * i386.h (TARGET_USE_GATHER): Define.
21146 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21148 2018-01-10 Martin Liska <mliska@suse.cz>
21151 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21152 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21154 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21155 CLEANUP_NO_PARTITIONING is not set.
21157 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21159 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21160 for vectors, as a partial revert of r254296.
21161 * rtl.h (const_vec_p): Delete.
21162 (const_vec_duplicate_p): Don't test for vector CONSTs.
21163 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21164 * expmed.c (make_tree): Likewise.
21167 * common.md (E, F): Use CONSTANT_P instead of checking for
21169 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21170 checking for CONST_VECTOR.
21172 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21174 PR middle-end/83575
21175 * predict.c (force_edge_cold): Handle in more sane way edges
21176 with no prediction.
21178 2018-01-09 Carl Love <cel@us.ibm.com>
21180 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21182 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21183 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21184 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21185 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21186 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21187 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21188 * config/rs6000/rs6000-protos.h: Add extern defition for
21189 rs6000_generate_float2_double_code.
21190 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21192 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21193 (float2_v2df): Add define_expand.
21195 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21198 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21199 op_mode in the force_to_mode call.
21201 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21203 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21204 instead of checking each element individually.
21205 (aarch64_evpc_uzp): Likewise.
21206 (aarch64_evpc_zip): Likewise.
21207 (aarch64_evpc_ext): Likewise.
21208 (aarch64_evpc_rev): Likewise.
21209 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21210 instead of checking each element individually. Return true without
21212 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21213 whether all selected elements come from the same input, instead of
21214 checking each element individually. Remove calls to gen_rtx_REG,
21215 start_sequence and end_sequence and instead assert that no rtl is
21218 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21220 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21221 order of HIGH and CONST checks.
21223 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21225 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21226 if the destination isn't an SSA_NAME.
21228 2018-01-09 Richard Biener <rguenther@suse.de>
21230 PR tree-optimization/83668
21231 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21233 (canonicalize_loop_form): ... here, renamed from ...
21234 (canonicalize_loop_closed_ssa_form): ... this and amended to
21235 swap successor edges for loop exit blocks to make us use
21236 the RPO order we need for initial schedule generation.
21238 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21240 PR tree-optimization/64811
21241 * match.pd: When optimizing comparisons with Inf, avoid
21242 introducing or losing exceptions from comparisons with NaN.
21244 2018-01-09 Martin Liska <mliska@suse.cz>
21247 * asan.c (shadow_mem_size): Add gcc_assert.
21249 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21251 Don't save registers in main().
21254 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21255 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21256 * config/avr/avr.c (avr_set_current_function): Don't error if
21257 naked, OS_task or OS_main are specified at the same time.
21258 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21260 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21262 * common/config/avr/avr-common.c (avr_option_optimization_table):
21263 Switch on -mmain-is-OS_task for optimizing compilations.
21265 2018-01-09 Richard Biener <rguenther@suse.de>
21267 PR tree-optimization/83572
21268 * graphite.c: Include cfganal.h.
21269 (graphite_transform_loops): Connect infinite loops to exit
21270 and remove fake edges at the end.
21272 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21274 * ipa-inline.c (edge_badness): Revert accidental checkin.
21276 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21279 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21280 symbols; not inline clones.
21282 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21285 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21286 hard registers. Formatting fixes.
21288 PR preprocessor/83722
21289 * gcc.c (try_generate_repro): Pass
21290 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21291 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21294 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21295 Kito Cheng <kito.cheng@gmail.com>
21297 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21298 (riscv_leaf_function_p): Delete.
21299 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21301 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21303 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21305 (do_ifelse): New function.
21306 (do_isel): New function.
21307 (do_sub3): New function.
21308 (do_add3): New function.
21309 (do_load_mask_compare): New function.
21310 (do_overlap_load_compare): New function.
21311 (expand_compare_loop): New function.
21312 (expand_block_compare): Call expand_compare_loop() when appropriate.
21313 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21314 option description.
21315 (-mblock-compare-inline-loop-limit): New option.
21317 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21320 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21321 Reverse order of second and third operands in first alternative.
21322 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21323 of first and second elements in UNSPEC_VPERMR vector.
21324 (altivec_expand_vec_perm_le): Likewise.
21326 2018-01-08 Jeff Law <law@redhat.com>
21328 PR rtl-optimizatin/81308
21329 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21330 (process_switch): If group_case_labels makes a change, then set
21332 (pass_convert_switch::execute): If a switch is converted, then
21333 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21335 PR rtl-optimization/81308
21336 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21339 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21341 PR target/83663 - Revert r255946
21342 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21343 generation for cases where splatting a value is not useful.
21344 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21345 across a vec_duplicate and a paradoxical subreg forming a vector
21346 mode to a vec_concat.
21348 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21350 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21351 -march=armv8.3-a variants.
21352 * config/arm/t-multilib: Likewise.
21353 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21355 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21357 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21359 (cceq_ior_compare_complement): Give it a name so I can use it, and
21360 change boolean_or_operator predicate to boolean_operator so it can
21361 be used to generate a crand.
21362 (eqne): New code iterator.
21363 (bd/bd_neg): New code_attrs.
21364 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21365 a single define_insn.
21366 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21367 decrement (bdnzt/bdnzf/bdzt/bdzf).
21368 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21369 with the new names of the branch decrement patterns, and added the
21370 names of the branch decrement conditional patterns.
21372 2018-01-08 Richard Biener <rguenther@suse.de>
21374 PR tree-optimization/83563
21375 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21378 2018-01-08 Richard Biener <rguenther@suse.de>
21380 PR middle-end/83713
21381 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21383 2018-01-08 Richard Biener <rguenther@suse.de>
21385 PR tree-optimization/83685
21386 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21387 references to abnormals.
21389 2018-01-08 Richard Biener <rguenther@suse.de>
21392 * dwarf2out.c (output_indirect_strings): Handle empty
21393 skeleton_debug_str_hash.
21394 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21396 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21398 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21399 (emit_store_direct): Likewise.
21400 (arc_trampoline_adjust_address): Likewise.
21401 (arc_asm_trampoline_template): New function.
21402 (arc_initialize_trampoline): Use asm_trampoline_template.
21403 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21404 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21405 * config/arc/arc.md (flush_icache): Delete pattern.
21407 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21409 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21410 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21413 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21416 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21417 by not USED_FOR_TARGET.
21418 (make_pass_resolve_sw_modes): Likewise.
21420 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21422 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21425 2018-01-08 Richard Biener <rguenther@suse.de>
21427 PR middle-end/83580
21428 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21430 2018-01-08 Richard Biener <rguenther@suse.de>
21432 PR middle-end/83517
21433 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21435 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21437 PR middle-end/81897
21438 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21439 basic blocks with a small number of successors.
21440 (convert_control_dep_chain_into_preds): Improve handling of
21442 (dump_predicates): Split apart into...
21443 (dump_pred_chain): ...here...
21444 (dump_pred_info): ...and here.
21445 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21446 (can_chain_union_be_invalidated_p): Improve check for invalidation
21448 (uninit_uses_cannot_happen): Avoid unnecessary if
21449 convert_control_dep_chain_into_preds yielded nothing.
21451 2018-01-06 Martin Sebor <msebor@redhat.com>
21453 PR tree-optimization/83640
21454 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21455 subtracting negative offset from size.
21456 (builtin_access::overlap): Adjust offset bounds of the access to fall
21457 within the size of the object if possible.
21459 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21461 PR rtl-optimization/83699
21462 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21463 extract_bit_field_as_subreg to cases in which the extracted
21464 value is also a vector.
21466 * lra-constraints.c (process_alt_operands): Test for the equivalence
21467 substitutions when detecting a possible reload cycle.
21469 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21472 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21473 by default if flag_selective_schedling{,2}. Formatting fixes.
21475 PR rtl-optimization/83682
21476 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21477 if it has non-VECTOR_MODE element mode.
21478 (vec_duplicate_p): Likewise.
21480 PR middle-end/83694
21481 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21482 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21484 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21487 * config/i386/i386-builtin.def
21488 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21489 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21490 Require also OPTION_MASK_ISA_AVX512F in addition to
21491 OPTION_MASK_ISA_GFNI.
21492 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21493 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21494 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21495 to OPTION_MASK_ISA_GFNI.
21496 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21497 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21498 OPTION_MASK_ISA_AVX512BW.
21499 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21500 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21501 addition to OPTION_MASK_ISA_GFNI.
21502 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21503 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21504 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21505 to OPTION_MASK_ISA_GFNI.
21506 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21507 a requirement for all ISAs rather than any of them with a few
21509 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21511 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21512 bitmasks to be enabled with 3 exceptions, instead of requiring any
21513 enabled ISA with lots of exceptions.
21514 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21515 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21516 Change avx512bw in isa attribute to avx512f.
21517 * config/i386/sgxintrin.h: Add license boilerplate.
21518 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21519 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21520 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21521 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21523 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21524 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21525 temporarily sse2 rather than sse if not enabled already.
21528 * config/i386/sse.md (VI248_VLBW): Rename to ...
21529 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21530 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21531 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21532 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21533 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21534 mode iterator instead of VI248_VLBW.
21536 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21538 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21539 (record_modified): Skip clobbers; add debug output.
21540 (param_change_prob): Use sreal frequencies.
21542 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21544 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21545 punt for user-aligned variables.
21547 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21549 * tree-chrec.c (chrec_contains_symbols): Return true for
21552 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21555 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21556 of (x|y) == x for BICS pattern.
21558 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21560 PR tree-optimization/83605
21561 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21562 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21565 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21567 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21568 * config/epiphany/rtems.h: New file.
21570 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21571 Uros Bizjak <ubizjak@gmail.com>
21574 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21575 QIreg_operand instead of register_operand predicate.
21576 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21577 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21578 comments instead of -fmitigate[-_]rop.
21580 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21583 * cgraphunit.c (symbol_table::compile): Switch to text_section
21584 before calling assembly_start debug hook.
21585 * run-rtl-passes.c (run_rtl_passes): Likewise.
21588 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21590 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21591 range_int_cst_p rather than !symbolic_range_p before calling
21592 extract_range_from_multiplicative_op_1.
21594 2018-01-04 Jeff Law <law@redhat.com>
21596 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21597 redundant test in assertion.
21599 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21601 * doc/rtl.texi: Document machine_mode wrapper classes.
21603 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21605 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21606 using tree_to_uhwi.
21608 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21610 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21611 the VEC_PERM_EXPR fold to fail.
21613 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21616 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21617 to switched_sections.
21619 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21622 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21623 test for d.testing.
21625 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21628 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21629 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21631 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21634 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21635 is BLKmode and bitpos not zero or mode change is needed.
21637 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21640 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21643 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21646 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21647 instead of MULT rtx. Update all corresponding splitters.
21648 (*saddl_se): Ditto.
21649 (*ssub<modesuffix>): Ditto.
21650 (*ssubl_se): Ditto.
21651 (*cmp_sadd_di): Update split patterns.
21652 (*cmp_sadd_si): Ditto.
21653 (*cmp_sadd_sidi): Ditto.
21654 (*cmp_ssub_di): Ditto.
21655 (*cmp_ssub_si): Ditto.
21656 (*cmp_ssub_sidi): Ditto.
21657 * config/alpha/predicates.md (const23_operand): New predicate.
21658 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21659 Look for ASHIFT, not MULT inner operand.
21660 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21662 2018-01-04 Martin Liska <mliska@suse.cz>
21664 PR gcov-profile/83669
21665 * gcov.c (output_intermediate_file): Add version to intermediate
21667 * doc/gcov.texi: Document new field 'version' in intermediate
21668 file format. Fix location of '-k' option of gcov command.
21670 2018-01-04 Martin Liska <mliska@suse.cz>
21673 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21675 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21677 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21679 2018-01-03 Martin Sebor <msebor@redhat.com>
21681 PR tree-optimization/83655
21682 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21683 checking calls with invalid arguments.
21685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21687 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21688 (vectorizable_mask_load_store): Delete.
21689 (vectorizable_call): Return false for masked loads and stores.
21690 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21691 instead of gimple_assign_rhs1.
21692 (vectorizable_load): Handle IFN_MASK_LOAD.
21693 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21697 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21699 (vectorizable_mask_load_store): ...here.
21700 (vectorizable_load): ...and here.
21702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21704 * tree-vect-stmts.c (vect_build_all_ones_mask)
21705 (vect_build_zero_merge_argument): New functions, split out from...
21706 (vectorizable_load): ...here.
21708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21710 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21712 (vectorizable_mask_load_store): ...here.
21713 (vectorizable_store): ...and here.
21715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21717 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21719 (vectorizable_mask_load_store): ...here.
21721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21723 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21724 (vect_model_store_cost): Take a vec_load_store_type instead of a
21726 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21727 (vect_model_store_cost): Take a vec_load_store_type instead of a
21729 (vectorizable_mask_load_store): Update accordingly.
21730 (vectorizable_store): Likewise.
21731 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21735 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21736 IFN_MASK_LOAD calls here rather than...
21737 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21740 Alan Hayward <alan.hayward@arm.com>
21741 David Sherwood <david.sherwood@arm.com>
21743 * expmed.c (extract_bit_field_1): For vector extracts,
21744 fall back to extract_bit_field_as_subreg if vec_extract
21747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21748 Alan Hayward <alan.hayward@arm.com>
21749 David Sherwood <david.sherwood@arm.com>
21751 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21752 they are variable or constant sized.
21753 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21754 slots for constant-sized data.
21756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21757 Alan Hayward <alan.hayward@arm.com>
21758 David Sherwood <david.sherwood@arm.com>
21760 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21761 handling COND_EXPRs with boolean comparisons, try to find a better
21762 basis for the mask type than the boolean itself.
21764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21766 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21767 is calculated and how it can be overridden.
21768 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21769 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21771 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21775 Alan Hayward <alan.hayward@arm.com>
21776 David Sherwood <david.sherwood@arm.com>
21778 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21779 Remove the mode argument.
21780 (aarch64_simd_valid_immediate): Remove the mode and inverse
21782 * config/aarch64/iterators.md (bitsize): New iterator.
21783 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21784 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21785 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21786 aarch64_simd_valid_immediate.
21787 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21788 (aarch64_reg_or_bic_imm): Likewise.
21789 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21790 with an insn_type enum and msl with a modifier_type enum.
21791 Replace element_width with a scalar_mode. Change the shift
21792 to unsigned int. Add constructors for scalar_float_mode and
21793 scalar_int_mode elements.
21794 (aarch64_vect_float_const_representable_p): Delete.
21795 (aarch64_can_const_movi_rtx_p)
21796 (aarch64_simd_scalar_immediate_valid_for_move)
21797 (aarch64_simd_make_constant): Update call to
21798 aarch64_simd_valid_immediate.
21799 (aarch64_advsimd_valid_immediate_hs): New function.
21800 (aarch64_advsimd_valid_immediate): Likewise.
21801 (aarch64_simd_valid_immediate): Remove mode and inverse
21802 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21803 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21804 and aarch64_float_const_representable_p on the result.
21805 (aarch64_output_simd_mov_immediate): Remove mode argument.
21806 Update call to aarch64_simd_valid_immediate and use of
21807 simd_immediate_info.
21808 (aarch64_output_scalar_simd_mov_immediate): Update call
21811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21812 Alan Hayward <alan.hayward@arm.com>
21813 David Sherwood <david.sherwood@arm.com>
21815 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21816 (mode_nunits): Likewise CONST_MODE_NUNITS.
21817 * machmode.def (ADJUST_NUNITS): Document.
21818 * genmodes.c (mode_data::need_nunits_adj): New field.
21819 (blank_mode): Update accordingly.
21820 (adj_nunits): New variable.
21821 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21823 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21824 listed in adj_nunits.
21825 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21826 listed in adj_nunits. Don't emit case statements for such modes.
21827 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21828 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21829 nothing if adj_nunits is nonnull.
21830 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21831 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21832 (emit_mode_fbit): Update use of print_maybe_const_decl.
21833 (emit_move_size): Likewise. Treat the array as non-const
21835 (emit_mode_adjustments): Handle adj_nunits.
21837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21839 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21840 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21841 (VECTOR_MODES): Use it.
21842 (make_vector_modes): Take the prefix as an argument.
21844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21845 Alan Hayward <alan.hayward@arm.com>
21846 David Sherwood <david.sherwood@arm.com>
21848 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21849 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21850 for MODE_VECTOR_BOOL.
21851 * machmode.def (VECTOR_BOOL_MODE): Document.
21852 * genmodes.c (VECTOR_BOOL_MODE): New macro.
21853 (make_vector_bool_mode): New function.
21854 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21856 * lto-streamer-in.c (lto_input_mode_table): Likewise.
21857 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21859 * stor-layout.c (int_mode_for_mode): Likewise.
21860 * tree.c (build_vector_type_for_mode): Likewise.
21861 * varasm.c (output_constant_pool_2): Likewise.
21862 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21863 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
21864 for MODE_VECTOR_BOOL.
21865 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21866 of mode class checks.
21867 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21868 instead of a list of mode class checks.
21869 (expand_vector_scalar_condition): Likewise.
21870 (type_for_widest_vector_mode): Handle BImode as an inner mode.
21872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21873 Alan Hayward <alan.hayward@arm.com>
21874 David Sherwood <david.sherwood@arm.com>
21876 * machmode.h (mode_size): Change from unsigned short to
21878 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21879 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21880 or if measurement_type is not polynomial.
21881 (fixed_size_mode::includes_p): Check for constant-sized modes.
21882 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21883 return a poly_uint16 rather than an unsigned short.
21884 (emit_mode_size): Change the type of mode_size from unsigned short
21885 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
21886 (emit_mode_adjustments): Cope with polynomial vector sizes.
21887 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21889 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21891 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21892 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21893 * caller-save.c (setup_save_areas): Likewise.
21894 (replace_reg_with_saved_mem): Likewise.
21895 * calls.c (emit_library_call_value_1): Likewise.
21896 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21897 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21898 (gen_lowpart_for_combine): Likewise.
21899 * convert.c (convert_to_integer_1): Likewise.
21900 * cse.c (equiv_constant, cse_insn): Likewise.
21901 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21902 (cselib_subst_to_values): Likewise.
21903 * dce.c (word_dce_process_block): Likewise.
21904 * df-problems.c (df_word_lr_mark_ref): Likewise.
21905 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21906 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21907 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21908 (rtl_for_decl_location): Likewise.
21909 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21910 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21911 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21912 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21913 (expand_expr_real_1): Likewise.
21914 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21915 (pad_below): Likewise.
21916 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21917 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21918 * ira.c (get_subreg_tracking_sizes): Likewise.
21919 * ira-build.c (ira_create_allocno_objects): Likewise.
21920 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21921 (ira_sort_regnos_for_alter_reg): Likewise.
21922 * ira-costs.c (record_operand_costs): Likewise.
21923 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21924 (resolve_simple_move): Likewise.
21925 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21926 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21927 (lra_constraints): Likewise.
21928 (CONST_POOL_OK_P): Reject variable-sized modes.
21929 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21930 (add_pseudo_to_slot, lra_spill): Likewise.
21931 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21932 * optabs-query.c (get_best_extraction_insn): Likewise.
21933 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21934 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21935 (expand_mult_highpart, valid_multiword_target_p): Likewise.
21936 * recog.c (offsettable_address_addr_space_p): Likewise.
21937 * regcprop.c (maybe_mode_change): Likewise.
21938 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21939 * regrename.c (build_def_use): Likewise.
21940 * regstat.c (dump_reg_info): Likewise.
21941 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21942 (find_reloads, find_reloads_subreg_address): Likewise.
21943 * reload1.c (eliminate_regs_1): Likewise.
21944 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21945 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21946 (simplify_binary_operation_1, simplify_subreg): Likewise.
21947 * targhooks.c (default_function_arg_padding): Likewise.
21948 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21949 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21950 (verify_gimple_assign_ternary): Likewise.
21951 * tree-inline.c (estimate_move_cost): Likewise.
21952 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21953 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21954 (get_address_cost_ainc): Likewise.
21955 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21956 (vect_supportable_dr_alignment): Likewise.
21957 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21958 (vectorizable_reduction): Likewise.
21959 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21960 (vectorizable_operation, vectorizable_load): Likewise.
21961 * tree.c (build_same_sized_truth_vector_type): Likewise.
21962 * valtrack.c (cleanup_auto_inc_dec): Likewise.
21963 * var-tracking.c (emit_note_insn_var_location): Likewise.
21964 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21965 (ADDR_VEC_ALIGN): Likewise.
21967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21968 Alan Hayward <alan.hayward@arm.com>
21969 David Sherwood <david.sherwood@arm.com>
21971 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
21973 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21974 or if measurement_type is polynomial.
21975 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
21976 * combine.c (make_extraction): Likewise.
21977 * dse.c (find_shift_sequence): Likewise.
21978 * dwarf2out.c (mem_loc_descriptor): Likewise.
21979 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
21980 (extract_bit_field, extract_low_bits): Likewise.
21981 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
21982 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
21983 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
21984 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
21985 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21986 * reload.c (find_reloads): Likewise.
21987 * reload1.c (alter_reg): Likewise.
21988 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21989 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21990 * tree-if-conv.c (predicate_mem_writes): Likewise.
21991 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21992 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21993 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21994 * valtrack.c (dead_debug_insert_temp): Likewise.
21995 * varasm.c (mergeable_constant_section): Likewise.
21996 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21999 Alan Hayward <alan.hayward@arm.com>
22000 David Sherwood <david.sherwood@arm.com>
22002 * expr.c (expand_assignment): Cope with polynomial mode sizes
22003 when assigning to a CONCAT.
22005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22006 Alan Hayward <alan.hayward@arm.com>
22007 David Sherwood <david.sherwood@arm.com>
22009 * machmode.h (mode_precision): Change from unsigned short to
22011 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22013 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22014 or if measurement_type is not polynomial.
22015 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22016 in which the mode is already known to be a scalar_int_mode.
22017 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22018 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22020 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22021 for GET_MODE_PRECISION.
22022 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22023 for GET_MODE_PRECISION.
22024 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22026 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22027 (expand_field_assignment, make_extraction): Likewise.
22028 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22029 (get_last_value): Likewise.
22030 * convert.c (convert_to_integer_1): Likewise.
22031 * cse.c (cse_insn): Likewise.
22032 * expr.c (expand_expr_real_1): Likewise.
22033 * lra-constraints.c (simplify_operand_subreg): Likewise.
22034 * optabs-query.c (can_atomic_load_p): Likewise.
22035 * optabs.c (expand_atomic_load): Likewise.
22036 (expand_atomic_store): Likewise.
22037 * ree.c (combine_reaching_defs): Likewise.
22038 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22039 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22040 * tree.h (type_has_mode_precision_p): Likewise.
22041 * ubsan.c (instrument_si_overflow): Likewise.
22043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22044 Alan Hayward <alan.hayward@arm.com>
22045 David Sherwood <david.sherwood@arm.com>
22047 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22048 polynomial numbers of units.
22049 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22050 (valid_vector_subparts_p): New function.
22051 (build_vector_type): Remove temporary shim and take the number
22052 of units as a poly_uint64 rather than an int.
22053 (build_opaque_vector_type): Take the number of units as a
22054 poly_uint64 rather than an int.
22055 * tree.c (build_vector_from_ctor): Handle polynomial
22056 TYPE_VECTOR_SUBPARTS.
22057 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22058 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22059 (build_vector_from_val): If the number of units is variable,
22060 use build_vec_duplicate_cst for constant operands and
22061 VEC_DUPLICATE_EXPR otherwise.
22062 (make_vector_type): Remove temporary is_constant ().
22063 (build_vector_type, build_opaque_vector_type): Take the number of
22064 units as a poly_uint64 rather than an int.
22065 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22067 * cfgexpand.c (expand_debug_expr): Likewise.
22068 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22069 (store_constructor, expand_expr_real_1): Likewise.
22070 (const_scalar_mask_from_tree): Likewise.
22071 * fold-const-call.c (fold_const_reduction): Likewise.
22072 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22073 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22074 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22075 (fold_relational_const): Likewise.
22076 (native_interpret_vector): Likewise. Change the size from an
22077 int to an unsigned int.
22078 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22079 TYPE_VECTOR_SUBPARTS.
22080 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22081 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22082 duplicating a non-constant operand into a variable-length vector.
22083 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22084 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22085 * ipa-icf.c (sem_variable::equals): Likewise.
22086 * match.pd: Likewise.
22087 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22088 * print-tree.c (print_node): Likewise.
22089 * stor-layout.c (layout_type): Likewise.
22090 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22091 * tree-cfg.c (verify_gimple_comparison): Likewise.
22092 (verify_gimple_assign_binary): Likewise.
22093 (verify_gimple_assign_ternary): Likewise.
22094 (verify_gimple_assign_single): Likewise.
22095 * tree-pretty-print.c (dump_generic_node): Likewise.
22096 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22097 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22098 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22099 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22100 (vect_shift_permute_load_chain): Likewise.
22101 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22102 (expand_vector_condition, optimize_vector_constructor): Likewise.
22103 (lower_vec_perm, get_compute_type): Likewise.
22104 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22105 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22106 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22107 (vect_recog_mask_conversion_pattern): Likewise.
22108 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22109 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22110 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22111 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22112 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22113 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22114 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22115 (supportable_widening_operation): Likewise.
22116 (supportable_narrowing_operation): Likewise.
22117 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22119 * varasm.c (output_constant): Likewise.
22121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22122 Alan Hayward <alan.hayward@arm.com>
22123 David Sherwood <david.sherwood@arm.com>
22125 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22126 so that both the length == 3 and length != 3 cases set up their
22127 own permute vectors. Add comments explaining why we know the
22128 number of elements is constant.
22129 (vect_permute_load_chain): Likewise.
22131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22132 Alan Hayward <alan.hayward@arm.com>
22133 David Sherwood <david.sherwood@arm.com>
22135 * machmode.h (mode_nunits): Change from unsigned char to
22137 (ONLY_FIXED_SIZE_MODES): New macro.
22138 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22139 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22140 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22142 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22143 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22144 or if measurement_type is not polynomial.
22145 * genmodes.c (ZERO_COEFFS): New macro.
22146 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22148 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22149 Use ZERO_COEFFS when emitting initializers.
22150 * data-streamer.h (bp_pack_poly_value): New function.
22151 (bp_unpack_poly_value): Likewise.
22152 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22153 for GET_MODE_NUNITS.
22154 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22155 for GET_MODE_NUNITS.
22156 * tree.c (make_vector_type): Remove temporary shim and make
22157 the real function take the number of units as a poly_uint64
22158 rather than an int.
22159 (build_vector_type_for_mode): Handle polynomial nunits.
22160 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22161 * emit-rtl.c (const_vec_series_p_1): Likewise.
22162 (gen_rtx_CONST_VECTOR): Likewise.
22163 * fold-const.c (test_vec_duplicate_folding): Likewise.
22164 * genrecog.c (validate_pattern): Likewise.
22165 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22166 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22167 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22168 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22169 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22170 * rtlanal.c (subreg_get_info): Likewise.
22171 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22172 (vect_grouped_load_supported): Likewise.
22173 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22174 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22175 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22176 (simplify_const_unary_operation, simplify_binary_operation_1)
22177 (simplify_const_binary_operation, simplify_ternary_operation)
22178 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22179 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22180 instead of CONST_VECTOR_NUNITS.
22181 * varasm.c (output_constant_pool_2): Likewise.
22182 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22183 explicit-encoded elements in the XVEC for variable-length vectors.
22185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22187 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22190 Alan Hayward <alan.hayward@arm.com>
22191 David Sherwood <david.sherwood@arm.com>
22193 * coretypes.h (fixed_size_mode): Declare.
22194 (fixed_size_mode_pod): New typedef.
22195 * builtins.h (target_builtins::x_apply_args_mode)
22196 (target_builtins::x_apply_result_mode): Change type to
22197 fixed_size_mode_pod.
22198 * builtins.c (apply_args_size, apply_result_size, result_vector)
22199 (expand_builtin_apply_args_1, expand_builtin_apply)
22200 (expand_builtin_return): Update accordingly.
22202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22204 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22205 * cselib.c (cselib_hash_rtx): Likewise.
22206 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22207 CONST_VECTOR encoding.
22209 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22210 Jeff Law <law@redhat.com>
22213 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22214 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22215 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22216 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22219 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22220 explicitly probe *sp in a noreturn function if there were any callee
22221 register saves or frame pointer is needed.
22223 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22226 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22227 BLKmode for ternary, binary or unary expressions.
22230 * var-tracking.c (delete_vta_debug_insn): New inline function.
22231 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22232 insns from get_insns () to NULL instead of each bb separately.
22233 Use delete_vta_debug_insn. No longer static.
22234 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22235 delete_vta_debug_insns callers.
22236 * rtl.h (delete_vta_debug_insns): Declare.
22237 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22238 instead of variable_tracking_main.
22240 2018-01-03 Martin Sebor <msebor@redhat.com>
22242 PR tree-optimization/83603
22243 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22244 arguments past the endof the argument list in functions declared
22245 without a prototype.
22246 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22247 Avoid checking when arguments are null.
22249 2018-01-03 Martin Sebor <msebor@redhat.com>
22252 * doc/extend.texi (attribute const): Fix a typo.
22253 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22254 issuing -Wsuggest-attribute for void functions.
22256 2018-01-03 Martin Sebor <msebor@redhat.com>
22258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22259 offset_int::from instead of wide_int::to_shwi.
22260 (maybe_diag_overlap): Remove assertion.
22261 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22262 * gimple-ssa-sprintf.c (format_directive): Same.
22263 (parse_directive): Same.
22264 (sprintf_dom_walker::compute_format_length): Same.
22265 (try_substitute_return_value): Same.
22267 2018-01-03 Jeff Law <law@redhat.com>
22269 PR middle-end/83654
22270 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22271 non-constant residual for zero at runtime and avoid probing in
22272 that case. Reorganize code for trailing problem to mirror handling
22275 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22277 PR tree-optimization/83501
22278 * tree-ssa-strlen.c (get_string_cst): New.
22279 (handle_char_store): Call get_string_cst.
22281 2018-01-03 Martin Liska <mliska@suse.cz>
22283 PR tree-optimization/83593
22284 * tree-ssa-strlen.c: Include tree-cfg.h.
22285 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22286 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22287 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22289 (strlen_dom_walker::before_dom_children): Call
22290 gimple_purge_dead_eh_edges. Dump tranformation with details
22292 (strlen_dom_walker::before_dom_children): Update call by adding
22293 new argument cleanup_eh.
22294 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22296 2018-01-03 Martin Liska <mliska@suse.cz>
22299 * cif-code.def (VARIADIC_THUNK): New enum value.
22300 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22303 2018-01-03 Jan Beulich <jbeulich@suse.com>
22305 * sse.md (mov<mode>_internal): Tighten condition for when to use
22306 vmovdqu<ssescalarsize> for TI and OI modes.
22308 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22310 Update copyright years.
22312 2018-01-03 Martin Liska <mliska@suse.cz>
22315 * ipa-visibility.c (function_and_variable_visibility): Skip
22316 functions with noipa attribure.
22318 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22320 * gcc.c (process_command): Update copyright notice dates.
22321 * gcov-dump.c (print_version): Ditto.
22322 * gcov.c (print_version): Ditto.
22323 * gcov-tool.c (print_version): Ditto.
22324 * gengtype.c (create_file): Ditto.
22325 * doc/cpp.texi: Bump @copying's copyright year.
22326 * doc/cppinternals.texi: Ditto.
22327 * doc/gcc.texi: Ditto.
22328 * doc/gccint.texi: Ditto.
22329 * doc/gcov.texi: Ditto.
22330 * doc/install.texi: Ditto.
22331 * doc/invoke.texi: Ditto.
22333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22335 * vector-builder.h (vector_builder::m_full_nelts): Change from
22336 unsigned int to poly_uint64.
22337 (vector_builder::full_nelts): Update prototype accordingly.
22338 (vector_builder::new_vector): Likewise.
22339 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22340 (vector_builder::operator ==): Likewise.
22341 (vector_builder::finalize): Likewise.
22342 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22343 Take the number of elements as a poly_uint64 rather than an
22345 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22346 from unsigned int to poly_uint64.
22347 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22348 (vec_perm_indices::new_vector): Likewise.
22349 (vec_perm_indices::length): Likewise.
22350 (vec_perm_indices::nelts_per_input): Likewise.
22351 (vec_perm_indices::input_nelts): Likewise.
22352 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22353 number of elements per input as a poly_uint64 rather than an
22354 unsigned int. Use the original encoding for variable-length
22355 vectors, rather than clamping each individual element.
22356 For the second and subsequent elements in each pattern,
22357 clamp the step and base before clamping their sum.
22358 (vec_perm_indices::series_p): Handle polynomial element counts.
22359 (vec_perm_indices::all_in_range_p): Likewise.
22360 (vec_perm_indices_to_tree): Likewise.
22361 (vec_perm_indices_to_rtx): Likewise.
22362 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22363 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22364 (tree_vector_builder::new_binary_operation): Handle polynomial
22365 element counts. Return false if we need to know the number
22366 of elements at compile time.
22367 * fold-const.c (fold_vec_perm): Punt if the number of elements
22368 isn't known at compile time.
22370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22372 * vec-perm-indices.h (vec_perm_builder): Change element type
22373 from HOST_WIDE_INT to poly_int64.
22374 (vec_perm_indices::element_type): Update accordingly.
22375 (vec_perm_indices::clamp): Handle polynomial element_types.
22376 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22377 (vec_perm_indices::all_in_range_p): Likewise.
22378 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22380 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22381 polynomial vec_perm_indices element types.
22382 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22383 * fold-const.c (fold_vec_perm): Likewise.
22384 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22385 * tree-vect-generic.c (lower_vec_perm): Likewise.
22386 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22387 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22388 element type to HOST_WIDE_INT.
22390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22391 Alan Hayward <alan.hayward@arm.com>
22392 David Sherwood <david.sherwood@arm.com>
22394 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22395 rather than an int. Use plus_constant.
22396 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22397 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22400 Alan Hayward <alan.hayward@arm.com>
22401 David Sherwood <david.sherwood@arm.com>
22403 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22404 a HOST_WIDE_INT to a poly_int64.
22406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22407 Alan Hayward <alan.hayward@arm.com>
22408 David Sherwood <david.sherwood@arm.com>
22410 * calls.c (load_register_parameters): Cope with polynomial
22411 mode sizes. Require a constant size for BLKmode parameters
22412 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22413 forces a parameter to be padded at the lsb end in order to
22414 fill a complete number of words, require the parameter size
22415 to be ordered wrt UNITS_PER_WORD.
22417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22418 Alan Hayward <alan.hayward@arm.com>
22419 David Sherwood <david.sherwood@arm.com>
22421 * reload1.c (spill_stack_slot_width): Change element type
22422 from unsigned int to poly_uint64_pod.
22423 (alter_reg): Treat mode sizes as polynomial.
22425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22426 Alan Hayward <alan.hayward@arm.com>
22427 David Sherwood <david.sherwood@arm.com>
22429 * reload.c (complex_word_subreg_p): New function.
22430 (reload_inner_reg_of_subreg, push_reload): Use it.
22432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22433 Alan Hayward <alan.hayward@arm.com>
22434 David Sherwood <david.sherwood@arm.com>
22436 * lra-constraints.c (process_alt_operands): Reject matched
22437 operands whose sizes aren't ordered.
22438 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22445 that the mode size is in the set {1, 2, 4, 8, 16}.
22447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22448 Alan Hayward <alan.hayward@arm.com>
22449 David Sherwood <david.sherwood@arm.com>
22451 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22452 Use plus_constant instead of gen_rtx_PLUS.
22454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22455 Alan Hayward <alan.hayward@arm.com>
22456 David Sherwood <david.sherwood@arm.com>
22458 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22459 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22460 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22461 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22462 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22463 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22464 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22465 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22466 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22467 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22469 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22470 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22471 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22472 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22473 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22474 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22475 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22476 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22477 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22478 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22480 * expr.c (emit_move_resolve_push): Treat the input and result
22481 of PUSH_ROUNDING as a poly_int64.
22482 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22483 (emit_push_insn): Likewise.
22484 * lra-eliminations.c (mark_not_eliminable): Likewise.
22485 * recog.c (push_operand): Likewise.
22486 * reload1.c (elimination_effects): Likewise.
22487 * rtlanal.c (nonzero_bits1): Likewise.
22488 * calls.c (store_one_arg): Likewise. Require the padding to be
22489 known at compile time.
22491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22492 Alan Hayward <alan.hayward@arm.com>
22493 David Sherwood <david.sherwood@arm.com>
22495 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22496 Use plus_constant instead of gen_rtx_PLUS.
22498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22499 Alan Hayward <alan.hayward@arm.com>
22500 David Sherwood <david.sherwood@arm.com>
22502 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22503 rather than an int.
22505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22506 Alan Hayward <alan.hayward@arm.com>
22507 David Sherwood <david.sherwood@arm.com>
22509 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22510 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22511 via stack temporaries. Treat the mode size as polynomial too.
22513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22514 Alan Hayward <alan.hayward@arm.com>
22515 David Sherwood <david.sherwood@arm.com>
22517 * expr.c (expand_expr_real_2): When handling conversions involving
22518 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22519 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22520 as a poly_uint64 too.
22522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22523 Alan Hayward <alan.hayward@arm.com>
22524 David Sherwood <david.sherwood@arm.com>
22526 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22529 Alan Hayward <alan.hayward@arm.com>
22530 David Sherwood <david.sherwood@arm.com>
22532 * combine.c (can_change_dest_mode): Handle polynomial
22533 REGMODE_NATURAL_SIZE.
22534 * expmed.c (store_bit_field_1): Likewise.
22535 * expr.c (store_constructor): Likewise.
22536 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22537 and polynomial REGMODE_NATURAL_SIZE.
22538 (gen_lowpart_common): Likewise.
22539 * reginfo.c (record_subregs_of_mode): Likewise.
22540 * rtlanal.c (read_modify_subreg_p): Likewise.
22542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22543 Alan Hayward <alan.hayward@arm.com>
22544 David Sherwood <david.sherwood@arm.com>
22546 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22547 numbers of elements.
22549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22550 Alan Hayward <alan.hayward@arm.com>
22551 David Sherwood <david.sherwood@arm.com>
22553 * match.pd: Cope with polynomial numbers of vector elements.
22555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22556 Alan Hayward <alan.hayward@arm.com>
22557 David Sherwood <david.sherwood@arm.com>
22559 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22560 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
22567 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22568 (ipa_simd_modify_function_body): Likewise.
22570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22571 Alan Hayward <alan.hayward@arm.com>
22572 David Sherwood <david.sherwood@arm.com>
22574 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22575 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22576 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22577 (expand_vector_condition, vector_element): Likewise.
22578 (subparts_gt): New function.
22579 (get_compute_type): Use subparts_gt.
22580 (count_type_subparts): Delete.
22581 (expand_vector_operations_1): Use subparts_gt instead of
22582 count_type_subparts.
22584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22585 Alan Hayward <alan.hayward@arm.com>
22586 David Sherwood <david.sherwood@arm.com>
22588 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22589 (vect_compile_time_alias): ...this new function. Do the calculation
22590 on poly_ints rather than trees.
22591 (vect_prune_runtime_alias_test_list): Update call accordingly.
22593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22594 Alan Hayward <alan.hayward@arm.com>
22595 David Sherwood <david.sherwood@arm.com>
22597 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22599 (vect_schedule_slp_instance): Likewise.
22601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22602 Alan Hayward <alan.hayward@arm.com>
22603 David Sherwood <david.sherwood@arm.com>
22605 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22606 constant and extern definitions for variable-length vectors.
22607 (vect_get_constant_vectors): Note that the number of units
22608 is known to be constant.
22610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22611 Alan Hayward <alan.hayward@arm.com>
22612 David Sherwood <david.sherwood@arm.com>
22614 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22615 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
22623 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22626 Alan Hayward <alan.hayward@arm.com>
22627 David Sherwood <david.sherwood@arm.com>
22629 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22630 vectors as polynomial. Use build_index_vector for
22631 IFN_GOMP_SIMD_LANE.
22633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22634 Alan Hayward <alan.hayward@arm.com>
22635 David Sherwood <david.sherwood@arm.com>
22637 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22638 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22639 for variable-length vectors.
22640 (vectorizable_mask_load_store): Treat the number of units as
22641 polynomial, asserting that it is constant if the condition has
22642 already been enforced.
22643 (vectorizable_store, vectorizable_load): Likewise.
22645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22646 Alan Hayward <alan.hayward@arm.com>
22647 David Sherwood <david.sherwood@arm.com>
22649 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22650 of units as polynomial. Punt if we can't tell at compile time
22651 which vector contains the final result.
22653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22654 Alan Hayward <alan.hayward@arm.com>
22655 David Sherwood <david.sherwood@arm.com>
22657 * tree-vect-loop.c (vectorizable_induction): Treat the number
22658 of units as polynomial. Punt on SLP inductions. Use an integer
22659 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22660 cast of such a series for variable-length floating-point
22663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22664 Alan Hayward <alan.hayward@arm.com>
22665 David Sherwood <david.sherwood@arm.com>
22667 * tree.h (build_index_vector): Declare.
22668 * tree.c (build_index_vector): New function.
22669 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22670 of units as polynomial, forcibly converting it to a constant if
22671 vectorizable_reduction has already enforced the condition.
22672 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22673 to create a {1,2,3,...} vector.
22674 (vectorizable_reduction): Treat the number of units as polynomial.
22675 Choose vectype_in based on the largest scalar element size rather
22676 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
22684 number of units as polynomial.
22686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22687 Alan Hayward <alan.hayward@arm.com>
22688 David Sherwood <david.sherwood@arm.com>
22690 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22691 * target.def (autovectorize_vector_sizes): Return the vector sizes
22692 by pointer, using vector_sizes rather than a bitmask.
22693 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22694 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22695 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22697 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22698 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22699 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22700 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22701 * omp-general.c (omp_max_vf): Likewise.
22702 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22703 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22704 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22705 * tree-vect-slp.c (vect_slp_bb): Likewise.
22706 * doc/tm.texi: Regenerate.
22707 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22709 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22710 the vector size as a poly_uint64 rather than an unsigned int.
22711 (current_vector_size): Change from an unsigned int to a poly_uint64.
22712 (get_vectype_for_scalar_type): Update accordingly.
22713 * tree.h (build_truth_vector_type): Take the size and number of
22714 units as a poly_uint64 rather than an unsigned int.
22715 (build_vector_type): Add a temporary overload that takes
22716 the number of units as a poly_uint64 rather than an unsigned int.
22717 * tree.c (make_vector_type): Likewise.
22718 (build_truth_vector_type): Take the number of units as a poly_uint64
22719 rather than an unsigned int.
22721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22722 Alan Hayward <alan.hayward@arm.com>
22723 David Sherwood <david.sherwood@arm.com>
22725 * target.def (get_mask_mode): Take the number of units and length
22726 as poly_uint64s rather than unsigned ints.
22727 * targhooks.h (default_get_mask_mode): Update accordingly.
22728 * targhooks.c (default_get_mask_mode): Likewise.
22729 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22730 * doc/tm.texi: Regenerate.
22732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22733 Alan Hayward <alan.hayward@arm.com>
22734 David Sherwood <david.sherwood@arm.com>
22736 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22737 * omp-general.c (omp_max_vf): Likewise.
22738 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22739 (expand_omp_simd): Handle polynomial safelen.
22740 * omp-low.c (omplow_simd_context): Add a default constructor.
22741 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22742 (lower_rec_simd_input_clauses): Update accordingly.
22743 (lower_rec_input_clauses): Likewise.
22745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22746 Alan Hayward <alan.hayward@arm.com>
22747 David Sherwood <david.sherwood@arm.com>
22749 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22750 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22751 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22752 (vect_analyze_slp_cost): Likewise.
22753 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22754 (vect_model_load_cost): Likewise.
22756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22757 Alan Hayward <alan.hayward@arm.com>
22758 David Sherwood <david.sherwood@arm.com>
22760 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22761 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22762 from an unsigned int * to a poly_uint64_pod *.
22763 (calculate_unrolling_factor): New function.
22764 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22767 Alan Hayward <alan.hayward@arm.com>
22768 David Sherwood <david.sherwood@arm.com>
22770 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22771 from an unsigned int to a poly_uint64.
22772 (_loop_vec_info::slp_unrolling_factor): Likewise.
22773 (_loop_vec_info::vectorization_factor): Change from an int
22775 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22776 (vect_get_num_vectors): New function.
22777 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22778 (vect_get_num_copies): Use vect_get_num_vectors.
22779 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22780 to an unsigned int *.
22781 (vect_analyze_data_refs): Change min_vf from an int * to a
22783 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22784 than an unsigned HOST_WIDE_INT.
22785 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22786 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22787 to an unsigned int *.
22788 (vect_analyze_data_ref_dependences): Likewise.
22789 (vect_compute_data_ref_alignment): Handle polynomial vf.
22790 (vect_enhance_data_refs_alignment): Likewise.
22791 (vect_prune_runtime_alias_test_list): Likewise.
22792 (vect_shift_permute_load_chain): Likewise.
22793 (vect_supportable_dr_alignment): Likewise.
22794 (dependence_distance_ge_vf): Take the vectorization factor as a
22795 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22796 (vect_analyze_data_refs): Change min_vf from an int * to a
22798 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22799 vfm1 as a poly_uint64 rather than an int. Make the same change
22800 for the returned bound_scalar.
22801 (vect_gen_vector_loop_niters): Handle polynomial vf.
22802 (vect_do_peeling): Likewise. Update call to
22803 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22804 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22806 * tree-vect-loop.c (vect_determine_vectorization_factor)
22807 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22808 (vect_get_known_peeling_cost): Likewise.
22809 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22810 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22811 (vect_transform_loop): Likewise. Use the lowest possible VF when
22812 updating the upper bounds of the loop.
22813 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22814 rather than an int.
22815 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22816 polynomial unroll factors.
22817 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22818 (vect_make_slp_decision): Likewise.
22819 (vect_supported_load_permutation_p): Likewise, and polynomial
22821 (vect_analyze_slp_cost): Handle polynomial vf.
22822 (vect_slp_analyze_node_operations): Likewise.
22823 (vect_slp_analyze_bb_1): Likewise.
22824 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22825 than an unsigned HOST_WIDE_INT.
22826 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22827 (vectorizable_load): Handle polynomial vf.
22828 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22830 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22833 Alan Hayward <alan.hayward@arm.com>
22834 David Sherwood <david.sherwood@arm.com>
22836 * match.pd: Handle bit operations involving three constants
22837 and try to fold one pair.
22839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22841 * tree-vect-loop-manip.c: Include gimple-fold.h.
22842 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22843 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22844 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22845 Add a path that uses a step of VF instead of 1, but disable it
22847 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22848 and niters_no_overflow parameters. Update calls to
22849 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22850 Create a new SSA name if the latter choses to use a ste other
22851 than zero, and return it via niters_vector_mult_vf_var.
22852 * tree-vect-loop.c (vect_transform_loop): Update calls to
22853 vect_do_peeling, vect_gen_vector_loop_niters and
22854 slpeel_make_loop_iterate_ntimes.
22855 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22856 (vect_gen_vector_loop_niters): Update declarations after above changes.
22858 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
22860 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22861 128-bit round to integer instructions.
22862 (ceil<mode>2): Likewise.
22863 (btrunc<mode>2): Likewise.
22864 (round<mode>2): Likewise.
22866 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22868 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22869 unaligned VSX load/store on P8/P9.
22870 (expand_block_clear): Allow the use of unaligned VSX
22871 load/store on P8/P9.
22873 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22875 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22877 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22878 swap associated with both a load and a store.
22880 2018-01-02 Andrew Waterman <andrew@sifive.com>
22882 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22883 * config/riscv/riscv.md (clear_cache): Use it.
22885 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
22887 * web.c: Remove out-of-date comment.
22889 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22891 * expr.c (fixup_args_size_notes): Check that any existing
22892 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22893 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22894 (emit_single_push_insn): ...here.
22896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22898 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22899 (const_vector_encoded_nelts): New function.
22900 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22901 (const_vector_int_elt, const_vector_elt): Declare.
22902 * emit-rtl.c (const_vector_int_elt_1): New function.
22903 (const_vector_elt): Likewise.
22904 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22905 of CONST_VECTOR_ELT.
22907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22909 * expr.c: Include rtx-vector-builder.h.
22910 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22911 directly on the tree encoding.
22912 (const_vector_from_tree): Likewise.
22913 * optabs.c: Include rtx-vector-builder.h.
22914 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22915 sequence of "u" values.
22916 * vec-perm-indices.c: Include rtx-vector-builder.h.
22917 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22918 directly on the vec_perm_indices encoding.
22920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22922 * doc/rtl.texi (const_vector): Describe new encoding scheme.
22923 * Makefile.in (OBJS): Add rtx-vector-builder.o.
22924 * rtx-vector-builder.h: New file.
22925 * rtx-vector-builder.c: Likewise.
22926 * rtl.h (rtx_def::u2): Add a const_vector field.
22927 (CONST_VECTOR_NPATTERNS): New macro.
22928 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22929 (CONST_VECTOR_DUPLICATE_P): Likewise.
22930 (CONST_VECTOR_STEPPED_P): Likewise.
22931 (CONST_VECTOR_ENCODED_ELT): Likewise.
22932 (const_vec_duplicate_p): Check for a duplicated vector encoding.
22933 (unwrap_const_vec_duplicate): Likewise.
22934 (const_vec_series_p): Check for a non-duplicated vector encoding.
22935 Say that the function only returns true for integer vectors.
22936 * emit-rtl.c: Include rtx-vector-builder.h.
22937 (gen_const_vec_duplicate_1): Delete.
22938 (gen_const_vector): Call gen_const_vec_duplicate instead of
22939 gen_const_vec_duplicate_1.
22940 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22941 (gen_const_vec_duplicate): Use rtx_vector_builder.
22942 (gen_const_vec_series): Likewise.
22943 (gen_rtx_CONST_VECTOR): Likewise.
22944 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22945 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22946 Build a new vector rather than modifying a CONST_VECTOR in-place.
22947 (handle_special_swappables): Update call accordingly.
22948 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22949 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22950 Build a new vector rather than modifying a CONST_VECTOR in-place.
22951 (handle_special_swappables): Update call accordingly.
22953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22955 * simplify-rtx.c (simplify_const_binary_operation): Use
22956 CONST_VECTOR_ELT instead of XVECEXP.
22958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22960 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22961 the selector elements to be different from the data elements
22962 if the selector is a VECTOR_CST.
22963 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22964 ssizetype for the selector.
22966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22968 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
22969 before testing each element individually.
22970 * tree-vect-generic.c (lower_vec_perm): Likewise.
22972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22974 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
22975 * selftest-run-tests.c (selftest::run_tests): Call it.
22976 * vector-builder.h (vector_builder::operator ==): New function.
22977 (vector_builder::operator !=): Likewise.
22978 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
22979 (vec_perm_indices::all_from_input_p): New function.
22980 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22981 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
22982 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
22983 instead of reading the VECTOR_CST directly. Detect whether both
22984 vector inputs are the same before constructing the vec_perm_indices,
22985 and update the number of inputs argument accordingly. Use the
22986 utility functions added above. Only construct sel2 if we need to.
22988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22990 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22991 the broadcast of the low byte.
22992 (expand_mult_highpart): Use an explicit encoding for the permutes.
22993 * optabs-query.c (can_mult_highpart_p): Likewise.
22994 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22995 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22996 (vectorizable_bswap): Likewise.
22997 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22998 explicit encoding for the power-of-2 permutes.
22999 (vect_permute_store_chain): Likewise.
23000 (vect_grouped_load_supported): Likewise.
23001 (vect_permute_load_chain): Likewise.
23003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23005 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23006 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23007 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23008 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23009 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23010 (vect_gen_perm_mask_any): Likewise.
23012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23014 * int-vector-builder.h: New file.
23015 * vec-perm-indices.h: Include int-vector-builder.h.
23016 (vec_perm_indices): Redefine as an int_vector_builder.
23017 (auto_vec_perm_indices): Delete.
23018 (vec_perm_builder): Redefine as a stand-alone class.
23019 (vec_perm_indices::vec_perm_indices): New function.
23020 (vec_perm_indices::clamp): Likewise.
23021 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23022 (vec_perm_indices::new_vector): New function.
23023 (vec_perm_indices::new_expanded_vector): Update for new
23024 vec_perm_indices class.
23025 (vec_perm_indices::rotate_inputs): New function.
23026 (vec_perm_indices::all_in_range_p): Operate directly on the
23027 encoded form, without computing elided elements.
23028 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23029 encoding. Update for new vec_perm_indices class.
23030 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23031 the given vec_perm_builder.
23032 (expand_vec_perm_var): Update vec_perm_builder constructor.
23033 (expand_mult_highpart): Use vec_perm_builder instead of
23034 auto_vec_perm_indices.
23035 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23036 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23037 or double series encoding as appropriate.
23038 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23039 vec_perm_indices instead of auto_vec_perm_indices.
23040 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23041 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23042 (vect_permute_store_chain): Likewise.
23043 (vect_grouped_load_supported): Likewise.
23044 (vect_permute_load_chain): Likewise.
23045 (vect_shift_permute_load_chain): Likewise.
23046 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23047 (vect_transform_slp_perm_load): Likewise.
23048 (vect_schedule_slp_instance): Likewise.
23049 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23050 (vectorizable_mask_load_store): Likewise.
23051 (vectorizable_bswap): Likewise.
23052 (vectorizable_store): Likewise.
23053 (vectorizable_load): Likewise.
23054 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23055 vec_perm_indices instead of auto_vec_perm_indices. Use
23056 tree_to_vec_perm_builder to read the vector from a tree.
23057 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23058 vec_perm_builder instead of a vec_perm_indices.
23059 (have_whole_vector_shift): Use vec_perm_builder and
23060 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23061 truncation to calc_vec_perm_mask_for_shift.
23062 (vect_create_epilog_for_reduction): Likewise.
23063 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23064 from auto_vec_perm_indices to vec_perm_indices.
23065 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23066 instead of changing individual elements.
23067 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23068 the vector in d.perm.
23069 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23070 from auto_vec_perm_indices to vec_perm_indices.
23071 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23072 instead of changing individual elements.
23073 (arm_vectorize_vec_perm_const): Use new_vector to install
23074 the vector in d.perm.
23075 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23076 Update vec_perm_builder constructor.
23077 (rs6000_expand_interleave): Likewise.
23078 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23079 (rs6000_expand_interleave): Likewise.
23081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23083 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23084 to qimode could truncate the indices.
23085 * optabs.c (expand_vec_perm_var): Likewise.
23087 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23089 * Makefile.in (OBJS): Add vec-perm-indices.o.
23090 * vec-perm-indices.h: New file.
23091 * vec-perm-indices.c: Likewise.
23092 * target.h (vec_perm_indices): Replace with a forward class
23094 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23095 * optabs.h: Include vec-perm-indices.h.
23096 (expand_vec_perm): Delete.
23097 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23098 (expand_vec_perm_const): Declare.
23099 * target.def (vec_perm_const_ok): Replace with...
23100 (vec_perm_const): ...this new hook.
23101 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23102 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23103 * doc/tm.texi: Regenerate.
23104 * optabs.def (vec_perm_const): Delete.
23105 * doc/md.texi (vec_perm_const): Likewise.
23106 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23107 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23108 expand_vec_perm for constant permutation vectors. Assert that
23109 the mode of variable permutation vectors is the integer equivalent
23110 of the mode that is being permuted.
23111 * optabs-query.h (selector_fits_mode_p): Declare.
23112 * optabs-query.c: Include vec-perm-indices.h.
23113 (selector_fits_mode_p): New function.
23114 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23115 is defined, instead of checking whether the vec_perm_const_optab
23116 exists. Use targetm.vectorize.vec_perm_const instead of
23117 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23118 fit in the vector mode before using a variable permute.
23119 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23120 vec_perm_indices instead of an rtx.
23121 (expand_vec_perm): Replace with...
23122 (expand_vec_perm_const): ...this new function. Take the selector
23123 as a vec_perm_indices rather than an rtx. Also take the mode of
23124 the selector. Update call to shift_amt_for_vec_perm_mask.
23125 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23126 Use vec_perm_indices::new_expanded_vector to expand the original
23127 selector into bytes. Check whether the indices fit in the vector
23128 mode before using a variable permute.
23129 (expand_vec_perm_var): Make global.
23130 (expand_mult_highpart): Use expand_vec_perm_const.
23131 * fold-const.c: Includes vec-perm-indices.h.
23132 * tree-ssa-forwprop.c: Likewise.
23133 * tree-vect-data-refs.c: Likewise.
23134 * tree-vect-generic.c: Likewise.
23135 * tree-vect-loop.c: Likewise.
23136 * tree-vect-slp.c: Likewise.
23137 * tree-vect-stmts.c: Likewise.
23138 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23140 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23141 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23142 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23143 (aarch64_vectorize_vec_perm_const): ...this new function.
23144 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23145 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23146 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23147 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23148 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23149 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23150 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23152 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23153 check for NEON modes.
23154 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23155 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23156 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23157 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23159 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23160 the old VEC_PERM_CONST conditions.
23161 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23162 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23163 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23164 (ia64_vectorize_vec_perm_const_ok): Merge into...
23165 (ia64_vectorize_vec_perm_const): ...this new function.
23166 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23167 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23168 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23169 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23170 * config/mips/mips.c (mips_expand_vec_perm_const)
23171 (mips_vectorize_vec_perm_const_ok): Merge into...
23172 (mips_vectorize_vec_perm_const): ...this new function.
23173 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23174 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23175 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23176 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23177 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23178 (rs6000_expand_vec_perm_const): Delete.
23179 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23181 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23182 (altivec_expand_vec_perm_const_le): Take each operand individually.
23183 Operate on constant selectors rather than rtxes.
23184 (altivec_expand_vec_perm_const): Likewise. Update call to
23185 altivec_expand_vec_perm_const_le.
23186 (rs6000_expand_vec_perm_const): Delete.
23187 (rs6000_vectorize_vec_perm_const_ok): Delete.
23188 (rs6000_vectorize_vec_perm_const): New function.
23189 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23190 an element count and rtx array.
23191 (rs6000_expand_extract_even): Update call accordingly.
23192 (rs6000_expand_interleave): Likewise.
23193 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23194 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23195 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23196 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23197 (rs6000_expand_vec_perm_const): Delete.
23198 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23199 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23200 (altivec_expand_vec_perm_const_le): Take each operand individually.
23201 Operate on constant selectors rather than rtxes.
23202 (altivec_expand_vec_perm_const): Likewise. Update call to
23203 altivec_expand_vec_perm_const_le.
23204 (rs6000_expand_vec_perm_const): Delete.
23205 (rs6000_vectorize_vec_perm_const_ok): Delete.
23206 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23207 reference to the SPE evmerge intructions.
23208 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23209 an element count and rtx array.
23210 (rs6000_expand_extract_even): Update call accordingly.
23211 (rs6000_expand_interleave): Likewise.
23212 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23213 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23215 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23219 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23220 vector mode and that that mode matches the mode of the data
23222 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23223 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23224 directly using expand_vec_perm_1 when forcing selectors into
23226 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23230 * optabs-query.h (can_vec_perm_p): Delete.
23231 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23232 * optabs-query.c (can_vec_perm_p): Split into...
23233 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23234 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23235 particular selector is valid.
23236 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23237 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23238 (vect_grouped_load_supported): Likewise.
23239 (vect_shift_permute_load_chain): Likewise.
23240 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23241 (vect_transform_slp_perm_load): Likewise.
23242 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23243 (vectorizable_bswap): Likewise.
23244 (vect_gen_perm_mask_checked): Likewise.
23245 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23246 implementations of variable permutation vectors into account
23247 when deciding which selector to use.
23248 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23249 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23250 with a false third argument.
23251 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23252 to test whether the constant selector is valid and can_vec_perm_var_p
23253 to test whether a variable selector is valid.
23255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23257 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23258 * optabs-query.c (can_vec_perm_p): Likewise.
23259 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23260 instead of vec_perm_indices.
23261 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23262 (vect_gen_perm_mask_checked): Likewise,
23263 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23264 (vect_gen_perm_mask_checked): Likewise,
23266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23268 * optabs-query.h (qimode_for_vec_perm): Declare.
23269 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23270 (qimode_for_vec_perm): ...this new function.
23271 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23273 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23275 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23276 does not have a conditional at the top.
23278 2018-01-02 Richard Biener <rguenther@suse.de>
23280 * ipa-inline.c (big_speedup_p): Fix expression.
23282 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23285 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23288 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23292 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23293 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23294 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23295 cond_taken_branch_cost 3->4.
23297 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23299 PR tree-optimization/83581
23300 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23301 TODO_cleanup_cfg if any changes have been made.
23303 PR middle-end/83608
23304 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23305 convert_modes if target mode has the right side, but different mode
23308 PR middle-end/83609
23309 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23310 last argument when extracting from CONCAT. If either from_real or
23311 from_imag is NULL, use expansion through memory. If result is not
23312 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23313 the parts directly to inner mode, if even that fails, use expansion
23316 PR middle-end/83623
23317 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23318 check for bswap in mode rather than HImode and use that in expand_unop
23321 Copyright (C) 2018 Free Software Foundation, Inc.
23323 Copying and distribution of this file, with or without modification,
23324 are permitted in any medium without royalty provided the copyright
23325 notice and this notice are preserved.