* expmed.h (canonicalize_comparison): New declaration.
[official-gcc.git] / gcc / ChangeLog
blob0e1125595f14aa83eaa1507c50859b485f6fb095
1 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
3         * expmed.h (canonicalize_comparison): New declaration.
4         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6         * optabs.c (prepare_cmp_insn): Likewise.
7         * rtl.h (unsigned_condition_p): New function which checks if a
8         comparison operator is unsigned.
10 2018-08-16  Nathan Sidwell  <nathan@acm.org>
12         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
13         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
15 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
17         PR target/84711
18         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
19         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
20         (mov<mov>): ..this and enable unconditionally.
22 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
24         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
26 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
28         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
29         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
30         "Common" with "Target".
32 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
34         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
35         * doc/invoke.texi (mmitigate-rop): Remove.
36         * config/i386/i386.c: Do not include "regrename.h".
37         (ix86_rop_should_change_byte_p, reg_encoded_number)
38         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
39         Remove.
40         (ix86_reorg): Remove call to ix86_mitigate_rop.
41         * config/i386/i386.md (attr "modrm_class"): Remove.
42         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
43         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
44         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
46 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
48         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
49         allow folding of mergeh() and mergel() for the float and double types.
50         (fold_mergehl_helper): Rework to handle building a permute tree
51         for float vectors.
53 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
55         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
56         for TARGET_SSE.
58 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
60         * common.opt (fdiagnostics-show-labels): New option.
61         * diagnostic-show-locus.c (class layout_range): Add field
62         "m_label".
63         (class layout): Add field "m_show_labels_p".
64         (layout_range::layout_range): Add param "label" and use it to
65         initialize m_label.
66         (make_range): Pass in NULL for new "label" param of layout_range's
67         ctor.
68         (layout::layout): Initialize m_show_labels_p.
69         (layout::maybe_add_location_range): Pass in loc_range->m_label
70         when constructing layout_range instances.
71         (struct line_label): New struct.
72         (layout::print_any_labels): New member function.
73         (layout::print_line): Call it if label-printing is enabled.
74         (selftest::test_one_liner_labels): New test.
75         (selftest::test_diagnostic_show_locus_one_liner): Call it.
76         * diagnostic.c (diagnostic_initialize): Initialize
77         context->show_labels_p.
78         * diagnostic.h (struct diagnostic_context): Add field
79         "show_labels_p".
80         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
81         -fno-diagnostics-show-labels.
82         * dwarf2out.c (gen_producer_string): Add
83         OPT_fdiagnostics_show_labels to the ignored options.
84         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
85         param.
86         (gcc_rich_location::maybe_add_expr): Likewise.
87         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
88         label" param, defaulting to NULL.
89         (gcc_rich_location::add_expr): Add "label" param.
90         (gcc_rich_location::maybe_add_expr): Likewise.
91         (class text_range_label): New class.
92         (class range_label_for_type_mismatch): New class.
93         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
94         of format_warning_va.
95         (fmtwarn_n): Likewise for new params of format_warning_n_va.
96         * lto-wrapper.c (merge_and_complain): Add
97         OPT_fdiagnostics_show_labels to the "pick one setting" options.
98         (append_compiler_options): Likewise to the dropped options.
99         (append_diag_options): Likewise to the passed-on options.
100         * opts.c (common_handle_option): Handle the new option.
101         * selftest-diagnostic.c
102         (test_diagnostic_context::test_diagnostic_context): Enable
103         show_labels_p.
104         * substring-locations.c: Include "gcc-rich-location.h".
105         (format_warning_n_va): Add "fmt_label" and "param_label" params
106         and use them as appropriate.
107         (format_warning_va): Add "fmt_label" and "param_label" params,
108         passing them on to format_warning_n_va.
109         (format_warning_at_substring): Likewise.
110         (format_warning_at_substring_n): Likewise.
111         * substring-locations.h (format_warning_va): Add "fmt_label" and
112         "param_label" params.
113         (format_warning_n_va): Likewise.
114         (format_warning_at_substring): Likewise.
115         (format_warning_at_substring_n): Likewise.
116         * toplev.c (general_init): Initialize global_dc->show_labels_p.
118 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
120         PR testsuite/86519
121         * builtins.c (expand_builtin_memcmp): Do not expand the call
122         when overflow is detected.
124 2018-08-15  Martin Sebor  <msebor@redhat.com>
126         PR tree-optimization/71625
127         * config/aarch64/aarch64-builtins.c
128         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
130 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
132         * config/s390/s390.c (s390_reorg): Remove loop.
134 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
136         * config/darwin.c
137          (darwin_function_switched_text_sections): Delete.
138         * gcc/config/darwin.h
139          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
141 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
143         PR target/81685
144         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
145         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
146         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
148 2018-08-15  Martin Liska  <mliska@suse.cz>
150         PR tree-optimization/86925
151         * predict.c (expr_expected_value_1): When taking
152         later predictor, assign also probability.
153         Use fold_build2_initializer_loc in order to fold
154         the expression in -frounding-math.
156 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
158         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
159         patterns.
160         (expand_vec_perm_1): Try the new method.
162 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
164         PR target/86547
165         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
166         Check whether lra_live_max_point is 0 before dividing.
168 2018-08-14  Martin Sebor  <msebor@redhat.com>
170         PR tree-optimization/86650
171         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
172         (vrp_prop::check_mem_ref): Same.
174 2018-08-13  Liu Hao <lh_mouse@126.com>
176         * pretty-print.c (eat_esc_sequence): Swap the foreground and
177         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
178         and clear it thereafter, as it only works for DBCS.
180 2018-08-13  Liu Hao <lh_mouse@126.com>
182         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
183         handle returned by _get_osf_handle().
185 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
187         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
188         for folding vec_perm.
190 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
192         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
193         Add support for gimple-folding of vec_pack() and vec_unpack()
194         intrinsics.
196 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
198         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
199         vec_xst variants to the list.
200         (rs6000_gimple_fold_builtin): Add support for folding unaligned
201         vector loads and stores.
203 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
205         * config.gcc (rs6000-ibm-aix4.x): Delete.
206         (rs6000-ibm-aix5.1): Delete.
207         (rs6000-ibm-aix5.2): Delete.
208         (rs6000-ibm-aix5.3): Delete.
209         * config/rs6000/aix43.h: Delete.
210         * config/rs6000/aix51.h: Delete.
211         * config/rs6000/aix52.h: Delete.
212         * config/rs6000/t-aix43: Delete.
214 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
216         * config/s390/s390.c (s390_decompose_constant_pool_ref):
217         New function.
218         (s390_decompose_address): Factor out constant pool ref
219         decomposition.
221 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
223         * config/nds32/nds32-predicates.c
224         (nds32_can_use_bclr_p): Change return type as bool.
225         (nds32_can_use_bset_p): Ditto.
226         (nds32_can_use_btgl_p): Ditto.
227         (nds32_can_use_bitci_p): Ditto.
228         * config/nds32/nds32-protos.h
229         (nds32_can_use_bclr_p): Change declaration.
230         (nds32_can_use_bset_p): Ditto.
231         (nds32_can_use_btgl_p): Ditto.
232         (nds32_can_use_bitci_p): Ditto.
234 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
236         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
237         Support -msched-prolog-epilog option.
238         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
240 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
242         * common/config/nds32/nds32-common.c
243         (nds32_option_optimization_table): Enalbe -malways-align.
245 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
247         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
248         extra_headers.
249         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
250         OPT_misr_secure_ case.
251         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
252         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
253         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
254         secure attribute.
255         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
256         (nds32_isr_info): New field security_level.
257         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
258         * config/nds32/nds32.md (return_internal): Consider critical attribute.
259         * config/nds32/nds32.opt (misr-secure): New option.
260         * config/nds32/nds32_init.inc: New file.
261         * config/nds32/nds32_isr.h: New file.
263 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
265         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
266         Update comment for atomic instructions.
267         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
268         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
269         Remove.
270         (atomic_loaddi): Revise fence expansion to only emit fence prior to
271         load for __ATOMIC_SEQ_CST model.
272         (atomic_loaddi_1): Remove float register target.
273         (atomic_storedi): Handle CONST_INT values.
274         (atomic_storedi_1): Remove float register source.  Add special case
275         for zero value.
276         (memory_barrier): New expander and insn.
278 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
280         PR tree-optimization/86835
281         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
282         new_stmt after def_gsi, make sure to insert new_square_stmt after
283         that stmt, not 2 stmts before it.
285 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
287         PR target/82418
288         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
289         instead of SWI48.
291 2018-08-10  Martin Liska  <mliska@suse.cz>
293         PR target/83610
294         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
295         function type.
296         * builtins.c (expand_builtin_expect_with_probability):
297         New function.
298         (expand_builtin_expect_with_probability): New function.
299         (build_builtin_expect_predicate): Add new argumnet probability
300         for BUILT_IN_EXPECT_WITH_PROBABILITY.
301         (fold_builtin_expect):
302         (fold_builtin_2):
303         (fold_builtin_3):
304         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
305         * builtins.h (fold_builtin_expect): Set new argument.
306         * doc/extend.texi: Document __builtin_expect_with_probability.
307         * doc/invoke.texi: Likewise.
308         * gimple-fold.c (gimple_fold_call): Pass new argument.
309         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
310         also BUILT_IN_EXPECT_WITH_PROBABILITY.
311         * predict.c (get_predictor_value): New function.
312         (expr_expected_value): Add new argument probability. Assume
313         that predictor and probability are always non-null.
314         (expr_expected_value_1): Likewise.  For __builtin_expect and
315         __builtin_expect_with_probability set probability.  Handle
316         combination in binary expressions.
317         (tree_predict_by_opcode): Simplify code by simply calling
318         get_predictor_value.
319         (pass_strip_predict_hints::execute): Add handling of
320         BUILT_IN_EXPECT_WITH_PROBABILITY.
321         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
322         new predictor.
323         * tree.h (DECL_BUILT_IN_P): New function.
325 2018-08-10  Martin Liska  <mliska@suse.cz>
327         PR tree-optimization/85799
328         * passes.def: Add argument for pass_strip_predict_hints.
329         * predict.c (class pass_strip_predict_hints): Add new argument
330         early_p.
331         (strip_predictor_early): New function.
332         (pass_strip_predict_hints::execute): Call the function to
333         strip predictors.
334         (strip_predict_hints): New function.
335         * predict.def: Fix comment.
337 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
339         * Makefile.in: Clarify which tm.texi to copy over to assert the
340         right to grant a GFDL license for all.
342 2018-08-09  Jeff Law  <law@redhat.com>
344         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
345         unused variable.
347 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
349         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
350         prototype.
352 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
354         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
355         reductions for variable-length vectors.
357 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
359         PR other/84889
360         * common.opt (fdiagnostics-show-line-numbers): New option.
361         * diagnostic-show-locus.c (class layout): Add fields
362         "m_show_line_numbers_p" and "m_linenum_width";
363         (num_digits): New function.
364         (test_num_digits): New function.
365         (layout::layout): Initialize new fields.  Update m_x_offset
366         logic to handle any left margin.
367         (layout::print_source_line): Print line number when requested.
368         (layout::start_annotation_line): New member function.
369         (layout::print_annotation_line): Call it.
370         (layout::print_leading_fixits): Likewise.
371         (layout::print_trailing_fixits): Likewise.  Update calls to
372         move_to_column for new parameter.
373         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
374         it to potentially call start_annotation_line.
375         (layout::show_ruler): Call start_annotation_line.
376         (selftest::test_line_numbers_multiline_range): New selftest.
377         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
378         and selftest::test_line_numbers_multiline_range.
379         * diagnostic.c (diagnostic_initialize): Initialize
380         show_line_numbers_p.
381         * diagnostic.h (struct diagnostic_context): Add field
382         "show_line_numbers_p".
383         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
384         -fno-diagnostics-show-line-numbers.
385         * dwarf2out.c (gen_producer_string): Add
386         OPT_fdiagnostics_show_line_numbers to the ignored options.
387         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
388         one setting" options.
389         (append_compiler_options): Likewise to the dropped options.
390         (append_diag_options): Likewise to the passed-on options.
391         * opts.c (common_handle_option): Handle the new option.
392         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
394 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
396         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
397         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
398         third argument of type "const signed char" to descriptions of
399         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
400         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
401         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
402         __builtin_bcdsub_ov functions.
404 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
406         PR tree-optimization/86858
407         * tree-vect-loop.c (vect_is_simple_reduction): Restore
408         flow_bb_inside_loop_p calls.
410 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
412         PR tree-optimization/86871
413         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
414         instead of gimple_assign_lhs.
416 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
418         PR target/86887
419         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
420         register constraint to operand 0.
421         (add<mode>3_carryinC): Likewise.
422         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
424 2018-08-09  Martin Liska  <mliska@suse.cz>
426         PR c/86895
427         * common.opt: Remove extra line.
429 2018-08-09  Martin Liska  <mliska@suse.cz>
431         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
432         at the end of a line, make first letter capital and end up
433         a sentence with a dot.
434         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
435         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
436         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
437         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
438         (PARAM_MAX_ISL_OPERATIONS): Likewise.
439         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
440         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
441         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
442         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
443         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
444         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
445         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
446         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
447         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
448         (PARAM_TREE_REASSOC_WIDTH): Likewise.
449         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
450         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
451         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
453 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
455         PR target/84332
456         * config/s390/s390.c (s390_option_override_internal): Reduce the
457         stack-clash-protection-probe-interval param if it would be too big
458         for z900.
460 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
462         PR target/46179
463         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
464         * config/m68k/m68k.c (handle_move_double): Don't call
465         m68k_final_prescan_insn.
466         (m68k_adjust_decorated_operand): Renamed from
467         m68k_final_prescan_insn, remove first and third operand and
468         simplify.
469         (print_operand): Call it.
470         (print_operand_address): Call it.
472 2018-08-08  Nathan Sidwell  <nathan@acm.org>
474         * diagnostic.c (diagnostic_report_current_module): Use
475         linemap_included_from & linemap_included_from_linemap.
477 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
479         * config/aarch64/aarch64-cores.def: Add phecda core.
480         * config/aarch64/aarch64-tune.md: Regenerate.
481         * doc/invoke.texi: Add phecda core.
483 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
485         PR target/85295
486         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
487         definitions.
488         * config/s390/s390.md ("movti"): Add more alternatives for
489         constant to GPR copies.
491 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
493         * config/s390/s390.c: Fix whitespace damage throughout the file.
494         * config/s390/s390.h: Likewise.
495         * config/s390/tpf.h: Likewise.
497 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
499         * config/s390/s390.c (s390_loadrelative_operand_p):
500         Remove TARGET_CPU_ZARCH usages.
501         (s390_rtx_costs): Likewise.
502         (s390_legitimate_constant_p): Likewise.
503         (s390_cannot_force_const_mem): Likewise.
504         (legitimate_reload_constant_p): Likewise.
505         (s390_preferred_reload_class): Likewise.
506         (legitimize_pic_address): Likewise.
507         (legitimize_tls_address): Likewise.
508         (s390_split_branches): Removed.
509         (s390_add_execute): Removed.
510         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
511         (s390_mainpool_start): Likewise.
512         (s390_mainpool_finish): Likewise.
513         (s390_mainpool_cancel): Removed.
514         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
515         (s390_chunkify_cancel): Likewise.
516         (s390_return_addr_rtx): Likewise.
517         (s390_register_info): Remove split_branches_pending_p uages.
518         (s390_optimize_register_info): Likewise.
519         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
520         split_branches_pending_p usages.
521         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
522         (s390_load_got): Likewise.
523         (s390_expand_split_stack_prologue): Likewise.
524         (output_asm_nops): Likewise.
525         (s390_function_profiler): Likewise.
526         (s390_emit_call): Likewise.
527         (s390_conditional_register_usage): Likewise.
528         (s390_optimize_prologue): Likewise.
529         (s390_reorg): Remove TARGET_CPU_ZARCH and
530         split_branches_pending_p usages.
531         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
532         usages.
533         (s390_output_indirect_thunk_function): Likewise.
534         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
535         (TARGET_CPU_ZARCH_P): Removed.
536         (struct machine_function): Remove split_branches_pending_p.
537         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
539 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
541         * common/config/s390/s390-common.c (processor_flags_table):
542         Remove flags.
543         * config.gcc: Remove with_arch/with_tune support.
544         * config/s390/2064.md: Remove cpu attribute comparisons.
545         * config/s390/driver-native.c (s390_host_detect_local_cpu):
546         Remove MTN.
547         * config/s390/linux.h (ASM_SPEC):
548         Remove -march support.
549         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
550         Use a table to get an arch level.
551         * config/s390/s390-opts.h (enum processor_type):
552         Remove enum values.
553         * config/s390/s390.c
554         (processor_table): Remove entries, add arch_level values.
555         (s390_issue_rate): Remove cases.
556         (s390_option_override): Adjust
557         s390_option_override_internal() call.
558         (s390_option_override_internal): Remove deprecation warning.
559         (s390_valid_target_attribute_tree): Adjust
560         s390_option_override_internal() call.
561         * config/s390/s390.h (struct s390_processor):
562         Share with s390-c.c, add arch_level field.
563         * config/s390/s390.md:
564         Remove occurrences in cpu attribute.
565         * config/s390/s390.opt: Remove -march/-mtune support.
566         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
567         * doc/invoke.texi: Remove deprecation warning.
569 2018-08-08  Luis Machado  <luis.machado@linaro.org>
571         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
572         global.
573         (qdf24xx_tunings): Set vector cost structure to
574         qdf24xx_vector_cost.
576         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
577         <register_sextend>: Set to 3.
579 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
581         PR target/86838
582         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
583         * config/aarch64/aarch64-simd.md
584         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
585         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
586         (aarch64_frecpx<mode>): ...this new pattern.
587         * config/aarch64/aarch64-simd-builtins.def: Remove comment
588         about aarch64_frecp<FRECP:frecp_suffix><mode>.
590 2018-08-07  Martin Liska  <mliska@suse.cz>
592         PR middle-end/83023
593         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
594         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
595         * predict.def (PRED_MALLOC_NONNULL): New predictor.
596         * doc/extend.texi: Document that malloc attribute adds
597         hit to compiler.
599 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
601         PR target/86785
602         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
603         Define to speculation_safe_value_not_needed.
605 2018-08-06  Jeff Law  <law@redhat.com>
607         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
608         the vr_values instance to cprop_into_stmt.
609         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
610         (cprop_operand): Also query EVRP to determine if OP is a constant.
612 2018-08-06  Nathan Sidwell  <nathan@acm.org>
614         * diagnostic.c (diagnostic_report_current_module): Reroll
615         included-at loop.  Translate text.
617 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
619         * function-tests.c (selftest::test_expansion_to_rtl): Call
620         free_after_compilation.
622 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
624         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
626 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
628         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
629         loops with memory block operations from getting unrolled.
631 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
633         PR target/86807
634         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
635         Define to speculation_safe_value_not_needed.
637 2018-08-06  Jeff Law  <law@redhat.com>
639         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
640         assert.
642 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
644         PR target/86662
645         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
646         with all enabled __intN types.
648         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
650 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
652         * alias.c (record_set): Check for clobber high.
653         * cfgexpand.c (expand_gimple_stmt): Likewise.
654         * combine-stack-adj.c (single_set_for_csa): Likewise.
655         * combine.c (find_single_use_1): Likewise.
656         (set_nonzero_bits_and_sign_copies): Likewise.
657         (get_combine_src_dest): Likewise.
658         (is_parallel_of_n_reg_sets): Likewise.
659         (try_combine): Likewise.
660         (record_dead_and_set_regs_1): Likewise.
661         (reg_dead_at_p_1): Likewise.
662         (reg_dead_at_p): Likewise.
663         * dce.c (deletable_insn_p): Likewise.
664         (mark_nonreg_stores_1): Likewise.
665         (mark_nonreg_stores_2): Likewise.
666         * df-scan.c (df_find_hard_reg_defs): Likewise.
667         (df_uses_record): Likewise.
668         (df_get_call_refs): Likewise.
669         * dwarf2out.c (mem_loc_descriptor): Likewise.
670         * haifa-sched.c (haifa_classify_rtx): Likewise.
671         * ira-build.c (create_insn_allocnos): Likewise.
672         * ira-costs.c (scan_one_insn): Likewise.
673         * ira.c (equiv_init_movable_p): Likewise.
674         (rtx_moveable_p): Likewise.
675         (interesting_dest_for_shprep): Likewise.
676         * jump.c (mark_jump_label_1): Likewise.
677         * postreload-gcse.c (record_opr_changes): Likewise.
678         * postreload.c (reload_cse_simplify): Likewise.
679         (struct reg_use): Add source expr.
680         (reload_combine): Check for clobber high.
681         (reload_combine_note_use): Likewise.
682         (reload_cse_move2add): Likewise.
683         (move2add_note_store): Likewise.
684         * print-rtl.c (print_pattern): Likewise.
685         * recog.c (decode_asm_operands): Likewise.
686         (store_data_bypass_p): Likewise.
687         (if_test_bypass_p): Likewise.
688         * regcprop.c (kill_clobbered_value): Likewise.
689         (kill_set_value): Likewise.
690         * reginfo.c (reg_scan_mark_refs): Likewise.
691         * reload1.c (maybe_fix_stack_asms): Likewise.
692         (eliminate_regs_1): Likewise.
693         (elimination_effects): Likewise.
694         (mark_not_eliminable): Likewise.
695         (scan_paradoxical_subregs): Likewise.
696         (forget_old_reloads_1): Likewise.
697         * reorg.c (find_end_label): Likewise.
698         (try_merge_delay_insns): Likewise.
699         (redundant_insn): Likewise.
700         (own_thread_p): Likewise.
701         (fill_simple_delay_slots): Likewise.
702         (fill_slots_from_thread): Likewise.
703         (dbr_schedule): Likewise.
704         * resource.c (update_live_status): Likewise.
705         (mark_referenced_resources): Likewise.
706         (mark_set_resources): Likewise.
707         * rtl.c (copy_rtx): Likewise.
708         * rtlanal.c (reg_referenced_p): Likewise.
709         (single_set_2): Likewise.
710         (noop_move_p): Likewise.
711         (note_stores): Likewise.
712         * sched-deps.c (sched_analyze_reg): Likewise.
713         (sched_analyze_insn): Likewise.
715 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
717         * cse.c (invalidate_reg): New function extracted from...
718         (invalidate): ...here.
719         (canonicalize_insn): Check for clobber high.
720         (invalidate_from_clobbers): invalidate clobber highs.
721         (invalidate_from_sets_and_clobbers): Likewise.
722         (count_reg_usage): Check for clobber high.
723         (insn_live_p): Likewise.
724         * cselib.c (cselib_expand_value_rtx_1):Likewise.
725         (cselib_invalidate_regno): Check for clobber in setter.
726         (cselib_invalidate_rtx): Pass through setter.
727         (cselib_invalidate_rtx_note_stores):
728         (cselib_process_insn): Check for clobber high.
729         * cselib.h (cselib_invalidate_rtx): Add operand.
731 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
733         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
734         (mark_not_eliminable): Likewise.
735         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
736         * lra-lives.c (process_bb_lives): Check for clobber high.
737         * lra.c (new_insn_reg): Remember clobber highs.
738         (collect_non_operand_hard_regs): Check for clobber high.
739         (lra_set_insn_recog_data): Likewise.
740         (add_regs_to_insn_regno_info): Likewise.
741         (lra_update_insn_regno_info): Likewise.
743 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
745         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
746         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
748 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
750         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
751         (copy_insn_1): Likewise.
752         (gen_hard_reg_clobber_high): New gen function.
753         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
754         * genemit.c (gen_exp): Likewise.
755         (gen_emit_seq): Pass through info.
756         (gen_insn): Check for CLOBBER_HIGH.
757         (gen_expand): Pass through info.
758         (gen_split): Likewise.
759         (output_add_clobbers): Likewise.
760         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
761         (remove_clobbers): Likewise.
762         * rtl.h (gen_hard_reg_clobber_high): New declaration.
764 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
766         * doc/rtl.texi (clobber_high): Add.
767         (parallel): Add in clobber high
768         * rtl.c (rtl_check_failed_code3): Add function.
769         * rtl.def (CLOBBER_HIGH): Add expression.
770         * rtl.h (RTL_CHECKC3): Add macro.
771         (rtl_check_failed_code3): Add declaration.
772         (XC3EXP): Add macro.
774 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
776         PR target/86386
777         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
778         cfun->machine->max_used_stack_alignment if needed.
780 2018-08-04  Martin Sebor  <msebor@redhat.com>
782         PR tree-optimization/86571
783         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
784         NaN output to 4.
786 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
788         PR target/86799
789         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
790         Define.
792 2018-08-03  Jeff Law  <law@redhat.com>
794         PR target/86795
795         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
796         Define to speculation_safe_value_not_needed.
798 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
800         * doc/gcov.texi (-x): Remove duplicate "to".
801         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
802         (-Wif-not-aligned): Remove duplicate "is".
803         (-flto): Remove duplicate "the".
804         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
805         duplicate "v5.00.b".
806         (MSP430 Options): Remove duplicate "and" from the description
807         of "-mgprel-sec=regexp".
808         (x86 Options): Remove duplicate copies of "vmldLog102" and
809         vmlsLog104 from description of "-mveclibabi=type".
811 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
813         * internal-fn.h (first_commutative_argument): Declare.
814         * internal-fn.c (first_commutative_argument): New function.
815         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
816         restrictions for pattern statements.  Use first_commutative_argument
817         to look for commutative operands in calls to internal functions.
819 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
821         * Makefile.in (wide-int-range.o): New.
822         * tree-vrp.c: Move all the wide_int_* functions to...
823         * wide-int-range.cc: ...here.
824         * tree-vrp.h: Move all the wide_int_* prototypes to...
825         * wide-int-range.h: ...here.
827 2018-08-03  Tom de Vries  <tdevries@suse.de>
829         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
830         UI_NONE.
831         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
832         * except.c (output_function_exception_table): Do early exit if
833         targetm_common.except_unwind_info (&global_options) == UI_NONE.
835 2018-08-03  Martin Liska  <mliska@suse.cz>
837         * predict.c (dump_prediction): Change to 2 digits
838         in fraction part.
840 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
842         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
843         neon_dup_q to...
844         (falkor_am_1_gtov_gtov): ... a new insn reservation.
846 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
848         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
849         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
850         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
851         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
852         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
854 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
856         * diagnostic-show-locus.c (diagnostic_show_locus): Use
857         pp_take_prefix when saving the existing prefix.
858         * diagnostic.c (diagnostic_append_note): Likewise.
859         * langhooks.c (lhd_print_error_function): Likewise.
860         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
861         param's type.  Free the existing prefix.
862         (pp_take_prefix): New function.
863         (pretty_printer::pretty_printer): Drop the prefix parameter.
864         Rename the length parameter to match the comment.
865         (pretty_printer::~pretty_printer): Free the prefix.
866         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
867         parameter.
868         (struct pretty_printer): Drop the "const" from "prefix" field's
869         type and clarify memory management.
870         (pp_set_prefix): Drop the "const" from the 2nd param.
871         (pp_take_prefix): New decl.
873 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
875         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
876         for word_mode_ok here instead of passing as argument.
877         (expand_block_compare): Change select_block_compare_mode() call.
878         (expand_strncmp_gpr_sequence): New function.
879         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
881 2018-08-02  Jeff Law  <law@redhat.com>
883         PR target/86790
884         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
885         Define to speculation_safe_value_not_needed.
887         PR target/86784
888         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
889         Define to speculation_safe_value_not_needed.
891 2018-08-02  Tom de Vries  <tdevries@suse.de>
893         PR target/86660
894         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
895         function.  Return UI_TARGET unconditionally.
896         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
897         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
899 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
901         * genemit.c (print_overload_test): Fix typo.
903 2018-08-02  Richard Biener  <rguenther@suse.de>
905         PR tree-optimization/86816
906         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
907         which checks for value availability before querying it.
908         (gvn_uses_equal): Use it.
909         (same_succ_hash): Likewise.
910         (gimple_equal_p): Likewise.
912 2018-08-02  Nick Clifton  <nickc@redhat.com>
914         PR target/86813
915         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
916         Define to speculation_safe_value_not_needed.
918         PR target/86810
919         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
920         Define to speculation_safe_value_not_needed.
922         PR target/86810
923         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
924         Define to speculation_safe_value_not_needed.
926         PR target/86803
927         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
928         Define to speculation_safe_value_not_needed.
930         PR target/86797
931         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
932         Define to speculation_safe_value_not_needed.
934         PR target/86791
935         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
936         Define to speculation_safe_value_not_needed.
938         PR target/86789
939         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
940         Define to speculation_safe_value_not_needed.
942         PR target/86787
943         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
944         Define to speculation_safe_value_not_needed.
946         PR target/86782
947         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
948         speculation_safe_value_not_needed.
950         PR target/86781
951         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
952         to speculation_safe_value_not_needed.
954 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
956         * doc/md.texi: Expand the documentation of instruction names
957         to mention port-local uses.  Document '@' in pattern names.
958         * read-md.h (overloaded_instance, overloaded_name): New structs.
959         (mapping): Declare.
960         (md_reader::handle_overloaded_name): New member function.
961         (md_reader::get_overloads): Likewise.
962         (md_reader::m_first_overload): New member variable.
963         (md_reader::m_next_overload_ptr): Likewise.
964         (md_reader::m_overloads_htab): Likewise.
965         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
966         m_next_overload_ptr and m_overloads_htab.
967         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
968         (get_mode_token, get_code_token, get_int_token): New functions.
969         (map_attr_string): Add an optional argument that passes back
970         the associated iterator.
971         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
972         (md_reader::handle_overloaded_name, add_overload_instance): New
973         functions.
974         (apply_iterators): Handle '@' names.  Report an error if '@'
975         is used without iterators.
976         (initialize_iterators): Initialize the new iterator_group fields.
977         * genopinit.c (handle_overloaded_code_for)
978         (handle_overloaded_gen): New functions.
979         (main): Use them to print declarations of maybe_code_for_* and
980         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
981         * genemit.c (print_overload_arguments, print_overload_test)
982         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
983         (main): Use it to print definitions of maybe_code_for_* and
984         maybe_gen_* functions.
985         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
986         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
987         instead of explicit mode checks.
988         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
989         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
990         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
991         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
992         (aarch64_expand_compare_and_swap): Likewise
993         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
994         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
995         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
996         (aarch64_constant_pool_reload_icode): Delete.
997         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
998         instead of aarch64_constant_pool_reload_icode.  Use
999         code_for_aarch64_reload_mov instead of explicit mode checks.
1000         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1001         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1002         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1003         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1004         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1005         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1006         (aarch64_atomic_load_op_code): Delete.
1007         (aarch64_emit_atomic_load_op): Likewise.
1008         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1009         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
1010         instead of aarch64_emit_atomic_load_op.
1011         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1012         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1013         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1014         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1015         character before the pattern name.
1016         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1017         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1018         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1019         (aarch64_frecps<mode>): Likewise.
1020         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1021         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1022         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1023         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1024         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1026 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
1028         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1029         Allow HFmode constants if TARGET_FP_F16INST.
1031 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
1033         PR target/86014
1034         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1035         No longer check last store for clobber of address register.
1037 2018-08-02  Martin Liska  <mliska@suse.cz>
1039         PR gcov-profile/86817
1040         * gcov.c (process_all_functions): New function.
1041         (main): Call it.
1042         (process_file): Move functions processing to
1043         process_all_functions.
1045 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
1047         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1048         "const" to the "gimple *" and "rtx_insn *" parameters.
1049         * dumpfile.h (dump_user_location_t::dump_user_location_t):
1050         Likewise.
1051         (dump_location_t::dump_location_t): Likewise.
1053 2018-08-01  Martin Sebor  <msebor@redhat.com>
1055         PR tree-optimization/86650
1056         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1057         rather than a "gcall *".  Directly pass the data of interest
1058         to percent_K_format, rather than building a temporary CALL_EXPR
1059         to hold it.
1060         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1061         (gimple_fold_builtin_strncat): Adjust.
1062         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1063         gcall* argument with gimple*.
1064         * gimple-ssa-warn-restrict.c (check_call): Same.
1065         (wrestrict_dom_walker::before_dom_children): Same.
1066         (builtin_access::builtin_access): Same.
1067         (check_bounds_or_overlap): Same
1068         (maybe_diag_overlap): Same.
1069         (maybe_diag_offset_bounds): Same.
1070         * tree-diagnostic.c (default_tree_printer): Move usage of
1071         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1072         to this callsite.
1073         * tree-pretty-print.c (percent_K_format): Add argument.
1074         * tree-pretty-print.h: Add argument.
1075         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1076         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1077         (maybe_diag_stxncpy_trunc): Same.
1078         (handle_builtin_stxncpy): Same.
1079         (handle_builtin_strcat): Same.
1081 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1083         * match.pd: Optimise pointer range checks.
1085 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1087         PR tree-optimization/86758
1088         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1089         to remove pattern statements.
1091 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1093         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1094         result of dfs_enumerate_from when constructing stmt_vec_infos,
1095         instead of additionally calling get_loop_body.
1097 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1099         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1100         parameter.
1101         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1102         When creating an iv, assert that the step is not known to be zero.
1103         (vect_setup_realignment): Update call accordingly.
1104         * tree-vect-stmts.c (vectorizable_store): Likewise.
1105         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
1107 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1109         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1110         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1111         (vectorizable_reduction): Likewise.
1112         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1113         (vect_detect_hybrid_slp_stmts): Likewise.
1114         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1116 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
1118         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1119         (wide_int_set_zero_nonzero_bits): ...this.
1120         (zero_nonzero_bits_from_vr): Rename to...
1121         (vrp_set_zero_nonzero_bits): ...this.
1122         (extract_range_from_multiplicative_op_1): Abstract wide int
1123         code...
1124         (wide_int_range_multiplicative_op): ...here.
1125         (extract_range_from_binary_expr_1): Extract wide int binary
1126         operations into their own functions.
1127         (wide_int_range_lshift): New.
1128         (wide_int_range_can_optimize_bit_op): New.
1129         (wide_int_range_shift_undefined_p): New.
1130         (wide_int_range_bit_xor): New.
1131         (wide_int_range_bit_ior): New.
1132         (wide_int_range_bit_and): New.
1133         (wide_int_range_trunc_mod): New.
1134         (extract_range_into_wide_ints): New.
1135         (vrp_shift_undefined_p): New.
1136         (extract_range_from_multiplicative_op): New.
1137         (vrp_can_optimize_bit_op): New.
1138         * tree-vrp.h (value_range::dump): New.
1139         (wide_int_range_multiplicative_op): New.
1140         (wide_int_range_lshift):New.
1141         (wide_int_range_shift_undefined_p): New.
1142         (wide_int_range_bit_xor): New.
1143         (wide_int_range_bit_ior): New.
1144         (wide_int_range_bit_and): New.
1145         (wide_int_range_trunc_mod): New.
1146         (zero_nonzero_bits_from_bounds): Rename to...
1147         (wide_int_set_zero_nonzero_bits): ...this.
1148         (zero_nonzero_bits_from_vr): Rename to...
1149         (vrp_set_zero_nonzero_bits): ...this.
1150         (range_easy_mask_min_max): Rename to...
1151         (wide_int_range_can_optimize_bit_op): this.
1152         * vr-values.c (simplify_bit_ops_using_ranges): Rename
1153         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1155 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1157         * tree-vectorizer.h (vect_orig_stmt): New function.
1158         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1159         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1160         (vect_create_epilog_for_reduction): Likewise.
1161         (vectorizable_live_operation): Likewise.
1162         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1163         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1164         * tree-vect-stmts.c (vectorizable_call): Likewise.
1165         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1167 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1169         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1170         argument.
1171         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1172         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1173         (vect_transform_loop): Likewise.
1174         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1176 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1178         * tree-vectorizer.h (vect_schedule_slp): Return void.
1179         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1180         (vect_schedule_slp): Likewise.
1182 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1184         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1185         argument.
1186         (vect_transform_loop): Update calls accordingly.  Schedule SLP
1187         instances before the main loop, if any exist.
1189 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
1191         PR tree-optimization/86749
1192         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1193         If the lhs is used in a COND_EXPR, check that it is being used
1194         as the "then" or "else" value.
1196 2018-08-01  Tom de Vries  <tdevries@suse.de>
1198         PR target/86800
1199         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1200         speculation_safe_value_not_needed.
1202 2018-08-01  Richard Biener  <rguenther@suse.de>
1204         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1205         as base and offset.
1207 2018-08-01  Martin Liska  <mliska@suse.cz>
1209         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1210         format how successful transformation is dumped.
1211         (gimple_mod_pow2_value_transform): Likewise.
1212         (gimple_mod_subtract_transform): Likewise.
1213         (gimple_stringops_transform): Likewise.
1215 2018-08-01  Martin Liska  <mliska@suse.cz>
1217         PR value-prof/35543
1218         * value-prof.c (interesting_stringop_to_profile_p):
1219         Simplify the code and add BUILT_IN_MEMMOVE.
1220         (gimple_stringops_transform): Likewise.
1222 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
1224         * config/aarch64/aarch64-simd.md
1225         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1226         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1227         use GPI iterator instead of SI mode.
1229 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
1231         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1232         rs6000_speculation_barrier.
1233         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1234         new barrier pattern name.
1236 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
1238         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1239         (speculation_barrier): New insn.
1241 2018-08-01  Richard Biener  <rguenther@suse.de>
1243         PR bootstrap/86724
1244         * graphite.h: Include isl/id.h and isl/space.h to allow build
1245         with ISL 0.20.
1247 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
1249         PR target/86651
1250         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1251         mode for COFF targets.
1252         * defaults.h (TARGET_COFF): Define.
1253         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1254         TARGET_COFF): Define.
1255         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1256         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1257         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1259 2018-07-31  Alexandre Oliva <oliva@adacore.com>
1261         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1262         * gimple-streamer-out.c (output_bb): Save it.
1263         * lto-streamer-in.c (input_struct_function_base): Restore
1264         instance discriminator if available.  Create map on demand.
1265         * lto-streamer-out.c (output_struct_function_base): Save it if
1266         available.
1267         * final.c (decl_to_instance_map): Document LTO strategy.
1269 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
1270             Olivier Hainque  <hainque@adacore.com>
1272         * debug.h (decl_to_instance_map_t): New type.
1273         (decl_to_instance_map): Declare.
1274         (maybe_create_decl_to_instance_map): New inline function.
1275         * final.c (bb_discriminator, last_bb_discriminator): New statics,
1276         to track basic block discriminators.
1277         (final_start_function_1): Initialize them.
1278         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1279         bb_discriminator.
1280         (decl_to_instance_map): New variable.
1281         (map_decl_to_instance, maybe_set_discriminator): New functions.
1282         (notice_source_line): Set discriminator.
1284 2018-07-31  Ian Lance Taylor  <iant@golang.org>
1286         * targhooks.c (default_have_speculation_safe_value): Add
1287         ATTRIBUTE_UNUSED.
1289 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
1291         * dump-context.h: Include "pretty-print.h".
1292         (dump_context::refresh_dumps_are_enabled): New decl.
1293         (dump_context::emit_item): New decl.
1294         (class dump_context): Add fields "m_test_pp" and
1295         "m_test_pp_flags".
1296         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1297         (temp_dump_context::get_dumped_text): New decl.
1298         (class temp_dump_context): Add field "m_pp".
1299         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1300         (dump_context::refresh_dumps_are_enabled): ...and add a test for
1301         m_test_pp.
1302         (set_dump_file): Update for above change.
1303         (set_alt_dump_file): Likewise.
1304         (dump_loc): New overload, taking a pretty_printer *.
1305         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
1306         to any test pretty-printer.
1307         (make_item_for_dump_gimple_stmt): New function, adapted from
1308         optinfo::add_gimple_stmt.
1309         (dump_context::dump_gimple_stmt): Call it, and use the result,
1310         eliminating the direct usage of dump_file and alt_dump_file in
1311         favor of indirectly using them via emit_item.
1312         (make_item_for_dump_gimple_expr): New function, adapted from
1313         optinfo::add_gimple_expr.
1314         (dump_context::dump_gimple_expr): Call it, and use the result,
1315         eliminating the direct usage of dump_file and alt_dump_file in
1316         favor of indirectly using them via emit_item.
1317         (make_item_for_dump_generic_expr): New function, adapted from
1318         optinfo::add_tree.
1319         (dump_context::dump_generic_expr): Call it, and use the result,
1320         eliminating the direct usage of dump_file and alt_dump_file in
1321         favor of indirectly using them via emit_item.
1322         (make_item_for_dump_printf_va): New function, adapted from
1323         optinfo::add_printf_va.
1324         (make_item_for_dump_printf): New function.
1325         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1326         and use the result, eliminating the direct usage of dump_file and
1327         alt_dump_file in favor of indirectly using them via emit_item.
1328         (make_item_for_dump_dec): New function.
1329         (dump_context::dump_dec): Call it, and use the result,
1330         eliminating the direct usage of dump_file and alt_dump_file in
1331         favor of indirectly using them via emit_item.
1332         (make_item_for_dump_symtab_node): New function, adapted from
1333         optinfo::add_symtab_node.
1334         (dump_context::dump_symtab_node): Call it, and use the result,
1335         eliminating the direct usage of dump_file and alt_dump_file in
1336         favor of indirectly using them via emit_item.
1337         (dump_context::begin_scope): Reimplement, avoiding direct usage
1338         of dump_file and alt_dump_file in favor of indirectly using them
1339         via emit_item.
1340         (dump_context::emit_item): New member function.
1341         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1342         Set up test pretty-printer on the underlying context.  Call
1343         refresh_dumps_are_enabled.
1344         (temp_dump_context::~temp_dump_context): Call
1345         refresh_dumps_are_enabled.
1346         (temp_dump_context::get_dumped_text): New member function.
1347         (selftest::verify_dumped_text): New function.
1348         (ASSERT_DUMPED_TEXT_EQ): New macro.
1349         (selftest::test_capture_of_dump_calls): Run all tests twice, with
1350         and then without optinfo enabled.  Add uses of
1351         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
1352         * dumpfile.h: Update comment for the dump_* API.
1353         * optinfo-emit-json.cc
1354         (selftest::test_building_json_from_dump_calls): Update for new
1355         param for temp_dump_context ctor.
1356         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1357         and "m_owned" field.
1358         (optinfo_item::~optinfo_item): Likewise.
1359         (optinfo::add_item): New member function.
1360         (optinfo::emit): Update comment.
1361         (optinfo::add_string): Delete.
1362         (optinfo::add_printf): Delete.
1363         (optinfo::add_printf_va): Delete.
1364         (optinfo::add_gimple_stmt): Delete.
1365         (optinfo::add_gimple_expr): Delete.
1366         (optinfo::add_tree): Delete.
1367         (optinfo::add_symtab_node): Delete.
1368         (optinfo::add_dec): Delete.
1369         * optinfo.h (class dump_context): New forward decl.
1370         (optinfo::add_item): New decl.
1371         (optinfo::add_string): Delete.
1372         (optinfo::add_printf): Delete.
1373         (optinfo::add_printf_va): Delete.
1374         (optinfo::add_gimple_stmt): Delete.
1375         (optinfo::add_gimple_expr): Delete.
1376         (optinfo::add_tree): Delete.
1377         (optinfo::add_symtab_node): Delete.
1378         (optinfo::add_dec): Delete.
1379         (optinfo::add_poly_int): Delete.
1380         (optinfo_item::optinfo_item): Remove "owned" param.
1381         (class optinfo_item): Remove field "m_owned".
1383 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1385         PR middle-end/86705
1386         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1387         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1388         requested variable alignment.
1389         (expand_one_ssa_partition): Likewise.
1390         (expand_one_var): Likewise.
1392 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1394         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1395         to speculation_safe_value_not_needed.
1397 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1399         * targhooks.h (speculation_safe_value_not_needed): New prototype.
1400         * targhooks.c (speculation_safe_value_not_needed): New function.
1401         * target.def (have_speculation_safe_value): Update documentation.
1402         * doc/tm.texi: Regenerated.
1404 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1406         * config/aarch64/iterators.md (ALLI_TI): New iterator.
1407         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1408         expand.
1409         (despeculate_copy<ALLI:mode>_insn): New insn.
1410         (despeculate_copyti_insn): New insn.
1411         (despeculate_simple<ALLI:mode>): New insn
1412         (despeculate_simpleti): New insn.
1413         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1414         function.
1415         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1416         aarch64_speculation_safe_value.
1417         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1419 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1421         * config/aarch64/aarch64-speculation.cc: New file.
1422         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1423         before pass_reorder_blocks.
1424         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1425         prototype.
1426         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1427         X14 and X15 when tracking speculation.
1428         * config/aarch64/aarch64.md (register name constants): Add
1429         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1430         (unspec): Add UNSPEC_SPECULATION_TRACKER.
1431         (speculation_barrier): New insn attribute.
1432         (cmp<mode>): Allow SP in comparisons.
1433         (speculation_tracker): New insn.
1434         (speculation_barrier): Add speculation_barrier attribute.
1435         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1436         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1437         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1439 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1441         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1442         aarch64_track_speculation is true.
1443         (tb<optab><mode>1): Likewise.
1444         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1445         generate CB[N]Z when tracking speculation.
1446         (aarch64_split_compare_and_swap): Likewise.
1447         (aarch64_split_atomic_op): Likewise.
1449 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1451         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1453 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1455         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1456         (speculation_barrier): New insn.
1458 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1460         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1461         * config/arm/arm.md (speculation_barrier): New expand.
1462         (speculation_barrier_insn): New pattern.
1464 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
1466         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1467         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1468         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1469         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1470         list.
1471         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1472         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1473         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1474         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1475         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1476         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1477         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1478         * builtins.c (expand_speculation_safe_value): New function.
1479         (expand_builtin): Call it.
1480         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1481         * doc/extend.texi: Document __builtin_speculation_safe_value.
1482         * doc/md.texi: Document "speculation_barrier" pattern.
1483         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1484         TARGET_HAVE_SPECULATION_SAFE_VALUE.
1485         * doc/tm.texi: Regenerated.
1486         * target.def (have_speculation_safe_value, speculation_safe_value): New
1487         hooks.
1488         * targhooks.c (default_have_speculation_safe_value): New function.
1489         (default_speculation_safe_value): New function.
1490         * targhooks.h (default_have_speculation_safe_value): Add prototype.
1491         (default_speculation_safe_value): Add prototype.
1493 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
1495         * dump-context.h (dump_context::dump_loc): New decl.
1496         * dumpfile.c (dump_context::dump_loc): New member function.
1497         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1498         and dump_gimple_stmt.
1499         (dump_context::dump_gimple_expr_loc): Likewise, using
1500         dump_gimple_expr.
1501         (dump_context::dump_generic_expr_loc): Likewise, using
1502         dump_generic_expr.
1503         (dump_context::dump_printf_loc_va): Likewise, using
1504         dump_printf_va.
1505         (dump_context::begin_scope): Explicitly using the global function
1506         "dump_loc", rather than the member function.
1508 2018-07-31  Martin Sebor  <msebor@redhat.com>
1510         PR tree-optimization/86741
1511         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1513 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
1515         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1516         depend on whether prefetch instructions will be emitted or not.
1517         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1518         will be emitted or not.
1519         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1520         (TARGET_SETMEM_PFD): New macros.
1522 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1524         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1525         (NULL_STMT_VEC_INFO): Delete.
1526         (stmt_vec_info::operator*): Likewise.
1527         (stmt_vec_info::operator gimple *): Likewise.
1528         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1529         of NULL_STMT_VEC_INFO.
1530         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1531         (vect_reassociating_reduction_p): Likewise.
1532         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1533         (vectorizable_store): Likewise.
1534         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1535         (vec_info::free_stmt_vec_infos): Likewise.
1537 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1539         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1540         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1542 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1544         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1545         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1546         (vec_info::free_stmt_vec_info): New private member functions.
1547         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1548         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1549         * tree-parloops.c (gather_scalar_reductions): Remove calls to
1550         set_stmt_vec_info_vec and free_stmt_vec_infos.
1551         * tree-vect-loop.c (_loop_vec_info): Remove call to
1552         set_stmt_vec_info_vec.
1553         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1554         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1555         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1556         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1557         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
1558         assignments in {vec_info::,}new_stmt_vec_info that are redundant
1559         with the clearing in the xcalloc.
1560         (stmt_vec_info_vec): Delete.
1561         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1562         (vectorize_loops): Likewise.
1563         (vec_info::~vec_info): Remove argument from call to
1564         free_stmt_vec_infos.
1565         (vec_info::add_stmt): Remove vinfo argument from call to
1566         new_stmt_vec_info.
1568 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1570         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1571         rather than a gimple stmt.
1572         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
1573         information for pattern statements when passed the original
1574         statement; instead wait to be passed the pattern statement itself.
1575         Don't call set_vinfo_for_stmt here.
1576         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1577         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1578         stmt_vec_infos here.
1579         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1580         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1581         stmt_vec_infos entry.
1583 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1585         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1586         * tree-vectorizer.c (vec_info::replace_stmt): New function.
1587         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1588         * tree-vect-stmts.c (vectorizable_call): Likewise.
1589         (vectorizable_simd_clone_call): Likewise.
1591 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1593         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1594         * tree-vectorizer.c (vec_info::remove_stmt): New function.
1595         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1596         * tree-vect-loop.c (vect_transform_loop): Likewise.
1597         * tree-vect-slp.c (vect_schedule_slp): Likewise.
1598         * tree-vect-stmts.c (vect_remove_stores): Likewise.
1600 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1602         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1603         (vect_dr_stmt): Delete.
1604         * tree-vectorizer.c (vec_info::lookup_dr): New function.
1605         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1606         of DR_VECT_AUX.
1607         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1608         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1609         (vect_verify_datarefs_alignment, vect_peeling_supportable)
1610         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1611         (vect_analyze_data_refs): Likewise.
1612         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
1613         argument.
1614         (vect_find_same_alignment_drs): Likewise.
1615         (vect_slp_analyze_node_dependences): Update calls accordingly.
1616         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
1617         instead of DR_VECT_AUX.
1618         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1619         of a vector data references.  Use vec_info::lookup_dr instead of
1620         DR_VECT_AUX.
1621         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1622         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
1623         instead of DR_VECT_AUX.
1625 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1627         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1628         dr_vec_info.
1629         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1630         accordingly.
1631         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1632         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1633         (vect_gen_prolog_loop_niters): Likewise.
1635 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1637         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1638         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1639         (vect_known_alignment_in_bytes, vect_dr_behavior)
1640         (vect_get_scalar_dr_size): Take references as dr_vec_infos
1641         instead of data_references.  Update calls to other routines for
1642         which the same change has been made.
1643         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1644         dr_vec_infos instead of stmt_vec_infos.
1645         (vect_analyze_data_ref_dependence): Update call accordingly.
1646         (vect_slp_analyze_data_ref_dependence)
1647         (vect_record_base_alignments): Use DR_VECT_AUX.
1648         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1649         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1650         (vector_alignment_reachable_p, vect_get_data_access_cost)
1651         (vect_peeling_supportable, vect_analyze_group_access_1)
1652         (vect_analyze_group_access, vect_analyze_data_ref_access)
1653         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1654         (vect_compile_time_alias, vect_small_gap_p)
1655         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1656         (vect_supportable_dr_alignment): Take references as dr_vec_infos
1657         instead of data_references.  Update calls to other routines for
1658         which the same change has been made.
1659         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1660         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1661         (vect_slp_analyze_and_verify_node_alignment)
1662         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1663         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1664         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
1665         above changes.
1666         (_vect_peel_info::dr): Replace with...
1667         (_vect_peel_info::dr_info): ...this new field.
1668         (vect_peeling_hash_get_most_frequent)
1669         (vect_peeling_hash_choose_best_peeling): Update accordingly.
1670         (vect_peeling_hash_get_lowest_cost):
1671         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
1672         routines for which the same change has been made.
1673         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
1674         data_reference.
1675         * tree-vect-loop-manip.c (get_misalign_in_elems)
1676         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
1677         above changes.
1678         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1679         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1680         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1681         (get_group_load_store_type, get_negative_load_store_type)
1682         (vect_get_data_ptr_increment, vectorizable_store)
1683         (vectorizable_load): Likewise.
1684         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1685         Update calls to other routines for which the same change has been made.
1687 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1689         * tree-vectorizer.h (vec_info::move_dr): New member function.
1690         (dataref_aux): Rename to...
1691         (dr_vec_info): ...this and add "dr" and "stmt" fields.
1692         (_stmt_vec_info::dr_aux): Update accordingly.
1693         (_stmt_vec_info::data_ref_info): Delete.
1694         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1695         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1696         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1697         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1698         of data_ref.
1699         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
1700         (STMT_VINFO_DR_INFO): New macro.
1701         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1702         (set_dr_misalignment): Update after rename of dataref_aux.
1703         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
1704         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1705         initialization of STMT_VINFO_DATA_REF.
1706         * tree-vectorizer.c (vec_info::move_dr): New function.
1707         * tree-vect-patterns.c (vect_recog_bool_pattern)
1708         (vect_recog_mask_conversion_pattern)
1709         (vect_recog_gather_scatter_pattern): Use it.
1710         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1711         the "dr" and "stmt" fields of dr_vec_info instead of
1712         STMT_VINFO_DATA_REF.
1714 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1716         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1717         (is_pattern_stmt_p): Use it.
1718         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1719         on pattern statements.
1721 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1723         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1724         original stmt as a stmt_vec_info rather than a gimple stmt.
1725         (vect_pattern_recog_1): Take the statement directly as a
1726         stmt_vec_info, rather than via a gimple_stmt_iterator.
1727         Update call to vect_mark_pattern_stmts.
1728         (vect_pattern_recog): Update calls accordingly.
1730 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1732         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1733         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1734         a vect_def_type for the first argument.
1735         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1736         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
1737         operand if it isn't defined by a vectorized statement.
1738         (vect_build_gather_load_calls): Remove the mask_dt argument and
1739         update calls to vect_get_vec_def_for_stmt_copy.
1740         (vectorizable_bswap): Likewise the dt argument.
1741         (vectorizable_call): Update calls to vectorizable_bswap and
1742         vect_get_vec_def_for_stmt_copy.
1743         (vectorizable_simd_clone_call, vectorizable_assignment)
1744         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1745         (vectorizable_comparison): Update calls to
1746         vect_get_vec_def_for_stmt_copy.
1747         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
1748         vect_is_simple_use.
1749         (vect_get_loop_based_defs): Remove dt argument and update call
1750         to vect_get_vec_def_for_stmt_copy.
1751         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1752         and vect_get_vec_def_for_stmt_copy.
1753         (vectorizable_load): Update calls to vect_build_gather_load_calls
1754         and vect_get_vec_def_for_stmt_copy.
1755         * tree-vect-loop.c (vect_create_epilog_for_reduction)
1756         (vectorizable_reduction, vectorizable_live_operation): Update calls
1757         to vect_get_vec_def_for_stmt_copy.
1759 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1761         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1762         and gimple stmt arguments with a stmt_vec_info.
1763         (vect_record_base_alignments): Update calls accordingly.
1764         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1765         and gimple stmt arguments with a stmt_vec_info.
1766         (vect_build_slp_tree_1): Remove vinfo argument and update call
1767         to vect_record_max_nunits.
1768         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1769         and vect_record_max_nunits.
1771 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1773         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1774         file and take a stmt_vec_info instead of a gimple stmt.
1775         (supportable_widening_operation, vect_finish_replace_stmt)
1776         (vect_finish_stmt_generation, vect_get_store_rhs)
1777         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1778         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1779         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1780         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1781         (vect_create_data_ref_ptr, bump_vector_ptr)
1782         (vect_permute_store_chain, vect_setup_realignment)
1783         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1784         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1785         (vectorizable_reduction, vectorizable_induction)
1786         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1787         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1788         than gimple stmts as arguments.
1789         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1790         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1791         (can_group_stmts_p, vect_check_gather_scatter)
1792         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1793         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1794         (vect_permute_load_chain, vect_shift_permute_load_chain)
1795         (vect_transform_grouped_load)
1796         (vect_record_grouped_load_vectors): Likewise.
1797         * tree-vect-loop.c (vect_fixup_reduc_chain)
1798         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1799         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1800         (vectorizable_reduction, vectorizable_induction)
1801         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1802         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1803         (vect_get_load_store_mask): Likewise.
1804         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1805         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1806         * tree-vect-stmts.c (vect_mark_relevant)
1807         (is_simple_and_all_uses_invariant)
1808         (exist_non_indexing_operands_for_use_p, process_use)
1809         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1810         (vect_get_vec_def_for_operand, vect_get_vec_defs)
1811         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1812         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1813         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1814         (get_negative_load_store_type, get_load_store_type)
1815         (vect_check_load_store_mask, vect_check_store_rhs)
1816         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1817         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1818         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1819         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1820         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1821         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1822         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1823         (supportable_widening_operation): Likewise.
1825 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1827         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1828         a stmt_vec_info instead of a gcall.
1829         (vect_check_gather_scatter): Update call accordingly.
1830         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1831         of a gphi.
1832         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1833         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1834         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1835         instead of a gimple stmt.
1836         (vect_transform_loop): Update calls accordingly.
1837         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1838         stmt_vec_infos instead of gimple stmts.
1839         (vect_analyze_slp_instance): Update use accordingly.
1840         * tree-vect-stmts.c (read_vector_array, write_vector_array)
1841         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1842         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1843         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1844         (vect_gen_widened_results_half, vect_get_loop_based_defs)
1845         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1846         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1847         down to subroutines.
1849 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1851         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1852         of the worklist from a vector of gimple stmts to a vector of
1853         stmt_vec_infos.
1854         * tree-vect-stmts.c (vect_mark_relevant, process_use)
1855         (vect_mark_stmts_to_be_vectorized): Likewise
1857 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1859         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1860         statement before passing it to vect_analyze_stmt.
1861         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1862         the chain of phi vector definitions.  Track the exit phi via its
1863         stmt_vec_info.
1864         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1865         STMT_VINFO_REDUC_DEF.
1866         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1867         stmt_vec_infos to handle the statement chains.
1868         (vect_get_slp_defs): Record the first statement in the node
1869         using a stmt_vec_info.
1870         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1871         statements here and pass their stmt_vec_info down to subroutines.
1872         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1873         down to vect_finish_stmt_generation.
1874         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1875         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1876         stmt_vec_infos to subroutines.
1877         (vect_remove_stores): Use stmt_vec_infos to handle the statement
1878         chains.
1880 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1882         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1883         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1884         (vect_permute_store_chain, vect_setup_realignment)
1885         (vect_permute_load_chain, vect_shift_permute_load_chain)
1886         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1887         stmts internally, and when passing values to other vectorizer routines.
1888         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1889         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
1890         (vect_analyze_loop_operations, get_initial_def_for_reduction)
1891         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
1892         (vectorizable_reduction, vectorizable_induction)
1893         (vectorizable_live_operation, vect_transform_loop_stmt)
1894         (vect_transform_loop): Likewise.
1895         * tree-vect-patterns.c (vect_reassociating_reduction_p)
1896         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
1897         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
1898         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1899         (vect_slp_analyze_node_operations_1): Likewise.
1900         * tree-vect-stmts.c (vect_mark_relevant, process_use)
1901         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
1902         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
1903         (vect_finish_stmt_generation_1, get_group_load_store_type)
1904         (get_load_store_type, vect_build_gather_load_calls)
1905         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1906         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1907         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1908         (vectorizable_store, vectorizable_load, vectorizable_condition)
1909         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1910         (supportable_widening_operation): Likewise.
1911         (vect_get_vector_types_for_stmt): Likewise.
1912         * tree-vectorizer.h (vect_dr_behavior): Likewise.
1914 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1916         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1917         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1918         (vect_permute_store_chain, vect_permute_load_chain)
1919         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1920         repeated stmt_vec_info lookups.
1921         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1922         (vect_update_ivs_after_vectorizer): Likewise.
1923         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
1924         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1925         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
1926         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1927         (vect_bb_slp_scalar_cost): Likewise.
1928         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
1930 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1932         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1933         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
1934         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
1935         to get gassigns and gcalls, rather than operating on generc gimple
1936         stmts.
1937         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
1938         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
1939         (vectorizable_load, vect_analyze_stmt): Likewise.
1940         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
1942 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1944         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
1945         return stmt_vec_infos rather than gimple stmts.  Do not accept
1946         null arguments.
1947         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
1948         of a gimple stmt.
1949         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1950         Update use of get_later_stmt.
1951         (vect_get_constant_vectors): Update call accordingly.
1952         (vect_schedule_slp_instance): Likewise
1953         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
1954         (vect_slp_analyze_instance_dependence): Likewise.
1955         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
1957 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1959         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
1960         (stmt_info_for_cost::stmt_info): ...this new field.
1961         (add_stmt_costs): Update accordingly.
1962         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1963         (vect_get_known_peeling_cost): Likewise.
1964         (vect_estimate_min_profitable_iters): Likewise.
1965         * tree-vect-stmts.c (record_stmt_cost): Likewise.
1967 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1969         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
1970         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1971         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1972         accordingly.
1973         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
1975 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1977         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
1978         a gimple stmt to a stmt_vec_info.
1979         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1981 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1983         * tree-vectorizer.h (vec_info::grouped_stores): Change from
1984         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1985         (_loop_vec_info::reduction_chains): Likewise.
1986         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
1987         accordingly.
1988         * tree-vect-slp.c (vect_analyze_slp): Likewise.
1990 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
1992         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
1993         a gimple stmt to a stmt_vec_info.
1994         (_stmt_vec_info::next_element): Likewise.
1995         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
1996         (vect_slp_analyze_and_verify_node_alignment)
1997         (vect_analyze_group_access_1, vect_analyze_group_access)
1998         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
1999         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2000         (vect_supportable_dr_alignment): Update accordingly.
2001         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2002         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2003         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2004         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2005         * tree-vect-slp.c (vect_build_slp_tree_1)
2006         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2007         (vect_split_slp_store_group, vect_analyze_slp_instance)
2008         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2009         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2010         (get_group_load_store_type, get_load_store_type)
2011         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2012         (vect_transform_stmt, vect_remove_stores): Likewise.
2014 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2016         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2017         than a gimple stmt.
2018         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2019         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2020         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2021         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2022         (vector_alignment_reachable_p, vect_get_data_access_cost)
2023         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2024         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2025         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2026         (vect_analyze_group_access_1, vect_analyze_group_access)
2027         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2028         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2029         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2030         result of vect_dr_stmt and use the stmt_vec_info instead of
2031         the associated gimple stmt.
2032         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2033         (vect_gen_prolog_loop_niters): Likewise.
2034         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2036 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2038         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2039         to a vec<stmt_vec_info>.
2040         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2041         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2042         vec<stmt_vec_info>.
2043         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2044         to a vec<stmt_vec_info>.
2045         (bst_traits::value_type, bst_traits::value_type): Likewise.
2046         (bst_traits::hash): Update accordingly.
2047         (vect_get_and_check_slp_defs): Change the stmts parameter from
2048         a vec<gimple *> to a vec<stmt_vec_info>.
2049         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2050         (vect_build_slp_tree): Likewise.
2051         (vect_build_slp_tree_2): Likewise.  Update uses of
2052         SLP_TREE_SCALAR_STMTS.
2053         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2054         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2055         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2056         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2057         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2058         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2059         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2060         (vect_get_constant_vectors, vect_get_slp_defs)
2061         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2062         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2063         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2064         instead of gimple stmts.
2065         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2066         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2067         (vect_slp_analyze_instance_dependence): Update uses of
2068         SLP_TREE_SCALAR_STMTS.
2069         (vect_slp_analyze_and_verify_node_alignment): Likewise.
2070         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2071         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2072         (get_initial_defs_for_reduction): Likewise.
2073         (vect_create_epilog_for_reduction): Likewise.
2074         (vectorize_fold_left_reduction): Likewise.
2075         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2076         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2077         (can_vectorize_live_stmts): Likewise.
2079 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2081         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2082         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2083         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2084         than gimple stmts.
2085         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2086         of a gimple stmt.
2087         (gather_scalar_reductions): Update after above interface changes.
2088         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2089         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2090         than gimple stmts.
2091         (vect_force_simple_reduction): Likewise.
2092         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2093         LOOP_VINFO_REDUCTIONS.
2094         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2096 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2098         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2099         a gimple stmt to a stmt_vec_info.
2100         * tree-vect-loop.c (vect_active_double_reduction_p)
2101         (vect_force_simple_reduction, vectorizable_reduction): Update
2102         accordingly.
2104 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2106         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2107         vec<gimple *> to a vec<stmt_vec_info>.
2108         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2109         the reduction_phis argument from a vec<gimple *> to a
2110         vec<stmt_vec_info>.
2111         (vectorizable_reduction): Likewise the phis local variable that
2112         is passed to vect_create_epilog_for_reduction.  Update for new type
2113         of SLP_TREE_VEC_STMTS.
2114         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2115         (vectorizable_live_operation): Likewise.
2116         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2117         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2119 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2121         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2122         a gimple stmt to a stmt_vec_info.
2123         (vectorizable_condition, vectorizable_live_operation)
2124         (vectorizable_reduction, vectorizable_induction): Pass back the
2125         vectorized statement as a stmt_vec_info.
2126         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2127         use of STMT_VINFO_VEC_STMT.
2128         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2129         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2130         as stmt_vec_infos rather than gimple stmts.
2131         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2132         to a stmt_vec_info.
2133         (vectorizable_live_operation): Likewise.
2134         (vectorizable_reduction, vectorizable_induction): Likewise,
2135         updating use of STMT_VINFO_VEC_STMT.
2136         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2137         of STMT_VINFO_VEC_STMT.
2138         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2139         (vectorizable_simd_clone_call, vectorizable_conversion)
2140         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2141         (vectorizable_store, vectorizable_load, vectorizable_condition)
2142         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2143         from a gimple stmt to a stmt_vec_info.
2144         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
2145         pointer to a stmt_vec_info to the vectorizable_* routines.
2147 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2149         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2150         a gimple stmt to a stmt_vec_info.
2151         (is_pattern_stmt_p): Update accordingly.
2152         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2153         (vect_record_grouped_load_vectors): Likewise.
2154         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2155         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2156         (vect_model_reduction_cost): Likewise.
2157         (vect_create_epilog_for_reduction): Likewise.
2158         (vectorizable_reduction, vectorizable_induction): Likewise.
2159         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2160         Return the stmt_vec_info for the pattern statement.
2161         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2162         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2163         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2164         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2165         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2166         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2167         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2168         (free_stmt_vec_info, vect_is_simple_use): Likewise.
2170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2172         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2173         (vect_finish_stmt_generation): Likewise.
2174         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2175         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2176         (vect_build_gather_load_calls): Use the return value of the above
2177         functions instead of a separate call to vinfo_for_stmt.  Use narrow
2178         scopes for the input gimple stmt and wider scopes for the associated
2179         stmt_vec_info.  Use vec_info::lookup_def when setting these
2180         stmt_vec_infos from an SSA_NAME definition.
2181         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2182         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2183         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2184         (vectorizable_store, vectorizable_load, vectorizable_condition)
2185         (vectorizable_comparison): Likewise.
2186         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2187         (vectorizable_reduction): Likewise.
2189 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2191         * tree-vectorizer.h (vect_is_simple_use): Add an optional
2192         stmt_vec_info * parameter before the optional gimple **.
2193         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2194         (process_use, vect_get_vec_def_for_operand_1): Update callers.
2195         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2196         * tree-vect-loop.c (vectorizable_reduction): Likewise.
2197         (vectorizable_live_operation): Likewise.
2198         * tree-vect-patterns.c (type_conversion_p): Likewise.
2199         (vect_look_through_possible_promotion): Likewise.
2200         (vect_recog_rotate_pattern): Likewise.
2201         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2203 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2205         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2206         a typedef to a wrapper class.
2207         (NULL_STMT_VEC_INFO): New macro.
2208         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2209         (stmt_vec_info::operator*): New function.
2210         (stmt_vec_info::operator gimple *): Likewise.
2211         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2212         (add_stmt_costs): Likewise.
2213         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2214         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2215         (vect_get_known_peeling_cost): Likewise.
2216         (vect_estimate_min_profitable_iters): Likewise.
2217         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2218         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2219         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2220         (vectorizable_store, free_stmt_vec_infos): Likewise.
2221         (new_stmt_vec_info): Change return type of xcalloc to
2222         _stmt_vec_info *.
2224 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2226         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2227         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2228         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2229         a single_imm_use-based sequence.
2230         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2232 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2234         * tree-vectorizer.h (vec_info::lookup_def): Declare.
2235         * tree-vectorizer.c (vec_info::lookup_def): New function.
2236         * tree-vect-patterns.c (vect_get_internal_def): Use it.
2237         (vect_widened_op_tree): Likewise.
2238         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2239         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2240         (vectorizable_reduction): Likewise.
2241         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2242         of a gimple *.
2243         (vect_is_slp_reduction): Update calls accordingly.  Use
2244         vec_info::lookup_def.
2245         (vect_is_simple_reduction): Likewise
2246         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2248 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2250         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2251         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2252         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2253         of vinfo_for_stmt.
2254         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2255         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2256         (vect_update_vf_for_slp, vect_analyze_loop_operations)
2257         (vect_is_slp_reduction, vectorizable_induction)
2258         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2259         * tree-vect-patterns.c (vect_init_pattern_stmt):
2260         (vect_determine_min_output_precision_1, vect_determine_precisions)
2261         (vect_pattern_recog): Likewise.
2262         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2263         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2264         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2265         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2266         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2267         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
2268         info field from a loop to a loop_vec_info.
2270 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2272         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2273         (vec_info::add_stmt): Declare.
2274         * tree-vectorizer.c (vec_info::add_stmt): New function.
2275         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2276         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2277         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2278         (vectorizable_induction): Likewise.
2279         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2280         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2281         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2282         (vectorizable_load): Likewise.
2283         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2284         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2285         (vect_recog_gather_scatter_pattern): Likewise.
2286         (append_pattern_def_seq): Likewise.  Remove a check that is
2287         performed by add_stmt itself.
2289 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2291         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2292         which make_ssa_name was called with new_stmt before new_stmt
2293         had been created.
2295 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2297         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2298         split out from...
2299         (vect_is_slp_reduction): ...here...
2300         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
2301         that are already known to be false.
2303 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2305         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2306         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
2307         STMT_VINFO_NUM_SLP_USES when it's true.
2308         (vect_free_slp_instance): Add a final_p parameter and pass it to
2309         vect_free_slp_tree.
2310         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2311         (vect_analyze_slp_instance): Likewise.
2312         (vect_slp_analyze_operations): Likewise.
2313         (vect_slp_analyze_bb_1): Likewise.
2314         * tree-vectorizer.c (vec_info): Likewise.
2315         * tree-vect-loop.c (vect_transform_loop): Likewise.
2317 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2319         * tree-vect-loop.c (vectorizable_reduction): Assert that the
2320         function is not called for second and subsequent members of
2321         a reduction group.
2323 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
2325         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2326         cases for nested loops from here to ...
2327         (vect_create_epilog_for_reduction): ...here.  Only call
2328         vect_is_simple_use for inner-loop reductions.
2330 2018-07-31  Martin Liska  <mliska@suse.cz>
2332         PR gcov-profile/85338
2333         PR gcov-profile/85350
2334         PR gcov-profile/85372
2335         * profile.c (struct location_triplet): New.
2336         (struct location_triplet_hash): Likewise.
2337         (output_location): Do not output a BB that
2338         is already recorded for a line.
2339         (branch_prob): Use streamed_locations.
2341 2018-07-31  Martin Liska  <mliska@suse.cz>
2343         PR gcov-profile/85370
2344         * coverage.c (coverage_begin_function): Do not mark target
2345         clones as artificial functions.
2347 2018-07-31  Martin Liska  <mliska@suse.cz>
2349         PR gcov-profile/83813
2350         PR gcov-profile/84758
2351         PR gcov-profile/85217
2352         PR gcov-profile/85332
2353         * profile.c (branch_prob): Do not record GOTO expressions
2354         for GIMPLE statements which locations are already streamed.
2356 2018-07-31  Olivier Hainque  <hainque@adacore.com>
2358         * gcc.c (handle_spec_function): Accept a soft_matched_part
2359         argument, as do_spec_1.  Pass it down to ...
2360         (eval_spec_function): Accept a soft_matched_part argument,
2361         and pass it down to ...
2362         (do_spec_2): Accept a soft_matched_part argument, and pass
2363         it down to do_spec_1.
2364         (do_spec_1): Pass soft_matched_part to handle_spec_function.
2365         (handle_braces): Update call to handle_spec_function.
2366         (driver::set_up_specs): Update calls to do_spec_2.
2367         (compare_debug_dump_opt_spec_function): Likewise.
2368         (compare_debug_self_opt_spec_function): Likewise.
2370 2018-07-31  Olivier Hainque  <hainque@adacore.com>
2372         * common.opt (nolibc): New option.
2373         * doc/invoke.texi (Link Options): Document it.
2374         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2375         * config/alpha/linux.h: Likewise.
2376         * config/arc/elf.h: Likewise.
2377         * config/arm/uclinux-elf.h: Likewise.
2378         * config/arm/unknown-elf.h: Likewise.
2379         * config/avr/avrlibc.h: Likewise.
2380         * config/bfin/bfin.h: Likewise.
2381         * config/bfin/linux.h: Likewise.
2382         * config/bfin/uclinux.h: Likewise.
2383         * config/darwin.h: Likewise.
2384         * config/darwin10.h: Likewise.
2385         * config/darwin12.h: Likewise.
2386         * config/gnu-user.h: Likewise.
2387         * config/lm32/uclinux-elf.h: Likewise.
2388         * config/pa/pa-hpux11.h: Likewise.
2389         * config/pa/pa64-hpux.h: Likewise.
2390         * config/sparc/sparc.h: Likewise.
2392 2018-07-31  Olivier Hainque  <hainque@adacore.com>
2394         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2395         undefined variables.
2397 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
2399         PR target/86640
2400         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2401         instead of GEN_INT.
2403 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2405         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2406         terminated string literal.
2408 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
2410         PR rtl-optimization/85160
2411         * combine.c (is_just_move): New function.
2412         (try_combine): Allow combining two instructions into two if neither of
2413         the original instructions was a move.
2415 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
2417         PR target/86673
2418         * doc/extend.texi (Global Register Variables): Discourage use of type
2419         qualifiers.
2420         (Local Register Variables): Likewise.
2422 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
2424         PR tree-optimization/86506
2425         * hwint.h (ceil_log2): Resync with hwint.c implementation.
2427 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
2429         PR target/86547
2430         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2431         hard_regno, make sure no insn between `from` and `to` clobbers it.
2433 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
2434             Tom de Vries  <tdevries@suse.de>
2436         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2437         (PTX_DEFAULT_RUNTIME_DIM): ... this.
2438         (nvptx_goacc_validate_dims): Set default worker and gang dims to
2439         PTX_DEFAULT_RUNTIME_DIM.
2440         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2442 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
2444         * config/pa/pa.c (pa_output_addr_vec): Align address table.
2445         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2446         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2448 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
2450         * config/rs6000/constraints.md (wG constraint): Delete, no longer
2451         used.
2452         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2453         predicate to reflect toc fusion has been deleted.
2454         (toc_fusion_mem_raw): Delete, no longer used.
2455         (toc_fusion_mem_wrapped): Likewise.
2456         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2457         fusion mask bit.
2458         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2459         Delete, no longer used.
2460         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2461         meant to be used for toc fusion.
2462         (rs6000_debug_print_mode): Delete toc fusion debugging.
2463         (rs6000_debug_reg_global): Likewise.
2464         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2465         fusion and secondary reload support that were never used.
2466         (rs6000_option_override_internal): Delete TOC fusion, that was only
2467         partially defined, and it did not work unless you also used the
2468         -mcmodel= switch.
2469         (rs6000_legitimate_address_p): Delete TOC fusion support.
2470         (rs6000_opt_masks): Likewise.
2471         (fusion_wrap_memory_address): Delete function, no longer used.
2472         (fusion_split_address); Delete TOC fusion support.
2473         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2474         longer used with toc fusion being deleted.
2475         (TARGET_TOC_FUSION_FP): Likewise.
2476         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2477         UNSPEC.
2478         (toc fusion spliter): Delete TOC fusion support.
2479         (toc_fusionload_<mode>): Likewise.
2480         (toc_fusionload_di): Likewise.
2481         (fusion_gpr_load_<mode>): Delete generator function, this insn no
2482         longer needs to be named.  Rename predicate to delete TOC fusion.
2483         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2484         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2485         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2486         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2487         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2489 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2491         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2492         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
2493         __int128 in built-in function prototypes.
2494         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2495         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2497 2018-07-27  Martin Sebor  <msebor@redhat.com>
2499         PR tree-optimization/86696
2500         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2501         types, including enums.
2502         (handle_char_store): Be prepared for the above function to fail.
2504 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
2506         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2507         when optimization level is lower than 2 or optimize for size.
2509 2018-07-26  Martin Sebor  <msebor@redhat.com>
2511         PR tree-optimization/86043
2512         PR tree-optimization/86042
2513         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2514         (get_string_cst_length): Rename...
2515         (get_min_string_length): ...to this.  Add argument.
2516         (handle_char_store): Extend to handle multi-character stores by
2517         MEM_REF.
2518         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
2519         * tree.h (initializer_zerop): Add argument.
2521 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
2523         PR middle-end/86660
2524         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2525         declare target to variables if they have always,{to,from,tofrom} map
2526         kinds.
2528 2018-07-26  Martin Liska  <mliska@suse.cz>
2530         PR lto/86548
2531         * lto-wrapper.c: Add linker_output as prefix
2532         for ltrans_output_file.
2534 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
2536         PR rtl-optimization/85805
2537         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2538         value for hard registers if that was written in the same mode.
2540 2018-07-26  Martin Liska  <mliska@suse.cz>
2542         PR gcov-profile/86536
2543         * gcov.c (format_gcov): Use printf format %.*f directly
2544         and do not handle special values.
2546 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
2548         * common/config/arc/arc-common.c (arc_option_optimization_table):
2549         Update default optimizations for size.
2551 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
2553         * config/arc/arc.md (movsf_insn): Add short instruction selection.
2554         * config/arc/constraints.md (CfZ): New constraint.
2555         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2556         (subsf3_fpu): Likewise.
2557         (cmpsf_fpu): Likewise.
2558         (cmpsf_fpu_uneq): Likewise.
2560 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
2562         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2563         accesses as well.
2564         (arc_is_uncached_mem_p): uncached applies to both the variable and
2565         the pointer.
2567 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
2569         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2570         register names.
2572 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
2574         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2575         field "m_scopes" from vec to auto_vec.
2577 2018-07-25  Martin Liska  <mliska@suse.cz>
2579         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2580         return type.
2582 2018-07-25  Richard Biener  <rguenther@suse.de>
2584         PR debug/86654
2585         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2586         special wrt context_die late.
2587         (gen_subprogram_die): Re-use DIEs in local scope.
2589 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
2591         PR tree-optimization/86644
2592         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
2594 2018-07-25  Martin Liska  <mliska@suse.cz>
2596         PR middle-end/86645
2597         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2598         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2600 2018-07-25  Martin Liska  <mliska@suse.cz>
2602         PR sanitizer/79635
2603         * params.def: Explain ASan abbreviation and provide
2604         a documentation link.
2606 2018-07-24  Martin Sebor  <msebor@redhat.com>
2608         PR tree-optimization/86622
2609         PR tree-optimization/86532
2610         * builtins.h (string_length): Declare.
2611         * builtins.c (c_strlen): Correct handling of non-constant offsets.
2612         (check_access): Be prepared for non-constant length ranges.
2613         (string_length): Make extern.
2614         * expr.c (string_constant): Only handle the minor non-constant
2615         array index.  Use string_constant to compute the length of
2616         a generic string constant.
2618 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
2620         PR tree-optimization/86618
2621         * tree-vect-stmts.c (vectorizable_call): Don't take the address
2622         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2624 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
2626         PR tree-optimization/86636
2627         * json.cc (json::object::set): Fix comment.  Add assertions.
2628         (json::array::append): Move here from json.h.  Add comment and an
2629         assertion.
2630         (json::string::string): Likewise.
2631         * json.h (json::array::append): Move to json.cc.
2632         (json::string::string): Likewise.
2633         * optinfo-emit-json.cc
2634         (optrecord_json_writer::impl_location_to_json): Assert that we
2635         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2636         wrapper around it.  Expand the location once, rather than three
2637         times.
2638         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2639         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2640         wrappers.
2641         (optrecord_json_writer::optinfo_to_json): Likewise, in four
2642         places.  Fix some overlong lines.
2644 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
2646         * config/aarch64/aarch64-simd.md
2647         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2648         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2649         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2650         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2651         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2652         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2653         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2654         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2655         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2657 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
2659         PR middle-end/86627
2660         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2661         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
2662         and abs_d == d, do the power of two handling if profitable.
2664 2018-07-24  Richard Biener  <rguenther@suse.de>
2666         * match.pd: Add BIT_FIELD_REF canonicalizations.
2668 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2670         PR c/86617
2671         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2673 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2675         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2676         terminated STRING_CST object.
2678 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2680         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2682 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
2684         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2685         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2686         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2687         the elements into a register.
2688         (rs6000_split_v4si_init_di_reg): Delete.
2689         (rs6000_split_v4si_init): Delete.
2690         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2691         (vsx_init_v4si): Rewrite as a define_expand.
2693 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
2695         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2696         zero_extend argument from memory): New.
2698 2018-07-22  Martin Sebor  <msebor@redhat.com>
2700         PR bootstrap/86621
2701         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2702         diagnosing calls with unknown arguments unless -Walloca-larger-than
2703         is restricted to less than PTRDIFF_MAX bytes.
2705 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
2707         * doc/gcov.texi (Invoking Gcov): Editorial changes.
2709 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
2711         * pretty-print.c (text_info::set_location): Remove redundant
2712         "line_table" parameter from call to rich_location::set_range.
2714 2018-07-20  Martin Sebor  <msebor@redhat.com>
2716         PR middle-end/82063
2717         * builtins.c (expand_builtin_alloca): Adjust.
2718         * calls.c (alloc_max_size): Simplify.
2719         * cgraphunit.c (cgraph_node::expand): Adjust.
2720         * common.opt (larger_than_size, warn_frame_larger_than): Remove
2721         variables.
2722         (frame_larger_than_size): Same.
2723         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2724         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2725         Initialize.
2726         * doc/invoke.texi (GCC Command Options): Document option arguments.
2727         Explain byte-size arguments and suffixes.
2728         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2729         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2730         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2731         * doc/options.texi (UInteger): Expand.
2732         (Host_Wide_Int, ByteSize): Document new properties.
2733         * final.c (final_start_function_1): Include sizes in an error message.
2734         * function.c (frame_offset_overflow): Same.
2735         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2736         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2737         Diagnose unbounded alloca calls only for limits of less than
2738         PTRDIFF_MAX.
2739         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
2740         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
2741         for alloca(0).
2742         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
2743         only for limits of less than PTRDIFF_MAX.
2744         * langhooks-def.h (lhd_handle_option): Change function argument
2745         to HOST_WIDE_INT.
2746         * langhooks.c (lhd_handle_option): Same.
2747         * langhooks.h (handle_option): Same.
2748         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2749         ByteSize flags.
2750         (var_type, var_type_struct): Same.
2751         (var_set): Handle ByteSize flag.
2752         * optc-gen.awk: Add comments to output to ease debugging.  Make
2753         use of HOST_WIDE_INT where appropriate.
2754         * opts-gen-save.awk:  Use %lx to format unsigned long.
2755         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2756         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2757         arguments.  Parse bytes-size suffixes.
2758         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2759         (enum_value_to_arg): Same.
2760         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
2761         (handle_option): Adjust.
2762         (generate_option): Change function argument to HOST_WIDE_INT.
2763         (cmdline_handle_error): Adjust.
2764         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2765         (set_option): Change function argument to HOST_WIDE_INT.
2766         (option_enabled): Handle cl_host_wide_int.
2767         (get_option_state): Handle CLVC_SIZE.
2768         (control_warning_option): Same.
2769         * opts.c (common_handle_option): Change function argument to
2770         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
2771         OPT_Wvla_larger_than_.
2772         * opts.h (enum cl_var_type): Add an enumerator.
2773         * stor-layout.c (layout_decl): Print a more meaningful warning.
2774         * toplev.c (output_stack_usage): Adjust.
2776 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
2778         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2779         call to inline_expand_builtin_string_cmp.
2780         (expand_builtin_strcmp): Likewise.
2781         (expand_builtin_strncmp): Likewise.
2782         (inline_string_cmp): Delete the last parameter, change char_type_node
2783         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2784         two operands.
2785         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2786         the inlining expansion on target where the type of the call has same or
2787         narrower precision than unsigned char.
2789 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
2791         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2792         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2793         * common.opt (fsave-optimization-record): New option.
2794         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2795         * doc/invoke.texi (-fsave-optimization-record): New option.
2796         * dumpfile.c: Include "optinfo-emit-json.h".
2797         (struct kv_pair): Move to coretypes.h.
2798         (optgroup_options): Make non-static.
2799         (dump_context::end_scope): Call
2800         optimization_records_maybe_pop_dump_scope.
2801         * dumpfile.h (optgroup_options): New decl.
2802         * json.cc: New file.
2803         * json.h: New file.
2804         * optinfo-emit-json.cc: New file.
2805         * optinfo-emit-json.h: New file.
2806         * optinfo.cc: Include "optinfo-emit-json.h".
2807         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2808         (optinfo_enabled_p): Check optimization_records_enabled_p.
2809         (optinfo_wants_inlining_info_p): Likewise.
2810         * optinfo.h: Update comment.
2811         * profile-count.c (profile_quality_as_string): New function.
2812         * profile-count.h (profile_quality_as_string): New decl.
2813         (profile_count::quality): New accessor.
2814         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2815         and optinfo_emit_json_cc_tests.
2816         * selftest.h (selftest::json_cc_tests): New decl.
2817         (selftest::optinfo_emit_json_cc_tests): New decl.
2818         * toplev.c: Include "optinfo-emit-json.h".
2819         (compile_file): Call optimization_records_finish.
2820         (do_compile): Call optimization_records_start.
2821         * tree-ssa-live.c: Include optinfo.h.
2822         (remove_unused_scope_block_p): Retain inlining information if
2823         optinfo_wants_inlining_info_p returns true.
2825 2018-07-20  Richard Biener  <rguenther@suse.de>
2827         PR debug/86585
2828         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2829         to cover -flto-partition=none.
2831 2018-07-20  Martin Liska  <mliska@suse.cz>
2833         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2834         (get_decl_source_range): Remove unused function.
2836 2018-07-20  Richard Biener  <rguenther@suse.de>
2838         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2839         (struct vn_phi_s): Likewise.
2840         (struct vn_reference_s): Likewise.
2841         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2842         for searching the slot of an entry known to be in the hash itself.
2843         (vn_phi_hasher::equal): Likewise.
2844         (vn_reference_hasher::equal): Likewise.
2845         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2846         globals.
2847         (optimistic_info, current_info): Remove, keeping only valid_info.
2848         (vn_reference_lookup_1): Remove fallback lookup.
2849         (vn_reference_lookup_2): Likewise.
2850         (vn_nary_op_lookup_1): Likewise.
2851         (vn_phi_lookup): Likewise.
2852         (vn_nary_build_or_lookup_1): Make sure to not chain the built
2853         hash element.
2854         (vn_reference_insert): Adjust, chain the inserted hash element
2855         at last_inserted_ref.
2856         (vn_reference_insert_pieces): Likewise.
2857         (visit_reference_op_call): Likewise.
2858         (vn_nary_op_insert_into): Chain the inserted hash element at
2859         last_inserted_nary.
2860         (vn_nary_op_insert_pieces): Adjust.
2861         (vn_nary_op_insert): Likewise.
2862         (vn_nary_op_insert_stmt): Likewise.
2863         (vn_phi_insert): Adjust, chain the inserted hash element at
2864         last_inserted_phi.
2865         (process_scc): Remove clearing and copying the optimistic
2866         table.  Instead remove elements inserted during an optimistic
2867         iteration from the single table we maintain.
2868         (init_scc_vn): Adjust.
2869         (free_scc_vn): Likewise.
2870         (sccvn_dom_walker::record_cond): Likewise.
2871         (sccvn_dom_walker::after_dom_children): Likewise.
2873 2018-07-19  Martin Sebor  <msebor@redhat.com>
2875         PR tree-optimization/84047
2876         PR tree-optimization/83776
2877         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2878         (check_array_bounds): Call it.
2880 2018-07-19  Martin Sebor  <msebor@redhat.com>
2882         * align.h (align_flags): Use member initialization.
2884 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
2886         * Makefile.in (OBJS): Add optinfo.o.
2887         * coretypes.h (class symtab_node): New forward decl.
2888         (struct cgraph_node): New forward decl.
2889         (class varpool_node): New forward decl.
2890         * dump-context.h: New file.
2891         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
2892         "tree-pass.h".
2893         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
2894         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
2895         (set_alt_dump_file): Likewise.
2896         (dump_context::~dump_context): New dtor.
2897         (dump_gimple_stmt): Move implementation to...
2898         (dump_context::dump_gimple_stmt): ...this new member function.
2899         Add the stmt to any pending optinfo, creating one if need be.
2900         (dump_gimple_stmt_loc): Move implementation to...
2901         (dump_context::dump_gimple_stmt_loc): ...this new member function.
2902         Start a new optinfo and add the stmt to it.
2903         (dump_gimple_expr): Move implementation to...
2904         (dump_context::dump_gimple_expr): ...this new member function.
2905         Add the stmt to any pending optinfo, creating one if need be.
2906         (dump_gimple_expr_loc): Move implementation to...
2907         (dump_context::dump_gimple_expr_loc): ...this new member function.
2908         Start a new optinfo and add the stmt to it.
2909         (dump_generic_expr): Move implementation to...
2910         (dump_context::dump_generic_expr): ...this new member function.
2911         Add the tree to any pending optinfo, creating one if need be.
2912         (dump_generic_expr_loc): Move implementation to...
2913         (dump_context::dump_generic_expr_loc): ...this new member
2914         function.  Add the tree to any pending optinfo, creating one if
2915         need be.
2916         (dump_printf): Move implementation to...
2917         (dump_context::dump_printf_va): ...this new member function.  Add
2918         the text to any pending optinfo, creating one if need be.
2919         (dump_printf_loc): Move implementation to...
2920         (dump_context::dump_printf_loc_va): ...this new member function.
2921         Start a new optinfo and add the stmt to it.
2922         (dump_dec): Move implementation to...
2923         (dump_context::dump_dec): ...this new member function.  Add the
2924         value to any pending optinfo, creating one if need be.
2925         (dump_context::dump_symtab_node): New member function.
2926         (dump_context::get_scope_depth): New member function.
2927         (dump_context::begin_scope): New member function.
2928         (dump_context::end_scope): New member function.
2929         (dump_context::ensure_pending_optinfo): New member function.
2930         (dump_context::begin_next_optinfo): New member function.
2931         (dump_context::end_any_optinfo): New member function.
2932         (dump_context::s_current): New global.
2933         (dump_context::s_default): New global.
2934         (dump_scope_depth): Delete global.
2935         (dumpfile_ensure_any_optinfo_are_flushed): New function.
2936         (dump_symtab_node): New function.
2937         (get_dump_scope_depth): Reimplement in terms of dump_context.
2938         (dump_begin_scope): Likewise.
2939         (dump_end_scope): Likewise.
2940         (selftest::temp_dump_context::temp_dump_context): New ctor.
2941         (selftest::temp_dump_context::~temp_dump_context): New dtor.
2942         (selftest::verify_item): New function.
2943         (ASSERT_IS_TEXT): New macro.
2944         (ASSERT_IS_TREE): New macro.
2945         (ASSERT_IS_GIMPLE): New macro.
2946         (selftest::test_capture_of_dump_calls): New test.
2947         (selftest::dumpfile_c_tests): Call it.
2948         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
2949         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
2950         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
2951         descriptive comment.
2952         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
2953         (dump_node, dump_bb): Move these unrelated decls.
2954         (class dump_manager): Add leading comment.
2955         * optinfo.cc: New file.
2956         * optinfo.h: New file.
2958 2018-07-19  Michael Collison  <michael.collison@arm.com>
2959             Richard Henderson <rth@redhat.com>
2961         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
2962         (subti): Handle op1 zero.
2963         (subvti4, usub4ti4): New.
2964         (*sub<GPI>3_compare1_imm): New.
2965         (sub<GPI>3_carryinCV): New.
2966         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
2967         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
2969 2018-07-19  Michael Collison  <michael.collison@arm.com>
2970             Richard Henderson <rth@redhat.com>
2972         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
2973         (addti3): Create simpler code if low part is already known to be 0.
2974         (addvti4, uaddvti4): New.
2975         (*add<GPI>3_compareC_cconly_imm): New.
2976         (*add<GPI>3_compareC_cconly): New.
2977         (*add<GPI>3_compareC_imm): New.
2978         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
2979         handle constants within this pattern..
2980         (*add<GPI>3_compareV_cconly_imm): New.
2981         (*add<GPI>3_compareV_cconly): New.
2982         (*add<GPI>3_compareV_imm): New.
2983         (add<GPI>3_compareV): New.
2984         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
2985         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
2986         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
2987         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
2988         with 'comparison' operator.
2989         (*add<GPI>3_compareV_cconly_imm): Ditto.
2990         (*add<GPI>3_compareV_cconly): Ditto.
2991         (*add<GPI>3_compareV_imm): Ditto.
2992         (add<GPI>3_compareV): Ditto.
2993         (add<mode>3_carryinC): Ditto.
2994         (*add<mode>3_carryinC_zero): Ditto.
2995         (*add<mode>3_carryinC): Ditto.
2996         (add<mode>3_carryinV): Ditto.
2997         (*add<mode>3_carryinV_zero): Ditto.
2998         (*add<mode>3_carryinV): Ditto.
3000 2018-07-19  Michael Collison  <michael.collison@arm.com>
3001             Richard Henderson <rth@redhat.com>
3003         * config/aarch64/aarch64-modes.def (CC_V): New.
3004         * config/aarch64/aarch64-protos.h
3005         (aarch64_addti_scratch_regs): Declare
3006         (aarch64_subvti_scratch_regs): Declare.
3007         (aarch64_expand_subvti): Declare.
3008         (aarch64_gen_unlikely_cbranch): Declare
3009         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3010         for signed overflow using CC_Vmode.
3011         (aarch64_get_condition_code_1): Handle CC_Vmode.
3012         (aarch64_gen_unlikely_cbranch): New function.
3013         (aarch64_addti_scratch_regs): New function.
3014         (aarch64_subvti_scratch_regs): New function.
3015         (aarch64_expand_subvti): New function.
3017 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3019         * config/aarch64/aarch64-option-extensions.def: New entry for profile
3020         extension.
3021         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3022         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3023         extension.
3025 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3027         PR target/83009
3028         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3029         address check not strict.
3031 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3033         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3034         Umq with Umn.
3035         (store_pair_lanes<mode>): Likewise.
3036         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3037         enum value 'ADDR_QUERY_LDP_STP_N'.
3038         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3039         (aarch64_print_address_internal): Add declaration.
3040         (aarch64_print_ldpstp_address): Remove.
3041         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3042         (aarch64_print_operand): Change printing of 'y'.
3043         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3044         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3045         'true' rather than '1'.
3046         * config/aarch64/constraints.md (Uml): Likewise.
3047         (Uml): Rename to Umn.
3048         (Umq): Remove.
3050 2018-07-19  Richard Biener  <rguenther@suse.de>
3052         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3053         a trailing array.
3054         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3055         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3056         (vn_reference_hasher): Likewise.
3057         (struct vn_tables_s): Remove obstack and alloc-pool members.
3058         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3059         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3060         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3061         (vn_reference_insert_pieces): Likewise.
3062         (alloc_vn_nary_op_noinit): Adjust.
3063         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3064         (vn_phi_eq): Adjust.
3065         (shared_lookup_phiargs): Remove.
3066         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3067         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3068         (visit_reference_op_call): Likewise.
3069         (copy_nary, copy_phi, copy_reference): Remove.
3070         (process_scc): Rewind the obstack when iterating.  Do not
3071         copy the elements to valid_info but just move them from one
3072         hashtable to the other.
3073         (allocate_vn_table): Adjust.
3074         (free_vn_table): Likewise.
3075         (init_scc_vn): Likewise.
3076         (free_scc_vn): Likewise.
3078 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
3080         PR target/86560
3081         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3082         indirect_return as function type attribute.
3083         (ix86_attribute_table): Change indirect_return to function
3084         type attribute.
3085         * doc/extend.texi: Update indirect_return attribute.
3087 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
3089         * wide-int.h (widest2_int): New.
3090         * gimple-fold.c (arith_overflowed_p): Use it.
3091         * tree.h (widest2_int_cst): New.
3092         * tree-vrp.c (wide_int_binop_overflow): Rename from
3093         vrp_int_const_binop.
3094         Rewrite to work on trees.
3095         (extract_range_from_multiplicative_op_1): Abstract code to...
3096         (wide_int_range_min_max): ...here.
3097         (wide_int_range_cross_product): ...and here.
3098         (extract_range_from_binary_expr_1): Abstract overflow code to...
3099         (wide_int_range_mult_wrapping): ...here.
3100         * tree-vrp.h (wide_int_range_cross_product): New.
3101         (wide_int_range_mult_wrapping): New.
3103 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
3104             Julia Koval  <julia.koval@intel.com>
3106         * config/i386/x86-tune-costs.h (skylake_memcpy,
3107         skylake_memset): Replace rep_prefix with unrolling for size 512.
3109 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
3111         PR middle-end/86544
3112         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3113         comparision with EQ_EXPR in last stmt.
3115 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3117         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3118         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3119         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3120         previously known as "PowerPC AltiVec Built-in Functions".  Move
3121         some material to new subsubsections "PowerPC AltiVec Built-in
3122         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3123         ISA 2.07".
3124         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3125         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3126         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3127         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3129 2018-07-18  Richard Biener  <rguenther@suse.de>
3131         PR tree-optimization/86557
3132         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3133         EXACT_DIV_EXPR.
3135 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
3137         * config/s390/s390.c (s390_function_profiler): Generate CFI.
3139 2018-07-17  Jeff Law  <law@redhat.com>
3141         * config/arm/arm.c (get_label_padding): Update for recent
3142         changes to label_to_alignment.
3144         PR tree-optimization/86010
3145         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3147         * config/mips/mips.c (vr4130_align_insns): Update for recent
3148         changes to label_to_alignment.
3150         * config/frv/frv.c (frv_label_align): Update for recent changes
3151         to label_to_alignment.
3153         * config/nios2/nios2.c (nios2_label_align): Update for recent
3154         changes which dropped ALIGN_LABELS_LOG.
3156 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
3158         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3159         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3161 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
3163         * config/arc/arc.c (arc_label_align): Use align_labels instead of
3164         deprecated align_labels_log.
3166 2018-07-17  Richard Biener  <rguenther@suse.de>
3168         PR lto/86456
3169         * dwarf2out.c (init_sections_and_labels): Always generate
3170         a debug_line_str_section for early LTO debug.
3171         (dwarf2out_finish): Reset debug_line_str_hash output early.
3172         Bump counter for extra dwarf5 .debug_loc labels to not conflict
3173         with fat LTO part.
3174         (dwarf2out_early_finish): Output debug_line_str.
3176 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
3178         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3179         index register on z196 or later.
3181 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
3183         * config/s390/s390.c (s390_default_align): Set default function
3184         alignment to 16.
3185         (s390_override_options_after_change): Call s390_default align.
3186         (s390_option_override_internal): Call s390_default_align.
3187         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3189 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
3191         PR middle-end/86542
3192         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3193         to _looptemp_ clauses, other than the first two.
3195 2018-07-17  Martin Liska  <mliska@suse.cz>
3197         * opts.c: Do not enable OPT_falign_* for -Os.
3199 2018-07-17  Martin Liska  <mliska@suse.cz>
3201         * align.h (MAX_CODE_ALIGN): New.
3202         (MAX_CODE_ALIGN_VALUE): New.
3203         * common/config/i386/i386-common.c (ix86_handle_option):
3204         (MAX_CODE_ALIGN): Moved to align.h.
3205         * final.c (MAX_CODE_ALIGN): Likewise.
3206         * opts.c (parse_and_check_align_values):
3207         (MAX_CODE_ALIGN): Likewise.
3208         (MAX_CODE_ALIGN_VALUE): Likewise.
3210 2018-07-17  Martin Liska  <mliska@suse.cz>
3212         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3213         in order to fulfil coding style.
3214         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3215         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3216         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3217         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3218         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3219         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3220         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3221         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3222         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3224 2018-07-17  Martin Liska  <mliska@suse.cz>
3226         * align.h: New file.
3227         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3228         directly.
3229         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3230         align_flags of label_to_alignment.
3231         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3232         align_flags class.
3233         * config/m68k/m68k.c: Do not use removed align_labels_value and
3234         align_loops_value.
3235         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3236         (LOOP_ALIGN): Likewise.
3237         (LABEL_ALIGN): Likewise.
3238         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3239         Remove not used macro.
3240         (rs6000_loop_align): Change return type to align_flags.
3241         (rs6000_loop_align_max_skip): Remove.
3242         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3243         Change return type to align_flags.
3244         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3245         Remove not used macro.
3246         (rs6000_loop_align):  Change return type to align_flags.
3247         (rs6000_loop_align_max_skip): Remove.
3248         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3249         * config/rx/rx-protos.h (rx_align_for_label): Make it
3250         static function.
3251         * config/rx/rx.c (rx_align_for_label): Change return type
3252         to align_flags.
3253         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3254         macro definitions.
3255         into align_flags class.
3256         (LABEL_ALIGN): Likewise.
3257         (LOOP_ALIGN): Likewise.
3258         * config/s390/s390.c (s390_label_align): Use align_flags
3259         class member.
3260         (s390_asm_output_function_label): Likewise.
3261         * config/sh/sh.c (sh_override_options_after_change):
3262         Use align_flags class directly without macros.
3263         (find_barrier): Likewise.
3264         (barrier_align): Likewise.
3265         (sh_loop_align): Likewise.
3266         * config/spu/spu.c (spu_option_override):
3267         Use align_flags_tuple::get_value instead of removed macros.
3268         (spu_sched_init): Likewise.
3269         * config/spu/spu.h (GTY): Likewise.
3270         * config/visium/visium.c (visium_option_override):
3271         Set "8" as default secondary alignment.
3272         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3273         in order to guarantee secondary alignment of 8.
3274         * coretypes.h: Include align.h header file.
3275         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3276         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3277         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3278         * doc/tm.texi.in: Likewise.
3279         * final.c (struct label_alignment): Remove not used structure.
3280         (LABEL_ALIGN): Change type to align_flags.
3281         (LOOP_ALIGN): Likewise.
3282         (JUMP_ALIGN): Likewise.
3283         (default_loop_align_max_skip): Remove.
3284         (default_label_align_max_skip): Likewise.
3285         (default_jump_align_max_skip): Likewise.
3286         (default_label_align_after_barrier_max_skip):
3287         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3288         (LABEL_TO_MAX_SKIP): Remove.
3289         (label_to_alignment): Return align_flags type instead of integer.
3290         (label_to_max_skip): Remove.
3291         (align_fuzz): Use align_flags type.
3292         (compute_alignments): Use align_flags type and use align_flags::max
3293         to combine multiple alignments.
3294         (grow_label_align): Grow vec instead of C array.
3295         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3296         (shorten_branches):  Use align_flags type and use align_flags::max
3297         to combine multiple alignments.
3298         (final_scan_insn_1): Remove usage of secondary alignment that comes
3299         from label alignment, but instead use proper secondary alignment
3300         which is computed in grow_label_align.
3301         * flags.h (struct align_flags_tuple): Move to align.h.
3302         (struct align_flags): Likewise.
3303         (state_align_loops): Rename to align_loops.
3304         (state_align_jumps): Rename to align_jumps.
3305         (state_align_labels): Rename to align_labels.
3306         (state_align_functions): Rename to align_functions.
3307         (align_loops_log): Remove.
3308         (align_jumps_log): Remove.
3309         (align_labels_log): Remove.
3310         (align_functions_log): Remove.
3311         (align_loops_max_skip): Remove.
3312         (align_jumps_max_skip): Remove.
3313         (align_labels_max_skip): Remove.
3314         (align_functions_max_skip): Remove.
3315         (align_loops_value): Remove.
3316         (align_jumps_value): Remove.
3317         (align_labels_value): Remove.
3318         (align_functions_value): Remove.
3319         * output.h (label_to_alignment): Change return type to align_flags.
3320         (label_to_max_skip): Remove.
3321         * target.def: Remove loop_align_max_skip, label_align_max_skip,
3322         jump_align_max_skip macros.
3323         * targhooks.h (default_loop_align_max_skip): Remove.
3324         (default_label_align_max_skip): Likewise.
3325         (default_jump_align_max_skip): Likewise.
3326         (default_label_align_after_barrier_max_skip): Remove.
3327         * toplev.c (read_log_maxskip): Use ::normalize function.
3328         (parse_N_M): Remove not used argument and also call ::normalize.
3329         (parse_alignment_opts): Do not pass unused arguments.
3330         * varasm.c (assemble_start_function): Use directly align_functions
3331         instead of removed macros.
3332         * system.h: Do not poison removed macros.
3334 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
3336         PR middle-end/86539
3337         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3338         and cond temporaries don't have reference type if iterator has
3339         pointer type.  For init use &for_pre_body instead of pre_p if
3340         for_pre_body is non-empty.
3342 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
3344         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3345         double-double modes to SFmode directly directly.
3346         (trunc<mode>sf2_fprs): Delete.
3348 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
3350         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3351         for conversions between IFmode and the decimal floating point modes.
3352         (init_float128_ieee): Use the correct names for conversions between
3353         KFmode and the decimal floating point modes.
3355 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
3357         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3358         for the conversions between TDmode and IFmode.
3359         (init_float128_ieee): Use more correct names for the conversions
3360         between TDmode and KFmode.
3362 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
3364         PR tree-optimization/86526
3365         * builtins.c (expand_builtin_memcmp): Formatting fixes.
3366         (inline_expand_builtin_string_cmp): Likewise.
3367         (inline_string_cmp): Likewise.  Use c_readstr instead of
3368         builtin_memcpy_read_str.  Add unit_mode temporary.
3370 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3372         PR middle-end/86528
3373         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3374         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3376 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3378         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3379         Alphabetize prototypes of built-in functions, separating out
3380         built-in functions that are listed in this section but should be
3381         described elsewhere.
3383 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
3385         PR target/86511
3386         * expmed.c (emit_store_flag): Do not emit setcc followed by a
3387         conditional move when trapping comparison was split to a
3388         non-trapping one (and vice versa).
3390 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
3392         * config/s390/s390.c (s390_function_profiler): Generate nops
3393         instead of profiler call sequences.
3394         * config/s390/s390.opt: Add the new option.
3396 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
3398         * config/s390/s390.c (s390_function_profiler): Generate
3399         __mcount_loc section.
3400         * config/s390/s390.opt: Add the new option.
3402 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
3404         * common.opt: Add the new warning.
3405         * config/s390/s390.c (s390_function_profiler): Emit "brasl
3406         %r0,__fentry__" when -mfentry is specified.
3407         (s390_option_override_internal): Disallow -mfentry for 31-bit
3408         CPUs.
3409         * config/s390/s390.opt: Add the new option.
3411 2018-07-16  Richard Biener  <rguenther@suse.de>
3413         PR lto/86523
3414         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3415         for function-local FUNCTION_DECL and RESULT_DECL immediately.
3417 2018-07-16  Martin Liska  <mliska@suse.cz>
3419         PR ipa/86529
3420         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3421         to ::get_create.
3423 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
3425         * config/arc/arcHS.md: Update ARCHS scheduling rules.
3427 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
3429         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3430         for ARCHS4x.
3431         * config/arc/arc-cpus.def (hs4x): New cpu.
3432         (hs4xd): Likewise.
3433         * config/arc/arc-tables.opt: Regenerate.
3434         * config/arc/arc.c (arc_sched_issue_rate): New function.
3435         (TARGET_SCHED_ISSUE_RATE): Define.
3436         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3437         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3438         fpu_cvt.
3439         (attr tune): Add ARCHS4x tune values.
3440         (attr tune_dspmpy): Define.
3441         (*tst): Correct instruction type.
3442         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3443         * config/arc/arcHS4x.md: New file.
3444         * config/arc/fpu.md: Update instruction type attributes.
3445         * config/arc/t-multilib: Regenerate.
3447 2018-07-16  Tom de Vries  <tdevries@suse.de>
3449         PR debug/86455
3450         * var-tracking.c (vt_initialize): Fix pre_dec handling.
3452 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
3454         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3455         early clobber.
3457 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
3459         PR tree-optimization/86514
3460         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3461         conversion to a boolean type from a type with greater precision.
3463 2018-07-16  Tom de Vries  <tdevries@suse.de>
3465         * var-tracking.c (vt_initialize): Print adjusted insn slim if
3466         dump_flags request TDF_SLIM.
3468 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
3470         * fold-const.c (int_const_binop_1): Abstract...
3471         (wide_int_binop): ...wide int code here.
3472         (poly_int_binop): ...poly int code here.
3473         Abstract the rest of int_const_binop_1 into int_const_binop.
3474         * fold-const.h (wide_int_binop): New.
3475         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3476         Remove useless PLUS/MINUS_EXPR case.
3477         (zero_nonzero_bits_from_vr): Move wide int code...
3478         (zero_nonzero_bits_from_bounds): ...here.
3479         (extract_range_from_binary_expr_1): Move mask optimization code...
3480         (range_easy_mask_min_max): ...here.
3481         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3482         (range_easy_mask_min_max): New.
3484 2018-07-15  Jeff Law  <law@redhat.com>
3486         PR target/85993
3487         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3488         block.
3490 2018-07-14  Jim Wilson  <jimw@sifive.com>
3492         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3494 2018-07-14  Paul Koning  <ni1d@arrl.net>
3496         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3498 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
3500         * lto-streamer-out.c (copy_function_or_variable): Dump info about
3501         copying section.
3503 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
3504             Steve Munroe  <munroesj52@gmail.com>
3506         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3507         (_mm_andnot_si128): Likewise.
3508         (_mm_or_si128): Likewise.
3509         (_mm_xor_si128): Likewise.
3511 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
3513         PR middle-end/78809
3514         * builtins.c (expand_builtin_memcmp): Inline the calls first
3515         when result_eq is false.
3516         (expand_builtin_strcmp): Inline the calls first.
3517         (expand_builtin_strncmp): Likewise.
3518         (inline_string_cmp): New routine. Expand a string compare
3519         call by using a sequence of char comparison.
3520         (inline_expand_builtin_string_cmp): New routine. Inline expansion
3521         a call to str(n)cmp/memcmp.
3522         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3523         New option.
3524         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3526 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
3528         * config/arm/driver-arm.c: Include arm-native.h.
3529         (host_detect_local_cpu): Use auto-generated data tables.
3530         (vendors, arm_cpu_table): Delete.  Move part information to ...
3531         * config/arm/arm-cpus.in: ... here.
3532         * config/arm/parsecpu.awk (gen_native): New function.
3533         (vendor, part): New CPU fields.
3534         (END): Add support for building the native CPU detection tables.
3535         * config/arm/t-arm (arm-native.h): Add build rule.
3536         (driver-arm.o): Add dependency on arm-native.h.
3538 2018-07-13  Richard Biener  <rguenther@suse.de>
3540         PR middle-end/85974
3541         * match.pd (addr1 - addr2): Allow either of the operand to
3542         have a conversion.
3544 2018-07-13  Tom de Vries  <tdevries@suse.de>
3546         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3547         in remap_ssa_name.
3549 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
3551         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3552         arrays instead of numbered variables.
3554 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
3556         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3557         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3559 2018-07-13  Richard Biener  <rguenther@suse.de>
3561         PR debug/86452
3562         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3563         instead of get_context_die.
3565 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
3566             Richard Biener  <rguenther@suse.de>
3568         PR middle-end/86489
3569         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3570         that the loop latch destination where phi is defined.
3572 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
3574         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3575         (riscv_expand_epilogue): Add assertion to check interrupt mode.
3576         (riscv_set_current_function): Extract getting interrupt type to new
3577         function.
3578         (riscv_get_interrupt_type): New function.
3579         (riscv_merge_decl_attributes): New function, checking interrupt type is
3580         same.
3581         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3583 2018-07-12  Paul Koning  <ni1d@arrl.net>
3585         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3586         directive.
3588 2018-07-12  Paul Koning  <ni1d@arrl.net>
3590         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3591         zero reference, add doloop_end instead.
3592         * doc/md.texi (decrement_and_branch_until_zero): Remove.
3593         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
3594         detail for doloop_end.
3596 2018-07-12  Martin Sebor  <msebor@redhat.com>
3598         PR c/86453
3599         * attribs.c (decl_attributes): Reject conflicting attributes before
3600         calling attribute handlers.
3602 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
3604         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3605          parameter.
3606         (gcc::dump_manager::get_dump_file_name): likewise.
3607         (dump_begin): Likewise.
3608         * dumpfile.h (dump_begin): Update prototype.
3609         (gcc::dump_manager::get_dump_file_name,
3610         gcc::dump_manager::get_dump_file_name): Update prototype.
3612 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
3614         * internal-fn.h (vectorizable_internal_fn_p): New function.
3615         * tree-vect-slp.c (compatible_calls_p): Likewise.
3616         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
3617         to internal functions.
3618         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3620 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
3622         * fold-const.h (inverse_conditions_p): Declare.
3623         * fold-const.c (inverse_conditions_p): New function.
3624         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
3625         that test the inverse condition of a conditional internal function.
3626         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3627         * internal-fn.c (internal_fn_mask_index): Handle conditional
3628         internal functions.
3629         (vectorized_internal_fn_supported_p): New function.
3630         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3631         (any_pred_load_store): Replace with...
3632         (need_to_predicate): ...this new variable.
3633         (redundant_ssa_names): New variable.
3634         (ifcvt_can_use_mask_load_store): Move initial checks to...
3635         (ifcvt_can_predicate): ...this new function.  Handle tree codes
3636         for which a conditional internal function exists.
3637         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3638         instead of ifcvt_can_use_mask_load_store.  Update after variable
3639         name change.
3640         (predicate_load_or_store): New function, split out from
3641         predicate_mem_writes.
3642         (check_redundant_cond_expr): New function.
3643         (value_available_p): Likewise.
3644         (predicate_rhs_code): Likewise.
3645         (predicate_mem_writes): Rename to...
3646         (predicate_statements): ...this.  Use predicate_load_or_store
3647         and predicate_rhs_code.
3648         (combine_blocks, tree_if_conversion): Update after above name changes.
3649         (ifcvt_local_dce): Handle redundant_ssa_names.
3650         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3651         general conditional functions.
3652         * tree-vect-stmts.c (vectorizable_call): Likewise.
3654 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
3655             Alan Hayward  <alan.hayward@arm.com>
3656             David Sherwood  <david.sherwood@arm.com>
3658         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3659         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3660         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3661         plus and minus and convert them into IFN_COND_FMA-based sequences.
3662         (convert_mult_to_fma): Handle conditional plus and minus.
3664 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
3666         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3667         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3668         (cond_fnms_optab): New optabs.
3669         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3670         internal functions.
3671         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3672         * internal-fn.h (get_conditional_internal_fn): Declare.
3673         (get_unconditional_internal_fn): Likewise.
3674         * internal-fn.c (cond_ternary_direct): New macro.
3675         (expand_cond_ternary_optab_fn): Likewise.
3676         (direct_cond_ternary_optab_supported_p): Likewise.
3677         (FOR_EACH_COND_FN_PAIR): Likewise.
3678         (get_conditional_internal_fn): New function.
3679         (get_unconditional_internal_fn): Likewise.
3680         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3681         (gimple_match_op::gimple_match_op): Add a new overload for 5
3682         operands.
3683         (gimple_match_op::set_op): Likewise.
3684         (gimple_resimplify5): Declare.
3685         * genmatch.c (decision_tree::gen): Generate simplifications for
3686         5 operands.
3687         * gimple-match-head.c (gimple_simplify): Define an overload for
3688         5 operands.  Handle calls with 5 arguments in the top-level overload.
3689         (convert_conditional_op): Handle conversions from unconditional
3690         internal functions to conditional ones.
3691         (gimple_resimplify5): New function.
3692         (build_call_internal): Pass a fifth operand.
3693         (maybe_push_res_to_seq): Likewise.
3694         (try_conditional_simplification): Try converting conditional
3695         internal functions to unconditional internal functions.
3696         Handle 3-operand unconditional forms.
3697         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3698         Define ternary equivalents of the current rules for binary conditional
3699         internal functions.
3700         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3701         ternary operations.
3702         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3703         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3704         (optab): Handle them.
3705         (SVE_COND_FP_TERNARY): New int iterator.
3706         (sve_fmla_op, sve_fmad_op): New int attributes.
3707         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3708         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3709         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3711 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
3713         * target.def (preferred_else_value): New target hook.
3714         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3715         * doc/tm.texi: Regenerate.
3716         * targhooks.h (default_preferred_else_value): Declare.
3717         * targhooks.c (default_preferred_else_value): New function.
3718         * internal-fn.h (conditional_internal_fn_code): Declare.
3719         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3720         (get_conditional_internal_fn): Use it.
3721         (conditional_internal_fn_code): New function.
3722         * gimple-match.h (gimple_match_cond): New struct.
3723         (gimple_match_op): Add a cond member function.
3724         (gimple_match_op::gimple_match_op): Update all forms to take a
3725         gimple_match_cond.
3726         * genmatch.c (expr::gen_transform): Use the same condition as res_op
3727         for the suboperation, but don't specify a particular else_value.
3728         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3729         (visit_nary_op, visit_reference_op_load): Pass
3730         gimple_match_cond::UNCOND to the gimple_match_op constructor.
3731         * gimple-match-head.c: Include tree-eh.h
3732         (convert_conditional_op): New function.
3733         (maybe_resimplify_conditional_op): Likewise.
3734         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3735         (gimple_resimplify2): Likewise.
3736         (gimple_resimplify3): Likewise.
3737         (gimple_resimplify4): Likewise.
3738         (maybe_push_res_to_seq): Return null for conditional operations.
3739         (try_conditional_simplification): New function.
3740         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
3741         constructor.
3742         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3743         IFN_COND_* call.
3744         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3745         function.
3746         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3748 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
3750         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3751         DECL_FCONTEXT
3752         (hash_tree): Do not hash DECL_FCONTEXT
3753         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3754         Do not stream DECL_FCONTEXT.
3755         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3756         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3758 2018-07-12  Richard Biener  <rguenther@suse.de>
3760         PR debug/86462
3761         * dwarf2out.c (gen_block_die): Only output blocks when they have
3762         at least one !DECL_IGNORED_P variable.
3764 2018-07-12  Richard Biener  <rguenther@suse.de>
3766         PR target/84829
3767         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3768         Remove -mieee-fp handling.
3770 2018-07-12  Richard Biener  <rguenther@suse.de>
3772         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3773         left-over from last patch.
3775 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
3777         PR tree-optimization/86492
3778         * gimple-ssa-store-merging.c
3779         (imm_store_chain_info::coalesce_immediate_stores): Call
3780         check_no_overlap even for the merge_overlapping case.  Formatting fix.
3782 2018-07-12  Richard Biener  <rguenther@suse.de>
3784         PR middle-end/86479
3785         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3786         move possibly trapping operations into the conditional.
3788 2018-07-12  Richard Biener  <rguenther@suse.de>
3790         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3791         (vn_lookup_simplify_result): Remove recursion limit applied
3792         here.
3793         (vn_nary_build_or_lookup_1): Adjust.
3794         (try_to_simplify): Likewise.
3795         * gimple-match-head.c (gimple_resimplify1): Instead apply one
3796         here.
3797         (gimple_resimplify2): Likewise.
3798         (gimple_resimplify3): Likewise.
3799         (gimple_resimplify4): Likewise.
3801 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
3803         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3804         Use __mmask64 type instead of __mmask8 for __M argument.
3805         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3806         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3807         __U argument.
3808         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3809         __mmask16 for __M argument.
3810         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3811         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3812         to __mmask16 instead of __mmask8.
3813         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3814         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3815         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3816         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3817         instead of __mmask16 for __U argument.
3818         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3819         __mmask16 instead of __mmask8 for __U argument.
3820         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3821         __U argument.
3822         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3823         __mmask16.
3824         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3825         argument.
3826         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3827         __U argument.
3828         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3829         __mmask16.
3830         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3831         of __mmask16.
3832         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3833         __U argument.
3834         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3835         __U argument.
3836         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3837         __U argument.
3838         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3839         __U argument.
3840         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3841         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3842         return type as well as __M argument type and all casts from __mmask8
3843         to __mmask32.
3844         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3845         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3846         return type as well as __M argument type and all casts from __mmask8
3847         to __mmask16.
3848         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3849         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3850         return type as well as __M argument type and all casts from __mmask8
3851         to __mmask32.
3852         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3853         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3854         return type as well as __M argument type and all casts from __mmask8
3855         to __mmask16.
3856         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3857         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3858         __mmask16.
3860 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
3862         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3863         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3864         for __U argument.
3866 2018-07-11  Paul Koning  <ni1d@arrl.net>
3868         * doc/md.texi (define_subst): Document how multiple occurrences of
3869         the same argument in the replacement pattern are handled.
3871 2018-07-11  Paul Koning  <ni1d@arrl.net>
3873         * doc/extend.texi (Common Variable Attributes): Move "mode" into
3874         alphabetical order.
3875         (Common Type Attributes): Add "mode" attribute.
3877 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
3879         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3880         stream DECL_ORIGINAL_TYPE.
3881         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3882         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3883         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3884         Do not walk original type.
3885         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3886         external decls.
3887         (write_ts_decl_non_common_tree_pointers): Do not stream
3888         DECL_ORIGINAL_TYPE
3889         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
3890         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
3892 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
3894         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
3895         thread twice from the same starting edge.
3897 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
3899         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
3900         * gimple.c (gimple_call_nonnull_result_p): ...here...
3901         (gimple_call_nonnull_arg): ...and here.
3902         * gimple.h (gimple_call_nonnull_result_p): New.
3903         (gimple_call_nonnull_arg): New.
3905 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
3907         * config/arm/arm-cpus.in: Move information from fpu field of each
3908         cpu definition to the isa field.
3909         * config/arm/parsecpu.awk (fpu): Delete match rule.
3910         (gen_comm_data): Don't add bits from the CPU's FPU entry.
3912 2018-07-11  Richard Biener  <rguenther@suse.de>
3914         PR debug/86457
3915         * dwarf2out.c (init_sections_and_labels): Use
3916         output_asm_line_debug_info consistently.
3917         (dwarf2out_early_finish): Likewise.
3918         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3919         type units.
3921 2018-07-11  Richard Biener  <rguenther@suse.de>
3923         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
3924         Rework father_bb setting in a way to avoid propagating constants
3925         multiple times on a loop body.
3927 2018-07-10  Mark Wielaard  <mark@klomp.org>
3929         PR debug/86459
3930         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
3932 2018-07-10  Richard Biener  <rguenther@suse.de>
3934         * hash-map.h (hash_map::iterator::operator*): Return
3935         references to key and value.
3937 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
3939         PR c++/86443
3940         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
3941         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
3942         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
3943         (gimplify_omp_for): For composite loops, move outer
3944         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
3945         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
3946         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
3947         TREE_LIST for both the original class iterator and the "last" helper
3948         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
3949         loop, remember has_decl_expr from outer composite loops for the
3950         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
3952 2018-07-09  Martin Sebor  <msebor@redhat.com>
3954         PR middle-end/77357
3955         PR middle-end/86428
3956         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
3957         accessing implicitly initialized array elements.
3958         * expr.c (string_constant): Handle string initializers of
3959         character arrays within aggregates.
3960         * gimple-fold.c (fold_array_ctor_reference): Add argument.
3961         Store element offset.  As a special case, handle zero size.
3962         (fold_nonarray_ctor_reference): Same.
3963         (fold_ctor_reference): Add argument.  Store subobject offset.
3964         * gimple-fold.h (fold_ctor_reference): Add argument.
3966 2018-07-09  Paul Koning  <ni1d@arrl.net>
3968         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
3969         (pdp11_insn_cost): New function.
3970         (pdp11_md_asm_adjust): New function.
3971         (TARGET_INVALID_WITHIN_DOLOOP): Define.
3972         (pdp11_rtx_costs): Update to match machine better.
3973         (output_addr_const_pdp11): Correct format mismatch warnings.
3974         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
3975         * config/pdp11/pdp11.md: General change to add base_cost and/or
3976         length attributes for use by new pdp11_insn_cost function.
3977         (MIN_BRANCH): Correct definition.
3978         (MIN_SOB): Ditto.
3979         (doloop_end): Use standard pattern name for looping pattern.
3980         (doloop_end_nocc): New.
3981         (movsf): Add another constraint alternative.
3982         (zero_extendqihi2): Add constraint alternatives for not in place
3983         extend.
3984         (zero_extendhisi2): Remove.
3985         (shift patterns): Add CC handling variants.
3986         (bswaphi2): New.
3987         (bswapsi2): New.
3988         (rothi3): New.
3989         (define_peephole2): New peephole to recognize mov that sets CC for
3990         subsequent test.
3992 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3994         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
3995         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
3996         wi::add.
3998 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
4000         PR c/86420
4001         * real.c (real_nextafter): Return true if result is denormal.
4003 2018-07-09  Martin Liska  <mliska@suse.cz>
4005         * common.opt: Add back wrongly removed attribute.
4007 2018-07-09  Richard Biener  <rguenther@suse.de>
4009         PR debug/86413
4010         * dwarf2out.c (gen_block_die): For an early generated DIE
4011         always output high/low PC attributes.
4013 2018-07-09  Tom de Vries  <tdevries@suse.de>
4015         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4016         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4017         onto VAR_DECL with abstract origin.
4019 2018-07-07  Jim Wilson  <jimw@sifive.com>
4021         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4023 2018-07-07  Tom de Vries  <tdevries@suse.de>
4025         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4027 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
4029         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4030         overflow_type.
4031         (combine_bound): Use wide-int overflow calculation instead of
4032         rolling our own.
4033         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4034         overflow_type.
4035         * fold-const.c (int_const_binop_2): Same.
4036         (extract_muldiv_1): Same.
4037         (fold_div_compare): Same.
4038         (fold_abs_const): Same.
4039         * match.pd: Same.
4040         * poly-int.h (add): Same.
4041         (sub): Same.
4042         (neg): Same.
4043         (mul): Same.
4044         * predict.c (predict_iv_comparison): Same.
4045         * profile-count.c (slow_safe_scale_64bit): Same.
4046         * simplify-rtx.c (simplify_const_binary_operation): Same.
4047         * tree-chrec.c (tree_fold_binomial): Same.
4048         * tree-data-ref.c (split_constant_offset_1): Same.
4049         * tree-if-conv.c (idx_within_array_bound): Same.
4050         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4051         * tree-ssa-phiopt.c (minmax_replacement): Same.
4052         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4053         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4054         * vr-values.c (vr_values::adjust_range_with_scev): Same.
4055         * wide-int.cc (wi::add_large): Same.
4056         (wi::mul_internal): Same.
4057         (wi::sub_large): Same.
4058         (wi::divmod_internal): Same.
4059         * wide-int.h: Change overflow type to overflow_type for neg, add,
4060         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4061         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4062         mul_internal, divmod_internal.
4063         (overflow_type): New enum.
4064         (accumulate_overflow): New.
4066 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4068         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4069         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4071 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
4073         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4074         argument is checked for zero before entering loop, avoid checking again.
4076 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
4078         * gimplify.h (generic_expr_could_trap_p): Set as global function.
4079         * gimplify.h (generic_expr_could_trap_p): Likwise.
4080         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4082 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
4084         PR tree-optimization/86401
4085         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4086         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4087         (fold_bit_and_mask): ... here.  New helper function for match.pd.
4088         * fold-const.h (fold_bit_and_mask): Declare.
4089         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4091 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
4093         PR target/86324
4094         * target.def (translate_mode_attribute): New hook.
4095         * targhooks.h (default_translate_mode_attribute): Declare.
4096         * targhooks.c (default_translate_mode_attribute): New function.
4097         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4098         * doc/tm.texi: Regenerate.
4099         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4100         (rs6000_translate_mode_attribute): New function.
4102 2018-07-06  Paul Koning  <ni1d@arrl.net>
4104         * doc/md.texi (define_split): Document DONE and FAIL.
4105         (define_peephole2): Ditto.
4107 2018-07-05  Jeff Law  <law@redhat.com>
4109         PR tree-optimization/86010
4110         * tree-ssa-dse.c (compute_trims): More aggressively trim at
4111         both the head and tail of mem* and str* calls.
4113 2018-07-05  Jim Wilson  <jimw@sifive.com>
4115         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4117 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
4119         * config/aarch64/aarch64-simd.md: correct flags text for
4120         MIN_EXPR replacement.
4122 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
4124         * configure: Regenerated.
4126 2018-07-05  Carl Love  <cel@us.ibm.com>
4128         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4129         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4130         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4131         VSX_BUILTIN_DOUBLEL_V4SF.
4133 2018-07-05  Martin Sebor  <msebor@redhat.com>
4135         PR c++/86400
4136         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4137         than its domain to compute its the upper bound of a char array.
4139 2018-07-05  Nathan Sidwell  <nathan@acm.org>
4141         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4142         * doc/cpp.texi: Update comment.
4143         * doc/tm.texi: Rebuilt.
4144         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4145         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4146         * doc/extend.texi (Backwards Compatibility): Clarify it is system
4147         headers affected by extern "C".
4148         * system.h: Poison NO_IMPLICIT_EXTERN_C.
4149         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4150         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4151         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4152         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4153         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4154         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4155         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4156         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4157         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4158         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4159         config/sparc/sp64-elf.h, config/spu/spu.h,
4160         config/stormy16/stormy16.h, config/v850/v850.h,
4161         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4162         define NO_IMPLICIT_EXTERN_C.
4163         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4165 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
4167         PR target/84711
4168         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4169         instead of GET_MODE_SIZE when comparing Units.
4171 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
4173         PR target/84711
4174         * rtlanal.c (set_noop_p): Constrain on mode change,
4175         include hard-reg-set.h
4177 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
4179         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4181 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
4183         Revert
4184         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
4186         PR sanitizer/84250
4187         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4188         libasan.
4189         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4191 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
4193         PR sanitizer/84250
4194         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4195         libasan.
4196         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4198 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
4200         PR middle-end/86380
4201         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4203 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
4205         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4206         neg_*_op* variables.
4208 2018-07-04  Martin Liska  <mliska@suse.cz>
4210         * tree-switch-conversion.c: Define
4211         max_ratio_for_speed and max_ratio_for_size constants.
4213 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
4214             Martin Liska  <mliska@suse.cz>
4216         PR middle-end/66240
4217         PR target/45996
4218         PR c/84100
4219         * common.opt: Rename align options with 'str_' prefix.
4220         * common/config/i386/i386-common.c (set_malign_value): New
4221         function.
4222         (ix86_handle_option): Use it to set -falign-* options/
4223         * config/aarch64/aarch64-protos.h (struct tune_params): Change
4224         type from int to string.
4225         * config/aarch64/aarch64.c: Update default values from int
4226         to string.
4227         * config/alpha/alpha.c (alpha_override_options_after_change):
4228         Likewise.
4229         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4230         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4231         max skip conditionally.
4232         * config/i386/freebsd.h (SUBALIGN_LOG): New.
4233         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4234         max skip conditionally.
4235         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4236         max skip conditionally.
4237         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4238         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4239         max skip conditionally.
4240         * config/i386/i386.c (struct ptt): Change type from int to
4241         string.
4242         (ix86_default_align): Set default values.
4243         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4244         max skip conditionally.
4245         * config/i386/iamcu.h (SUBALIGN_LOG): New.
4246         (ASM_OUTPUT_MAX_SKIP_ALIGN):
4247         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4248         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4249         max skip conditionally.
4250         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4251         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4252         * config/i386/x86-64.h (SUBALIGN_LOG): New.
4253         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4254         max skip conditionally.
4255         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4256         * config/ia64/ia64.c (ia64_option_override): Set default values
4257         for alignment options.
4258         * config/m68k/m68k.c: Handle new str_align_* options.
4259         * config/mips/mips.c (mips_set_compression_mode): Change
4260         type of constants.
4261         (mips_option_override): Set default values for options.
4262         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4263         Likewise.
4264         * config/rs6000/rs6000.c (rs6000_option_override_internal):
4265         Likewise.
4266         * config/rx/rx.c (rx_option_override): Likewise.
4267         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4268         (LABEL_ALIGN): Use align_labels_log.
4269         (LOOP_ALIGN): Use align_loops_align.
4270         * config/s390/s390.c (s390_asm_output_function_label): Use new
4271         macros.
4272         * config/sh/sh.c (sh_override_options_after_change):
4273         Change type of constants.
4274         * config/spu/spu.c (spu_sched_init): Likewise.
4275         * config/sparc/sparc.c (sparc_option_override): Set default
4276         values for options.
4277         * config/visium/visium.c (visium_option_override): Likewise.
4278         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4279         emit p2align format with last argument if it's not needed.
4280         * doc/invoke.texi: Document extended format of -falign-*.
4281         * final.c: Use align_labels alignment.
4282         * flags.h (struct target_flag_state): Change type to use
4283         align_flags.
4284         (struct align_flags_tuple): New.
4285         (struct align_flags): Likewise.
4286         (align_loops_log): Redefine macro to use new types.
4287         (align_loops_max_skip): Redefine macro to use new types.
4288         (align_jumps_log): Redefine macro to use new types.
4289         (align_jumps_max_skip): Redefine macro to use new types.
4290         (align_labels_log): Redefine macro to use new types.
4291         (align_labels_max_skip): Redefine macro to use new types.
4292         (align_functions_log): Redefine macro to use new types.
4293         (align_loops): Redefine macro to use new types.
4294         (align_jumps): Redefine macro to use new types.
4295         (align_labels): Redefine macro to use new types.
4296         (align_functions): Redefine macro to use new types.
4297         (align_functions_max_skip): Redefine macro to use new types.
4298         (align_loops_value): New macro.
4299         (align_jumps_value): New macro.
4300         (align_labels_value): New macro.
4301         (align_functions_value): New macro.
4302         * function.c (invoke_set_current_function_hook): Propagate
4303         alignment values from flags to global variables default in
4304         topleev.h.
4305         * ipa-icf.c (sem_function::equals_wpa): Use
4306         cl_optimization_option_eq instead of memcmp.
4307         * lto-streamer.h (cl_optimization_stream_out): Support streaming
4308         of string types.
4309         (cl_optimization_stream_in): Likewise.
4310         * optc-save-gen.awk: Support strings in cl_optimization.
4311         * opth-gen.awk: Likewise.
4312         * opts.c (finish_options): Remove error checking of invalid
4313         value ranges.
4314         (MAX_CODE_ALIGN): Remove.
4315         (MAX_CODE_ALIGN_VALUE): Likewise.
4316         (parse_and_check_align_values): New function.
4317         (check_alignment_argument): Likewise.
4318         (common_handle_option): Use check_alignment_argument.
4319         * opts.h (parse_and_check_align_values): Declare.
4320         * toplev.c (init_alignments): Remove.
4321         (read_log_maxskip): New.
4322         (parse_N_M): Likewise.
4323         (parse_alignment_opts): Likewise.
4324         (backend_init_target): Remove usage of init_alignments.
4325         * toplev.h (parse_alignment_opts): Declare.
4326         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4327         argument.
4328         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4329         * tree.c (cl_option_hasher::equal): New.
4330         * varasm.c: Use new global macros.
4332 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
4334         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4335         Use a simpler align directive also if MAXSKIP = ALIGN-1.
4336         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4337         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4338         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4339         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4340         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4341         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4342         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4343         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4344         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4345         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4346         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4348 2018-07-04  Martin Liska  <mliska@suse.cz>
4349             Jonathan Wakely  <jwakely@redhat.com>
4351         * coverage.c: Use correct type.
4352         * doc/invoke.texi: Language correction.
4354 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
4356         PR target/85620
4357         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4358         ENDBRANCH for non-tail call which may return via indirect branch.
4359         * doc/extend.texi: Document indirect_return attribute.
4361 2018-07-03  Martin Sebor  <msebor@redhat.com>
4363         PR tree-optimization/86274
4364         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4365         precondition.
4366         (format_floating): Correct handling of infinities and NaNs.
4368 2018-07-03  Martin Sebor  <msebor@redhat.com>
4370         * print-tree.c (print_real_cst): New function.
4371         (print_node_brief): Call it.
4372         (print_node): Ditto.
4374 2018-07-03  Jeff Law  <law@redhat.com>
4376         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4377         into a single pattern.
4379         * config/h8300/h8300.md (ors code_iterator): New.
4380         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4381         a single pattern and single splitter.
4382         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4383         (iorqi3_1, xorqi3_1): Likewise.
4384         (iorqi3, xorqi3 expanders): Similarly.
4386         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4387         (movmd_internal) into a single pattern using the P mode iterator.
4388         (movmd splitters): Similarly.
4389         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4390         (movsd splitters): Similarly.
4392         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4393         ADDB, ADDW and ADDL into a single ADD attribute which selects the
4394         right table based on the size of the operand.
4395         * config/h8300/h8300.md (length_table): Corresponding changes. All
4396         references to "addb", "addw" and "addl" changed to "add".
4397         (btst patterns): Merge two variants into a single pattern.
4398         (tstqi, tsthi): Likewise.
4399         (addhi3_incdec, addsi3_incdec): Likewise.
4400         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4401         (mulhi3, mulsi3): Likewise.
4402         (udivhi3, udivsi3): Likewise.
4403         (divhi3, divsi3): Likewise.
4404         (andorqi3, andorhi3, andorsi3): Likewise.
4406 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
4408         PR target/85694
4409         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4410         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4412 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4414         PR tree-optimization/85694
4415         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4416         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4417         UNSPEC_URHADD.
4418         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4419         (<u>avg<mode>3_ceil): New patterns.
4421 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
4423         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4424         scan-tree-dump directive.
4425         * gcc.dg/vect/slp-perm-2.c: Likewise.
4426         * gcc.dg/vect/slp-perm-3.c: Likewise.
4427         * gcc.dg/vect/slp-perm-5.c: Likewise.
4428         * gcc.dg/vect/slp-perm-6.c: Likewise.
4429         * gcc.dg/vect/slp-perm-7.c: Likewise.
4430         * gcc.dg/vect/slp-perm-8.c: Likewise.
4432 2018-07-03  Marek Polacek  <polacek@redhat.com>
4434         PR middle-end/86202
4435         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4437 2018-07-03  Richard Biener  <rguenther@suse.de>
4439         PR ipa/86389
4440         * tree-ssa-structalias.c (find_func_clobbers): Properly
4441         handle indirect calls.
4443 2018-07-03  Jeff Law  <law@redhat.com>
4445         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4446         (shifts): New code iterator.
4447         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4448         expander.  Fix HImode handling on H8/SX.
4449         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4450         (subqi3, subhi3, subsi3 expanders): Likewise.
4451         (andqi3, andhi3, andsi3 expanders): Likewise.
4452         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4453         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4454         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4455         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4456         (zero_extendqihi2, zero_extendqisi2): Likewise.
4457         (extendqihi2, extendqisi2): Likewise.
4458         (rotlqi3, rotlhi3, rotlsi3): Likewise.
4459         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4460         (rotlqi3_1, rotlhi3_1): Likewise.
4461         (logicalhi3_sn, logicalsi3_sn): Likewise.
4462         (logicalhi3, logicalsi3): Likewise.
4464 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4466         * tree-vect-patterns.c (vect_recog_rotate_pattern)
4467         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4468         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4469         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4470         type to append_pattern_def_seq instead of creating a stmt_vec_info
4471         directly.
4472         (build_mask_conversion): Likewise.  Remove vinfo argument.
4473         (vect_add_conversion_to_patterm): Likewise, renaming to...
4474         (vect_add_conversion_to_pattern): ...this.
4475         (vect_recog_mask_conversion_pattern): Update call to
4476         build_mask_conversion.  Pass the vector type to
4477         append_pattern_def_seq here too.
4478         (vect_recog_gather_scatter_pattern): Update call to
4479         vect_add_conversion_to_pattern.
4481 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4483         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4484         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4485         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4486         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4487         STMT_VINFO_PATTERN_DEF_SEQ to null here.
4488         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4489         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4490         append_pattern_def_seq instead of new_pattern_def_seq.
4491         (vect_recog_divmod_pattern): Do both of the above.
4492         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4493         is null.
4495 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4497         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4498         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4499         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4500         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4501         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4502         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4503         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4504         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4505         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4506         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4507         parameter with a single stmt_vec_info.
4508         (vect_recog_func_ptr): Likewise.
4509         (vect_recog_gather_scatter_pattern): Likewise, folding in...
4510         (vect_try_gather_scatter_pattern): ...this.
4511         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4512         the stmt_vec_info of the statement to be matched.  Don't clear
4513         STMT_VINFO_RELATED_STMT.
4514         (vect_pattern_recog): Update call accordingly.
4516 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4518         PR tree-optimization/85694
4519         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4520         (uavgM3_ceil): Document new optabs.
4521         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4522         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4523         functions.
4524         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4525         (savg_ceil_optab): New optabs.
4526         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4527         (vect_vect_recog_func_ptrs): Add it.
4528         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4529         constant directly from the associated lhs.
4531 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4533         * tree-vect-patterns.c (vect_split_statement): New function.
4534         (vect_convert_input): Use it to try to split an existing cast.
4536 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4538         * poly-int.h (print_hex): New function.
4539         * dumpfile.h (dump_dec, dump_hex): Declare.
4540         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4541         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4542         min_input_precision, operation_precision and operation_sign.
4543         * tree-vect-patterns.c (vect_get_range_info): New function.
4544         (vect_same_loop_or_bb_p, vect_single_imm_use)
4545         (vect_operation_fits_smaller_type): Delete.
4546         (vect_look_through_possible_promotion): Add an optional
4547         single_use_p parameter.
4548         (vect_recog_over_widening_pattern): Rewrite to use new
4549         stmt_vec_info infomration.  Handle one operation at a time.
4550         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4551         (vect_truncatable_operation_p, vect_set_operation_type)
4552         (vect_set_min_input_precision): New functions.
4553         (vect_determine_min_output_precision_1): Likewise.
4554         (vect_determine_min_output_precision): Likewise.
4555         (vect_determine_precisions_from_range): Likewise.
4556         (vect_determine_precisions_from_users): Likewise.
4557         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4558         (vect_vect_recog_func_ptrs): Put over_widening first.
4559         Add cast_forwprop.
4560         (vect_pattern_recog): Call vect_determine_precisions.
4562 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
4564         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4565         statements that have been replaced by further pattern statements.
4566         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4568 2018-07-03  Richard Biener  <rguenther@suse.de>
4570         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4571         always set *dt.  Dump vectype in vectype overload.
4572         * dumpfile.h (dump_gimple_expr): New function.
4573         (dump_gimple_expr_loc): Likewise.
4574         * dumpfile.c (dump_gimple_expr): New function.
4575         (dump_gimple_expr_loc): Likewise.
4577 2018-07-02  Jeff Law  <law@redhat.com>
4579         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4580         the H8/300, H8/300H and H8/S variants into a single pattern.
4581         (movhi_h8300, movqi_h8300hs): Similarly.
4582         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4583         (QHI mode iterator): New.
4585         * config/h8300/h8300.md: Remove trailing whitespace.
4587 2018-07-02  Jim Wilson  <jimw@sifive.com>
4589         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4590         instead of emit_insn for interrupt returns.
4591         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4592         (riscv_sret, riscv_uret): Likewise.
4594 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
4596         * pretty-print.c (selftest::test_pp_format): Move save and restore
4597         of quotes to class auto_fix_quotes, and add an instance.
4598         * selftest.c: Include "intl.h".
4599         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4600         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4601         * selftest.h (selftest::auto_fix_quotes): New class.
4603 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
4605         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4606         (aarch64_sve_prepare_conditional_op): Remove.
4607         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4608         Allow aarch64_simd_reg_or_zero as select operand; remove
4609         the aarch64_sve_prepare_conditional_op call.
4610         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4611         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4612         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4613         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4614         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4615         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4616         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4617         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4618         and a splitters to match all of the *_any patterns.
4619         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4621         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4622         (SVE_COND_FP_BINARY_REV): Remove.
4623         (sve_int_op_rev, sve_fp_op_rev): New.
4624         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4625         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4626         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4627         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4628         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4629         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4630         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4631         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4632         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4634         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4635         Remove match_dup 1 from the inner unspec.
4636         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4638         * config/aarch64/aarch64.md (movprfx): New attr.
4639         (length): Default movprfx to 8.
4640         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4641         (*madd<SVE_I>, *msub<SVE_I): Likewise.
4642         (*<su>mul<SVE_I>3_highpart): Likewise.
4643         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4644         (*v<ASHIFT><SVE_I>3): Likewise.
4645         (*<su><MAXMIN><SVE_I>3): Likewise.
4646         (*<su><MAXMIN><SVE_F>3): Likewise.
4647         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4648         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4649         (*div<SVE_F>4): Likewise.
4651 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
4653         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4654         in dump string.
4656 2018-07-02  Richard Biener  <rguenther@suse.de>
4658         PR tree-optimization/86363
4659         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4660         memset argument refers to a non-variable address.
4662 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
4664         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4665         {PLUS,MINUS}_EXPR code to...
4666         (adjust_symbolic_bound): ...here,
4667         (combine_bound): ...here,
4668         (set_value_range_with_overflow): ...and here.
4670 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
4672         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4673         code...
4674         (extract_range_from_abs_expr): ...here.
4676 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
4678         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4679         -fno-omit-frame-pointer when not optimizing.
4681 2018-07-02  Martin Liska  <mliska@suse.cz>
4683         PR ipa/86279
4684         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4685         (propagate_nothrow): Likewise.
4687 2018-07-02  Martin Liska  <mliska@suse.cz>
4689         PR ipa/86323
4690         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4692 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
4694         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4695         function in r262149, changing "loc" param from source_location to
4696         const dump_location_t &.
4697         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4698         declaration, as above.
4700 2018-07-01  Paul Koning  <ni1d@arrl.net>
4702         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4703         -munit-asm, -mgnu-asm, -mdec-asm.
4704         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4705         (pdp11_output_labelref): New.
4706         (pdp11_output_def): New.
4707         (pdp11_output_addr_vec_elt): New.
4708         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
4709         %# and %@ format codes.
4710         (pdp11_option_override): New.
4711         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4712         (pdp11_output_ident): New.
4713         (pdp11_asm_named_section): New.
4714         (pdp11_asm_init_sections): New.
4715         (pdp11_file_start): New.
4716         (pdp11_file_end): New.
4717         (output_ascii): Use .ascii/.asciz for -mdec-asm.
4718         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
4719         %o, like %c but octal.
4720         (pdp11_option_override): New.
4721         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4722         -mdec-asm.
4723         (DATA_SECTION_ASM_OP): Ditto.
4724         (READONLY_DATA_SECTION_ASM_OP): New.
4725         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4726         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4727         (ASM_OUTPUT_LABELREF): Ditto.
4728         (ASM_OUTPUT_DEF): Ditto.
4729         (ASM_OUTPUT_EXTERNAL): New.
4730         (ASM_OUTPUT_SOURCE_FILENAME): New.
4731         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4732         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4733         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
4734         %# and %@ format codes.
4735         * config/pdp11/pdp11.opt (mgnu-asm): New.
4736         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4737         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4738         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4740 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
4742         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4743         dereferencing path[] beyond its length.
4744         (debug_path): New.
4745         (debug_all_paths): New.
4746         (rewire_first_differing_edge): New.
4747         (adjust_paths_after_duplication): New.
4748         (duplicate_thread_path): Call adjust_paths_after_duplication.
4749         Add new argument.
4750         (thread_through_all_blocks): Add new argument to
4751         duplicate_thread_path.
4753 2018-06-30  Jim Wilson  <jimw@sifive.com>
4755         * config/riscv/predicates.md (p2m1_shift_operand): New.
4756         (high_mask_shift_operand): New.
4757         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4758         pattern using p2m1_shift_operand.
4759         (lshsi3_zero_extend_3+2): New combiner pattern using
4760         high_mask_shift_operand.
4762 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4764         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4765         split out from...
4766         (vect_recog_rotate_pattern): ...here.
4767         (vect_convert_input): Try to insert casts of invariants in the
4768         preheader.
4769         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4770         preheader to be empty.
4772 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4774         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4775         vector type.  If given, install it in the new statement's
4776         STMT_VINFO_VECTYPE.
4777         (vect_element_precision): New function.
4778         (vect_unpromoted_value): New struct.
4779         (vect_unpromoted_value::vect_unpromoted_value): New function.
4780         (vect_unpromoted_value::set_op): Likewise.
4781         (vect_look_through_possible_promotion): Likewise.
4782         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4783         (vect_widened_op_tree, vect_convert_input): Likewise.
4784         (vect_convert_inputs, vect_convert_output): Likewise.
4785         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4786         to handle the optional cast of the multiplication result and
4787         vect_widened_op_tree to detect the widened multiplication itself.
4788         Do not require the input and output of promotion casts to have
4789         the same sign, but base the signedness of the operation on the
4790         input rather than the result.  If the pattern includes two
4791         promotions, check that those promotions have the same sign.
4792         Do not restrict the MULT_EXPR handling to a double-width result;
4793         handle quadruple-width results and wider.  Use vect_convert_inputs
4794         to convert the inputs to the common type.
4795         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
4796         to handle the optional cast of the ABS result.  Also allow a sign
4797         change or a sign extension between the ABS and MINUS.
4798         Use vect_widened_op_tree to detect the widened subtraction and use
4799         vect_convert_inputs to convert the inputs to the common type.
4800         (vect_handle_widen_op_by_const): Delete.
4801         (vect_recog_widen_op_pattern): New function.
4802         (vect_recog_widen_mult_pattern): Use it.
4803         (vect_recog_widen_shift_pattern): Likewise.
4804         (vect_recog_widen_sum_pattern): Use
4805         vect_look_through_possible_promotion to handle the promoted
4806         PLUS_EXPR operand.
4808 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4810         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4811         the containing gimple_seq *.
4812         * gimple-iterator.h (gsi_for_stmt): Declare it.
4813         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4814         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4815         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4816         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4817         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4818         checks.
4819         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4820         split out from...
4821         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
4822         statement being replaced is part of an existing pattern
4823         definition sequence, inserting the new pattern statements before
4824         the original one.
4825         (vect_pattern_recog_1): Don't return a bool.  If the statement
4826         is already part of a pattern, instead apply pattern matching
4827         to the pattern definition statements.  Don't clear the
4828         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4829         (vect_pattern_recog): Don't break after the first match;
4830         continue processing the pattern definition statements instead.
4831         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4833 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4835         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4836         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4837         (vect_recog_widen_sum_pattern): Use it.
4839 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4841         * tree-vect-loop.c (vectorizable_reduction): Assert that the
4842         phi is not a pattern statement and has not been replaced by
4843         a pattern statement.
4844         * tree-vect-patterns.c (type_conversion_p): Don't check
4845         STMT_VINFO_IN_PATTERN_P.
4846         (vect_recog_vector_vector_shift_pattern): Likewise.
4847         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4848         the pattern statement rather than the original statement; check
4849         directly for a WIDEN_MULT_EXPR here.
4850         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4851         vect_is_simple_use to return the pattern statement rather
4852         than the original statement; use is_pattern_stmt_p to check
4853         for such a pattern statement.
4854         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4855         to return the pattern statement rather than the original statement;
4856         don't do the same transformation here.
4857         (vect_is_simple_use): If the defining statement has been replaced
4858         by a pattern statement, return the pattern statement instead.
4859         Remove the corresponding (local) transformation from the vectype
4860         overload.
4862 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
4864         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4865         end and default to null.
4866         * tree-vect-loop.c (vect_create_epilog_for_reduction)
4867         (vectorizable_reduction): Update calls accordingly, dropping the
4868         gimple ** argument if the passed-back statement isn't needed.
4869         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4870         (vect_recog_rotate_pattern): Likewise.
4871         (vect_recog_mask_conversion_pattern): Likewise.
4872         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4873         (vect_mask_constant_operand_p): Likewise.
4874         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4875         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4876         (get_group_load_store_type, get_load_store_type): Likewise.
4877         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4878         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4879         (vectorizable_conversion, vectorizable_assignment): Likewise.
4880         (vectorizable_shift, vectorizable_operation): Likewise.
4881         (vectorizable_store, vect_is_simple_cond): Likewise.
4882         (vectorizable_condition, vectorizable_comparison): Likewise.
4883         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4884         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4885         and move it to the end.  Cope with null def_stmt_outs.
4887 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4889         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
4891 2018-06-29  Jeff Law  <law@redhat.com>
4893         * config/v850/v850.c (v850_legitimate_address_p): Handle large
4894         displacements for TARGET_V850E2V3 and newer.
4895         (TARGET_LRA_P): Remove.  Defaults to LRA now.
4896         * config/v850/v850.md (sign23byte_load): Remove.
4897         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
4898         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
4900 2018-06-29  Martin Liska  <mliska@suse.cz>
4902         PR lto/85759
4903         * coverage.c (coverage_init): Mangle full path name.
4904         * doc/invoke.texi: Document the change.
4905         * gcov-io.c (mangle_path): New.
4906         * gcov-io.h (mangle_path): Likewise.
4907         * gcov.c (mangle_name): Use mangle_path for path mangling.
4909 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4911         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4912         if starting source register is not even.
4914 2018-06-29  Martin Liska  <mliska@suse.cz>
4916         PR tree-optimization/86263
4917         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4918         Make edge redirection.
4920 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
4922         * dumpfile.c (dump_loc): Add indentation based on scope depth.
4923         (dump_scope_depth): New variable.
4924         (get_dump_scope_depth): New function.
4925         (dump_begin_scope): New function.
4926         (dump_end_scope): New function.
4927         * dumpfile.h (get_dump_scope_depth): New declaration.
4928         (dump_begin_scope): New declaration.
4929         (dump_end_scope): New declaration.
4930         (class auto_dump_scope): New class.
4931         (AUTO_DUMP_SCOPE): New macro.
4932         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
4933         AUTO_DUMP_SCOPE.
4935 2018-06-29  Richard Biener  <rguenther@suse.de>
4937         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
4938         compute_all_dependences succeeds.
4939         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
4940         exceed --param loop-max-datarefs-for-datadeps.
4942 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
4944         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
4946 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
4948         PR target/86348
4949         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
4950         alternative 0 in preferred_for_speed attribute.
4952 2018-06-28  Paul Koning  <ni1d@arrl.net>
4954         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
4955         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
4956         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
4957         * config/pdp11/pdp11.md: Correct "length" attribute calculation
4958         for shift insn patterns.
4960 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
4962         * cgraph.c (cgraph_node::get_body): Replace assignments to
4963         "dump_file" with calls to set_dump_file.
4964         * dumpfile.c (alt_dump_file): Make static, and group with...
4965         (alt_flags): ...this definition.
4966         (dumps_are_enabled): New variable.
4967         (refresh_dumps_are_enabled): New function.
4968         (set_dump_file): New function.
4969         (set_alt_dump_file): New function.
4970         (gcc::dump_manager::dump_start): Replace assignments to
4971         "dump_file" and "alt_dump_file" with calls to set_dump_file and
4972         set_alt_dump_file.
4973         (gcc::dump_manager::dump_finish): Likewise.
4974         * dumpfile.h (alt_dump_file): Delete decl.
4975         (dumps_are_enabled): New variable decl.
4976         (set_dump_file): New function decl.
4977         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
4978         global.
4979         * tree-nested.c (lower_nested_functions): Replace assignments to
4980         "dump_file" with calls to set_dump_file.
4982 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
4984         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4985         goto_locus of each outgoing edge of each basic block.
4987 2018-06-28  Richard Biener  <rguenther@suse.de>
4989         * dwarf2out.c (decl_scope_table): Remove.
4990         (push_decl_scope): Likewise.
4991         (pop_decl_scope): Likewise.
4992         (gen_type_die_for_member): Do not call push/pop_decl_scope.
4993         (gen_struct_or_union_type_die): Likewise.
4994         (gen_tagged_type_die): Likewise.
4995         (dwarf2out_init): Do not initialize decl_scope_table.
4996         (dwarf2out_c_finalize): Do not free it.
4998 2018-06-28  Richard Biener  <rguenther@suse.de>
5000         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5001         deciding whether to not re-use a DIE.
5003 2018-06-28  Richard Biener  <rguenther@suse.de>
5005         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5006         DW_AT_abstract_origin attribute.
5008 2018-06-28  Martin Liska  <mliska@suse.cz>
5010         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5011         Use newly introduced constants.
5012         * tree-switch-conversion.h (struct jump_table_cluster):
5013         Define max_ratio_for_size and max_ratio_for_speed.
5015 2018-06-28  Martin Liska  <mliska@suse.cz>
5017         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5018         Add new checking assert to catch invalid state.
5019         (jump_table_cluster::can_be_handled): Handle single case
5020         clusters.
5021         (jump_table_cluster::is_beneficial): Bail out for such case.
5022         (bit_test_cluster::find_bit_tests):
5023         Add new checking assert to catch invalid state.
5024         (bit_test_cluster::can_be_handled): Handle single case
5025         clusters.
5026         (bit_test_cluster::is_beneficial): Bail out for such case.
5027         (switch_decision_tree::analyze_switch_statement):
5028         Fix comment.
5030 2018-06-28  Martin Liska  <mliska@suse.cz>
5032         * common.opt: Introduce -completion option.
5033         * gcc.c (driver_handle_option): Handle it.
5034         (driver::main): Print completions if completion
5035         is set.
5036         * opt-suggestions.c (option_proposer::get_completions):
5037         New function.
5038         (option_proposer::suggest_completion): Likewise.
5039         (option_proposer::find_param_completions): Likewise.
5040         (verify_autocompletions): Likewise.
5041         (test_completion_valid_options): Likewise.
5042         (test_completion_valid_params): Likewise.
5043         (in_completion_p): Likewise.
5044         (empty_completion_p): Likewise.
5045         (test_completion_partial_match): Likewise.
5046         (test_completion_garbage): Likewise.
5047         (opt_proposer_c_tests): Likewise.
5048         * opt-suggestions.h: Declare new functions.
5049         * opts.c (common_handle_option): Handle OPT__completion_.
5050         * selftest-run-tests.c (selftest::run_tests): Add
5051         opt_proposer_c_tests.
5052         * selftest.c (assert_str_startswith): New.
5053         * selftest.h (assert_str_startswith): Likewise.
5054         (opt_proposer_c_tests): New.
5055         (ASSERT_STR_STARTSWITH): Likewise.
5057 2018-06-28  Martin Liska  <mliska@suse.cz>
5059         * Makefile.in: Add opt-suggestions.o.
5060         * gcc-main.c: Include opt-suggestions.h.
5061         * gcc.c (driver::driver): Likewise.
5062         (driver::~driver): Remove m_option_suggestions.
5063         (driver::build_option_suggestions): Moved to option_proposer.
5064         (driver::suggest_option): Likewise.
5065         (driver::handle_unrecognized_options): Use option_proposer.
5066         * gcc.h (class driver): Add new memver m_option_proposer.
5067         * opt-suggestions.c: New file.
5068         * opt-suggestions.h: New file.
5070 2018-06-28  Martin Liska  <mliska@suse.cz>
5072         * vec.h (class auto_string_vec): New (moved from auto_argvec).
5073         (auto_string_vec::~auto_string_vec): Likewise.
5075 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
5077         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5078         prevent_decl_creation_for_types fields up and add reset_location field.
5079         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5080         statement if id->reset_location is true.
5081         (copy_edges_for_bb): Do not set goto_locus on the new edges if
5082         id->reset_location is true.
5083         (copy_phis_for_bb): Force input_location on the arguments if
5084         id->reset_location is true.
5085         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5086         is set on the function to be inlined.
5088 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
5090         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5092 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
5094         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5095         registers for Pmode.
5096         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5097         hard registers for the clobbered pseudo.
5099 2018-06-27  Paul Koning  <ni1d@arrl.net>
5101         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5102         mutually exclusive options.
5103         * config/pdp11/constraints.md (h): New constraint.
5104         (O): Update definition to match shift code generation.
5105         (D): New constraint.
5106         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5107         (CCFP): Remove.
5108         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5109         function.
5110         (output_jump): Change arguments.
5111         (pdp11_fixed_cc_regs): New function.
5112         (pdp11_cc_mode): Ditto.
5113         (pdp11_expand_shift): Ditto.
5114         (pdp11_assemble_shift): Ditto.
5115         (pdp11_small_shift): Ditto.
5116         (pdp11_branch_cost): Remove.
5117         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5118         from output.
5119         (pdp11_register_move_cost): Update for CC registers.
5120         (pdp11_rtx_costs): Add case for LSHIFTRT.
5121         (pdp11_output_jump): Add CCNZ mode conditional branches.
5122         (notice_update_cc_on_set): Remove.
5123         (pdp11_cc_mode): New function.
5124         (simple_memory_operand): Correct pre/post decrement case.
5125         (no_side_effect_operand): New function.
5126         (pdp11_regno_reg_class): Add CC_REGS class.
5127         (pdp11_fixed_cc_regs): New function.
5128         (pdp11_small_shift): New function.
5129         (pdp11_expand_shift): New function to expand shift insns.
5130         (pdp11_assemble_shift): New function to output shifts.
5131         (pdp11_branch_cost): Remove.
5132         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5133         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5134         (WCHAR_TYPE): Ditto.
5135         (PTRDIFF_TYPE): Ditto.
5136         (ADJUST_INSN_LENGTH): New macro.
5137         (FIXED_REGISTERS): Add CC registers.
5138         (CALL_USED_REGISTERS): Ditto.
5139         (reg_class): Ditto.
5140         (REG_CLASS_NAMES): Ditto.
5141         (REG_CLASS_CONTENTS): Ditto.
5142         (SELECT_CC_MODE): Use new function.
5143         (TARGET_FLAGS_REGNUM): New macro.
5144         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5145         (cc0_reg_rtx): Remove.
5146         (CC_STATUS_MDEP): Remove.
5147         (CC_STATUS_MDEFP_INIT): Remove.
5148         (CC_IN_FPU): Remove.
5149         (NOTICE_UPDATE_CC): Remove.
5150         (REGISTER_NAMES): Add CC registers.
5151         (BRANCH_COST): Change to constant 1.
5152         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5153         handling.
5154         * config/pdp11/pdp11.opt (mbcopy): Remove.
5155         (mbcopy-builtin): Remove.
5156         (mbranch-cheap): Remove.
5157         (mbranch-expensive): Remove.
5158         * config/pdp11/predicates.md (expand_shift_operand): Update to
5159         match shift code generation.
5160         (ccnz_operator): New predicate.
5161         * doc/invoke.texi (PDP-11 Options): Remove deleted options
5162         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5163         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
5164         exclusive options.
5165         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
5166         description of O constraint.
5168 2018-06-27  Jeff Law  <law@redhat.com>
5169             Austin Law  <austinklaw@gmail.com>
5171         * config/v850/v850.md (addsi3_set_flags): New pattern.
5172         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5173         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5174         (zero_extendhisi2_v850_set_flags): Likewise.
5175         (zero_extendqisi2_v850_set_flags): Likewise.
5176         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5177         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5178         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5180         * config/v850/v850-protos.h (notice_update_cc): Remove.
5181         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5182         (v850_print_operand): Handle 'D' and "d".
5183         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5184         Add handling of arithmetic/logical operations compared against zero.
5185         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5186         Do not look at v850_compare_op, instead get mode from last argument.
5187         (v850_gen_compare): Remove
5188         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5189         after reload for prologue insns.
5190         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5191         patterns.
5192         (construct_save_jarl): Likewise.
5193         (TARGET_FLAGS_REGNUM): Define.
5194         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5195         (NOTICE_UPDATE_CC): Remove.
5196         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5197         than cc0.  Conditionalize on reload_completed.
5198         (cmpsi_insn, setfcc_insn): Likewise.
5199         (tst1 splitter): Turn into define_and_split which sets the flags
5200         after reload.
5201         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5202         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5203         (cstoresf4, cstoredf4): Clobber the flags.
5204         (cmpsi, cmpsf, cmpdf): Remove expanders.
5205         (setf_insn): Remove pattern.
5206         (addsi3): Turn into define_and_split which clobbers the flags after
5207         reload and a suitable pattern (addsi3_clobber_flags) for use after
5208         reload.
5209         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5210         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5211         (ashrsi3, ashrsi3_v850e2): Likewise.
5212         (bins): Clobber the flags.
5213         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5214         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5215         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5216         (call_value_internal_short, call_value_internal_long): Likewise.
5217         (callt_save_interrupt, callt_return_interrupt): Likewise.
5218         (save_interrupt, return_interrupt): Likewise.
5219         (callt_save_all_interrupt, save_all_interrupt): Likewise.
5220         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5221         (restore_all_interrupt, _restore_all_interrupt): Likewise.
5222         (All FP comparisons): Only allow after reload has completed.
5223         (trfsr): Likewise.
5224         (divh, divhu): Tweak output template.
5225         (branch_z_normal, branch_z_invert): Remove
5226         (branch_nz_normal, branch_nz_invert): Likewise.
5227         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5229         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5230         * config/v850/v850.c (notice_update_cc): Remove.
5231         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
5232         (CC_NO_CARRY): Likewise.
5233         (NOTICE_UPDATE_CC): Define to nothing.
5234         * config/v850/v850.md: Remove block comment on cc0 handling
5235         Remove "cc" attribute from all patterns.  Remove cc_status handling
5236         from all patterns.  Minor formatting fixes.
5238 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5240         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5241         (cortex-a76.cortex-a55): Likewise.
5242         * config/aarch64/aarch64-tune.md: Regenerate.
5243         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5244         cortex-a76.cortex-a55.
5246 2018-06-27  Jeff Law  <law@redhat.com>
5248         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5249         (MULTILIB_DIRNAMES): Similarly.
5251 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
5253         * gimple.h (gimple_return_retbnd): Delete.
5254         (gimple_return_set_retbnd): Likewise.
5255         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5256         gimple_return_set_retbnd.
5257         * gimple-pretty-print.c (dump_gimple_return): Remove call to
5258         gimple_return_retbnd and adjust.
5259         * tree-inline.h (struct copy_body_data): Remove retbnd field.
5260         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5261         Explicitly return NULL in a couple more cases.  Move assertion
5262         on debug statements and remove unreachable code.
5263         (reset_debug_binding): Do not test id->retbnd.
5264         (expand_call_inline): Do not set it.
5266 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5268         * configure.ac: Add --disable-gcov option.
5269         * configure: Regenerate.
5270         * Makefile.in: Honour @enable_gcov@.
5271         * doc/install.texi: Document --disable-gcov.
5273 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5275         * config/arm/arm-cpus.in (cortex-a76): New entry.
5276         (cortex-a76.cortex-a55): Likewise.
5277         * config/arm/arm-tables.opt: Regenerate.
5278         * config/arm/arm-tune.md: Likewise.
5279         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5280         * doc/invoke.texi (ARM Options): Document cortex-a76 and
5281         cortex-a76.cortex-a55.
5283 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
5285         PR target/85769
5286         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5288 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5290         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5291         comment.
5292         (EPILOGUE_USES): Likewise.
5294 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
5296         * tree-inline.c (remap_location): New function extracted from...
5297         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
5298         (copy_phis_for_bb): ...here.  Call remap_location.
5299         (copy_cfg_body): Adjust call to copy_edges_for_bb.
5301 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
5303         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5304         unaligned vsx for 16B memset.
5306 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
5308         PR target/86285
5309         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5310         ieee128_float_type_node to long_double_type_node unless
5311         TARGET_LONG_DOUBLE_128 is set.
5313 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
5315         * cfgloop.c (get_loop_location): Convert return type from
5316         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5317         by implicit construction from rtx_insn *, and using
5318         dump_user_location_t::from_function_decl for the fallback case.
5319         * cfgloop.h (get_loop_location): Convert return type from
5320         location_t to dump_user_location_t.
5321         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5322         dump_printf_loc to pass in a dump_location_t rather than a
5323         location_t, via the gimple stmt.
5324         * coverage.c (get_coverage_counts): Update calls to
5325         dump_printf_loc to pass in dump_location_t rather than a
5326         location_t.
5327         * doc/optinfo.texi (Dump types): Convert example of
5328         dump_printf_loc from taking "locus" to taking "insn".  Update
5329         description of the "_loc" calls to cover dump_location_t.
5330         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5331         "selftest.h".
5332         (dump_user_location_t::dump_user_location_t): New constructors,
5333         from gimple *stmt and rtx_insn *.
5334         (dump_user_location_t::from_function_decl): New function.
5335         (dump_loc): Make static.
5336         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5337         const dump_location_t &.
5338         (dump_generic_expr_loc): Delete.
5339         (dump_printf_loc): Convert param "loc" from location_t to
5340         const dump_location_t &.
5341         (selftest::test_impl_location): New function.
5342         (selftest::dumpfile_c_tests): New function.
5343         * dumpfile.h: Include "profile-count.h".
5344         (class dump_user_location_t): New class.
5345         (struct dump_impl_location_t): New struct.
5346         (class dump_location_t): New class.
5347         (dump_printf_loc): Convert 2nd param from source_location to
5348         const dump_location_t &.
5349         (dump_generic_expr_loc): Delete.
5350         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5351         const dump_location_t &.
5352         * gimple-fold.c (fold_gimple_assign): Update call to
5353         dump_printf_loc to pass in a dump_location_t rather than a
5354         location_t, via the gimple stmt.
5355         (gimple_fold_call): Likewise.
5356         * gimple-loop-interchange.cc
5357         (loop_cand::analyze_iloop_reduction_var): Update for change to
5358         check_reduction_path.
5359         (tree_loop_interchange::interchange): Update for change to
5360         find_loop_location.
5361         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5362         change in return-type of find_loop_location.
5363         (graphite_regenerate_ast_isl): Likewise.
5364         * graphite-optimize-isl.c (optimize_isl): Likewise.
5365         * graphite.c (graphite_transform_loops): Likewise.
5366         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5367         pass in a dump_location_t rather than a location_t, via the
5368         gimple stmt.
5369         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5370         * ipa.c (walk_polymorphic_call_targets): Likewise.
5371         * loop-unroll.c (report_unroll): Convert "locus" param from
5372         location_t to dump_location_t.
5373         (decide_unrolling): Update for change to get_loop_location's
5374         return type.
5375         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5376         location_t to dump_user_location_t.
5377         (grid_find_single_omp_among_assignments_1): Updates calls to
5378         dump_printf_loc to pass in a dump_location_t rather than a
5379         location_t, via the gimple stmt.
5380         (grid_parallel_clauses_gridifiable): Convert "tloc" from
5381         location_t to dump_location_t.  Updates calls to dump_printf_loc
5382         to pass in a dump_location_t rather than a location_t, via the
5383         gimple stmt.
5384         (grid_inner_loop_gridifiable_p): Likewise.
5385         (grid_dist_follows_simple_pattern): Likewise.
5386         (grid_gfor_follows_tiling_pattern): Likewise.
5387         (grid_target_follows_gridifiable_pattern): Likewise.
5388         (grid_attempt_target_gridification): Convert initialization
5389         of local "grid" from memset to zero-initialization; FIXME: does
5390         this require C++11?  Update call to dump_printf_loc to pass in a
5391         optinfo_location rather than a location_t, via the gimple stmt.
5392         * profile.c (read_profile_edge_counts): Updates call to
5393         dump_printf_loc to pass in a dump_location_t rather than a
5394         location_t
5395         (compute_branch_probabilities): Likewise.
5396         * selftest-run-tests.c (selftest::run_tests): Call
5397         dumpfile_c_tests.
5398         * selftest.h (dumpfile_c_tests): New decl.
5399         * tree-loop-distribution.c (pass_loop_distribution::execute):
5400         Update for change in return type of find_loop_location.
5401         * tree-parloops.c (parallelize_loops): Likewise.
5402         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5403         "locus" from location_t to dump_user_location_t.
5404         (canonicalize_loop_induction_variables): Likewise.
5405         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5406         for change in return type of find_loop_location.
5407         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5408         to dump_printf_loc to pass in a dump_location_t rather than a
5409         location_t, via the stmt.
5410         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5411         Likewise.
5412         * tree-vect-loop-manip.c (find_loop_location): Convert return
5413         type from source_location to dump_user_location_t.
5414         (vect_do_peeling): Update for above change.
5415         (vect_loop_versioning): Update for change in type of
5416         vect_location.
5417         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5418         from location_t to dump_user_location_t.
5419         (vect_estimate_min_profitable_iters): Update for change in type
5420         of vect_location.
5421         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5422         location_t to dump_location_t.
5423         (vect_slp_bb): Update for change in type of vect_location.
5424         * tree-vectorizer.c (vect_location): Convert from source_location
5425         to dump_user_location_t.
5426         (try_vectorize_loop_1): Update for change in vect_location's type.
5427         (vectorize_loops): Likewise.
5428         (increase_alignment): Likewise.
5429         * tree-vectorizer.h (vect_location): Convert from source_location
5430         to dump_user_location_t.
5431         (find_loop_location): Convert return type from source_location to
5432         dump_user_location_t.
5433         (check_reduction_path): Convert 1st param from location_t to
5434         dump_user_location_t.
5435         * value-prof.c (check_counter): Update call to dump_printf_loc to
5436         pass in a dump_user_location_t rather than a location_t; update
5437         call to error_at for change in type of "locus".
5438         (check_ic_target): Update call to dump_printf_loc to
5439         pass in a dump_user_location_t rather than a location_t, via the
5440         call_stmt.
5442 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
5444         * config/s390/s390.h (enum processor_flags): Do not use
5445         default tune parameter when -march was specified.
5447 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
5449         PR target/86314
5450         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5451         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5452         operands.
5454 2018-06-26  Richard Biener  <rguenther@suse.de>
5456         PR tree-optimization/86287
5457         PR bootstrap/86316
5458         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5459         (vect_analyze_loop): Initialize n_stmts.
5461 2018-06-26  Richard Biener  <rguenther@suse.de>
5463         PR middle-end/86271
5464         * fold-const.c (fold_convertible_p): Pointer extension
5465         isn't valid.
5467 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
5469         PR debug/86064
5470         * dwarf2out.c (loc_list_has_views): Adjust comments.
5471         (dw_loc_list): Split single cross-partition range with
5472         nonzero locview.
5474 2018-06-25  Jeff Law  <law@redhat.com>
5476         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5477         on -mbig-switch by default.
5479         * config/v850/predicates.md (const_float_1_operand): Fix match_code
5480         test.
5481         (const_float_0_operand): Remove unused predicate.
5482         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5483         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5484         (recipsf2): New expander.  Original pattern now called
5485         (recipsf2_insn).
5486         (recipdf2, recipdf2_insn): Similarly.
5487         (rsqrtsf2, rsqrtsf2_insn): Similarly
5488         (rsqrtdf2, rsqrtdf2_insn): Similarly
5490 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
5492         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5493         Simplify logic for FreeBSD (twice).
5495 2018-06-25  Martin Sebor  <msebor@redhat.com>
5497         PR tree-optimization/86204
5498         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5499         a strnlen result if it's less than the length of the string.
5501 2018-06-25  Martin Sebor  <msebor@redhat.com>
5503         PR tree-optimization/85700
5504         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5505         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5506         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5508 2018-06-25  Martin Sebor  <msebor@redhat.com>
5510         * doc/extend.texi (Zero-length arrays): Update and clarify.
5512 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
5514         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5515         added IEEE/IBM long double multilib support on PowerPC little
5516         endian Linux systems.
5517         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5518         (MULTILIB_DEFAULTS): Likewise.
5519         * config/rs6000/rs6000.c (rs6000_option_override_internal):
5520         Likewise.
5521         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5522         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5523         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5525 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
5527         PR middle-end/86311
5528         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5529         (REORDER_45): Likewise.
5531 2018-06-25  Jeff Law  <law@redhat.com>
5533         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5534         dividend to 32 bits.  Adjust length.
5535         (udivmodhi4): Cleanup output template.  Fix length.
5537 2018-06-25  Carl Love  <cel@us.ibm.com>
5539         * config/rs6000/vsx.md: Change word selector to prefered location.
5541 2018-06-25  Richard Biener  <rguenther@suse.de>
5543         PR tree-optimization/86304
5544         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5545         epilogue-if-converted loops as well.
5547 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
5549         * lto-section-out.c (lto_begin_section): Do not print section
5550         name for noaddr and unnumbered dumps.
5552 2018-06-25  Richard Biener  <rguenther@suse.de>
5554         * tree-vectorizer.h (struct vec_info_shared): New structure
5555         with parts split out from struct vec_info and loop_nest from
5556         struct _loop_vec_info.
5557         (struct vec_info): Adjust accordingly.
5558         (struct _loop_vec_info): Likewise.
5559         (LOOP_VINFO_LOOP_NEST): Adjust.
5560         (LOOP_VINFO_DATAREFS): Likewise.
5561         (LOOP_VINFO_DDRS): Likewise.
5562         (struct _bb_vec_info): Likewise.
5563         (BB_VINFO_DATAREFS): Likewise.
5564         (BB_VINFO_DDRS): Likewise.
5565         (struct _stmt_vec_info): Add dr_aux member.
5566         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5567         (DR_MISALIGNMENT_UNINITIALIZED): New.
5568         (set_dr_misalignment): Adjust.
5569         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5570         (vect_analyze_loop): Adjust prototype.
5571         (vect_analyze_loop_form): Likewise.
5572         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5573         Compute dependences lazily.
5574         (vect_record_base_alignments): Use shared datarefs/ddrs.
5575         (vect_verify_datarefs_alignment): Likewise.
5576         (vect_analyze_data_refs_alignment): Likewise.
5577         (vect_analyze_data_ref_accesses): Likewise.
5578         (vect_analyze_data_refs): Likewise.
5579         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5580         constructor parameter for shared part.
5581         (vect_analyze_loop_form): Pass in shared part and adjust.
5582         (vect_analyze_loop_2): Pass in storage for the number of
5583         stmts.  Move loop nest finding to the caller.  Compute
5584         datarefs lazily.
5585         (vect_analyze_loop): Pass in shared part.
5586         (vect_transform_loop): Verify shared datarefs are unchanged.
5587         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5588         constructor parameter for shared part.
5589         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5590         (vect_slp_bb): Verify shared datarefs are unchanged before
5591         transform.
5592         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5593         change.
5594         (new_stmt_vec_info): Initialize DR_AUX misalignment to
5595         DR_MISALIGNMENT_UNINITIALIZED.
5596         * tree-vectorizer.c (vec_info::vec_info): Add constructor
5597         parameter for shared part.
5598         (vec_info::~vec_info): Adjust.
5599         (vec_info_shared::vec_info_shared): New.
5600         (vec_info_shared::~vec_info_shared): Likewise.
5601         (vec_info_shared::save_datarefs): Likewise.
5602         (vec_info_shared::check_datarefs): Likewise.
5603         (try_vectorize_loop_1): Construct shared part live for analyses
5604         of a single loop for multiple vector sizes.
5605         * tree-parloops.c (gather_scalar_reductions): Adjust.
5607 2018-06-25  Richard Biener  <rguenther@suse.de>
5609         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5610         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5611         (vect_analyze_data_refs): Remove similar code from here and
5612         simplify accordingly.
5614 2018-06-25  Richard Biener  <rguenther@suse.de>
5616         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5617         for reverse storage order accesses rather than asserting
5618         they cannot happen here.
5620 2018-06-25  Tom de Vries  <tdevries@suse.de>
5622         PR debug/86257
5623         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5624         Use data16 instead of .byte for insn prefix.
5626 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
5628         PR C++/86082
5629         * parser.c (make_char_string_pack): Pass this literal chars
5630         through cpp_interpret_string.
5631         (cp_parser_userdef_numeric_literal): Check the result of
5632         make_char_string_pack.
5634 2018-06-24  Maya Rashish  <coypu@sdf.org>
5636         * ginclude/stddef.h: Simplify conditions around avoiding
5637         re-definition of __size_t.
5639 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
5641         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5642         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5644 2018-06-22  Maya Rashish  <coypu@sdf.org>
5646         * doc/invoke.texi (mno-fancy-math-387): Update for changes
5647         made to OpenBSD and NetBSD through the years.
5649 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5651         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5652         behavior of vec_pack (vector double, vector double) to match
5653         behavior of vec_float2 (vector double, vector double).
5655 2018-06-22  Olivier Hainque  <hainque@adacore.com>
5657         * gimplify.c (gimplify_function_tree): Prevent creation
5658         of a trampoline for the address of the current function
5659         passed to entry/exit instrumentation hooks.
5661 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
5663         PR target/86222
5664         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5665         correctly.
5667 2018-06-22  Martin Liska  <mliska@suse.cz>
5669         PR tree-optimization/86263
5670         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5671         Bail out if is_enabled is false.
5672         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5673         New declaration.
5674         (jump_table_cluster::is_enabled): New function.
5676 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
5678         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5679         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5680         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5681         (lto_input_ts_binfo_tree_pointers): Likewise.
5682         * tree-streamer-out.c (streamer_write_tree_bitfields,
5683         write_ts_binfo_tree_pointers): Likewise.
5684         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5686 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
5688         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5690 2018-06-22  Martin Liska  <mliska@suse.cz>
5692         * symbol-summary.h (get): Make it pure and inline move
5693         functionality from ::get function.
5694         (get): Remove and inline into ::get and ::get_create.
5695         (get_create): Move code from ::get function.
5697 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5699         PR target/85994
5700         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5701         -x assembler-with-cpp.
5703 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5705         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5706         _FILE_OFFSET_BITS=64 for C++.
5708 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
5710         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5711         conversion insn that shows up when pr85657-3.c is compiled using
5712         IEEE 128-bit long double.
5713         (neg<mode>2_internal): Use the correct mode to check whether the
5714         mode is IBM extended.
5715         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5716         multiply and divide external functions from being created more
5717         than once.
5719 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
5721         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5722         functions.
5723         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
5724         the edge can be forwarded.
5725         (cfg_layout_merge_blocks): Likewise.
5727 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
5729         * except.c (finish_eh_generation): Commit edge insertions only after
5730         the EH edges have been redirected from post-landing to landing pads.
5732 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
5734         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5735         create_tmp_var_for to create the FRAME decl.
5736         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5738 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
5740         * tree-inline.c (copy_edges_for_bb): Minor tweak.
5741         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5742         debug statement when resetting its value.
5743         (expand_call_inline): Copy the locus of the call onto the assignment
5744         of the return value, if any.  Use local variable in more cases.
5746 2018-06-21  Martin Liska  <mliska@suse.cz>
5748         * ipa-pure-const.c (propagate_nothrow): Use
5749         funct_state_summaries->get.
5750         (dump_malloc_lattice): Likewise.
5751         (propagate_malloc): Likewise.
5753 2018-06-21  Richard Biener  <rguenther@suse.de>
5755         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5756         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5757         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5758         comment.
5759         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5760         BLOCK_ABSTRACT_ORIGIN unconditionally.
5762 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
5764         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5765         deleting it.
5766         * ipa-reference.c (ipa_reference_c_finalize): Delete
5767         ipa_ref_opt_sum_summaries and set it to NULL.
5769 2018-06-21  Tom de Vries  <tdevries@suse.de>
5771         PR tree-optimization/85859
5772         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5773         test with comment from bb_no_side_effects_p.
5775 2018-06-21  Richard Biener  <rguenther@suse.de>
5777         PR tree-optimization/86232
5778         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5779         max for constant niter.
5781 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5783         * config/aarch64/aarch64-simd.md
5784         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5786 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5788         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5789         Make opernads of the unspec commutative.
5791 2018-06-21  Richard Biener  <rguenther@suse.de>
5793         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5794         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5795         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5796         (vect_analyze_data_ref_dependence): Re-order checks to deal with
5797         NULL DR_STEP.
5798         (vect_record_base_alignments): Do not record base alignment
5799         for gathers or scatters.
5800         (vect_compute_data_ref_alignment): Drop return value that is always
5801         true.  Bail out early for gathers or scatters.
5802         (vect_enhance_data_refs_alignment): Bail out early for gathers
5803         or scatters.
5804         (vect_find_same_alignment_drs): Likewise.
5805         (vect_analyze_data_refs_alignment): Remove dead code.
5806         (vect_slp_analyze_and_verify_node_alignment): Likewise.
5807         (vect_analyze_data_refs): For possible gathers or scatters do
5808         not create an alternate DR, just check their possible validity
5809         and mark them.  Adjust DECL_NONALIASED handling to not rely
5810         on DR_BASE_ADDRESS.
5811         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5812         update inits of gathers or scatters.
5813         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5814         Also copy gather/scatter flag to pattern vinfo.
5816 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5818         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5819         behavior of vec_packsu (vector unsigned long long, vector unsigned
5820         long long) to match behavior of vec_packs with same signature.
5822 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
5823             Thomas Schwinge <thomas@codesourcery.com>
5824             Cesar Philippidis  <cesar@codesourcery.com>
5826         * gimplify.c (gimplify_scan_omp_clauses): Add support for
5827         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5828         (gimplify_adjust_omp_clauses): Likewise.
5829         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5830         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5831         (gimplify_omp_target_update): Update handling of acc update and
5832         enter/exit data.
5833         * omp-low.c (install_var_field): Remove unused parameter
5834         base_pointers_restrict.
5835         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5836         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5837         FINALIZE}
5838         (omp_target_base_pointers_restrict_p): Delete.
5839         (scan_omp_target): Update call to scan_sharing_clauses.
5840         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5841         FINALIZE}.
5842         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5843         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5844         (convert_local_omp_clauses): Likewise.
5845         * tree-pretty-print.c (dump_omp_clause): Likewise.
5846         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
5847         FINALIZE}.
5848         (omp_clause_code_name): Likewise.
5850 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
5852         PR debug/86194
5853         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5854         be narrowed.
5856         PR tree-optimization/86231
5857         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
5858         anti-range don't overwrite *vr0min before using it to compute *vr0max.
5860 2018-06-20  Tom de Vries  <tdevries@suse.de>
5862         PR tree-optimization/86097
5863         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5864         iv type if signedness of iv type is not the same as that of *nit.
5866 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
5868         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
5869         EDGE_EH edges, verify they are all EDGE_EH.
5871 2018-06-20  Maya Rashish  <coypu@sdf.org>
5873         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5875 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5877         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5878         * config/aarch64/aarch64.c (xgene1_tunings): Add
5879         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5880         (aarch64_mode_valid_for_sched_fusion_p):
5881         Allow 16-byte modes.
5882         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5883         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5884         128-bit modes.
5885         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5886         New pattern.
5887         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
5888         * config/aarch64/iterators.md (VQ2): New mode iterator.
5890 2018-06-20  Martin Liska  <mliska@suse.cz>
5892         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5893         Change default ratio from 10 to 8.
5895 2018-06-20  Martin Liska  <mliska@suse.cz>
5897         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5898         New.
5899         (bit_test_cluster::find_bit_tests): Likewise.
5900         (switch_decision_tree::analyze_switch_statement): Find clusters.
5901         * tree-switch-conversion.h (struct jump_table_cluster): Document
5902         hierarchy.
5904 2018-06-20  Martin Liska  <mliska@suse.cz>
5906         * tree-switch-conversion.c (switch_conversion::collect):
5907         Record m_uniq property.
5908         (switch_conversion::expand): Bail out for special conditions.
5909         (group_cluster::~group_cluster): New.
5910         (group_cluster::group_cluster): Likewise.
5911         (group_cluster::dump): Likewise.
5912         (jump_table_cluster::emit): New.
5913         (switch_decision_tree::fix_phi_operands_for_edges): New.
5914         (struct case_node): Remove struct.
5915         (jump_table_cluster::can_be_handled): New.
5916         (case_values_threshold): Moved to header.
5917         (reset_out_edges_aux): Likewise.
5918         (jump_table_cluster::is_beneficial): New.
5919         (bit_test_cluster::can_be_handled): Likewise.
5920         (add_case_node): Remove.
5921         (bit_test_cluster::is_beneficial): New.
5922         (case_bit_test::cmp): New.
5923         (bit_test_cluster::emit): New.
5924         (expand_switch_as_decision_tree_p): Remove.
5925         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
5926         (fix_phi_operands_for_edge): Likewise.
5927         (switch_decision_tree::analyze_switch_statement): New.
5928         (compute_cases_per_edge): Move ...
5929         (switch_decision_tree::compute_cases_per_edge): ... here.
5930         (try_switch_expansion): Likewise.
5931         (switch_decision_tree::try_switch_expansion): Likewise.
5932         (record_phi_operand_mapping): Likewise.
5933         (switch_decision_tree::record_phi_operand_mapping): Likewise.
5934         (emit_case_decision_tree): Likewise.
5935         (switch_decision_tree::emit): Likewise.
5936         (balance_case_nodes): Likewise.
5937         (switch_decision_tree::balance_case_nodes): Likewise.
5938         (dump_case_nodes): Likewise.
5939         (switch_decision_tree::dump_case_nodes): Likewise.
5940         (emit_jump): Likewise.
5941         (switch_decision_tree::emit_jump): Likewise.
5942         (emit_cmp_and_jump_insns): Likewise.
5943         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
5944         (emit_case_nodes): Likewise.
5945         (switch_decision_tree::emit_case_nodes): Likewise.
5946         (conditional_probability): Remove.
5947         * tree-switch-conversion.h (enum cluster_type): New.
5948         (PRINT_CASE): New.
5949         (struct cluster): Likewise.
5950         (cluster::cluster): Likewise.
5951         (struct simple_cluster): Likewise.
5952         (simple_cluster::simple_cluster): Likewise.
5953         (struct group_cluster): Likewise.
5954         (struct jump_table_cluster): Likewise.
5955         (struct bit_test_cluster): Likewise.
5956         (struct min_cluster_item): Likewise.
5957         (struct case_tree_node): Likewise.
5958         (case_tree_node::case_tree_node): Likewise.
5959         (jump_table_cluster::case_values_threshold): Likewise.
5960         (struct case_bit_test): Likewise.
5961         (struct switch_decision_tree): Likewise.
5962         (struct switch_conversion): Likewise.
5963         (switch_decision_tree::reset_out_edges_aux): Likewise.
5965 2018-06-20  Martin Liska  <mliska@suse.cz>
5967         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
5968         (hoist_edge_and_branch_if_true): Likewise.
5969         (expand_switch_using_bit_tests_p): Likewise.
5970         (struct case_bit_test): Likewise.
5971         (case_bit_test_cmp): Likewise.
5972         (emit_case_bit_tests): Likewise.
5973         (switch_conversion::switch_conversion): New class.
5974         (struct switch_conv_info): Remove old struct.
5975         (collect_switch_conv_info): More to ...
5976         (switch_conversion::collect): ... this.
5977         (check_range): Likewise.
5978         (switch_conversion::check_range): Likewise.
5979         (check_all_empty_except_final): Likewise.
5980         (switch_conversion::check_all_empty_except_final): Likewise.
5981         (check_final_bb): Likewise.
5982         (switch_conversion::check_final_bb): Likewise.
5983         (create_temp_arrays): Likewise.
5984         (switch_conversion::create_temp_arrays): Likewise.
5985         (free_temp_arrays): Likewise.
5986         (gather_default_values): Likewise.
5987         (switch_conversion::gather_default_values): Likewise.
5988         (build_constructors): Likewise.
5989         (switch_conversion::build_constructors): Likewise.
5990         (constructor_contains_same_values_p): Likewise.
5991         (switch_conversion::contains_same_values_p): Likewise.
5992         (array_value_type): Likewise.
5993         (switch_conversion::array_value_type): Likewise.
5994         (build_one_array): Likewise.
5995         (switch_conversion::build_one_array): Likewise.
5996         (build_arrays): Likewise.
5997         (switch_conversion::build_arrays): Likewise.
5998         (gen_def_assigns): Likewise.
5999         (switch_conversion::gen_def_assigns): Likewise.
6000         (prune_bbs): Likewise.
6001         (switch_conversion::prune_bbs): Likewise.
6002         (fix_phi_nodes): Likewise.
6003         (switch_conversion::fix_phi_nodes): Likewise.
6004         (gen_inbound_check): Likewise.
6005         (switch_conversion::gen_inbound_check): Likewise.
6006         (process_switch): Use the newly created class.
6007         (switch_conversion::expand): New.
6008         (switch_conversion::~switch_conversion): New.
6009         * tree-switch-conversion.h: New file.
6011 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6013         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6014         tree-vect-patterns.c.
6015         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6016         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
6017         (vect_recog_sad_pattern): Likewise.
6018         (vect_recog_widen_sum_pattern): Likewise.
6019         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
6020         (vect_recog_widen_shift_pattern): Remove the type_in argument.
6021         (vect_recog_rotate_pattern): Likewise.
6022         (vect_recog_mult_pattern): Likewise.
6023         (vect_recog_vector_vector_shift_pattern): Likewise.
6024         (vect_recog_divmod_pattern): Likewise.
6025         (vect_recog_mixed_size_cond_pattern): Likewise.
6026         (vect_recog_bool_pattern): Likewise.
6027         (vect_recog_mask_conversion_pattern): Likewise.
6028         (vect_try_gather_scatter_pattern): Likewise.
6029         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
6030         (vect_recog_over_widening_pattern): Likewise.
6031         (vect_recog_gather_scatter_pattern): Likewise.
6032         (vect_recog_func_ptr): Move from tree-vectorizer.h
6033         (vect_vect_recog_func_ptrs): Move further down the file.
6034         (vect_recog_func): Likewise.  Remove the third argument.
6035         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6036         (vect_pattern_recog_1): Expect the pattern function to do any
6037         necessary target tests.  Also expect it to provide a vector type.
6038         Remove the type_in handling.
6040 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6042         * tree-vect-patterns.c (vect_pattern_detected): New function.
6043         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6044         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6045         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6046         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6047         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6048         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6049         (vect_recog_mask_conversion_pattern)
6050         (vect_try_gather_scatter_pattern): Likewise.
6052 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6054         * tree-vect-patterns.c (vect_get_internal_def): New function.
6055         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6056         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6057         (search_type_for_mask_1): Use it.
6059 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6061         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6062         redundant WIDEN_SUM_EXPR handling.
6063         (vect_recog_sad_pattern): Likewise.
6065 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6067         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6068         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6069         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6070         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6071         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6073 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6075         * tree-vect-stmts.c (vectorizable_call): Make sure that we
6076         use the stmt_vec_info of the original bb statement for the
6077         new zero assignment, even if the call is part of a pattern.
6079 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6081         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6082         that the sequence is attached to the original statement rather
6083         than the pattern statement.
6084         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6085         PATTERN_DEF_SEQ from the original statement rather than
6086         the main pattern statement.
6087         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6088         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6089         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6091 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
6093         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6094         definition statements before the early exit for statements that aren't
6095         live or relevant.
6096         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6097         split out from...
6098         (vect_transform_loop): ...here.  Process pattern definition
6099         statements without first checking whether the main pattern
6100         statement is live or relevant.
6102 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
6104         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6105         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6107 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
6109         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6110         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6111         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6112         (expand_block_compare): Change select_block_compare_mode call.
6113         (expand_strncmp_align_check): Use new functions, fix comment.
6114         (emit_final_str_compare_gpr): New function.
6115         (expand_strn_compare): Refactor and clean up code.
6116         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6118 2018-06-19  Tony Reix  <tony.reix@atos.com>
6119             Damien Bergamini  <damien.bergamini@atos.com>
6120             David Edelsohn  <dje.gcc@gmail.com>
6122         * collect2.c (static_obj): New variable.
6123         (static_libs): New variable.
6124         (is_in_list): Uncomment declaration.
6125         (main): Track AIX libraries linked statically.
6126         (is_in_list): Uncomment definition.
6127         (scan_prog_file): Don't add AIX shared libraries initializer
6128         to constructor list if linking statically.
6130 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
6132         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6133         constant.
6134         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6136 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
6138         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6139         blocks.
6141 2018-06-19  Martin Liska  <mliska@suse.cz>
6143         * config/i386/i386.c (ix86_can_inline_p): Do not use
6144         ipa_fn_summaries::get_create.
6145         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6146         get.
6147         (devirtualization_time_bonus): Likewise.
6148         (ipcp_propagate_stage): Likewise.
6149         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6150         (edge_set_predicate): Likewise.
6151         (evaluate_conditions_for_known_args): Likewise.
6152         (evaluate_properties_for_edge): Likewise.
6153         (ipa_call_summary::reset): Tranform to ...
6154         (ipa_call_summary::~ipa_call_summary): ... this.
6155         (ipa_fn_summary::reset): Transform to ...
6156         (ipa_fn_summary::~ipa_fn_summary): ... this.
6157         (ipa_fn_summary_t::remove): Rename to ...
6158         (ipa_fn_summary_t::remove_callees): ... this.
6159         (ipa_fn_summary_t::duplicate): Use placement new
6160         instead of memory copy.
6161         (ipa_call_summary_t::duplicate): Likewise.
6162         (ipa_call_summary_t::remove): Remove.
6163         (dump_ipa_call_summary): Change get_create to get.
6164         (ipa_dump_fn_summary): Dump only when summary exists.
6165         (analyze_function_body): Use symbol_summary::get instead
6166         of get_create.
6167         (compute_fn_summary): Likewise.
6168         (estimate_edge_devirt_benefit): Likewise.
6169         (estimate_edge_size_and_time): Likewise.
6170         (inline_update_callee_summaries): Likewise.
6171         (remap_edge_change_prob): Likewise.
6172         (remap_edge_summaries): Likewise.
6173         (ipa_merge_fn_summary_after_inlining): Likewise.
6174         (write_ipa_call_summary): Likewise.
6175         (ipa_fn_summary_write): Likewise.
6176         (ipa_free_fn_summary): Likewise.
6177         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6178         (struct ipa_call_summary): Likewise.
6179         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6180         of get_create.
6181         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6182         (estimate_size_after_inlining): Likewise.
6183         (estimate_growth): Likewise.
6184         (growth_likely_positive): Likewise.
6185         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6186         (inline_call): Likewise.
6187         * ipa-inline.c (caller_growth_limits): Likewise.
6188         (can_inline_edge_p): Likewise.
6189         (can_inline_edge_by_limits_p): Likewise.
6190         (compute_uninlined_call_time): Likewise.
6191         (compute_inlined_call_time): Likewise.
6192         (want_inline_small_function_p): Likewise.
6193         (edge_badness): Likewise.
6194         (update_caller_keys): Likewise.
6195         (update_callee_keys): Likewise.
6196         (inline_small_functions): Likewise.
6197         (inline_to_all_callers_1): Likewise.
6198         (dump_overall_stats): Likewise.
6199         (early_inline_small_functions): Likewise.
6200         (early_inliner): Likewise.
6201         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6202         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6203         * ipa-pure-const.c (malloc_candidate_p): Likewise.
6204         * ipa-split.c (execute_split_functions): Likewise.
6205         * symbol-summary.h: Likewise.
6206         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6208 2018-06-19  Richard Biener  <rguenther@suse.de>
6210         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6211         (vectorize_loops): ... here.  Fix dbgcnt handling.
6212         (try_vectorize_loop): Wrap try_vectorize_loop_1.
6214 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
6216         PR target/86197
6217         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6218         ieee128 argument takes up only one (vector) register, not two (floating
6219         point) registers.
6221 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
6223         * gimplify.c (gimplify_init_constructor): Really never clear for an
6224         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6226 2018-06-19  Richard Biener  <rguenther@suse.de>
6228         PR tree-optimization/86179
6229         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6230         after failed recognition.
6232 2018-06-18  Martin Sebor  <msebor@redhat.com>
6234         PR middle-end/85602
6235         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6236         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6237         Handle integer subtraction.
6238         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6239         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6241 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
6243         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6244         param from rtx to rtx_insn *.
6245         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6246         param.
6247         (frv_ifcvt_modify_insn): Likwise.
6248         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6249         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
6250         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6251         as_a <rtx_insn *> cast to local "unprotected_region" once
6252         it's been established that it's not NULL or pc_rtx.
6253         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6254         param "sethi" from rtx to rtx_insn *.
6255         (nds32_group_float_insns): Likewise for param "insn".
6256         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6257         param.
6258         (vax_output_int_subtract): Likewise.
6259         * config/vax/vax.c (vax_output_int_add): Likewise for param
6260         "insn".
6261         (vax_output_int_subtract): Likewise.
6262         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6263         (emit_pattern_after): Likewise for param "after".
6264         (emit_insn_after): Likewise.
6265         (emit_jump_insn_after): Likewise.
6266         (emit_call_insn_after): Likewise.
6267         (emit_debug_insn_after): Likewise.
6268         (emit_pattern_before): Likewise for param "before".
6269         (emit_insn_before): Likewise.
6270         (emit_jump_insn_before): Likewise.
6271         * final.c (get_insn_template): Likewise for param "insn", removing
6272         a cast.
6273         * output.h (get_insn_template): Likewise for 2nd param.
6274         * rtl.h (emit_insn_before): Likewise.
6275         (emit_jump_insn_before): Likewise.
6276         (emit_debug_insn_before_noloc): Likewise.
6277         (emit_insn_after): Likewise.
6278         (emit_jump_insn_after): Likewise.
6279         (emit_call_insn_after): Likewise.
6280         (emit_debug_insn_after): Likewise.
6281         (set_insn_deleted): Likewise for param.
6283 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
6285         PR target/85358
6286         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6287         floating point modes, so that IFmode is numerically greater than
6288         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6289         to declare the ordering.  This prevents IFmode from being
6290         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6291         machine.  Include rs6000-modes.h to share the fractional values
6292         between genmodes* and the rest of the compiler.
6293         (IFmode): Likewise.
6294         (KFmode): Likewise.
6295         (TFmode): Likewise.
6296         * config/rs6000/rs6000-modes.h: New file.
6297         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6298         meaning of rs6000_long_double_size so that 126..128 selects an
6299         appropriate 128-bit floating point type.
6300         (rs6000_option_override_internal): Likewise.
6301         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6302         (TARGET_LONG_DOUBLE_128): Change the meaning of
6303         rs6000_long_double_size so that 126..128 selects an appropriate
6304         128-bit floating point type.
6305         (LONG_DOUBLE_TYPE_SIZE): Update comment.
6306         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6307         source and destination to match the standard usage.
6308         (truncifkf2): Likewise.
6309         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6310         ISA 2.07 to use an explicit clobber, instead of passing in a
6311         temporary.
6312         (copysign<mode>3_soft): Likewise.
6314 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
6316         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6317         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6318         (vect_slp_analyze_instance_dependence): Likewise.
6319         (vect_enhance_data_refs_alignment): Likewise.
6320         (vect_analyze_data_refs_alignment): Likewise.
6321         (vect_slp_analyze_and_verify_instance_alignment
6322         (vect_analyze_data_ref_accesses): Likewise.
6323         (vect_prune_runtime_alias_test_list): Likewise.
6324         (vect_analyze_data_refs): Likewise.
6325         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6326         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6327         (vect_analyze_scalar_cycles_1): Likewise.
6328         (vect_get_loop_niters): Likewise.
6329         (vect_analyze_loop_form_1): Likewise.
6330         (vect_update_vf_for_slp): Likewise.
6331         (vect_analyze_loop_operations): Likewise.
6332         (vect_analyze_loop): Likewise.
6333         (vectorizable_induction): Likewise.
6334         (vect_transform_loop): Likewise.
6335         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6336         * tree-vect-slp.c (vect_analyze_slp): Likewise.
6337         (vect_make_slp_decision): Likewise.
6338         (vect_detect_hybrid_slp): Likewise.
6339         (vect_slp_analyze_operations): Likewise.
6340         (vect_slp_bb): Likewise.
6341         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6342         (vectorizable_bswap): Likewise.
6343         (vectorizable_call): Likewise.
6344         (vectorizable_simd_clone_call): Likewise.
6345         (vectorizable_conversion): Likewise.
6346         (vectorizable_assignment): Likewise.
6347         (vectorizable_shift): Likewise.
6348         (vectorizable_operation): Likewise.
6349         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6351 2018-06-18  Martin Sebor  <msebor@redhat.com>
6353         PR tree-optimization/81384
6354         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6355         * builtins.c (expand_builtin_strnlen): New function.
6356         (expand_builtin): Call it.
6357         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6358         * builtins.def (BUILT_IN_STRNLEN): New.
6359         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6360         Warn for bounds in excess of maximum object size.
6361         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6362         single-value ranges.  Handle strnlen.
6363         (handle_builtin_strlen): Handle strnlen.
6364         (strlen_check_and_optimize_stmt): Same.
6365         * doc/extend.texi (Other Builtins): Document strnlen.
6367 2018-06-18  Maya Rashish  <coypu@sdf.org>
6369         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6370         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6371         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6373         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6374         here to ...
6375         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6377 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6379         * tree.c (escaped_string::escape): Replace cast to char * by
6380         const_cast<char *> (unescaped).
6382 2018-06-18  Nick Clifton  <nickc@redhat.com>
6384         PR 84195
6385         * tree.c (escaped_string): New class.  Converts an unescaped
6386         string into its escaped equivalent.
6387         (warn_deprecated_use): Use the new class to convert the
6388         deprecation message, if present.
6389         (test_escaped_strings): New self test.
6390         (test_c_tests): Add test_escaped_strings.
6391         * doc/extend.texi (deprecated): Add a note that the
6392         deprecation message is affected by the -fmessage-length
6393         option, and that control characters will be escaped.
6394         (#pragma GCC error): Document this pragma.
6395         (#pragma GCC warning): Likewise.
6396         * doc/invoke.texi (-fmessage-length): Document this option's
6397         effect on the #warning and #error preprocessor directives and
6398         the deprecated attribute.
6400 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
6402         * tree.c (decl_value_expr_lookup): Revert latest change.
6403         (decl_value_expr_insert): Likewise.
6405 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
6407         * gimplify.c (nonlocal_vlas): Delete.
6408         (nonlocal_vla_vars): Likewise.
6409         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6410         referenced VLAs.
6411         (gimplify_body): Do not create and destroy nonlocal_vlas.
6412         * tree-nested.c: Include diagnostic.h.
6413         (use_pointer_in_frame): Tweak.
6414         (lookup_field_for_decl): Add assertion and declare the transformation.
6415         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6416         internal error when the reference is in a wrong context.  Do not
6417         create a debug decl by default.
6418         (note_nonlocal_block_vlas): Delete.
6419         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6420         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
6421         create a debug decl by default.
6422         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6423         call is in a wrong context.
6424         (fixup_vla_decls): New function.
6425         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
6426         debug variables were created.
6427         * tree.c (decl_value_expr_lookup): Add checking assertion.
6428         (decl_value_expr_insert): Likewise.
6430 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
6432         PR middle-end/82479
6433         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6434         * tree-scalar-evolution.c (interpret_expr): Likewise.
6435         (expression_expensive_p): Likewise.
6436         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6437         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6438         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6439         (ssa_defined_by_minus_one_stmt_p): New.
6441 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
6443         PR middle-end/64946
6444         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6445         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6446         * dojump.c (do_jump): Likewise.
6447         * expr.c (expand_expr_real_2): Check operand type's sign.
6448         * fold-const.c (const_unop): Handle ABSU_EXPR.
6449         (fold_abs_const): Likewise.
6450         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6451         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6452         (strip_sign_op_1): Likesise.
6453         * match.pd: Add new pattern to generate ABSU_EXPR.
6454         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6455         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6456         * tree-eh.c (operation_could_trap_helper_p): Likewise.
6457         * tree-inline.c (estimate_operator_cost): Likewise.
6458         * tree-pretty-print.c (dump_generic_node): Likewise.
6459         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6460         * tree.def (ABSU_EXPR): New.
6462 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
6464         PR middle-end/86095
6465         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6466         documented as preserved for backward compatibility only.
6467         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6469         PR rtl-optimization/86108
6470         * bb-reorder.c (create_forwarder_block): Renamed to ...
6471         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
6472         jump from new landing pad to the second part.
6473         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6474         Adjust callers.
6476 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
6478         PR middle-end/85878
6479         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6480         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6481         Only call store_expr for halves if the mode is the same.
6483         PR middle-end/86123
6484         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6485         Fix up comment formatting.
6487 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6489         * typed-splay-tree.h (typed_splay_tree::remove): New function.
6490         (typed_splay_tree::closure,
6491         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6492         (typed_splay_tree::typed_splay_tree,
6493         typed_splay_tree::operator =): Declared private.
6494         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6495         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6496         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6497         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6498         typed_splay_tree::splay_tree_splay,
6499         typed_splay_tree::splay_tree_foreach_helper,
6500         typed_splay_tree::splay_tree_insert,
6501         typed_splay_tree::splay_tree_remove,
6502         typed_splay_tree::splay_tree_lookup,
6503         typed_splay_tree::splay_tree_predecessor,
6504         typed_splay_tree::splay_tree_successor,
6505         typed_splay_tree::splay_tree_min,
6506         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6507         (typed_splay_tree::root, typed_splay_tree::comp,
6508         typed_splay_tree::delete_key,
6509         typed_splay_tree::delete_value): New data members.
6510         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6511         typed_splay_tree::remove.
6513 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
6515         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6516         -mginv and -mno-ginv to the assembler.
6517         * config/mips/mips.opt (-mcrc): New option.
6518         (-mginv): Likewise.
6519         * doc/invoke.text (-mcrc): Document.
6520         (-mginv): Likewise.
6522 2018-06-15  Nick Clifton  <nickc@redhat.com>
6524         PR 84195
6525         * tree.c (escaped_string): New class.  Converts an unescaped
6526         string into its escaped equivalent.
6527         (warn_deprecated_use): Use the new class to convert the
6528         deprecation message, if present.
6529         (test_escaped_strings): New self test.
6530         (test_c_tests): Add test_escaped_strings.
6531         * doc/extend.texi (deprecated): Add a note that the
6532         deprecation message is affected by the -fmessage-length
6533         option, and that control characters will be escaped.
6534         (#pragma GCC error): Document this pragma.
6535         (#pragma GCC warning): Likewise.
6536         * doc/invoke.texi (-fmessage-length): Document this option's
6537         effect on the #warning and #error preprocessor directives and
6538         the deprecated attribute.
6540 2018-06-15  Richard Biener  <rguenther@suse.de>
6542         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6543         here, also noting vector size used.
6544         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
6545         size used in MSG_OPTIMIZED_LOCATIONS dump.
6546         (pass_slp_vectorize::execute): Adjust.
6548 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
6550         PR target/85968
6551         * config/arc/arc.c (arc_return_address_register): Fix
6552         if-condition.
6554 2018-06-15  Richard Biener  <rguenther@suse.de>
6556         PR middle-end/86159
6557         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6558         leave useless conversion stripping to force_gimple_operand_gsi.
6559         (gimplify_build2): Likewise.
6560         (gimplify_build1): Likewise.
6562 2018-06-15  Richard Biener  <rguenther@suse.de>
6564         PR middle-end/86076
6565         * tree-cfg.c (move_stmt_op): unshare invariant addresses
6566         before adjusting their block.
6568 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
6570         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6571         multilibs for *-*-rtems*.
6572         * config/riscv/t-rtems: New file.
6574 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
6576         PR middle-end/86122
6577         * match.pd ((A +- CST1) +- CST2): Punt if last resort
6578         unsigned_type_for returns NULL.
6580         PR target/85945
6581         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6582         subregs of multi-word pseudos unless the float mode has word size.
6584 2018-06-14  Richard Biener  <rguenther@suse.de>
6586         PR middle-end/86139
6587         * tree-vect-generic.c (build_word_mode_vector_type): Remove
6588         duplicate and harmful type_hash_canon.
6589         * tree.c (type_hash_canon): Assert we didn't find ourselves.
6591 2018-06-14  Richard Biener  <rguenther@suse.de>
6593         PR ipa/86124
6594         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6595         NULL cgraph_node.
6597 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
6599         * config/rtems.h (STDINT_LONG32): Define.
6601 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
6602             Prachi Godbole  <prachi.godbole@imgtec.com>
6604         * config/mips/mips-cpus.def: Define P6600.
6605         * config/mips/mips-tables.opt: Regenerate.
6606         * config/mips/mips.c (mips_ucbranch_type): New enum.
6607         (mips_rtx_cost_data): Add support for P6600.
6608         (mips_issue_rate): Likewise.
6609         (mips_multipass_dfa_lookahead): Likewise.
6610         (mips_avoid_hazard): Likewise.
6611         (mips_reorg_process_insns): Likewise.
6612         (mips_classify_branch_p6600): New function.
6613         * config/mips/mips.h (TUNE_P6600): New define.
6614         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6615         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6616         * config/mips/mips.md: Include p6600.md.
6617         (processor): Add p6600.
6618         * config/mips/p6600.md: New file.
6619         * doc/invoke.texi: Add p6600 to supported architectures.
6621 2018-06-13  Martin Sebor  <msebor@redhat.com>
6623         PR tree-optimization/86114
6624         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6625         of integer types.
6626         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6628 2018-06-13  Richard Biener  <rguenther@suse.de>
6630         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6631         Properly set vector type of the intermediate stmt.
6632         * tree-vect-stmts.c (vectorizable_operation): The destination
6633         var always has vectype_out type.
6635 2018-06-13  Jeff Law  <law@redhat.com>
6637         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6638         integer 0 for argument to print_rtl_with_bb.
6639         (rl78_reorg): Likewise.
6641 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
6643         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6644         from rtx to rtx_insn *.
6645         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6646         "label".
6647         (add_sched_insns_for_speculation): Likewise for local "target",
6648         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6649         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6650         from rtx_insn ** to rtx_code_label **.
6651         (reorg_emit_nops): Likewise.
6652         (c6x_reorg): Likewise for local "call_labels".
6653         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6654         rtx to rtx_insn *.
6655         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6656         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6657         the loops over LABEL_REFS.
6658         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6659         braf_label.
6660         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6661         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6662         (split_branches): Strengthen local "olabel" from rtx to
6663         rtx_insn *, adding a safe_as_a cast.
6664         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6665         to "rtx_insn *".
6666         (add_insn_after): Likewise for first two params.
6667         (add_insn_before): Likewise.
6668         (remove_insn): Likewise for param.
6669         (emit_pattern_before_noloc): Likewise for second and third params.
6670         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6671         (emit_call_insn_before_noloc): Likewise.
6672         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6673         to "rtx_insn *".
6674         (emit_barrier_before): Likewise.
6675         (emit_label_before): Strengthen "label" param from "rtx" to
6676         "rtx_code_label *".  Strengthen "before" param from "rtx" to
6677         "rtx_insn *".
6678         (emit_insn_after_1): Strengthen "after" param from "rtx" to
6679         "rtx_insn *".
6680         (emit_pattern_after_noloc): Likewise.
6681         (emit_insn_after_noloc): Likewise.
6682         (emit_jump_insn_after_noloc): Likewise.
6683         (emit_call_insn_after_noloc): Likewise.
6684         (emit_debug_insn_after_noloc): Likewise.
6685         (emit_barrier_after): Likewise.
6686         (emit_label_after): Likewise for both params.
6687         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
6688         "loc" param from "int" to "location_t".
6689         (emit_insn_after_setloc): Likewise.
6690         (emit_jump_insn_after_setloc): Likewise.
6691         (emit_call_insn_after_setloc): Likewise.
6692         (emit_debug_insn_after_setloc): Likewise.
6693         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
6694         "loc" param from "int" to "location_t".
6695         (emit_pattern_before): Convert NULL_RTX to NULL.
6696         (emit_insn_before_setloc): Convert "loc" param from "int" to
6697         "location_t".
6698         (emit_jump_insn_before_setloc): Likewise.
6699         (emit_call_insn_before_setloc): Likewise.
6700         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6701         rtx_insn *.  Convert "loc" param from "int" to "location_t".
6702         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6703         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6704         Convert 3rd param from "int" to "location_t".
6705         (emit_barrier_before, emit_barrier_after, next_real_insn):
6706         Strengthen param from rtx to rtx_insn *.
6707         (emit_label_before): Strengthen 1st param from "rtx" to
6708         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
6709         "rtx_insn *".
6710         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6711         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6712         Strengthen 2nd param from "rtx" to "rtx_insn *".
6713         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6714         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6715         Likewise. Convert 3rd param from "int" to "location_t".
6716         (emit_label_after): Strengthen 1st param from "rtx" to
6717         "rtx_code_label *".
6718         (next_real_insn, remove_insn): Strengthen param from "rtx" to
6719         "rtx_insn *".
6720         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6721         from "rtx" to "rtx_insn *".
6723 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
6725         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6726         bodies streamed in with -Q.
6727         * dumpfile.c (dump_files): Add lto-stream-out dump file.
6728         * dumpfile.h (tree_dump_index): Add lto_stream_out.
6729         * gimple-streamer-out.c: Include gimple-pretty-print.h
6730         (output_bb): Dump stmts streamed.
6731         * lto-section-out.c: Include print-tree.h
6732         (lto_begin_section): Dump sections created.
6733         (lto_output_decl_index): Dump decl encoded.
6734         * lto-streamer-out.c: Include print-tree.h
6735         (create_output_block): Dump output block created.
6736         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6737         (output_function): Dump function output.
6738         (output_constructor): Dump constructor streamed.
6739         (write_global_stream): Output indexes encoded.
6740         (produce_asm_for_decls): Dump streams encoded.
6741         * lto-streamer.c (streamer_dump_file): New global var.
6742         * lto-streamer.h (streamer_dump_file): Declare.
6743         * passes.c (ipa_write_summaries): Initialize streamer dump.
6744         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6745         in.
6747 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
6749         PR target/86048
6750         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6751         offsets for register save directives.  Emit a second batch of save
6752         directives, if need be, when the function accesses prior frames.
6754 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6756         * config/arc/fpu.md (fmasf4): Force operand to register.
6757         (fnmasf4): Likewise.
6759 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6761         * config/arc/arc-protos.h (arc_pad_return): Remove.
6762         * config/arc/arc.c (machine_function): Remove force_short_suffix
6763         and size_reason.
6764         (arc_print_operand): Adjust printing of '&'.
6765         (arc_verify_short): Remove conditional printing of short suffix.
6766         (arc_final_prescan_insn): Remove reference to size_reason.
6767         (pad_return): New function.
6768         (arc_reorg): Call pad_return.
6769         (arc_pad_return): Remove.
6770         (arc_init_machine_status): Remove reference to force_short_suffix.
6771         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6772         (attr length): When attribute iscompact is true force to 2
6773         regardless; in the case of maybe check if we want to force the
6774         instruction to have 4 bytes length.
6775         (nopv): Change it to generate 4 byte long nop as well.
6776         (blockage): New pattern.
6777         (simple_return): Remove call to arc_pad_return.
6778         (p_return_i): Likewise.
6780 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6782         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6784 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6786         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6787         ARC cores.
6789 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6791         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6792         for ARC700 and ARCv2.
6794 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
6796         PR target/86076
6797         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6798         operands[2] instead of operands[1].
6801 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
6803         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6804         case, check whether the outer register overlaps an unallocatable
6805         register, not just whether it fits the required class.
6807 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
6809         * poly-int.h (can_div_trunc_p): Add new overload in which all values
6810         are poly_ints.
6811         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6812         (memrefs_conflict_p): Likewise.
6813         (init_alias_analysis): Likewise.
6814         * cfgexpand.c (expand_debug_expr): Likewise.
6815         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6816         * cse.c (fold_rtx): Likewise.
6817         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6818         * expr.c (emit_block_move_hints): Likewise.
6819         (clear_storage_hints, push_block, emit_push_insn): Likewise.
6820         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6821         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6822         (emit_group_store): Likewise.
6823         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
6824         to read the PRE/POST_MODIFY increment.
6825         * calls.c (store_one_arg): Use strip_offset.
6826         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6827         poly_int_rtx_p.
6828         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6829         by a VEC_SELECT.
6830         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6831         (simplify_binary_operation_1): Extend CONST_INT handling to
6832         poly_int_rtx_p.
6833         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6834         than a HOST_WIDE_INT.
6835         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6836         poly_int64.
6837         (adjust_mems, add_stores): Update accodingly.
6838         (vt_canonicalize_addr): Track polynomial offsets.
6839         (emit_note_insn_var_location): Likewise.
6840         (vt_add_function_parameter): Likewise.
6841         (vt_initialize): Likewise.
6843 2018-06-12  Jeff Law  <law@redhat.com>
6845         * config.gcc (alpha*-*-freebsd*): Remove.
6846         * config/alpha/freebsd.h: Remove.
6848 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
6850         PR other/69968
6851         * spellcheck-tree.c (levenshtein_distance): Rename to...
6852         (get_edit_distance): ...this, and update for underlying renaming.
6853         * spellcheck-tree.h (levenshtein_distance): Rename to...
6854         (get_edit_distance): ...this.
6855         * spellcheck.c (levenshtein_distance): Rename to...
6856         (get_edit_distance): ...this.  Convert from Levenshtein distance
6857         to Damerau-Levenshtein distance by supporting transpositions of
6858         adjacent characters.  Rename "v1" to "v_next" and "v0" to
6859         "v_one_ago".
6860         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6861         (selftest::test_edit_distance_unit_test_oneway): ...this, and
6862         update for underlying renaming.
6863         (selftest::levenshtein_distance_unit_test): Rename to...
6864         (selftest::test_get_edit_distance_unit): ...this, and update for
6865         underlying renaming.
6866         (selftest::test_find_closest_string): Add example from PR 69968
6867         where transposition helps
6868         (selftest::test_metric_conditions): Update for renaming.
6869         (selftest::test_metric_conditions): Likewise.
6870         (selftest::spellcheck_c_tests): Likewise.
6871         * spellcheck.h (levenshtein_distance): Rename both overloads to...
6872         (get_edit_distance): ...this.
6873         (best_match::consider): Update for renaming.
6875 2018-06-12  Martin Sebor  <msebor@redhat.com>
6877         PR tree-optimization/85259
6878         * builtins.c (compute_objsize): Handle constant offsets.
6879         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6880         true iff a warning has been issued.
6881         * gimple.h (gimple_nonartificial_location): New function.
6882         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6883         gimple_nonartificial_location and handle -Wno-system-headers.
6884         (handle_builtin_stxncpy): Same.
6886 2018-06-12  Martin Sebor  <msebor@redhat.com>
6888         PR c/85931
6889         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
6891 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
6893         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
6894         BUILTIN_VEC_XST entries for pointer to double and long long.
6896 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
6898         PR target/85990
6899         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6900         Update comments.
6901         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6902         Likewise.
6904 2018-06-12  Martin Liska  <mliska@suse.cz>
6906         * doc/options.texi: Document IntegerRange.
6908 2018-06-12  Martin Liska  <mliska@suse.cz>
6910         * config/i386/i386.opt: Make MPX-related options as Deprecated.
6911         * opt-functions.awk: Handle Deprecated flag.
6912         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6913         and report error.
6914         (read_cmdline_option): Report warning for a deprecated option.
6915         * opts.h (struct cl_option): Add new field cl_deprecated.
6916         (CL_ERR_DEPRECATED): New.
6918 2018-06-12  Martin Liska  <mliska@suse.cz>
6920         * doc/options.texi: Document Deprecated option flag.
6922 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6924         * config/arc/arc-arch.h (arc_extras): New enum.
6925         (arc_cpu_t):Add field extra.
6926         (arc_cpu_types): Consider the extras.
6927         * config/arc/arc-cpus.def: Add extras info.
6928         * config/arc/arc-opts.h (processor_type): Consider extra field.
6929         * config/arc/arc.c (arc_override_options): Handle extra field.
6931 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
6933         * config/arc/arc-arch.h: Update ARC_OPTX macro.
6934         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
6935         field.
6936         * config/arc/arc.c (arc_init): Update pic warning.
6937         (irq_range): Update irq range parsing warnings.
6938         (arc_override_options): Update various warning messages.
6939         (arc_handle_aux_attribute): Likewise.
6941 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
6943         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
6945 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6947         * doc/sourcebuild.texi: Document usage of line number 0 in verify
6948         compiler messages directives.
6950 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
6952         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
6953         * config/mips/mips-tables.opt: Regenerate.
6954         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
6955         mips64r6.
6956         * doc/invoke.texi: Document -march=i6500.
6958 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
6960         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
6961         (i6400_gpmul): Add cpu_unit.
6962         (i6400_gpdiv): Likewise.
6963         (i6400_msa_add_d): Update reservations.
6964         (i6400_msa_int_add) Likewise.
6965         (i6400_msa_short_logic3) Likewise.
6966         (i6400_msa_short_logic2) Likewise.
6967         (i6400_msa_short_logic) Likewise.
6968         (i6400_msa_move) Likewise.
6969         (i6400_msa_cmp) Likewise.
6970         (i6400_msa_short_float2) Likewise.
6971         (i6400_msa_div_d) Likewise.
6972         (i6400_msa_long_logic1) Likewise.
6973         (i6400_msa_long_logic2) Likewise.
6974         (i6400_msa_mult) Likewise.
6975         (i6400_msa_long_float2) Likewise.
6976         (i6400_msa_long_float4) Likewise.
6977         (i6400_msa_long_float5) Likewise.
6978         (i6400_msa_long_float8) Likewise.
6979         (i6400_fpu_fadd): Include frint type.
6980         (i6400_fpu_store): New define_insn_reservation.
6981         (i6400_fpu_load): Likewise.
6982         (i6400_fpu_move): Likewise.
6983         (i6400_fpu_fcmp): Likewise.
6984         (i6400_fpu_fmadd): Likewise.
6985         (i6400_int_mult): Include imul3nc type and update reservation.
6986         (i6400_int_div): Include idiv3 type and update reservation.
6987         (i6400_int_load): Update to check type not move_type.
6988         (i6400_int_store): Likewise.
6989         (i6400_int_prefetch): Set zero latency.
6991 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
6993         * gcc.c: Document new %@{...} sequence.
6994         (LINK_COMMAND_SPEC): Use it for the -L switches.
6995         (cpp_unique_options): Use it for the -I switches.
6996         (at_file_argbuf): New global variable.
6997         (in_at_file): Likewise.
6998         (alloc_args): Create at_file_argbuf.
6999         (clear_args): Truncate at_file_argbuf.
7000         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7001         (open_at_file): New function.
7002         (close_at_file): Likewise.
7003         (create_at_file): Delete.
7004         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7005         <'o'>: Likewise.
7006         <'@'>: New case.
7007         (validate_switches_from_spec): Deal with %@{...} sequence.
7008         (validate_switches): Likewise.
7009         (driver::finalize): Call clear_args.
7011 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
7013         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7015 2018-06-11  Martin Sebor  <msebor@redhat.com>
7017         * doc/invoke.texi (-Wall): List -Wc++17-compat.
7018         (Wno-class-memaccess): Add @opindex.
7019         (Wno-templates, Wno-multiple-inheritance): Same.
7020         (Wno-virtual-inheritance, Wno-namespaces): Same.
7021         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7022         (Wno-format-overflow, Wno-format-truncation): Same.
7023         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7024         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7025         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7026         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7027         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7028         (Wno-misspelled-isr): Same.
7030 2018-06-11  Martin Sebor  <msebor@redhat.com>
7032         * PR tree-optimization/86083
7033         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7035 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
7037         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7039 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
7041         PR target/85755
7042         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7043         on the correct operand.
7044         (*movdi_internal64): Ditto.
7046 2018-06-11  Martin Liska  <mliska@suse.cz>
7048         PR tree-optimization/86089
7049         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7051 2018-06-11  Julia Koval  <julia.koval@intel.com>
7053         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7054         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7055         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7057 2018-06-11  Olivier Hainque  <hainque@adacore.com>
7059         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7060         for Ada with strict dwarf2.
7062 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
7064         PR target/85755
7065         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7066         addresses.
7068 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
7070         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7072 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
7074         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7075         TARGET_ELF.
7077 2018-06-08  Martin Liska  <mliska@suse.cz>
7079         * tree-cfg.h (debug_function): Fix argument type to match
7080         implementation.
7082 2018-06-08  Martin Liska  <mliska@suse.cz>
7084         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7085         Remove usage of MPX-related (and removed) fields.
7086         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7088 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
7090         * cfg.c (debug): Use TDF_NONE rather than 0.
7091         * cfghooks.c (debug): Likewise.
7092         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7093         (struct dump_option_value_info): Convert to...
7094         (struct kv_pair): ...this template type.
7095         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7096         rather than 0.
7097         (optinfo_verbosity_options): Likewise.
7098         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7099         OPTGROUP_NONE.
7100         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7101         than int for "optgroup_flags" param.
7102         (dump_generic_expr_loc): Use dump_flags_t rather than int for
7103         "dump_kind" param.
7104         (dump_dec): Likewise.
7105         (dump_finish): Use TDF_NONE rather than 0.
7106         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7107         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
7108         than 0.  Update for change to option_ptr.
7109         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7110         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
7111         0.  Update for changes to optinfo_verbosity_options and
7112         optgroup_options.
7113         (opt_info_switch_p): Convert optgroup_flags from int to
7114         optgroup_flags_t.
7115         (dump_basic_block): Use dump_flags_t rather than int
7116         for "dump_kind" param.
7117         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7118         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7119         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7120         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7121         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7122         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7123         TDF_NONE): Convert from macros to...
7124         (enum dump_flag): ...this new enum.
7125         (dump_flags_t): Update to use enum.
7126         (operator|, operator&, operator~, operator|=, operator&=):
7127         Implement for dump_flags_t.
7128         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7129         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7130         Convert from macros to...
7131         (enum optgroup_flag): ...this new enum.
7132         (optgroup_flags_t): New typedef.
7133         (operator|, operator|=): Implement for optgroup_flags_t.
7134         (struct dump_file_info): Convert field "alt_flags" to
7135         dump_flags_t.  Convert field "optgroup_flags" to
7136         optgroup_flags_t.
7137         (dump_basic_block): Use dump_flags_t rather than int for param.
7138         (dump_generic_expr_loc): Likewise.
7139         (dump_dec): Likewise.
7140         (dump_register): Convert param "optgroup_flags" to
7141         optgroup_flags_t.
7142         (opt_info_enable_passes): Likewise.
7143         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7144         than 0.
7145         * gimple-pretty-print.c (debug): Likewise.
7146         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7147         (merged_store_group::apply_stores): Likewise.
7148         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7149         * gimple.c (verify_gimple_pp): Likewise.
7150         * graphite-poly.c (print_pbb_body): Likewise.
7151         * passes.c (pass_manager::register_one_dump_file): Convert
7152         local "optgroup_flags" to optgroup_flags_t.
7153         * print-tree.c (print_node): Use TDF_NONE rather than 0.
7154         (debug): Likewise.
7155         (debug_body): Likewise.
7156         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7157         to optgroup_flags_t.
7158         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7159         than 0.
7160         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7161         (convert_mult_to_fma): Likewise.
7162         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7163         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7164         * tree-vect-data-refs.c (dump_lower_bound): Convert param
7165         "dump_kind" to dump_flags_t.
7167 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
7169         * config/rs6000/rs6000.c (min, max): Delete.
7171 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
7173         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7174         -mabi=spe and -mabi=no-spe.
7176 2018-06-08  Martin Liska  <mliska@suse.cz>
7178         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7179         where we expect an existing summary.
7181 2018-06-08  Martin Liska  <mliska@suse.cz>
7183         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7184         * ipa-inline.h (estimate_edge_growth): Likewise.
7186 2018-06-08  Martin Liska  <mliska@suse.cz>
7188         * cgraph.c (function_version_hasher::hash): Use
7189         cgraph_node::get_uid ().
7190         (function_version_hasher::equal):
7191         * cgraph.h (cgraph_node::get_uid): New method.
7192         * ipa-inline.c (update_caller_keys): Use
7193         cgraph_node::get_uid ().
7194         (update_callee_keys): Likewise.
7195         * ipa-utils.c (searchc): Likewise.
7196         (ipa_reduced_postorder): Likewise.
7197         * lto-cgraph.c (input_node): Likewise.
7198         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7199         * symbol-summary.h (symtab_insertion): Likewise.
7200         (symtab_removal): Likewise.
7201         (symtab_duplication): Likewise.
7202         * tree-pretty-print.c (dump_function_header): Likewise.
7203         * tree-sra.c (convert_callers_for_node): Likewise.
7205 2018-06-08  Martin Liska  <mliska@suse.cz>
7207         * cgraph.c (symbol_table::create_edge): Always assign a new
7208         unique number.
7209         (symbol_table::free_edge): Do not recycle numbers.
7210         * cgraph.h (cgraph_edge::get): New method.
7211         * symbol-summary.h (symtab_removal): Use it.
7212         (symtab_duplication): Likewise.
7213         (call_summary::hashable_uid): Remove.
7215 2018-06-08  Martin Liska  <mliska@suse.cz>
7217         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7218         (initialize_growth_caches): Remove.
7219         (free_growth_caches): Likewise.
7220         (do_estimate_edge_time): Use edge_growth_cache.
7221         (do_estimate_edge_size): Likewise.
7222         (do_estimate_edge_hints): Likewise.
7223         * ipa-inline.c (reset_edge_caches): Likewise.
7224         (recursive_inlining): Likewise.
7225         (inline_small_functions): Likewise.
7226         * ipa-inline.h (initialize_growth_caches): Remove.
7227         (estimate_edge_size): Likewise.
7228         (estimate_edge_time): Likewise.
7229         (estimate_edge_hints): Likewise.
7230         (reset_edge_growth_cache): Likewise.
7231         * symbol-summary.h (call_summary::remove): New method.
7233 2018-06-08  Martin Liska  <mliska@suse.cz>
7235         * ipa-cp.c (class edge_clone_summary): New summary.
7236         (grow_edge_clone_vectors): Remove.
7237         (ipcp_edge_duplication_hook): Remove.
7238         (class edge_clone_summary_t): New call_summary class.
7239         (ipcp_edge_removal_hook): Remove.
7240         (edge_clone_summary_t::duplicate): New function.
7241         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7242         (create_specialized_node): Likewise.
7243         (ipcp_driver): Initialize edge_clone_summaries and do not
7244         register hooks.
7246 2018-06-08  Martin Liska  <mliska@suse.cz>
7248         * symbol-summary.h (get): New function.
7249         (call_summary::m_initialize_when_cloning): New class member.
7251 2018-06-08  Martin Liska  <mliska@suse.cz>
7253         * cgraph.c (cgraph_node::remove): Do not recycle uid.
7254         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7255         (symbol_table::allocate_cgraph_symbol): Do not set uid.
7256         * passes.c (uid_hash_t): Record removed_nodes by their uids.
7257         (remove_cgraph_node_from_order): Use the removed_nodes set.
7258         (do_per_function_toporder): Likwise.
7259         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7260         instead of summary_uid.
7261         (symtab_removal): Likewise.
7262         (symtab_duplication): Likewise.
7264 2018-06-08  Martin Liska  <mliska@suse.cz>
7266         * ipa-cp.c (ipcp_store_bits_results): Use
7267         ipcp_transformation_sum.
7268         (ipcp_store_vr_results): Likewise.
7269         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7270         to ...
7271         (ipcp_transformation_initialize): ... this.
7272         (ipa_set_node_agg_value_chain):
7273         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7274         (write_ipcp_transformation_info): Likewise.
7275         (read_ipcp_transformation_info): Likewise.
7276         (ipcp_update_bits): Likewise.
7277         (ipcp_update_vr): Likewise.
7278         (ipcp_transform_function): Likewise.
7279         * ipa-prop.h: Rename ipcp_transformation_summary to
7280         ipcp_transformation.
7281         (class ipcp_transformation_t): New function summary.
7282         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7283         (ipa_get_agg_replacements_for_node): Likewise.
7285 2018-06-08  Martin Liska  <mliska@suse.cz>
7287         * ipa-pure-const.c (struct funct_state_d): Do it class instead
7288         of struct.
7289         (class funct_state_summary_t): New function_summary class.
7290         (has_function_state): Remove.
7291         (get_function_state): Likewise.
7292         (set_function_state): Likewise.
7293         (add_new_function): Likewise.
7294         (funct_state_summary_t::insert): New function.
7295         (duplicate_node_data): Remove.
7296         (remove_node_data): Remove.
7297         (funct_state_summary_t::duplicate): New function.
7298         (register_hooks): Create new funct_state_summaries.
7299         (pure_const_generate_summary): Use it.
7300         (pure_const_write_summary): Likewise.
7301         (pure_const_read_summary): Likewise.
7302         (propagate_pure_const): Likewise.
7303         (propagate_nothrow): Likewise.
7304         (dump_malloc_lattice): Likewise.
7305         (propagate_malloc): Likewise.
7306         (execute): Do not register hooks, just remove summary
7307         instead.
7308         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7309         constructor.
7311 2018-06-08  Martin Liska  <mliska@suse.cz>
7313         * ipa-reference.c (remove_node_data): Remove.
7314         (duplicate_node_data): Likewise.
7315         (class ipa_ref_var_info_summary_t): New class.
7316         (class ipa_ref_opt_summary_t): Likewise.
7317         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7318         (get_reference_optimization_summary): Use
7319         ipa_ref_opt_sum_summaries.
7320         (set_reference_vars_info): Remove.
7321         (set_reference_optimization_summary): Likewise.
7322         (ipa_init): Create summaries.
7323         (init_function_info): Use function summary.
7324         (ipa_ref_opt_summary_t::duplicate): New function.
7325         (ipa_ref_opt_summary_t::remove): New function.
7326         (get_read_write_all_from_node): Fix GNU coding style.
7327         (propagate): Use function summary.
7328         (write_node_summary_p): Fix GNU coding style.
7329         (stream_out_bitmap): Likewise.
7330         (ipa_reference_read_optimization_summary): Use function summary.
7331         (ipa_reference_c_finalize): Do not release hooks.
7333 2018-06-08  Martin Liska  <mliska@suse.cz>
7335         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7336         (analyze_function_body): Extract multiple calls of get_create.
7337         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7338         * ipa-inline.c (recursive_inlining): Use ::get method.
7339         * ipa-inline.h (estimate_edge_growth): Likewise.
7341 2018-06-08  Martin Liska  <mliska@suse.cz>
7343         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7344         HSA_INVALID.
7345         (hsa_function_summary::hsa_function_summary): Use the new enum
7346         value.
7347         (hsa_gpu_implementation_p): Use hsa_summaries::get.
7348         * hsa-gen.c (hsa_get_host_function): Likewise.
7349         (get_brig_function_name): Likewise.
7350         * ipa-hsa.c (process_hsa_functions): Likewise.
7351         (ipa_hsa_write_summary): Likewise.
7352         * symbol-summary.h (symtab_duplication): Use ::get function/
7353         (get): New function.
7355 2018-06-08  Martin Liska  <mliska@suse.cz>
7357         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7358         of get.
7359         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7360         (hsa_register_kernel): Likewise.
7361         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7362         * hsa-gen.c (hsa_get_host_function): Likewise.
7363         (get_brig_function_name): Likewise.
7364         (generate_hsa): Likewise.
7365         (pass_gen_hsail::execute): Likewise.
7366         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7367         (devirtualization_time_bonus): Likewise.
7368         (ipcp_propagate_stage): Likewise.
7369         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7370         (edge_set_predicate): Likewise.
7371         (evaluate_conditions_for_known_args): Likewise.
7372         (evaluate_properties_for_edge): Likewise.
7373         (ipa_fn_summary::reset): Likewise.
7374         (ipa_fn_summary_t::duplicate): Likewise.
7375         (dump_ipa_call_summary): Likewise.
7376         (ipa_dump_fn_summary): Likewise.
7377         (analyze_function_body): Likewise.
7378         (compute_fn_summary): Likewise.
7379         (estimate_edge_devirt_benefit): Likewise.
7380         (estimate_edge_size_and_time): Likewise.
7381         (estimate_calls_size_and_time): Likewise.
7382         (estimate_node_size_and_time): Likewise.
7383         (inline_update_callee_summaries): Likewise.
7384         (remap_edge_change_prob): Likewise.
7385         (remap_edge_summaries): Likewise.
7386         (ipa_merge_fn_summary_after_inlining): Likewise.
7387         (ipa_update_overall_fn_summary): Likewise.
7388         (read_ipa_call_summary): Likewise.
7389         (inline_read_section): Likewise.
7390         (write_ipa_call_summary): Likewise.
7391         (ipa_fn_summary_write): Likewise.
7392         (ipa_free_fn_summary): Likewise.
7393         * ipa-hsa.c (process_hsa_functions): Likewise.
7394         (ipa_hsa_write_summary): Likewise.
7395         (ipa_hsa_read_section): Likewise.
7396         * ipa-icf.c (sem_function::merge): Likewise.
7397         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7398         (do_estimate_edge_time): Likewise.
7399         (estimate_size_after_inlining): Likewise.
7400         (estimate_growth): Likewise.
7401         (growth_likely_positive): Likewise.
7402         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7403         (inline_call): Likewise.
7404         * ipa-inline.c (caller_growth_limits): Likewise.
7405         (can_inline_edge_p): Likewise.
7406         (can_inline_edge_by_limits_p): Likewise.
7407         (compute_uninlined_call_time): Likewise.
7408         (compute_inlined_call_time): Likewise.
7409         (want_inline_small_function_p): Likewise.
7410         (edge_badness): Likewise.
7411         (update_caller_keys): Likewise.
7412         (update_callee_keys): Likewise.
7413         (recursive_inlining): Likewise.
7414         (inline_small_functions): Likewise.
7415         (inline_to_all_callers_1): Likewise.
7416         (dump_overall_stats): Likewise.
7417         (early_inline_small_functions): Likewise.
7418         (early_inliner): Likewise.
7419         * ipa-inline.h (estimate_edge_growth): Likewise.
7420         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7421         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7422         * ipa-prop.h (IPA_NODE_REF): Likewise.
7423         (IPA_EDGE_REF): Likewise.
7424         * ipa-pure-const.c (malloc_candidate_p): Likewise.
7425         (propagate_malloc): Likewise.
7426         * ipa-split.c (execute_split_functions): Likewise.
7427         * symbol-summary.h: Rename get to get_create.
7428         (get): Likewise.
7429         (get_create): Likewise.
7430         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7432 2018-06-08  Martin Liska  <mliska@suse.cz>
7434         * symbol-summary.h (release): Move definition out of class
7435         declaration.
7436         (symtab_removal): Likewise.
7437         (symtab_duplication): Likewise.
7439 2018-06-08  Martin Liska  <mliska@suse.cz>
7441         * symbol-summary.h (function_summary): Move constructor
7442         implementation out of class declaration.
7443         (release): Likewise.
7444         (symtab_insertion): Likewise.
7445         (symtab_removal): Likewise.
7446         (symtab_duplication): Likewise.
7447         (get): Likewise.
7449 2018-06-08  Martin Liska  <mliska@suse.cz>
7451         * Makefile.in: Remove support for MPX (macros, related functions,
7452         fields in cgraph_node, ...).
7453         * builtin-types.def (BT_BND): Likewise.
7454         (BT_FN_BND_CONST_PTR): Likewise.
7455         (BT_FN_CONST_PTR_BND): Likewise.
7456         (BT_FN_VOID_PTR_BND): Likewise.
7457         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7458         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7459         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7460         (expand_builtin_mempcpy_with_bounds): Likewise.
7461         (expand_builtin_memset_with_bounds): Likewise.
7462         (expand_builtin_memset_args): Likewise.
7463         (std_expand_builtin_va_start): Likewise.
7464         (expand_builtin): Likewise.
7465         (expand_builtin_with_bounds): Likewise.
7466         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7467         (DEF_LIB_BUILTIN_CHKP): Likewise.
7468         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7469         (DEF_CHKP_BUILTIN): Likewise.
7470         (BUILT_IN_MEMCPY): Likewise.
7471         (BUILT_IN_MEMMOVE): Likewise.
7472         (BUILT_IN_MEMPCPY): Likewise.
7473         (BUILT_IN_MEMSET): Likewise.
7474         (BUILT_IN_STPCPY): Likewise.
7475         (BUILT_IN_STRCAT): Likewise.
7476         (BUILT_IN_STRCHR): Likewise.
7477         (BUILT_IN_STRCPY): Likewise.
7478         (BUILT_IN_STRLEN): Likewise.
7479         (BUILT_IN_MEMCPY_CHK): Likewise.
7480         (BUILT_IN_MEMMOVE_CHK): Likewise.
7481         (BUILT_IN_MEMPCPY_CHK): Likewise.
7482         (BUILT_IN_MEMSET_CHK): Likewise.
7483         (BUILT_IN_STPCPY_CHK): Likewise.
7484         (BUILT_IN_STRCAT_CHK): Likewise.
7485         (BUILT_IN_STRCPY_CHK): Likewise.
7486         * calls.c (store_bounds): Likewise.
7487         (emit_call_1): Likewise.
7488         (special_function_p): Likewise.
7489         (maybe_warn_nonstring_arg): Likewise.
7490         (initialize_argument_information): Likewise.
7491         (finalize_must_preallocate): Likewise.
7492         (compute_argument_addresses): Likewise.
7493         (expand_call): Likewise.
7494         * cfgexpand.c (expand_call_stmt): Likewise.
7495         (expand_return): Likewise.
7496         (expand_gimple_stmt_1): Likewise.
7497         (pass_expand::execute): Likewise.
7498         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7499         (cgraph_node::remove): Likewise.
7500         (cgraph_node::dump): Likewise.
7501         (cgraph_node::verify_node): Likewise.
7502         * cgraph.h (chkp_function_instrumented_p): Likewise.
7503         (symtab_node::get_alias_target): Likewise.
7504         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7505         (cgraph_local_p): Likewise.
7506         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7507         (cgraph_edge::rebuild_references): Likewise.
7508         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7509         (walk_polymorphic_call_targets): Likewise.
7510         (cgraph_node::expand_thunk): Likewise.
7511         (symbol_table::output_weakrefs): Likewise.
7512         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7513         (ix86_handle_option): Likewise.
7514         * config/i386/constraints.md: Likewise.
7515         * config/i386/i386-builtin-types.def (BND): Likewise.
7516         (VOID): Likewise.
7517         (PVOID): Likewise.
7518         (ULONG): Likewise.
7519         * config/i386/i386-builtin.def (BDESC_END): Likewise.
7520         (BDESC_FIRST): Likewise.
7521         (BDESC): Likewise.
7522         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7523         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7524         * config/i386/i386.c (enum reg_class): Likewise.
7525         (ix86_target_string): Likewise.
7526         (ix86_option_override_internal): Likewise.
7527         (ix86_conditional_register_usage): Likewise.
7528         (ix86_valid_target_attribute_inner_p): Likewise.
7529         (ix86_set_indirect_branch_type): Likewise.
7530         (ix86_set_current_function): Likewise.
7531         (ix86_function_arg_regno_p): Likewise.
7532         (init_cumulative_args): Likewise.
7533         (ix86_function_arg_advance): Likewise.
7534         (ix86_function_arg): Likewise.
7535         (ix86_pass_by_reference): Likewise.
7536         (ix86_function_value_regno_p): Likewise.
7537         (ix86_function_value_1): Likewise.
7538         (ix86_function_value_bounds): Likewise.
7539         (ix86_return_in_memory): Likewise.
7540         (ix86_setup_incoming_vararg_bounds): Likewise.
7541         (ix86_va_start): Likewise.
7542         (indirect_thunk_need_prefix): Likewise.
7543         (print_reg): Likewise.
7544         (ix86_print_operand): Likewise.
7545         (ix86_expand_call): Likewise.
7546         (ix86_output_function_return): Likewise.
7547         (reg_encoded_number): Likewise.
7548         (BDESC_VERIFYS): Likewise.
7549         (ix86_init_mpx_builtins): Likewise.
7550         (ix86_init_builtins): Likewise.
7551         (ix86_emit_cmove): Likewise.
7552         (ix86_emit_move_max): Likewise.
7553         (ix86_expand_builtin): Likewise.
7554         (ix86_builtin_mpx_function): Likewise.
7555         (ix86_get_arg_address_for_bt): Likewise.
7556         (ix86_load_bounds): Likewise.
7557         (ix86_store_bounds): Likewise.
7558         (ix86_load_returned_bounds): Likewise.
7559         (ix86_store_returned_bounds): Likewise.
7560         (ix86_class_likely_spilled_p): Likewise.
7561         (ix86_hard_regno_mode_ok): Likewise.
7562         (x86_order_regs_for_local_alloc): Likewise.
7563         (ix86_mitigate_rop): Likewise.
7564         (ix86_bnd_prefixed_insn_p): Likewise.
7565         (ix86_mpx_bound_mode): Likewise.
7566         (ix86_make_bounds_constant): Likewise.
7567         (ix86_initialize_bounds): Likewise.
7568         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7569         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7570         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7571         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7572         (TARGET_CHKP_BOUND_MODE): Likewise.
7573         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7574         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7575         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7576         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7577         * config/i386/i386.h (TARGET_MPX): Likewise.
7578         (TARGET_MPX_P): Likewise.
7579         (VALID_BND_REG_MODE): Likewise.
7580         (FIRST_BND_REG): Likewise.
7581         (LAST_BND_REG): Likewise.
7582         (enum reg_class): Likewise.
7583         (BND_REG_P): Likewise.
7584         (BND_REGNO_P): Likewise.
7585         (BNDmode): Likewise.
7586         (ADJUST_INSN_LENGTH): Likewise.
7587         * config/i386/i386.md: Likewise.
7588         * config/i386/i386.opt: Likewise.
7589         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7590         (defined): Likewise.
7591         (LINK_MPX): Likewise.
7592         (MPX_SPEC): Likewise.
7593         (LIBMPX_SPEC): Likewise.
7594         (LIBMPXWRAPPERS_SPEC): Likewise.
7595         (CHKP_SPEC): Likewise.
7596         * config/i386/predicates.md: Likewise.
7597         * dbxout.c (dbxout_type): Likewise.
7598         * doc/extend.texi: Likewise.
7599         * doc/invoke.texi: Likewise.
7600         * doc/md.texi: Likewise.
7601         * doc/tm.texi: Likewise.
7602         * doc/tm.texi.in: Likewise.
7603         * dwarf2out.c (is_base_type): Likewise.
7604         (gen_formal_types_die): Likewise.
7605         (gen_subprogram_die): Likewise.
7606         (gen_type_die_with_usage): Likewise.
7607         (gen_decl_die): Likewise.
7608         (dwarf2out_late_global_decl): Likewise.
7609         * expr.c (expand_assignment): Likewise.
7610         (emit_storent_insn): Likewise.
7611         (store_expr_with_bounds): Likewise.
7612         (store_expr): Likewise.
7613         (expand_expr_real_1): Likewise.
7614         * expr.h (store_expr_with_bounds): Likewise.
7615         * function.c (use_register_for_decl): Likewise.
7616         (struct bounds_parm_data): Likewise.
7617         (assign_parms_augmented_arg_list): Likewise.
7618         (assign_parm_find_entry_rtl): Likewise.
7619         (assign_parm_is_stack_parm): Likewise.
7620         (assign_parm_load_bounds): Likewise.
7621         (assign_bounds): Likewise.
7622         (assign_parms): Likewise.
7623         (expand_function_start): Likewise.
7624         * gcc.c (CHKP_SPEC): Likewise.
7625         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7626         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7627         (wrestrict_dom_walker::check_call): Likewise.
7628         * gimple.c (gimple_build_call_from_tree): Likewise.
7629         * gimple.h (enum gf_mask): Likewise.
7630         (gimple_call_with_bounds_p): Likewise.
7631         (gimple_call_set_with_bounds): Likewise.
7632         * gimplify.c (gimplify_init_constructor): Likewise.
7633         * ipa-cp.c (initialize_node_lattices): Likewise.
7634         (propagate_constants_across_call): Likewise.
7635         (find_more_scalar_values_for_callers_subset): Likewise.
7636         * ipa-hsa.c (process_hsa_functions): Likewise.
7637         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7638         * ipa-icf.c (sem_function::merge): Likewise.
7639         * ipa-inline.c (early_inliner): Likewise.
7640         * ipa-pure-const.c (warn_function_noreturn): Likewise.
7641         (warn_function_cold): Likewise.
7642         (propagate_pure_const): Likewise.
7643         * ipa-ref.h (enum GTY): Likewise.
7644         * ipa-split.c (find_retbnd): Likewise.
7645         (consider_split): Likewise.
7646         (split_function): Likewise.
7647         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7648         * ipa.c (walk_polymorphic_call_targets): Likewise.
7649         (symbol_table::remove_unreachable_nodes): Likewise.
7650         (process_references): Likewise.
7651         (cgraph_build_static_cdtor_1): Likewise.
7652         * lto-cgraph.c (lto_output_node): Likewise.
7653         (output_refs): Likewise.
7654         (compute_ltrans_boundary): Likewise.
7655         (input_overwrite_node): Likewise.
7656         (input_node): Likewise.
7657         (input_cgraph_1): Likewise.
7658         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7659         * passes.c (pass_manager::execute_early_local_passes): Likewise.
7660         (class pass_chkp_instrumentation_passes): Likewise.
7661         (make_pass_chkp_instrumentation_passes): Likewise.
7662         * passes.def: Likewise.
7663         * rtl.h (struct GTY): Likewise.
7664         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7665         * stor-layout.c (layout_type): Likewise.
7666         * symtab.c: Likewise.
7667         * target.def: Likewise.
7668         * targhooks.c (default_chkp_bound_type): Likewise.
7669         (default_chkp_bound_mode): Likewise.
7670         (default_builtin_chkp_function): Likewise.
7671         (default_chkp_function_value_bounds): Likewise.
7672         (default_chkp_make_bounds_constant): Likewise.
7673         (default_chkp_initialize_bounds): Likewise.
7674         * targhooks.h (default_chkp_bound_type): Likewise.
7675         (default_chkp_bound_mode): Likewise.
7676         (default_builtin_chkp_function): Likewise.
7677         (default_chkp_function_value_bounds): Likewise.
7678         (default_chkp_make_bounds_constant): Likewise.
7679         (default_chkp_initialize_bounds): Likewise.
7680         * toplev.c (compile_file): Likewise.
7681         (process_options): Likewise.
7682         * tree-core.h (DEF_BUILTIN): Likewise.
7683         (DEF_BUILTIN_CHKP): Likewise.
7684         * tree-inline.c (declare_return_variable): Likewise.
7685         (remap_gimple_stmt): Likewise.
7686         (copy_bb): Likewise.
7687         (initialize_inlined_parameters): Likewise.
7688         (expand_call_inline): Likewise.
7689         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7690         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7691         (make_pass_ipa_chkp_produce_thunks): Likewise.
7692         (make_pass_chkp): Likewise.
7693         (make_pass_chkp_opt): Likewise.
7694         (make_pass_chkp_instrumentation_passes): Likewise.
7695         * tree-pretty-print.c (dump_generic_node): Likewise.
7696         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7697         * tree-ssa-dce.c (propagate_necessity): Likewise.
7698         (eliminate_unnecessary_stmts): Likewise.
7699         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7700         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7701         * tree-ssa-sccvn.h: Likewise.
7702         * tree-ssa-strlen.c (get_string_length): Likewise.
7703         (valid_builtin_call): Likewise.
7704         (adjust_last_stmt): Likewise.
7705         (handle_builtin_strchr): Likewise.
7706         (handle_builtin_strcpy): Likewise.
7707         (handle_builtin_stxncpy): Likewise.
7708         (handle_builtin_memcpy): Likewise.
7709         (handle_builtin_strcat): Likewise.
7710         (strlen_check_and_optimize_stmt): Likewise.
7711         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7712         * tree-streamer-in.c: Likewise.
7713         * tree-streamer.c (record_common_node): Likewise.
7714         * tree.c (tree_code_size): Likewise.
7715         (wide_int_to_tree_1): Likewise.
7716         (type_contains_placeholder_1): Likewise.
7717         (build_common_tree_nodes): Likewise.
7718         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7719         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7720         (POINTER_BOUNDS_P): Likewise.
7721         (BOUNDED_TYPE_P): Likewise.
7722         (BOUNDED_P): Likewise.
7723         (CALL_WITH_BOUNDS_P): Likewise.
7724         (pointer_bounds_type_node): Likewise.
7725         * value-prof.c (gimple_ic): Likewise.
7726         * var-tracking.c (vt_add_function_parameters): Likewise.
7727         * varasm.c (make_decl_rtl): Likewise.
7728         (assemble_start_function): Likewise.
7729         (output_constant): Likewise.
7730         (maybe_assemble_visibility): Likewise.
7731         * varpool.c (ctor_for_folding): Likewise.
7732         * chkp-builtins.def: Remove.
7733         * ipa-chkp.c: Remove.
7734         * ipa-chkp.h: Remove.
7735         * rtl-chkp.c: Remove.
7736         * rtl-chkp.h: Remove.
7737         * tree-chkp-opt.c: Remove.
7738         * tree-chkp.c: Remove.
7739         * tree-chkp.h: Remove.
7741 2018-06-07  Carl Love  <cel@us.ibm.com>
7743         * config/rs6000/vsx.md (vextract_fp_from_shorth,
7744         vextract_fp_from_shortl): Add BE support.
7746 2018-06-07  Paul Koning  <ni1d@arrl.net>
7748         * compare-elim.c (try_merge_compare): Don't merge compare if
7749         address contains a side effect.
7750         (try_eliminate_compare): Likewise.
7752 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
7754         * config.gcc: Support "tremont".
7755         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7756         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7757         PROCESSOR_TREMONT.
7758         * config/i386/i386.c (m_TREMONT): Define.
7759         (processor_target_table): Add "tremont".
7760         (PTA_TREMONT): Define.
7761         (ix86_lea_outperforms): Add TARGET_TREMONT.
7762         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7763         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7764         and M_INTEL_GOLDMONT_PLUS.
7765         (fold_builtin_cpu): Add "tremont".
7766         (ix86_add_stmt_cost): Add TARGET_TREMONT.
7767         (ix86_option_override_internal): Add "tremont".
7768         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7769         (processor_type): Add PROCESSOR_TREMONT.
7770         * config/i386/x86-tune.def: Add m_TREMONT.
7771         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7773 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7775         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7776         symbol defined for msp430i* devices to be lower case.
7778 2018-06-07  Richard Biener  <rguenther@suse.de>
7780         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7781         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
7782         Properly wrap signed arithmetic if overflow wraps.
7784 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
7786         PR tree-optimization/69615
7787         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7788         of a cast from a same precision integral SSA_NAME in a bb dominated
7789         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
7790         cast to utype if rhs2 has already a compatible type.
7792 2018-06-07  Richard Biener  <rguenther@suse.de>
7794         PR tree-optimization/85935
7795         * graphite-scop-detection.c (find_params_in_bb): Analyze
7796         condition operands with respect to the correct loop.  Assert
7797         the analysis doesn't fail.
7799 2018-06-04  Carl Love  <cel@us.ibm.com>
7801         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7802         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
7803         as it is slightly cheaper.
7804         (first_match_or_eos_index_<mode>):
7805         Calculate index using natural element order.
7806         (first_match_index_<mode>):
7807         Calculate index using natural element order.
7808         (first_match_or_eos_index_<mode>):
7809         Calculate index using natural order.
7810         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7811         for BE and LE modes.
7812         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7813         P9V_BUILTIN_VCLZLSBB_V16QI.
7814         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7815         specific.
7817 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7819         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7820         indentation and line wrap for many prototypes.  Add missing
7821         @smallexample directives around block of prototypes for vec_xl and
7822         vec_xst.
7824 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
7826         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7827         track if we pass or return IEEE 128-bit floating point.
7828         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7829         C++ mangling that is compatible with GCC 8.1.
7830         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7831         (init_cumulative_args): Note if we pass or return IEEE 128-bit
7832         floating point types.
7833         (rs6000_function_arg_advance_1): Likewise.
7834         (rs6000_mangle_type): Optionally generate mangled names that match
7835         what GCC 8.1 generated for IEEE 128-bit floating point types.
7836         (rs6000_globalize_decl_name): If we have an external function that
7837         passes or returns IEEE 128-bit types, generate a weak reference
7838         from the mangled name used in GCC 8.1 to the current mangled
7839         name.
7840         (rs6000_init_builtins): Make __ibm128 use the long double type if
7841         long double is IBM extended double.  Make __float128 use the long
7842         double type if long double is IEEE 128-bit.
7844         PR target/85657
7845         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7846         macro for __ibm128 built-in functions.
7847         (PACK_IF): Add __ibm128 pack/unpack functions.
7848         (UNPACK_IF): Likewise.
7849         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7850         enable long double built-in functions if long double is IEEE
7851         128-bit floating point.
7852         (rs6000_invalid_builtin): Update long double built-in function
7853         error message.
7854         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7855         functions, adjust the built-in function to use the long double
7856         built-in function if __ibm128 and long double are the same type.
7857         * doc/extend.texi (PowerPC builtins): Update documention for
7858         __builtin_{,un}pack_longdouble.  Add documentation for
7859         __builtin_{,un}pack_ibm128.
7861 2018-06-06  Jim Wilson  <jimw@sifive.com>
7863         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7864         (struct machine_function): New field interrupt_mode.
7865         (riscv_handle_type_attribute): New function.  Add forward declaration.
7866         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7867         (riscv_expand_epilogue): Check interrupt_mode field.
7868         (riscv_set_current_function): Check interrupt attribute args and
7869         set interrupt_mode field.
7870         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7871         (riscv_sret, riscv_uret): New.
7872         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7873         new arguments to interrupt attribute.
7875 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
7877         PR target/63177
7878         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7879         Don't handle -mcpu=power8 if -mpower9-vector is also used.
7881 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7883         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7884         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7885         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7886         several redundant entries.
7888 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
7890         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
7891         type from "rtx" to "rtx_insn *".
7892         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
7893         for local "call_insn", removing cast.
7894         (ix86_expand_call): Likewise, introducing a "call_insn" local.
7896 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
7898         PR tree-optimization/86066
7899         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
7900         for BIT_INSERT_EXPR stores.
7902 2018-06-06  Richard Biener  <rguenther@suse.de>
7904         PR tree-optimization/86062
7905         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7906         component refs ontop
7907         of to be offsetted base.
7909 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7911         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7912         to be static and remove check on interrupt attribute name.
7914 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7916         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7917         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7919 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
7921         PR target/79924
7922         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
7923         second argument.
7924         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
7925         Remove second argument, change how error is called.
7926         (aarch64_layout_arg): Remove second argument from
7927         aarch64_err_no_fpadvsimd call.
7928         (aarch64_init_cumulative_args): Ditto.
7929         (aarch64_gimplify_va_arg_expr): Ditto.
7930         * config/aarch64/aarch64.md (mov<mode>): Ditto.
7932 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
7934         * config/i386/i386.md (simple_return_indirect_internal): New expander.
7935         (*simple_return_indirect_internal<mode>): Rename from
7936         simple_return_indirect_internal.  Use W mode iterator.
7937         (rstorssp): New expander.
7938         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
7939         (clrssbsy): New expander.
7940         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
7942 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
7944         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
7945         __typeof__.
7946         (cmse_check_pointed_object): Likewise.
7948 2018-06-05  Martin Liska  <mliska@suse.cz>
7950         PR gcov-profile/47618
7951         * doc/invoke.texi: Document how -fprofile-dir format
7952         is extended.
7954 2018-06-05  Richard Biener  <rguenther@suse.de>
7956         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
7957         removal pretend DOM info isn't available so we do not update
7958         it and only remove edges, not dominated blocks.  Actually free
7959         DOM info in case we removed something.  Remove unreachable blocks.
7960         (mfb_keep_latches): Work with either DOM info or marked backedges.
7961         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
7962         first.  Mark backedges if DOM info isn't available.
7963         (Re-)compute DOM info after cleanup_control_flow_pre.
7965 2018-06-05  Richard Biener  <rguenther@suse.de>
7967         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
7968         (locus_discrim_hasher::hash): Adjust.
7969         (locus_discrim_hasher::equal): Likewise.
7970         (next_discriminator_for_locus): Work on line directly.
7971         (same_line_p): Pass in expanded locus1 as well.
7972         (assign_discriminators): Avoid redundant location expansions.
7974 2018-06-05  Richard Biener  <rguenther@suse.de>
7976         PR tree-optimization/86046
7977         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
7978         if required after clearing TREE_ADDRESSABLE.
7980 2018-06-05  Richard Biener  <rguenther@suse.de>
7982         PR tree-optimization/86047
7983         * tree-ssa-loop.c (for_each_index): Glob handling of all
7984         decls and constants and really handle all of them.
7986 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7988         PR target/81497
7989         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
7990         qualifier_void_pointer and qualifier_const_void_pointer.
7991         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7992         (arm_init_builtins): Handle the above.
7993         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
7994         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
7995         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
7996         void intrinsics.
7998 2018-06-05  Martin Liska  <mliska@suse.cz>
8000         * auto-profile.c (read_autofdo_file): Do not use
8001         gcov_ctr_summary struct.
8002         (afdo_callsite_hot_enough_for_early_inline): Likewise.
8003         * coverage.c (struct counts_entry): Likewise.
8004         (read_counts_file): Read just single summary entry.
8005         (get_coverage_counts): Use gcov_summary struct.
8006         * coverage.h (get_coverage_counts): Likewise.
8007         * gcov-dump.c (dump_working_sets): Likewise.
8008         (tag_summary): Dump just single summary.
8009         * gcov-io.c (gcov_write_summary): Write just histogram
8010         summary.
8011         (gcov_read_summary): Read just single summary.
8012         (compute_working_sets): Use gcov_summary struct.
8013         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8014         of GCOV_COUNTERS_SUMMABLE.
8015         (GCOV_COUNTERS_SUMMABLE): Remove.
8016         (GCOV_FIRST_VALUE_COUNTER): Replace with
8017         GCOV_COUNTER_V_INTERVAL.
8018         (struct gcov_ctr_summary): Remove.
8019         (struct gcov_summary): Directly use fields of former
8020         gcov_ctr_summary.
8021         (compute_working_sets): Use gcov_summary struct.
8022         * gcov.c (read_count_file): Do not use ctrs fields.
8023         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8024         struct.
8025         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8026         struct.
8027         * profile.c: Likewise.
8028         * profile.h: Likewise.
8030 2018-06-05  Martin Liska  <mliska@suse.cz>
8032         PR gcov-profile/84846
8033         * gcov.c (output_lines): Print working directory only
8034         in intermediate format.
8036 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
8038         * config/s390/s390-builtin-types.def: Add void function type.
8039         * config/s390/s390-builtins.def: Use the function type for the
8040         tbeginc builtin.
8042 2018-06-04  Jim Wilson  <jimw@sifive.com>
8044         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8045         to int.
8046         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8047         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
8048         handle EH_RETURN_DATA_REGNO registers properly.
8049         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8050         (riscv_expand_epilogue): Update comment.  Change argument name and
8051         type.  Update code to use new name and type.  Pass new args to
8052         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
8053         EXCEPTION_RETURN.
8054         * config/riscv/riscv.md (NORMAL_RETURN): New.
8055         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8056         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8057         (eh_return): Call gen_eh_return_internal and emit barrier.
8058         (eh_return_internal): Call riscv_expand_epilogue.
8060 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
8062         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8063         bit_insertion field and declare can_be_merged_into method.
8064         (merged_store_group::can_be_merged_into): New method.
8065         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8066         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
8067         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8069 2018-06-04  Richard Biener  <rguenther@suse.de>
8071         PR tree-optimization/85955
8072         * builtins.c (fold_builtin_sincos): Convert pointers to
8073         destination to appropriate type before dereferencing.
8075 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
8077         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8079 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
8081         * expr.c (expand_expr_real_1): Force the operand into memory if
8082         its TYPE_MODE is BLKmode and if there is no integer mode for
8083         the number of bits being extracted.
8085 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
8087         PR target/85832
8088         PR target/86036
8089         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8090         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8092 2018-06-04  Richard Biener  <rguenther@suse.de>
8094         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8095         (cleanup_tree_cfg_noloop): ... single caller.  Do
8096         start_recording_case_labels later.
8098 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
8100         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8101         to _IMMINTRIN_H_INCLUDED.
8102         * config/i386/pconfigintrin.h: Ditto.
8103         * config/i386/waitpkgintrin.h: Ditto.
8104         * config/i386/immintrin.h: Add includes for sgxintrin.h,
8105         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8106         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8107         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8108         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8109         waitpkgintrin.h and cldemoteintrin.h.
8111 2018-06-04  Richard Biener  <rguenther@suse.de>
8113         PR tree-optimization/86038
8114         * tracer.c (find_best_successor): Check probability for
8115         being initialized, bail out if not.
8117 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
8119         PR target/86003
8120         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8121         of bits to ignore when comparing architectures.
8123 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
8125         PR tree-optimization/69615
8126         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8127         maximum or minimum of the type, try to merge it also as if
8128         range1 is + [-, x - 1] or + [x + 1, -].
8130         PR c++/86025
8131         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8133 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
8135         PR tree-optimization/86034
8136         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8137         the unsigned bitfield type in a bit insertion sequence if it does not
8138         have a larger precision than the bitfield size.
8139         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8141 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
8143         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8145 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
8147         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8148         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8149         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8150         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8152 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
8154         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8155         Disable -fdelete-null-pointer-checks for ELF toolchain.
8157 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
8158             Kito Cheng  <kito.cheng@gmail.com>
8160         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8161         (nds32le-*-*, nds32be-*-*): Integrate checking process.
8162         (nds32*-*-*): Add glibc and uclibc conditions.
8163         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8164         (TARGET_EXCEPT_UNWIND_INFO): Define.
8165         * config/nds32/elf.h: New file.
8166         * config/nds32/linux.h: New file.
8167         * config/nds32/nds32-elf.opt: New file.
8168         * config/nds32/nds32-linux.opt: New file.
8169         * config/nds32/nds32-fp-as-gp.c
8170         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8171         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8172         TARGET_LINUX_ABI.
8173         (nds32_asm_file_end): Ditto.
8174         (nds32_print_operand): Ditto.
8175         (nds32_insert_attributes): Ditto.
8176         (nds32_init_libfuncs): New function.
8177         (TARGET_HAVE_TLS): Define.
8178         (TARGET_INIT_LIBFUNCS): Define.
8179         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8180         spec content.
8181         (TARGET_ELF): Apply different mcmodel setting.
8182         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8183         been migrated into elf.h and linux.h files.
8184         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8185         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8186         (mcmodel): The content has been migrated into nds32-elf.opt and
8187         nds32-linux.opt files.
8188         * config/nds32/t-elf: New file.
8189         * config/nds32/t-linux: New file.
8191 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
8192             Shiva Chen  <shiva0217@gmail.com>
8194         * config/nds32/constants.md (unspec_volatile_element): Add
8195         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8196         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8197         optimization.
8198         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8199         (make_pass_nds32_fp_as_gp): Declare.
8200         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8201         optmization pass.
8202         (nds32_asm_function_end_prologue): Remove unused asm output.
8203         (nds32_asm_function_begin_epilogue): Remove unused asm output.
8204         (nds32_asm_file_start): Output necessary fp_as_gp information.
8205         (nds32_option_override): Adjust register usage.
8206         (nds32_expand_prologue): Consider fp_as_gp situation.
8207         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8208         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8209         (epilogue): Ditto.
8210         (return): Ditto.
8211         (simple_return): Ditto.
8212         (omit_fp_begin): Output special directive for fp_as_gp.
8213         (omit_fp_end): Output special directive for fp_as_gp.
8214         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8215         mforbid-fp-as-gp): New options.
8217 2018-06-01  Mark Wielaard  <mark@klomp.org>
8219         * dwarf2out.c (dwarf2out_finish): Remove generation of
8220         DW_AT_loclists_base.
8222 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
8224         * gimple-ssa-store-merging.c: Include gimple-fold.h.
8225         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8226         (struct merged_store_group): Add bit_insertion field.
8227         (dump_char_array): Use standard hexadecimal format.
8228         (merged_store_group::merged_store_group): Set bit_insertion to false.
8229         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
8230         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
8231         also print the mask in the dump file.
8232         (pass_store_merging::gate): Minor tweak.
8233         (imm_store_chain_info::coalesce_immediate): Fix wrong association
8234         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
8235         stores with INTEGER_CST stores.
8236         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8237         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8238         and use it throughout.  Generate bit insertion sequences if need be.
8239         (pass_store_merging::process_store): Remove redundant condition.
8240         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8242 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
8244         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8245         the 128-bit floating point types.  Fix function comment.
8247 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8249         * config/aarch64/aarch64-simd.md
8250         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8251         mnemonics.
8252         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8253         mnemonics.
8255 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
8257         PR tree-optimization/85989
8258         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8259         variable.
8260         (backprop::intersect_uses): Check it when deciding whether this
8261         is a backedge reference.
8262         (backprop::process_block): Add each phi to m_visited_phis
8263         after visiting it, then clear it at the end.
8265 2018-06-01  Richard Biener  <rguenther@suse.de>
8267         * tree-vectorizer.h (vect_dr_stmt): New function.
8268         (vect_get_load_cost): Adjust.
8269         (vect_get_store_cost): Likewise.
8270         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8271         Use vect_dr_stmt instead of DR_SMTT.
8272         (vect_record_base_alignments): Likewise.
8273         (vect_calculate_target_alignment): Likewise.
8274         (vect_compute_data_ref_alignment): Likewise and make static.
8275         (vect_update_misalignment_for_peel): Likewise.
8276         (vect_verify_datarefs_alignment): Likewise.
8277         (vector_alignment_reachable_p): Likewise.
8278         (vect_get_data_access_cost): Likewise.  Pass down
8279         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8280         (vect_get_peeling_costs_all_drs): Likewise.
8281         (vect_peeling_hash_get_lowest_cost): Likewise.
8282         (vect_enhance_data_refs_alignment): Likewise.
8283         (vect_find_same_alignment_drs): Likewise.
8284         (vect_analyze_data_refs_alignment): Likewise.
8285         (vect_analyze_group_access_1): Likewise.
8286         (vect_analyze_group_access): Likewise.
8287         (vect_analyze_data_ref_access): Likewise.
8288         (vect_analyze_data_ref_accesses): Likewise.
8289         (vect_vfa_segment_size): Likewise.
8290         (vect_small_gap_p): Likewise.
8291         (vectorizable_with_step_bound_p): Likewise.
8292         (vect_prune_runtime_alias_test_list): Likewise.
8293         (vect_analyze_data_refs): Likewise.
8294         (vect_supportable_dr_alignment): Likewise.
8295         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8296         (vect_gen_prolog_loop_niters): Likewise.
8297         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8298         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8299         modify DR_STMT.
8300         (vect_recog_mask_conversion_pattern): Likewise.
8301         (vect_try_gather_scatter_pattern): Likewise.
8302         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8303         to vect_get_store_cost.
8304         (vect_get_store_cost): Get stmt_info instead of DR.
8305         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8306         (vect_get_load_cost): Get stmt_info instead of DR.
8308 2018-06-01  Richard Biener  <rguenther@suse.de>
8310         PR middle-end/86017
8311         * gimple-fold.c (var_decl_component_p): Also allow offsetted
8312         vars wrapped in MEM_REFs.
8314 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
8316         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8317         Fix subreg tests so that we only return a choice between
8318         GENERAL_REGS and FP_REGS if the original classes included both.
8320 2018-06-01  Richard Biener  <rguenther@suse.de>
8322         PR ipa/85960
8323         * tree-ssa-structalias.c (get_function_part_constraint):
8324         Handle NULL fi->decl.
8325         (find_func_aliases_for_call): Properly handle indirect
8326         fi from direct call.
8327         (find_func_clobbers): Likewise.
8328         (ipa_pta_execute): Likewise.
8329         (create_variable_info_for): For functions that are ifunc_resolver
8330         resolve to a varinfo that contains the result of the resolver call.
8331         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8332         aliases.
8334 2018-05-31  Michael Collison  <michael.collison@arm.com>
8336         * config/aarch64/aarch64.md:
8337         (*fix_to_zero_extenddfdi2): New pattern.
8338         * gcc.target/aarch64/fix_extend1.c: New testcase.
8340 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
8342         PR middle-end/78809
8343         PR middle-end/83026
8344         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8345         and BUILT_IN_STRNCMP_EQ.
8346         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8347         BUILT_IN_STRNCMP_EQ.
8348         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8349         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8350         (gimple_fold_builtin): Likewise.
8351         * tree-ssa-strlen.c (compute_string_length): New function.
8352         (determine_min_obsize): New function.
8353         (handle_builtin_string_cmp): New function to handle calls to
8354         string compare functions.
8355         (strlen_optimize_stmt): Add handling to builtin string compare
8356         calls.
8357         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8358         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8359         * tree.c (build_common_builtin_nodes): Add new defines of
8360         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8362 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
8364         PR target/85984
8365         * bb-reorder.c (pass_partition_blocks::gate): Return false for
8366         functions with naked attribute.
8368 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
8370         * config/i386/sse.md (avx_vec_concat<mode>):
8371         Substitute concat_tg_mode mode attribute with xtg_mode.
8372         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8373         (concat_tg_mode): Remove mode attribute.
8375 2018-05-31  Martin Sebor  <msebor@redhat.com>
8377         PR c/82063
8378         * calls.c (alloc_max_size): Correct a logic error/typo.
8379         Treat excessive arguments as infinite.  Warn for invalid arguments.
8380         * doc/invoke.texi (-Walloc-size-larger-than): Update.
8382 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
8384         PR target/85829
8385         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8386         and movx for Haswell.
8388 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
8389             Cesar Philippidis  <cesar@codesourcery.com>
8391         PR middle-end/85879
8392         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8393         when emitting error on private/firstprivate reductions.
8394         * omp-low.c (lower_omp_target): Avoid reference-type processing
8395         on pointers for firstprivate clause.
8397 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
8399         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8400         (st1x2): Likewise.
8401         (st1x3): Likewise.
8402         * config/aarch64/aarch64-simd.md
8403         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8404         (aarch64_ld1_x3_<mode>): Likewise
8405         (aarch64_st1x2<VALLDIF:mode>): Likewise
8406         (aarch64_st1_x2_<mode>): Likewise
8407         (aarch64_st1x3<VALLDIF:mode>): Likewise
8408         (aarch64_st1_x3_<mode>): Likewise
8409         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8410         (vld1_s8_x3): Likewise.
8411         (vld1_u16_x3): Likewise.
8412         (vld1_s16_x3): Likewise.
8413         (vld1_u32_x3): Likewise.
8414         (vld1_s32_x3): Likewise.
8415         (vld1_u64_x3): Likewise.
8416         (vld1_s64_x3): Likewise.
8417         (vld1_f16_x3): Likewise.
8418         (vld1_f32_x3): Likewise.
8419         (vld1_f64_x3): Likewise.
8420         (vld1_p8_x3): Likewise.
8421         (vld1_p16_x3): Likewise.
8422         (vld1_p64_x3): Likewise.
8423         (vld1q_u8_x3): Likewise.
8424         (vld1q_s8_x3): Likewise.
8425         (vld1q_u16_x3): Likewise.
8426         (vld1q_s16_x3): Likewise.
8427         (vld1q_u32_x3): Likewise.
8428         (vld1q_s32_x3): Likewise.
8429         (vld1q_u64_x3): Likewise.
8430         (vld1q_s64_x3): Likewise.
8431         (vld1q_f16_x3): Likewise.
8432         (vld1q_f32_x3): Likewise.
8433         (vld1q_f64_x3): Likewise.
8434         (vld1q_p8_x3): Likewise.
8435         (vld1q_p16_x3): Likewise.
8436         (vld1q_p64_x3): Likewise.
8437         (vst1_s64_x2): Likewise.
8438         (vst1_u64_x2): Likewise.
8439         (vst1_f64_x2): Likewise.
8440         (vst1_s8_x2): Likewise.
8441         (vst1_p8_x2): Likewise.
8442         (vst1_s16_x2): Likewise.
8443         (vst1_p16_x2): Likewise.
8444         (vst1_s32_x2): Likewise.
8445         (vst1_u8_x2): Likewise.
8446         (vst1_u16_x2): Likewise.
8447         (vst1_u32_x2): Likewise.
8448         (vst1_f16_x2): Likewise.
8449         (vst1_f32_x2): Likewise.
8450         (vst1_p64_x2): Likewise.
8451         (vst1q_s8_x2): Likewise.
8452         (vst1q_p8_x2): Likewise.
8453         (vst1q_s16_x2): Likewise.
8454         (vst1q_p16_x2): Likewise.
8455         (vst1q_s32_x2): Likewise.
8456         (vst1q_s64_x2): Likewise.
8457         (vst1q_u8_x2): Likewise.
8458         (vst1q_u16_x2): Likewise.
8459         (vst1q_u32_x2): Likewise.
8460         (vst1q_u64_x2): Likewise.
8461         (vst1q_f16_x2): Likewise.
8462         (vst1q_f32_x2): Likewise.
8463         (vst1q_f64_x2): Likewise.
8464         (vst1q_p64_x2): Likewise.
8465         (vst1_s64_x3): Likewise.
8466         (vst1_u64_x3): Likewise.
8467         (vst1_f64_x3): Likewise.
8468         (vst1_s8_x3): Likewise.
8469         (vst1_p8_x3): Likewise.
8470         (vst1_s16_x3): Likewise.
8471         (vst1_p16_x3): Likewise.
8472         (vst1_s32_x3): Likewise.
8473         (vst1_u8_x3): Likewise.
8474         (vst1_u16_x3): Likewise.
8475         (vst1_u32_x3): Likewise.
8476         (vst1_f16_x3): Likewise.
8477         (vst1_f32_x3): Likewise.
8478         (vst1_p64_x3): Likewise.
8479         (vst1q_s8_x3): Likewise.
8480         (vst1q_p8_x3): Likewise.
8481         (vst1q_s16_x3): Likewise.
8482         (vst1q_p16_x3): Likewise.
8483         (vst1q_s32_x3): Likewise.
8484         (vst1q_s64_x3): Likewise.
8485         (vst1q_u8_x3): Likewise.
8486         (vst1q_u16_x3): Likewise.
8487         (vst1q_u32_x3): Likewise.
8488         (vst1q_u64_x3): Likewise.
8489         (vst1q_f16_x3): Likewise.
8490         (vst1q_f32_x3): Likewise.
8491         (vst1q_f64_x3): Likewise.
8492         (vst1q_p64_x3): Likewise.
8494 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8496         * config/msp430/msp430.c (msp430_output_labelref): Prepend
8497         user_label_prefix to name.
8499         * tree-core.h: Update comment about the format of NAME string
8500         passed to handler in attribute_spec.
8502         * config/msp430/msp430.md: Remove erroneous subreg expression from
8503         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8504         zero_extend{q,h}isi2.
8506 2018-05-30  Borislav Petkov  <bp@suse.de>
8508         * doc/extend.texi: Document some architecture specific
8509         constraints and sort entries.
8511 2018-05-30  Martin Sebor  <msebor@redhat.com>
8513         PR middle-end/85369
8514         * builtins.c (expand_builtin_stpcpy_1): New function.
8515         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8516         only if the former succeeds.
8518 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
8520         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8521         in saphira.
8523 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
8525         * doc/invoke.texi (-flinker-output): Document
8527 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
8529         * passes.c (ipa_write_summaries): Only modify statements if body
8530         is in memory.
8531         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8532         incrementally linking.
8533         (ipa_passes): Likewise.
8534         * lto-cgraph.c (lto_output_node): When incrementally linking do not
8535         pass down resolution info.
8536         * common.opt (flag_incremental_link): Update info.
8537         * gcc.c (plugin specs): Turn flinker-output=* to
8538         -plugin-opt=-linker-output-known
8539         * toplev.c (compile_file): Also cut compilation when doing incremental
8540         link.
8541         * flag-types. (enum lto_partition_model): Add
8542         LTO_LINKER_OUTPUT_NOLTOREL.
8543         (invoke.texi): Add -flinker-output docs.
8544         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8545         link same way as WPA; do not stream in dead initializers.
8547         * dwarf2out.c (dwarf2out_die_ref_for_decl,
8548         darf2out_register_external_decl): Support incremental link.
8550 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
8552         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8554 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
8556         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8557         it down to simple_object_copy_lto_debug_sections.
8558         (run_gcc): Determine incremental LTO link time and configure
8559         lto1 into non-wpa mode, disable renaming of debug sections.
8561 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8563         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8564         descriptions of various incorrectly documented functions.
8566 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8568         Revert:
8569         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8570         address check not strict.
8572 2018-05-30  Richard Biener  <rguenther@suse.de>
8574         PR tree-optimization/85964
8575         * tracer.c (better_p): Drop initialized count check, we only
8576         call the function with initialized counts now.
8577         (find_best_successor): Do find a best edge if one
8578         has uninitialized count.
8579         (find_best_predecessor): Likewise.  Do BB frequency check only
8580         if count is initialized.
8582 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
8584         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8585         (aarch64_ldrstr_offset_compare): New.
8586         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8587         load/store orderings.
8588         (aarch64_gen_adjusted_ldpstp): Likewise.
8590 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
8592         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8593         Check for subset of GENERAL_REGS and FP_REGS.
8594         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8595         r=w alternative.
8597 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
8599         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8600         and wi::to_poly_offset.  Add the current offset and then check
8601         whether the sum fits, rather than using an unchecked addition of
8602         a checked term.  Check for a shwi rather than a uhwi.
8603         * expr.c (get_bit_range): Use tree_to_poly_uint64.
8604         (store_constructor): Use poly_int_tree_p.
8605         (expand_expr_real_1): Likewise.
8606         * function.c (assign_temp): Likewise.
8607         * fold-const.c (const_binop): Use poly_int_tree_p and
8608         wi::to_poly_offset.
8609         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
8610         division.
8611         * ipa-icf-gimple.c (func_checker::compare_operand): Use
8612         to_poly_offset for MEM offsets.
8613         * ipa-icf.c (sem_variable::equals): Likewise.
8614         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8615         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8616         wi::to_poly_offset for BIT_FIELD_REF offsets.
8617         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8618         wi::to_poly_offset.
8619         * var-tracking.c (emit_note_insn_var_location): Use
8620         tree_to_poly_uint64.
8622 2018-05-29  Jim Wilson  <jimw@sifive.com>
8624         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8626 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
8628         PR target/85950
8629         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8630         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8631         sequence.
8632         (sse4_1_round<mode>2): Use nonimmediate_operand
8633         for operand 1 predicate.
8635 2018-05-29  Martin Sebor  <msebor@redhat.com>
8636             Richard Biener  <rguenther@suse.de>
8638         PR testsuite/85888
8639         * calls.c (get_size_range): Call determine_value_range instead
8640         of get_value_range..
8641         * tree-vrp.h (determine_value_range): Declared new function.
8642         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8644 2018-05-29  Richard Biener  <rguenther@suse.de>
8646         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8647         sure to use non-pattern stmts for get_earlier_stmt arguments.
8648         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8649         called on pattern stmts.
8650         (get_later_stmt): Likewise.
8652 2018-05-29  Martin Liska  <mliska@suse.cz>
8654         PR gcov-profile/85759
8655         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8656         env variables.
8658 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
8660         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8661         VEC_UNPACK_*_EXPR.
8662         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8663         VEC_PACK_*_EXPR.
8665         PR target/85918
8666         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8667         VEC_PACK_FLOAT_EXPR): New tree codes.
8668         * tree-pretty-print.c (op_code_prio): Handle
8669         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8670         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8671         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8672         * tree-inline.c (estimate_operator_cost): Likewise.
8673         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8674         * fold-const.c (const_binop): Likewise.
8675         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8676         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8677         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8678         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8679         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8680         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8681         * expr.c (expand_expr_real_2): Likewise.
8682         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8683         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8684         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8685         optabs.
8686         * optabs.c (expand_widen_pattern_expr): For
8687         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8688         sign from result type rather than operand's type.
8689         (expand_binop_directly): For vec_packu_float_optab and
8690         vec_packs_float_optab allow result type to be different from operand's
8691         type.
8692         * optabs-tree.c (optab_for_tree_code): Handle
8693         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8694         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
8695         * tree-vect-generic.c (expand_vector_operations_1):  Handle
8696         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8697         VEC_PACK_FLOAT_EXPR.
8698         * tree-vect-stmts.c (supportable_widening_operation): Handle
8699         FIX_TRUNC_EXPR.
8700         (supportable_narrowing_operation): Handle FLOAT_EXPR.
8701         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8702         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8703         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
8704         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8705         mode attributes.
8706         (vec_pack<floatprefix>_float_<mode>): New expander.
8707         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8708         attributes.
8709         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8710         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8711         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8712         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8713         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8714         Document.
8715         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8716         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8717         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8718         VEC_PACK_FLOAT_EXPR): Document.
8720 2018-05-29  Richard Biener  <rguenther@suse.de>
8722         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8723         member.
8724         (stmt_vec_info_vec): Make pointer.
8725         (init_stmt_vec_info_vec): Remove.
8726         (free_stmt_vec_info_vec): Likewise.
8727         (set_stmt_vec_info_vec): New function.
8728         (free_stmt_vec_infos): Likewise.
8729         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8730         (set_vinfo_for_stmt): Likewise.
8731         (get_earlier_stmt): Likewise.
8732         (get_later_stmt): Likewise.
8733         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8734         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8735         (vec_info::~vec_info): Free stmt_vec_infos.
8736         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8737         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8738         (pass_slp_vectorize::execute): Likewise.
8739         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8740         (free_stmt_vec_info_vec): Likewise.
8741         (set_stmt_vec_info_vec): New function.
8742         (free_stmt_vec_infos): Likewise.
8743         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8744         the global stmt_vec_info_vec.
8745         * tree-parloops.c (gather_scalar_reductions): Use
8746         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8747         vector.
8749 2018-05-29  Richard Biener  <rguenther@suse.de>
8751         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8753 2018-05-29  Martin Liska  <mliska@suse.cz>
8754             David Malcolm  <dmalcolm@redhat.com>
8756         * vec.c (test_reverse): New.
8757         (vec_c_tests): Add new test.
8758         * vec.h (vl_ptr>::reverse): New function.
8760 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
8762         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8764         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8765         and later.
8767 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8769         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8771 2018-05-28  Richard Biener  <rguenther@suse.de>
8773         PR tree-optimization/85933
8774         * tree-vect-data-refs.c (vect_record_base_alignments): Only
8775         look at stmts marked as vectorizable.
8777 2018-05-28  Richard Biener  <rguenther@suse.de>
8779         PR tree-optimization/85934
8780         * tree-vect-generic.c (expand_vector_operations_1): Hoist
8781         vector boolean check before scalar optimization.
8783 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
8785         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8786         for armv5te.
8788 2018-05-28  Mark Wielaard  <mark@klomp.org>
8790         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8791         if it is an expression containing a minus sign.
8793 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
8795         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8797 2018-05-27  Paul Koning  <ni1d@arrl.net>
8799         * config/pdp11/pdp11.md (truncsihi2): Remove.
8801 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
8802             Chung-Ju Wu  <jasonwucj@gmail.com>
8804         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8805         implementation.
8806         (unaligned_store_dw): Ditto.
8807         * config/nds32/nds32-memory-manipulation.c
8808         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8809         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8810         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8811         (emit_setmem_word_loop): Rename to ...
8812         (emit_setmem_doubleword_loop): ... this.
8813         (nds32_gen_dup_4_byte_to_word_value): New function.
8814         (nds32_gen_dup_8_byte_to_double_word_value): New function.
8815         (nds32_expand_setmem_loop): Refine implementation.
8816         (nds32_expand_setmem_loop_v3m): Ditto.
8817         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8818         pattern.
8820 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
8822         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8824 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
8826         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8827         (nds32_init_machine_status): Initialize machine->attr_naked_p and
8828         machine->attr_no_prologue_p.
8829         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8830         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8831         (nds32_expand_epilogue): Consider attr_naked_p.
8832         (nds32_expand_epilogue_v3pop): Likewise.
8833         (nds32_can_use_return_insn): Likewise.
8834         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8835         attr_no_prologue_p fields.
8836         * config/nds32/nds32.opt (mret-in-naked-func): New option.
8838 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
8840         PR target/85918
8841         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8842         attributes.
8843         * config/i386/sse.md
8844         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8845         Rename to ...
8846         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8847         ... this.
8848         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8849         Rename to ...
8850         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8851         ... this.
8852         (*<floatsuffix>floatv2div2sf2): Rename to ...
8853         (*float<floatunssuffix>v2div2sf2): ... this.
8854         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8855         (float<floatunssuffix>v2div2sf2_mask): ... this.
8856         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8857         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8858         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8859         to ...
8860         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8861         ... this.
8862         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8863         Rename to ...
8864         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8865         ... this.
8866         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8867         Rename to ...
8868         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8869         ... this.
8870         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8871         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8872         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8873         gen_ufix_truncv8dfv8si2.
8874         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8875         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8876         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8877         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8878         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8879         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8880         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8881         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8883 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
8885         PR target/85900
8886         PR target/85345
8887         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
8889 2018-05-25  Jim Wilson  <jimw@sifive.com>
8891         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
8892         * config/riscv/riscv.c (struct machine_function): Add
8893         interrupt_handler_p and attribute_checked_p fields.
8894         (riscv_attribute_table): Add interrupt.
8895         (riscv_interrupt_type_p): New.
8896         (riscv_save_reg_p): Save extra regs for interrupt handler.
8897         (riscv_use_save_libcall): Return false  for interrupt handler.
8898         (riscv_first_stack_step): Add forward declaration.
8899         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
8900         for interrupt handler with large frame.  Use it for saved reg list.
8901         (riscv_expand_prologue): Move flag_stack_usage_info support to
8902         eliminate duplication.
8903         (riscv_expand_epilogue): Generate mret for interrupt handler.
8904         (riscv_epilogue_uses): New.
8905         (riscv_can_use_return_insn): Return false for interrupt handler.
8906         (riscv_function_ok_for_sibcall): Likewise.
8907         (riscv_set_current_function): Add interrupt handler support.
8908         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8909         * config/riscv/riscv.md (UNSPECV_MRET): New.
8910         (GP_REGNUM): New.
8911         (riscv_frflags, riscv_fsflags): Use tab after opcode.
8912         (riscv_mret): New.
8913         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8915 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
8917         PR tree-optimization/85712
8918         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8919         this candidate has already been replaced in-situ by a copy.
8921 2018-05-25  Jason Merrill  <jason@redhat.com>
8923         PR c++/80485 - inline function non-zero address.
8924         * symtab.c (nonzero_address): Check DECL_COMDAT.
8926 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
8928         PR target/83628
8929         * config/alpha/alpha.md (ashlsi3): New insn pattern.
8930         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
8931         extension of SImode operation.  Use const123_operand predicate.
8932         (*saddsi_1): Remove.
8933         (*saddl_se_1): Ditto.
8934         (*ssubsi_1): Ditto.
8935         (*ssubl_se_1): Ditto.
8936         * config/alpha/predicates.md (const123_operand): New predicate.
8937         * config/alpha/constraints.md (P): Use IN_RANGE.
8939 2018-05-25  Richard Biener  <rguenther@suse.de>
8941         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
8942         defaulted to true.
8943         (ref_maybe_used_by_stmt_p): Likewise.
8944         (stmt_may_clobber_ref_p): Likewise.
8945         (stmt_may_clobber_ref_p_1): Likewise.
8946         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
8947         and pass it along.
8948         (ref_maybe_used_by_stmt_p): Likewise.
8949         (stmt_may_clobber_ref_p): Likewise.
8950         (stmt_may_clobber_ref_p_1): Likewise.
8951         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
8952         the alias oracle to disambiguate DRs with stmts DR analysis
8953         couldn't handle.
8954         (vect_analyze_data_refs): Do not give up on not analyzable
8955         DRs for BB vectorization.  Remove code truncating the dataref
8956         vector.
8958 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
8960         PR target/85832
8961         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8962         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
8963         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
8965 2018-05-25  Richard Biener  <rguenther@suse.de>
8967         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
8968         function, combining stmt data ref gathering and fatal analysis
8969         parts.
8970         (vect_analyze_data_refs): Remove now redudnant code and simplify.
8971         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
8972         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
8973         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
8974         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
8976 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
8978         PR tree-optimization/85720
8979         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
8980         SCC if all partitions are builtins.
8981         (version_loop_by_alias_check): New parameter.  Generate cancelable
8982         runtime alias check if all partitions are builtins.
8983         (distribute_loop): Update call to above function.
8985 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
8987         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
8988         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
8989         (parm_default_def_partition_arg): Ditto.
8990         (set_parm_default_def_partition): Ditto.
8991         (get_parm_default_def_partitions): Ditto and make it static.
8992         (get_undefined_value_partitions): Ditto and make it static.
8993         (remove_ssa_form): Refactor call to init_var_map here.
8994         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
8995         computation for loop region.
8996         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
8997         (register_default_def): Delete.
8998         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
8999         (parm_default_def_partition_arg): Ditto.
9000         (set_parm_default_def_partition): Ditto.
9001         (get_parm_default_def_partitions): Ditto and make it static.
9002         (get_undefined_value_partitions): Ditto and make it static.
9003         (coalesce_with_default, coalesce_with_default): Update comment.
9004         (create_coalesce_list_for_region): New func factored out from
9005         create_outofssa_var_map.
9006         (populate_coalesce_list_for_outofssa): New func factored out from
9007         create_outofssa_var_map and coalesce_ssa_name.
9008         (create_outofssa_var_map): Delete.
9009         (coalesce_ssa_name): Refactor to support live range computation.
9010         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9011         (get_parm_default_def_partitions): Delete.
9012         (get_undefined_value_partitions): Ditto.
9013         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9014         computation for loop region.
9015         (new_tree_live_info, loe_visit_block): Ditto.
9016         (live_worklist, set_var_live_on_entry): Ditto.
9017         (calculate_live_on_exit, verify_live_on_entry): Ditto.
9018         * tree-ssa-live.h (struct _var_map): New fields.
9019         (init_var_map): Change decl.
9020         (region_contains_p): New.
9022 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
9024         * tree-ssa-live.h (live_merge_and_clear): Delete.
9026 2018-05-25  Richard Biener  <rguenther@suse.de>
9028         PR c++/85912
9029         * tree-dump.c (dequeue_and_dump): Remove access to removed
9030         operand 2 of a SWITCH_EXPR.
9032 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9034         * doc/sourcebuild.texi (vect_double_cond_arith): Include
9035         multiplication and division.
9036         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9037         (cond_udiv@var{m}, cond_umod@var{m}): Document.
9038         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9039         (cond_udiv_optab, cond_umod_optab): New optabs.
9040         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9041         (IFN_COND_RDIV): New internal functions.
9042         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9043         TRUNC_MOD_EXPR and RDIV_EXPR.
9044         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9045         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9046         New unspecs.
9047         (SVE_INT_BINARY): Include mult.
9048         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9049         (optab, sve_int_op): Handle mult.
9050         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9051         UNSPEC_COND_DIV.
9052         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9053         for SVE_INT_BINARY_SD.
9055 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9057         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9058         (optab, sve_int_op): Handle div and udiv.
9059         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9060         for SVE_INT_BINARY_SD.
9061         (*<optab><mode>3): New insn for the same.
9063 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9065         * tree-vect-patterns.c: Include predict.h.
9066         (vect_recog_divmod_pattern): Restrict check for division support
9067         to when optimizing for size.
9069 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9071         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9072         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9073         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9074         (gimple_match_op::set_op): Likewise.
9075         (gimple_resimplify4): Declare.
9076         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9077         (expr::gen_transform): Likewise.
9078         (decision_tree::gen): Generate a simplification routine for 4 operands.
9079         * gimple-match-head.c (gimple_simplify): Add an overload for
9080         4 operands.  In the top-level function, handle up to 4 call
9081         arguments and call gimple_resimplify4.
9082         (gimple_resimplify4): New function.
9083         (build_call_internal): Pass a fourth operand.
9084         (maybe_push_to_seq): Likewise.
9085         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9086         Fold VEC_COND_EXPRs of an operation and a default value into
9087         an IFN_COND_* function if possible.
9088         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9089         New unspecs.
9090         (SVE_COND_FP_BINARY): Include them.
9091         (optab, sve_fp_op): Handle them.
9092         (SVE_INT_BINARY_REV): New code iterator.
9093         (SVE_COND_FP_BINARY_REV): New int iterator.
9094         (commutative): New int attribute.
9095         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9096         Declare.
9097         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9098         function.
9099         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9100         (*cond_<optab><mode>): New patterns for reversed operands.
9102 2018-05-25  Richard Biener  <rguenther@suse.de>
9104         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9105         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9106         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9107         (STMT_VINFO_GROUPED_ACCESS): Adjust.
9108         * tree-vect-data-refs.c (everywhere): Adjust users.
9109         * tree-vect-loop.c (everywhere): Likewise.
9110         * tree-vect-slp.c (everywhere): Likewise.
9111         * tree-vect-stmts.c (everywhere): Likewise.
9112         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9114 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9116         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9117         Rename to...
9118         (gcc_cv_as_section_exclude): ... this.
9119         Try Solaris as #exclude syntax.
9120         * configure: Regenerate.
9121         * config.in: Regenerate.
9122         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9123         SECTION_EXCLUDE.
9124         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9125         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9127         * varasm.c (default_elf_asm_named_section): Don't check if
9128         HAVE_GAS_SECTION_EXCLUDE is defined.
9130 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9132         * doc/md.texi: Update the documentation of the cond_* optabs
9133         to mention the new final operand.  Fix GET_MODE_NUNITS call.
9134         Describe the scalar case too.
9135         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9136         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9137         instead of 2.
9138         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9139         (get_conditional_internal_fn): Update comment.
9140         * tree-vect-loop.c (vectorizable_reduction): Pass the original
9141         accumulator value as a final argument to conditional functions.
9142         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9143         a define_expand and add an "else" operand.  Assert for now that
9144         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
9145         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9146         (*cond_<optab><mode>): New patterns.
9147         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9148         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9149         (UNSPEC_COND_EOR): Delete.
9150         (optab): Remove associated mappings.
9151         (SVE_INT_BINARY): New code iterator.
9152         (sve_int_op): Remove int attribute and add "minus" to the code
9153         attribute.
9154         (SVE_COND_INT_OP): Delete.
9155         (SVE_COND_FP_OP): Rename to...
9156         (SVE_COND_FP_BINARY): ...this.
9158 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
9160         * optabs.c (can_reuse_operands_p): New function.
9161         (maybe_legitimize_operands): Try to reuse the results for
9162         earlier operands.
9164 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
9166         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9167         Add {q} suffix to insn mnemonic.
9169 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9171         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9172         (msp430_warn_func_return): New.
9174 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
9176         * fold-const.c (tree_nonzero_bits): New function.
9177         * fold-const.h (tree_nonzero_bits): Likewise.
9178         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9179         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9181 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
9183         PR target/85900
9184         PR target/85345
9185         * varasm.c (assemble_alias): Check ifunc_resolver only on
9186         FUNCTION_DECL.
9188 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
9190         PR target/85903
9191         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9192         when memory input operand is handled.
9194 2018-05-24  Luis Machado  <luis.machado@linaro.org>
9196         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9197         global.
9198         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9200 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
9202         * match.pd: Delay FMA folds until after vectorization.
9204 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9206         PR target/83009
9207         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9208         address check not strict.
9210 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
9212         * gimple-match.h (gimple_match_op): New class.
9213         (mprts_hook): Replace parameters with a gimple_match_op *.
9214         (maybe_build_generic_op): Likewise.
9215         (gimple_simplified_result_is_gimple_val): Replace parameters with
9216         a const gimple_match_op *.
9217         (gimple_simplify): Replace code_helper * and tree * parameters with
9218         a gimple_match_op * parameter.
9219         (gimple_resimplify1): Replace code_helper *, tree and tree *
9220         parameters with a gimple_match_op * parameter.
9221         (gimple_resimplify2): Likewise.
9222         (gimple_resimplify3): Likewise.
9223         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9224         parameters with a gimple_match_op * parameter.
9225         * gimple-match-head.c (gimple_simplify): Change prototypes of
9226         auto-generated functions to take a gimple_match_op * instead of
9227         separate code_helper * and tree * parameters.  Make the same
9228         change in the top-level overload and update calls to the
9229         gimple_resimplify routines.  Update calls to the auto-generated
9230         functions and to maybe_push_res_to_seq in the publicly-facing
9231         operation-specific gimple_simplify overloads.
9232         (gimple_match_op::MAX_NUM_OPS): Define.
9233         (gimple_resimplify1): Replace rcode and ops with a single res_op
9234         parameter.  Update call to gimple_simplify.
9235         (gimple_resimplify2): Likewise.
9236         (gimple_resimplify3): Likewise.
9237         (mprts_hook): Replace parameters with a gimple_match_op *.
9238         (maybe_build_generic_op): Likewise.
9239         (build_call_internal): Replace type, nargs and ops with
9240         a gimple_match_op *.
9241         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9242         with a single gimple_match_op *.  Update calls to mprts_hook,
9243         build_call_internal and gimple_simplified_result_is_gimple_val.
9244         Factor out code that is common to the tree_code and combined_fn cases.
9245         * genmatch.c (expr::gen_transform): Replace tem_code and
9246         tem_ops with a gimple_match_op called tem_op.  Update calls
9247         to the gimple_resimplify functions and maybe_push_res_to_seq.
9248         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9249         res_ops.  Update call to the gimple_resimplify functions.
9250         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9251         (decision_tree::gen): Make the functions take a gimple_match_op *
9252         called res_op instead of separate res_code and res_ops parameters.
9253         Update call accordingly.
9254         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9255         and ops with a single res_op parameter.  Update calls to
9256         maybe_build_generic_op and maybe_push_res_to_seq.
9257         (fold_stmt_1): Update calls to gimple_simplify and
9258         replace_stmt_with_simplification.
9259         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9260         and gimple_simplified_result_is_gimple_val.
9261         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9262         gimple_simplify.
9263         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9264         with a gimple_match_op *.
9265         (vn_nary_build_or_lookup): Likewise.  Update call to
9266         vn_nary_build_or_lookup_1.
9267         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9268         gimple_match_op *.  Update calls to the gimple_resimplify routines
9269         and to gimple_simplified_result_is_gimple_val.
9270         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9271         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9272         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9273         (visit_nary_op): Likewise.
9274         (visit_reference_op_load): Likewise.
9276 2018-05-23  Luis Machado  <luis.machado@linaro.org>
9278         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9279         modifier for printing the step amount.
9281 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
9283         PR target/78849
9284         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9285         types.
9287 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
9289         * doc/sourcebuild.texi (Endianness): New subsubsection.
9291 2018-05-23  Luis Machado  <luis.machado@linaro.org>
9293         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9294         <prefetch_dynamic_strides>: New const bool field.
9295         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9296         prefetch_dynamic_strides.
9297         (exynosm1_prefetch_tune): Likewise.
9298         (thunderxt88_prefetch_tune): Likewise.
9299         (thunderx_prefetch_tune): Likewise.
9300         (thunderx2t99_prefetch_tune): Likewise.
9301         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9302         false.
9303         (aarch64_override_options_internal): Update to set
9304         PARAM_PREFETCH_DYNAMIC_STRIDES.
9305         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9306         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9307         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9308         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9309         prefetch-dynamic-strides setting.
9311 2018-05-23  Luis Machado  <luis.machado@linaro.org>
9313         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9314         <minimum_stride>: New const int field.
9315         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9316         minimum_stride field defaulting to -1.
9317         (exynosm1_prefetch_tune): Likewise.
9318         (thunderxt88_prefetch_tune): Likewise.
9319         (thunderx_prefetch_tune): Likewise.
9320         (thunderx2t99_prefetch_tune): Likewise.
9321         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9322         <default_opt_level>: Set to 3.
9323         (aarch64_override_options_internal): Update to set
9324         PARAM_PREFETCH_MINIMUM_STRIDE.
9325         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9326         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9327         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9328         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9329         stride is constant and is below the minimum stride threshold.
9331 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9333         * config/arm/arm-cpus.in (mode26): Delete.
9334         (armv4): Delete mode26 reference.
9335         * config/arm/arm.c (arm_configure_build_target): Delete use of
9336         isa_bit_mode26.
9338 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
9340         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9341         New insn pattern.
9342         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9343         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
9344         for non-SSE modes.
9345         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
9346         (floatunsdidf2): Ditto.
9348 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
9350         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9351         (fixuns_trunc<mode>si2_avx512f): Ditto.
9352         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9353         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9354         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9356 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
9358         PR rtl-optimization/79985
9359         * df-scan.c (df_insn_refs_collect): Remove special case for
9360         global registers and asm statements.
9362 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
9364         * extend.texi (Global Register Variables): Rewrite the bullet list.
9365         Note that the register is available for allocation. Note that access
9366         via inline asm must use constraints. Add note about async-signal
9367         handlers. Remove paragraph about automagic register selection.
9369 2018-05-23  Richard Biener  <rguenther@suse.de>
9371         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9372         of fixed offset from memset VN.
9374 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
9376         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9377         first_interp field.
9378         (alloc_cand_and_find_basis): Initialize first_interp field.
9379         (slsr_process_mul): Modify first_interp field.
9380         (slsr_process_add): Likewise.
9381         (slsr_process_cast): Modify first_interp field for each new
9382         interpretation.
9383         (slsr_process_copy): Likewise.
9384         (dump_candidate): Dump first_interp field.
9385         (replace_mult_candidate): Process all interpretations, not just
9386         subsequent ones.
9387         (replace_rhs_if_not_dup): Likewise.
9388         (replace_one_candidate): Likewise.
9390 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
9392         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9393         Add new boolean.
9394         (aarch64_needs_frame_chain): New function.
9395         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9397 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
9399         PR target/84882
9400         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9401         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9402         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9403         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9404         as true for strict-align.
9405         (aarch64_can_inline_p): Perform checks even when callee has no
9406         attributes to check for strict alignment.
9407         * doc/extend.texi (AArch64 Function Attributes): Document
9408         no-strict-align.
9409         * doc/invoke.texi: (AArch64 Options): Likewise.
9411 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
9413         PR tree-optimization/85853
9414         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9415         the handling of the root of the node to...
9416         (vect_slp_analyze_node_operations_1): ...this new function,
9417         and run the whole thing with the child nodes' def types
9418         set according to their SLP node's def type.
9420 2018-05-23  Richard Biener  <rguenther@suse.de>
9422         PR middle-end/85874
9423         * tree-data-ref.c (create_runtime_alias_checks): Defer
9424         and ignore overflow warnings.
9426 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
9428         PR tree-optimization/85822
9429         * tree-vrp.c (is_masked_range_test): Fix handling of negative
9430         constants.
9432 2018-05-23  Richard Biener  <rguenther@suse.de>
9434         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9435         memset constants via native_interpret_expr.
9437 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
9439         PR target/85345
9440         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9441         attribute.
9442         (cgraph_node::create_alias): Likewise.
9443         (cgraph_node::get_availability): Check ifunc_resolver instead
9444         of looking up ifunc attribute.
9445         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9446         * varasm.c (do_assemble_alias): Likewise.
9447         (assemble_alias): Likewise.
9448         (default_binds_local_p_3): Likewise.
9449         * cgraph.h (cgraph_node): Add ifunc_resolver.
9450         (cgraph_node::only_called_directly_or_aliased_p): Return false
9451         for IFUNC resolver.
9452         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9453         attribute.
9454         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9455         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9456         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9457         instead of looking up ifunc attribute.
9459 2018-05-22  Luis Machado  <luis.machado@linaro.org>
9461         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9463 2018-05-22  Martin Sebor  <msebor@redhat.com>
9465         PR middle-end/85359
9466         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9467         only when expasion succeeds.
9468         (expand_builtin_strcmp): Same.
9469         (expand_builtin_strncmp): Same.
9471 2018-05-22  Martin Sebor  <msebor@redhat.com>
9473         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9475 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
9476             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9478         * config/aarch64/aarch64-ldpstp.md: Replace uses of
9479         aarch64_mem_pair_operand with memory_operand and delete operand swapping
9480         code.
9481         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9482         Add check for legitimate_address.
9483         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9484         (aarch64_swap_ldrstr_operands): New.
9485         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9486         Define prototype.
9488 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
9489             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9491         * config/aarch64/aarch64.md: New patterns to generate stp
9492         and ldp.
9493         (store_pair_sw, store_pair_dw): New patterns to generate stp for
9494         single words and double words.
9495         (load_pair_sw, load_pair_dw): Likewise.
9496         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9497         Delete.
9498         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9499         Delete.
9500         * config/aarch64/aarch64-ldpstp.md: Modify peephole
9501         for different mode ldpstp and add peephole for merged zero stores.
9502         Likewise for loads.
9503         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9504         Add size check.
9505         (aarch64_gen_store_pair): Rename calls to match new patterns.
9506         (aarch64_gen_load_pair): Rename calls to match new patterns.
9507         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9508         (load_pair<DREG:mode><DREG2:mode>): ... This.
9509         (store_pair<mode>): Rename to...
9510         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9511         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9512         New mode iterators.
9513         (V_INT_EQUIV): Handle SImode.
9514         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9515         New predicate.
9517 2018-05-22  Martin Sebor  <msebor@redhat.com>
9519         PR c/85623
9520         * calls.c (maybe_warn_nonstring_arg): Use string length to set
9521         or ajust the presumed bound on an operation to avoid unnecessary
9522         warnings.
9524 2018-05-22  Martin Sebor  <msebor@redhat.com>
9526         PR tree-optimization/85826
9527         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9528         assuming that a DECL necesarily has a constant size.
9530 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
9532         PR middle-end/85862
9533         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9535 2018-05-22  Richard Biener  <rguenther@suse.de>
9537         PR tree-optimization/85834
9538         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9539         non-constant and non-zero memset arguments.
9541 2018-05-22  Martin Liska  <mliska@suse.cz>
9543         PR ipa/85607
9544         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9546 2018-05-22  Richard Biener  <rguenther@suse.de>
9548         PR tree-optimization/85863
9549         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9550         comparisons when vectype is specified.
9551         (vectorizable_condition): Do not specify vectype for
9552         vect_is_simple_cond when SLP vectorizing.
9554 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
9556         PR target/85657
9557         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9558         define __ibm128 as long double.
9559         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9560         as a distinct type when IEEE 128-bit support is enabled.
9561         (init_float128_ieee): Fix up conversions between IFmode and IEEE
9562         128-bit types to use the correct functions.
9563         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9564         convert between 128-bit floating point types that have different
9565         modes but the same representation, instead of using gen_lowpart to
9566         makean alias.
9567         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9568         KFmode.
9569         (IFKF_reg): New attributes to give the register constraints for
9570         IFmode and KFmode.
9571         (extend<mode>tf2_internal): New insns to mark an explicit
9572         conversion between 128-bit floating point types that have a
9573         different mode but share the same representation.
9575 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
9577         PR tree-optimization/85814
9578         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9579         a null return from get_strinfo when unsharing the next
9580         strinfo in the chain.
9582 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
9584         PR gcc/84923
9585         * varasm.c (weak_finish): Clean up weak_decls.
9587 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9589         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9590         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9591         UNSPEC_UADALP values.
9592         * config/aarch64/iterators.md (ABAL): New int iterator.
9593         (ABDL2): Likewise.
9594         (ADALP): Likewise.
9595         (sur): Add mappings for the above.
9596         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9597         New define_insn.
9598         (aarch64_<sur>abal<mode>_4): Likewise.
9599         (aarch64_<sur>adalp<mode>_3): Likewise.
9600         (<sur>sadv16qi): New define_expand.
9602 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
9604         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9605         (*movdf_internal): Ditto.
9606         (*rcpsf2_sse): Ditto.
9607         (*rsqrtsf2_sse): Ditto.
9608         (*sqrt<mode>2_sse): Ditto.
9610 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
9612         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9613         eor3q<mode>4.
9614         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9615         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9616         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9617         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9618         vbcaxq_s64): New.
9619         * config/aarch64/arm_neon.h: Likewise.
9620         * config/aarch64/iterators.md (VQ_I): New.
9622 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
9624         * config.gcc: Add arc/t-multilib-linux to tmake_file for
9625         arc*-*-linux*.
9626         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9627         MULTILIB_DIRNAMES
9629 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
9631         * config/nds32/constraints.md (S): New constraint.
9632         * config/nds32/nds32.md (call_internal): Use constraint S.
9633         (call_value_internal): Likewise.
9634         (sibcall_internal): Likewise.
9635         (sibcall_value_internal): Likewise.
9637 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
9638             Chung-Ju Wu  <jasonwucj@gmail.com>
9640         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9641         into consideration.
9643 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
9644             Chung-Ju Wu  <jasonwucj@gmail.com>
9646         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9647         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9648         (nds32_rtx_costs_impl): Simplify.
9649         (nds32_address_cost_impl): Simplify.
9650         (nds32_init_rtx_costs): New function.
9651         (nds32_rtx_costs_speed_prefer): Likewise.
9652         (nds32_rtx_costs_size_prefer): Likewise.
9653         (nds32_address_cost_speed_prefer): Likewise.
9654         (nds32_address_cost_speed_fwprop): Likewise.
9655         (nds32_address_cost_size_prefer): Likewise.
9656         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9657         * config/nds32/nds32.c (nds32_option_override): Use
9658         nds32_init_rtx_costs function.
9660 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
9662         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9663         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9664         (TARGET_PIPELINE_N8): Likewise.
9665         (TARGET_PIPELINE_N10): Likewise.
9666         (TARGET_PIPELINE_N13): Likewise.
9667         (TARGET_PIPELINE_GRAYWOLF): Likewise.
9669 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
9671         * config/nds32/nds32-fpu.md: Update copyright year.
9673 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9675         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9677 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9679         * config/nds32/nds32.c
9680         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9681         * config/nds32/nds32.opt (minline-asm-r15): New option.
9683 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9685         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9686         MASK_HW_ABS.
9687         * config/nds32/nds32.md (abssi2): New pattern.
9689 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
9691         * config/i386/i386.md (rex64namesuffix): New mode attribute.
9692         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9693         Merge insn pattern from sse_cvtsi2ss<round_name> and
9694         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9695         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9696         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9697         using SWI48 mode iterator.
9698         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9699         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9700         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9701         pattern from sse_cvttss2si<round_saeonly_name>
9702         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9703         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9704         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9705         using SWI48 mode iterator.
9706         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9707         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9708         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9709         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9710         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9711         using SWI48 mode iterator.
9712         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9713         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9714         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9715         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9716         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9717         SWI48 mode iterator.
9718         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9719         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9720         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9721         pattern from sse_cvttsd2si<round_saeonly_name>
9722         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9724 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9726         * config/nds32/nds32-md-auxiliary.c
9727         (nds32_valid_smw_lwm_base_p): Refine.
9728         (nds32_output_smw_single_word): Refine.
9729         (nds32_output_smw_double_word): New.
9730         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9732 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9734         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9735         (nds32_output_stack_pop): Refine.
9736         (nds32_expand_unaligned_load): Refine.
9737         (nds32_expand_unaligned_store): Refine.
9739 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
9740             Chung-Ju Wu  <jasonwucj@gmail.com>
9742         * config/nds32/constants.md: Add TP_REGNUM constant.
9743         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9744         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9745         UNSPEC_ADD32.
9746         * config/nds32/nds32-doubleword.md: Consider flag_pic.
9747         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9748         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9749         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9750         and PIC code generation.
9751         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9752         code generation.
9753         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9754         optimization.
9755         * config/nds32/nds32.md: Support TLS and PIC.
9756         * config/nds32/nds32.c: Support TLS and PIC.
9757         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9758         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9759         predicate.
9761 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
9763         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9764         mode with E_ prefix.
9766 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
9767             Chung-Ju Wu  <jasonwucj@gmail.com>
9769         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9770         * config/nds32/nds32-md-auxiliary.c
9771         (symbolic_reference_mentioned_p): New.
9772         (nds32_legitimize_ict_address): New.
9773         (nds32_expand_ict_move): New.
9774         (nds32_indirect_call_referenced_p): New.
9775         (nds32_symbol_binds_local_p): Delete.
9776         (nds32_long_call_p): Modify.
9777         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9778         * config/nds32/nds32-protos.h
9779         (symbolic_reference_mentioned_p): Declare.
9780         (nds32_legitimize_ict_address): Declare.
9781         (nds32_expand_ict_move): Declare.
9782         (nds32_indirect_call_referenced_p): Declare.
9783         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9784         (nds32_relax_group): Use nds32_ict_const_p as condition.
9785         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9786         (nds32_asm_file_start): Output ict_model directive in asm code.
9787         (nds32_legitimate_address_p): Consider indirect call.
9788         (nds32_print_operand): Consider indirect call.
9789         (nds32_print_operand_address): Consider indirect call.
9790         (nds32_insert_attributes): Handle "indirect_call" attribute.
9791         (TARGET_LEGITIMATE_ADDRESS_P): Define.
9792         (TARGET_LEGITIMATE_CONSTANT_P): Define.
9793         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9794         (TARGET_DELEGITIMIZE_ADDRESS): Define.
9795         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9796         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9797         (TARGET_ICT_MODEL_SMALL): Define.
9798         (TARGET_ICT_MODEL_LARGE): Define.
9799         * config/nds32/nds32.md (movsi): Consider ict model.
9800         (call, call_value): Consider ict model.
9801         (sibcall, sibcall_value): Consider ict model.
9802         * config/nds32/nds32.opt (mict-model): New option.
9803         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9804         model.
9806 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
9807             Monk Chiang  <sh.chiang04@gmail.com>
9808             Jim Wilson <jimw@sifive.com>
9810         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9811         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
9812         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9813         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9814         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
9815         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9816         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9817         compute save_libcall_adjustment properly.
9818         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9819         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9820         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9821         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9822         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9823         (ABI_SPEC): Handle mabi=ilp32e.
9824         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9825         (RVE): Add RVE mask.
9826         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9827         <-march>: Add rv32e as an example.
9829 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
9831         PR c++/82899
9832         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9833         (intra_create_variable_infos): Handle C++ constructors.
9835 2018-05-18  Martin Liska  <mliska@suse.cz>
9837         * passes.def: Remove a redundant pass.
9839 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
9841         PR bootstrap/85838
9842         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9844 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9846         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9847         (ARMv4): Update.
9848         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9849         (ARMv6m): Update.
9850         (armv2, armv2a, armv3, armv3m): Delete architectures.
9851         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9852         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9853         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9854         Delete cpus.
9855         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9856         (*mulsidi3adddi): Likewise.
9857         (mulsidi3): Likewise.
9858         (*mulsidi3_nov6): Likewise.
9859         (umulsidi3): Likewise.
9860         (umulsidi3_nov6): Likewise.
9861         (umaddsidi4): Likewise.
9862         (*umulsidi3adddi): Likewise.
9863         (smulsi3_highpart): Likewise.
9864         (*smulsi3_highpart_nov6): Likewise.
9865         (umulsi3_highpart): Likewise.
9866         (*umulsi3_highpart_nov6): Likewise.
9867         * config/arm/arm.h (arm_arch3m): Delete.
9868         * config/arm/arm.c (arm_arch3m): Delete.
9869         (arm_option_override_internal): Update armv3-related comment.
9870         (arm_configure_build_target): Delete use of isa_bit_mode32.
9871         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9872         (arm_rtx_costs_internal): Delete check of arm_arch3m.
9873         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9874         (mulsa3): Likewise.
9875         (mulusa3): Likewise.
9876         * config/arm/arm-protos.h (arm_arch3m): Delete.
9877         * config/arm/arm-tables.opt: Regenerate.
9878         * config/arm/arm-tune.md: Likewise.
9879         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9880         deleted architectures.
9882 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9884         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9885         (armv5t, armv5te): New features.
9886         (ARMv5, ARMv5e): Delete fgroups.
9887         (ARMv5t, ARMv5te): Adjust for above changes.
9888         (ARMv6m): Likewise.
9889         (armv5, armv5e): Delete arches.
9890         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
9891         arm_arch5.
9892         (*call_reg_arm): Likewise.
9893         (*call_value_reg_armv5): Likewise.
9894         (*call_value_reg_arm): Likewise.
9895         (*call_symbol): Likewise.
9896         (*call_value_symbol): Likewise.
9897         (*sibcall_insn): Likewise.
9898         (*sibcall_value_insn): Likewise.
9899         (clzsi2): Likewise.
9900         (prefetch): Likewise.
9901         (define_split and define_peephole2 dependent on arm_arch5):
9902         Likewise.
9903         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
9904         arm_arch5e.
9905         (TARGET_ARM_QBIT): Likewise.
9906         (TARGET_DSP_MULTIPLY): Likewise.
9907         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
9908         (arm_arch5, arm_arch5e): Delete.
9909         (arm_arch5t, arm_arch5te): Declare.
9910         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9911         (arm_arch5t): Declare.
9912         (arm_option_reconfigure_globals): Update for the above.
9913         (arm_options_perform_arch_sanity_checks): Update comment, replace
9914         use of arm_arch5 with arm_arch5t.
9915         (use_return_insn): Likewise.
9916         (arm_emit_call_insn): Likewise.
9917         (output_return_instruction): Likewise.
9918         (arm_final_prescan_insn): Likewise.
9919         (arm_coproc_builtin_available): Likewise.
9920         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9921         arm_arch5e with arm_arch5t and arm_arch5te.
9922         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
9923         (arm_arch5t, arm_arch5te): Declare.
9924         * config/arm/arm-tables.opt: Regenerate.
9925         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
9926         * config/arm/t-multilib: Likewise.
9927         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
9928         instead of arm_arch5.
9929         (*call_reg_thumb1): Likewise.
9930         (*call_value_reg_thumb1_v5): Likewise.
9931         (*call_value_reg_thumb1): Likewise.
9932         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
9933         unreachable path.
9934         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
9936 2018-05-18  Martin Liska  <mliska@suse.cz>
9938         PR gcov-profile/84846
9939         * doc/gcov.texi: Document -t option of gcov tool.
9941 2018-05-18  Martin Liska  <mliska@suse.cz>
9943         PR gcov-profile/84846
9944         * gcov.c (print_usage): Add new -t option.
9945         (process_args): Handle the option.
9946         (generate_results): Use stdout as output when requested by
9947         the option.
9949 2018-05-18  Martin Liska  <mliska@suse.cz>
9951         PR gcov-profile/84846
9952         * coverage.c (coverage_init): Write PWD to .gcno file.
9953         * doc/gcov.texi: Document how working directory is printed.
9954         * gcov-dump.c (dump_gcov_file): Print PWD.
9955         * gcov.c (output_intermediate_file): Likewise.
9956         (read_graph_file): Read PWD string.
9957         (output_lines): Print PWD.
9959 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9961         PR middle-end/85817
9962         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
9963         for retval and return false if all args to phi are zero.
9965 2018-05-18  Richard Biener  <rguenther@suse.de>
9967         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
9968         method.
9969         (evrp_dom_walker::before_dom_children): Call it.
9971 2018-05-18  Richard Biener  <rguenther@suse.de>
9973         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
9974         results when processing array refs with variable index.
9976 2018-05-18  Toon Moene  <toon@moene.org>
9978         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9979         directly after that of -floop-interchange. Indicate that both
9980         options are enabled by default when specifying -O3.
9982 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9984         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
9985         iterator.  Delete separate integer-mode vec_set<mode> expander.
9986         (aarch64_simd_vec_setv2di): Delete.
9987         (vec_setv2di): Delete.
9988         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
9989         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
9990         the "w, r" alternative.
9992 2018-05-18  Martin Liska  <mliska@suse.cz>
9994         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
9995         * tree-pass.h (make_pass_lower_switch_O0): New function.
9996         * tree-switch-conversion.c (node_has_low_bound): Remove.
9997         (node_has_high_bound): Likewise.
9998         (node_is_bounded): Likewise.
9999         (class pass_lower_switch): Make it a template type and create
10000         two instances.
10001         (pass_lower_switch::execute): Add template argument.
10002         (make_pass_lower_switch): New function.
10003         (make_pass_lower_switch_O0): New function.
10004         (do_jump_if_equal): Remove.
10005         (emit_case_nodes): Simplify to just handle all 3 cases and leave
10006         all the hard work to tree optimization passes.
10008 2018-05-18  Martin Liska  <mliska@suse.cz>
10010         * dbgcnt.c (limit_low): Renamed from limit.
10011         (limit_high): New variable.
10012         (dbg_cnt_is_enabled): Check for upper limit.
10013         (dbg_cnt): Adjust dumping.
10014         (dbg_cnt_set_limit_by_index): Add new argument for high
10015         value.
10016         (dbg_cnt_set_limit_by_name): Likewise.
10017         (dbg_cnt_process_single_pair): Parse new format.
10018         (dbg_cnt_process_opt): Use strtok.
10019         (dbg_cnt_list_all_counters): Remove 'value' and add
10020         'limit_high'.
10021         * doc/invoke.texi: Document changes.
10023 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
10025         * doc/sourcebuild.texi (scalar_all_fma): Document.
10026         * tree.def (FMA_EXPR): Delete.
10027         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10028         * internal-fn.c (ternary_direct): New macro.
10029         (expand_ternary_optab_fn): Likewise.
10030         (direct_ternary_optab_supported_p): Likewise.
10031         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10032         * builtins.c (fold_builtin_fma): Delete.
10033         (fold_builtin_3): Don't call it.
10034         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10035         * expr.c (expand_expr_real_2): Likewise.
10036         * fold-const.c (operand_equal_p): Likewise.
10037         (fold_ternary_loc): Likewise.
10038         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10039         * gimple.c (DEFTREECODE): Likewise.
10040         * gimplify.c (gimplify_expr): Likewise.
10041         * optabs-tree.c (optab_for_tree_code): Likewise.
10042         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10043         * tree-eh.c (operation_could_trap_p): Likewise.
10044         (stmt_could_throw_1_p): Likewise.
10045         * tree-inline.c (estimate_operator_cost): Likewise.
10046         * tree-pretty-print.c (dump_generic_node): Likewise.
10047         (op_code_prio): Likewise.
10048         * tree-ssa-loop-im.c (stmt_cost): Likewise.
10049         * tree-ssa-operands.c (get_expr_operands): Likewise.
10050         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10051         * fold-const-call.h (fold_fma): Delete.
10052         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10053         CFN_FNMA and CFN_FNMS.
10054         (fold_fma): Delete.
10055         * genmatch.c (combined_fn): New enum.
10056         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10057         (commutative_op): New function.
10058         (commutate): Use it.  Handle more than 2 operands.
10059         (dt_operand::gen_gimple_expr): Use commutative_op.
10060         (parser::parse_expr): Allow :c to be used with non-binary
10061         operators if the commutative operand is known.
10062         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10063         CFN_FMS, CFN_FNMA and CFN_FNMS.
10064         (backprop::process_assign_use): Remove FMA_EXPR handling.
10065         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10066         (gen_hsa_fma): New function.
10067         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10068         IFN_FNMA and IFN_FNMS.
10069         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10070         * gimple-fold.h (follow_all_ssa_edges): Declare.
10071         * gimple-fold.c (follow_all_ssa_edges): New function.
10072         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10073         gimple_build interface and use follow_all_ssa_edges to fold the result.
10074         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10075         instead of checking for optabs directly.
10076         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10077         rather than FMA_EXPRs.
10078         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10079         call to IFN_FMA instead of an FMA_EXPR.
10081 2018-05-17  Jim Wilson  <jimw@sifive.com>
10083         * expr.c (do_tablejump): When converting index to Pmode, if we have a
10084         sign extended promoted subreg, and the range does not have the sign bit
10085         set, then do a sign extend.
10087         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10088         test, check for sign extended subreg and/or constant operands, and
10089         do a sign extend in that case.
10091 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
10093         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10094         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10095         Add untyped.
10096         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10097         Change logics_shift_reg to logics_shift_imm.
10098         (thunderx2t99_fp_loadpair_basic): Delete.
10099         (thunderx2t99_fp_storepair_basic): Delete.
10100         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10101         (thunderx2t99_asimd_polynomial): Delete.
10102         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10103         and neon_fp_mul_d_scalar_q.
10104         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10105         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10106         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10107         (thunderx2t99_asimd_lut): Add missing tbl types.
10108         (thunderx2t99_asimd_ext): Delete.
10109         (thunderx2t99_asimd_load1_1_mult): Delete.
10110         (thunderx2t99_asimd_load1_2_mult): Delete.
10111         (thunderx2t99_asimd_load1_ldp): New.
10112         (thunderx2t99_asimd_load1): New.
10113         (thunderx2t99_asimd_load2): Add missing *load2* types.
10114         (thunderx2t99_asimd_load3): New.
10115         (thunderx2t99_asimd_load4): New.
10116         (thunderx2t99_asimd_store1_1_mult): Delete.
10117         (thunderx2t99_asimd_store1_2_mult): Delete.
10118         (thunderx2t99_asimd_store2_mult): Delete.
10119         (thunderx2t99_asimd_store2_onelane): Delete.
10120         (thunderx2t99_asimd_store_stp): New.
10121         (thunderx2t99_asimd_store1): New.
10122         (thunderx2t99_asimd_store2): New.
10123         (thunderx2t99_asimd_store3): New.
10124         (thunderx2t99_asimd_store4): New.
10126 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
10128         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10129         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
10131 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
10132             Segher Boessenkool  <segher@kernel.crashing.org>
10134         PR target/85698
10135         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10136         operand.
10138 2018-05-17  Richard Biener  <rguenther@suse.de>
10140         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10141         for pruning loop and prune defs feeding only already visited PHIs.
10143 2018-05-17  Richard Biener  <rguenther@suse.de>
10145         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10147 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
10148             Richard Biener  <rguenther@suse.de>
10150         PR tree-optimization/85793
10151         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10152         for VMAT_ELEMENTWISE.
10154 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
10156         * internal-fn.h (lookup_internal_fn): Declare
10157         * internal-fn.c (lookup_internal_fn): New function.
10158         * gimple.c (gimple_build_call_from_tree): Handle calls to
10159         internal functions.
10160         * gimple-pretty-print.c (dump_gimple_call): Print "." before
10161         internal function names.
10162         * tree-pretty-print.c (dump_generic_node): Likewise.
10163         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10165 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
10167         * gimple-fold.h (gimple_build): Make the function forms take
10168         combined_fn rather than built_in_function.
10169         (gimple_simplify): Likewise.
10170         * gimple-match-head.c (gimple_simplify): Likewise.
10171         * gimple-fold.c (gimple_build): Likewise.
10172         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10173         rather than gimple_build_call_internal.
10174         (get_initial_defs_for_reduction): Likewise.
10175         (vect_create_epilog_for_reduction): Likewise.
10176         (vectorizable_live_operation): Likewise.
10178 2018-05-17  Martin Liska  <mliska@suse.cz>
10180         * gimple-ssa-sprintf.c (format_directive): Do not use
10181         space in between 'G_' and '('.
10183 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
10185         PR target/85323
10186         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10187         even if the mask is not all ones.
10189         PR target/85323
10190         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10191         vector.
10192         (ix86_gimple_fold_builtin): Likewise.
10194         PR target/85323
10195         * config/i386/i386.c: Include tree-vector-builder.h.
10196         (ix86_vector_shift_count): New function.
10197         (ix86_fold_builtin): Fold shift builtins by scalar count.
10198         (ix86_gimple_fold_builtin): Likewise.
10200         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10201         _mm512_setzero): New intrinsics.
10203 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
10204             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10206         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10207         code generation for cases where splatting a value is not useful.
10208         * simplify-rtx.c (simplify_ternary_operation): Simplify
10209         vec_merge across a vec_duplicate and a paradoxical subreg forming
10210         a vector mode to a vec_concat.
10212 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
10214         * config.gcc: Support "goldmont-plus".
10215         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10216         "goldmont-plus".
10217         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10218         PROCESSOR_GOLDMONT_PLUS.
10219         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10220         (processor_target_table): Add "goldmont-plus".
10221         (PTA_GOLDMONT_PLUS): Define.
10222         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10223         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10224         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10225         (fold_builtin_cpu): Add "goldmont-plus".
10226         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10227         (ix86_option_override_internal): Add "goldmont-plus".
10228         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10229         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10230         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10231         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10233 2018-05-17  Richard Biener  <rguenther@suse.de>
10235         PR tree-optimization/85757
10236         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10237         remove defs that only feed that PHI from further processing.
10239 2018-05-16  Jim Wilson  <jimw@sifive.com>
10241         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10242         asterisk to name.
10243         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10245 2018-05-16  Mark Wielaard  <mark@klomp.org>
10247         * dwarf2out.c (count_index_strings): New function.
10248         (output_indirect_strings): Call count_index_strings and generate
10249         header for dwarf_version >= 5.
10251 2018-05-16  Mark Wielaard  <mark@klomp.org>
10253         * dwarf2out.c (dwarf_FORM): New function.
10254         (set_indirect_string): Use dwarf_FORM.
10255         (reset_indirect_string): Likewise.
10256         (size_of_die): Likewise.
10257         (value_format): Likewise.
10258         (output_die): Likewise.
10259         (add_skeleton_AT_string): Likewise.
10260         (output_macinfo_op): Likewise.
10261         (index_string): Likewise.
10262         (output_index_string_offset): Likewise.
10263         (output_index_string): Likewise.
10264         (count_index_strings): Likewise.
10266 2018-05-16  Carl Love  <cel@us.ibm.com>
10268         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10269         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10271 2018-05-16  Martin Jambor  <mjambor@suse.cz>
10273         * ipa-prop.c (ipa_free_all_edge_args): Remove.
10274         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10276 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
10278         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10279         (fnma<mode>4): Likewise.
10280         (fms<mode>4): Likewise.
10281         (fnms<mode>4): Likewise.
10282         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10283         (aarch64_fnma<mode>4): Likewise.
10284         (aarch64_fms<mode>4): Likewise.
10285         (aarch64_fnms<mode>4): Likewise.
10286         (aarch64_fnmadd<mode>4): Likewise.
10288 2018-05-16  Jason Merrill  <jason@redhat.com>
10290         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
10292 2018-05-16  Richard Biener  <rguenther@suse.de>
10294         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10295         (dump_stmt_cost): Declare.
10296         (add_stmt_cost): Dump cost we add.
10297         (add_stmt_costs): New function.
10298         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10299         No longer exported.
10300         (vect_analyze_stmt): Adjust prototype.
10301         (vectorizable_condition): Likewise.
10302         (vectorizable_live_operation): Likewise.
10303         (vectorizable_reduction): Likewise.
10304         (vectorizable_induction): Likewise.
10305         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10306         cost vector to pass to vectorizable_ and record afterwards.
10307         (vect_model_reduction_cost): Take cost vector argument and adjust.
10308         (vect_model_induction_cost): Likewise.
10309         (vectorizable_reduction): Likewise.
10310         (vectorizable_induction): Likewise.
10311         (vectorizable_live_operation): Likewise.
10312         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10313         SLP_TREE_NUMBER_OF_VEC_STMTS.
10314         (vect_analyze_slp_cost_1): Remove.
10315         (vect_analyze_slp_cost): Likewise.
10316         (vect_slp_analyze_node_operations): Take visited args and
10317         a target cost vector.  Avoid processing already visited stmt sets.
10318         (vect_slp_analyze_operations): Use a local cost vector to gather
10319         costs and register those of non-discarded instances.
10320         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10321         (vect_schedule_slp_instance): Remove copying of
10322         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
10323         zero.
10324         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10325         adding cost.  Record cost entry location.
10326         (vect_prologue_cost_for_slp_op): Function to compute cost of
10327         a constant or invariant generated for SLP vect in the prologue,
10328         split out from vect_analyze_slp_cost_1.
10329         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
10330         (vect_model_promotion_demotion_cost): Likewise.
10331         (vect_model_store_cost): Likewise, make static.
10332         (vect_model_load_cost): Likewise.
10333         (vectorizable_bswap): Add cost vector arg and adjust.
10334         (vectorizable_call): Likewise.
10335         (vectorizable_simd_clone_call): Likewise.
10336         (vectorizable_conversion): Likewise.
10337         (vectorizable_assignment): Likewise.
10338         (vectorizable_shift): Likewise.
10339         (vectorizable_operation): Likewise.
10340         (vectorizable_store): Likewise.
10341         (vectorizable_load): Likewise.
10342         (vectorizable_condition): Likewise.
10343         (vectorizable_comparison): Likewise.
10344         (can_vectorize_live_stmts): Likewise.
10345         (vect_analyze_stmt): Likewise.
10346         (vect_transform_stmt): Adjust calls to vectorizable_*.
10347         * tree-vectorizer.c: Include gimple-pretty-print.h.
10348         (dump_stmt_cost): New function.
10350 2018-05-16  Richard Biener  <rguenther@suse.de>
10352         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10353         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10354         * tree-ssa-dse.c: Include tree-ssa-loop.h.
10355         (check_name): New callback.
10356         (dse_classify_store): Track cycles via a visited bitmap of PHI
10357         defs and simplify handling of in-loop and across loop dead stores
10358         and properly fail for loop-variant refs.  Handle byte-tracking with
10359         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10360         limiting the walk.
10362 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
10364         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10365         (vect_get_mask_type_for_stmt): Likewise.
10366         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10367         split out from...
10368         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
10369         to determine the statement's vector type and the vector type that
10370         should be used for calculating nunits.  Deal with cases in which
10371         the type has to be deferred.
10372         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10373         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10374         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10375         (vect_determine_vf_for_stmt): New functions, split out from...
10376         (vect_determine_vectorization_factor): ...here.
10377         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10378         (vect_get_mask_type_for_stmt): New functions, split out from
10379         vect_determine_vectorization_factor.
10381 2018-05-16  Richard Biener  <rguenther@suse.de>
10383         * tree-cfg.c (verify_gimple_assign_ternary): Properly
10384         verify the [VEC_]COND_EXPR embedded comparison.
10386 2018-05-15  Martin Sebor  <msebor@redhat.com>
10388         PR tree-optimization/85753
10389         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10390         RECORD_TYPE in addition to ARRAY_TYPE.
10392 2018-05-15  Martin Sebor  <msebor@redhat.com>
10394         PR middle-end/85643
10395         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10397 2018-05-15  Richard Biener  <rguenther@suse.de>
10399         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10400         add by_clobber_p one.  Change algorithm to collect all defs
10401         representing uses we need to walk and try reducing them to
10402         a single one before failing.
10403         (dse_dom_walker::dse_optimize_stmt): Adjust.
10405 2018-05-13  Mark Wielaard  <mark@klomp.org>
10407         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10408         (size_of_loc_descr): Likewise.
10409         (output_loc_operands): Likewise.
10410         (output_loc_operands_raw): Likewise.
10411         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10412         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10413         (hash_loc_operands): Likewise.
10414         (compare_loc_operands): Likewise.
10416 2018-05-14  Mark Wielaard  <mark@klomp.org>
10418         * dwarf2out.c (count_index_addrs): New function.
10419         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10421 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10423         PR tree-optimization/83648
10424         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10425         return value as malloc candidate.
10427 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10429         PR ipa/85734
10430         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10431         param as true in call to suggest_attribute.
10433 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
10435         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10436         -mreadonly-in-sdata.
10438 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10440         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10441         New pattern.
10442         (aarch64_crypto_aesd_fused): Likewise.
10444 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
10446         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10447         (movsi_aarch64): Likewise.
10448         (load_pairsi): Likewise.
10449         (load_pairdi): Likewise.
10450         (store_pairsi): Likewise.
10451         (store_pairdi): Likewise.
10452         (load_pairsf): Likewise.
10453         (load_pairdf): Likewise.
10454         (store_pairsf): Likewise.
10455         (store_pairdf): Likewise.
10456         (zero_extend): Likewise.
10457         (trunc): Swap alternatives.
10458         (fcvt_target): Add '?' to prefer w over r.
10460 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
10462         PR target/85756
10463         * config/i386/i386.md: Disallow non-commutative arithmetics in
10464         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10465         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
10466         in the peephole2 before it.
10468 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
10470         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10471         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10472         (ix86_handle_option): Handle -mcldemote.
10473         * config.gcc: New header.
10474         * config/i386/cldemoteintrin.h: New file.
10475         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10476         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10477         -mcldemote.
10478         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10479         OPTION_MASK_ISA_CLDEMOTE.
10480         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10481         (ix86_valid_target_attribute_inner_p): Ditto.
10482         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10483         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10484         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10485         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10486         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10487         (cldemote): New.
10488         * config/i386/i386.opt: Add -mcldemote.
10489         * config/i386/x86intrin.h: New header.
10490         * doc/invoke.texi: Add -mcldemote.
10492 2018-05-14  Richard Biener  <rguenther@suse.de>
10494         * doc/match-and-simplify.texi: Adjust :s documentation.
10496 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
10498         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10499         intended memcpy size.
10500         (REORDER_45): Likewise.
10502 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
10504         * sort.cc: New file.
10505         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10506         * vec.c (qsort_chk): Use gcc_qsort.
10507         * Makefile.in (OBJS-libcommon): Add sort.o.
10508         (build/sort.o): New target.  Use it...
10509         (BUILD_RTL): ... here, and...
10510         (build/gencfn-macros): ... here, and...
10511         (build/genmatch): ... here.
10513 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
10514             Chung-Ju Wu  <jasonwucj@gmail.com>
10516         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10517         * config/nds32/nds32-graywolf.md: New file.
10518         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10519         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10520         pipeline.
10521         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10522         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10523         * config/nds32/nds32.md (pipeline_model): Add graywolf.
10524         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10525         * config/nds32/pipelines.md: Include n15 settings.
10527 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
10528             Chung-Ju Wu  <jasonwucj@gmail.com>
10530         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10531         * config/nds32/nds32-n13.md: New file.
10532         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10533         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10534         pipeline.
10535         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10536         * config/nds32/nds32.md (pipeline_model): Add n13.
10537         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10538         * config/nds32/pipelines.md: Include n13 settings.
10540 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
10541             Chung-Ju Wu  <jasonwucj@gmail.com>
10543         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10544         * config/nds32/nds32-n10.md: New file.
10545         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10546         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10547         pipeline.
10548         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10549         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10550         * config/nds32/nds32.md (pipeline_model): Add n10.
10551         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10552         * config/nds32/pipelines.md: Include n10 settings.
10554 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
10555             Kito Cheng  <kito.cheng@gmail.com>
10556             Chung-Ju Wu  <jasonwucj@gmail.com>
10558         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10559         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10560         Add enum values for DSP extension instructions.
10561         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10562         New constraints.
10563         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10564         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10565         New code iterators.
10566         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10567         * config/nds32/nds32-dspext.md: New file for DSP implementation.
10568         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10569         * config/nds32/nds32-intrinsic.md: Likewise.
10570         * config/nds32/nds32_intrinsic.h: Likewise.
10571         * config/nds32/nds32-md-auxiliary.c: Likewise.
10572         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10573         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10574         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10575         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10576         * config/nds32/nds32-protos.h: New declarations for DSP extension.
10577         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10578         TYPE_DMAC in switch statement.
10579         * config/nds32/nds32.c: New checking and implementation for DSP
10580         extension instructions.
10581         * config/nds32/nds32.h: Likewise.
10582         * config/nds32/nds32.md: Likewise.
10583         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10584         * config/nds32/predicates.md: Implement new predicates for DSP
10585         extension.
10587 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
10589         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10590         Reformat alternatives and attributes so it is easier to identify
10591         which constraints/attributes go with which instruction.
10592         (mov<mode>_hardfloat32, FMOVE64): Likewise.
10593         (mov<mode>_softfloat32, FMOVE64): Likewise.
10594         (mov<mode>_hardfloat64, FMOVE64): Likewise.
10595         (mov<mode>_softfloat64, FMOVE64): Likewise.
10597 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10599         * doc/extend.texi (PowerPC Built-in Functions): Rename this
10600         subsection.
10601         (Basic PowerPC Built-in Functions): The new name of the
10602         subsection previously known as "PowerPC Built-in Functions".
10603         (Basic PowerPC Built-in Functions Available on all Configurations):
10604         New subsubsection.
10605         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10606         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10607         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10608         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10610 2018-05-11  Martin Jambor  <mjambor@suse.cz>
10612         PR ipa/85655
10613         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10614         single const.
10616 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
10618         PR target/85733
10619         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10621 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
10623         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10624         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10625         (ix86_handle_option): Handle -mwaitpkg.
10626         * config.gcc: New header.
10627         * config/i386/cpuid.h (bit_WAITPKG): New bit.
10628         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10629         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10630         function type.
10631         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10632         OPTION_MASK_ISA_WAITPKG.
10633         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10634         (ix86_option_override_internal): Add PTA_WAITPKG.
10635         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10636         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10637         IX86_BUILTIN_TPAUSE.
10638         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10639         __builtin_ia32_umwait and __builtin_ia32_tpause.
10640         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10641         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10642         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10643         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10644         UNSPECV_TPAUSE): New.
10645         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10646         * config/i386/i386.opt: Add -mwaitpkg.
10647         * config/i386/waitpkgintrin.h: New file.
10648         * config/i386/x86intrin.h: New header.
10649         * doc/invoke.texi: Add -mwaitpkg.
10651 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
10653         PR target/85606
10654         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10655         equivalent.
10656         (cortex-m0): Use armv6s-m isa.
10657         (cortex-m0plus): Likewise.
10658         (cortex-m1): Likewise.
10659         (cortex-m0.small-multiply): Likewise.
10660         (cortex-m0plus.small-multiply): Likewise.
10661         (cortex-m1.small-multiply): Likewise.
10663 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
10664             Jakub Jelinek  <jakub@redhat.com>
10666         PR tree-optimization/85692
10667         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10668         source permute as well.
10670 2018-05-11  Martin Liska  <mliska@suse.cz>
10672         PR sanitizer/85556
10673         * doc/extend.texi: Document LLVM style format for no_sanitize
10674         attribute.
10676 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
10678         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10679         mode_supports_vsx_dform_quad to mode_supports_dq_form.
10680         (mode_supports_vsx_dform_quad): Likewise.
10681         (mode_supports_vmx_dform): Move these functions to be next to the
10682         other mode_supports functions.
10683         (mode_supports_dq_form): Likewise.
10684         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10685         mode_supports_dq_form.
10686         (reg_offset_addressing_ok_p): Likewise.
10687         (offsettable_ok_by_alignment): Likewise.
10688         (rs6000_legitimate_offset_address_p): Likewise.
10689         (legitimate_lo_sum_address_p): Likewise.
10690         (rs6000_legitimize_address): Likewise.
10691         (rs6000_legitimize_reload_address): Likewise.
10692         (rs6000_secondary_reload_inner): Likewise.
10693         (rs6000_preferred_reload_class): Likewise.
10694         (rs6000_output_move_128bit): Likewise.
10696 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
10698         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10699         Generate SImode target register for null target.
10700         <case IX86_BUILTIN_XGETBV>: Ditto.
10701         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10702         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10704 2018-05-10  Carl Love  <cel@us.ibm.com>
10706         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10707         dcbtt and dcbtstt if operands[2] is 0.
10709 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
10711         PR target/85693
10712         * config/i386/sse.md (usadv64qi): New expander.
10714 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
10716         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10717         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10718         -maltivec=be support.
10719         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10720         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10721         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10722         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10723         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10724         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10725         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10726         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10727         altivec_vsumsws): Adjust.
10728         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10729         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10730         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10731         support.
10732         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10733         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10734         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10735         (altivec_lve<VI_char>x): Delete expand.
10736         (*altivec_lve<VI_char>x_internal): Rename to...
10737         (altivec_lve<VI_char>x): ... this.
10738         (altivec_lvxl_<mode>): Delete expand.
10739         (*altivec_lvxl_<mode>_internal): Rename to ...
10740         (altivec_lvxl_<mode>): ... this.
10741         (altivec_stvxl_<mode>): Delete expand.
10742         (*altivec_stvxl_<mode>_internal): Rename to ...
10743         (altivec_stvxl_<mode>): ... this.
10744         (altivec_stve<VI_char>x): Delete expand.
10745         (*altivec_stve<VI_char>x_internal): Rename to ...
10746         (altivec_stve<VI_char>x): ... this.
10747         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10748         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10749         reduc_plus_scal_<mode>): Adjust.
10750         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10751         comment.
10752         (rs6000_cpu_cpp_builtins): Adjust.
10753         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10754         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10755         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10756         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10757         -maltivec=be support.
10758         (rs6000_split_vec_extract_var): Adjust.
10759         (rs6000_split_v4si_init): Adjust.
10760         (swap_selector_for_mode): Delete.
10761         (altivec_expand_lvx_be, altivec_expand_stvx_be,
10762         altivec_expand_stvex_be): Delete.
10763         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10764         -maltivec=be support.
10765         (rs6000_gimple_fold_builtin): Ditto.
10766         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10767         Adjust.
10768         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10769         (TARGET_DIRECT_MOVE_64BIT): Adjust.
10770         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10771         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10772         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10773         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10774         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10775         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10776         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10777         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10778         anonymous split): Adjust.
10779         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10780         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10782 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
10784         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10785         when --with-gxx-include-dir is also specified.
10786         * configure: Regenerate.
10788 2018-05-09  Jim Wilson  <jimw@sifive.com>
10790         PR target/84797
10791         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10792         * config/riscv/t-withmultilib: New.
10793         * config/riscv/withmultilib.h: New.
10794         * doc/install.texi: Document RISC-V --with-multilib-list support.
10796 2018-05-09  Richard Biener  <rguenther@suse.de>
10798         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10799         vector.
10800         (vect_bb_vectorization_profitable_p): Adjust.  Compute
10801         actual scalar cost using the cost vector and the add_stmt_cost
10802         machinery.
10804 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
10806         PR rtl-optimization/85645
10807         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10808         in the REG_CFA_REGISTER note for LR, don't leave it empty.
10810 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
10812         PR rtl-optimization/85645
10813         * shrink-wrap.c (spread_components): Return a boolean saying if
10814         anything was changed.
10815         (try_shrink_wrapping_separate): Iterate spread_components until
10816         nothing changes anymore.
10818 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
10820         PR rtl-optimization/85645
10821         * regrename.c (build_def_use): Also kill the chains that include the
10822         destination of a REG_CFA_REGISTER note.
10824 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
10826         PR rtl-optimization/85645
10827         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10828         insn that has a REG_CFA_REGISTER note.
10830 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
10832         * cfgexpand.c (expand_clobber): New function.
10833         (expand_gimple_stmt_1): Use it.
10834         * tree-vect-stmts.c (vect_clobber_variable): New function,
10835         split out from...
10836         (vectorizable_simd_clone_call): ...here.
10837         (vectorizable_store): Emit a clobber either side of an
10838         IFN_STORE_LANES sequence.
10839         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10841 2018-05-09  Tom de Vries  <tom@codesourcery.com>
10843         PR target/85626
10844         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10845         (define_insn "trap_if_false"): Add exit after trap.
10847 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
10849         PR rtl-optimization/85638
10850         * bb-reorder.c: Include common/common-target.h.
10851         (create_forwarder_block): New function extracted from...
10852         (fix_up_crossing_landing_pad): ...here.  Rename into...
10853         (dw2_fix_up_crossing_landing_pad): ...this.
10854         (sjlj_fix_up_crossing_landing_pad): New function.
10855         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10856         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10857         from both partitions and exit the loop after one iteration.
10859 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10861         Revert:
10862         * doc/extend.texi (PowerPC Built-in Functions): Rename this
10863         subsection.
10864         (Basic PowerPC Built-in Functions): The new name of the
10865         subsection previously known as "PowerPC Built-in Functions".
10866         (Basic PowerPC Built-in Functions Available on all Configurations):
10867         New subsubsection.
10868         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10869         subsubsection.
10870         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10871         subsubsection.
10872         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10873         subsubsection.
10874         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10875         subsubsection.
10877 2018-05-08  Jim Wilson  <jimw@sifive.com>
10879         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10880         (LD_EMUL_SUFFIX): New.
10881         (LINK_SPEC): Use it.
10883 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10885         * doc/extend.texi (PowerPC Built-in Functions): Rename this
10886         subsection.
10887         (Basic PowerPC Built-in Functions): The new name of the
10888         subsection previously known as "PowerPC Built-in Functions".
10889         (Basic PowerPC Built-in Functions Available on all Configurations):
10890         New subsubsection.
10891         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10892         subsubsection.
10893         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10894         subsubsection.
10895         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10896         subsubsection.
10897         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10898         subsubsection.
10900 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
10902         PR target/85683
10903         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
10904         after cmpelim optimization.
10906 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
10908         * config.gcc: Support "goldmont".
10909         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10910         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10911         PROCESSOR_GOLDMONT.
10912         * config/i386/i386.c (m_GOLDMONT): Define.
10913         (processor_target_table): Add "goldmont".
10914         (PTA_GOLDMONT): Define.
10915         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10916         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10917         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10918         (fold_builtin_cpu): Add "goldmont".
10919         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10920         (ix86_option_override_internal): Add "goldmont".
10921         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
10922         (processor_type): Add PROCESSOR_GOLDMONT.
10923         * config/i386/i386.md: Add CPU "glm".
10924         * config/i386/glm.md: New file.
10925         * config/i386/x86-tune.def: Add m_GOLDMONT.
10926         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
10928 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
10930         PR target/85572
10931         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
10932         E_V4DImode.
10933         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
10934         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
10935         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
10937         PR target/85317
10938         * config/i386/i386.c (ix86_fold_builtin): Handle
10939         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
10941         PR target/85480
10942         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
10943         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
10945 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
10947         PR target/85658
10948         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
10949         (check_arch): Likewise.
10950         (check_fpu): Return the result rather than printing it.
10951         (end arch): Fix operator precedence.
10952         (end cpu): Likewise.
10953         (END): Print the result from check_fpu.
10955 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
10956             Alan Hayward  <alan.hayward@arm.com>
10957             David Sherwood  <david.sherwood@arm.com>
10959         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
10960         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
10961         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
10962         (*fcmuo<mode>_and): New patterns.
10964 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
10966         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
10967         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
10968         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
10969         (cmp_op, sve_imm_con): New code attributes.
10970         (SVE_COND_INT_CMP, imm_con): Delete.
10971         (cmp_op): Remove above unspecs from int attribute.
10972         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
10973         to...
10974         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
10975         comparison-specific unspecs.
10976         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
10977         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
10978         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
10979         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
10980         (*vec_fcm<cmp_op><mode>): Rename to...
10981         (*fcm<cmp_op><mode>): ...this and adjust likewise.
10982         (*vec_fcmuo<mode>): Rename to...
10983         (*fcmuo<mode>): ...this and adjust likewise.
10984         (*pred_fcm<cmp_op><mode>): New pattern.
10985         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
10986         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
10987         functions.
10988         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
10989         and UNORDERED.
10990         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
10991         (aarch64_emit_sve_predicated_cond): New function.
10992         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
10993         (aarch64_emit_unspec_cond_or): Replace with...
10994         (aarch64_emit_sve_or_conds): ...this new function.  Use
10995         aarch64_emit_sve_ptrue_op for the individual comparisons and
10996         aarch64_emit_binop to OR them together.
10997         (aarch64_emit_inverted_unspec_cond): Replace with...
10998         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
10999         aarch64_emit_sve_ptrue_op for the comparison and
11000         aarch64_emit_unop to invert the result.
11001         (aarch64_expand_sve_vec_cmp_float): Update after the above
11002         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
11004 2018-05-07  Nathan Sidwell  <nathan@acm.org>
11006         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11007         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11008         (Backwards Compatibility): Likewise.
11010 2018-05-07  Luis Machado  <luis.machado@linaro.org>
11012         PR bootstrap/85681
11013         Revert:
11014         2018-05-07  Luis Machado  <luis.machado@linaro.org>
11016         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11017         <prefetch_dynamic_strides>: New const bool field.
11018         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11019         prefetch_dynamic_strides.
11020         (exynosm1_prefetch_tune): Likewise.
11021         (thunderxt88_prefetch_tune): Likewise.
11022         (thunderx_prefetch_tune): Likewise.
11023         (thunderx2t99_prefetch_tune): Likewise.
11024         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11025         to false.
11026         (aarch64_override_options_internal): Update to set
11027         PARAM_PREFETCH_DYNAMIC_STRIDES.
11028         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11029         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11030         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11031         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11032         prefetch-dynamic-strides setting.
11034         2018-05-07  Luis Machado  <luis.machado@linaro.org>
11036         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11037         <minimum_stride>: New const int field.
11038         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11039         minimum_stride field.
11040         (exynosm1_prefetch_tune): Likewise.
11041         (thunderxt88_prefetch_tune): Likewise.
11042         (thunderx_prefetch_tune): Likewise.
11043         (thunderx2t99_prefetch_tune): Likewise.
11044         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11045         (aarch64_override_options_internal): Update to set
11046         PARAM_PREFETCH_MINIMUM_STRIDE.
11047         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11048         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11049         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11050         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11051         stride is constant and is below the minimum stride threshold.
11053 2018-05-07  Luis Machado  <luis.machado@linaro.org>
11055         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11056         to 512.
11058 2018-05-07  Luis Machado  <luis.machado@linaro.org>
11060         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11061         <prefetch_dynamic_strides>: New const bool field.
11062         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11063         prefetch_dynamic_strides.
11064         (exynosm1_prefetch_tune): Likewise.
11065         (thunderxt88_prefetch_tune): Likewise.
11066         (thunderx_prefetch_tune): Likewise.
11067         (thunderx2t99_prefetch_tune): Likewise.
11068         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11069         to false.
11070         (aarch64_override_options_internal): Update to set
11071         PARAM_PREFETCH_DYNAMIC_STRIDES.
11072         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11073         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11074         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11075         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11076         prefetch-dynamic-strides setting.
11078 2018-05-07  Luis Machado  <luis.machado@linaro.org>
11080         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11081         <minimum_stride>: New const int field.
11082         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11083         minimum_stride field.
11084         (exynosm1_prefetch_tune): Likewise.
11085         (thunderxt88_prefetch_tune): Likewise.
11086         (thunderx_prefetch_tune): Likewise.
11087         (thunderx2t99_prefetch_tune): Likewise.
11088         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11089         (aarch64_override_options_internal): Update to set
11090         PARAM_PREFETCH_MINIMUM_STRIDE.
11091         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11092         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11093         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11094         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11095         stride is constant and is below the minimum stride threshold.
11097 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
11099         PR c++/85659
11100         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11101         the type is addressable.  Don't force op into register if it has
11102         BLKmode.
11104 2018-05-05  Roland McGrath  <mcgrathr@google.com>
11106         PR other/77609
11107         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11108         any section for which we don't know a specific type it should have,
11109         regardless of name.  Previously this was done only for the exact
11110         names ".init_array", ".fini_array", and ".preinit_array".
11111         (default_elf_asm_named_section): Add comment about
11112         relationship with default_section_type_flags and SECTION_NOTYPE.
11113         (get_section): Don't consider it a type conflict if one side has
11114         SECTION_NOTYPE and the other doesn't, as long as neither has the
11115         SECTION_BSS et al used in the default_section_type_flags logic.
11117 2018-05-05  Tom de Vries  <tom@codesourcery.com>
11119         PR target/85653
11120         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11121         (workaround_barsyncs): New function.
11122         (nvptx_reorg): Use workaround_barsyncs.
11123         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11124         (define_expand "nvptx_membar_cta"): New define_expand.
11125         (define_insn "*nvptx_membar_cta"): New insn.
11127 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
11129         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11130         To improve optimization opportunities.
11131         * builtin-types.def: The new needed builtin types for the above.
11133 2018-05-04  Richard Biener  <rguenther@suse.de>
11135         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11136         * gimple-ssa-store-merging.c
11137         (imm_store_chain_info::output_merged_store): Remove redundant create,
11138         release split_store vector contents on failure.
11139         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11140         scalar stmt vector on cache hit.
11142 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
11144         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11145         Xilinx FP support.
11146         * config.gcc (powerpc-xilinx-eabi*): Remove.
11147         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11148         support.
11149         (fusion_addis_mem_combo_load): Ditto.
11150         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11151         FP support.
11152         (rs6000_cpu_cpp_builtins): Ditto.
11153         * config/rs6000/rs6000-linux.c
11154         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11155         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11156         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11157         support.
11158         (rs6000_setup_reg_addr_masks): Ditto.
11159         (rs6000_init_hard_regno_mode_ok): Ditto.
11160         (rs6000_option_override_internal): Ditto.
11161         (legitimate_lo_sum_address_p): Ditto.
11162         (rs6000_legitimize_address): Ditto.
11163         (rs6000_legitimize_reload_address): Ditto.
11164         (rs6000_legitimate_address_p): Ditto.
11165         (abi_v4_pass_in_fpr): Ditto.
11166         (setup_incoming_varargs): Ditto.
11167         (rs6000_gimplify_va_arg): Ditto.
11168         (rs6000_split_multireg_move): Ditto.
11169         (rs6000_savres_strategy): Ditto.
11170         (rs6000_emit_prologue_components): Ditto.
11171         (rs6000_emit_epilogue_components): Ditto.
11172         (rs6000_emit_prologue): Ditto.
11173         (rs6000_emit_epilogue): Ditto.
11174         (rs6000_elf_file_end): Ditto.
11175         (rs6000_function_value): Ditto.
11176         (rs6000_libcall_value): Ditto.
11177         * config/rs6000/rs6000.h: Ditto.
11178         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11179         (TARGET_MINMAX): ... this.  New.
11180         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11181         * config/rs6000/rs6000.md: Remove Xilinx FP support.
11182         (*movsi_internal1_single): Delete.
11183         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11184         mfpu=, mxilinx-fpu): Delete.
11185         * config/rs6000/singlefp.h: Delete.
11186         * config/rs6000/sysv4.h: Remove Xilinx FP support.
11187         * config/rs6000/t-rs6000: Ditto.
11188         * config/rs6000/t-xilinx: Delete.
11189         * config/rs6000/titan.md: Adjust for fp_type removal.
11190         * config/rs6000/vsx.md: Remove Xilinx FP support.
11191         (VStype_simple): Delete.
11192         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11193         * config/rs6000/xfpu.h: Delete.
11194         * config/rs6000/xfpu.md: Delete.
11195         * config/rs6000/xilinx.h: Delete.
11196         * config/rs6000/xilinx.opt: Delete.
11197         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11198         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11200 2018-05-04  Tom de Vries  <tom@codesourcery.com>
11202         PR libgomp/85639
11203         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11204         if ignore == 0.
11206 2018-05-04  Richard Biener  <rguenther@suse.de>
11208         PR middle-end/85627
11209         * tree-complex.c (update_complex_assignment): We are always in SSA form.
11210         (expand_complex_div_wide): Likewise.
11211         (expand_complex_operations_1): Likewise.
11212         (expand_complex_libcall): Preserve EH info of the original stmt.
11213         (tree_lower_complex): Handle removed blocks.
11214         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11215         on complex multiplication and division libcall builtins.
11217 2018-05-04  Richard Biener  <rguenther@suse.de>
11219         PR middle-end/85574
11220         * fold-const.c (negate_expr_p): Restrict negation of operand
11221         zero of a division to when we know that can happen without
11222         overflow.
11223         (fold_negate_expr_1): Likewise.
11225 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
11227         PR libstdc++/85466
11228         * real.h (real_nextafter): Declare.
11229         * real.c (real_nextafter): New function.
11230         * fold-const-call.c (fold_const_nextafter): New function.
11231         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11232         CASE_CFN_NEXTTOWARD.
11233         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11234         even when arg1_mode is different from arg0_mode.
11236 2018-05-03  Nathan Sidwell  <nathan@acm.org>
11238         * doc/extend.texi (Deprecated Features): Remove
11239         -ffriend-injection.
11240         (Backwards Compatibility): Likewise.
11241         * doc/invoke.texi (C++ Language Options): Likewise.
11242         (C++ Dialect Options): Likewise.
11244 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
11246         PR target/85530
11247         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11248         _mm512_mask_mullox_epi64): New intrinsics.
11250 2018-05-03  Tom de Vries  <tom@codesourcery.com>
11252         PR testsuite/85106
11253         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11254         dump files): Add offload-tree.
11256 2018-05-03  Richard Biener  <rguenther@suse.de>
11258         PR tree-optimization/85615
11259         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11260         to loops not nested in BBs loop father to avoid creating multi-entry
11261         loops.
11263 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11265         PR tree-optimization/70291
11266         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11267         arguments.  Change return type to tree.  Emit libcall as a new
11268         statement rather than replacing existing one when inplace_p is true.
11269         (expand_complex_multiplication_components): New function.
11270         (expand_complex_multiplication): Expand floating-point complex
11271         multiplication using the above.
11272         (expand_complex_division): Rename inner_type parameter to type.
11273         Update expand_complex_libcall call-site.
11274         (expand_complex_operations_1): Update expand_complex_multiplication
11275         and expand_complex_division call-sites.
11277 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
11279         PR target/85582
11280         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11281         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11282         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11283         the highest significant bit of the shift count mask is clear.  In
11284         check whether and[sq]i3 is needed verify that all significant bits
11285         of the shift count other than the highest are set.
11287 2018-05-02  Tom de Vries  <tom@codesourcery.com>
11289         PR libgomp/82428
11290         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11291         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11292         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11293         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11294         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11295         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11296         __builtin_goacc_parlevel_size.
11298 2018-05-02  Richard Biener  <rguenther@suse.de>
11300         PR tree-optimization/85597
11301         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11302         do not use split vect_get_vec_defs call but call vect_get_slp_defs
11303         directly.
11305 2018-05-02  Tom de Vries  <tom@codesourcery.com>
11307         PR testsuite/85106
11308         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11309         dump files): Add ltrans-tree.
11311 2018-05-02  Tom de Vries  <tom@codesourcery.com>
11313         PR testsuite/85106
11314         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11315         dump files): Add wpa-ipa.
11317 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
11319         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
11320         powerpc*-*-linux*paired* target.
11321         * config/rs6000/750cl.h: Delete.
11322         * config/rs6000/paired.h: Delete.
11323         * config/rs6000/paired.md: Delete.
11324         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11325         float support.
11326         * config/rs6000/rs6000-builtin.def: Remove paired float support.
11327         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11328         comment.  Remove paired float support.
11329         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11330         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11331         VECTOR_PAIRED.
11332         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11333         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11334         declarations.
11335         * config/rs6000/rs6000.c: Remove paired float support.
11336         (paired_expand_vector_init, paired_expand_vector_move,
11337         paired_emit_vector_compare, paired_emit_vector_cond_expr,
11338         (paired_expand_lv_builtin, paired_expand_stv_builtin,
11339         paired_expand_builtin, paired_expand_predicate_builtin,
11340         paired_init_builtins): Delete.
11341         * config/rs6000/rs6000.h: Remove paired float support.
11342         * config/rs6000/rs6000.md: Remove paired float support.
11343         (move_from_CR_ov_bit): Delete.
11344         * config/rs6000/rs6000.opt (mpaired): Delete.
11345         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11346         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11348 2018-05-02  Richard Biener  <rguenther@suse.de>
11350         PR middle-end/85567
11351         * gimplify.c (gimplify_save_expr): When in SSA form allow
11352         SAVE_EXPRs to compute to SSA vars.
11354 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
11356         PR target/85582
11357         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11358         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11359         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11360         clobber operands[2], instead use a new pseudo.  Formatting fixes.
11362 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
11364         PR tree-optimization/85586
11365         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11366         exit early for statements in the same group if the accesses are
11367         not strided.
11369 2018-05-02  Tom de Vries  <tom@codesourcery.com>
11371         PR lto/85451
11372         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11373         error message.
11375 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
11377         PR tree-optimization/85143
11378         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11380 2018-05-01  Tom de Vries  <tom@codesourcery.com>
11382         PR lto/85451
11383         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11384         not found" error message.
11386 2018-05-01  Tom de Vries  <tom@codesourcery.com>
11388         PR other/83786
11389         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11390         * vec.c (test_ordered_remove_if): New function.
11391         (vec_c_tests): Call test_ordered_remove_if.
11392         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11393         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11394         * tree-vect-patterns.c (vect_pattern_recog_1): Use
11395         VEC_ORDERED_REMOVE_IF.
11397 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11399         PR tree-optimization/82665
11400         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11401         pointer subtraction where arguments come from a memchr call.
11403 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
11405         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11406         --push-state --as-needed and --pop-state instead of --as-needed and
11407         --no-as-needed if ld supports it.
11408         * configure: Regenerated.
11410         PR web/85578
11411         * doc/install.texi2html: Replace _002d with - and _002a with * in
11412         generated html files using sed.
11414 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
11416         PR c++/85523
11417         * gcc-rich-location.c (blank_line_before_p): New function.
11418         (use_new_line): New function.
11419         (gcc_rich_location::add_fixit_insert_formatted): New function.
11420         * gcc-rich-location.h
11421         (gcc_rich_location::add_fixit_insert_formatted): New function.
11423 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
11425         * selftest.c (assert_streq): Rename "expected" and "actual" to
11426         "val1" and "val2".  Extend NULL-handling to cover both inputs
11427         symmetrically, while still requiring both to be non-NULL for a pass.
11428         * selftest.h (assert_streq): Rename "expected" and "actual" to
11429         "val1" and "val2".
11430         (ASSERT_EQ): Likewise.
11431         (ASSERT_EQ_AT): Likewise.
11432         (ASSERT_KNOWN_EQ): Likewise.
11433         (ASSERT_KNOWN_EQ_AT): Likewise.
11434         (ASSERT_NE): Likewise.
11435         (ASSERT_MAYBE_NE): Likewise.
11436         (ASSERT_MAYBE_NE_AT): Likewise.
11437         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
11438         the assertion to pass.
11439         (ASSERT_STREQ_AT): Likewise.
11441 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
11443         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11444         interaction with -pie.
11446 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
11448         * selftest.h: Fix alphabetization of per-source-file selftest
11449         declarations.
11451 2018-04-30  Jason Merrill  <jason@redhat.com>
11453         PR c++/61982 - dead stores to destroyed objects.
11454         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11455         of clobber.
11457 2018-04-30  Jason Merrill  <jason@redhat.com>
11459         * tree.c (build_clobber): New.
11460         * tree.h: Declare it.
11461         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11463 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
11465         * diagnostic-show-locus.c (layout::layout): Update for
11466         location_get_source_line returning a char_span.
11467         (struct char_span): Move to input.h.
11468         (struct correction): Update for fields in char_span becoming
11469         private.
11470         (struct source_line): Update for location_get_source_line
11471         returning a char_span.
11472         (layout::print_line): Likewise.
11473         * edit-context.c (edited_file::print_content): Likewise.
11474         (edited_file::print_diff_hunk): Likewise.
11475         (edited_file::print_run_of_changed_lines): Likewise.
11476         (edited_file::get_num_lines): Likewise.
11477         (edited_line::edited_line): Likewise.
11478         * final.c (asm_show_source): Likewise.
11479         * input.c (location_get_source_line): Convert return type
11480         from const char * to char_span, losing the final "line_len"
11481         param.
11482         (dump_location_info): Update for the above.
11483         (get_substring_ranges_for_loc): Likewise.  Use a char_span
11484         when handling the literal within the line.
11485         (test_reading_source_line): Update for location_get_source_line
11486         returning a char_span.
11487         * input.h (class char_span): Move here from
11488         diagnostic-show-locus.c, converting from a struct to a class.
11489         Make data members private.
11490         (char_span::operator bool): New.
11491         (char_span::length): New.
11492         (char_span::get_buffer): New.
11493         (char_span::operator[]): New.
11494         (char_span::subspan): Make const.
11495         (char_span::xstrdup): New.
11496         (location_get_source_line): Convert return type from const char *
11497         to char_span, losing the final "line_size" param.
11499 2018-04-30  Jan Hubicka  <jh@suse.cz>
11501         * lto-wrapper.c (ltrans_priorities): New static var.
11502         (cmp_priority): New.
11503         (run_gcc): Read priorities and if doing parallel build order
11504         the Makefile by them.
11506 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
11508         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11510 2018-04-30  Richard Biener  <rguenther@suse.de>
11512         * tree-cfg.c (verify_address): Remove base argument, add
11513         flag whether to check TREE_ADDRESSABLE and do that.
11514         (verify_expr): Remove.
11515         (verify_types_in_gimple_reference): Add pieces from verify_expr.
11516         (verify_gimple_assign_single): Likewise.
11517         (verify_gimple_switch): Likewise.
11518         (verify_expr_location_1): Dereference tp once.  Add (disabled)
11519         piece from verify_expr.
11520         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11522 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
11524         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11526 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
11528         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11529         (small_data_pattern): Likewise.
11530         (arc_rewrite_small_data): Likewise.
11531         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11532         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11533         (get_symbol_alignment): New function.
11534         (legitimate_small_data_address_p): Likewise.
11535         (legitimate_scaled_address): Update, call
11536         legitimate_small_data_address_p.
11537         (output_sdata): New static variable.
11538         (arc_print_operand): Update how we handle small data operands.
11539         (arc_print_operand_address): Likewise.
11540         (arc_legitimate_address_p): Update, use
11541         legitimate_small_data_address_p.
11542         (arc_rewrite_small_data_p): Remove.
11543         (arc_rewrite_small_data_1): Likewise.
11544         (arc_rewrite_small_data): Likewise.
11545         (small_data_pattern): Likewise.
11546         (compact_sda_memory_operand): Update to use
11547         legitimate_small_data_address_p and get_symbol_alignment.
11548         (prepare_move_operands): Don't rewite sdata pattern.
11549         (prepare_extend_operands): Remove.
11550         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11551         pattern.
11552         (zero_extendqisi2): Likewise.
11553         (zero_extendhisi2): Likewise.
11554         (extendqihi2): Likewise.
11555         (extendqisi2): Likewise.
11556         (extendhisi2): Likewise.
11557         (addsi3): Likewise.
11558         (subsi3): Likewise.
11559         (andsi3): Likewise.
11560         * config/arc/constraints.md (Usd): Change it to memory constraint.
11562 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
11564         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11565         as source of std instructions.
11566         * config/arc/arc.md (movsi_insn): Update pattern predicate to
11567         allow 6-bit constants as source for store instructions.
11568         (movdi_insn): Update instruction pattern to allow 6-bit constants
11569         as source for store instructions.
11571 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
11573         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11575 2018-04-30  Nathan Sidwell  <nathan@acm.org>
11576             Sandra Loosemore <sandra@codesourcery.com>
11578         * dumpfile.c (dump_open): Allow '-' for stdout.
11579         * doc/invoke.texi (Developer Options): Document dump filename
11580         determination early.  Document stdin/stdout selection.
11582 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
11584         Microblaze Target: PIC data text relative
11586         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11587         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11588         Add declaration.
11589         * config/microblaze/microblaze.h (microblaze_constant_address_p):
11590         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11591         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11592         New addressing mode for data-text relative position indepenedent code.
11593         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11594         'ADDRESS_SYMBOLIC_TXT_REL'.
11595         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11596         (microblaze_legitimate_pic_operand): Exclude function calls from
11597         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11598         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11599         addresses cases.
11600         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11601         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11602         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11603         for 'address + offset'.
11604         (microblaze_expand_prologue): Add new function prologue call for
11605         'r20' assignation.
11606         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11607         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11608         table in case of TARGET_PIC_DATA_TEXT_REL.
11609         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11610         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11611         Add new macros 'UNSPEC_TEXT',
11612         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11613         + exclude function calls from 'UNSPEC_PLT' in case of data text
11614         relative mode.
11615         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11616         new target hook for generating address diff vector tables in case of
11617         flag_pic.
11618         * doc/tm.texi : Regenerate.
11619         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11620         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11621         of addr diff vector generation.
11622         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11623         target hook definition.
11624         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11625         Add default function for generate_pic_addr_diff_vec -> flag_pic.
11626         * doc/invoke.texi (Add new pic option): Add new microblaze pic
11627         option for data text relative.
11629 2018-04-30  Richard Biener  <rguenther@suse.de>
11631         * tree-chrec.h (evolution_function_is_constant_p): Remove
11632         redundant check.
11633         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11635 2018-04-30  Richard Biener  <rguenther@suse.de>
11637         PR bootstrap/85571
11638         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11640 2018-04-30  Richard Biener  <rguenther@suse.de>
11642         PR tree-optimization/28364
11643         PR tree-optimization/85275
11644         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11645         copying first exit test.
11647 2018-04-28  Mark Wielaard  <mark@klomp.org>
11649         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11650         dwarf_version >= 5.
11651         (dwarf_AT): Handle DW_AT_addr_base.
11652         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11654 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
11656         PR target/84431
11657         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11658         (*ashl<dwi>3_doubleword_mask_1): Ditto.
11659         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11660         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11662 2018-04-28  Richard Biener  <rguenther@suse.de>
11664         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11665         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11666         to reflect use.  Only add interesting stmts.
11668 2018-04-27  Martin Jambor  <mjambor@suse.cz>
11670         PR ipa/85549
11671         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11672         the jump function allows for passing through aggregate values.
11674 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
11676         * input.h (in_system_header_at): Convert from macro to inline
11677         function.
11678         (from_macro_expansion_at): Likewise.
11679         (from_macro_definition_at): Likewise.
11681 2018-04-27  Jeff Law  <law@redhat.com>
11683         * config.gcc: Mark tile* targets as deprecated/obsolete.
11685 2018-04-27  Richard Biener  <rguenther@suse.de>
11687         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11688         fix for ILP32.
11690 2018-04-27  Richard Biener  <rguenther@suse.de>
11692         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11694 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
11696         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11697         with Yd constraint. Set "preferred_for_speed" attribute from
11698         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11699         with Yd constraint.
11700         (*movdi_internal): Ditto.
11701         (movti_interunit splitters): Remove
11702         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11703         (movdi_interunit splitters): Ditto.
11704         * config/i386/constraints.md (Ye): Remove.
11705         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11707 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11709         PR target/85512
11710         * config/aarch64/constraints.md (Usg): Limit to 31.
11711         (Usj): Limit to 63.
11713 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
11715         PR tree-optimization/85529
11716         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11717         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11718         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11719         zero extension or masking of the MSB bit.
11720         (optimize_range_tests): Add FIRST_BB argument, pass it through
11721         to optimize_range_tests_var_bound.
11722         (maybe_optimize_range_tests, reassociate_bb): Adjust
11723         optimize_range_tests callers.
11725 2018-04-26  Richard Biener  <rguenther@suse.de>
11726             Jakub Jelinek  <jakub@redhat.com>
11728         * cgraph.h (symbol_table): Just declare debug method here.
11729         * symtab.c (symbol_table::debug): Define.
11731 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
11733         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11735 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
11737         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11738         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11739         (*movdi_internal): Substitute Yi and Yj constraint with x
11740         and Ym and Yn constraint with y constraint.  Update "isa"
11741         attribute and set "preferred_for_speed" attribute from
11742         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11743         (*movsi_internal): Ditto.
11744         (*movdf_internal): Ditto.
11745         (*movsf_internal): Ditto.
11746         (*zero_extendsidi2): Ditto.
11747         * config/i386/sse.md (vec_set<mode>_0): Ditto.
11748         (sse2_loadld): Ditto.
11749         (*vec_extract<ssevecmodelower>_0): Ditto.
11750         (*vec_extractv4si_0_zext_sse4): Ditto.
11751         (vec_concatv2di): Ditto.
11752         (*vec_dup<mode>): Ditto.
11753         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11754         * config/i386/constraints.md (Yi): Remove.
11755         (Yj): Remove.
11756         (Ym): Remove.
11757         (Yn): Remove.
11759 2018-04-26  Nathan Sidwell  <nathan@acm.org>
11761         * dumpfile.c (dump_open): New.
11762         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11763         (dump_finish): Detect stdio/stderr by value not name.
11765 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
11767         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11769 2018-04-26  Tom de Vries  <tom@codesourcery.com>
11771         PR target/84952
11772         * config/nvptx/nvptx.c (verify_neutering_jumps)
11773         (verify_neutering_labels): New function
11774         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11776 2018-04-26  Tom de Vries  <tom@codesourcery.com>
11778         PR target/84025
11779         * config/nvptx/nvptx.c (needs_neutering_p): New function.
11780         (nvptx_single): Use needs_neutering_p to skip over insns that do not
11781         need neutering.
11783 2018-04-26  Richard Biener <rguenther@suse.de>
11784             Tom de Vries  <tom@codesourcery.com>
11786         PR lto/85422
11787         * lto-streamer-out.c (output_function): Fixup loops if required to match
11788         discovery done in the reader.
11790 2018-04-26  Richard Biener  <rguenther@suse.de>
11792         PR tree-optimization/85116
11793         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11794         have a loop exit from the single latch predecessor.  Remove
11795         case of header with just condition.
11796         (ch_base::copy_headers): Exclude infinite loops from any
11797         processing.
11798         (pass_ch::execute): Record exits.
11800 2018-04-26  Richard Biener  <rguenther@suse.de>
11802         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11803         prologue cost vector and pass it to vect_get_load_cost.
11804         (vect_get_peeling_costs_all_drs): Likewise.
11805         (vect_peeling_hash_get_lowest_cost): Likewise.
11806         (vect_enhance_data_refs_alignment): Likewise.
11808 2018-04-26  Richard Biener  <rguenther@suse.de>
11810         PR middle-end/85450
11811         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11812         checking of integer<->pointer conversions.
11813         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11814         sign-/zero-extending pointer types.
11815         (expand_omp_for_static_chunk): Likewise.
11817 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
11818             Jean Lee  <xiaoyur347@gmail.com>
11820         * config/mips/mips.c (mips_asan_shadow_offset): New function.
11821         (TARGET_ASAN_SHADOW_OFFSET): Define.
11822         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11823         true for -fsanitize=address.
11825 2018-04-25  Mark Wielaard  <mark@klomp.org>
11827         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11828         shorter ones.
11830 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
11832         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11833         than "alu", remove explicit "memory" and "imm_disp" attributes.
11834         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11836         PR middle-end/85414
11837         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11838         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11839         gen_lowpart_no_emit.
11841 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
11843         PR target/85473
11844         * config/i386/i386.c (ix86_expand_builtin): Change memory
11845         operand to XI, extend p0 to Pmode.
11846         * config/i386/i386.md: Change unspec volatile and operand
11847         1 mode to XI, change operand 0 mode to P.
11849 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
11851         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11852         GET_MODE_MASK before any checking.
11853         (nds32_can_use_bset_p): Likewise.
11854         (nds32_can_use_btgl_p): Likewise.
11856 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
11858         * config/nds32/nds32-doubleword.md: New define_split pattern for
11859         illegal register number.
11861 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
11863         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11865 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
11867         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11869 2018-04-25  Richard Biener  <rguenther@suse.de>
11871         * lto-streamer.h (LTO_major_version): Bump to 8.
11873 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
11875         * BASE-VER: Set to 9.0.0.
11877 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
11879         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11880         in __abskf2 and __powikf2.
11882 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11884         PR target/85512
11885         * config/aarch64/constraints.md (Usg, Usj): New constraints.
11886         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11887         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
11888         Use the above on operand 2.  Reindent.
11889         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
11891 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
11893         PR target/85485
11894         * common/config/i386/i386-common.c (ix86_handle_option): Don't
11895         handle OPT_mcet.
11896         * config/i386/i386.opt (mcet): Removed.
11897         * doc/install.texi: Remove -mcet documentation.
11898         * doc/invoke.texi: Likewise.
11900 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
11902         PR target/85485
11903         * doc/install.texi: Remove -mcet from bootstrap-cet.
11905 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
11907         PR target/85511
11908         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11909         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11910         if TARGET_64BIT.
11912         PR target/85503
11913         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11914         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11915         containing a CONST_VECTOR.
11917 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
11919         * doc/install.texi: Update newlib dependency for nvptx.
11921 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
11923         PR target/85508
11924         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
11925         instead of INTVAL when shifting x left.
11927 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
11929         PR tree-optimization/85478
11930         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
11931         vect_grouped_store_supported for single element vectors.
11933 2018-04-24  Richard Biener  <rguenther@suse.de>
11935         PR target/85491
11936         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
11937         load cost increase to the case of non-constant step.
11939 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
11941         PR target/84828
11942         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
11943         destination if any_malformed_asm.
11945 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
11947         PR middle-end/85496
11948         * expr.c (store_field): In the bitfield case, if the value comes from
11949         a function call and is returned in registers by means of a PARALLEL,
11950         do not change the mode of the temporary unless BLKmode and VOIDmode.
11952 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
11954         PR rtl-optimization/85423
11955         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
11956         dependencies to debug insns when the previous insn is non-debug.
11958 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
11960         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
11961         enums into a single definition.
11962         (fls): Fix predicates and printing.
11963         (seti): Likewise.
11965 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
11967         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
11968         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
11969         and short u6 immediate.
11970         (check_if_valid_sleep_operand): Remove.
11971         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
11973 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
11975         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
11976         flag_always_save_lp condition.
11977         * config/nds32/nds32.opt (malways-save-lp): New option.
11979 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
11981         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
11982         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
11983         * config/nds32/nds32.h
11984         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
11985         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
11987 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
11989         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
11990         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
11992 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
11993             Chung-Ju Wu  <jasonwucj@gmail.com>
11995         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
11996         Declare.
11997         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
11998         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12000 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
12002         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12004 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
12006         * config/nds32/nds32-protos.h (nds32_data_alignment,
12007         nds32_local_alignment): Declare.
12008         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12009         nds32_local_alignment): New functions.
12010         (TARGET_CONSTANT_ALIGNMENT): Define.
12011         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12013 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
12015         * config/nds32/nds32.c
12016         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12017         (TARGET_MODES_TIEABLE_P): Likewise.
12019 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
12021         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12022         level Ofast and Og.
12024 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
12025             Chung-Ju Wu  <jasonwucj@gmail.com>
12027         * config/nds32/constants.md (unspec_volatile_element): Add enum values
12028         for unaligned access.
12029         * config/nds32/nds32-intrinsic.c: Implementation of expanding
12030         unaligned access.
12031         * config/nds32/nds32-intrinsic.md: Likewise.
12032         * config/nds32/nds32_intrinsic.h: Likewise.
12033         * config/nds32/nds32.h (nds32_builtins): Likewise.
12034         * config/nds32/nds32.opt (munaligned-access): New option.
12035         * config/nds32/nds32.c (nds32_asm_file_start): Display
12036         flag_unaligned_access status.
12038 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
12040         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12041         -mno-relax is present.
12042         * config/riscv/linux.h (LINK_SPEC): Ditto.
12044 2018-04-20  Martin Sebor  <msebor@redhat.com>
12046         PR c/85365
12047         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12048         for null pointers.
12049         (gimple_fold_builtin_stxcpy_chk): Same.
12050         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12052 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
12054         PR target/85456
12055         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12056         __powikf2 when long double is IEEE 128-bit.
12058 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
12060         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12061         step to make sure stack always aligned.
12063 2018-04-20  Carl Love  <cel@us.ibm.com>
12065         PR target/83402
12066         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12067         size check for arg0.
12069 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
12070             Tom de Vries  <tom@codesourcery.com>
12072         PR target/85445
12073         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12074         Emit insns for calls too.
12075         (nvptx_find_par): Always look for worker-level predecessor insn.
12076         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
12077         calls.
12078         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12079         (nvptx_process_pars): Propagate frames for calls.
12081 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
12083         PR target/85469
12084         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12085         Removed.
12086         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12087         (ix86_handle_option): Don't handle OPT_mibt.
12088         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12089         __SHSTK__.
12090         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12091         has_ibt and ibt.
12092         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12093         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12094         (ix86_target_macros): Define __CET__ with flag_cf_protection
12095         for -fcf-protection.
12096         * config/i386/i386.c (isa2_opts): Remove -mibt.
12097         * config/i386/i386.h (TARGET_IBT): Removed.
12098         (TARGET_IBT_P): Likewise.
12099         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12100         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12101         * config/i386/i386.opt (mcet): Update help message.
12102         (mshstk): Likewise.
12103         (mibt): Removed.
12104         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
12105         -mcet as an alias for -mshstk.
12107 2018-04-20  Richard Biener <rguenther@suse.de>
12109         PR middle-end/85475
12110         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12111         complexity by forcing a single use of the multiply operand.
12113 2018-04-20  Martin Jambor  <mjambor@suse.cz>
12115         ipa/85449
12116         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12117         recursion dependency to only apply to non-clones.
12119 2018-04-20  Martin Jambor  <mjambor@suse.cz>
12121         ipa/85447
12122         * ipa-cp.c (create_specialized_node): Check that clones of
12123         self-recursive edges exist during IPA-CP.
12125 2018-04-19  Toon Moene  <toon@moene.org>
12127         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12128         by -O3.
12130 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
12132         PR tree-optimization/85467
12133         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12134         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
12135         VECTOR_CST element to type.
12137 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
12139         PR target/85397
12140         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12141         * config/i386/i386.md (builtin_setjmp_setup): Removed.
12142         (builtin_longjmp): Likewise.
12143         (save_stack_nonlocal): New pattern.
12144         (restore_stack_nonlocal): Likewise.
12146 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
12148         PR target/85404
12149         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12150         Replace ASM_OUTPUT_LABEL with fprintf.
12152 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
12154         PR target/85417
12155         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12156         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12157         * config/i386/i386-c.c (ix86_target_macros_internal): Also
12158         define __IBT__ and __SHSTK__ for -fcf-protection.
12159         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12160         TARGET_IBT.
12161         (ix86_trampoline_init): Likewise.
12162         (x86_output_mi_thunk): Likewise.
12163         (ix86_notrack_prefixed_insn_p): Likewise.
12164         (ix86_option_override_internal): Don't disallow -fcf-protection.
12165         * config/i386/i386.md (rdssp<mode>): Also enable for
12166         -fcf-protection.
12167         (incssp<mode>): Likewise.
12168         (nop_endbr): Likewise.
12169         * config/i386/i386.opt (mcet): Change help message to built-in
12170         functions only.
12171         (mibt): Likewise.
12172         (mshstk): Likewise.
12173         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12174         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
12175         enable CET built-in functions.
12177 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
12179         * common/config/i386/i386-common.c
12180         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12181         OPTION_MASK_ISA_MOVDIRI_UNSET,
12182         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12183         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12184         * config.gcc (movdirintrin.h): New header.
12185         * config/i386/cpuid.h (bit_MOVDIRI,
12186         bit_MOVDIR64B): New bits.
12187         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12188         and -mmvodir64b.
12189         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12190         (VOID, PVOID, PCVOID)): New function types.
12191         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12192         __builtin_ia32_directstoreu_u64,
12193         __builtin_ia32_movdir64b): New builtins.
12194         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12195         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12196         and -mmovdiri.
12197         (ix86_valid_target_attribute_inner_p): Ditto.
12198         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12199         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12200         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12201         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12202         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12203         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12204         (movdiri<mode>, movdir64b_<mode>): New.
12205         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12206         * config/i386/immintrin.h: Include movdirintrin.h.
12207         * config/i386/movdirintrin.h: New file.
12208         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12210 2018-04-19  Richard Biener  <rguenther@suse.de>
12212         PR middle-end/85455
12213         * cfg.c (clear_bb_flags): When loop state says we have
12214         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12216 2018-04-19  Richard Biener  <rguenther@suse.de>
12218         PR tree-optimization/84737
12219         * tree-vect-data-refs.c (vect_copy_ref_info): New function
12220         copying restrict info.
12221         (vect_setup_realignment): Use it.
12222         * tree-vectorizer.h (vect_copy_ref_info): Declare.
12223         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12224         the first DR to all generated stores.
12225         (vectorizable_load): Likewise for loads.
12227 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
12229         PR tree-optimization/85446
12230         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12231         the integral and pointer types to have the same precision.
12233         * doc/install.texi: Document --disable-cet being the default and
12234         --enable-cet=auto.
12236 2018-04-18  Martin Liska  <mliska@suse.cz>
12238         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12239         style.
12241 2018-04-18  Martin Liska  <mliska@suse.cz>
12243         Revert
12244         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
12246         PR ipa/83983
12247         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12248         arguments if they are comparable.
12250 2018-04-18  Martin Liska  <mliska@suse.cz>
12252         Revert
12253         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
12255         PR lto/84805
12256         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12257         incomplete types.
12259 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
12261         PR target/85388
12262         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12263         ENDBR after calling __morestack.
12265 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
12267         PR jit/85384
12268         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12269         by using gcc_base_ver to generate a gcc_driver_version, and use
12270         it when generating GCC_DRIVER_NAME.
12271         * configure: Regenerate.
12273 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
12275         PR target/81084
12276         * config.gcc: Obsolete powerpc*-*-*spe*.
12278 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
12280         PR debug/84637
12281         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12282         (stabstr_D): Change type of unum from unsigned int to
12283         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
12284         type.
12286 2018-04-17  Jim Wilson  <jimw@sifive.com>
12288         PR 84856
12289         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12290         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12291         Set arg_pointer_offset after using pretend_args_size.
12293 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
12295         PR rtl-optimization/85431
12296         * dse.c (record_store): Ignore zero width stores.
12298         PR sanitizer/85230
12299         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
12300         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12301         __builtin_stack_restore rather than after it.
12302         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12303         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12304         argument instead of virtual_dynamic_stack_rtx.
12306 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12308         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12309         New prototype.
12310         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12311         Add note to error message to explain internal mapping of overloaded
12312         built-in function name to non-overloaded built-in function name.
12313         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12314         function.
12316 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
12318         PR target/85424
12319         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12320         where the inputs overlap with the output.
12322 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
12324         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12325         (=v, v) alternative and explicit "memory" attribute.
12326         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
12327         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12328         attributes.
12329         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12330         "sselog1" type instead of "sselog".
12331         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12332         "sselog".  Remove explicit "memory" attribute.
12333         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12334         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12335         attributes.
12336         (vec_extract_hi_v32hi): Merge all alternatives into one, use
12337         "sselog1" type instead of "sselog".  Remove explicit "memory"
12338         attribute.
12339         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12340         use "sselog1" type instead of "sselog".  Remove explicit "memory"
12341         attribute.
12342         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12343         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12344         attributes.
12345         (vec_extract_hi_v64qi): Merge all alternatives into one, use
12346         "sselog1" type instead of "sselog".  Remove explicit "memory"
12347         attribute.
12348         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12349         use "sselog1" type instead of "sselog".  Remove explicit "memory"
12350         attribute.
12352         PR target/85430
12353         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12355         PR middle-end/85414
12356         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12357         on a SUBREG.
12359 2018-04-17  Martin Jambor  <mjambor@suse.cz>
12361         PR ipa/85421
12362         * ipa-cp.c (create_specialized_node): Call
12363         expand_all_artificial_thunks if necessary.
12365 2018-04-17  Martin Liska  <mliska@suse.cz>
12367         PR lto/85405
12368         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12369         in message, remote space in between '_G' and '('.
12371 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
12373         PR target/85281
12374         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12375         avx512f_vmcmp<mode>3<round_saeonly_name>,
12376         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12377         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12378         avx512f_rndscale<mode><round_saeonly_name>,
12379         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12380         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12381         Use %<iptr>2 instead of %2 for -masm=intel.
12382         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12383         avx512f_vcvttss2usi<round_saeonly_name>,
12384         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12385         -masm=intel.
12386         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12387         avx512f_vcvttsd2usi<round_saeonly_name>,
12388         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12389         Use %q1 instead of %1 for -masm=intel.
12390         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12391         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12392         of %3 for -masm=intel.
12393         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12394         -masm=intel.
12395         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12396         -masm=intel.
12397         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12398         -masm=intel.
12399         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12400         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12401         %g1.
12402         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12403         -masm=intel.
12404         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12405         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12406         %g1 and one with %0 and %1.
12407         (avx512er_vmrcp28<mode><round_saeonly_name>,
12408         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12409         %1 for -masm=intel.
12410         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12411         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12412         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12413         of %0 and %{%4%} for -masm=intel.
12414         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12415         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12416         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12417         order of %0 and %{%5%}%{z%} for -masm=intel.
12419 2018-04-17  Jan Hubicka  <jh@suse.cz>
12421         PR lto/85405
12422         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12424 2018-04-17  Martin Liska  <mliska@suse.cz>
12426         PR ipa/85329
12427         * multiple_target.c (create_dispatcher_calls): Set apostrophes
12428         for target_clone error message.  Make default implementation
12429         clone to be a local declaration.
12430         (separate_attrs): Add new argument and check for an empty
12431         string.
12432         (expand_target_clones): Handle it.
12433         (ipa_target_clone): Make redirection just for target_clones
12434         functions.
12436 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
12437             Tom de Vries  <tom@codesourcery.com>
12439         PR middle-end/84955
12440         * omp-expand.c (expand_oacc_for): Add dummy false branch for
12441         tiled basic blocks without omp continue statements.
12443 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
12445         PR target/83660
12446         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12447         vec_extract expression as having side effects to make sure it gets
12448         a cleanup point.
12450 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
12452         PR target/85403
12453         * config/i386/i386.c (get_builtin_code_for_version): Check
12454         error_mark_node.
12456 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
12458         PR target/84331
12459         * config.gcc: Support "skylake".
12460         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12461         PROCESSOR_SKYLAKE.
12462         * config/i386/i386.c (m_SKYLAKE): Define.
12463         (processor_target_table): Add "skylake".
12464         (ix86_option_override_internal): Add "skylake".
12465         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12466         PROCESSOR_CANNONLAKE.
12467         (get_builtin_code_for_version): Fix priority for
12468         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12469         PROCESSOR_SKYLAKE-AVX512.
12470         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12471         (processor_type): Add PROCESSOR_SKYLAKE.
12473 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
12474             Jason Merrill  <jason@redhat.com>
12476         PR c++/85112
12477         * convert.c (convert_to_integer_1): Use direct recursion for
12478         enumeral types and types with a precision less than the number
12479         of bits in their mode.
12481 2018-04-16  Julia Koval  <julia.koval@intel.com>
12483         PR target/84413
12484         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12485         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12487 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
12489         PR target/85293
12490         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12491         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12492         and -mno-direct-move.
12494 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
12496         PR target/83402
12497         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12498         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12499         Ensure negative shifts result in {0}.
12501 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
12503         PR rtl-optimization/79916
12504         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12505         regs (if any) to define how to gnerate SD moves when LRA is in
12506         progress.
12508 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
12510         PR rtl-optimization/85393
12511         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12512         * except.c (expand_dw2_landing_pad_for_region): Make static.
12513         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12514         a label and unconditional jump to old_bb, rather than
12515         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12516         basic block.
12518         PR rtl-optimization/85376
12519         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12520         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12521         instead of a specific value.
12523 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
12524             Bin Cheng  <bin.cheng@arm.com>
12526         PR tree-optimization/82965
12527         PR tree-optimization/83991
12528         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12529         by_profile_only parameter.
12530         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12531         information if the loop was predicted to iterate too many times.
12532         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12534 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
12536         PR lto/71991
12537         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12538         always inline.
12540 2018-04-13  Martin Liska  <mliska@suse.cz>
12541             Jakub Jelinek  <jakub@redhat.com>
12543         PR middle-end/81657
12544         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12545         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12546         * builtins.c (expand_builtin_memory_copy_args): Use
12547         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12548         handle dest_addr == pc_rtx.
12550 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
12552         PR target/85291
12553         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12554         asked to not generate direct moves.
12555         (fix_trunc<mode>si2_stfiwx): Similar.
12556         (fix_trunc<mode>si2_internal): Similar.
12558 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
12560         PR debug/83157
12561         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12562         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12563         lookup if dest in some wider mode is known to be const0_rtx and
12564         if so, record permanent equivalence for it to be ZERO_EXTEND of
12565         the narrower mode destination.
12567 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
12569         * lto-streamer-out.c (output_function): Revert 259346.
12570         * omp-expand.c (expand_oacc_for): Likewise.
12572 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
12574         PR rtl-optimization/85354
12575         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12576         * sel-sched.c (sel_global_init): ... here.
12578 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
12580         PR target/85238
12581         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12582         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12583         mode for PE-COFF targets.
12584         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12585         (i386_pe_asm_lto_end): Likewise.
12586         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12587         (TARGET_ASM_LTO_END): Likewise.
12588         * config/i386/winnt.c (saved_debug_info_level): New static variable.
12589         (i386_pe_asm_lto_start): New function.
12590         (i386_pe_asm_lto_end): Likewise.
12592 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
12593             Richard Biener  <rguenther@suse.de>
12595         PR middle-end/84955
12596         * lto-streamer-out.c (output_function): Fix CFG loop state before
12597         streaming out.
12598         * omp-expand.c (expand_oacc_for): Handle calls to internal
12599         functions like regular functions.
12601 2018-04-12  Richard Biener  <rguenther@suse.de>
12603         PR lto/85371
12604         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12605         for the early LTO debug to properly generate references to it
12606         during DIE emission.  Do not re-use that for the skeleton for
12607         split-dwarf.
12608         (dwarf2out_early_finish): Likewise.
12610 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
12612         PR target/85328
12613         * config/i386/sse.md
12614         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12615         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12616         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12617         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12618         and output is a reg, avoid creating invalid lowpart subreg, but
12619         instead split into a 512-bit move.  Don't split if not AVX512VL,
12620         input is xmm16+ reg and output is a mem.
12621         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12622         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12623         xmm16+ reg and output is a mem.
12625 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
12627         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12628         also for flag_dwarf2_cfi_asm.
12630 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
12632         PR rtl-optimization/85342
12633         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12634         a bool scalar var inside of the loop instead.  Don't try to update
12635         recog_data.operand after failed apply_change_group.
12637 2018-04-12  Tom de Vries  <tom@codesourcery.com>
12639         PR target/85296
12640         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12641         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
12642         array with flexible array member as array without given dimension.
12643         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12644         argument for undefined param to true.
12646 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
12648         PR target/85321
12649         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12650         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12651         from PowerPC section.
12652         * config/rs6000/sysv4.opt (mcall-): Improve help text.
12653         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12654         help text that is too long.
12655         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12656         help text that is too long.
12657         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12658         help text that is too long.
12660 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
12662         * config/alpha/alpha.md (stack_probe_internal): Rename
12663         from "probe_stack".  Update all callers.
12665 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
12667         PR rtl-optimization/84566
12668         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12669         sched_macro_fuse_insns.
12671 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
12673         PR target/84301
12674         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12675         (compute_block_dependences): ... from here.
12677 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
12679         PR tree-optimization/85331
12680         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12681         from int to HOST_WIDE_INT.
12683 2018-04-11  Martin Jambor  <mjambor@suse.cz>
12685         PR ipa/84149
12686         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12687         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12688         not the same as the source val.
12689         (cgraph_edge_brings_value_p): New parameter.
12690         (gather_edges_for_value): Pass destination value to
12691         cgraph_edge_brings_value_p.
12692         (perhaps_add_new_callers): Likewise.
12693         (get_info_about_necessary_edges): Likewise and exclude values brought
12694         only by self-recursive edges.
12695         (create_specialized_node): Redirect only clones of self-calling edges.
12696         (+self_recursive_pass_through_p): New function.
12697         (find_more_scalar_values_for_callers_subset): Use it.
12698         (find_aggregate_values_for_callers_subset): Likewise.
12699         (known_aggs_to_agg_replacement_list): Removed.
12700         (decide_whether_version_node): Re-calculate known constants for all
12701         remaining context clones.
12703 2018-04-11  Richard Biener  <rguenther@suse.de>
12705         PR lto/85339
12706         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12707         from early DWARF output.
12708         (dwarf2out_early_finish): Output line info unconditionally into
12709         early DWARF and add reference to it.
12711 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
12713         PR target/85281
12714         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12715         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12716         other than V2DFmode using iptr mode attribute.
12717         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12719 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
12721         PR rtl-optimization/84659
12722         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12724 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
12726         PR debug/85302
12727         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12728         SIZEP is NULL.
12729         (output_loc_list): Pass address of a dummy size variable even in the
12730         locview handling loop.
12731         (index_location_lists): Add comment on why skip_loc_list_entry can't
12732         call size_of_locs.
12734 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12736         PR target/85261
12737         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12738         into register.
12740 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
12742         PR target/85321
12743         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12744         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12745         and -mstring-compare-inline-limit.
12747 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
12749         PR target/85287
12750         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12751         for stack clash protection in a register whenever we need it to be in
12752         a register.
12754 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
12756         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12757         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12759 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
12761         PR target/85321
12762         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12763         the help text.
12764         (mlong-double-): Ditto.
12765         * config/rs6000/sysv4.opt (msdata=): Ditto.
12766         (mtls-size=): Ditto.
12768 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12770         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12771         erroneous entries for
12772         "vector int vec_ldl (int, long int *)", and
12773         "vector unsigned int vec_ldl (int, unsigned long int *)".
12774         Add comments and entries for
12775         "vector bool char vec_ldl (int, bool char *)",
12776         "vector bool short vec_ldl (int, bool short *)",
12777         "vector bool int vec_ldl (int, bool int *)",
12778         "vector bool long long vec_ldl (int, bool long long *)",
12779         "vector pixel vec_ldl (int, pixel *)",
12780         "vector long long vec_ldl (int, long long *)",
12781         "vector unsigned long long vec_ldl (int, unsigned long long *)".
12782         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12783         type tree bool_long_long_type_node and correct definition of
12784         bool_V2DI_type_node to make reference to this new type tree.
12785         (rs6000_mangle_type): Replace erroneous reference to
12786         bool_long_type_node with bool_long_long_type_node.
12787         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12788         comments to emphasize sign distinctions for char and int types and
12789         replace RS6000_BTI_bool_long constant with
12790         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
12791         use of RS6000_BTI_pixel.
12792         (bool_long_type_node): Remove this macro definition.
12793         (bool_long_long_type_node): New macro definition
12795 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
12797         PR rtl-optimization/85300
12798         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12799         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12800         simplify_unary_operation fails.
12802 2018-04-10  Martin Liska  <mliska@suse.cz>
12804         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12805         cgraph_edge and ipa_ref.
12807 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
12809         PR target/85177
12810         PR target/85255
12811         * config/i386/sse.md
12812         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12813         computation of the VEC_MERGE selector from mask.
12814         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12815         Fix decoding of the VEC_MERGE selector into mask.
12817 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
12819         PR tree-optimization/85286
12820         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12822 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
12824         * final.c (final_1): Set insn_last_address as well as
12825         insn_current_address.
12827 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12829         PR target/85173
12830         * explow.c (emit_stack_probe): Call validize_mem on memory location
12831         before passing it to gen_probe_stack.  Create address operand and
12832         legitimize it for the probe_stack_address case.
12834 2018-04-09  Jan Hubicka  <jh@suse.cz>
12836         PR lto/85078
12837         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12838         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12839         * tree.c (free_lang_data_in_type): Fix handling of binfos;
12840         walk basetypes.
12841         (free_lang_data): Rebuild type inheritance graph.
12843 2018-04-09  Martin Sebor  <msebor@redhat.com>
12845         * invoke.texi (-finline-small-functions): Mention other optimization
12846         options.
12847         (-findirect-inlining, -fpartial-inlining): Same.
12848         (-finline-functions-called-once): Same.
12849         (-freorder-blocks-and-partition): Same.
12851 2018-04-09  Jan Hubicka  <jh@suse.cz>
12853         PR rtl/84058
12854         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12855         jumps; choose last target that matches the criteria (i.e.
12856         no partition changes for non-crossing jumps).
12857         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12858         support for redirecting crossing jumps to non-crossing.
12860 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
12862         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12863         also for naked functions.
12865 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
12867         * config/arc/arc.md (add_shift): New pattern.
12868         (add_shift2): Likewise.
12869         (sub_shift): Likewise.
12870         (sub_shift_cmp0_noout): Likewise.
12871         (compare_si_ashiftsi): Likewise.
12872         (xbfu_cmp0_noout): New combine pattern.
12873         (xbfu_cmp0"): Likewise.
12874         (movsi_set_cc_insn): Place the predicable variant first.
12875         (commutative_binary_cmp0_noout): Remove clobber.
12876         (commutative_binary_cmp0): New pattern.
12877         (noncommutative_binary_cmp0): Likewise.
12878         (noncommutative_binary_cmp0_noout): Likewise.
12879         (noncommutative_binary_comparison_result_used): Removed.
12880         (rsub_cmp0): New pattern.
12881         (rsub_cmp0_noout): Likewise.
12882         (extzvsi): Changed, keep only meaningful variants.
12883         (SQH, SEZ): New iterators.
12884         (SQH_postfix): New mode attribute.
12885         (SEZ_prefix): New code attribute.
12886         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12887         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
12888         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
12889         of numerical value.
12890         (noncommutative_operator): Check the availability of barrel
12891         shifter option.
12893 2018-04-09  Richard Biener  <rguenther@suse.de>
12895         PR tree-optimization/85284
12896         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
12897         Only use the niter constraining form of simple_iv when the exit
12898         is always executed.
12900 2018-04-09  Tom de Vries  <tom@codesourcery.com>
12902         PR target/84041
12903         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12904         (define_expand "*memory_barrier"): New define_expand.
12905         (define_insn "memory_barrier"): New insn.
12907 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
12909         PR rtl-optimization/80463
12910         PR rtl-optimization/83972
12911         PR rtl-optimization/83480
12913         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12914         correct producer for the insn.
12915         (tidy_control_flow): Fixup seqnos in case of debug insns.
12917 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
12919         PR rtl-optimization/83913
12921         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
12922         different sched-times when merging exprs.
12924 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
12926         PR rtl-optimization/83962
12928         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
12929         tidy_fallthru_edge and tidy_control_flow.
12931 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
12933         PR rtl-optimization/83530
12935         * sel-sched.c (force_next_insn): New global variable.
12936         (remove_insn_for_debug): When force_next_insn is true, also leave only
12937         next insn in the ready list.
12938         (sel_sched_region): When the region wasn't scheduled, make another pass
12939         over it with force_next_insn set to 1.
12941 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
12943         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
12944         into tm_file.
12945         * config/nds32/constants.md (unspec_volatile_element): Add enum values
12946         for interrupt control.
12947         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12948         functions for interrupt control.
12949         * config/nds32/nds32-intrinsic.md: Likewise.
12950         * config/nds32/nds32_intrinsic.h: Likewise.
12951         * config/nds32/nds32.h (nds32_builtins): Likewise.
12953 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
12955         * config/nds32/nds32.c (nds32_init_machine_status,
12956         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
12957         strict_aligned_p field.
12958         (nds32_expand_to_rtl_hook): New function.
12959         (TARGET_EXPAND_TO_RTL_HOOK): Define.
12960         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
12962 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
12963             Chung-Ju Wu  <jasonwucj@gmail.com>
12965         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
12966         * config/nds32/nds32-n7.md: New file.
12967         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
12968         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
12969         pipeline.
12970         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
12971         * config/nds32/nds32.md (pipeline_model): Add n7.
12972         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
12973         * config/nds32/pipelines.md: Include n7 settings.
12975 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
12976             Chung-Ju Wu  <jasonwucj@gmail.com>
12978         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
12979         * config/nds32/nds32-e8.md: New file.
12980         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
12981         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
12982         pipeline.
12983         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
12984         * config/nds32/nds32.md (pipeline_model): Add e8.
12985         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
12986         * config/nds32/pipelines.md: Include e8 settings.
12988 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
12989             Chung-Ju Wu  <jasonwucj@gmail.com>
12991         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
12992         * config/nds32/nds32-n8.md: New file.
12993         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
12994         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
12995         pipeline.
12996         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
12997         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
12998         * config/nds32/nds32.md (pipeline_model): Add n8.
12999         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13000         * config/nds32/pipelines.md: Include n8 settings.
13002 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
13003             Chung-Ju Wu  <jasonwucj@gmail.com>
13005         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13006         * config/nds32/nds32-n9-2r1w.md: New file.
13007         * config/nds32/nds32-n9-3r2w.md: New file.
13008         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13009         nds32_register_ports): New or modify for cpu n9.
13010         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13011         pipeline.
13012         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13013         * config/nds32/nds32-utils.c: New file.
13014         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13015         TARGET_MUL_SLOW): Define.
13016         * config/nds32/nds32.md (pipeline_model): New attribute.
13017         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13018         New options that support cpu n9.
13019         * config/nds32/pipelines.md: Include n9 settings.
13020         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13022 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
13024         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13025         information if necessary.
13026         (output_cond_branch_compare_zero): Likewise.
13027         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13028         (nds32_target_alignment): Refine for alignment.
13029         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13030         (FUNCTION_BOUNDARY): Modify.
13031         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13032         align case.
13033         * config/nds32/nds32.opt (malways-align, malign-functions): New.
13035 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
13037         * config/nds32/constants.md (unspec_volatile_element): Add values for
13038         TLB operation and data prefetch.
13039         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13040         functions for TLB operation and data prefetch.
13041         * config/nds32/nds32-intrinsic.md: Likewise.
13042         * config/nds32/nds32_intrinsic.h: Likewise.
13043         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13044         (nds32_print_operand): Likewise.
13045         * config/nds32/nds32.h (nds32_builtins): Likewise.
13047 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
13048         Andrew Pinski <pinsika@gcc.gnu.org>
13050         PR middle-end/82976
13051         * match.pd: Use constant_boolean_node of correct type instead of
13052         boolean_true_node or boolean_false_node for simplifying
13053         pointer comparisons to zero.
13055 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
13057         PR tree-optimization/80021
13058         * tree.c (verify_type_variant): Make error call in verify_variant_match
13059         translatable and remove final full stop.
13061 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
13063         * config/nds32/constants.md (unspec_volatile_element): Add
13064         UNSPEC_VOLATILE_EH_RETURN.
13065         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13066         nds32_output_stack_pop): Support dwarf exception handling process.
13067         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13068         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13069         exception handling process.
13070         (nds32_compute_stack_frame): Likewise.
13071         (nds32_return_addr_rtx): Likewise.
13072         (nds32_initial_elimination_offset): Likewise.
13073         (nds32_expand_prologue): Likewise.
13074         (nds32_expand_epilogue): Likewise.
13075         (nds32_dynamic_chain_address): New function.
13076         * config/nds32/nds32.h (machine_function): Add fields for dwarf
13077         exception handling.
13078         (DYNAMIC_CHAIN_ADDRESS): Define.
13079         (EH_RETURN_DATA_REGNO): Define.
13080         (EH_RETURN_STACKADJ_RTX): Define.
13081         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13082         patterns for dwarf exception handling.
13084 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
13086         * config/nds32/nds32.h: Clean up obsolete macros.
13088 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
13090         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13091         Add enum values for particular instructions.
13092         * config/nds32/nds32-intrinsic.c: Implementation of expanding
13093         particular intrinsic functions.
13094         * config/nds32/nds32-intrinsic.md: Likewise.
13095         * config/nds32/nds32_intrinsic.h: Likewise.
13096         * config/nds32/nds32.h (nds32_builtins): Likewise.
13097         * config/nds32/nds32.md (type): Add pbsad and pbsada.
13098         (btst, ave): New patterns for particular instructions.
13100 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
13102         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13103         Add enum values for atomic load/store and memory sync.
13104         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13105         and memory sync.
13106         * config/nds32/nds32-intrinsic.md: Likewise.
13107         * config/nds32/nds32_intrinsic.h: Likewise.
13108         * config/nds32/nds32.h (nds32_builtins): Likewise.
13110 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
13112         PR tree-optimization/85257
13113         * fold-const.c (native_encode_vector): If not all elts could fit
13114         and off is -1, return 0 rather than offset.
13115         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13116         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13117         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
13118         adjust buffer in native_interpret_expr call.
13120 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
13122         * config/nds32/constants.md (unspec_volatile_element): Add cache
13123         control enum values.
13124         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13125         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13126         * config/nds32/nds32.c (nds32_cctl_names): New.
13127         (nds32_print_operand): Handle cache control register names.
13128         * config/nds32/nds32.h (nds32_builtins): New enum values.
13129         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13130         macros.
13131         * config/nds32/nds32.md (type): Add mmu.
13132         * config/nds32/pipelines.md (simple_insn): Add mmu.
13134 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
13136         * config/nds32/nds32.md (type): Remove call.
13137         * config/nds32/pipelines.md (simple_insn): Likewise.
13139 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
13141         * config/nds32/constants.md (unspec_volatile_element): Add
13142         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13143         UNSPEC_VOLATILE_FMFCFG.
13144         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13145         description for fmfcfg and fmfcsr.
13146         (bdesc_1arg): Add fmtcsr.
13147         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13148         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13149         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13150         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13151         unspec_fmfcfg): New patterns.
13152         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13153         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13154         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13155         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13156         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13157         __nds32__fmfcfg): Define.
13159 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
13161         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13162         intrinsic register names.
13163         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13164         intrinsic register enum values and macros.
13166 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
13168         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13169         for load/store addressing form.
13170         (nds32_print_operand_address): Likewise.
13172 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
13174         PR target/85196
13175         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13176         based on LABEL_REF.  Remove useless assertion.
13177         (pic_address_needs_scratch): Fix formatting.
13178         (sparc_legitimize_pic_address): Minor tweaks.
13179         (sparc_delegitimize_address): Adjust assertion accordingly.
13180         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13181         into symbolic_operand.
13182         (movsi_high_pic_label_ref): Likewise.
13183         (movsi_lo_sum_pic_label_ref): Likewise.
13184         (movdi_pic_label_ref): Likewise.
13185         (movdi_high_pic_label_ref): Likewise.
13186         (movdi_lo_sum_pic_label_ref): Likewise.
13188 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
13190         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13191         custom LIB_SPEC setup.
13193 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
13194             Kito Cheng  <kito.cheng@gmail.com>
13196         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13197         * config/riscv/freebsd.h: New.
13199 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
13201         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13202         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13203         file.
13205 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
13206             Kito Cheng  <kito.cheng@gmail.com>
13208         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13209         nds32_output_call, nds32_symbol_binds_local_p): New functions.
13210         * config/nds32/nds32-protos.h (nds32_output_call,
13211         nds32_output_return): Declare.
13212         * config/nds32/nds32.md: Refine all the call and return patterns.
13214 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
13216         PR debug/85252
13217         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13218         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13220         PR rtl-optimization/84872
13221         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13222         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13223         EDGE_CROSSING edge.
13225 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
13227         * expr.c (copy_blkmode_to_reg): Revert 254862.
13228         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13230 2018-04-06  Richard Biener  <rguenther@suse.de>
13232         PR middle-end/85244
13233         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13234         after seeing a component reference with an adjacent field.  Treat
13235         refs to arrays at struct end of external decls similar to
13236         refs to unconstrained commons.
13238 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
13240         PR sanitizer/85213
13241         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13242         look through SAVE_EXPRs with non-side-effects argument.  Adjust
13243         recursive calls.
13244         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13245         save_p here.
13247 2018-04-06  Richard Biener  <rguenther@suse.de>
13249         PR middle-end/85180
13250         * alias.c (find_base_term): New wrapper around find_base_term
13251         unwinding CSELIB_VAL_PTR changes.
13252         (find_base_term): Do not restore CSELIB_VAL_PTR during the
13253         recursion.
13255 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13257         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13258         instructions.
13259         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13260         constant definitions.
13261         ("nop"): lr 0,0 -> nopr r0
13262         ("nop_lr0", "nop_lr1"): New insn definitions.
13264 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
13266         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13267         NDS32_V3PUSH_AVAILABLE_P macro.
13269 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
13270             Chung-Ju Wu  <jasonwucj@gmail.com>
13272         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13273         (nds32*-*-*): Add float and fpu_config into supported_defaults.
13274         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13275         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13276         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13277         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13278         * config/nds32/constraints.md: New constraints and checking for hard
13279         float configuration.
13280         * config/nds32/iterators.md: New mode iterator and attribute for hard
13281         float configuration.
13282         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13283         patterns.
13284         * config/nds32/nds32-fpu.md: New file.
13285         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13286         deal with hard float code generation.
13287         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13288         ARCH_V3S.
13289         (abi_type, float_reg_number): New enum type.
13290         * config/nds32/nds32-predicates.c: New predicates for hard float.
13291         * config/nds32/nds32-protos.h: Declare functions for hard float.
13292         * config/nds32/nds32.c: Implementation for hard float configuration.
13293         * config/nds32/nds32.h: Definitions for hard float configuration.
13294         * config/nds32/nds32.md: Include hard float machine description and
13295         modify patterns for hard float configuration.
13296         * config/nds32/nds32.opt: New options for hard float configuration.
13297         * config/nds32/predicates.md: New predicates for hard float
13298         configuration.
13300 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
13302         * common/config/nds32/nds32-common.c
13303         (nds32_option_optimization_table): Enable -mreleax-hint by default.
13305 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
13307         PR middle-end/85195
13308         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13309         CONSTRUCTOR_ELT (ctor, ...)->value.
13311 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
13313         PR target/85193
13314         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13316 2018-04-05  Tom de Vries  <tom@codesourcery.com>
13318         PR target/85204
13319         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13320         cond jump.
13322 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
13323             Kito Cheng  <kito.cheng@gmail.com>
13325         * config/nds32/constraints.md (U33): Fine-tune checking condition.
13326         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13327         * config/nds32/nds32.h (nds32_16bit_address_type): Add
13328         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13330 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
13331             Kito Cheng  <kito.cheng@gmail.com>
13333         * config/nds32/constraints.md (Ufe): New memory constraint.
13334         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13335         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13336         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13337         operands.
13338         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13339         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13341 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13343         * config/nds32/nds32.md: Use optimize_size in the condition for
13344         alu-shift instructions.
13346 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13348         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13350 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13352         * config/nds32/nds32.md (negsi2): Refine pattern.
13354 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
13355             Chung-Ju Wu  <jasonwucj@gmail.com>
13357         * config/nds32/iterators.md (shift_rotate): New code iterator.
13358         (shift): New code attribute.
13359         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13360         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13361         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13362         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13363         bit-wise operations.
13364         (andsi3, *andsi3): Ditto.
13365         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13366         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13367         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13368         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13369         nds32_ior_operand, nds32_xor_operand): New predicates.
13371 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13373         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13374         (addsi3, subsi3): ... this.
13376 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13378         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13380 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13382         * config/nds32/nds32.md: Adjust indention.
13384 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
13386         * config/nds32/nds32.md (feature): New attribute.
13388 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
13390         * config/nds32/nds32.md (subtype): New attribute.
13392 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
13394         PR target/85203
13395         * config/arm/arm-builtins.c (arm_expand_builtin): Change
13396         expansion to perform a bitwise AND of the argument followed by a
13397         boolean negation of the result.
13399 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
13401         PR rtl-optimization/84878
13402         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13403         the basic block.  Assert the use reference is not artificial and that
13404         it has an associated insn.
13406 2018-04-04  Michael Matz  <matz@suse.de>
13408         * builtins.c (compute_objsize): Pass correct operand
13409         to array_at_struct_end_p.
13411 2018-04-04  Richard Biener  <rguenther@suse.de>
13413         PR lto/85176
13414         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13415         from contexts for DINFO_LEVEL_TERSE and below.
13417 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
13419         * config/nds32/nds32-doubleword.md (move_<mode>): Require
13420         resiter_operand condition.
13421         * config/nds32/nds32.md (*move<mode>): Ditto.
13423 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
13424             Monk Chiang  <sh.chiang04@gmail.com>
13426         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13428 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
13430         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13432 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
13433             Kito Cheng  <kito.cheng@gmail.com>
13435         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13436         nds32_cond_code_str, output_cond_branch,
13437         output_cond_branch_compare_zero, nds32_expand_cbranch,
13438         nds32_expand_cstore, nds32_expand_movcc,
13439         nds32_output_cbranchsi4_equality_zero,
13440         nds32_output_cbranchsi4_equality_reg,
13441         nds32_output_cbranchsi4_equality_reg_or_const_int,
13442         nds32_output_cbranchsi4_greater_less_zero: New functions.
13443         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13444         nds32_expand_cstore, nds32_expand_movcc,
13445         nds32_output_cbranchsi4_equality_zero,
13446         nds32_output_cbranchsi4_equality_reg,
13447         nds32_output_cbranchsi4_equality_reg_or_const_int,
13448         nds32_output_cbranchsi4_greater_less_zero): Declare.
13449         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13450         nds32_rimm11s_operand): New predicates.
13451         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13452         * config/nds32/nds32.md: Rewrite all the branch and conditional move
13453         patterns.
13455 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
13457         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13458         * config/nds32/nds32.md: Ditto.
13459         * config/nds32/pipelines.md: Ditto.
13461 2018-04-04  Richard Biener  <rguenther@suse.de>
13463         PR tree-optimization/85168
13464         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13465         propagating abnormals.
13467 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
13469         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13471 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
13472             Kito Cheng  <kito.cheng@gmail.com>
13474         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13475         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13476         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13477         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13478         * config/nds32/nds32.md (sibcall_internal): New.
13479         (sibcall_register): Remove.
13480         (sibcall_immediate): Remove.
13481         (sibcall_value_internal): New.
13482         (sibcall_value_register): Remove.
13483         (sibcall_value_immediate): Remove.
13484         * config/nds32/predicates.md (nds32_general_register_operand): New.
13485         (nds32_call_address_operand): New.
13487 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
13489         PR rtl-optimization/85167
13490         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13491         bb_defs if *split_p, instead preinitialize it to NULL.
13493         PR tree-optimization/85156
13494         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13495         evaluating the argument multiple times.
13497 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
13499         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13500         than vector.
13501         (_mm_cvtpd_ps): Likewise.
13502         (_mm_cvttpd_epi32): Likewise.
13503         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13504         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13505         vector, pixel, and bool following altivec.h include.
13507 2018-04-03  Martin Sebor  <msebor@redhat.com>
13509         * doc/extend.texi (Common Function Attributes): Clarify.
13510         (const attribute): Likewise.
13511         (pure attribute): Likewise.
13513 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
13515         PR target/85169
13516         * config/i386/i386.c (ix86_expand_vector_set): Use
13517         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
13519 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
13521         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13522         instructions when changing rounding bits to preserve precision bits
13523         in the x87 control word.
13525 2018-04-03  Martin Liska  <mliska@suse.cz>
13527         PR tree-optimization/82491
13528         * rtl.h (strip_offset_and_add): Replace += suboffset with
13529         poly_uint64 () + suboffset.
13531 2018-03-29  Martin Liska  <mliska@suse.cz>
13532             Martin Jambor  <mjambor@suse.cz>
13534         PR ipa/84947
13535         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13536         param_type is not an integral or pointer type.
13538 2018-04-03  Richard Biener  <rguenther@suse.de>
13540         * sese.h (recompute_all_dominators): Remove.
13542 2018-04-02  Martin Sebor  <msebor@redhat.com>
13544         * doc/invoke.texi (-Wrestrict): Fix typos.
13546 2018-04-02  Jim Wilson  <jimw@sifive.com>
13548         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13549         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13550         (<optab>di3, <optab>si3_extend): Likewise.
13551         (<optab>si3_mask, <optab>si3_mask_1): New.
13552         (<optab>di3_mask, <optab>di3_mask_1): New.
13553         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13554         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13555         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13557 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
13559         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13560         example.
13562 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13564         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13565         (nds32_canonicalize_comparison): New function.
13567 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
13568             Kito Cheng  <kito.cheng@gmail.com>
13569             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
13571         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13572         * config/nds32/constants.md (unspec_volatile_element): Add
13573         UNSPEC_VOLATILE_RELAX_GROUP.
13574         * config/nds32/nds32-relax-opt.c: New file.
13575         * config/nds32/nds32-predicates.c
13576         (nds32_symbol_load_store_p): New function.
13577         * config/nds32/nds32-protos.h
13578         (nds32_symbol_load_store_p): Declare function.
13579         (make_pass_nds32_relax_opt): Declare new rtl pass function.
13580         * config/nds32/nds32.c
13581         (nds32_register_pass): New function to register pass.
13582         (nds32_register_passes): New function to register passes.
13583         * config/nds32/nds32.md (relax_group): New pattern.
13584         * config/nds32/nds32.opt (mrelax-hint): New option.
13585         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13587 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
13589         * config/nds32/t-nds32: Modify files dependency.
13591 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
13593         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13594         (PROFILE_HOOK): Define its implementation.
13596 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
13598         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13599         type and 32-bit size.
13601 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
13603         PR middle-end/85090
13604         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13605         (V_128_256): New mode iterator.
13606         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13607         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13608         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13609         of V.
13610         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13611         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13613 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
13615         PR target/83315
13616         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13617         NaN inputs correctly.
13619 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
13621         PR target/80546
13622         * config/rs6000/vsx.md (??r): New mode attribute.
13623         (*vsx_mov<mode>_64bit): Use it.
13624         (*vsx_mov<mode>_32bit): Likewise.
13626 2018-03-30  Martin Sebor  <msebor@redhat.com>
13628         PR tree-optimization/84818
13629         * builtins.c (check_access): Use warning_n.
13631 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13633         PR target/83822
13634         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13635         condition.
13636         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13637         condition.
13639 2018-03-30  Julia Koval  <julia.koval@intel.com>
13641         PR target/84413
13642         * x86-tune.def (movx, partial_reg_dependency): Enable for
13643         m_SKYLAKE_AVX512.
13645 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
13647         PR inline-asm/84985
13648         * lra-constraints.c (process_alt_operands): Move setting
13649         this_alternative_matches below.
13651 2018-03-29  Martin Liska  <mliska@suse.cz>
13653         PR lto/84995.
13654         * doc/invoke.texi: Document how LTO works with debug info.
13655         Describe auto-load support of binutils.  Mention 'x86-64'
13656         as valid option value of -march option.
13658 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
13660         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13662         PR c/85094
13663         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13664         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13665         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13666         checking.
13668 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
13670         PR target/84912
13671         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13672         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13673         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13674         for RS6000_BTM_POWERPC64.
13675         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13676         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13677         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13678         definition.
13679         (DIVDE): Use it.
13680         (DIVDEU): Likewise.
13682 2018-03-28  Carl Love  <cel@us.ibm.com>
13684         Revert
13685         2017-09-27  Carl Love  <cel@us.ibm.com>
13687         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13688         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13689         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13690         fctiw instruction.
13692 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13694         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13695         instead of __vector bool.
13696         (_mm_max_pu8): Likewise.
13697         (_mm_min_pi16): Likewise.
13699 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
13701         PR target/84912
13702         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13703         (DIVWEUO): Likewise.
13704         (DIVDEO): Likewise.
13705         (DIVDEUO): Likewise.
13706         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13707         DIVWEUO and DIVDEUO.
13708         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13709         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13710         (div_extend): Likewise.
13711         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13712         builtin function.
13713         (__builtin_divweuo): Likewise.
13714         (__builtin_divdeo): Likewise.
13715         (__builtin_divdeuo): Likewise.
13717 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
13719         PR target/85095
13720         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13721         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13723         PR tree-optimization/82004
13724         * gimple-match-head.c (optimize_pow_to_exp): New function.
13725         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13726         Don't fold to exp if optimize_pow_to_exp is false.
13728 2018-03-28  Martin Liska  <mliska@suse.cz>
13730         PR other/84819
13731         * calls.c (initialize_argument_information): Fix trailing space.
13732         * common.opt: Fix typo and provide better explanation for
13733         -fsanitize-coverage option.
13734         * config/i386/i386.opt: Fix typo.
13736 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
13737             Martin Liska  <mliska@suse.cz>
13739         PR sanitizer/85081
13740         * gimplify.c (asan_poison_variable): Don't do the check for
13741         gimplify_omp_ctxp here.
13742         (gimplify_decl_expr): Do it here.
13743         (gimplify_target_expr): Likewise.
13745 2018-03-28  Martin Liska  <mliska@suse.cz>
13747         PR target/84988
13748         * config/i386/i386.c (ix86_function_arg_advance): Do not call
13749         chkp_type_bounds_count if MPX is not enabled.
13751 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
13753         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13755 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
13757         PR target/84914
13758         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13759         function to create the function decl for complex long double
13760         multiply and divide for -mabi=ieeelongdouble.
13761         (init_float128_ieee): Call it.
13763 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
13765         PR target/85044
13766         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13767         -fcf-protection=branch -mibt.
13768         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13770 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13772         PR target/81863
13773         * config/arm/arm.c (arm_valid_symbolic_address): Handle
13774         arm_word_relocations.
13776 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
13778         PR target/85056
13779         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13780         extern array declarations.
13782 2018-03-27  Richard Biener  <rguenther@suse.de>
13784         PR middle-end/84067
13785         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13786         explicit single_use checks.
13788 2018-03-27  Richard Biener  <rguenther@suse.de>
13790         PR tree-optimization/85082
13791         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13792         Valueize the VUSE.
13794 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13796         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13797         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13798         Turn on fasynchronous-unwind-tables and funwind-tables.
13800 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
13802         PR target/85073
13803         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13804         (*bmi_blsr_<mode>_ccz): Ditto.
13806 2018-03-26  Tom de Vries  <tom@codesourcery.com>
13808         PR tree-optimization/85063
13809         * omp-general.c (offloading_function_p): New function.  Factor out
13810         of ...
13811         * omp-offload.c (pass_omp_target_link::gate): ... here.
13812         * omp-general.h (offloading_function_p): Declare.
13813         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13814         with attribute omp declare target for offloading functions.
13816 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
13818         PR tree-optimization/84005
13819         * tree-data-ref.h (get_base_for_alignment): Declare.
13820         * tree-data-ref.c (get_base_for_alignment_1): New function.
13821         (get_base_for_alignment): Likewise.
13822         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13823         get_base_for_alignment to find a suitable base object, instead
13824         of always using drb->base_address.
13826 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
13828         PR inline-asm/85022
13829         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13830         known size by default.
13832 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
13834         PR inline-asm/85030
13835         * lra-constraints.c (process_alt_operands): Don't match BLKmode
13836         and non BLKmode operands.
13838 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13840         PR target/85026
13841         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13842         Clean up attributes.
13844 2018-03-23  Richard Biener  <rguenther@suse.de>
13846         PR debug/85020
13847         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13848         we are going to emit early debug for LTO.
13850 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
13852         PR inline-asm/85034
13853         * function.c (match_asm_constraints_1): Don't optimize if input
13854         doesn't satisfy general_operand predicate for output's mode.
13856         PR inline-asm/85022
13857         * alias.c (write_dependence_p): Don't require for x_canonicalized
13858         non-VOIDmode if x has VOIDmode.
13860         PR sanitizer/85029
13861         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13862         just don't try to optimize it rather than assert it never happens.
13864 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13866         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13867         macro expansions for definition of ST_INTERNAL_<mode> and
13868         LD_INTERNAL_<mode> builtins.
13869         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13870         Remove prototype.
13871         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13872         function.
13873         (altivec_expand_st_builtin): Likewise.
13874         (altivec_expand_builtin): Remove calls to deleted functions.
13875         (rs6000_address_for_altivec): Delete this function.
13876         * config/rs6000/vector.md: Remove expands for
13877         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13879 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
13881         PR target/84826
13882         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13883         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13884         re-computing once computed.
13885         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13886         (arm_init_machine_status): Initialize
13887         machine->static_chain_stack_bytes.
13889 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13891         PR target/84760
13892         * doc/extend.texi: Add four new prototypes for vec_ld.
13893         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
13894         definitions for more logical presentation.
13895         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
13896         entries for V1TI variants of __builtin_altivec_ld builtin.
13897         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
13898         handling of V1TI variant of LVX icode pattern.
13899         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
13900         (rs6000_gimple_fold_builtin): Likewise.
13901         (altivec_init_builtins): Add code to define
13902         __builtin_altivec_lvx_v1ti function.
13904 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
13906         PR inline-asm/84941
13907         * function.c (match_asm_constraints_1): Don't do the optimization
13908         if input isn't a REG, SUBREG, MEM or constant.
13910 2018-03-22  Tom de Vries  <tom@codesourcery.com>
13912         PR tree-optimization/84956
13913         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13914         bb_has_abnormal_pred.
13916 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
13918         PR sanitizer/85018
13919         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13920         DECL_INITIAL (decl) to decl at the end.
13921         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
13922         adjust the comment.
13924 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
13926         * doc/extend.texi (__builtin_tgmath): Document when complex
13927         integer types are treated as _Complex _Float64.
13929 2018-03-21  Tom de Vries  <tom@codesourcery.com>
13931         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
13933 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
13935         PR tree-optimization/84960
13936         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
13937         if it is ENTRY block, move them into single succ of ENTRY in that case.
13939 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
13941         PR tree-optimization/84811
13942         * poly-int.h (poly_span_traits): Remove the T3 parameter and
13943         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
13944         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13945         (known_subrange_p): Update accordingly.  Cast each value involved
13946         in the size comparison, rather than casting the result of the
13947         subtraction.
13949 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
13951         PR tree-optimization/84982
13952         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
13953         by flipping the least significant bit rather than all bits from
13954         bitpos to bitpos + bitsize - 1.
13956 2018-03-21  Nathan Sidwell  <nathan@acm.org>
13958         * doc/extend.texi (Deprecated Features): Remove mention of
13959         long-deleted deprecations.
13961 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13963         PR jit/84288
13964         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
13965         * configure: Regenerate.
13967 2018-03-21  Tom de Vries  <tom@codesourcery.com>
13969         PR tree-optimization/83126
13970         * tree-parloops.c (num_phis): New function.
13971         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
13973 2018-03-21  Nathan Sidwell  <nathan@acm.org>
13975         * doc/extend.texi (Deprecated Features): Update deprecated flags,
13976         mention anon-struct/union members and trailing attributes.
13978 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
13980         PR tree-optimization/84969
13981         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
13982         builtin memset partitions if they set different rhs values.
13984 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
13986         PR rtl-optimization/84989
13987         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
13988         VEC_DUPLICATE with scalar result mode.
13990 2018-03-21  Martin Liska  <mliska@suse.cz>
13992         PR ipa/84963
13993         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
13994         not intended return statement.
13996 2018-03-21  Martin Liska  <mliska@suse.cz>
13998         PR target/84988
13999         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14000         (chkp_find_bound_slots_1): Limit number of iterations.
14002 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
14004         PR target/84838
14005         * Minor grammar fixes for x86 options.
14007 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
14009         PR debug/84875
14010         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14011         holding REG_CFA_RESTORE notes, instead turn them into a USE.
14013 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
14015         PR target/83789
14016         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14017         (altivec_lvx_<mode>_1op): Likewise.
14018         (altivec_stvx_<mode>_2op): Likewise.
14019         (altivec_stvx_<mode>_1op): Likewise.
14020         (altivec_lvx_<VM2:mode>): New define_expand.
14021         (altivec_stvx_<VM2:mode>): Likewise.
14022         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14023         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14024         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14025         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14026         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14027         (rs6000_gen_lvx): Likewise.
14028         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14029         (altivec_expand_stv_builtin): Likewise.
14030         (altivec_expand_builtin): Likewise.
14031         * config/rs6000/vector.md: Likewise.
14033 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14035         PR target/82518
14036         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14037         BYTES_BIG_ENDIAN.
14039 2018-03-20  Richard Biener  <rguenther@suse.de>
14041         PR target/84986
14042         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14043         sign-conversions as zero, fall back to standard scalar_stmt
14044         cost for the rest.
14046 2018-03-20  Martin Liska  <mliska@suse.cz>
14048         PR ipa/84825
14049         * predict.c (rebuild_frequencies): Handle case when we have
14050         PROFILE_ABSENT, but flag_guess_branch_prob is false.
14052 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
14054         PR target/84990
14055         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14056         flag_section_anchors.
14057         * varasm.c (use_blocks_for_decl_p): Remove hack for
14058         dw2_force_const_mem.
14060         PR target/84845
14061         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14062         to ...
14063         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
14064         be created, use lowpart_subreg of operands[0] rather than operands[0]
14065         itself.
14066         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14067         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14068         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14069         and n constraint instead of aarch64_shift_imm_di and Usd.
14070         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14071         (*aarch64_<optab>_reg_minus<mode>3): ... this.
14073 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
14075         PR target/82989
14076         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14077         to favor GPR over NEON registers.
14078         (<shift>di3_neon): Likewise.
14080 2018-03-20  Tom de Vries  <tom@codesourcery.com>
14082         PR target/84952
14083         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14084         (nvptx_process_pars): Emit bar.sync asap and alap.
14086 2018-03-20  Tom de Vries  <tom@codesourcery.com>
14088         PR target/84954
14089         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14090         seen_label if seen_label is already set.
14092 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
14094         PR target/84945
14095         * config/i386/i386.c (fold_builtin_cpu): For features above 31
14096         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14097         Use 1U instead of 1.  Formatting fixes.
14099         PR c/84953
14100         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14101         instead of TREE_TYPE (s1) for the return value.
14103 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
14105         PR tree-optimization/84946
14106         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14107         bitsize + bitsize in poly_uint64 rather than poly_int64.
14109         PR sanitizer/78651
14110         * dwarf2asm.c: Include fold-const.c.
14111         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14112         of decl rather than decl itself.
14114         PR rtl-optimization/84643
14115         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14117 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
14119         PR sanitizer/78651
14120         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14121         calling assemble_variable.
14123 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
14125         PR target/81647
14126         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14127         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14129 2018-03-19  Jim Wilson  <jimw@sifive.com>
14131         PR bootstrap/84856
14132         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14133         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14134         (riscv_first_stack_step): Likewise.
14135         (riscv_option_override): Use STACK_BOUNDARY instead of
14136         MIN_STACK_BOUNDARY.
14137         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14138         MIN_STACK_BOUNDARY.
14139         (BIGGEST_ALIGNMENT): Set to 128.
14140         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14141         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14142         STACK_BOUNDARY.
14144 2018-03-19  Richard Biener  <rguenther@suse.de>
14146         PR tree-optimization/84933
14147         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14148         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14150 2018-03-19  Richard Biener  <rguenther@suse.de>
14152         PR tree-optimization/84859
14153         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14154         (cond_if_else_store_replacement): Perform sinking operation on
14155         single-store BBs regardless of MAX_STORES_TO_SINK setting.
14156         Generalize what a BB with a single eligible store is.
14158 2018-03-19  Richard Biener  <rguenther@suse.de>
14160         PR tree-optimization/84929
14161         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14162         chrec_is_positive against non-chrec arg.
14164 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
14166         PR target/84711
14167         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14169 2018-03-18  Martin Liska  <mliska@suse.cz>
14171         PR rtl-optimization/84635
14172         * regrename.c (build_def_use): Use matches_mode only when
14173         matches >= 0.
14175 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
14177         PR tree-optimization/84913
14178         * tree-vect-loop.c (vectorizable_reduction): Don't try to
14179         vectorize chains of COND_EXPRs.
14181 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
14183         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14185 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
14187         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14189 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
14191         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14193 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
14194             Kito Cheng  <kito.cheng@gmail.com>
14196         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14197         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14198         (nds32_adjust_reg_alloc_order): New function.
14199         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14201 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
14203         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14204         nds32_print_operand, nds32_print_operand_address): Use
14205         HOST_WIDE_INT_PRINT_DEC instead.
14207 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
14209         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14211 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
14213         PR target/84902
14214         * config/i386/i386.c (initial_ix86_tune_features,
14215         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14216         unsigned long long.
14217         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14218         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14219         rather than 1u << ix86_tune.  Formatting fix.
14220         (ix86_option_override_internal): Change ix86_arch_mask from
14221         unsigned int to unsigned HOST_WIDE_INT, initialize to
14222         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14223         (ix86_function_specific_restore): Likewise.
14225 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
14227         PR target/84899
14228         * postreload.c (reload_combine_recognize_pattern): Perform
14229         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14230         truncate_int_for_mode the result for the destination's mode.
14232         PR c/84909
14233         * hsa-gen.c (mem_type_for_type): Fix comment typo.
14234         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14235         Likewise.
14236         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14237         Likewise.
14239 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
14241         PR target/84876
14242         * lra-assigns.c (lra_split_hard_reg_for): Don't use
14243         regno_allocno_class_array and sorted_pseudos.
14244         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14245         insns where regno is used.
14247 2018-03-16  Martin Liska  <mliska@suse.cz>
14249         PR ipa/84833
14250         * multiple_target.c (create_dispatcher_calls): Redirect
14251         reference in the symbol table.
14253 2018-03-16  Martin Liska  <mliska@suse.cz>
14255         PR ipa/84722
14256         * multiple_target.c (create_dispatcher_calls): Redirect also
14257         an alias.
14259 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
14261         PR c++/79937
14262         PR c++/82410
14263         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14264         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14265         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14267 2018-03-16  Julia Koval  <julia.koval@intel.com>
14269         * doc/invoke.texi (Skylake Server): Add CLWB.
14270         Cannonlake): Remove CLWB.
14272 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
14274         PR tree-optimization/84841
14275         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14276         1 << 3.
14277         (FLOAT_ONE_CONST_TYPE): Define.
14278         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14279         (sort_by_operand_rank): Put entries with higher constant_type last
14280         rather than first to match comments.
14282 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
14284         * config/nios2/nios2.md (movsi_internal): Fix thinko in
14285         split predicate.
14287 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
14289         PR c++/79085
14290         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14291         check and use address of target always.
14293 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
14295         PR target/84574
14296         * config/i386/i386.c (indirect_thunk_needed): Update comments.
14297         (indirect_thunk_bnd_needed): Likewise.
14298         (indirect_thunks_used): Likewise.
14299         (indirect_thunks_bnd_used): Likewise.
14300         (indirect_return_needed): New.
14301         (indirect_return_bnd_needed): Likewise.
14302         (output_indirect_thunk_function): Add a bool argument for
14303         function return.
14304         (output_indirect_thunk_function): Don't generate alias for
14305         function return thunk.
14306         (ix86_code_end): Call output_indirect_thunk_function to generate
14307         function return thunks.
14308         (ix86_output_function_return): Set indirect_return_bnd_needed
14309         and indirect_return_needed instead of indirect_thunk_bnd_needed
14310         and indirect_thunk_needed.
14312 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
14314         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14315         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14316         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14318 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
14319             Paul Hua <paul.hua.gm@gmail.com>
14321         PR c/84852
14322         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14324 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
14326         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14327         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14328         resp. SFmode cases.
14330 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
14332         PR target/84711
14333         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14334         instead of GET_MODE_SIZE when comparing Units.
14336 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
14338         PR target/68256
14339         * varasm.c (hash_section): Return an unchangeble hash value
14340         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14341         Return !aarch64_can_use_per_function_literal_pools_p ().
14343 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
14345         PR target/84860
14346         * optabs.c (emit_conditional_move): Pass address of cmode's copy
14347         rather than address of cmode as last argument to prepare_cmp_insn.
14349 2018-03-15  Julia Koval  <julia.koval@intel.com>
14351         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14352         F_AVX512VNNI, F_AVX512BITALG): New.
14354 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
14356         PR target/83451
14357         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
14358         insn for floating-point loads and stores.
14360 2018-03-14  Carl Love  <cel@us.ibm.com>
14362         * config/rs6000/rs6000-c.c: Add macro definitions for
14363         ALTIVEC_BUILTIN_VEC_PERMXOR.
14364         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14365         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14366         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14367         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14368         UNSPEC_VPERMXOR.
14369         * config/doc/extend.texi: Add prototypes for vec_permxor.
14371 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
14373         PR c/84852
14374         * diagnostic-show-locus.c (class layout_point): Convert m_line
14375         from int to linenum_type.
14376         (line_span::comparator): Use linenum "compare" function when
14377         comparing line numbers.
14378         (test_line_span): New function.
14379         (layout_range::contains_point): Convert param "row" from int to
14380         linenum_type.
14381         (layout_range::intersects_line_p): Likewise.
14382         (layout::will_show_line_p): Likewise.
14383         (layout::print_source_line): Likewise.
14384         (layout::should_print_annotation_line_p): Likewise.
14385         (layout::print_annotation_line): Likewise.
14386         (layout::print_leading_fixits): Likewise.
14387         (layout::annotation_line_showed_range_p): Likewise.
14388         (struct line_corrections): Likewise for field m_row.
14389         (line_corrections::line_corrections): Likewise for param "row".
14390         (layout::print_trailing_fixits): Likewise.
14391         (layout::get_state_at_point): Likewise.
14392         (layout::get_x_bound_for_row): Likewise.
14393         (layout::print_line): Likewise.
14394         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14395         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14396         * input.c (selftest::test_linenum_comparisons): New function.
14397         (selftest::input_c_tests): Call it.
14398         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14399         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14400         * selftest.h (ASSERT_GT): New macro.
14401         (ASSERT_GT_AT): New macro.
14402         (ASSERT_LT): New macro.
14403         (ASSERT_LT_AT): New macro.
14405 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
14407         PR rtl-optimization/84780
14408         * combine.c (distribute_links): Don't make a link based on pc_rtx.
14410 2018-03-14  Martin Liska  <mliska@suse.cz>
14412         * tree.c (record_node_allocation_statistics): Use
14413         get_stats_node_kind.
14414         (get_stats_node_kind): New function extracted from
14415         record_node_allocation_statistics.
14416         (free_node): Use get_stats_node_kind.
14418 2018-03-14  Richard Biener  <rguenther@suse.de>
14420         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14421         that the value-set of ANTIC_IN doesn't grow.
14423         Revert
14424         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14425         member.
14426         (BB_VISITED_WITH_VISITED_SUCCS): New define.
14427         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14429 2018-03-14  Julia Koval  <julia.koval@intel.com>
14431         * config.gcc (icelake-client, icelake-server): New.
14432         (icelake): Remove.
14433         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14434         (initial_ix86_arch_features): Ditto.
14435         (PTA_SKYLAKE): Add SGX.
14436         (PTA_ICELAKE): Remove.
14437         (PTA_ICELAKE_CLIENT): New.
14438         (PTA_ICELAKE_SERVER): New.
14439         (ix86_option_override_internal): Split up icelake on icelake client and
14440         icelake server.
14441         (get_builtin_code_for_version): Ditto.
14442         (fold_builtin_cpu): Ditto.
14443         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14444         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14445         * config/i386/i386.h (processor_type): Ditto.
14446         * doc/invoke.texi: Ditto.
14448 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
14450         PR sanitizer/83392
14451         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14452         INTEGER_CST offset, add it together with bitpos / 8 and
14453         sign extend based on POINTER_SIZE.
14455         PR target/84844
14456         Revert
14457         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
14459         PR target/78090
14460         * config/i386/constraints.md (Yc): New register constraint.
14461         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14462         Use Yc constraint for alternative 2 of operand 0.  Remove
14463         preferred_for_speed attribute.
14465 2018-03-14  Richard Biener  <rguenther@suse.de>
14467         PR tree-optimization/84830
14468         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14469         with the old one to avoid oscillations.
14471 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
14473         PR target/83712
14474         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14475         pseudos.
14476         (assign_by_spills): Return a flag of reload assignment failure.
14477         Do not process the reload assignment failures.  Do not spill other
14478         reload pseudos if they has the same reg class.  Update n if
14479         necessary.
14480         (lra_assign): Add a return arg.  Set up from the result of
14481         assign_by_spills call.
14482         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14483         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
14484         usage_insns if it is not NULL.
14485         (spill_hard_reg_in_range): New function.
14486         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14487         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14488         function prototypes.
14489         (lra_assign): Change prototype.
14490         * lra.c (lra): Add code to deal with fails by splitting hard reg
14491         live ranges.
14493 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
14495         * config/riscv/riscv.opt (mrelax): New option.
14496         * config/riscv/riscv.c (riscv_file_start): Emit ".option
14497         "norelax" when riscv_mrelax is disabled.
14498         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14500 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14502         PR target/84743
14503         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14504         reassociation for int modes.
14506 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
14508         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14509         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14510         for big-endian.
14511         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14512         * config/aarch64/aarch64-sve.md
14513         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14514         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14515         (*extend<mode><Vwide>2): Rename to...
14516         (aarch64_sve_extend<mode><Vwide>2): ...this.
14517         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14518         renaming the old pattern to...
14519         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
14520         unsigned packs.
14521         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14522         define_expand, renaming the old pattern to...
14523         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14524         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14525         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14526         account when deciding which SVE instruction the optab should use.
14527         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14529 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
14531         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14532         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14533         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14534         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14535         (tlsdesc_small_<mode>): Turn a define_expand and use
14536         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
14537         (tlsdesc_small_advsimd_<mode>): ...this.
14538         (tlsdesc_small_sve_<mode>): New pattern.
14540 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
14542         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14543         (UNSPEC_UMUL_HIGHPART): New constants.
14544         (MUL_HIGHPART): New int iteraor.
14545         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14546         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14547         define_expand.
14548         (*<su>mul<mode>3_highpart): New define_insn.
14550 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
14552         PR lto/84805
14553         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14554         incomplete types.
14556 2018-03-13  Martin Liska  <mliska@suse.cz>
14558         PR ipa/84658.
14559         * (sem_item_optimizer::sem_item_optimizer): Initialize new
14560         vector.
14561         (sem_item_optimizer::~sem_item_optimizer): Release it.
14562         (sem_item_optimizer::merge_classes): Register variable aliases.
14563         (sem_item_optimizer::fixup_pt_set): New function.
14564         (sem_item_optimizer::fixup_points_to_sets): Likewise.
14565         * ipa-icf.h: Declare new variables and functions.
14567 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
14569         PR middle-end/84834
14570         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14571         integer_pow2p@2 and test integer_pow2p in condition.
14572         (A < 0 ? C : 0): Similarly for @1.
14574         PR middle-end/84831
14575         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14576         characters starting at p contain '\0' character, don't look beyond
14577         that.
14579         PR target/84827
14580         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14581         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14583         PR target/84828
14584         * reg-stack.c (change_stack): Change update_end var from int to
14585         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14586         also call set_block_for_insn on the newly added insns and rescan.
14588         PR target/84786
14589         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14590         on the last operand.
14592         PR c++/84704
14593         * tree.c (stabilize_reference_1): Return save_expr (e) for
14594         STATEMENT_LIST even if it doesn't have side-effects.
14596 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
14598         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14600 2018-03-12  Renlin Li  <renlin.li@arm.com>
14602         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14603         aarch64_output_scalar_simd_mov_immediate.
14605 2018-03-12  Martin Sebor  <msebor@redhat.com>
14607         PR tree-optimization/83456
14608         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14609         for perfectly overlapping calls to memcpy.
14610         (gimple_fold_builtin_memory_chk): Same.
14611         (gimple_fold_builtin_strcpy): Handle no-warning.
14612         (gimple_fold_builtin_stxcpy_chk): Same.
14613         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14615 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
14617         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14618         parameter.  Use it for SFmode.
14619         (rs6000_function_arg_advance_1): Adjust.
14620         (rs6000_function_arg): Adjust.
14621         (rs6000_gimplify_va_arg): Pass false for that new parameter.
14623 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
14625         PR rtl-optimization/84169
14626         PR rtl-optimization/84780
14627         * combine.c (can_combine_p): Check for a 2-insn combination whether
14628         the destination register is used between the two insns, too.
14630 2018-03-12  Richard Biener  <rguenther@suse.de>
14632         PR tree-optimization/84803
14633         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14634         for refs DR analysis didn't process.
14636 2018-03-12  Richard Biener  <rguenther@suse.de>
14638         PR tree-optimization/84777
14639         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14640         force-vectorize loops ignore whether we are optimizing for size.
14642 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
14644         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14645         (TARGET_MD_ASM_ADJUST): Define.
14647 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
14648             Kito Cheng  <kito.cheng@gmail.com>
14649             Chung-Ju Wu  <jasonwucj@gmail.com>
14651         * config/nds32/nds32.c (nds32_compute_stack_frame,
14652         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14653         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14654         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14655         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14656         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14657         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14658         * config/nds32/nds32.md (prologue, epilogue): Use macro
14659         NDS32_V3PUSH_AVAILABLE_P to do checking.
14661 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
14663         PR debug/58150
14664         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14665         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14666         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
14667         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
14668         addition of most attributes on !orig_type_die or the attribute not
14669         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14671 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
14672             Chung-Ju Wu  <jasonwucj@gmail.com>
14674         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14675         __NDS32_VH__ macro.
14676         * config/nds32/nds32.opt (mvh): New option.
14678 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
14679             Chung-Ju Wu  <jasonwucj@gmail.com>
14681         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14682         function.
14683         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14684         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14685         definition.
14687 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
14688             Chung-Ju Wu  <jasonwucj@gmail.com>
14690         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14691         function.
14692         * config/nds32/nds32-multiple.md (strlensi): New pattern.
14693         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14695 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
14696             Kito Cheng  <kito.cheng@gmail.com>
14697             Chung-Ju Wu  <jasonwucj@gmail.com>
14699         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14700         UNSPEC_FFMISM and UNSPEC_FLMISM.
14701         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14702         for ffb, ffmism and flmism.
14703         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14704         (unspec_ffmism): Ditto.
14705         (unspec_flmism): Ditto.
14706         (nds32_expand_builtin_impl): Check if string extension is available.
14707         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14708         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14710 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
14712         Reverting patch:
14713         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
14715         PR target/83712
14716         * lra-assigns.c (assign_by_spills): Return a flag of reload
14717         assignment failure.  Do not process the reload assignment
14718         failures.  Do not spill other reload pseudos if they has the same
14719         reg class.
14720         (lra_assign): Add a return arg.  Set up from the result of
14721         assign_by_spills call.
14722         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14723         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
14724         usage_insns if it is not NULL.
14725         (spill_hard_reg_in_range): New function.
14726         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14727         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14728         function prototypes.
14729         (lra_assign): Change prototype.
14730         * lra.c (lra): Add code to deal with fails by splitting hard reg
14731         live ranges.
14733 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
14735         PR target/84807
14736         * config/i386/i386.opt: Replace Enforcment with Enforcement.
14738 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
14740         PR debug/84620
14741         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14742         (dw_val_node): Add val_symbolic_view.
14743         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14744         (symview_upper_bound): New.
14745         (new_line_info_table): Initialize symviews_since_reset.
14746         (dwarf2out_source_line): Count symviews_since_reset and set
14747         symview_upper_bound.
14748         (dw_val_equal_p): Handle symview.
14749         (add_AT_symview): New.
14750         (print_dw_val): Handle symview.
14751         (attr_checksum, attr_checksum_ordered): Likewise.
14752         (same_dw_val_p, size_of_die): Likewise.
14753         (value_format, output_die): Likewise.
14754         (add_high_low_attributes): Use add_AT_symview for entry_view.
14755         (dwarf2out_finish): Reset symview_upper_bound, clear
14756         zero_view_p.
14758 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
14760         PR target/83969
14761         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14762         Add strict argument and use it.
14763         (rs6000_split_multireg_move): Update for new strict argument.
14764         (mem_operand_gpr): Disallow all non-offsettable addresses.
14765         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14767 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
14769         PR target/84772
14770         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14771         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14772         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14774         PR c++/84767
14775         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14776         decl, use remap_type if we want to use the type.
14778 2018-03-09  Martin Sebor  <msebor@redhat.com>
14780         PR tree-optimization/84526
14781         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14782         Remove dead code.
14783         (builtin_access::generic_overlap): Be prepared to handle non-array
14784         base objects.
14786 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
14788         PR rtl-optimization/84682
14789         * lra-constraints.c (process_address_1): Check is_address flag
14790         for address constraints.
14791         (process_alt_operands): Likewise.
14792         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14793         preprocess_constraints.
14794         * recog.h (preprocess_constraints): Add oploc parameter.
14795         Adjust callers.
14796         * recog.c (preprocess_constraints): Test address_operand for
14797         CT_ADDRESS constraints.
14799 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
14801         PR target/83712
14802         * lra-assigns.c (assign_by_spills): Return a flag of reload
14803         assignment failure.  Do not process the reload assignment
14804         failures.  Do not spill other reload pseudos if they has the same
14805         reg class.
14806         (lra_assign): Add a return arg.  Set up from the result of
14807         assign_by_spills call.
14808         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14809         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
14810         usage_insns if it is not NULL.
14811         (spill_hard_reg_in_range): New function.
14812         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14813         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14814         function prototypes.
14815         (lra_assign): Change prototype.
14816         * lra.c (lra): Add code to deal with fails by splitting hard reg
14817         live ranges.
14819 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14821         PR target/83193
14822         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14823         Accept complain bool parameter.  Only emit errors if it is true.
14824         (arm_parse_cpu_option_name): Likewise.
14825         (arm_target_thumb_only): Adjust callers of the above.
14826         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14827         prototype to take a default true bool parameter.
14828         (arm_parse_arch_option_name): Likewise.
14830 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
14831             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
14833         PR jit/64089
14834         PR jit/84288
14835         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14836         * configure: Regenerate.
14837         * configure.ac ("linker --version-script option"): New.
14838         ("linker soname option"): New.
14840 2018-03-09  Richard Biener  <rguenther@suse.de>
14842         PR tree-optimization/84775
14843         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14844         immediate uses of predicate stmts and mark them modified.
14846         Revert
14847         PR tree-optimization/84178
14848         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14849         to caller.
14850         (version_loop_for_if_conversion): Delay update_ssa call.
14851         (tree_if_conversion): Delay update_ssa until after predicate
14852         insertion.
14854 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
14856         PR target/84763
14857         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14858         when the function accesses prior frames.
14860 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
14862         PR debug/84456
14863         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14864         gen_llsym, otherwise call maybe_gen_llsym.
14866         PR inline-asm/84742
14867         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14868         has ',' character inside of it.
14870 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14872         PR target/84748
14873         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14874         as clobbering CC_REGNUM.
14876 2018-03-08  Richard Biener  <rguenther@suse.de>
14878         PR middle-end/84552
14879         * tree-scalar-evolution.c: Include tree-into-ssa.h.
14880         (follow_copies_to_constant): Do not follow SSA names registered
14881         for update.
14883 2018-03-08  Richard Biener  <rguenther@suse.de>
14885         PR tree-optimization/84178
14886         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14887         to caller.
14888         (version_loop_for_if_conversion): Delay update_ssa call.
14889         (tree_if_conversion): Delay update_ssa until after predicate
14890         insertion.
14892 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
14894         PR tree-optimization/84178
14895         * tree-if-conv.c (release_bb_predicate): Remove the
14896         the assertion that the stmts have NULL use_ops.
14897         Discard the statements, asserting that they haven't
14898         yet been added to a BB.
14900 2018-03-08  Richard Biener  <rguenther@suse.de>
14902         PR tree-optimization/84746
14903         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
14904         (phi_translate): Pass in destination ANTIC_OUT set.
14905         (phi_translate_1): Likewise.  For a simplified result lookup
14906         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
14907         (phi_translate_set): Adjust.
14908         (do_pre_regular_insertion): Likewise.
14909         (do_pre_partial_partial_insertion): Likewise.
14911 2018-03-08  Martin Liska  <mliska@suse.cz>
14913         PR gcov-profile/84735
14914         * doc/gcov.texi: Document usage of profile files.
14915         * gcov-io.h: Document changes in the format.
14917 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
14919         PR debug/84404
14920         PR debug/84408
14921         * dwarf2out.c (struct dw_line_info_table): Update comments for
14922         view == -1.
14923         (FORCE_RESET_NEXT_VIEW): New.
14924         (FORCE_RESETTING_VIEW_P): New.
14925         (RESETTING_VIEW_P): Check for -1 too.
14926         (ZERO_VIEW_P): Likewise.
14927         (new_line_info_table): Force-reset next view.
14928         (dwarf2out_begin_function): Likewise.
14929         (dwarf2out_source_line): Simplify zero_view_p initialization.
14930         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
14931         view directly.  Omit view when omitting .loc at line 0.
14933 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
14935         PR tree-optimization/84740
14936         * tree-switch-conversion.c (process_switch): Call build_constructors
14937         only if info.phi_count is non-zero.
14939         PR tree-optimization/84739
14940         * tree-tailcall.c (find_tail_calls): Check call arguments against
14941         DECL_ARGUMENTS (current_function_decl) rather than
14942         DECL_ARGUMENTS (func) when checking for tail recursion.
14944 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
14946         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
14947         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
14948         Volker Reichelt's entry and add entries for people that perform
14949         GCC fuzzy testing and report numerous bugs.
14951 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
14953         PR target/82411
14954         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
14955         readonly data in sdata, if that is disabled.
14956         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
14957         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
14958         -mreadonly-in-sdata option.
14960 2018-03-07  Martin Sebor  <msebor@redhat.com>
14962         PR tree-optimization/84468
14963         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
14964         basic block when looking for nul assignment.
14966 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
14968         PR target/84277
14969         * except.h (output_function_exception_table): Adjust prototype.
14970         * except.c (output_function_exception_table): Remove FNNAME parameter
14971         and add SECTION parameter.  Ouput one part of the table at a time.
14972         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
14973         the first part of the exception table and emit unwind directives.
14974         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
14975         (i386_pe_seh_cold_init): Likewise.
14976         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
14977         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
14978         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
14979         (ix86_output_call_insn): Emit a nop in one more case for SEH.
14980         * config/i386/winnt.c: Include except.h.
14981         (struct seh_frame_state): Add reg_offset, after_prologue and
14982         in_cold_section fields.
14983         (i386_pe_seh_end_prologue): Set seh->after_prologue.
14984         (i386_pe_seh_cold_init): New function.
14985         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
14986         to seh->in_cold_section.
14987         (seh_emit_push): Record the offset of the push.
14988         (seh_emit_save): Record the offet of the save.
14989         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
14990         Test seh->after_prologue to disregard the epilogue.
14991         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
14992         (i386_pe_end_cold_function): New function.
14994 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
14996         PR fortran/84565
14997         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
14998         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15000         PR c++/84704
15001         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15002         on tmp_var.
15003         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15004         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15006         PR middle-end/84723
15007         * multiple_target.c: Include tree-inline.h and intl.h.
15008         (expand_target_clones): Diagnose and fail if node->definition and
15009         !tree_versionable_function_p (node->decl).
15011 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
15013         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15014         sprint_ul.
15015         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15016         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15017         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15019 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
15021         PR target/84710
15022         * combine.c (try_combine): Use reg_or_subregno instead of handling
15023         just paradoxical SUBREGs and REGs.
15025 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
15027         * config/arc/arc.c (arc_finalize_pic): Remove function.
15028         (arc_must_save_register): We use single base PIC register, remove
15029         checks to save/restore the PIC register.
15030         (arc_expand_prologue): Likewise.
15031         * config/arc/arc-protos.h (arc_set_default_type_attributes):
15032         Remove.
15033         (arc_verify_short): Likewise.
15034         (arc_attr_type): Likewise.
15035         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15036         (walk_stores): Likewise.
15037         (arc_address_cost): Make it static.
15038         (arc_verify_short): Likewise.
15039         (branch_dest): Likewise.
15040         (arc_attr_type): Likewise.
15041         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15042         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15043         (arc_final_prescan_insn): Remove inserting the nops due to
15044         hardware hazards.  It is done in reorg step.
15045         (insn_length_variant_t): Remove.
15046         (insn_length_parameters_t): Likewise.
15047         (arc_insn_length_parameters): Likewise.
15048         (arc_get_insn_variants): Likewise.
15049         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15051 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
15053         PR inline-asm/84683
15054         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15055         assertion failure.
15057         PR tree-optimization/84687
15058         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15059         on new_node->decl.
15060         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15062 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
15064         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15065         Rename to ppc_speculation_barrier.
15066         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15067         __builtin_ppc_speculation_barrier.
15069 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
15071         PR target/84700
15072         * combine.c (combine_simplify_rtx): Don't try to simplify if
15073         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15074         are equal to x.
15076 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
15078         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15079         to 32 bytes when compiling for POWER9.
15081 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
15083         PR target/84564
15084         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15085         regparm >= 3 with no arg reg available also for calls with
15086         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
15088         PR target/84524
15089         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15090         orig,vex.
15091         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
15093 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
15095         PR target/84264
15096         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15098 2018-03-05  Richard Biener  <rguenther@suse.de>
15100         PR tree-optimization/84486
15101         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15102         When inserting a __builtin_assume_aligned call set the LHS
15103         SSA name alignment info accordingly.
15105 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
15107         PR tree-optimization/84114
15108         * config/aarch64/aarch64.c (aarch64_reassociation_width)
15109         Avoid reassociation of FLOAT_MODE addition.
15111 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
15113         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15114         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15115         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15116         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15117         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15118         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15119         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15120         and -mwbnoinvd.
15121         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15122         __builtin_ia32_wbinvd): New builtins.
15123         (SPECIAL_ARGS2): New.
15124         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15125         (SPECIAL_ARGS2): New.
15126         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15127         (ix86_valid_target_attribute_inner_p): Ditto.
15128         (ix86_init_mmx_sse_builtins): Add special_args2.
15129         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15130         TARGET_WBNOINVD_P): New.
15131         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15132         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15133         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15134         * config/i386/immintrin.h (_wbinvd): New intrinsic.
15135         * config/i386/pconfigintrin.h: New file.
15136         * config/i386/wbnoinvdintrin.h: Ditto.
15137         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15138         wbnoinvdintrin.h.
15139         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15141 2018-03-05  Richard Biener  <rguenther@suse.de>
15143         PR tree-optimization/84670
15144         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15145         member.
15146         (BB_VISITED_WITH_VISITED_SUCCS): New define.
15147         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15148         (compute_antic_aux): Only assert the number of values in ANTIC_IN
15149         doesn't grow if all successors (recursively) were visited at least
15150         once.
15152 2018-03-05  Richard Biener  <rguenther@suse.de>
15154         PR tree-optimization/84650
15155         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15156         if executed in the loop pipeline.
15158 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
15160         * doc/configfiles.texi (Configuration Files): Move info about
15161         conditionalizing $target-protos.h to...
15162         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
15163         differs from $target-protos.h.
15165 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
15166             Chung-Ju Wu  <jasonwucj@gmail.com>
15168         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15169         * config/nds32/nds32-multiple.md (setmemsi): Define.
15170         * config/nds32/nds32-memory-manipulation.c
15171         (nds32_gen_dup_4_byte_to_word_value): New.
15172         (emit_setmem_word_loop): New.
15173         (emit_setmem_byte_loop): New.
15174         (nds32_expand_setmem_loop): New.
15175         (nds32_expand_setmem_loop_v3m): New.
15176         (nds32_expand_setmem_unroll): New.
15177         (nds32_expand_setmem): New.
15179 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
15180             Chung-Ju Wu  <jasonwucj@gmail.com>
15182         * config/nds32/nds32-memory-manipulation.c
15183         (nds32_emit_load_store): New.
15184         (nds32_emit_post_inc_load_store): New.
15185         (nds32_emit_mem_move): New.
15186         (nds32_emit_mem_move_block): New.
15187         (nds32_expand_movmemsi_loop_unknown_size): New.
15188         (nds32_expand_movmemsi_loop_known_size): New.
15189         (nds32_expand_movmemsi_loop): New.
15190         (nds32_expand_movmemsi_unroll): New.
15191         (nds32_expand_movmemqi): Rename ...
15192         (nds32_expand_movmemsi): ... to this.
15193         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15194         (movmemsi): ... to this.
15195         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15196         (nds32_expand_movmemsi): ... to this.
15198 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
15199             Monk Chiang  <sh.chiang04@gmail.com>
15200             Chung-Ju Wu  <jasonwucj@gmail.com>
15202         * config/nds32/nds32-protos.h
15203         (nds32_expand_load_multiple): New arguments.
15204         (nds32_expand_store_multiple): Ditto.
15205         (nds32_valid_multiple_load_store): Rename ...
15206         (nds32_valid_multiple_load_store_p): ... to this.
15207         * config/nds32/nds32-memory-manipulation.c
15208         (nds32_expand_load_multiple): Refine implementation.
15209         (nds32_expand_store_multiple): Ditto.
15210         * config/nds32/nds32-multiple.md
15211         (load_multiple): Update nds32_expand_load_multiple interface.
15212         (store_multiple): Update nds32_expand_store_multiple interface.
15213         * config/nds32/nds32-predicates.c
15214         (nds32_valid_multiple_load_store): Rename ...
15215         (nds32_valid_multiple_load_store_p): ... to this and refine
15216         implementation.
15217         * config/nds32/predicates.md
15218         (nds32_load_multiple_and_update_address_operation): New predicate.
15219         (nds32_store_multiple_and_update_address_operation): New predicate.
15221 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
15222             Chung-Ju Wu  <jasonwucj@gmail.com>
15224         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15225         (combo): New attribute.
15226         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15228 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15230         * config/nds32/nds32.opt: Change -mcmodel= default value.
15232 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
15233             Monk Chiang  <sh.chiang04@gmail.com>
15234             Chung-Ju Wu  <jasonwucj@gmail.com>
15236         * config/nds32/constants.md (unspec_element): New enum.
15237         * config/nds32/constraints.md (Umw): New constraint.
15238         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15239         * config/nds32/nds32-intrinsic.md: Likewise.
15240         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15241         (nds32_valid_smw_lwm_base_p): New.
15242         (nds32_output_smw_single_word): New.
15243         (nds32_output_lmw_single_word): New.
15244         (nds32_expand_unaligned_load): New.
15245         (nds32_expand_unaligned_store): New.
15246         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15247         (nds32_output_smw_single_word): Declare.
15248         (nds32_output_lmw_single_word): Declare.
15249         (nds32_expand_unaligned_load): Declare.
15250         (nds32_expand_unaligned_store): Declare.
15251         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15252         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15253         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15254         NDS32_BUILTIN_UASTORE_DW.
15255         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15256         predicate.
15258 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
15259             Kito Cheng  <kito.cheng@gmail.com>
15260             Chung-Ju Wu  <jasonwucj@gmail.com>
15262         * config/nds32/nds32-intrinsic.c
15263         (nds32_expand_builtin_null_ftype_reg): Delete.
15264         (nds32_expand_builtin_reg_ftype_imm): Ditto.
15265         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15266         (nds32_read_argument): New.
15267         (nds32_legitimize_target): Ditto.
15268         (nds32_legitimize_argument): Ditto.
15269         (nds32_check_constant_argument): Ditto.
15270         (nds32_expand_unop_builtin): Ditto.
15271         (nds32_expand_unopimm_builtin): Ditto.
15272         (nds32_expand_binop_builtin): Ditto.
15273         (nds32_builtin_decl_impl): Ditto.
15274         (builtin_description): Ditto.
15275         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15276         (nds32_init_builtins_impl): Ditto.
15277         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15278         (nds32_builtin_decl): New.
15279         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15280         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15282 2018-03-02  Jeff Law  <law@redhat.com>
15284         * reorg.c (stop_search_p): Handle DEBUG_INSN.
15285         (redundant_insn, fill_simple_delay_slots): Likewise.
15286         (fill_slots_from_thread): Likewise.
15287         * resource.c (mark_referenced_resources): Likewise.
15288         (mark_set_resources, find_dead_or_set_registers): Likewise.
15290 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
15292         * substring-locations.h (format_warning_va): Formatting fix for
15293         ATTRIBUTE_GCC_DIAG.
15294         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15295         argument.
15296         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15297         * substring-locations.c: Include intl.h.
15298         (format_warning_va): Turned into small wrapper around
15299         format_warning_n_va, renamed to ...
15300         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15301         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15302         use ngettext.
15303         (format_warning_at_substring_n): New function.
15304         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15305         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
15306         format_warning_at_substring with just a shorter name instead of
15307         const function pointer.
15308         (fmtwarn_n): New function.
15309         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15310         appropriate, get rid of all the fmtstr temporaries, move conditionals
15311         with G_() wrapped string literals directly into fmtwarn arguments,
15312         cast dir.len to (int), formatting fixes.
15314 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
15316         * doc/invoke.texi: Remove "Cilk Plus" references.
15318 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
15319             Richard Biener  <rguenther@suse.de>
15321         PR ipa/84628
15322         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15323         for error or warning attributes if CALL_FROM_THUNK_P is set.
15324         Formatting fixes.
15326 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
15328         PR target/56540
15329         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15330         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15332         PR target/56540
15333         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15334         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15336         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15337         instead of -1U in last predictors element's probability member.
15339 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
15341         PR ipa/83983
15342         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15343         arguments if they are comparable.
15345 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
15347         PR tree-optimization/84634
15348         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15349         masks and masked_loop_p with a single loop_masks, making sure it's
15350         null for bb vectorization.
15352 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
15354         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15355         (vect_analyze_data_ref_access): Use loop->safe_len rather than
15356         loop->force_vectorize to check whether there is no alias.
15358 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
15360         PR target/84614
15361         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15362         prototypes.
15363         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15364         comments.
15365         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15366         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15367         instead of a loop around prev_real_insn.
15368         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15369         prev_real_insn.
15371         PR inline-asm/84625
15372         * config/i386/i386.c (ix86_print_operand): Use conditional
15373         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15374         zero vector.
15376 2018-03-02  Richard Biener  <rguenther@suse.de>
15378         PR tree-optimization/84427
15379         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15380         (bitmap_set_subtract_values): Rewrite to handle multiple
15381         exprs per value.
15382         (clean): Likewise.
15383         (prune_clobbered_mems): Likewise.
15384         (phi_translate): Take edge instead of pred/phiblock.
15385         (phi_translate_1): Likewise.
15386         (phi_translate_set): Likewise.  Insert all translated
15387         exprs for a value into the set, keeping possibly multiple
15388         expressions per value.
15389         (compute_antic_aux): Adjust for phi_translate changes.
15390         When intersecting union the expressions and prune those
15391         not in the final value set, keeping possibly multiple
15392         expressions per value.  Do not use value-insertion
15393         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15394         all expressions.  Add verification that the value-sets
15395         only shrink during iteration.
15396         (compute_partial_antic_aux): Adjust for the phi_translate changes.
15397         (do_pre_regular_insertion): Likewise.
15398         (do_pre_partial_partial_insertion): Likewise.
15400 2018-03-02  Richard Biener  <rguenther@suse.de>
15402         PR target/82005
15403         * config/darwin.c (saved_debug_info_level): New static global.
15404         (darwin_asm_lto_start): Disable debug info generation for LTO out.
15405         (darwin_asm_lto_end): Restore debug info generation settings.
15407 2018-03-01  Martin Liska  <mliska@suse.cz>
15409         PR sanitizer/82484
15410         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15411         volatile arguments.
15413 2018-03-01  Richard Biener  <rguenther@suse.de>
15415         PR debug/84645
15416         * dwarf2out.c (gen_variable_die): Properly handle late VLA
15417         type annotation with LTO when debug was disabled at compile-time.
15419 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
15421         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15422         XINT with INTVAL.
15423         (mips_final_postscan_insn): Likewise.
15425 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
15427         PR rtl-optimization/84528
15428         * alias.c (init_alias_target): Add commentary.
15429         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15430         a unique base value if the frame pointer is not eliminated
15431         to the stack pointer.
15433 2018-03-01  Tom de Vries  <tom@codesourcery.com>
15435         PR rtl-optimization/83327
15436         * lra-int.h (hard_regs_spilled_into): Declare.
15437         * lra.c (hard_regs_spilled_into): Define.
15438         (init_reg_info): Init hard_regs_spilled_into.
15439         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15440         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15441         (process_bb_lives): Handle hard_regs_spilled_into.
15442         (lra_create_live_ranges_1): Before doing liveness propagation, clear
15443         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15445 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
15447         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15448         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15449         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15450         * config/rs6000/aix72.h: New file.
15452 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
15454         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15455         instead of warning_at with conditional singular and plural messages
15456         where possible.
15458         PR target/52991
15459         * stor-layout.c (update_alignment_for_field): For
15460         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15461         && !DECL_PACKED (field), do the alignment update, just use
15462         only desired_align instead of MAX (type_align, desired_align)
15463         as the alignment.
15464         (place_field): Don't do known_align < desired_align handling
15465         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15466         is non-NULL, instead do it after rli->prev_field handling and
15467         only if not within a bitfield word.  For DECL_PACKED (field)
15468         use type_align of BITS_PER_UNIT.
15470 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
15472         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15473         superfluous parentheses and trailing spaces.
15475 2018-02-28  Richard Biener  <rguenther@suse.de>
15477         PR tree-optimization/84584
15478         * graphite-scop-detection.c (scop_detection::add_scop): Discard
15479         SCoPs with fake exit edge.
15481 2018-02-28  Martin Liska  <mliska@suse.cz>
15483         PR testsuite/84597
15484         * timevar.c (timer::print): Fix format to properly print 100%
15485         values.
15487 2018-02-28  Richard Biener  <rguenther@suse.de>
15489         PR middle-end/84607
15490         * genmatch.c (capture_info::walk_match): Do not mark
15491         captured expressions without operands as expr_p given
15492         they act more like predicates and should be subject to
15493         "lost tail" side-effect preserving.
15495 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
15497         PR rtl-optimization/81611
15498         * auto-inc-dec.c (attempt_change): Move dead note from
15499         mem_insn if it's the next use of regno
15500         (find_address): Take address use of reg holding
15501         non-incremented value.  Add parm to limit search to the named
15502         reg only.
15503         (merge_in_block): Attempt to use a mem insn that is the next
15504         use of the original regno.
15506 2018-02-27  Martin Sebor  <msebor@redhat.com>
15508         PR c++/83871
15509         * doc/invoke.texi (-Wmissing-attributes): New option.
15510         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15512 2018-02-27  Martin Sebor  <msebor@redhat.com>
15514         PR translation/84207
15515         * diagnostic-core.h (warning_n, error_n, inform_n): Change
15516         n argument to unsigned HOST_WIDE_INT.
15517         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15518         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
15519         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15520         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15522 2018-02-27  Richard Biener  <rguenther@suse.de>
15524         PR tree-optimization/84512
15525         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15526         Do not use the estimate returned from record_stmt_cost for
15527         the scalar iteration cost but sum properly using add_stmt_cost.
15529 2018-02-27  Richard Biener  <rguenther@suse.de>
15531         PR tree-optimization/84466
15532         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15533         Adjust last change to less strictly validate use operands.
15535 2018-02-27  Martin Liska  <mliska@suse.cz>
15537         PR gcov-profile/84548
15538         * gcov.c (process_file): Allow partial overlap and consider it
15539         also as group functions.
15540         (output_lines): Properly calculate range of lines for a group.
15542 2018-02-27  Martin Liska  <mliska@suse.cz>
15544         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15545         'ggc' suffixes.  Change first column width.
15546         (timer::print): Fix formatting of the column.
15548 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
15550         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15551         preserve inline entry blocks for the sake of debug inline
15552         entry point markers alone.
15553         (remove_unused_locals): Suggest in comments a better place to
15554         force the preservation of inline entry blocks that are
15555         otherwise unused, but do not preserve them.
15557 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
15559         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15561 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
15563         PR target/84039
15564         * config/i386/constraints.md (Bs): Replace
15565         ix86_indirect_branch_register with
15566         TARGET_INDIRECT_BRANCH_REGISTER.
15567         (Bw): Likewise.
15568         * config/i386/i386.md (indirect_jump): Likewise.
15569         (tablejump): Likewise.
15570         (*sibcall_memory): Likewise.
15571         (*sibcall_value_memory): Likewise.
15572         Peepholes of indirect call and jump via memory: Likewise.
15573         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15574         (*sibcall_value_GOT_32): Likewise.
15575         * config/i386/predicates.md (indirect_branch_operand): Likewise.
15576         (GOT_memory_operand): Likewise.
15577         (call_insn_operand): Likewise.
15578         (sibcall_insn_operand): Likewise.
15579         (GOT32_symbol_operand): Likewise.
15580         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15582 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
15584         PR rtl-optimization/83496
15585         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15586         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
15587         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15588         redundant insn, if any.
15589         (relax_delay_slots): Likewise.
15590         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15592 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
15594         PR tree-optimization/83965
15595         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15596         that grouped statements are part of a reduction chain.  Return
15597         true if the statement is not marked as a reduction itself but
15598         is part of a group.
15599         (vect_recog_dot_prod_pattern): Don't check whether the statement
15600         is part of a group here.
15601         (vect_recog_sad_pattern): Likewise.
15602         (vect_recog_widen_sum_pattern): Likewise.
15604 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
15606         PR debug/84545
15607         * final.c (rest_of_clean_state): Also look for calls inside sequences.
15609 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
15611         PR target/84530
15612         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15613         the bool argument.
15614         (ix86_output_indirect_function_return): New prototype.
15615         (ix86_split_simple_return_pop_internal): Likewise.
15616         * config/i386/i386.c (indirect_return_via_cx): New.
15617         (indirect_return_via_cx_bnd): Likewise.
15618         (indirect_thunk_name): Handle return va CX_REG.
15619         (output_indirect_thunk_function): Create alias for
15620         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15621         (ix86_output_indirect_jmp): Remove the bool argument.
15622         (ix86_output_indirect_function_return): New function.
15623         (ix86_split_simple_return_pop_internal): Likewise.
15624         * config/i386/i386.md (*indirect_jump): Don't pass false
15625         to ix86_output_indirect_jmp.
15626         (*tablejump_1): Likewise.
15627         (simple_return_pop_internal): Change it to define_insn_and_split.
15628         Call ix86_split_simple_return_pop_internal to split it for
15629         -mfunction-return=.
15630         (simple_return_indirect_internal): Call
15631         ix86_output_indirect_function_return instead of
15632         ix86_output_indirect_jmp.
15634 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
15636         PR bootstrap/84405
15637         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15638         memset and value initialization afterwards.
15640 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
15642         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15644 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
15646         PR target/84521
15647         * common/config/aarch64/aarch64-common.c
15648         (aarch_option_optimization_table[]): Switch
15649         off fomit-frame-pointer
15651 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
15652             Chung-Ju Wu  <jasonwucj@gmail.com>
15654         * config/nds32/nds32-multiple.md (load_multiple): Disallow
15655         volatile memory.
15656         (store_multiple): Ditto.
15658 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
15660         * config.gcc: Add --with-cpu support for nds32 target.
15661         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15662         * config/nds32/nds32.opt: Add -mcpu= option.
15664 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
15666         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15667         isel=yes): Warn for these deprecated options.
15669 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
15671         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15672         ISA_2_5_MASKS_EMBEDDED.
15674 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
15676         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15677         p->max as pointers rather than using iterative_hash_expr.
15679 2018-02-23  Carl Love  <cel@us.ibm.com>
15681         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15682         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15683         BU_P8V_OVERLOAD_2.
15684         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15685         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15686         P8V_BUILTIN_VEC_VUNSIGNED2.
15688 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
15690         PR target/81572
15691         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15692         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15693         LRA_UNKNOWN_ALT.
15694         * lra-constraints.c (curr_insn_transform): Set up
15695         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
15696         LRA_UNKNOWN_ALT.
15697         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15698         * lra-eliminations.c (spill_pseudos): Ditto.
15699         (process_insn_for_elimination): Ditto.
15700         * lra-lives.c (reg_early_clobber_p): Use the new macros.
15701         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15702         LRA_NON_CLOBBERED_ALT.
15704 2018-02-22  Martin Sebor  <msebor@redhat.com>
15706         PR tree-optimization/84480
15707         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15708         to maybe_diag_stxncpy_trunc.  Call it.
15709         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15710         from gimple_fold_builtin_strcpy.  Print inlining stack.
15711         (handle_builtin_stxncpy): Print inlining stack.
15712         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15714 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
15716         PR target/84176
15717         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15718         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15719         and -fcheck-pointer-bounds are used together.
15720         (indirect_thunk_prefix): New enum.
15721         (indirect_thunk_need_prefix): New function.
15722         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
15723         "_nt" instead of "_bnd" for NOTRACK prefix.
15724         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15725         (output_indirect_thunk_function): Likewise.
15726         (): Likewise.
15727         (ix86_code_end): Update output_indirect_thunk_function calls.
15728         (ix86_output_indirect_branch_via_reg): Replace
15729         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15730         (ix86_output_indirect_branch_via_push): Likewise.
15731         (ix86_output_function_return): Likewise.
15732         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15733         incompatible with -fcf-protection=branch and
15734         -fcheck-pointer-bounds.
15736 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
15738         PR target/83335
15739         * config/aarch64/aarch64.c (aarch64_print_address_internal):
15740         Change gcc_assert call to output_operand_lossage.
15742 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
15744         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15746 2018-02-22  DJ Delorie  <dj@redhat.com>
15747             Sebastian Perta  <sebastian.perta@renesas.com>
15748             Oleg Endo  <olegendo@gcc.gnu.org>
15750         * config/rx/rx.c (rx_rtx_costs): New function.
15751         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15753 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
15755         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15757 2018-02-22  Martin Liska  <mliska@suse.cz>
15759         PR driver/83193
15760         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15761         Add "native" as a possible value.
15763 2018-02-22  Martin Liska  <mliska@suse.cz>
15765         PR driver/83193
15766         * config/i386/i386.c (ix86_option_override_internal):
15767         Add "native" as a possible value for -march and -mtune.
15769 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
15771         PR target/84502
15772         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15773         to all type variants.
15775         PR tree-optimization/84503
15776         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15777         width as info->bitpos + info->bitsize - start.
15778         (merged_store_group::merge_overlapping): Simplify width computation.
15779         (check_no_overlap): New function.
15780         (imm_store_chain_info::try_coalesce_bswap): Compute expected
15781         start + width and last_order of the group, fail if check_no_overlap
15782         fails.
15783         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15784         to group if check_no_overlap fails.
15786 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
15788         * config/rs6000/altivec.md: Delete contraint arguments to
15789         define_expand, define_split, and define_peephole2, and in
15790         define_insn_and_split if always unused.
15791         * config/rs6000/darwin.md: Ditto.
15792         * config/rs6000/dfp.md: Ditto.
15793         * config/rs6000/rs6000.md: Ditto.
15794         * config/rs6000/sync.md: Ditto.
15795         * config/rs6000/vector.md: Ditto.
15796         * config/rs6000/vsx.md: Ditto.
15798 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
15800         * config/rs6000/altivec.md: Write output control strings as braced
15801         blocks instead of double-quoted strings.
15802         * config/rs6000/darwin.md: Ditto.
15803         * config/rs6000/rs6000.md: Ditto.
15804         * config/rs6000/vector.md: Ditto.
15805         * config/rs6000/vsx.md: Ditto.
15807 2018-02-21  Jason Merrill  <jason@redhat.com>
15809         PR c++/84314 - ICE with templates and fastcall attribute.
15810         * attribs.c (build_type_attribute_qual_variant): Remove assert.
15812 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
15814         * ipa-cp.c (determine_versionability): Fix comment typos.
15816 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
15818         PR c/84229
15819         * ipa-cp.c (determine_versionability): Do not version functions caling
15820         va_arg_pack.
15822 2018-02-21  Martin Liska  <mliska@suse.cz>
15824         PR driver/83193
15825         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15826         Add "native" as a possible value.
15827         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
15828         the macro when native cpu detection is available.
15830 2018-02-21  Martin Liska  <mliska@suse.cz>
15832         PR driver/83193
15833         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15834         Add "native" as a possible value.
15835         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15836         when native cpu detection is available.
15838 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
15839             Martin Sebor  <msebor@redhat.com>
15841         PR tree-optimization/84478
15842         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15843         false.
15844         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15845         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15846         support which is conservatively correct, for 2 only stay conservative
15847         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
15848         argument to the 2 argument get_range_strlen, adjust 6 arg
15849         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15850         false.
15851         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15852         (gimple_fold_builtin_strlen): Pass true as last argument to
15853         get_range_strlen.
15855 2018-02-20  Martin Sebor  <msebor@redhat.com>
15857         PR middle-end/84095
15858         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15859         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
15860         (builtin_memref::builtin_memref): Factor out parts into
15861         set_base_and_offset and call it.
15863 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
15865         PR middle-end/84406
15866         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15867         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15868         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
15869         search at the associated MODE_INT.
15871 2018-02-20  Jeff Law  <law@redhat.com>
15873         PR middle-end/82123
15874         PR tree-optimization/81592
15875         PR middle-end/79257
15876         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15877         for range data rather than using global data.
15878         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15879         range data rather than using global data.
15880         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15881         pass it to children as needed.
15882         (struct directive::fmtresult): Similarly.
15883         (struct directive::set_width): Similarly.
15884         (struct directive::set_precision): Similarly.
15885         (format_integer, format_directive, parse_directive): Similarly.
15886         (format_none): Accept unnamed vr_values parameter.
15887         (format_percent, format_floating, format_character): Similarly.
15888         (format_string, format_plain): Similarly.
15889         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
15890         the EVRP range analyzer for range data rather than using global data.
15891         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
15892         gimple-ssa-evrp-analyze.h
15893         (class sprintf_dom_walker): Add after_dom_children member function.
15894         Add evrp_range_analyzer member.
15895         (sprintf_dom_walker::before_dom_children): Call into the EVRP
15896         range analyzer as needed.
15897         (sprintf_dom_walker::after_dom_children): New member function.
15898         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
15899         if not optimizing.
15900         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
15901         (evrp_range_analyzer::pop_to_marker): Likewise.
15903 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
15905         PR tree-optimization/84419
15906         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
15907         with the required type if its current type is compatible but
15908         different.
15910 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
15912         PR middle-end/82004
15913         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15914         after vectorization.
15916 2018-02-20  Martin Liska  <mliska@suse.cz>
15918         PR driver/83193
15919         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15920         possible values if we don't have a hint.
15922 2018-02-20  Martin Liska  <mliska@suse.cz>
15924         PR c/84310
15925         PR target/79747
15926         * final.c (shorten_branches): Build align_tab array with one
15927         more element.
15928         * opts.c (finish_options): Add alignment option limit check.
15929         (MAX_CODE_ALIGN): Likewise.
15930         (MAX_CODE_ALIGN_VALUE): Likewise.
15931         * doc/invoke.texi: Document maximum allowed option value for
15932         all -falign-* options.
15934 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
15936         PR target/84146
15937         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
15938         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
15939         * var-tracking.c (emit_note_insn_var_location): Remove all references
15940         to NOTE_INSN_CALL_ARG_LOCATION.
15941         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
15942         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
15943         Use copy_rtx_if_shared.
15944         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
15945         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
15946         (dwarf2out_var_location): Remove handling of
15947         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
15948         on call_insn.
15949         * final.c (final_scan_insn): Remove all references to
15950         NOTE_INSN_CALL_ARG_LOCATION.
15951         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
15952         before dumping final insns.
15953         * except.c (emit_note_eh_region_end): Remove all references to
15954         NOTE_INSN_CALL_ARG_LOCATION.
15955         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
15956         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
15957         * config/arc/arc.c (hwloop_optimize): Likewise.
15958         * config/arm/arm.c (create_fix_barrier): Likewise.
15959         * config/s390/s390.c (s390_chunkify_start): Likewise.
15960         * config/sh/sh.c (find_barrier): Likewise.
15961         * config/i386/i386.c (rest_of_insert_endbranch,
15962         ix86_seh_fixup_eh_fallthru): Likewise.
15963         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
15964         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
15965         * config/frv/frv.c (frv_function_prologue): Likewise.
15966         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
15967         reg note.
15968         (note_outside_basic_block_p): Remove all references to
15969         NOTE_INSN_CALL_ARG_LOCATION.
15970         * gengtype.c (adjust_field_rtx_def): Likewise.
15971         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
15972         Likewise.
15973         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
15974         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
15976         PR c++/84444
15977         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
15978         is ADDR_EXPR.
15980         PR tree-optimization/84452
15981         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
15982         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
15983         is NULL.
15985 2018-02-19  Martin Liska  <mliska@suse.cz>
15987         PR sanitizer/82183
15988         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
15990 2018-02-19  Martin Liska  <mliska@suse.cz>
15991             Richard Sandiford  <richard.sandiford@linaro.org>
15993         PR tree-optimization/82491
15994         * gimple-fold.c (get_base_constructor): Make earlier bail out
15995         to prevent ubsan.
15997 2018-02-19  Carl Love  <cel@us.ibm.com>
15999         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16000         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16001         BU_P8V_OVERLOAD_1.
16002         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16003         P8V_BUILTIN_VEC_NEG.
16005 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
16007         * config/rl78/rl78.md (movdf): New define expand.
16009 2018-02-19  Martin Liska  <mliska@suse.cz>
16011         PR other/80589
16012         * doc/invoke.texi: Fix typo.
16013         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16015 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
16017         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16018         handle rs6000_single_float and rs6000_double_float specially for
16019         e500 family CPUs.
16021 2018-02-16  Jeff Law  <law@redhat.com>
16023         * config/rx/rx.c (add_pop_cfi_notes): New function.;
16024         (pop_regs): Use it.
16026 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
16028         PR ipa/84425
16029         * ipa-inline.c (inline_small_functions): Fix a typo.
16031 2018-02-16  Nathan Sidwell  <nathan@acm.org>
16033         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16035 2018-02-16  Carl Love  <cel@us.ibm.com>
16037         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16038         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16039         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16040         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16041         expansion to P8V_BUILTIN_VEC_FLOAT2.
16043 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
16045         PR rtl-optimization/70023
16046         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16047         src_regno into account.
16049 2018-02-16  Carl Love  <cel@us.ibm.com>
16051         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16052         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16053         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16054         * config/rs6000/rs6000.c: Remove case statements for
16055         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16056         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16057         and P9V_BUILTIN_VEC_VINSERT4B.
16058         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16059         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16060         * config/rs6000/vsx.md:
16061         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16062         vec_insert4b.
16064 2018-02-16  Carl Love  <cel@us.ibm.com>
16066         * config/rs6000/altivec.h: Add builtin names vec_extract4b
16067         vec_insert4b.
16068         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16069         definitions.
16070         * config/rs6000/rs6000-c.c: Add the definitions for
16071         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16072         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16073         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16074         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
16075         definition for insert4b and define insn *insert3b_internal.
16076         * doc/extend.texi: Add documentation for vec_extract4b.
16078 2018-02-16  Nathan Sidwell  <nathan@acm.org>
16080         * doc/extend.texi (Backwards Compatibility): Mention friend
16081         injection.  Note for-scope is deprecated.
16082         * doc/invoke.texi (-ffriend-injection): Deprecate.
16084 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
16086         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16087         that moved to I2, also allow destinations that are a paradoxical
16088         subreg (instead of a normal reg).
16090 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
16092         PR target/83831
16093         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16094         to QImode.
16096 2018-02-16  Richard Biener  <rguenther@suse.de>
16098         PR tree-optimization/84037
16099         PR tree-optimization/84016
16100         PR target/82862
16101         * config/i386/i386.c (ix86_builtin_vectorization_cost):
16102         Adjust vec_construct for the fact we need additional higher latency
16103         128bit inserts for AVX256 and AVX512 vector builds.
16104         (ix86_add_stmt_cost): Scale vector construction cost for
16105         elementwise loads.
16107 2018-02-16  Richard Biener  <rguenther@suse.de>
16109         PR tree-optimization/84417
16110         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16111         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16112         (non_rewritable_lvalue_p): Likewise, use poly-ints.
16114 2018-02-16  Martin Liska  <mliska@suse.cz>
16116         PR sanitizer/84307
16117         * internal-fn.def (ASAN_CHECK): Set proper flags.
16118         (ASAN_MARK): Likewise.
16120 2018-02-16  Julia Koval  <julia.koval@intel.com>
16122         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16123         from PTA_CANNONLAKE.
16125 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
16127         PR target/84272
16128         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16129         Use ++iter rather than iter++ for std::list iterators.
16130         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
16131         defer deleting them until all nodes in the forest are processed.  Do
16132         free even leaf nodes.  Change to_process into auto_vec.
16134         PR bootstrap/84405
16135         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16136         * vec.h (vec_default_construct): Use memset instead of placement new
16137         if BROKEN_VALUE_INITIALIZATION is defined.
16138         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16139         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16140         is defined.
16142         PR rtl-optimization/83723
16143         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16144         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
16145         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
16146         recursive calls.
16147         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16148         callers.
16149         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16151 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
16153         PR rtl-optimization/81443
16154         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16155         from inner REGs to paradoxical SUBREGs.
16157 2018-02-16  Richard Biener  <rguenther@suse.de>
16159         PR tree-optimization/84399
16160         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16161         For operands we can analyze at their definition make sure we can
16162         analyze them at each use as well.
16164 2018-02-16  Richard Biener  <rguenther@suse.de>
16166         PR tree-optimization/84190
16167         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16168         volatile accesses if the decl isn't volatile.
16170 2018-02-15  Jason Merrill  <jason@redhat.com>
16172         PR c++/84314 - ICE with templates and fastcall attribute.
16173         * attribs.c (build_type_attribute_qual_variant): Don't clobber
16174         TYPE_CANONICAL on an existing type.
16176 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
16178         PR tree-optimization/84383
16179         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16180         dstoff nor call operand_equal_p if dstbase is NULL.
16182         PR tree-optimization/84334
16183         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16184         also a CONSTANT_CLASS_P, punt.
16186 2018-02-14  Jim Wilson  <jimw@sifive.com>
16188         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16189         first SMALL_OPERAND check.  New local min_second_step.  Move assert
16190         to where locals are set.  Add TARGET_RVC support.
16191         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16193 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
16195         * doc/invoke.texi: Correct -Wformat-overflow code sample.
16197 2018-02-14  Martin Sebor  <msebor@redhat.com>
16199         PR tree-optimization/83698
16200         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16201         arrays constrain the offset range to their bounds.
16202         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16203         (builtin_access::overlap): Avoid setting the size of overlap if it's
16204         already been set.
16205         (maybe_diag_overlap): Also consider arrays when deciding what values
16206         of offsets to include in diagnostics.
16208 2018-02-14  Martin Sebor  <msebor@redhat.com>
16210         PR c/84108
16211         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16212         that correspond to the kind of a declaration.
16214 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
16216         PR target/83984
16217         * config/pa/pa.md: Load address of PIC label using the linkage table
16218         if the label is nonlocal.
16220 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16222         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16223         warning message if user requests -maltivec=be.
16224         * doc/invoke.texi: Document deprecation of -maltivec=be.
16226 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
16228         PR target/84220
16229         * config/rs6000/rs6000-c.c: Update definitions for
16230         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16231         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16233 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
16235         PR target/84239
16236         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16237         add _get_ssp intrinsics. Remove argument from
16238         __builtin_ia32_rdssp[d|q].
16239         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16240         * config/i386/i386-builtin.def: Remove argument from
16241         __builtin_ia32_rdssp[d|q].
16242         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16243         ix86_expand_special_args_builtin for _rdssp[d|q].
16244         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16245         Clear register before usage.
16246         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16247         Add documentation for new _get_ssp and _inc_ssp intrinsics.
16249 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
16251         PR tree-optimization/84357
16252         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16253         operand 1 of an ARRAY_REF too.
16255 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
16257         PR target/83831
16258         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16259         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16260         declarations.
16261         (set_of_reg): New struct.
16262         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16263         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16264         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16265         functions.
16266         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16267         Split into bitclr, bitset, bitinvert patterns if appropriate.
16268         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16269         use rx_fuse_in_memory_bitop.
16270         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16271         to named insn, correct maximum insn length.
16273 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
16275         PR target/79242
16276         * machmode.def: Define a complex mode for PARTIAL_INT.
16277         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16278         MODE_PARTIAL_INT.
16279         * doc/rtl.texi: Document CSPImode.
16280         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16281         handling.
16282         (msp430_hard_regno_nregs_with_padding): Likewise.
16284 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
16286         PR target/84279
16287         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16289 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
16291         PR rtl-optimization/84169
16292         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
16293         we generated a parallel as new i3 and we split that to new i2 and i3
16294         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
16295         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
16296         those to i2, not i1.  Partially rewrite this scan code.
16298 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
16300         PR c/82210
16301         * stor-layout.c (place_field): For variable length fields, adjust
16302         offset_align afterwards not just based on the field's alignment,
16303         but also on the size.
16305         PR middle-end/84309
16306         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16307         of exps and logs in the use_exp2 case.
16309 2018-02-13  Jeff Law  <law@redhat.com>
16311         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16312         entry for "vector".
16314         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16315         ARGS as unused.
16317 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
16319         PR debug/84342
16320         PR debug/84319
16321         * common.opt (gas-loc-support, gas-locview-support): New.
16322         (ginline-points, ginternal-reset-location-views): New.
16323         * doc/invoke.texi: Document them.  Use @itemx where intended.
16324         (gvariable-location-views): Adjust.
16325         * target.def (reset_location_view): New.
16326         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16327         (TARGET_RESET_LOCATION_VIEW): New.
16328         * doc/tm.texi: Rebuilt.
16329         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16330         (dwarf2out_default_as_locview_support): New.
16331         (output_asm_line_debug_info): Use option variables.
16332         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16333         (output_loc_list): Likewise.
16334         (add_high_low_attributes): Check option variables.
16335         Don't output entry view attribute in strict mode.
16336         (gen_inlined_subroutine_die): Check option variables.
16337         (dwarf2out_inline_entry): Likewise.
16338         (init_sections_and_labels): Likewise.
16339         (dwarf2out_early_finish): Likewise.
16340         (maybe_reset_location_view): New, from...
16341         (dwarf2out_var_location): ... here.  Call it.
16342         * debug.h (dwarf2out_default_as_loc_support): Declare.
16343         (dwarf2out_default_as_locview_support): Declare.
16344         * hooks.c (hook_int_rtx_insn_0): New.
16345         * hooks.h (hook_int_rtx_insn_0): Declare.
16346         * toplev.c (process_options): Take -gas-loc-support and
16347         -gas-locview-support from dwarf2out.  Enable
16348         -gvariable-location-views by default only with locview
16349         assembler support.  Enable -ginternal-reset-location-views by
16350         default only if the target defines the corresponding hook.
16351         Enable -ginline-points by default if location views are
16352         enabled; force it disabled if statement frontiers are
16353         disabled.
16354         * tree-inline.c (expand_call_inline): Check option variables.
16355         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16357 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
16359         PR tree-optimization/84321
16360         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16361         handling.  Also check whether the anti-range contains any values
16362         that satisfy the mask; switch to a VR_RANGE if not.
16364 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
16366         PR sanitizer/84340
16367         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16369 2018-02-13  Martin Jambor  <mjambor@suse.cz>
16371         PR c++/83990
16372         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16373         of call statements, also set location of a load to a temporary.
16375 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
16377         * config/rl78/rl78.c (add_vector_labels): New function.
16378         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16379         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16380         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16381         which checks that no arguments are passed.
16382         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16383         * doc/extend.texi: Documentation for the new attribute.
16385 2018-02-13  Andreas Schwab  <schwab@suse.de>
16387         * config/riscv/linux.h (CPP_SPEC): Define.
16389 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
16391         PR target/84335
16392         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16393         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16394         OPTION_MASK_ISA_AES as first argument to def_builtin_const
16395         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16396         instead of OPTION_MASK_ISA_PCLMUL as first argument to
16397         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16398         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16399         temporarily for AES and PCLMUL builtins.
16401         PR tree-optimization/84339
16402         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16403         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16404         Formatting fixes.
16406         PR middle-end/84309
16407         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16408         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16409         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16410         inline function.
16411         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16412         inline function.
16413         * omp-simd-clone.h: New file.
16414         * omp-simd-clone.c: Include omp-simd-clone.h.
16415         (expand_simd_clones): No longer static.
16416         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16417         cgraph.h and omp-simd-clone.h.
16418         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16419         (vect_recog_widen_shift_pattern): Formatting fix.
16420         (vect_pattern_recog_1): Don't check optab for calls.
16422         PR target/84336
16423         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16424         operands[2] into a REG before using gen_lowpart on it.
16426 2018-02-12  Jeff Law  <law@redhat.com>
16428         PR target/83760
16429         * config/sh/sh.c (find_barrier): Consider a sibling call
16430         a barrier as well.
16432         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16433         successfully back substituting a reg.
16435 2018-02-12  Richard Biener  <rguenther@suse.de>
16437         PR tree-optimization/84037
16438         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16439         parameter, move visited init to caller.
16440         (vect_slp_analyze_operations): Separate cost from validity
16441         check, initialize visited once for all instances.
16442         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16443         for all instances.
16444         * tree-vect-stmts.c (vect_model_simple_cost): Make early
16445         out an assert.
16446         (vect_model_promotion_demotion_cost): Likewise.
16447         (vectorizable_bswap): Guard cost modeling with !slp_node
16448         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16449         SLP stmts.
16450         (vectorizable_call): Likewise.
16451         (vectorizable_conversion): Likewise.
16452         (vectorizable_assignment): Likewise.
16453         (vectorizable_shift): Likewise.
16454         (vectorizable_operation): Likewise.
16455         (vectorizable_store): Likewise.
16456         (vectorizable_load): Likewise.
16457         (vectorizable_condition): Likewise.
16458         (vectorizable_comparison): Likewise.
16460 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
16462         PR sanitizer/84307
16463         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16464         (ASAN_MARK): Fix fnspec to account for return value, change pointer
16465         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16467 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
16469         PR middle-end/83665
16470         * params.def (inline-min-speedup): Increase from 8 to 15.
16471         (max-inline-insns-auto): Decrease from 40 to 30.
16472         * ipa-split.c (consider_split): Add some buffer for function to
16473         be considered inlining candidate.
16474         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16475         default values.
16477 2018-02-12  Richard Biener  <rguenther@suse.de>
16479         PR tree-optimization/84037
16480         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16481         matched stmts if we cannot swap the non-matched ones.
16483 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
16485         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16486         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16487         _mm_maskz_scalef_round_ss): New intrinsics.
16488         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16489         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16490         __builtin_ia32_scalefss_round): Remove.
16491         (__builtin_ia32_scalefsd_mask_round,
16492         __builtin_ia32_scalefss_mask_round): New intrinsics.
16493         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16494         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16495         ((match_operand:VF_128 2 "<round_nimm_predicate>"
16496         "<round_constraint>")): Changed to ...
16497         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16498         "<round_scalar_constraint>")): ... this.
16499         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16500         %0, %1, %2<round_op3>}"): Changed to ...
16501         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16502         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16503         %2<round_scalar_mask_op3>}"): ... this.
16504         * config/i386/subst.md (round_scalar_nimm_predicate): New.
16506 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
16508         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16509         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16510         (_mm_maskz_sqrt_round_ss): New intrinsics.
16511         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16512         (__builtin_ia32_sqrtsd_mask_round)
16513         (__builtin_ia32_sqrtss_mask_round): New builtins.
16514         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16515         (__builtin_ia32_sqrtss_round): Remove.
16516         (__builtin_ia32_sqrtsd_mask_round)
16517         (__builtin_ia32_sqrtss_mask_round): New builtins.
16518         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16519         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16520         ((match_operand:VF_128 1 "vector_operand"
16521         "xBm,<round_constraint>")): Changed to ...
16522         ((match_operand:VF_128 1 "vector_operand"
16523         "xBm,<round_scalar_constraint>")): ... this.
16524         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16525         %0, %2, %<iptr>1<round_op3>}): Changed to ...
16526         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16527         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16528         %<iptr>1<round_scalar_mask_op3>}): ... this.
16529         ((set_attr "prefix" "<round_prefix>")): Changed to ...
16530         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16532 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
16534         PR target/84266
16535         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16536         Cast vec_cmpeq result to correct type.
16537         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16538         Cast vec_cmpgt result to correct type.
16540 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
16542         * final.c (final_scan_insn_1): Renamed from...
16543         (final_scan_insn): ... this.  New wrapper, to recover
16544         seen from the outermost call in recursive ones.
16545         * config/sparc/sparc.c (output_return): Drop seen from call.
16546         (output_sibcall): Likewise.
16547         * config/visium/visium.c (output_branch): Likewise.
16549 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
16551         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16552         function label.
16554 2018-02-10  Alan Modra  <amodra@gmail.com>
16556         PR target/84300
16557         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16558         Specify LR as an input.
16560 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
16562         PR sanitizer/83987
16563         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16564         remove_member_access_dummy_vars): New functions.
16565         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16566         lower_omp_1, execute_lower_omp): Use them.
16568         PR rtl-optimization/84308
16569         * shrink-wrap.c (spread_components): Release todo vector.
16571 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
16573         PR rtl-optimization/57193
16574         * ira-color.c (struct allocno_color_data): Add member
16575         conflict_allocno_hard_prefs.
16576         (update_conflict_allocno_hard_prefs): New.
16577         (bucket_allocno_compare_func): Add a preference based on
16578         conflict_allocno_hard_prefs.
16579         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16580         (color_allocnos): Remove a dead code.  Initiate
16581         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
16583 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
16585         PR target/84226
16586         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16587         constraint from =wa to wa.  Avoid a subreg on the output operand,
16588         instead use a pseudo and subreg it in a move.
16589         (p9_xxbrd_<mode>): Changed to ...
16590         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16591         (p9_xxbrd_v2df): New expander.
16592         (p9_xxbrw_<mode>): Changed to ...
16593         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16594         (p9_xxbrw_v4sf): New expander.
16596 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
16598         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16600 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
16602         PR target/83926
16603         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16604         multiply in 32-bit mode.
16605         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16606         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16607         mode.
16609 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
16611         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16612         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16613         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16614         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16616 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
16618         PR lto/84213
16619         * dwarf2out.c (is_trivial_indirect_ref): New function.
16620         (dwarf2out_late_global_decl): Do not generate a location
16621         attribute for variables that have a non-trivial DECL_VALUE_EXPR
16622         and that are not defined in the current unit.
16624 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
16626         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16627         instead of a libcall for UNORDERED.
16629 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
16631         PR target/82641
16632         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16633         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16635 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
16637         PR target/PR84295
16638         * config/s390/s390.c (s390_set_current_function): Invoke
16639         s390_indirect_branch_settings also if fndecl didn't change.
16641 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
16643         * config/rs6000/rs6000.md (blockage): Set length to zero.
16645 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
16647         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16649 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
16651         PR sanitizer/84285
16652         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16653         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16654         -static-lib*san.
16656         PR debug/84252
16657         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16658         PARALLEL incoming that failed vt_get_decl_and_offset check.
16660         PR middle-end/84237
16661         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16662         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16663         TREE_READONLY bit.
16664         (get_variable_section): For decls in named .bss* sections pass true as
16665         second argument to bss_initializer_p.
16667 2018-02-09  Marek Polacek  <polacek@redhat.com>
16668             Jakub Jelinek  <jakub@redhat.com>
16670         PR c++/83659
16671         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16672         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
16673         Sync some changes from cxx_fold_indirect_ref.
16675 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
16677         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16678         markers.
16679         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16680         (BLOCK_INLINE_ENTRY_LABEL): New.
16681         (dwarf2out_var_location): Disregard inline entry markers.
16682         (inline_entry_data): New struct.
16683         (inline_entry_data_hasher): New hashtable type.
16684         (inline_entry_data_hasher::hash): New.
16685         (inline_entry_data_hasher::equal): New.
16686         (inline_entry_data_table): New variable.
16687         (add_high_low_attributes): Add DW_AT_entry_pc and
16688         DW_AT_GNU_entry_view attributes if a pending entry is found
16689         in inline_entry_data_table.  Add old entry_pc attribute only
16690         if debug nonbinding markers are disabled.
16691         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16692         markers are enabled.
16693         (block_within_block_p, dwarf2out_inline_entry): New.
16694         (dwarf2out_finish): Check that no entries remained in
16695         inline_entry_data_table.
16696         * final.c (reemit_insn_block_notes): Handle inline entry notes.
16697         (final_scan_insn, notice_source_line): Likewise.
16698         (rest_of_clean_state): Skip inline entry markers.
16699         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16700         markers.
16701         * gimple.c (gimple_build_debug_inline_entry): New.
16702         * gimple.h (enum gimple_debug_subcode): Add
16703         GIMPLE_DEBUG_INLINE_ENTRY.
16704         (gimple_build_debug_inline_entry): Declare.
16705         (gimple_debug_inline_entry_p): New.
16706         (gimple_debug_nonbind_marker_p): Adjust.
16707         * insn-notes.def (INLINE_ENTRY): New.
16708         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16709         inline entry marker notes.
16710         (print_insn): Likewise.
16711         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16712         (INSN_DEBUG_MARKER_KIND): Likewise.
16713         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16714         * tree-inline.c (expand_call_inline): Build and insert
16715         debug_inline_entry stmt.
16716         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16717         inline entry blocks early, if nonbind markers are enabled.
16718         (dump_scope_block): Dump fragment info.
16719         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16720         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16721         (gimple_build_debug_inline_entry): New.
16722         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16723         Enable/disable inline entry points too.
16724         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16725         (DEBUG_INSN): Describe inline entry markers.
16727         * common.opt (gvariable-location-views): New.
16728         (gvariable-location-views=incompat5): New.
16729         * config.in: Rebuilt.
16730         * configure: Rebuilt.
16731         * configure.ac: Test assembler for view support.
16732         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16733         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16734         * dwarf2out.c (var_loc_view): New typedef.
16735         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16736         (dwarf2out_locviews_in_attribute): New.
16737         (dwarf2out_locviews_in_loclist): New.
16738         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16739         (enum dw_line_info_opcode): Add LI_adv_address.
16740         (struct dw_line_info_table): Add view.
16741         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16742         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16743         (zero_view_p): New variable.
16744         (ZERO_VIEW_P): New macro.
16745         (output_asm_line_debug_info): New.
16746         (struct var_loc_node): Add view.
16747         (add_AT_view_list, AT_loc_list): New.
16748         (add_var_loc_to_decl): Add view param.  Test it against last.
16749         (new_loc_list): Add view params.  Record them.
16750         (AT_loc_list_ptr): Handle loc and view lists.
16751         (view_list_to_loc_list_val_node): New.
16752         (print_dw_val): Handle dw_val_class_view_list.
16753         (size_of_die): Likewise.
16754         (value_format): Likewise.
16755         (loc_list_has_views): New.
16756         (gen_llsym): Set vl_symbol too.
16757         (maybe_gen_llsym, skip_loc_list_entry): New.
16758         (dwarf2out_maybe_output_loclist_view_pair): New.
16759         (output_loc_list): Output view list or entries too.
16760         (output_view_list_offset): New.
16761         (output_die): Handle dw_val_class_view_list.
16762         (output_dwarf_version): New.
16763         (output_compilation_unit_header): Use it.
16764         (output_skeleton_debug_sections): Likewise.
16765         (output_rnglists, output_line_info): Likewise.
16766         (output_pubnames, output_aranges): Update version comments.
16767         (output_one_line_info_table): Output view numbers in asm comments.
16768         (dw_loc_list): Determine current endview, pass it to new_loc_list.
16769         Call maybe_gen_llsym.
16770         (loc_list_from_tree_1): Adjust.
16771         (add_AT_location_description): Create view list attribute if
16772         needed, check it's absent otherwise.
16773         (convert_cfa_to_fb_loc_list): Adjust.
16774         (maybe_emit_file): Call output_asm_line_debug_info for test.
16775         (dwarf2out_var_location): Reset views as needed.  Precompute
16776         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
16777         attribute.  Set view.
16778         (new_line_info_table): Reset next view.
16779         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16780         (dwarf2out_source_line): Likewise.  Output view resets and labels to
16781         the assembler, or select appropriate line info opcodes.
16782         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16783         (optimize_string_length): Catch it.  Adjust.
16784         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
16785         dw_val_class_view_list, and remove it if no longer needed.
16786         (hash_loc_list): Hash view numbers.
16787         (loc_list_hasher::equal): Compare them.
16788         (optimize_location_lists): Check whether a view list symbol is
16789         needed, and whether the locview attribute is present, and
16790         whether they match.  Remove the locview attribute if no longer
16791         needed.
16792         (index_location_lists): Call skip_loc_list_entry for test.
16793         (dwarf2out_finish): Call output_asm_line_debug_info for test.
16794         Use output_dwarf_version.
16795         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16796         (struct dw_val_node): Add val_view_list.
16797         * final.c (SEEN_NEXT_VIEW): New.
16798         (set_next_view_needed): New.
16799         (clear_next_view_needed): New.
16800         (maybe_output_next_view): New.
16801         (final_start_function): Rename to...
16802         (final_start_function_1): ... this.  Take pointer to FIRST,
16803         add SEEN parameter.  Emit param bindings in the initial view.
16804         (final_start_function): Reintroduce SEEN-less interface.
16805         (final): Rename to...
16806         (final_1): ... this.  Take SEEN parameter.  Output final pending
16807         next view at the end.
16808         (final): Reintroduce seen-less interface.
16809         (final_scan_insn): Output pending next view before switching
16810         sections or ending a block.  Mark the next view as needed when
16811         outputting variable locations.  Notify debug backend of section
16812         changes, and of location view changes.
16813         (rest_of_handle_final): Adjust.
16814         * toplev.c (process_options): Autodetect value for debug variable
16815         location views option.  Warn on incompat5 without -gdwarf-5.
16816         * doc/invoke.texi (gvariable-location-views): New.
16817         (gvariable-location-views=incompat5): New.
16818         (gno-variable-location-views): New.
16820 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
16822         PR tree-optimization/84136
16823         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16824         that the result of find_edge is non-NULL.
16826 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
16828         PR target/83008
16829         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16830         storing integer register in SImode.  Fix cost of 256 and 512
16831         byte aligned SSE register store.
16833 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
16835         * config/i386/i386.c (ix86_multiplication_cost): Fix
16836         multiplication cost for TARGET_AVX512DQ.
16838 2018-02-08  Marek Polacek  <polacek@redhat.com>
16840         PR tree-optimization/84238
16841         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16842         get_range_strlen.
16844 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
16846         PR tree-optimization/84265
16847         * tree-vect-stmts.c (vectorizable_store): Don't treat
16848         VMAT_CONTIGUOUS accesses as grouped.
16849         (vectorizable_load): Likewise.
16851 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
16853         PR tree-optimization/81635
16854         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16855         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16856         (test_round_for_mask): New functions.
16857         (wide_int_cc_tests): Call test_round_for_mask.
16858         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16859         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16860         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16861         range returned by get_range_info.
16863 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
16865         PR ipa/81360
16866         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16867         * symtab.c: Include builtins.h
16868         (symtab_node::output_to_lto_symbol_table_p): Move here
16869         from lto-streamer-out.c:output_symbol_p.
16870         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16871         (output_symbol_p): Move all logic to symtab.c
16872         (produce_symtab): Update.
16874 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
16876         * config/s390/s390-opts.h (enum indirect_branch): Define.
16877         * config/s390/s390-protos.h (s390_return_addr_from_memory)
16878         (s390_indirect_branch_via_thunk)
16879         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16880         (enum s390_indirect_branch_type): Define.
16881         * config/s390/s390.c (struct s390_frame_layout, struct
16882         machine_function): Remove.
16883         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16884         (indirect_branch_table_label_no, indirect_branch_table_name):
16885         Define variables.
16886         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16887         (enum s390_indirect_branch_option): Define.
16888         (s390_return_addr_from_memory): New function.
16889         (s390_handle_string_attribute): New function.
16890         (s390_attribute_table): Add new attribute handler.
16891         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
16892         (s390_indirect_branch_via_thunk): New function.
16893         (s390_indirect_branch_via_inline_thunk): New function.
16894         (s390_function_ok_for_sibcall): When jumping via thunk disallow
16895         sibling call optimization for non z10 compiles.
16896         (s390_emit_call): Force indirect branch target to be a single
16897         register.  Add r1 clobber for non-z10 compiles.
16898         (s390_emit_epilogue): Emit return jump via return_use expander.
16899         (s390_reorg): Handle JUMP_INSNs as execute targets.
16900         (s390_option_override_internal): Perform validity checks for the
16901         new command line options.
16902         (s390_indirect_branch_attrvalue): New function.
16903         (s390_indirect_branch_settings): New function.
16904         (s390_set_current_function): Invoke s390_indirect_branch_settings.
16905         (s390_output_indirect_thunk_function):  New function.
16906         (s390_code_end): Implement target hook.
16907         (s390_case_values_threshold): Implement target hook.
16908         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16909         macros.
16910         * config/s390/s390.h (struct s390_frame_layout)
16911         (struct machine_function): Move here from s390.c.
16912         (TARGET_INDIRECT_BRANCH_NOBP_RET)
16913         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16914         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16915         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16916         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16917         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16918         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16919         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16920         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16921         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
16922         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
16923         (mnemonic attribute): Add values which aren't recognized
16924         automatically.
16925         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
16926         pattern for branch conversion.  Fix mnemonic attribute.
16927         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
16928         indirect branch via thunk if requested.
16929         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
16930         ("*indirect_jump"): Disable for branch conversion using out of
16931         line thunks.
16932         ("indirect_jump_via_thunk<mode>_z10")
16933         ("indirect_jump_via_thunk<mode>")
16934         ("indirect_jump_via_inlinethunk<mode>_z10")
16935         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
16936         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
16937         ("casesi_jump_via_inlinethunk<mode>_z10")
16938         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
16939         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
16940         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
16941         ("*indirect2_jump"): Disable for branch conversion.
16942         ("casesi_jump"): Turn into expander and expand patterns for branch
16943         conversion.
16944         ("return_use"): New expander.
16945         ("*return"): Emit return via thunk and rename it to ...
16946         ("*return<mode>"): ... this one.
16947         * config/s390/s390.opt: Add new options and and enum for the
16948         option values.
16950 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
16952         * lra-constraints.c (match_reload): Unconditionally use
16953         gen_lowpart_SUBREG, rather than selecting between that
16954         and equivalent gen_rtx_SUBREG code.
16956 2018-02-08  Richard Biener  <rguenther@suse.de>
16958         PR tree-optimization/84233
16959         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
16960         changed flag instead of boguously re-using phi_inserted.
16962 2018-02-08  Martin Jambor  <mjambor@suse.cz>
16964         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
16965         static local variables.
16967 2018-02-08  Richard Biener  <rguenther@suse.de>
16969         PR tree-optimization/84278
16970         * tree-vect-stmts.c (vectorizable_store): When looking for
16971         smaller vector types to perform grouped strided loads/stores
16972         make sure the mode is supported by the target.
16973         (vectorizable_load): Likewise.
16975 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
16977         * config/aarch64/aarch64.c (aarch64_components_for_bb):
16978         Increase LDP/STP opportunities by adding adjacent callee-saves.
16980 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
16982         PR rtl-optimization/84068
16983         PR rtl-optimization/83459
16984         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
16986 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
16988         PR tree-optimization/84224
16989         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
16990         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
16991         non-zero arguments.
16993 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
16995         PR target/84113
16996         * config/rs6000/altivec.md (*restore_world): Remove LR use.
16997         * config/rs6000/predicates.md (restore_world_operation): Adjust op
16998         count, remove one USE.
17000 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
17002         * doc/install.texi (Configuration): Document the
17003         --with-long-double-format={ibm,ieee} PowerPC configuration
17004         options.
17006         PR target/84154
17007         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17008         Convert from define_expand to be define_insn_and_split.  Rework
17009         float/double/_Float128 conversions to QI/HI/SImode to work with
17010         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
17011         conversions to QI/HImode types did a store and then a load to
17012         truncate the value.  For conversions to VSX registers, don't split
17013         the insn, instead emit the code directly.  Use the code iterator
17014         any_fix to combine signed and unsigned conversions.
17015         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17016         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17017         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17018         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17019         (fix_<mode>di2_hw): Likewise.
17020         (fixuns_<mode>di2_hw): Likewise.
17021         (fix_<mode>si2_hw): Likewise.
17022         (fixuns_<mode>si2_hw): Likewise.
17023         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17024         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17025         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17026         fix<uns>_trunc<SFDF:mode>si2_p8.
17027         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17028         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17029         (fix<uns>_<mode>_mem): Likewise.
17030         (fctiw<u>z_<mode>_mem): Likewise.
17031         (fix<uns>_<mode>_mem): Likewise.
17032         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17033         the register allocator from doing a direct move to the GPRs to do
17034         a store, and instead use the ISA 3.0 store byte/half-word from
17035         vector register instruction.  For IEEE 128-bit floating point,
17036         also optimize stores of 32-bit ints.
17037         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17039 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
17041         * genextract.c (push_pathstr_operand): New function to support
17042         [a-zA-Z].
17043         (walk_rtx): Call push_pathstr_operand.
17044         (print_path): Support [a-zA-Z].
17046 2018-02-07  Richard Biener  <rguenther@suse.de>
17048         PR tree-optimization/84037
17049         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17050         (cse_and_gimplify_to_preheader): Declare.
17051         (vect_get_place_in_interleaving_chain): Likewise.
17052         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17053         ivexpr_map.
17054         (_loop_vec_info::~_loop_vec_info): Delete it.
17055         (cse_and_gimplify_to_preheader): New function.
17056         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17057         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17058         (vectorizable_load): Likewise.  For grouped stores always base
17059         the IV on the first element.
17060         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17061         condition before gimplifying.
17063 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
17065         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17066         *DIV_EXPR and *MOD_EXPR.
17068 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
17070         PR target/84248
17071         * config/i386/i386.c (ix86_option_override_internal): Mask out
17072         the CF_SET bit when checking -fcf-protection.
17074 2018-02-07  Tom de Vries  <tom@codesourcery.com>
17076         PR libgomp/84217
17077         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17078         enough.
17080 2018-02-07  Richard Biener  <rguenther@suse.de>
17082         PR tree-optimization/84204
17083         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17084         this place.
17086         PR tree-optimization/84205
17087         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17088         special-case isl_ast_op_zdiv_r.
17090         PR tree-optimization/84223
17091         * graphite-scop-detection.c (gather_bbs::before_dom_children):
17092         Only add conditions from within the region.
17093         (gather_bbs::after_dom_children): Adjust.
17095 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
17097         PR target/84209
17098         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17099         * config/avr/avr.md: Only post-reload split REG-REG moves if
17100         either register is GENERAL_REG_P.
17102 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
17104         PR tree-optimization/84235
17105         * tree-ssa-scopedtables.c
17106         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17107         if the subtraction is performed in floating point type where NaNs are
17108         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17109         build 1.  Formatting fix.
17111 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
17113         PR target/84146
17114         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17115         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17116         and skip it regardless of bb boundaries.  Use CALL_P macro,
17117         don't test INSN_P (insn) together with CALL_P or JUMP_P check
17118         unnecessarily, formatting fix.
17120 2018-02-06  Michael Collison  <michael.collison@arm.com>
17122         * config/arm/thumb2.md:
17123         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17124         (*thumb_mov_notscc): Ditto.
17126 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
17128         PR target/84154
17129         * config/rs6000/rs6000.md (su code attribute): Use "u" for
17130         unsigned_fix, not "s".
17132 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17134         * configure.ac (gcc_fn_eh_frame_ro): New function.
17135         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17136         correct .eh_frame permissions.
17137         * configure: Regenerate.
17139 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
17141         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17142         irrelevant options.
17144 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17146         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17147         Display warning message for -mno-speculate-indirect-jumps.
17149 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
17151         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17152         Undocumented.
17153         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17155 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
17157         PR tree-optimization/84225
17158         * tree-eh.c (find_trapping_overflow): Only call
17159         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17161 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
17163         PR target/84145
17164         * config/i386/i386.c: Reimplement the check of possible options
17165         -mibt/-mshstk conbination. Change error messages.
17166         * doc/invoke.texi: Fix a typo: remove extra '='.
17168 2018-02-06  Marek Polacek  <polacek@redhat.com>
17170         PR tree-optimization/84228
17171         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17173 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
17175         PR target/82641
17176         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17177         emitted arch directives.
17178         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17179         __ARM_FEATURE_COPROC before changing architectures.
17181 2018-02-06  Richard Biener  <rguenther@suse.de>
17183         * config/i386/i386.c (print_reg): Fix typo.
17184         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17186 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
17188         * configure: Regenerate.
17190 2018-02-05  Martin Sebor  <msebor@redhat.com>
17192         PR tree-optimization/83369
17193         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17194         inlining context.
17196 2018-02-05  Martin Liska  <mliska@suse.cz>
17198         * doc/invoke.texi: Cherry-pick upstream r323995.
17200 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
17202         * ira.c (ira_init_register_move_cost): Adjust comment.
17204 2018-02-05  Martin Liska  <mliska@suse.cz>
17206         PR gcov-profile/84137
17207         * doc/gcov.texi: Fix typo in documentation.
17209 2018-02-05  Martin Liska  <mliska@suse.cz>
17211         PR gcov-profile/83879
17212         * doc/gcov.texi: Document necessity of --dynamic-list-data when
17213         using dlopen functionality.
17215 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
17217         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17218         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17219         _mm_maskz_range_ss, _mm_mask_range_round_ss,
17220         _mm_maskz_range_round_ss): New intrinsics.
17221         (__builtin_ia32_rangesd128_round)
17222         (__builtin_ia32_rangess128_round): Remove.
17223         (__builtin_ia32_rangesd128_mask_round,
17224         __builtin_ia32_rangess128_mask_round): New builtins.
17225         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17226         __builtin_ia32_rangess128_round): Remove.
17227         (__builtin_ia32_rangesd128_mask_round,
17228         __builtin_ia32_rangess128_mask_round): New builtins.
17229         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17230         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17231         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17232         "<round_saeonly_constraint>")): Changed to ...
17233         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17234         "<round_saeonly_scalar_constraint>")): ... this.
17235         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17236         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17237         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17238         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17239         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17241 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
17243         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17244         options.
17245         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17246         Remove all values except native, 8540 and 8548.
17248 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
17250         * config/i386/i386.c (ix86_output_function_return): Pass
17251         INVALID_REGNUM, instead of -1, as invalid register number to
17252         indirect_thunk_name and output_indirect_thunk.
17254 2018-02-02  Julia Koval  <julia.koval@intel.com>
17256         * config.gcc: Add -march=icelake.
17257         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17258         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17259         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17260         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17261         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17262         (processor_target_table): Add icelake.
17263         (ix86_option_override_internal): Handle new PTAs.
17264         (get_builtin_code_for_version): Handle icelake.
17265         (M_INTEL_COREI7_ICELAKE): New.
17266         (fold_builtin_cpu): Handle icelake.
17267         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17268         * doc/invoke.texi: Add -march=icelake.
17270 2018-02-02  Julia Koval  <julia.koval@intel.com>
17272         * config/i386/i386.c (ix86_option_override_internal): Change flags type
17273         to wide_int_bitmask.
17274         * wide-int-bitmask.h: New.
17276 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
17278         PR target/84066
17279         * config/i386/i386.md: Replace Pmode with word_mode in
17280         builtin_setjmp_setup and builtin_longjmp to support x32.
17282 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
17284         PR target/56010
17285         PR target/83743
17286         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17287         #include "opts.h".
17288         (rs6000_supported_cpu_names): New static variable.
17289         (linux_cpu_translation_table): Likewise.
17290         (elf_platform) <cpu>: Define new static variable and use it.
17291         Translate kernel AT_PLATFORM name to canonical name if needed.
17292         Error if platform name is unknown.
17294 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
17296         PR target/84089
17297         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17299 2018-02-01  Jeff Law  <law@redhat.com>
17301         PR target/84128
17302         * config/i386/i386.c (release_scratch_register_on_entry): Add new
17303         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
17304         the scratch if RELEASE_VIA_POP is false.
17305         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17306         If we have to save a temporary register, decrement SIZE appropriately.
17307         Pass new arguments to release_scratch_register_on_entry.
17308         (ix86_adjust_stack_and_probe): Likewise.
17309         (ix86_emit_probe_stack_range): Pass new arguments to
17310         release_scratch_register_on_entry.
17312 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
17314         PR rtl-optimization/84157
17315         * combine.c (change_zero_ext): Use REG_P predicate in
17316         front of HARD_REGISTER_P predicate.
17318 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
17320         * config/avr/avr.c (avr_option_override): Move disabling of
17321         -fdelete-null-pointer-checks to...
17322         * common/config/avr/avr-common.c (avr_option_optimization_table):
17323         ...here.
17325 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17327         PR tree-optimization/81635
17328         * tree-data-ref.c (split_constant_offset_1): For types that
17329         wrap on overflow, try to use range info to prove that wrapping
17330         cannot occur.
17332 2018-02-01  Renlin Li  <renlin.li@arm.com>
17334         PR target/83370
17335         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17336         TAILCALL_ADDR_REGS.
17337         (aarch64_register_move_cost): Likewise.
17338         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17339         TAILCALL_ADDR_REGS.
17340         (REG_CLASS_NAMES): Likewise.
17341         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17342         TAILCALL_ADDR_REGS. Remove IP registers.
17343         * config/aarch64/aarch64.md (Ucs): Update register constraint.
17345 2018-02-01  Richard Biener  <rguenther@suse.de>
17347         * domwalk.h (dom_walker::dom_walker): Add additional constructor
17348         for specifying RPO order and allow NULL for that.
17349         * domwalk.c (dom_walker::dom_walker): Likewise.
17350         (dom_walker::walk): Handle NULL RPO order.
17351         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17352         in RPO order.
17353         (rewrite_update_dom_walker): Likewise.
17354         (mark_def_dom_walker): Likewise.
17356 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17358         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17359         (aarch64_maybe_expand_sve_subreg_move): Declare.
17360         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17361         * config/aarch64/predicates.md (aarch64_any_register_operand): New
17362         predicate.
17363         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17364         that are semantically a reverse operation.
17365         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17366         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17367         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17368         functions.
17369         (aarch64_can_change_mode_class): For big-endian, forbid changes
17370         between two SVE modes if they have different element sizes.
17372 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17374         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17375         the TImode handling for big-endian targets.
17377 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17379         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17380         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
17381         not just bytes.
17382         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17383         Remove BSWAP handing for big-endian targets and use the form of
17384         LD1RQ appropariate for the mode.
17386 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17388         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17389         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17390         duplicated element.
17392 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
17394         PR tearget/83845
17395         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17396         check for operands that need to go through aarch64_sve_reload_be.
17398 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
17400         PR tree-optimization/81661
17401         PR tree-optimization/84117
17402         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17403         * tree-eh.c: Include gimplify.h.
17404         (find_trapping_overflow, replace_trapping_overflow,
17405         rewrite_to_non_trapping_overflow): New functions.
17406         * tree-vect-loop.c: Include tree-eh.h.
17407         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17408         * tree-data-ref.c: Include tree-eh.h.
17409         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17411 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
17413         PR rtl-optimization/84123
17414         * combine.c (change_zero_ext): Check if hard register satisfies
17415         can_change_dest_mode before calling gen_lowpart_SUBREG.
17417 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
17419         PR target/82444
17420         * ira.c (ira_init_register_move_cost): Remove assert.
17422 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
17424         PR rtl-optimization/84071
17425         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17426         * doc/tm.texi: Regenerate.
17428 2018-01-31  Richard Biener  <rguenther@suse.de>
17430         PR tree-optimization/84132
17431         * tree-data-ref.c (analyze_miv_subscript): Properly
17432         check whether evolution_function_is_affine_multivariate_p
17433         before calling gcd_of_steps_may_divide_p.
17435 2018-01-31  Julia Koval  <julia.koval@intel.com>
17437         PR target/83618
17438         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17439         * config/i386/i386.md (rdpid_rex64) New.
17440         (rdpid): Make 32bit only.
17442 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
17444         PR lto/84105
17445         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17446         an IDENTIFIER_NODE for FUNCTION_TYPE's.
17448 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
17450         Revert
17451         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
17453         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17455 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
17457         PR rtl-optimization/84071
17458         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17459         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17461 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
17463         * config/arc/arc.c (arc_handle_aux_attribute): New function.
17464         (arc_attribute_table): Add 'aux' attribute.
17465         (arc_in_small_data_p): Consider aux like variables.
17466         (arc_is_aux_reg_p): New function.
17467         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17468         (arc_get_aux_arg): New function.
17469         (prepare_move_operands): Handle aux-register access.
17470         (arc_handle_aux_attribute): New function.
17471         * doc/extend.texi (ARC Variable attributes): Add subsection.
17473 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
17475         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17476         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17477         (arc_attribute_table): Add 'uncached' attribute.
17478         (arc_print_operand): Print '.di' flag for uncached memory
17479         accesses.
17480         (arc_in_small_data_p): Do not consider for small data the uncached
17481         types.
17482         (arc_is_uncached_mem_p): New function.
17483         * config/arc/predicates.md (compact_store_memory_operand): Check
17484         for uncached memory accesses.
17485         (nonvol_nonimm_operand): Likewise.
17486         * doc/extend.texi (ARC Type Attribute): New subsection.
17488 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
17490         PR c/84100
17491         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17492         falign-loops=): Add Optimization flag.
17494 2018-01-30  Jeff Law  <law@redhat.com>
17496         PR target/84064
17497         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17498         INT_REGISTERS_SAVED.  Check it prior to calling
17499         get_scratch_register_on_entry.
17500         (ix86_adjust_stack_and_probe): Similarly.
17501         (ix86_emit_probe_stack_range): Similarly.
17502         (ix86_expand_prologue): Corresponding changes.
17504 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17506         PR target/40411
17507         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17508         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17510 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
17512         PR target/84112
17513         * lra-constraints.c (curr_insn_transform): Process AND in the
17514         address.
17516 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
17518         PR rtl-optimization/83986
17519         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17520         dependence against last_pending_memory_flush in addition to
17521         pending_jump_insns.
17523 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
17525         PR tree-optimization/81611
17526         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17527         copies.
17529 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17531         PR target/83758
17532         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17533         a reg rtx.
17535 2018-01-30  Richard Biener  <rguenther@suse.de>
17536             Jakub Jelinek  <jakub@redhat.com>
17538         PR tree-optimization/84111
17539         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17540         inner loops added during recursion, as they don't have up-to-date
17541         SSA form.
17543 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
17545         PR ipa/81360
17546         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17547         (can_inline_edge_by_limits_p): ... here.
17548         (can_early_inline_edge_p, check_callers,
17549         update_caller_keys, update_callee_keys, recursive_inlining,
17550         add_new_edges_to_heap, speculation_useful_p,
17551         inline_small_functions,
17552         inline_small_functions, flatten_function,
17553         inline_to_all_callers_1): Update.
17555 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
17557         * profile-count.c (profile_count::combine_with_ipa_count): Handle
17558         zeros correctly.
17560 2018-01-30  Richard Biener  <rguenther@suse.de>
17562         PR tree-optimization/83008
17563         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17564         invariant and constant vector uses in stmts when they need
17565         more than one stmt.
17567 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17569         PR bootstrap/84017
17570         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17571         * configure: Regenerate.
17573 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
17575         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17576         pattern.
17577         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17578         Use gen_rtx_REG rather than gen_lowpart.
17580 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
17582         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17583         rather than 0 when creating partial subregs.
17585 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
17587         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17588         of usage.
17590 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
17592         PR target/81550
17593         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17594         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17595         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17596         flags.  This restores the settings used before the 2017-07-24.
17597         Turning off pre increment/decrement/modify allows IVOPTS to
17598         optimize DF/SF loops where the index is an int.
17600 2018-01-29  Richard Biener  <rguenther@suse.de>
17601             Kelvin Nilsen  <kelvin@gcc.gnu.org>
17603         PR bootstrap/80867
17604         * tree-vect-stmts.c (vectorizable_call): Don't call
17605         targetm.vectorize_builtin_md_vectorized_function if callee is
17606         NULL.
17608 2018-01-22  Carl Love  <cel@us.ibm.com>
17610         * doc/extend.tex: Fix typo in second arg in
17611         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17613 2018-01-29  Richard Biener  <rguenther@suse.de>
17615         PR tree-optimization/84086
17616         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17617         (flush_ssaname_freelist): When SSA names were released reset
17618         the SCEV hash table.
17620 2018-01-29  Richard Biener  <rguenther@suse.de>
17622         PR tree-optimization/84057
17623         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17624         removed paths when removing edges.
17626 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
17628         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17629         -mfunction-return=@var{choice}.
17631 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17633         PR diagnostic/84034
17634         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17635         Handle CR like TAB.
17636         (layout::print_source_line): Likewise.
17637         (test_get_line_width_without_trailing_whitespace): Add test cases.
17639 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
17641         PR middle-end/84040
17642         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17643         debug insns.
17645 2018-01-26  Jim Wilson  <jimw@sifive.com>
17647         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17649         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17650         specified.
17652 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17654         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17655         and CMP + SUB-immediate -> SUBS.
17657 2018-01-26  Martin Sebor  <msebor@redhat.com>
17659         PR tree-optimization/83896
17660         * tree-ssa-strlen.c (get_string_len): Rename...
17661         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
17662         Avoid assuming length is constant.
17663         (handle_char_store): Use HOST_WIDE_INT for string length.
17665 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
17667         PR target/81763
17668         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17669         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17671 2018-01-26  Richard Biener  <rguenther@suse.de>
17673         PR rtl-optimization/84003
17674         * dse.c (record_store): Only record redundant stores when
17675         the earlier store aliases at least all accesses the later one does.
17677 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
17679         PR rtl-optimization/83985
17680         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17681         REG_CFA_RESTORE insns.
17682         (delete_unmarked_insns): Don't ignore separate shrink wrapping
17683         REG_CFA_RESTORE insns here.
17685         PR c/83989
17686         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17687         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17689 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
17691         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17692         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17693         (arc_init): Likewise.
17694         (arc_override_options): Likewise.
17695         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17696         value.
17697         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17698         support.
17699         * config/arc/arc.h (TARGET_DBNZ): Define.
17700         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17701         properly set the tune attribute.
17702         (dbnz): Use TARGET_DBNZ guard.
17703         * config/arc/arc.opt (mtune): Add core3 option.
17705 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
17707         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17708         recognize new pic like addresses.
17709         (arc_delegitimize_address): Clean up.
17711 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
17713         * config/arc/arc-arches.def: Option mrf16 valid for all
17714         architectures.
17715         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17716         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17717         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17718         * config/arc/arc-tables.opt: Regenerate.
17719         * config/arc/arc.c (arc_conditional_register_usage): Handle
17720         reduced register file case.
17721         (arc_file_start): Set must have build attributes.
17722         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17723         mrf16 option value.
17724         * config/arc/arc.opt (mrf16): Add new option.
17725         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17726         * config/arc/genmultilib.awk: Handle new mrf16 option.
17727         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17728         * config/arc/t-multilib: Regenerate.
17729         * doc/invoke.texi (ARC Options): Document mrf16 option.
17731 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
17733         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17734         * config/arc/arc.c (arc_handle_secure_attribute): New function.
17735         (arc_attribute_table): Add 'secure_call' attribute.
17736         (arc_print_operand): Print secure call operand.
17737         (arc_function_ok_for_sibcall): Don't optimize tail calls when
17738         secure.
17739         (arc_is_secure_call_p): New function.  * config/arc/arc.md
17740         (call_i): Add support for sjli instruction.
17741         (call_value_i): Likewise.
17742         * config/arc/constraints.md (Csc): New constraint.
17744 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
17745             John Eric Martin  <John.Martin@emmicro-us.com>
17747         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17748         * config/arc/arc.c (_arc_jli_section): New struct.
17749         (arc_jli_section): New type.
17750         (rc_jli_sections): New static variable.
17751         (arc_handle_jli_attribute): New function.
17752         (arc_attribute_table): Add jli_always and jli_fixed attribute.
17753         (arc_file_end): New function.
17754         (TARGET_ASM_FILE_END): Define.
17755         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17756         (arc_add_jli_section): New function.
17757         (jli_call_scan): Likewise.
17758         (arc_reorg): Call jli_call_scan.
17759         (arc_output_addsi): Remove 'S' from printing asm operand.
17760         (arc_is_jli_call_p): New function.
17761         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17762         operand.
17763         (movhi_insn): Likewise.
17764         (movsi_insn): Likewise.
17765         (movsi_set_cc_insn): Likewise.
17766         (loadqi_update): Likewise.
17767         (load_zeroextendqisi_update): Likewise.
17768         (load_signextendqisi_update): Likewise.
17769         (loadhi_update): Likewise.
17770         (load_zeroextendhisi_update): Likewise.
17771         (load_signextendhisi_update): Likewise.
17772         (loadsi_update): Likewise.
17773         (loadsf_update): Likewise.
17774         (movsicc_insn): Likewise.
17775         (bset_insn): Likewise.
17776         (bxor_insn): Likewise.
17777         (bclr_insn): Likewise.
17778         (bmsk_insn): Likewise.
17779         (bicsi3_insn): Likewise.
17780         (cmpsi_cc_c_insn): Likewise.
17781         (movsi_ne): Likewise.
17782         (movsi_cond_exec): Likewise.
17783         (clrsbsi2): Likewise.
17784         (norm_f): Likewise.
17785         (normw): Likewise.
17786         (swap): Likewise.
17787         (divaw): Likewise.
17788         (flag): Likewise.
17789         (sr): Likewise.
17790         (kflag): Likewise.
17791         (ffs): Likewise.
17792         (ffs_f): Likewise.
17793         (fls): Likewise.
17794         (call_i): Remove 'S' asm letter, add jli instruction.
17795         (call_value_i): Likewise.
17796         * config/arc/arc.op (mjli-always): New option.
17797         * config/arc/constraints.md (Cji): New constraint.
17798         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17799         operand.
17800         (subsf3_fpx): Likewise.
17801         (mulsf3_fpx): Likewise.
17802         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17803         asm operand.
17804         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17805         function attrbutes.
17806         * doc/invoke.texi (ARC): Document mjli-always option.
17808 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
17810         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17811         avoid addition with 0 and use incw and decw where possible.
17813 2018-01-26  Richard Biener  <rguenther@suse.de>
17815         PR tree-optimization/81082
17816         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17817         association if it requires casting to unsigned.
17818         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17819         from fold_plusminus_mult_expr to catch important cases late when
17820         range info is available.
17822 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17824         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17825         * configure.ac (hidden_linkonce): New test.
17826         * configure: Regenerate.
17827         * config.in: Regenerate.
17829 2018-01-26  Julia Koval  <julia.koval@intel.com>
17831         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17832         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17833         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17834         _mm_mask_bitshuffle_epi64_mask): Fix type.
17835         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17836         USI_FTYPE_V4DI_V4DI_USI): Remove.
17837         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17838         __builtin_ia32_vpshufbitqmb256_mask,
17839         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17840         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17841         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17843 2018-01-26  Alan Modra  <amodra@gmail.com>
17845         PR target/84033
17846         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17847         UNSPEC_VBPERMQ.  Sort other unspecs.
17849 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
17851         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17853 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
17855         PR middle-end/83055
17856         * predict.c (drop_profile): Do not push/pop cfun; update also
17857         node->count.
17858         (handle_missing_profiles): Fix logic looking for zero profiles.
17860 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
17862         PR middle-end/83977
17863         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17864         on functions with #pragma omp declare simd or functions with simd
17865         attribute.
17866         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17867         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17868         Remove trailing \n from warning_at calls.
17870 2018-01-25  Tom de Vries  <tom@codesourcery.com>
17872         PR target/84028
17873         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17874         for neutered workers.
17876 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
17878         PR target/68467
17879         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17880         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17882 2018-01-24  Jeff Law  <law@redhat.com>
17884         PR target/83994
17885         * i386.c (get_probe_interval): Move to earlier point.
17886         (ix86_compute_frame_layout): If -fstack-clash-protection and
17887         the frame is larger than the probe interval, then use pushes
17888         to save registers rather than reg->mem moves.
17889         (ix86_expand_prologue): Remove conditional for int_registers_saved
17890         assertion.
17892 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
17894         PR target/84014
17895         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
17896         min/max for never referenced object.
17898 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
17900         PR middle-end/83977
17901         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
17902         here.
17903         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
17904         attributes from DECL_ATTRIBUTES (decl) without affecting
17905         DECL_ATTRIBUTES (current_function_decl).
17906         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
17907         functions with non-NULL DECL_ABSTRACT_ORIGIN.
17909 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
17911         PR tree-optimization/83979
17912         * fold-const.c (fold_comparison): Use constant_boolean_node
17913         instead of boolean_{true,false}_node.
17915 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
17917         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17918         with zero counts.
17920 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17922         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17923         Simplify the clause that sets the length attribute.
17924         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17925         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
17926         clause that sets the length attribute.
17927         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17929 2018-01-24  Tom de Vries  <tom@codesourcery.com>
17931         PR target/83589
17932         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
17933         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
17934         Add strict parameter.
17935         (prevent_branch_around_nothing): Insert dummy insn between branch to
17936         label and label with no ptx insn inbetween.
17937         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
17939 2018-01-24  Tom de Vries  <tom@codesourcery.com>
17941         PR target/81352
17942         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17943         for neutered threads in warp.
17944         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
17946 2018-01-24  Richard Biener  <rguenther@suse.de>
17948         PR tree-optimization/83176
17949         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
17950         operands.
17952 2018-01-24  Richard Biener  <rguenther@suse.de>
17954         PR tree-optimization/82819
17955         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
17956         code generating pluses that are no-ops in the target precision.
17958 2018-01-24  Richard Biener  <rguenther@suse.de>
17960         PR middle-end/84000
17961         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
17963 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
17965         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
17966         to merge probabilities.
17967         * predict.c (probably_never_executed): Also mark as cold functions
17968         with global 0 profile and guessed local profile.
17969         * profile-count.c (profile_probability::combine_with_count): New
17970         member function.
17971         * profile-count.h (profile_probability::operator*,
17972         profile_probability::operator*=, profile_probability::operator/,
17973         profile_probability::operator/=): Reduce precision to adjusted
17974         and set value to guessed on contradictory divisions.
17975         (profile_probability::combine_with_freq): Remove.
17976         (profile_probability::combine_wiht_count): Declare.
17977         (profile_count::force_nonzero):: Set to adjusted.
17978         (profile_count::probability_in):: Set quality to adjusted.
17979         * tree-ssa-tail-merge.c (replace_block_by): Use
17980         combine_with_count.
17982 2018-01-23  Andrew Waterman  <andrew@sifive.com>
17983             Jim Wilson  <jimw@sifive.com>
17985         * config/riscv/riscv.c (riscv_stack_boundary): New.
17986         (riscv_option_override): Set riscv_stack_boundary.  Handle
17987         riscv_preferred_stack_boundary_arg.
17988         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
17989         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
17990         (STACK_BOUNDARY): Set to riscv_stack_boundary.
17991         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
17992         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
17993         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
17995 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
17997         PR target/83905
17998         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
17999         of struct ix86_frame.
18000         (ix86_expand_epilogue): Likewise.  Add a local variable for
18001         the reg_save_offset field in struct ix86_frame.
18003 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
18005         PR tree-optimization/82604
18006         * tree-loop-distribution.c (enum partition_kind): New enum item
18007         PKIND_PARTIAL_MEMSET.
18008         (partition_builtin_p): Support above new enum item.
18009         (generate_code_for_partition): Ditto.
18010         (compute_access_range): Differentiate cases that equality can be
18011         proven at all loops, the innermost loops or no loops.
18012         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18013         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
18014         (finalize_partitions, distribute_loop): Don't fuse partition of
18015         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18016         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18017         parloop is enabled.
18019 2018-01-23  Martin Liska  <mliska@suse.cz>
18021         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18022         order to ignore the predictor.
18023         (PRED_POLYMORPHIC_CALL): Likewise.
18024         (PRED_RECURSIVE_CALL): Likewise.
18026 2018-01-23  Martin Liska  <mliska@suse.cz>
18028         * tree-profile.c (tree_profiling): Print function header to
18029         aware reader which function we are working on.
18030         * value-prof.c (gimple_find_values_to_profile): Do not print
18031         not interesting value histograms.
18033 2018-01-23  Martin Liska  <mliska@suse.cz>
18035         * profile-count.h (enum profile_quality): Add
18036         profile_uninitialized as the first value. Do not number values
18037         as they are zero based.
18038         (profile_count::verify): Update sanity check.
18039         (profile_probability::verify): Likewise.
18041 2018-01-23  Nathan Sidwell  <nathan@acm.org>
18043         * doc/invoke.texi (ffor-scope): Deprecate.
18045 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
18047         PR tree-optimization/83510
18048         * domwalk.c (set_all_edges_as_executable): New function.
18049         (dom_walker::dom_walker): Convert bool param
18050         "skip_unreachable_blocks" to enum reachability.  Move setup of
18051         edge flags to set_all_edges_as_executable and only do it when
18052         reachability is REACHABLE_BLOCKS.
18053         * domwalk.h (enum dom_walker::reachability): New enum.
18054         (dom_walker::dom_walker): Convert bool param
18055         "skip_unreachable_blocks" to enum reachability.
18056         (set_all_edges_as_executable): New decl.
18057         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
18058         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18059         "reachability".
18060         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18061         but converting true to REACHABLE_BLOCKS.
18062         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18063         * tree-vrp.c
18064         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18065         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18066         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18067         REACHABLE_BLOCKS.
18068         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18069         if check_all_array_refs will be called.
18071 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
18073         * tree.c (selftest::test_location_wrappers): Add more test
18074         coverage.
18076 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
18078         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18079         (selftest::test_bit_in_range): Likewise.
18081 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
18083         PR testsuite/83888
18084         * doc/sourcebuild.texi (vect_float): Say that the selector
18085         only describes the situation when -funsafe-math-optimizations is on.
18086         (vect_float_strict): Document.
18088 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
18090         PR tree-optimization/83965
18091         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18092         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18093         instead of checking only for a reduction.
18094         (vect_recog_widen_sum_pattern): Likewise.
18096 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
18098         * predict.c (probably_never_executed): Only use precise profile info.
18099         (compute_function_frequency): Skip after inlining hack since we now
18100         have quality checking.
18102 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
18104         * profile-count.h (profile_probability::very_unlikely,
18105         profile_probability::unlikely, profile_probability::even): Set
18106         precision to guessed.
18108 2018-01-23  Richard Biener  <rguenther@suse.de>
18110         PR tree-optimization/83963
18111         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18112         Properly terminate dominator walk when crossing the exit edge not
18113         when visiting its source block.
18115 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
18117         PR c++/83918
18118         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18119         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18121 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
18123         PR tree-optimization/83957
18124         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
18125         semicolon after for body surrounded by braces.
18127         PR tree-optimization/83081
18128         * profile-count.h (profile_probability::split): New method.
18129         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18130         Use profile_probability::split.
18131         (do_compare_rtx_and_jump): Fix adjustment of probabilities
18132         when splitting a single conditional jump into 2.
18134 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
18136         PR tree-optimization/69452
18137         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18138         decl.
18140 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18142         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18143         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18144         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18146 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18148         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18149         declaration.
18150         * config/rl78/rl78.md (movdi): New define_expand.
18151         * config/rl78/rl78.c (rl78_split_movdi): New function.
18153 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
18155         PR target/83862
18156         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18157         no longer used.
18158         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18159         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18160         128-bit to produce an UNSPEC move to get the double word with the
18161         signbit and then a shift directly to do signbit.
18162         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18163         implementation with a new version that just does either a direct
18164         move or a regular move.  Move memory interface to separate insns.
18165         Move insns so they are next to the expander.
18166         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18167         with signbit move.  Split big and little endian case.
18168         (signbit<mode>2_dm_mem_le): Likewise.
18169         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18170         (signbit<mode>2_dm2): Likewise.
18172 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18174         * config/rl78/rl78.md (anddi3): New define_expand.
18176 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18178         * config/rl78/rl78.md (umindi3): New define_expand.
18180 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18182         * config/rl78/rl78.md (smindi3): New define_expand.
18184 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18186         * config/rl78/rl78.md (smaxdi3): New define_expand.
18188 2018-01-22  Carl Love  <cel@us.ibm.com>
18190         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18191         LVX_V1TI): Add macro expansion.
18192         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18193         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18194         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18195         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18196         Change check to determine if the instruction is a byte reversing
18197         entry.  Fix typo in comment.
18198         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18199         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18200         Add def_builtin calls for new builtins.
18201         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18202         Add define_insn expansion.
18204 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18206         * config/rl78/rl78.md (umaxdi3): New define_expand.
18208 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
18210         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18211         for non-QImode registers.
18213 2018-01-22  Richard Biener  <rguenther@suse.de>
18215         PR tree-optimization/83963
18216         * graphite-scop-detection.c (scop_detection::get_sese): Delay
18217         including the loop exit block.
18218         (scop_detection::merge_sese): Likewise.
18219         (scop_detection::add_scop): Do it here instead.
18221 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18223         * doc/sourcebuild.texi (arm_softfloat): Document.
18225 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
18227         PR gcc/77734
18228         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18229         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18230         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18232 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18233             David Edelsohn  <dje.gcc@gmail.com>
18235         PR target/83946
18236         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18237         Change "crset eq" to "crset 2".
18238         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18239         (*call_indirect_aix<mode>_nospec): Likewise.
18240         (*call_value_indirect_aix<mode>_nospec): Likewise.
18241         (*call_indirect_elfv2<mode>_nospec): Likewise.
18242         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18243         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18244         change assembly output from . to $.
18245         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18246         (indirect_jump<mode>_nospec): Change assembly output from . to $.
18247         (*tablejump<mode>_internal1_nospec): Likewise.
18249 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
18251         PR target/80870
18252         * config/sh/sh_optimize_sett_clrt.cc:
18253         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18255 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
18257         PR tree-optimization/83940
18258         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18259         offset_dt to vect_constant_def rather than vect_unknown_def_type.
18260         (vect_check_load_store_mask): Add a mask_dt_out parameter and
18261         use it to pass back the definition type.
18262         (vect_check_store_rhs): Likewise rhs_dt_out.
18263         (vect_build_gather_load_calls): Add a mask_dt argument and use
18264         it instead of a call to vect_is_simple_use.
18265         (vectorizable_store): Update calls to vect_check_load_store_mask
18266         and vect_check_store_rhs.  Use the dt returned by the latter instead
18267         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
18268         instead of calls to vect_is_simple_use.  Pass the scalar rather
18269         than the vector operand to vect_is_simple_use when handling
18270         second and subsequent copies of an rhs value.
18271         (vectorizable_load): Update calls to vect_check_load_store_mask
18272         and vect_build_gather_load_calls.  Use the cached mask_dt and
18273         gs_info.offset_dt instead of calls to vect_is_simple_use.
18275 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
18277         PR middle-end/83945
18278         * tree-emutls.c: Include gimplify.h.
18279         (lower_emutls_2): New function.
18280         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18281         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18282         it before further processing.
18284         PR target/83930
18285         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18286         UINTVAL (trueop1) instead of INTVAL (op1).
18288 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
18290         PR debug/81570
18291         PR debug/83728
18292         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18293         INCOMING_FRAME_SP_OFFSET if not defined.
18294         (scan_trace): Add ENTRY argument.  If true and
18295         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18296         emit a note to adjust the CFA offset.
18297         (create_cfi_notes): Adjust scan_trace callers.
18298         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18299         INCOMING_FRAME_SP_OFFSET in the CIE.
18300         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18301         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18302         Likewise.
18303         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18304         * doc/tm.texi: Regenerated.
18306 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18308         PR rtl-optimization/83147
18309         * lra-constraints.c (remove_inheritance_pseudos): Use
18310         lra_substitute_pseudo_within_insn.
18312 2018-01-19  Tom de Vries  <tom@codesourcery.com>
18313             Cesar Philippidis  <cesar@codesourcery.com>
18315         PR target/83920
18316         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18318 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
18320         PR target/83790
18321         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18322         spaces for function labels.
18324 2018-01-19  Martin Liska  <mliska@suse.cz>
18326         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18327         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18328         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18329         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18330         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18331         (PRED_CONST_RETURN): Change from 69 to 65.
18332         (PRED_NULL_RETURN): Change from 91 to 71.
18333         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18334         (PRED_LOOP_GUARD): Change from 66 to 73.
18336 2018-01-19  Martin Liska  <mliska@suse.cz>
18338         * predict.c (predict_insn_def): Add new assert.
18339         (struct branch_predictor): Change type to signed integer.
18340         (test_prediction_value_range): Amend test to cover
18341         PROB_UNINITIALIZED.
18342         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18343         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18344         (PRED_LOOP_ITERATIONS_MAX): Likewise.
18345         (PRED_LOOP_IV_COMPARE): Likewise.
18346         * predict.h (PROB_UNINITIALIZED): Define new constant.
18348 2018-01-19  Martin Liska  <mliska@suse.cz>
18350         * predict.c (dump_prediction): Add new format for
18351         analyze_brprob.py script which is enabled with -details
18352         suboption.
18353         * profile-count.h (precise_p): New function.
18355 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
18357         PR tree-optimization/83922
18358         * tree-vect-loop.c (vect_verify_full_masking): Return false if
18359         there are no statements that need masking.
18360         (vect_active_double_reduction_p): New function.
18361         (vect_analyze_loop_operations): Use it when handling phis that
18362         are not in the loop header.
18364 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
18366         PR tree-optimization/83914
18367         * tree-vect-loop.c (vectorizable_induction): Don't convert
18368         init_expr or apply the peeling adjustment for inductions
18369         that are nested within the vectorized loop.
18371 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18373         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18374         instead of NEG.
18376 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
18378         PR sanitizer/81715
18379         PR testsuite/83882
18380         * function.h (gimplify_parameters): Add gimple_seq * argument.
18381         * function.c: Include gimple.h and options.h.
18382         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18383         for the added local temporaries if needed.
18384         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18385         if there are any parameter cleanups, wrap whole body into a
18386         try/finally with the cleanups.
18388 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
18390         PR target/82964
18391         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18392         Use GET_MODE_CLASS for scalar floating point.
18394 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
18396         PR ipa/82256
18397         patch by PaX Team
18398         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18399         Fix call of call_cgraph_insertion_hooks.
18401 2018-01-18  Martin Sebor  <msebor@redhat.com>
18403         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18405 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
18407         PR ipa/83619
18408         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18409         frequencies.
18411 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
18413         PR other/70268
18414         * common.opt: (-ffile-prefix-map): New option.
18415         * opts.c (common_handle_option): Defer it.
18416         * opts-global.c (handle_common_deferred_options): Handle it.
18417         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18418         * file-prefix-map.h: New file.
18419         (remap_debug_filename, add_debug_prefix_map): ...here.
18420         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18421         * final.c (debug_prefix_map, add_debug_prefix_map
18422         remap_debug_filename): Move to...
18423         * file-prefix-map.c: New file.
18424         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18425         generalize, get rid of alloca(), use strrchr() instead of strchr().
18426         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18427         Implement in terms of add_prefix_map().
18428         (remap_macro_filename, remap_debug_filename): Implement in term of
18429         remap_filename().
18430         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18431         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18432         * dbxout.c: Include file-prefix-map.h.
18433         * varasm.c: Likewise.
18434         * vmsdbgout.c: Likewise.
18435         * xcoffout.c: Likewise.
18436         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18437         * doc/cppopts.texi (-fmacro-prefix-map): Document.
18438         * doc/invoke.texi (-ffile-prefix-map): Document.
18439         (-fdebug-prefix-map): Update description.
18441 2018-01-18  Martin Liska  <mliska@suse.cz>
18443         * config/i386/i386.c (indirect_thunk_name): Document that also
18444         lfence is emitted.
18445         (output_indirect_thunk): Document why both instructions
18446         (pause and lfence) are generated.
18448 2018-01-18  Richard Biener  <rguenther@suse.de>
18450         PR tree-optimization/83887
18451         * graphite-scop-detection.c
18452         (scop_detection::get_nearest_dom_with_single_entry): Remove.
18453         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18454         (scop_detection::merge_sese): Re-implement with a flood-fill
18455         algorithm that properly finds a SESE region if it exists.
18457 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
18459         PR c/61240
18460         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18461         pointer_diff optimizations use view_convert instead of convert.
18463 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18465         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18466         Generate different code for -mno-speculate-indirect-jumps.
18467         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18468         (*call_indirect_aix<mode>): Disable for
18469         -mno-speculate-indirect-jumps.
18470         (*call_indirect_aix<mode>_nospec): New define_insn.
18471         (*call_value_indirect_aix<mode>): Disable for
18472         -mno-speculate-indirect-jumps.
18473         (*call_value_indirect_aix<mode>_nospec): New define_insn.
18474         (*sibcall_nonlocal_sysv<mode>): Generate different code for
18475         -mno-speculate-indirect-jumps.
18476         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18478 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18480         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18481         long double type, set the flags for noting the default long double
18482         type, even if we don't pass or return a long double type.
18484 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
18486         PR ipa/83051
18487         * ipa-inline.c (flatten_function): Do not overwrite final inlining
18488         failure.
18490 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
18492         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18493         support for merge[hl].
18494         (fold_mergehl_helper): New helper function.
18495         (tree-vector-builder.h): New #include for tree_vector_builder usage.
18496         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18497         (altivec_vmrglw_direct): Add xxmrglw insn.
18499 2018-01-17  Andrew Waterman  <andrew@sifive.com>
18501         * config/riscv/riscv.c (riscv_conditional_register_usage): If
18502         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18504 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
18506         PR lto/83121
18507         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18508         call the lto_location_cache before reading the
18509         DECL_SOURCE_LOCATION of the types.
18511 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
18512             Richard Sandiford  <richard.sandiford@linaro.org>
18514         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18515         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18516         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18517         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18518         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18519         Add declaration.
18520         * config/aarch64/constraints.md (aarch64_movti_operand):
18521         Limit immediates.
18522         * config/aarch64/predicates.md (Uti): Add new constraint.
18524 2018-01-17  Carl Love  <cel@us.ibm.com>
18526         * config/rs6000/vsx.md (define_expand xl_len_r,
18527         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18528         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18529         lxvll.
18530         (define_expand, define_insn): Move the shift left from  the
18531         define_insn to the define_expand for lxvl and stxvl instructions.
18532         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18533         and XL_LEN_R definitions to PURE.
18535 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
18537         * config/i386/i386.c (indirect_thunk_name): Declare regno
18538         as unsigned int.  Compare regno with INVALID_REGNUM.
18539         (output_indirect_thunk): Ditto.
18540         (output_indirect_thunk_function): Ditto.
18541         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
18542         in the call to output_indirect_thunk_function.
18544 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
18546         PR middle-end/83884
18547         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18548         rather than the size of inner_type to determine the stack slot size
18549         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18551 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
18553         PR target/83546
18554         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18555         to PTA_SILVERMONT.
18557 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
18559         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18560         endian Linux systems to optionally enable multilibs for selecting
18561         the long double type if the user configured an explicit type.
18562         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18563         have no long double multilibs if not defined.
18564         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18565         warn if the user used -mabi={ieee,ibm}longdouble and we built
18566         multilibs for long double.
18567         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18568         appropriate multilib option.
18569         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18570         multilib options.
18571         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18572         for building long double multilibs.
18573         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18575 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
18577         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18578         copies.
18580         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18581         64 bits.
18582         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18583         128 bits.
18585         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18586         variables.
18588         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18589         return value.
18591 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
18593         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18594         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18596 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18598         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18599         different rtl trees depending on TARGET_64BIT.
18600         (rs6000_gen_lvx): Likewise.
18602 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
18604         * config/visium/visium.md (nop): Tweak comment.
18605         (hazard_nop): Likewise.
18607 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18609         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18610         -mspeculate-indirect-jumps.
18611         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18612         for -mno-speculate-indirect-jumps.
18613         (*call_indirect_elfv2<mode>_nospec): New define_insn.
18614         (*call_value_indirect_elfv2<mode>): Disable for
18615         -mno-speculate-indirect-jumps.
18616         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18617         (indirect_jump): Emit different RTL for
18618         -mno-speculate-indirect-jumps.
18619         (*indirect_jump<mode>): Disable for
18620         -mno-speculate-indirect-jumps.
18621         (*indirect_jump<mode>_nospec): New define_insn.
18622         (tablejump): Emit different RTL for
18623         -mno-speculate-indirect-jumps.
18624         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18625         (tablejumpsi_nospec): New define_expand.
18626         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18627         (tablejumpdi_nospec): New define_expand.
18628         (*tablejump<mode>_internal1): Disable for
18629         -mno-speculate-indirect-jumps.
18630         (*tablejump<mode>_internal1_nospec): New define_insn.
18631         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18632         option.
18634 2018-01-16  Artyom Skrobov tyomitch@gmail.com
18636         * caller-save.c (insert_save): Drop unnecessary parameter.  All
18637         callers updated.
18639 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
18640             Richard Biener  <rguenth@suse.de>
18642         PR libgomp/83590
18643         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18644         return early, inline manually is_gimple_sizepos.  Make sure if we
18645         call gimplify_expr we don't end up with a gimple constant.
18646         * tree.c (variably_modified_type_p): Don't return true for
18647         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
18648         * gimplify.h (is_gimple_sizepos): Remove.
18650 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
18652         PR tree-optimization/83857
18653         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18654         vectorizable_live_operation for pure SLP statements.
18655         (vectorizable_live_operation): Handle PHIs.
18657 2018-01-16  Richard Biener  <rguenther@suse.de>
18659         PR tree-optimization/83867
18660         * tree-vect-stmts.c (vect_transform_stmt): Precompute
18661         nested_in_vect_loop_p since the scalar stmt may get invalidated.
18663 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
18665         PR c/83844
18666         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18667         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18668         If off is not INTEGER_CST, issue a may not be aligned warning
18669         rather than isn't aligned.  Use isn%'t rather than isn't.
18670         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18671         into MULT_EXPR.
18672         <case MULT_EXPR>: Improve the case when bottom and one of the
18673         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18674         operand, in that case check if the other operand is multiple of
18675         bottom divided by the INTEGER_CST operand.
18677 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
18679         PR target/83858
18680         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18681         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18682         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18683         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18684         * config/pa/pa.c (pa_function_arg_advance): Likewise.
18685         (pa_function_arg, pa_arg_partial_bytes): Likewise.
18686         (pa_function_arg_size): New function.
18688 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
18690         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18691         in a separate statement.
18693 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
18695         PR tree-optimization/83847
18696         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18697         group gathers and scatters.
18699 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
18701         PR rtl-optimization/86620
18702         * params.def (max-sched-ready-insns): Bump minimum value to 1.
18704         PR rtl-optimization/83213
18705         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18706         to last if both are JUMP_INSNs.
18708         PR tree-optimization/83843
18709         * gimple-ssa-store-merging.c
18710         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18711         store_immediate_info for bswap/nop orig_stores.
18713 2018-01-15  Andrew Waterman  <andrew@sifive.com>
18715         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18716         !TARGET_MUL.
18717         <UDIV>: Increase cost if !TARGET_DIV.
18719 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
18721         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18722         (define_attr "cr_logical_3op"): New.
18723         (cceq_ior_compare): Adjust.
18724         (cceq_ior_compare_complement): Adjust.
18725         (*cceq_rev_compare): Adjust.
18726         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18727         (is_cracked_insn): Adjust.
18728         (insn_must_be_first_in_group): Adjust.
18729         * config/rs6000/40x.md: Adjust.
18730         * config/rs6000/440.md: Adjust.
18731         * config/rs6000/476.md: Adjust.
18732         * config/rs6000/601.md: Adjust.
18733         * config/rs6000/603.md: Adjust.
18734         * config/rs6000/6xx.md: Adjust.
18735         * config/rs6000/7450.md: Adjust.
18736         * config/rs6000/7xx.md: Adjust.
18737         * config/rs6000/8540.md: Adjust.
18738         * config/rs6000/cell.md: Adjust.
18739         * config/rs6000/e300c2c3.md: Adjust.
18740         * config/rs6000/e500mc.md: Adjust.
18741         * config/rs6000/e500mc64.md: Adjust.
18742         * config/rs6000/e5500.md: Adjust.
18743         * config/rs6000/e6500.md: Adjust.
18744         * config/rs6000/mpc.md: Adjust.
18745         * config/rs6000/power4.md: Adjust.
18746         * config/rs6000/power5.md: Adjust.
18747         * config/rs6000/power6.md: Adjust.
18748         * config/rs6000/power7.md: Adjust.
18749         * config/rs6000/power8.md: Adjust.
18750         * config/rs6000/power9.md: Adjust.
18751         * config/rs6000/rs64.md: Adjust.
18752         * config/rs6000/titan.md: Adjust.
18754 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
18756         * config/i386/predicates.md (indirect_branch_operand): Rewrite
18757         ix86_indirect_branch_register logic.
18759 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
18761         * config/i386/constraints.md (Bs): Update
18762         ix86_indirect_branch_register check.  Don't check
18763         ix86_indirect_branch_register with GOT_memory_operand.
18764         (Bw): Likewise.
18765         * config/i386/predicates.md (GOT_memory_operand): Don't check
18766         ix86_indirect_branch_register here.
18767         (GOT32_symbol_operand): Likewise.
18769 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
18771         * config/i386/predicates.md (constant_call_address_operand):
18772         Rewrite ix86_indirect_branch_register logic.
18773         (sibcall_insn_operand): Likewise.
18775 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
18777         * config/i386/constraints.md (Bs): Replace
18778         ix86_indirect_branch_thunk_register with
18779         ix86_indirect_branch_register.
18780         (Bw): Likewise.
18781         * config/i386/i386.md (indirect_jump): Likewise.
18782         (tablejump): Likewise.
18783         (*sibcall_memory): Likewise.
18784         (*sibcall_value_memory): Likewise.
18785         Peepholes of indirect call and jump via memory: Likewise.
18786         * config/i386/i386.opt: Likewise.
18787         * config/i386/predicates.md (indirect_branch_operand): Likewise.
18788         (GOT_memory_operand): Likewise.
18789         (call_insn_operand): Likewise.
18790         (sibcall_insn_operand): Likewise.
18791         (GOT32_symbol_operand): Likewise.
18793 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
18795         PR middle-end/83837
18796         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18797         type rather than type addr's type points to.
18798         (expand_omp_atomic_mutex): Likewise.
18799         (expand_omp_atomic): Likewise.
18801 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
18803         PR target/83839
18804         * config/i386/i386.c (output_indirect_thunk_function): Use
18805         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18806         for  __x86_return_thunk.
18808 2018-01-15  Richard Biener  <rguenther@suse.de>
18810         PR middle-end/83850
18811         * expmed.c (extract_bit_field_1): Fix typo.
18813 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18815         PR target/83687
18816         * config/arm/iterators.md (VF): New mode iterator.
18817         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18818         Remove integer-related logic from pattern.
18819         (neon_vabd<mode>_3): Likewise.
18821 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
18823         PR middle-end/82694
18824         * common.opt (fstrict-overflow): No longer an alias.
18825         (fwrapv-pointer): New option.
18826         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18827         also for pointer types based on flag_wrapv_pointer.
18828         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18829         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18830         opts->x_flag_wrapv got set.
18831         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18832         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18833         POINTER_TYPE_OVERFLOW_UNDEFINED.
18834         * match.pd: Likewise in address comparison pattern.
18835         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18837 2018-01-15  Richard Biener  <rguenther@suse.de>
18839         PR lto/83804
18840         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18841         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
18842         Reset type names to their identifier if their TYPE_DECL doesn't
18843         have linkage (and thus is used for ODR and devirt).
18844         (save_debug_info_for_decl): Remove.
18845         (save_debug_info_for_type): Likewise.
18846         (add_tree_to_fld_list): Adjust.
18847         * tree-pretty-print.c (dump_generic_node): Make dumping of
18848         type names more robust.
18850 2018-01-15  Richard Biener  <rguenther@suse.de>
18852         * BASE-VER: Bump to 8.0.1.
18854 2018-01-14  Martin Sebor  <msebor@redhat.com>
18856         PR other/83508
18857         * builtins.c (check_access): Avoid warning when the no-warning bit
18858         is set.
18860 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
18862         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18863         * ira-color (allocno_hard_regs_compare): Likewise.
18865 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
18867         PR target/83013
18868         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18869         Use .pushsection/.popsection.
18871 2018-01-14  Martin Sebor  <msebor@redhat.com>
18873         PR c++/81327
18874         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18876 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
18878         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18879         entry from extra_headers.
18880         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18881         extra_headers, make the list bitwise identical to the i?86-*-* one.
18883 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
18885         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18886         -mcmodel=large with -mindirect-branch=thunk,
18887         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
18888         -mfunction-return=thunk-extern.
18889         * doc/invoke.texi: Document -mcmodel=large is incompatible with
18890         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
18891         -mfunction-return=thunk and -mfunction-return=thunk-extern.
18893 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
18895         * config/i386/i386.c (print_reg): Print the name of the full
18896         integer register without '%'.
18897         (ix86_print_operand): Handle 'V'.
18898         * doc/extend.texi: Document 'V' modifier.
18900 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
18902         * config/i386/constraints.md (Bs): Disallow memory operand for
18903         -mindirect-branch-register.
18904         (Bw): Likewise.
18905         * config/i386/predicates.md (indirect_branch_operand): Likewise.
18906         (GOT_memory_operand): Likewise.
18907         (call_insn_operand): Likewise.
18908         (sibcall_insn_operand): Likewise.
18909         (GOT32_symbol_operand): Likewise.
18910         * config/i386/i386.md (indirect_jump): Call convert_memory_address
18911         for -mindirect-branch-register.
18912         (tablejump): Likewise.
18913         (*sibcall_memory): Likewise.
18914         (*sibcall_value_memory): Likewise.
18915         Disallow peepholes of indirect call and jump via memory for
18916         -mindirect-branch-register.
18917         (*call_pop): Replace m with Bw.
18918         (*call_value_pop): Likewise.
18919         (*sibcall_pop_memory): Replace m with Bs.
18920         * config/i386/i386.opt (mindirect-branch-register): New option.
18921         * doc/invoke.texi: Document -mindirect-branch-register option.
18923 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
18925         * config/i386/i386-protos.h (ix86_output_function_return): New.
18926         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
18927         set function_return_type.
18928         (indirect_thunk_name): Add ret_p to indicate thunk for function
18929         return.
18930         (output_indirect_thunk_function): Pass false to
18931         indirect_thunk_name.
18932         (ix86_output_indirect_branch_via_reg): Likewise.
18933         (ix86_output_indirect_branch_via_push): Likewise.
18934         (output_indirect_thunk_function): Create alias for function
18935         return thunk if regno < 0.
18936         (ix86_output_function_return): New function.
18937         (ix86_handle_fndecl_attribute): Handle function_return.
18938         (ix86_attribute_table): Add function_return.
18939         * config/i386/i386.h (machine_function): Add
18940         function_return_type.
18941         * config/i386/i386.md (simple_return_internal): Use
18942         ix86_output_function_return.
18943         (simple_return_internal_long): Likewise.
18944         * config/i386/i386.opt (mfunction-return=): New option.
18945         (indirect_branch): Mention -mfunction-return=.
18946         * doc/extend.texi: Document function_return function attribute.
18947         * doc/invoke.texi: Document -mfunction-return= option.
18949 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
18951         * config/i386/i386-opts.h (indirect_branch): New.
18952         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
18953         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
18954         with local indirect jump when converting indirect call and jump.
18955         (ix86_set_indirect_branch_type): New.
18956         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
18957         (indirectlabelno): New.
18958         (indirect_thunk_needed): Likewise.
18959         (indirect_thunk_bnd_needed): Likewise.
18960         (indirect_thunks_used): Likewise.
18961         (indirect_thunks_bnd_used): Likewise.
18962         (INDIRECT_LABEL): Likewise.
18963         (indirect_thunk_name): Likewise.
18964         (output_indirect_thunk): Likewise.
18965         (output_indirect_thunk_function): Likewise.
18966         (ix86_output_indirect_branch_via_reg): Likewise.
18967         (ix86_output_indirect_branch_via_push): Likewise.
18968         (ix86_output_indirect_branch): Likewise.
18969         (ix86_output_indirect_jmp): Likewise.
18970         (ix86_code_end): Call output_indirect_thunk_function if needed.
18971         (ix86_output_call_insn): Call ix86_output_indirect_branch if
18972         needed.
18973         (ix86_handle_fndecl_attribute): Handle indirect_branch.
18974         (ix86_attribute_table): Add indirect_branch.
18975         * config/i386/i386.h (machine_function): Add indirect_branch_type
18976         and has_local_indirect_jump.
18977         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
18978         to true.
18979         (tablejump): Likewise.
18980         (*indirect_jump): Use ix86_output_indirect_jmp.
18981         (*tablejump_1): Likewise.
18982         (simple_return_indirect_internal): Likewise.
18983         * config/i386/i386.opt (mindirect-branch=): New option.
18984         (indirect_branch): New.
18985         (keep): Likewise.
18986         (thunk): Likewise.
18987         (thunk-inline): Likewise.
18988         (thunk-extern): Likewise.
18989         * doc/extend.texi: Document indirect_branch function attribute.
18990         * doc/invoke.texi: Document -mindirect-branch= option.
18992 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
18994         PR ipa/83051
18995         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
18997 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
18999         * ipa-inline.c (want_inline_small_function_p): Return false if
19000         inlining has already failed with CIF_FINAL_ERROR.
19001         (update_caller_keys): Call want_inline_small_function_p before
19002         can_inline_edge_p.
19003         (update_callee_keys): Likewise.
19005 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19007         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19008         New function.
19009         (rs6000_quadword_masked_address_p): Likewise.
19010         (quad_aligned_load_p): Likewise.
19011         (quad_aligned_store_p): Likewise.
19012         (const_load_sequence_p): Add comment to describe the outer-most loop.
19013         (mimic_memory_attributes_and_flags): New function.
19014         (rs6000_gen_stvx): Likewise.
19015         (replace_swapped_aligned_store): Likewise.
19016         (rs6000_gen_lvx): Likewise.
19017         (replace_swapped_aligned_load): Likewise.
19018         (replace_swapped_load_constant): Capitalize argument name in
19019         comment describing this function.
19020         (rs6000_analyze_swaps): Add a third pass to search for vector loads
19021         and stores that access quad-word aligned addresses and replace
19022         with stvx or lvx instructions when appropriate.
19023         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19024         New function prototype.
19025         (rs6000_quadword_masked_address_p): Likewise.
19026         (rs6000_gen_lvx): Likewise.
19027         (rs6000_gen_stvx): Likewise.
19028         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19029         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19030         when memory address is aligned.
19031         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19032         this split to select lvx instruction when memory address is aligned.
19033         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19034         instruction when memory address is aligned.
19035         (*vsx_le_perm_load_v16qi): Likewise.
19036         (four unnamed splitters): Modify to select the stvx instruction
19037         when memory is aligned.
19039 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
19041         * predict.c (determine_unlikely_bbs): Handle correctly BBs
19042         which appears in the queue multiple times.
19044 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19045             Alan Hayward  <alan.hayward@arm.com>
19046             David Sherwood  <david.sherwood@arm.com>
19048         * tree-vectorizer.h (vec_lower_bound): New structure.
19049         (_loop_vec_info): Add check_nonzero and lower_bounds.
19050         (LOOP_VINFO_CHECK_NONZERO): New macro.
19051         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19052         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19053         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19054         fields.  Make seg_len the distance travelled, not including the
19055         access size.
19056         (dr_direction_indicator): Declare.
19057         (dr_zero_step_indicator): Likewise.
19058         (dr_known_forward_stride_p): Likewise.
19059         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19060         tree-ssanames.h.
19061         (runtime_alias_check_p): Allow runtime alias checks with
19062         variable strides.
19063         (operator ==): Compare access_size and align.
19064         (prune_runtime_alias_test_list): Rework for new distinction between
19065         the access_size and seg_len.
19066         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
19067         segment lengths.
19068         (get_segment_min_max): New function.
19069         (create_intersect_range_checks): Use it.
19070         (dr_step_indicator): New function.
19071         (dr_direction_indicator): Likewise.
19072         (dr_zero_step_indicator): Likewise.
19073         (dr_known_forward_stride_p): Likewise.
19074         * tree-loop-distribution.c (data_ref_segment_size): Return
19075         DR_STEP * (niters - 1).
19076         (compute_alias_check_pairs): Update call to the dr_with_seg_len
19077         constructor.
19078         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19079         (vect_preserves_scalar_order_p): New function, split out from...
19080         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
19081         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19082         (vect_vfa_access_size): New function.
19083         (vect_vfa_align): Likewise.
19084         (vect_compile_time_alias): Take access_size_a and access_b arguments.
19085         (dump_lower_bound): New function.
19086         (vect_check_lower_bound): Likewise.
19087         (vect_small_gap_p): Likewise.
19088         (vectorizable_with_step_bound_p): Likewise.
19089         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19090         depencies if the vectorization factor is 1.  Convert the checks
19091         for nonzero steps into checks on the bounds of DR_STEP.  Try using
19092         a bunds check for variable steps if the minimum required step is
19093         relatively small. Update calls to the dr_with_seg_len
19094         constructor and to vect_compile_time_alias.
19095         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19096         function.
19097         (vect_loop_versioning): Call it.
19098         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19099         when retrying.
19100         (vect_estimate_min_profitable_iters): Account for any bounds checks.
19102 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19103             Alan Hayward  <alan.hayward@arm.com>
19104             David Sherwood  <david.sherwood@arm.com>
19106         * doc/sourcebuild.texi (vect_scatter_store): Document.
19107         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19108         optabs.
19109         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19110         Document.
19111         * genopinit.c (main): Add supports_vec_scatter_store and
19112         supports_vec_scatter_store_cached to target_optabs.
19113         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19114         IFN_MASK_SCATTER_STORE.
19115         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19116         functions.
19117         * internal-fn.h (internal_store_fn_p): Declare.
19118         (internal_fn_stored_value_index): Likewise.
19119         * internal-fn.c (scatter_store_direct): New macro.
19120         (expand_scatter_store_optab_fn): New function.
19121         (direct_scatter_store_optab_supported_p): New macro.
19122         (internal_store_fn_p): New function.
19123         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19124         IFN_MASK_SCATTER_STORE.
19125         (internal_fn_mask_index): Likewise.
19126         (internal_fn_stored_value_index): New function.
19127         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19128         for scatter stores.
19129         * optabs-query.h (supports_vec_scatter_store_p): Declare.
19130         * optabs-query.c (supports_vec_scatter_store_p): New function.
19131         * tree-vectorizer.h (vect_get_store_rhs): Declare.
19132         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19133         true for scatter stores.
19134         (vect_gather_scatter_fn_p): Handle scatter stores too.
19135         (vect_check_gather_scatter): Consider using scatter stores if
19136         supports_vec_scatter_store_p.
19137         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19138         scatter stores too.
19139         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19140         internal_fn_stored_value_index.
19141         (check_load_store_masking): Handle scatter stores too.
19142         (vect_get_store_rhs): Make public.
19143         (vectorizable_call): Use internal_store_fn_p.
19144         (vectorizable_store): Handle scatter store internal functions.
19145         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19146         when deciding whether the end of the group has been reached.
19147         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19148         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19149         (mask_scatter_store<mode>): New insns.
19151 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19152             Alan Hayward  <alan.hayward@arm.com>
19153             David Sherwood  <david.sherwood@arm.com>
19155         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19156         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19157         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19158         function.
19159         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19160         Use vect_truncate_gather_scatter_offset if we can't treat the
19161         operation as a normal gather load or scatter store.
19162         (get_group_load_store_type): Take the gather_scatter_info
19163         as argument.  Try using a gather load or scatter store for
19164         single-element groups.
19165         (get_load_store_type): Update calls to get_group_load_store_type
19166         and vect_use_strided_gather_scatters_p.
19168 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19169             Alan Hayward  <alan.hayward@arm.com>
19170             David Sherwood  <david.sherwood@arm.com>
19172         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19173         optional tree argument.
19174         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19175         null target hooks.
19176         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19177         but continue to use the current value as a fallback.
19178         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19179         to compare the updates.
19180         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19181         (get_load_store_type): Use it when handling a strided access.
19182         (vect_get_strided_load_store_ops): New function.
19183         (vect_get_data_ptr_increment): Likewise.
19184         (vectorizable_load): Handle strided gather loads.  Always pass
19185         a step to vect_create_data_ref_ptr and bump_vector_ptr.
19187 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19188             Alan Hayward  <alan.hayward@arm.com>
19189             David Sherwood  <david.sherwood@arm.com>
19191         * doc/md.texi (gather_load@var{m}): Document.
19192         (mask_gather_load@var{m}): Likewise.
19193         * genopinit.c (main): Add supports_vec_gather_load and
19194         supports_vec_gather_load_cached to target_optabs.
19195         * optabs-tree.c (init_tree_optimization_optabs): Use
19196         ggc_cleared_alloc to allocate target_optabs.
19197         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19198         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19199         functions.
19200         * internal-fn.h (internal_load_fn_p): Declare.
19201         (internal_gather_scatter_fn_p): Likewise.
19202         (internal_fn_mask_index): Likewise.
19203         (internal_gather_scatter_fn_supported_p): Likewise.
19204         * internal-fn.c (gather_load_direct): New macro.
19205         (expand_gather_load_optab_fn): New function.
19206         (direct_gather_load_optab_supported_p): New macro.
19207         (direct_internal_fn_optab): New function.
19208         (internal_load_fn_p): Likewise.
19209         (internal_gather_scatter_fn_p): Likewise.
19210         (internal_fn_mask_index): Likewise.
19211         (internal_gather_scatter_fn_supported_p): Likewise.
19212         * optabs-query.c (supports_at_least_one_mode_p): New function.
19213         (supports_vec_gather_load_p): Likewise.
19214         * optabs-query.h (supports_vec_gather_load_p): Declare.
19215         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19216         and memory_type field.
19217         (NUM_PATTERNS): Bump to 15.
19218         * tree-vect-data-refs.c: Include internal-fn.h.
19219         (vect_gather_scatter_fn_p): New function.
19220         (vect_describe_gather_scatter_call): Likewise.
19221         (vect_check_gather_scatter): Try using internal functions for
19222         gather loads.  Recognize existing calls to a gather load function.
19223         (vect_analyze_data_refs): Consider using gather loads if
19224         supports_vec_gather_load_p.
19225         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19226         (vect_get_gather_scatter_offset_type): Likewise.
19227         (vect_convert_mask_for_vectype): Likewise.
19228         (vect_add_conversion_to_patterm): Likewise.
19229         (vect_try_gather_scatter_pattern): Likewise.
19230         (vect_recog_gather_scatter_pattern): New pattern recognizer.
19231         (vect_vect_recog_func_ptrs): Add it.
19232         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19233         internal_fn_mask_index and internal_gather_scatter_fn_p.
19234         (check_load_store_masking): Take the gather_scatter_info as an
19235         argument and handle gather loads.
19236         (vect_get_gather_scatter_ops): New function.
19237         (vectorizable_call): Check internal_load_fn_p.
19238         (vectorizable_load): Likewise.  Handle gather load internal
19239         functions.
19240         (vectorizable_store): Update call to check_load_store_masking.
19241         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19242         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19243         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19244         (aarch64_gather_scale_operand_d): New predicates.
19245         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19246         (mask_gather_load<mode>): New insns.
19248 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19249             Alan Hayward  <alan.hayward@arm.com>
19250             David Sherwood  <david.sherwood@arm.com>
19252         * optabs.def (fold_left_plus_optab): New optab.
19253         * doc/md.texi (fold_left_plus_@var{m}): Document.
19254         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19255         * internal-fn.c (fold_left_direct): Define.
19256         (expand_fold_left_optab_fn): Likewise.
19257         (direct_fold_left_optab_supported_p): Likewise.
19258         * fold-const-call.c (fold_const_fold_left): New function.
19259         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19260         * tree-parloops.c (valid_reduction_p): New function.
19261         (gather_scalar_reductions): Use it.
19262         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19263         (vect_finish_replace_stmt): Declare.
19264         * tree-vect-loop.c (fold_left_reduction_fn): New function.
19265         (needs_fold_left_reduction_p): New function, split out from...
19266         (vect_is_simple_reduction): ...here.  Accept reductions that
19267         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19268         (vect_force_simple_reduction): Also store the reduction type in
19269         the assignment's STMT_VINFO_REDUC_TYPE.
19270         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19271         (merge_with_identity): New function.
19272         (vect_expand_fold_left): Likewise.
19273         (vectorize_fold_left_reduction): Likewise.
19274         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
19275         scalar phi in place for it.  Check for target support and reject
19276         cases that would reassociate the operation.  Defer the transform
19277         phase to vectorize_fold_left_reduction.
19278         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19279         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19280         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19282 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19284         * tree-if-conv.c (predicate_mem_writes): Remove redundant
19285         call to ifc_temp_var.
19287 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19288             Alan Hayward  <alan.hayward@arm.com>
19289             David Sherwood  <david.sherwood@arm.com>
19291         * target.def (legitimize_address_displacement): Take the original
19292         offset as a poly_int.
19293         * targhooks.h (default_legitimize_address_displacement): Update
19294         accordingly.
19295         * targhooks.c (default_legitimize_address_displacement): Likewise.
19296         * doc/tm.texi: Regenerate.
19297         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19298         as an argument, moving assert of ad->disp == ad->disp_term to...
19299         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
19300         Try calling targetm.legitimize_address_displacement before expanding
19301         the address rather than afterwards, and adjust for the new interface.
19302         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19303         Match the new hook interface.  Handle SVE addresses.
19304         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19305         new hook interface.
19307 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19309         * Makefile.in (OBJS): Add early-remat.o.
19310         * target.def (select_early_remat_modes): New hook.
19311         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19312         * doc/tm.texi: Regenerate.
19313         * targhooks.h (default_select_early_remat_modes): Declare.
19314         * targhooks.c (default_select_early_remat_modes): New function.
19315         * timevar.def (TV_EARLY_REMAT): New timevar.
19316         * passes.def (pass_early_remat): New pass.
19317         * tree-pass.h (make_pass_early_remat): Declare.
19318         * early-remat.c: New file.
19319         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19320         function.
19321         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19323 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19324             Alan Hayward  <alan.hayward@arm.com>
19325             David Sherwood  <david.sherwood@arm.com>
19327         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19328         vfm1 with a bound_epilog parameter.
19329         (vect_do_peeling): Update calls accordingly, and move the prologue
19330         call earlier in the function.  Treat the base bound_epilog as 0 for
19331         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
19332         this base when peeling for gaps.
19333         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19334         with fully-masked loops.
19335         (vect_estimate_min_profitable_iters): Handle the single peeled
19336         iteration in that case.
19338 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19339             Alan Hayward  <alan.hayward@arm.com>
19340             David Sherwood  <david.sherwood@arm.com>
19342         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19343         single-element interleaving even if the size is not a power of 2.
19344         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19345         accesses for single-element interleaving if the group size is
19346         not a power of 2.
19348 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19349             Alan Hayward  <alan.hayward@arm.com>
19350             David Sherwood  <david.sherwood@arm.com>
19352         * doc/md.texi (fold_extract_last_@var{m}): Document.
19353         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19354         * optabs.def (fold_extract_last_optab): New optab.
19355         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19356         * internal-fn.c (fold_extract_direct): New macro.
19357         (expand_fold_extract_optab_fn): Likewise.
19358         (direct_fold_extract_optab_supported_p): Likewise.
19359         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19360         * tree-vect-loop.c (vect_model_reduction_cost): Handle
19361         EXTRACT_LAST_REDUCTION.
19362         (get_initial_def_for_reduction): Do not create an initial vector
19363         for EXTRACT_LAST_REDUCTION reductions.
19364         (vectorizable_reduction): Leave the scalar phi in place for
19365         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
19366         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
19367         epilogue code for EXTRACT_LAST_REDUCTION and defer the
19368         transform phase to vectorizable_condition.
19369         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19370         split out from...
19371         (vect_finish_stmt_generation): ...here.
19372         (vect_finish_replace_stmt): New function.
19373         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19374         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19375         pattern.
19376         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19378 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19379             Alan Hayward  <alan.hayward@arm.com>
19380             David Sherwood  <david.sherwood@arm.com>
19382         * doc/md.texi (extract_last_@var{m}): Document.
19383         * optabs.def (extract_last_optab): New optab.
19384         * internal-fn.def (EXTRACT_LAST): New internal function.
19385         * internal-fn.c (cond_unary_direct): New macro.
19386         (expand_cond_unary_optab_fn): Likewise.
19387         (direct_cond_unary_optab_supported_p): Likewise.
19388         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19389         loops using EXTRACT_LAST.
19390         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19391         (extract_last_<mode>): ...this optab.
19392         (vec_extract<mode><Vel>): Update accordingly.
19394 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19395             Alan Hayward  <alan.hayward@arm.com>
19396             David Sherwood  <david.sherwood@arm.com>
19398         * target.def (empty_mask_is_expensive): New hook.
19399         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19400         * doc/tm.texi: Regenerate.
19401         * targhooks.h (default_empty_mask_is_expensive): Declare.
19402         * targhooks.c (default_empty_mask_is_expensive): New function.
19403         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19404         if the target says that empty masks are expensive.
19405         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19406         New function.
19407         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19409 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19410             Alan Hayward  <alan.hayward@arm.com>
19411             David Sherwood  <david.sherwood@arm.com>
19413         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19414         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19415         (vect_use_loop_mask_for_alignment_p): New function.
19416         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19417         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19418         niters_skip argument.  Make sure that the first niters_skip elements
19419         of the first iteration are inactive.
19420         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19421         Update call to vect_set_loop_masks_directly.
19422         (get_misalign_in_elems): New function, split out from...
19423         (vect_gen_prolog_loop_niters): ...here.
19424         (vect_update_init_of_dr): Take a code argument that specifies whether
19425         the adjustment should be added or subtracted.
19426         (vect_update_init_of_drs): Likewise.
19427         (vect_prepare_for_masked_peels): New function.
19428         (vect_do_peeling): Skip prologue peeling if we're using a mask
19429         instead.  Update call to vect_update_inits_of_drs.
19430         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19431         mask_skip_niters.
19432         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19433         alignment.  Do not include the number of peeled iterations in
19434         the minimum threshold in that case.
19435         (vectorizable_induction): Adjust the start value down by
19436         LOOP_VINFO_MASK_SKIP_NITERS iterations.
19437         (vect_transform_loop): Call vect_prepare_for_masked_peels.
19438         Take the number of skipped iterations into account when calculating
19439         the loop bounds.
19440         * tree-vect-stmts.c (vect_gen_while_not): New function.
19442 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19443             Alan Hayward  <alan.hayward@arm.com>
19444             David Sherwood  <david.sherwood@arm.com>
19446         * doc/sourcebuild.texi (vect_fully_masked): Document.
19447         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19448         default value to 0.
19449         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19450         split out from...
19451         (vect_analyze_loop_2): ...here. Don't check the vectorization
19452         factor against the number of loop iterations if the loop is
19453         fully-masked.
19455 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19456             Alan Hayward  <alan.hayward@arm.com>
19457             David Sherwood  <david.sherwood@arm.com>
19459         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19460         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19461         (dump_groups): Update accordingly.
19462         (iv_use::mem_type): New member variable.
19463         (address_p): New function.
19464         (record_use): Add a mem_type argument and initialize the new
19465         mem_type field.
19466         (record_group_use): Add a mem_type argument.  Use address_p.
19467         Remove obsolete null checks of base_object.  Update call to record_use.
19468         (find_interesting_uses_op): Update call to record_group_use.
19469         (find_interesting_uses_cond): Likewise.
19470         (find_interesting_uses_address): Likewise.
19471         (get_mem_type_for_internal_fn): New function.
19472         (find_address_like_use): Likewise.
19473         (find_interesting_uses_stmt): Try find_address_like_use before
19474         calling find_interesting_uses_op.
19475         (addr_offset_valid_p): Use the iv mem_type field as the type
19476         of the addressed memory.
19477         (add_autoinc_candidates): Likewise.
19478         (get_address_cost): Likewise.
19479         (split_small_address_groups_p): Use address_p.
19480         (split_address_groups): Likewise.
19481         (add_iv_candidate_for_use): Likewise.
19482         (autoinc_possible_for_pair): Likewise.
19483         (rewrite_groups): Likewise.
19484         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19485         (determine_group_iv_cost): Update after split of USE_ADDRESS.
19486         (get_alias_ptr_type_for_ptr_address): New function.
19487         (rewrite_use_address): Rewrite address uses in calls that were
19488         identified by find_address_like_use.
19490 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19491             Alan Hayward  <alan.hayward@arm.com>
19492             David Sherwood  <david.sherwood@arm.com>
19494         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19495         TARGET_MEM_REFs.
19496         * gimple-expr.h (is_gimple_addressable: Likewise.
19497         * gimple-expr.c (is_gimple_address): Likewise.
19498         * internal-fn.c (expand_call_mem_ref): New function.
19499         (expand_mask_load_optab_fn): Use it.
19500         (expand_mask_store_optab_fn): Likewise.
19502 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19503             Alan Hayward  <alan.hayward@arm.com>
19504             David Sherwood  <david.sherwood@arm.com>
19506         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19507         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19508         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19509         (cond_umax@var{mode}): Document.
19510         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19511         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19512         (cond_umin_optab, cond_umax_optab): New optabs.
19513         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19514         (COND_IOR, COND_XOR): New internal functions.
19515         * internal-fn.h (get_conditional_internal_fn): Declare.
19516         * internal-fn.c (cond_binary_direct): New macro.
19517         (expand_cond_binary_optab_fn): Likewise.
19518         (direct_cond_binary_optab_supported_p): Likewise.
19519         (get_conditional_internal_fn): New function.
19520         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19521         Cope with reduction statements that are vectorized as calls rather
19522         than assignments.
19523         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19524         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19525         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19526         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19527         (UNSPEC_COND_EOR): New unspecs.
19528         (optab): Add mappings for them.
19529         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19530         (sve_int_op, sve_fp_op): New int attributes.
19532 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19533             Alan Hayward  <alan.hayward@arm.com>
19534             David Sherwood  <david.sherwood@arm.com>
19536         * optabs.def (while_ult_optab): New optab.
19537         * doc/md.texi (while_ult@var{m}@var{n}): Document.
19538         * internal-fn.def (WHILE_ULT): New internal function.
19539         * internal-fn.h (direct_internal_fn_supported_p): New override
19540         that takes two types as argument.
19541         * internal-fn.c (while_direct): New macro.
19542         (expand_while_optab_fn): New function.
19543         (convert_optab_supported_p): Likewise.
19544         (direct_while_optab_supported_p): New macro.
19545         * wide-int.h (wi::udiv_ceil): New function.
19546         * tree-vectorizer.h (rgroup_masks): New structure.
19547         (vec_loop_masks): New typedef.
19548         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19549         and fully_masked_p.
19550         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19551         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19552         (vect_max_vf): New function.
19553         (slpeel_make_loop_iterate_ntimes): Delete.
19554         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19555         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19556         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19557         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19558         internal-fn.h, stor-layout.h and optabs-query.h.
19559         (vect_set_loop_mask): New function.
19560         (add_preheader_seq): Likewise.
19561         (add_header_seq): Likewise.
19562         (interleave_supported_p): Likewise.
19563         (vect_maybe_permute_loop_masks): Likewise.
19564         (vect_set_loop_masks_directly): Likewise.
19565         (vect_set_loop_condition_masked): Likewise.
19566         (vect_set_loop_condition_unmasked): New function, split out from
19567         slpeel_make_loop_iterate_ntimes.
19568         (slpeel_make_loop_iterate_ntimes): Rename to..
19569         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
19570         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19571         (vect_do_peeling): Update call accordingly.
19572         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19573         loops.
19574         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19575         mask_compare_type, can_fully_mask_p and fully_masked_p.
19576         (release_vec_loop_masks): New function.
19577         (_loop_vec_info): Use it to free the loop masks.
19578         (can_produce_all_loop_masks_p): New function.
19579         (vect_get_max_nscalars_per_iter): Likewise.
19580         (vect_verify_full_masking): Likewise.
19581         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19582         retries, and free the mask rgroups before retrying.  Check loop-wide
19583         reasons for disallowing fully-masked loops.  Make the final decision
19584         about whether use a fully-masked loop or not.
19585         (vect_estimate_min_profitable_iters): Do not assume that peeling
19586         for the number of iterations will be needed for fully-masked loops.
19587         (vectorizable_reduction): Disable fully-masked loops.
19588         (vectorizable_live_operation): Likewise.
19589         (vect_halve_mask_nunits): New function.
19590         (vect_double_mask_nunits): Likewise.
19591         (vect_record_loop_mask): Likewise.
19592         (vect_get_loop_mask): Likewise.
19593         (vect_transform_loop): Handle the case in which the final loop
19594         iteration might handle a partial vector.  Call vect_set_loop_condition
19595         instead of slpeel_make_loop_iterate_ntimes.
19596         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19597         (check_load_store_masking): New function.
19598         (prepare_load_store_mask): Likewise.
19599         (vectorizable_store): Handle fully-masked loops.
19600         (vectorizable_load): Likewise.
19601         (supportable_widening_operation): Use vect_halve_mask_nunits for
19602         booleans.
19603         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19604         (vect_gen_while): New function.
19605         * config/aarch64/aarch64.md (umax<mode>3): New expander.
19606         (aarch64_uqdec<mode>): New insn.
19608 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19609             Alan Hayward  <alan.hayward@arm.com>
19610             David Sherwood  <david.sherwood@arm.com>
19612         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19613         (reduc_xor_scal_optab): New optabs.
19614         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19615         (reduc_xor_scal_@var{m}): Document.
19616         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19617         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19618         internal functions.
19619         * fold-const-call.c (fold_const_call): Handle them.
19620         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19621         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19622         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19623         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19624         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19625         (UNSPEC_XORV): New unspecs.
19626         (optab): Add entries for them.
19627         (BITWISEV): New int iterator.
19628         (bit_reduc_op): New int attributes.
19630 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19631             Alan Hayward  <alan.hayward@arm.com>
19632             David Sherwood  <david.sherwood@arm.com>
19634         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19635         * internal-fn.def (VEC_SHL_INSERT): New internal function.
19636         * optabs.def (vec_shl_insert_optab): New optab.
19637         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19638         (duplicate_and_interleave): Likewise.
19639         * tree-vect-loop.c: Include internal-fn.h.
19640         (neutral_op_for_slp_reduction): New function, split out from
19641         get_initial_defs_for_reduction.
19642         (get_initial_def_for_reduction): Handle option 2 for variable-length
19643         vectors by loading the neutral value into a vector and then shifting
19644         the initial value into element 0.
19645         (get_initial_defs_for_reduction): Replace the code argument with
19646         the neutral value calculated by neutral_op_for_slp_reduction.
19647         Use gimple_build_vector for constant-length vectors.
19648         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19649         but the first group_size elements have a neutral value.
19650         Use duplicate_and_interleave otherwise.
19651         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19652         Update call to get_initial_defs_for_reduction.  Handle SLP
19653         reductions for variable-length vectors by creating one vector
19654         result for each scalar result, with the elements associated
19655         with other scalar results stubbed out with the neutral value.
19656         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19657         Require IFN_VEC_SHL_INSERT for double reductions on
19658         variable-length vectors, or SLP reductions that have
19659         a neutral value.  Require can_duplicate_and_interleave_p
19660         support for variable-length unchained SLP reductions if there
19661         is no neutral value, such as for MIN/MAX reductions.  Also require
19662         the number of vector elements to be a multiple of the number of
19663         SLP statements when doing variable-length unchained SLP reductions.
19664         Update call to vect_create_epilog_for_reduction.
19665         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19666         and remove initial values.
19667         (duplicate_and_interleave): Make public.
19668         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19669         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19671 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19672             Alan Hayward  <alan.hayward@arm.com>
19673             David Sherwood  <david.sherwood@arm.com>
19675         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19676         (can_duplicate_and_interleave_p): New function.
19677         (vect_get_and_check_slp_defs): Take the vector of statements
19678         rather than just the current one.  Remove excess parentheses.
19679         Restriction rejectinon of vect_constant_def and vect_external_def
19680         for variable-length vectors to boolean types, or types for which
19681         can_duplicate_and_interleave_p is false.
19682         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19683         (duplicate_and_interleave): New function.
19684         (vect_get_constant_vectors): Use gimple_build_vector for
19685         constant-length vectors and suitable variable-length constant
19686         vectors.  Use duplicate_and_interleave for other variable-length
19687         vectors.  Don't defer the update when inserting new statements.
19689 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19690             Alan Hayward  <alan.hayward@arm.com>
19691             David Sherwood  <david.sherwood@arm.com>
19693         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19694         min_profitable_iters doesn't go negative.
19696 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19697             Alan Hayward  <alan.hayward@arm.com>
19698             David Sherwood  <david.sherwood@arm.com>
19700         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19701         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19702         * optabs.def (vec_mask_load_lanes_optab): New optab.
19703         (vec_mask_store_lanes_optab): Likewise.
19704         * internal-fn.def (MASK_LOAD_LANES): New internal function.
19705         (MASK_STORE_LANES): Likewise.
19706         * internal-fn.c (mask_load_lanes_direct): New macro.
19707         (mask_store_lanes_direct): Likewise.
19708         (expand_mask_load_optab_fn): Handle masked operations.
19709         (expand_mask_load_lanes_optab_fn): New macro.
19710         (expand_mask_store_optab_fn): Handle masked operations.
19711         (expand_mask_store_lanes_optab_fn): New macro.
19712         (direct_mask_load_lanes_optab_supported_p): Likewise.
19713         (direct_mask_store_lanes_optab_supported_p): Likewise.
19714         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19715         parameter.
19716         (vect_load_lanes_supported): Likewise.
19717         * tree-vect-data-refs.c (strip_conversion): New function.
19718         (can_group_stmts_p): Likewise.
19719         (vect_analyze_data_ref_accesses): Use it instead of checking
19720         for a pair of assignments.
19721         (vect_store_lanes_supported): Take a masked_p parameter.
19722         (vect_load_lanes_supported): Likewise.
19723         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19724         vect_store_lanes_supported and vect_load_lanes_supported.
19725         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19726         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19727         parameter.  Don't allow gaps for masked accesses.
19728         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
19729         and vect_load_lanes_supported.
19730         (get_load_store_type): Take a masked_p parameter and update
19731         call to get_group_load_store_type.
19732         (vectorizable_store): Update call to get_load_store_type.
19733         Handle IFN_MASK_STORE_LANES.
19734         (vectorizable_load): Update call to get_load_store_type.
19735         Handle IFN_MASK_LOAD_LANES.
19737 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19738             Alan Hayward  <alan.hayward@arm.com>
19739             David Sherwood  <david.sherwood@arm.com>
19741         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19742         modes for SVE.
19743         * config/aarch64/aarch64-protos.h
19744         (aarch64_sve_struct_memory_operand_p): Declare.
19745         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19746         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19747         (VPRED, vpred): Handle SVE structure modes.
19748         * config/aarch64/constraints.md (Utx): New constraint.
19749         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19750         (aarch64_sve_struct_nonimmediate_operand): New predicates.
19751         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19752         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19753         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19754         structure modes.  Split into pieces after RA.
19755         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19756         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19757         New patterns.
19758         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19759         SVE structure modes.
19760         (aarch64_classify_address): Likewise.
19761         (sizetochar): Move earlier in file.
19762         (aarch64_print_operand): Handle SVE register lists.
19763         (aarch64_array_mode): New function.
19764         (aarch64_sve_struct_memory_operand_p): Likewise.
19765         (TARGET_ARRAY_MODE): Redefine.
19767 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19768             Alan Hayward  <alan.hayward@arm.com>
19769             David Sherwood  <david.sherwood@arm.com>
19771         * target.def (array_mode): New target hook.
19772         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19773         * doc/tm.texi: Regenerate.
19774         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19775         * hooks.c (hook_optmode_mode_uhwi_none): New function.
19776         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19777         targetm.array_mode.
19778         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
19779         type sizes.
19781 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19782             Alan Hayward  <alan.hayward@arm.com>
19783             David Sherwood  <david.sherwood@arm.com>
19785         * fold-const.c (fold_binary_loc): Check the argument types
19786         rather than the result type when testing for a vector operation.
19788 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19790         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19791         * doc/tm.texi: Regenerate.
19793 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
19794             Alan Hayward  <alan.hayward@arm.com>
19795             David Sherwood  <david.sherwood@arm.com>
19797         * doc/invoke.texi (-msve-vector-bits=): Document new option.
19798         (sve): Document new AArch64 extension.
19799         * doc/md.texi (w): Extend the description of the AArch64
19800         constraint to include SVE vectors.
19801         (Upl, Upa): Document new AArch64 predicate constraints.
19802         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19803         enum.
19804         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19805         (msve-vector-bits=): New option.
19806         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19807         SVE when these are disabled.
19808         (sve): New extension.
19809         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19810         modes.  Adjust their number of units based on aarch64_sve_vg.
19811         (MAX_BITSIZE_MODE_ANY_MODE): Define.
19812         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19813         aarch64_addr_query_type.
19814         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19815         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19816         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19817         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19818         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19819         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19820         (aarch64_simd_imm_zero_p): Delete.
19821         (aarch64_check_zero_based_sve_index_immediate): Declare.
19822         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19823         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19824         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19825         (aarch64_sve_float_mul_immediate_p): Likewise.
19826         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19827         rather than an rtx.
19828         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19829         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19830         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19831         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19832         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19833         (aarch64_regmode_natural_size): Likewise.
19834         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19835         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19836         left one place.
19837         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19838         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19839         for VG and the SVE predicate registers.
19840         (V_ALIASES): Add a "z"-prefixed alias.
19841         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19842         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19843         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19844         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19845         (REG_CLASS_NAMES): Add entries for them.
19846         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
19847         and the predicate registers.
19848         (aarch64_sve_vg): Declare.
19849         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19850         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19851         (REGMODE_NATURAL_SIZE): Define.
19852         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19853         SVE macros.
19854         * config/aarch64/aarch64.c: Include cfgrtl.h.
19855         (simd_immediate_info): Add a constructor for series vectors,
19856         and an associated step field.
19857         (aarch64_sve_vg): New variable.
19858         (aarch64_dbx_register_number): Handle VG and the predicate registers.
19859         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19860         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19861         (VEC_ANY_DATA, VEC_STRUCT): New constants.
19862         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19863         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19864         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19865         (aarch64_get_mask_mode): New functions.
19866         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19867         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19868         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
19869         predicate modes and predicate registers.  Explicitly restrict
19870         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
19871         to store a vector mode if it is recognized by
19872         aarch64_classify_vector_mode.
19873         (aarch64_regmode_natural_size): New function.
19874         (aarch64_hard_regno_caller_save_mode): Return the original mode
19875         for predicates.
19876         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19877         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19878         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19879         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19880         functions.
19881         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
19882         does not overlap dest if the function is frame-related.  Handle
19883         SVE constants.
19884         (aarch64_split_add_offset): New function.
19885         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19886         them aarch64_add_offset.
19887         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
19888         and update call to aarch64_sub_sp.
19889         (aarch64_add_cfa_expression): New function.
19890         (aarch64_expand_prologue): Pass extra temporary registers to the
19891         functions above.  Handle the case in which we need to emit new
19892         DW_CFA_expressions for registers that were originally saved
19893         relative to the stack pointer, but now have to be expressed
19894         relative to the frame pointer.
19895         (aarch64_output_mi_thunk): Pass extra temporary registers to the
19896         functions above.
19897         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
19898         IP0 and IP1 values for SVE frames.
19899         (aarch64_expand_vec_series): New function.
19900         (aarch64_expand_sve_widened_duplicate): Likewise.
19901         (aarch64_expand_sve_const_vector): Likewise.
19902         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
19903         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
19904         into the register, rather than emitting a SET directly.
19905         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
19906         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
19907         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
19908         (offset_9bit_signed_scaled_p): New functions.
19909         (aarch64_replicate_bitmask_imm): New function.
19910         (aarch64_bitmask_imm): Use it.
19911         (aarch64_cannot_force_const_mem): Reject expressions involving
19912         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
19913         (aarch64_classify_index): Handle SVE indices, by requiring
19914         a plain register index with a scale that matches the element size.
19915         (aarch64_classify_address): Handle SVE addresses.  Assert that
19916         the mode of the address is VOIDmode or an integer mode.
19917         Update call to aarch64_classify_symbol.
19918         (aarch64_classify_symbolic_expression): Update call to
19919         aarch64_classify_symbol.
19920         (aarch64_const_vec_all_in_range_p): New function.
19921         (aarch64_print_vector_float_operand): Likewise.
19922         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
19923         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
19924         and the FP immediates 1.0 and 0.5.
19925         (aarch64_print_address_internal): Handle SVE addresses.
19926         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
19927         (aarch64_regno_regclass): Handle predicate registers.
19928         (aarch64_secondary_reload): Handle big-endian reloads of SVE
19929         data modes.
19930         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
19931         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
19932         (aarch64_convert_sve_vector_bits): New function.
19933         (aarch64_override_options): Use it to handle -msve-vector-bits=.
19934         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19935         rather than an rtx.
19936         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
19937         Handle SVE vector and predicate modes.  Accept VL-based constants
19938         that need only one temporary register, and VL offsets that require
19939         no temporary registers.
19940         (aarch64_conditional_register_usage): Mark the predicate registers
19941         as fixed if SVE isn't available.
19942         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
19943         Return true for SVE vector and predicate modes.
19944         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
19945         rather than an unsigned int.  Handle SVE modes.
19946         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
19947         SVE modes.
19948         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
19949         if SVE is enabled.
19950         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19951         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19952         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19953         (aarch64_sve_float_mul_immediate_p): New functions.
19954         (aarch64_sve_valid_immediate): New function.
19955         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
19956         Explicitly reject structure modes.  Check for INDEX constants.
19957         Handle PTRUE and PFALSE constants.
19958         (aarch64_check_zero_based_sve_index_immediate): New function.
19959         (aarch64_simd_imm_zero_p): Delete.
19960         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
19961         vector modes.  Accept constants in the range of CNT[BHWD].
19962         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
19963         ask for an Advanced SIMD mode.
19964         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
19965         (aarch64_simd_vector_alignment): Handle SVE predicates.
19966         (aarch64_vectorize_preferred_vector_alignment): New function.
19967         (aarch64_simd_vector_alignment_reachable): Use it instead of
19968         the vector size.
19969         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
19970         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
19971         functions.
19972         (MAX_VECT_LEN): Delete.
19973         (expand_vec_perm_d): Add a vec_flags field.
19974         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
19975         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
19976         (aarch64_evpc_ext): Don't apply a big-endian lane correction
19977         for SVE modes.
19978         (aarch64_evpc_rev): Rename to...
19979         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
19980         (aarch64_evpc_rev_global): New function.
19981         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
19982         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
19983         MAX_VECT_LEN.
19984         (aarch64_evpc_sve_tbl): New function.
19985         (aarch64_expand_vec_perm_const_1): Update after rename of
19986         aarch64_evpc_rev.  Handle SVE permutes too, trying
19987         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
19988         than aarch64_evpc_tbl.
19989         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
19990         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
19991         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
19992         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
19993         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
19994         (aarch64_expand_sve_vcond): New functions.
19995         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
19996         of aarch64_vector_mode_p.
19997         (aarch64_dwarf_poly_indeterminate_value): New function.
19998         (aarch64_compute_pressure_classes): Likewise.
19999         (aarch64_can_change_mode_class): Likewise.
20000         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20001         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20002         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20003         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20004         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20005         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20006         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20007         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20008         constraints.
20009         (Dn, Dl, Dr): Accept const as well as const_vector.
20010         (Dz): Likewise.  Compare against CONST0_RTX.
20011         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20012         of "vector" where appropriate.
20013         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20014         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20015         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20016         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20017         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20018         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20019         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20020         (v_int_equiv): Extend to SVE modes.
20021         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20022         mode attributes.
20023         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20024         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20025         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20026         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20027         (SVE_COND_FP_CMP): New int iterators.
20028         (perm_hilo): Handle the new unpack unspecs.
20029         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20030         attributes.
20031         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20032         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20033         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20034         (aarch64_equality_operator, aarch64_constant_vector_operand)
20035         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20036         (aarch64_sve_nonimmediate_operand): Likewise.
20037         (aarch64_sve_general_operand): Likewise.
20038         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20039         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20040         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20041         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20042         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20043         (aarch64_sve_float_arith_immediate): Likewise.
20044         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20045         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20046         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20047         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20048         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20049         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20050         (aarch64_sve_float_arith_operand): Likewise.
20051         (aarch64_sve_float_arith_with_sub_operand): Likewise.
20052         (aarch64_sve_float_mul_operand): Likewise.
20053         (aarch64_sve_vec_perm_operand): Likewise.
20054         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20055         (aarch64_mov_operand): Accept const_poly_int and const_vector.
20056         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20057         as well as const_vector.
20058         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20059         in file.  Use CONST0_RTX and CONSTM1_RTX.
20060         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
20061         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20062         Use aarch64_simd_imm_zero.
20063         * config/aarch64/aarch64-sve.md: New file.
20064         * config/aarch64/aarch64.md: Include it.
20065         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20066         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20067         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20068         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20069         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20070         (sve): New attribute.
20071         (enabled): Disable instructions with the sve attribute unless
20072         TARGET_SVE.
20073         (movqi, movhi): Pass CONST_POLY_INT operaneds through
20074         aarch64_expand_mov_immediate.
20075         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20076         CNT[BHSD] immediates.
20077         (movti): Split CONST_POLY_INT moves into two halves.
20078         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20079         Split additions that need a temporary here if the destination
20080         is the stack pointer.
20081         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20082         (*add<mode>3_poly_1): New instruction.
20083         (set_clobber_cc): New expander.
20085 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
20087         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20088         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
20089         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20090         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20091         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
20092         Change innermode from fixed_mode_size to machine_mode.
20093         (simplify_subreg): Update call accordingly.  Handle a constant-sized
20094         subreg of a variable-length CONST_VECTOR.
20096 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
20097             Alan Hayward  <alan.hayward@arm.com>
20098             David Sherwood  <david.sherwood@arm.com>
20100         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20101         (add_offset_to_base): New function, split out from...
20102         (create_mem_ref): ...here.  When handling a scale other than 1,
20103         check first whether the address is valid without the offset.
20104         Add it into the base if so, leaving the index and scale as-is.
20106 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
20108         PR c++/83778
20109         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20110         fold_for_warn before checking if arg2 is INTEGER_CST.
20112 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
20114         * config/rs6000/predicates.md (load_multiple_operation): Delete.
20115         (store_multiple_operation): Delete.
20116         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20117         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20118         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20119         guarded by TARGET_STRING.
20120         (rs6000_output_load_multiple): Delete.
20121         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20122         OPTION_MASK_STRING / TARGET_STRING handling.
20123         (print_operand) <'N', 'O'>: Add comment that these are unused now.
20124         (const rs6000_opt_masks) <"string">: Change mask to 0.
20125         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20126         (MASK_STRING): Delete.
20127         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20128         parts.  Simplify.
20129         (load_multiple): Delete.
20130         (*ldmsi8): Delete.
20131         (*ldmsi7): Delete.
20132         (*ldmsi6): Delete.
20133         (*ldmsi5): Delete.
20134         (*ldmsi4): Delete.
20135         (*ldmsi3): Delete.
20136         (store_multiple): Delete.
20137         (*stmsi8): Delete.
20138         (*stmsi7): Delete.
20139         (*stmsi6): Delete.
20140         (*stmsi5): Delete.
20141         (*stmsi4): Delete.
20142         (*stmsi3): Delete.
20143         (movmemsi_8reg): Delete.
20144         (corresponding unnamed define_insn): Delete.
20145         (movmemsi_6reg): Delete.
20146         (corresponding unnamed define_insn): Delete.
20147         (movmemsi_4reg): Delete.
20148         (corresponding unnamed define_insn): Delete.
20149         (movmemsi_2reg): Delete.
20150         (corresponding unnamed define_insn): Delete.
20151         (movmemsi_1reg): Delete.
20152         (corresponding unnamed define_insn): Delete.
20153         * config/rs6000/rs6000.opt (mno-string): New.
20154         (mstring): Replace by deprecation warning stub.
20155         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20157 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
20159         * regrename.c (regrename_do_replace): If replacing the same
20160         reg multiple times, try to reuse last created gen_raw_REG.
20162         PR debug/81155
20163         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20164         main to workaround a bug in GDB.
20166 2018-01-12  Tom de Vries  <tom@codesourcery.com>
20168         PR target/83737
20169         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20171 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
20173         PR rtl-optimization/80481
20174         * ira-color.c (get_cap_member): New function.
20175         (allocnos_conflict_by_live_ranges_p): Use it.
20176         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20177         (setup_slot_coalesced_allocno_live_ranges): Ditto.
20179 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
20181         PR target/83628
20182         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20183         (*saddl_se_1): Ditto.
20184         (*ssubsi_1): Ditto.
20185         (*ssubl_se_1): Ditto.
20187 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
20189         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20190         rather than wi::to_widest for DR_INITs.
20191         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20192         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20193         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20194         INTEGER_CSTs.
20195         (vect_analyze_group_access_1): Note that here.
20197 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
20199         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20200         polynomial type sizes.
20202 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
20204         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20205         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20206         (gimple_add_tmp_var): Likewise.
20208 2018-01-12  Martin Liska  <mliska@suse.cz>
20210         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20211         (gimple_alloc_sizes): Likewise.
20212         (dump_gimple_statistics): Use PRIu64 in printf format.
20213         * gimple.h: Change uint64_t to int.
20215 2018-01-12  Martin Liska  <mliska@suse.cz>
20217         * tree-core.h: Use uint64_t instead of int.
20218         * tree.c (tree_node_counts): Likewise.
20219         (tree_node_sizes): Likewise.
20220         (dump_tree_statistics): Use PRIu64 in printf format.
20222 2018-01-12  Martin Liska  <mliska@suse.cz>
20224         * Makefile.in: As qsort_chk is implemented in vec.c, add
20225         vec.o to linkage of gencfn-macros.
20226         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20227         passing the info to record_node_allocation_statistics.
20228         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20229         and pass the info.
20230         * ggc-common.c (struct ggc_usage): Add operator== and use
20231         it in operator< and compare function.
20232         * mem-stats.h (struct mem_usage): Likewise.
20233         * vec.c (struct vec_usage): Remove operator< and compare
20234         function. Can be simply inherited.
20236 2018-01-12  Martin Jambor  <mjambor@suse.cz>
20238         PR target/81616
20239         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20240         * tree-ssa-math-opts.c: Include domwalk.h.
20241         (convert_mult_to_fma_1): New function.
20242         (fma_transformation_info): New type.
20243         (fma_deferring_state): Likewise.
20244         (cancel_fma_deferring): New function.
20245         (result_of_phi): Likewise.
20246         (last_fma_candidate_feeds_initial_phi): Likewise.
20247         (convert_mult_to_fma): Added deferring logic, split actual
20248         transformation to convert_mult_to_fma_1.
20249         (math_opts_dom_walker): New type.
20250         (math_opts_dom_walker::after_dom_children): New method, body moved
20251         here from pass_optimize_widening_mul::execute, added deferring logic
20252         bits.
20253         (pass_optimize_widening_mul::execute): Moved most of code to
20254         math_opts_dom_walker::after_dom_children.
20255         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20256         * config/i386/i386.c (ix86_option_override_internal): Added
20257         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20259 2018-01-12  Richard Biener  <rguenther@suse.de>
20261         PR debug/83157
20262         * dwarf2out.c (gen_variable_die): Do not reset old_die for
20263         inline instance vars.
20265 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
20267         PR target/81819
20268         * config/rx/rx.c (rx_is_restricted_memory_address):
20269         Handle SUBREG case.
20271 2018-01-12  Richard Biener  <rguenther@suse.de>
20273         PR tree-optimization/80846
20274         * target.def (split_reduction): New target hook.
20275         * targhooks.c (default_split_reduction): New function.
20276         * targhooks.h (default_split_reduction): Declare.
20277         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20278         target requests first reduce vectors by combining low and high
20279         parts.
20280         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20281         (get_vectype_for_scalar_type_and_size): Export.
20282         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20283         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20284         * doc/tm.texi: Regenerate.
20285         * config/i386/i386.c (ix86_split_reduction): Implement
20286         TARGET_VECTORIZE_SPLIT_REDUCTION.
20288 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
20290         PR target/83368
20291         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20292         in PIC mode except for TARGET_VXWORKS_RTP.
20293         * config/sparc/sparc.c: Include cfgrtl.h.
20294         (TARGET_INIT_PIC_REG): Define.
20295         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20296         (sparc_pic_register_p): New predicate.
20297         (sparc_legitimate_address_p): Use it.
20298         (sparc_legitimize_pic_address): Likewise.
20299         (sparc_delegitimize_address): Likewise.
20300         (sparc_mode_dependent_address_p): Likewise.
20301         (gen_load_pcrel_sym): Remove 4th parameter.
20302         (load_got_register): Adjust call to above.  Remove obsolete stuff.
20303         (sparc_expand_prologue): Do not call load_got_register here.
20304         (sparc_flat_expand_prologue): Likewise.
20305         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20306         (sparc_use_pseudo_pic_reg): New function.
20307         (sparc_init_pic_reg): Likewise.
20308         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20309         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20311 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
20313         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20314         Add item for branch_cost.
20316 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
20318         PR rtl-optimization/83565
20319         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20320         not extend the result to a larger mode for rotate operations.
20321         (num_sign_bit_copies1): Likewise.
20323 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20325         PR target/40411
20326         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20327         -symbolic.
20328         Use values-Xc.o for -pedantic.
20329         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20331 2018-01-12  Martin Liska  <mliska@suse.cz>
20333         PR ipa/83054
20334         * ipa-devirt.c (final_warning_record::grow_type_warnings):
20335         New function.
20336         (possible_polymorphic_call_targets): Use it.
20337         (ipa_devirt): Likewise.
20339 2018-01-12  Martin Liska  <mliska@suse.cz>
20341         * profile-count.h (enum profile_quality): Use 0 as invalid
20342         enum value of profile_quality.
20344 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
20346         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20347         -mext-string options.
20349 2018-01-12  Richard Biener  <rguenther@suse.de>
20351         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20352         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20353         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20354         Likewise.
20355         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20357 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
20359         * configure.ac (--with-long-double-format): Add support for the
20360         configuration option to change the default long double format on
20361         PowerPC systems.
20362         * config.gcc (powerpc*-linux*-*): Likewise.
20363         * configure: Regenerate.
20364         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20365         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20366         used without modification.
20368 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20370         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20371         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20372         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20373         MISC_BUILTIN_SPEC_BARRIER.
20374         (rs6000_init_builtins): Likewise.
20375         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20376         enum value.
20377         (speculation_barrier): New define_insn.
20378         * doc/extend.texi: Document __builtin_speculation_barrier.
20380 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
20382         PR target/83203
20383         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20384         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20385         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20386         iterators.
20387         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
20388         integral modes instead of "ss" and "sd".
20389         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20390         vectors with 32-bit and 64-bit elements.
20391         (vecdupssescalarmodesuffix): New mode attribute.
20392         (vec_dup<mode>): Use it.
20394 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
20396         PR target/83330
20397         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20398         frame if argument is passed on stack.
20400 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
20402         PR target/82682
20403         * ree.c (combine_reaching_defs): Optimize also
20404         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20405         reg2=any_extend(exp); reg1=reg2;, formatting fix.
20407 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
20409         PR middle-end/83189
20410         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20412 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
20414         PR middle-end/83718
20415         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20416         after they are computed.
20418 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
20420         PR tree-optimization/83695
20421         * gimple-loop-linterchange.cc
20422         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20423         reset cached scev information after interchange.
20424         (pass_linterchange::execute): Remove call to scev_reset_htab.
20426 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20428         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20429         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20430         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20431         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20432         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20433         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20434         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20435         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20436         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20437         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20438         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20439         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20440         (V_lane_reg): Likewise.
20441         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20442         New define_expand.
20443         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20444         (vfmal_lane_low<mode>_intrinsic,
20445         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20446         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20447         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20448         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20449         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20450         vfmsl_lane_high<mode>_intrinsic): New define_insns.
20452 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20454         * config/arm/arm-cpus.in (fp16fml): New feature.
20455         (ALL_SIMD): Add fp16fml.
20456         (armv8.2-a): Add fp16fml as an option.
20457         (armv8.3-a): Likewise.
20458         (armv8.4-a): Add fp16fml as part of fp16.
20459         * config/arm/arm.h (TARGET_FP16FML): Define.
20460         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20461         when appropriate.
20462         * config/arm/arm-modes.def (V2HF): Define.
20463         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20464         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20465         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20466         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20467         vfmsl_low, vfmsl_high): New set of builtins.
20468         * config/arm/iterators.md (PLUSMINUS): New code iterator.
20469         (vfml_op): New code attribute.
20470         (VFMLHALVES): New int iterator.
20471         (VFML, VFMLSEL): New mode attributes.
20472         (V_reg): Define mapping for V2HF.
20473         (V_hi, V_lo): New mode attributes.
20474         (VF_constraint): Likewise.
20475         (vfml_half, vfml_half_selector): New int attributes.
20476         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20477         define_expand.
20478         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20479         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20480         New define_insn.
20481         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20482         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20483         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20484         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
20485         documentation.
20486         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20487         Document new effective target and option set.
20489 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20491         * config/arm/arm-cpus.in (armv8_4): New feature.
20492         (ARMv8_4a): New fgroup.
20493         (armv8.4-a): New arch.
20494         * config/arm/arm-tables.opt: Regenerate.
20495         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20496         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20497         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20498         Add matching rules for -march=armv8.4-a and extensions.
20499         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20501 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
20503         PR target/81821
20504         * config/rx/rx.md (BW): New mode attribute.
20505         (sync_lock_test_and_setsi): Add mode suffix to insn output.
20507 2018-01-11  Richard Biener  <rguenther@suse.de>
20509         PR tree-optimization/83435
20510         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20511         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20512         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20514 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20515             Alan Hayward  <alan.hayward@arm.com>
20516             David Sherwood  <david.sherwood@arm.com>
20518         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20519         field.
20520         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
20521         (aarch64_print_address_internal): Use it to check for a zero offset.
20523 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20524             Alan Hayward  <alan.hayward@arm.com>
20525             David Sherwood  <david.sherwood@arm.com>
20527         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20528         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20529         Return a poly_int64 rather than a HOST_WIDE_INT.
20530         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20531         rather than a HOST_WIDE_INT.
20532         * config/aarch64/aarch64.h (aarch64_frame): Protect with
20533         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
20534         hard_fp_offset, frame_size, initial_adjust, callee_offset and
20535         final_offset from HOST_WIDE_INT to poly_int64.
20536         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20537         to_constant when getting the number of units in an Advanced SIMD
20538         mode.
20539         (aarch64_builtin_vectorized_function): Check for a constant number
20540         of units.
20541         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20542         GET_MODE_SIZE.
20543         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20544         attribute instead of GET_MODE_NUNITS.
20545         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20546         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20547         GET_MODE_SIZE for fixed-size registers.
20548         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20549         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20550         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20551         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20552         (aarch64_print_operand, aarch64_print_address_internal)
20553         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20554         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20555         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20556         Handle polynomial GET_MODE_SIZE.
20557         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
20558         wider than SImode without modification.
20559         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20560         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20561         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20562         passing and returning SVE modes.
20563         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20564         rather than GEN_INT.
20565         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20566         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20567         (aarch64_allocate_and_probe_stack_space): Likewise.
20568         (aarch64_layout_frame): Cope with polynomial offsets.
20569         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20570         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
20571         polynomial offsets.
20572         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20573         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20574         poly_int64 rather than a HOST_WIDE_INT.
20575         (aarch64_get_separate_components, aarch64_process_components)
20576         (aarch64_expand_prologue, aarch64_expand_epilogue)
20577         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20578         (aarch64_anchor_offset): New function, split out from...
20579         (aarch64_legitimize_address): ...here.
20580         (aarch64_builtin_vectorization_cost): Handle polynomial
20581         TYPE_VECTOR_SUBPARTS.
20582         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20583         GET_MODE_NUNITS.
20584         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20585         number of elements from the PARALLEL rather than the mode.
20586         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20587         rather than GET_MODE_BITSIZE.
20588         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20589         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20590         (aarch64_expand_vec_perm_const_1): Handle polynomial
20591         d->perm.length () and d->perm elements.
20592         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
20593         Apply to_constant to d->perm elements.
20594         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20595         polynomial CONST_VECTOR_NUNITS.
20596         (aarch64_move_pointer): Take amount as a poly_int64 rather
20597         than an int.
20598         (aarch64_progress_pointer): Avoid temporary variable.
20599         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20600         the mode attribute instead of GET_MODE.
20602 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20603             Alan Hayward  <alan.hayward@arm.com>
20604             David Sherwood  <david.sherwood@arm.com>
20606         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20607         x exists before using it.
20608         (aarch64_add_constant_internal): Rename to...
20609         (aarch64_add_offset_1): ...this.  Replace regnum with separate
20610         src and dest rtxes.  Handle the case in which they're different,
20611         including when the offset is zero.  Replace scratchreg with an rtx.
20612         Use 2 additions if there is no spare register into which we can
20613         move a 16-bit constant.
20614         (aarch64_add_constant): Delete.
20615         (aarch64_add_offset): Replace reg with separate src and dest
20616         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
20617         Use aarch64_add_offset_1.
20618         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20619         an rtx rather than an int.  Take the delta as a poly_int64
20620         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
20621         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20622         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20623         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20624         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20625         and aarch64_add_sp.
20626         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20627         aarch64_add_constant.
20629 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20631         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20632         Use scalar_float_mode.
20634 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20636         * config/aarch64/aarch64-simd.md
20637         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20638         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20639         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20640         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20641         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20642         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20643         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20644         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20645         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20646         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20648 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20650         PR target/83514
20651         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20652         targ_options->x_arm_arch_string is non NULL.
20654 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
20656         * config/aarch64/aarch64.h
20657         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
20659 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
20661         PR target/82096
20662         * expmed.c (emit_store_flag_force): Swap if const op0
20663         and change VOIDmode to mode of op0.
20665 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
20667         PR rtl-optimization/83761
20668         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20669         than bytes to mode_for_size.
20671 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
20673         PR middle-end/83189
20674         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20675         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20676         profile.
20678 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
20680         PR middle-end/83575
20681         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20682         when in layout mode.
20683         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20684         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20685         partition fixup.
20687 2018-01-10  Michael Collison  <michael.collison@arm.com>
20689         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20690         * config/aarch64/aarch64-option-extension.def: Add
20691         AARCH64_OPT_EXTENSION of 'fp16fml'.
20692         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20693         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20694         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20695         * config/aarch64/constraints.md (Ui7): New constraint.
20696         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20697         (VFMLA_SEL_W): Ditto.
20698         (f16quad): Ditto.
20699         (f16mac1): Ditto.
20700         (VFMLA16_LOW): New int iterator.
20701         (VFMLA16_HIGH): Ditto.
20702         (UNSPEC_FMLAL): New unspec.
20703         (UNSPEC_FMLSL): Ditto.
20704         (UNSPEC_FMLAL2): Ditto.
20705         (UNSPEC_FMLSL2): Ditto.
20706         (f16mac): New code attribute.
20707         * config/aarch64/aarch64-simd-builtins.def
20708         (aarch64_fmlal_lowv2sf): Ditto.
20709         (aarch64_fmlsl_lowv2sf): Ditto.
20710         (aarch64_fmlalq_lowv4sf): Ditto.
20711         (aarch64_fmlslq_lowv4sf): Ditto.
20712         (aarch64_fmlal_highv2sf): Ditto.
20713         (aarch64_fmlsl_highv2sf): Ditto.
20714         (aarch64_fmlalq_highv4sf): Ditto.
20715         (aarch64_fmlslq_highv4sf): Ditto.
20716         (aarch64_fmlal_lane_lowv2sf): Ditto.
20717         (aarch64_fmlsl_lane_lowv2sf): Ditto.
20718         (aarch64_fmlal_laneq_lowv2sf): Ditto.
20719         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20720         (aarch64_fmlalq_lane_lowv4sf): Ditto.
20721         (aarch64_fmlsl_lane_lowv4sf): Ditto.
20722         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20723         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20724         (aarch64_fmlal_lane_highv2sf): Ditto.
20725         (aarch64_fmlsl_lane_highv2sf): Ditto.
20726         (aarch64_fmlal_laneq_highv2sf): Ditto.
20727         (aarch64_fmlsl_laneq_highv2sf): Ditto.
20728         (aarch64_fmlalq_lane_highv4sf): Ditto.
20729         (aarch64_fmlsl_lane_highv4sf): Ditto.
20730         (aarch64_fmlalq_laneq_highv4sf): Ditto.
20731         (aarch64_fmlsl_laneq_highv4sf): Ditto.
20732         * config/aarch64/aarch64-simd.md:
20733         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20734         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20735         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20736         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20737         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20738         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20739         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20740         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20741         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20742         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20743         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20744         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20745         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20746         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20747         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20748         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20749         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20750         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20751         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20752         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20753         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20754         (vfmlsl_low_u32): Ditto.
20755         (vfmlalq_low_u32): Ditto.
20756         (vfmlslq_low_u32): Ditto.
20757         (vfmlal_high_u32): Ditto.
20758         (vfmlsl_high_u32): Ditto.
20759         (vfmlalq_high_u32): Ditto.
20760         (vfmlslq_high_u32): Ditto.
20761         (vfmlal_lane_low_u32): Ditto.
20762         (vfmlsl_lane_low_u32): Ditto.
20763         (vfmlal_laneq_low_u32): Ditto.
20764         (vfmlsl_laneq_low_u32): Ditto.
20765         (vfmlalq_lane_low_u32): Ditto.
20766         (vfmlslq_lane_low_u32): Ditto.
20767         (vfmlalq_laneq_low_u32): Ditto.
20768         (vfmlslq_laneq_low_u32): Ditto.
20769         (vfmlal_lane_high_u32): Ditto.
20770         (vfmlsl_lane_high_u32): Ditto.
20771         (vfmlal_laneq_high_u32): Ditto.
20772         (vfmlsl_laneq_high_u32): Ditto.
20773         (vfmlalq_lane_high_u32): Ditto.
20774         (vfmlslq_lane_high_u32): Ditto.
20775         (vfmlalq_laneq_high_u32): Ditto.
20776         (vfmlslq_laneq_high_u32): Ditto.
20777         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20778         (AARCH64_FL_FOR_ARCH8_4): New.
20779         (AARCH64_ISA_F16FML): New ISA flag.
20780         (TARGET_F16FML): New feature flag for fp16fml.
20781         (doc/invoke.texi): Document new fp16fml option.
20783 2018-01-10  Michael Collison  <michael.collison@arm.com>
20785         * config/aarch64/aarch64-builtins.c:
20786         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20787         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20788         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20789         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20790         (AARCH64_ISA_SHA3): New ISA flag.
20791         (TARGET_SHA3): New feature flag for sha3.
20792         * config/aarch64/iterators.md (sha512_op): New int attribute.
20793         (CRYPTO_SHA512): New int iterator.
20794         (UNSPEC_SHA512H): New unspec.
20795         (UNSPEC_SHA512H2): Ditto.
20796         (UNSPEC_SHA512SU0): Ditto.
20797         (UNSPEC_SHA512SU1): Ditto.
20798         * config/aarch64/aarch64-simd-builtins.def
20799         (aarch64_crypto_sha512hqv2di): New builtin.
20800         (aarch64_crypto_sha512h2qv2di): Ditto.
20801         (aarch64_crypto_sha512su0qv2di): Ditto.
20802         (aarch64_crypto_sha512su1qv2di): Ditto.
20803         (aarch64_eor3qv8hi): Ditto.
20804         (aarch64_rax1qv2di): Ditto.
20805         (aarch64_xarqv2di): Ditto.
20806         (aarch64_bcaxqv8hi): Ditto.
20807         * config/aarch64/aarch64-simd.md:
20808         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20809         (aarch64_crypto_sha512su0qv2di): Ditto.
20810         (aarch64_crypto_sha512su1qv2di): Ditto.
20811         (aarch64_eor3qv8hi): Ditto.
20812         (aarch64_rax1qv2di): Ditto.
20813         (aarch64_xarqv2di): Ditto.
20814         (aarch64_bcaxqv8hi): Ditto.
20815         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20816         (vsha512h2q_u64): Ditto.
20817         (vsha512su0q_u64): Ditto.
20818         (vsha512su1q_u64): Ditto.
20819         (veor3q_u16): Ditto.
20820         (vrax1q_u64): Ditto.
20821         (vxarq_u64): Ditto.
20822         (vbcaxq_u16): Ditto.
20823         * config/arm/types.md (crypto_sha512): New type attribute.
20824         (crypto_sha3): Ditto.
20825         (doc/invoke.texi): Document new sha3 option.
20827 2018-01-10  Michael Collison  <michael.collison@arm.com>
20829         * config/aarch64/aarch64-builtins.c:
20830         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20831         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20832         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20833         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20834         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20835         (AARCH64_ISA_SM4): New ISA flag.
20836         (TARGET_SM4): New feature flag for sm4.
20837         * config/aarch64/aarch64-simd-builtins.def
20838         (aarch64_sm3ss1qv4si): Ditto.
20839         (aarch64_sm3tt1aq4si): Ditto.
20840         (aarch64_sm3tt1bq4si): Ditto.
20841         (aarch64_sm3tt2aq4si): Ditto.
20842         (aarch64_sm3tt2bq4si): Ditto.
20843         (aarch64_sm3partw1qv4si): Ditto.
20844         (aarch64_sm3partw2qv4si): Ditto.
20845         (aarch64_sm4eqv4si): Ditto.
20846         (aarch64_sm4ekeyqv4si): Ditto.
20847         * config/aarch64/aarch64-simd.md:
20848         (aarch64_sm3ss1qv4si): Ditto.
20849         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20850         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20851         (aarch64_sm4eqv4si): Ditto.
20852         (aarch64_sm4ekeyqv4si): Ditto.
20853         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20854         (sm3part_op): Ditto.
20855         (CRYPTO_SM3TT): Ditto.
20856         (CRYPTO_SM3PART): Ditto.
20857         (UNSPEC_SM3SS1): New unspec.
20858         (UNSPEC_SM3TT1A): Ditto.
20859         (UNSPEC_SM3TT1B): Ditto.
20860         (UNSPEC_SM3TT2A): Ditto.
20861         (UNSPEC_SM3TT2B): Ditto.
20862         (UNSPEC_SM3PARTW1): Ditto.
20863         (UNSPEC_SM3PARTW2): Ditto.
20864         (UNSPEC_SM4E): Ditto.
20865         (UNSPEC_SM4EKEY): Ditto.
20866         * config/aarch64/constraints.md (Ui2): New constraint.
20867         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20868         * config/arm/types.md (crypto_sm3): New type attribute.
20869         (crypto_sm4): Ditto.
20870         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20871         (vsm3tt1aq_u32): Ditto.
20872         (vsm3tt1bq_u32): Ditto.
20873         (vsm3tt2aq_u32): Ditto.
20874         (vsm3tt2bq_u32): Ditto.
20875         (vsm3partw1q_u32): Ditto.
20876         (vsm3partw2q_u32): Ditto.
20877         (vsm4eq_u32): Ditto.
20878         (vsm4ekeyq_u32): Ditto.
20879         (doc/invoke.texi): Document new sm4 option.
20881 2018-01-10  Michael Collison  <michael.collison@arm.com>
20883         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20884         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20885         (AARCH64_FL_FOR_ARCH8_4): New.
20886         (AARCH64_FL_V8_4): New flag.
20887         (doc/invoke.texi): Document new armv8.4-a option.
20889 2018-01-10  Michael Collison  <michael.collison@arm.com>
20891         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20892         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
20893         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
20894         * config/aarch64/aarch64-option-extension.def: Add
20895         AARCH64_OPT_EXTENSION of 'sha2'.
20896         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
20897         (crypto): Disable sha2 and aes if crypto disabled.
20898         (crypto): Enable aes and sha2 if enabled.
20899         (simd): Disable sha2 and aes if simd disabled.
20900         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
20901         New flags.
20902         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
20903         (TARGET_SHA2): New feature flag for sha2.
20904         (TARGET_AES): New feature flag for aes.
20905         * config/aarch64/aarch64-simd.md:
20906         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
20907         conditional on TARGET_AES.
20908         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20909         (aarch64_crypto_sha1hsi): Make pattern conditional
20910         on TARGET_SHA2.
20911         (aarch64_crypto_sha1hv4si): Ditto.
20912         (aarch64_be_crypto_sha1hv4si): Ditto.
20913         (aarch64_crypto_sha1su1v4si): Ditto.
20914         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20915         (aarch64_crypto_sha1su0v4si): Ditto.
20916         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20917         (aarch64_crypto_sha256su0v4si): Ditto.
20918         (aarch64_crypto_sha256su1v4si): Ditto.
20919         (doc/invoke.texi): Document new aes and sha2 options.
20921 2018-01-10  Martin Sebor  <msebor@redhat.com>
20923         PR tree-optimization/83781
20924         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
20925         as string arrays.
20927 2018-01-11  Martin Sebor  <msebor@gmail.com>
20928             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20930         PR tree-optimization/83501
20931         PR tree-optimization/81703
20933         * tree-ssa-strlen.c (get_string_cst): Rename...
20934         (get_string_len): ...to this.  Handle global constants.
20935         (handle_char_store): Adjust.
20937 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
20938             Jim Wilson  <jimw@sifive.com>
20940         * config/riscv/riscv-protos.h (riscv_output_return): New.
20941         * config/riscv/riscv.c (struct machine_function): New naked_p field.
20942         (riscv_attribute_table, riscv_output_return),
20943         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
20944         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
20945         (riscv_compute_frame_info): Only compute frame->mask if not a naked
20946         function.
20947         (riscv_expand_prologue): Add early return for naked function.
20948         (riscv_expand_epilogue): Likewise.
20949         (riscv_function_ok_for_sibcall): Return false for naked function.
20950         (riscv_set_current_function): New.
20951         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
20952         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
20953         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
20954         * doc/extend.texi (RISC-V Function Attributes): New.
20956 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
20958         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
20959         check for 128-bit long double before checking TCmode.
20960         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
20961         128-bit long doubles before checking TFmode or TCmode.
20962         (FLOAT128_IBM_P): Likewise.
20964 2018-01-10  Martin Sebor  <msebor@redhat.com>
20966         PR tree-optimization/83671
20967         * builtins.c (c_strlen): Unconditionally return zero for the empty
20968         string.
20969         Use -Warray-bounds for warnings.
20970         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
20971         for non-constant array indices with COMPONENT_REF, arrays of
20972         arrays, and pointers to arrays.
20973         (gimple_fold_builtin_strlen): Determine and set length range for
20974         non-constant character arrays.
20976 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
20978         PR middle-end/81897
20979         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
20980         empty blocks.
20982 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
20984         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
20986 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
20988         PR target/83399
20989         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
20990         VECTOR_MEM_ALTIVEC_OR_VSX_P.
20991         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
20992         indexed_or_indirect_operand predicate.
20993         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
20994         (*vsx_le_perm_load_v8hi): Likewise.
20995         (*vsx_le_perm_load_v16qi): Likewise.
20996         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
20997         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
20998         (*vsx_le_perm_store_v8hi): Likewise.
20999         (*vsx_le_perm_store_v16qi): Likewise.
21000         (eight unnamed splitters): Likewise.
21002 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
21004         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21005         * config/rs6000/emmintrin.h: Likewise.
21006         * config/rs6000/mmintrin.h: Likewise.
21007         * config/rs6000/xmmintrin.h: Likewise.
21009 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
21011         PR c++/43486
21012         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21013         "public_flag".
21014         * tree.c (tree_nop_conversion): Return true for location wrapper
21015         nodes.
21016         (maybe_wrap_with_location): New function.
21017         (selftest::check_strip_nops): New function.
21018         (selftest::test_location_wrappers): New function.
21019         (selftest::tree_c_tests): Call it.
21020         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21021         (maybe_wrap_with_location): New decl.
21022         (EXPR_LOCATION_WRAPPER_P): New macro.
21023         (location_wrapper_p): New inline function.
21024         (tree_strip_any_location_wrapper): New inline function.
21026 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
21028         PR target/83735
21029         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21030         stack_realign_offset for the largest alignment of stack slot
21031         actually used.
21032         (ix86_find_max_used_stack_alignment): New function.
21033         (ix86_finalize_stack_frame_flags): Use it.  Set
21034         max_used_stack_alignment if we don't realign stack.
21035         * config/i386/i386.h (machine_function): Add
21036         max_used_stack_alignment.
21038 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
21040         * config/arm/arm.opt (-mbranch-cost): New option.
21041         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21042         account.
21044 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
21046         PR target/83629
21047         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21048         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21050 2018-01-10  Richard Biener  <rguenther@suse.de>
21052         PR debug/83765
21053         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21054         early out so it also covers the case where we have a non-NULL
21055         origin.
21057 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
21059         PR tree-optimization/83753
21060         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21061         for non-strided grouped accesses if the number of elements is 1.
21063 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
21065         PR target/81616
21066         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21067         * i386.h (TARGET_USE_GATHER): Define.
21068         * x86-tune.def (X86_TUNE_USE_GATHER): New.
21070 2018-01-10  Martin Liska  <mliska@suse.cz>
21072         PR bootstrap/82831
21073         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21074         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21075         partitioning.
21076         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21077         CLEANUP_NO_PARTITIONING is not set.
21079 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
21081         * doc/rtl.texi: Remove documentation of (const ...) wrappers
21082         for vectors, as a partial revert of r254296.
21083         * rtl.h (const_vec_p): Delete.
21084         (const_vec_duplicate_p): Don't test for vector CONSTs.
21085         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21086         * expmed.c (make_tree): Likewise.
21088         Revert:
21089         * common.md (E, F): Use CONSTANT_P instead of checking for
21090         CONST_VECTOR.
21091         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21092         checking for CONST_VECTOR.
21094 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
21096         PR middle-end/83575
21097         * predict.c (force_edge_cold): Handle in more sane way edges
21098         with no prediction.
21100 2018-01-09  Carl Love  <cel@us.ibm.com>
21102         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21103         V4SI, V4SF types.
21104         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21105         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21106         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21107         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
21108         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21109         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
21110         * config/rs6000/rs6000-protos.h: Add extern defition for
21111         rs6000_generate_float2_double_code.
21112         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21113         function.
21114         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21115         (float2_v2df): Add define_expand.
21117 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
21119         PR target/83628
21120         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21121         op_mode in the force_to_mode call.
21123 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
21125         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21126         instead of checking each element individually.
21127         (aarch64_evpc_uzp): Likewise.
21128         (aarch64_evpc_zip): Likewise.
21129         (aarch64_evpc_ext): Likewise.
21130         (aarch64_evpc_rev): Likewise.
21131         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21132         instead of checking each element individually.  Return true without
21133         generating rtl if
21134         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21135         whether all selected elements come from the same input, instead of
21136         checking each element individually.  Remove calls to gen_rtx_REG,
21137         start_sequence and end_sequence and instead assert that no rtl is
21138         generated.
21140 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
21142         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21143         order of HIGH and CONST checks.
21145 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
21147         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21148         if the destination isn't an SSA_NAME.
21150 2018-01-09  Richard Biener  <rguenther@suse.de>
21152         PR tree-optimization/83668
21153         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21154         move prologue...
21155         (canonicalize_loop_form): ... here, renamed from ...
21156         (canonicalize_loop_closed_ssa_form): ... this and amended to
21157         swap successor edges for loop exit blocks to make us use
21158         the RPO order we need for initial schedule generation.
21160 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
21162         PR tree-optimization/64811
21163         * match.pd: When optimizing comparisons with Inf, avoid
21164         introducing or losing exceptions from comparisons with NaN.
21166 2018-01-09  Martin Liska  <mliska@suse.cz>
21168         PR sanitizer/82517
21169         * asan.c (shadow_mem_size): Add gcc_assert.
21171 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
21173         Don't save registers in main().
21175         PR target/83738
21176         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21177         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21178         * config/avr/avr.c (avr_set_current_function): Don't error if
21179         naked, OS_task or OS_main are specified at the same time.
21180         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21181         OS_main.
21182         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21183         attribute.
21184         * common/config/avr/avr-common.c (avr_option_optimization_table):
21185         Switch on -mmain-is-OS_task for optimizing compilations.
21187 2018-01-09  Richard Biener  <rguenther@suse.de>
21189         PR tree-optimization/83572
21190         * graphite.c: Include cfganal.h.
21191         (graphite_transform_loops): Connect infinite loops to exit
21192         and remove fake edges at the end.
21194 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
21196         * ipa-inline.c (edge_badness): Revert accidental checkin.
21198 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
21200         PR ipa/80763
21201         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21202         symbols; not inline clones.
21204 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
21206         PR target/83507
21207         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21208         hard registers.  Formatting fixes.
21210         PR preprocessor/83722
21211         * gcc.c (try_generate_repro): Pass
21212         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21213         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21214         do_report_bug.
21216 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
21217             Kito Cheng  <kito.cheng@gmail.com>
21219         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21220         (riscv_leaf_function_p): Delete.
21221         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21223 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21225         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21226         function.
21227         (do_ifelse): New function.
21228         (do_isel): New function.
21229         (do_sub3): New function.
21230         (do_add3): New function.
21231         (do_load_mask_compare): New function.
21232         (do_overlap_load_compare): New function.
21233         (expand_compare_loop): New function.
21234         (expand_block_compare): Call expand_compare_loop() when appropriate.
21235         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21236         option description.
21237         (-mblock-compare-inline-loop-limit): New option.
21239 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21241         PR target/83677
21242         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21243         Reverse order of second and third operands in first alternative.
21244         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21245         of first and second elements in UNSPEC_VPERMR vector.
21246         (altivec_expand_vec_perm_le): Likewise.
21248 2018-01-08  Jeff Law  <law@redhat.com>
21250         PR rtl-optimizatin/81308
21251         * tree-switch-conversion.c (cfg_altered): New file scoped static.
21252         (process_switch): If group_case_labels makes a change, then set
21253         cfg_altered.
21254         (pass_convert_switch::execute): If a switch is converted, then
21255         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
21257         PR rtl-optimization/81308
21258         * recog.c (split_all_insns): Conditionally cleanup the CFG after
21259         splitting insns.
21261 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
21263         PR target/83663 - Revert r255946
21264         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21265         generation for cases where splatting a value is not useful.
21266         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21267         across a vec_duplicate and a paradoxical subreg forming a vector
21268         mode to a vec_concat.
21270 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21272         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21273         -march=armv8.3-a variants.
21274         * config/arm/t-multilib: Likewise.
21275         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
21277 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21279         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21280         to generate rtl.
21281         (cceq_ior_compare_complement): Give it a name so I can use it, and
21282         change boolean_or_operator predicate to boolean_operator so it can
21283         be used to generate a crand.
21284         (eqne): New code iterator.
21285         (bd/bd_neg): New code_attrs.
21286         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21287         a single define_insn.
21288         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21289         decrement (bdnzt/bdnzf/bdzt/bdzf).
21290         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21291         with the new names of the branch decrement patterns, and added the
21292         names of the branch decrement conditional patterns.
21294 2018-01-08  Richard Biener  <rguenther@suse.de>
21296         PR tree-optimization/83563
21297         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21298         cache.
21300 2018-01-08  Richard Biener  <rguenther@suse.de>
21302         PR middle-end/83713
21303         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21305 2018-01-08  Richard Biener  <rguenther@suse.de>
21307         PR tree-optimization/83685
21308         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21309         references to abnormals.
21311 2018-01-08  Richard Biener  <rguenther@suse.de>
21313         PR lto/83719
21314         * dwarf2out.c (output_indirect_strings): Handle empty
21315         skeleton_debug_str_hash.
21316         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21318 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
21320         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21321         (emit_store_direct): Likewise.
21322         (arc_trampoline_adjust_address): Likewise.
21323         (arc_asm_trampoline_template): New function.
21324         (arc_initialize_trampoline): Use asm_trampoline_template.
21325         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21326         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21327         * config/arc/arc.md (flush_icache): Delete pattern.
21329 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
21331         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21332         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21333         munaligned-access.
21335 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
21337         PR target/83681
21338         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21339         by not USED_FOR_TARGET.
21340         (make_pass_resolve_sw_modes): Likewise.
21342 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
21344         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21345         USED_FOR_TARGET.
21347 2018-01-08  Richard Biener  <rguenther@suse.de>
21349         PR middle-end/83580
21350         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21352 2018-01-08  Richard Biener  <rguenther@suse.de>
21354         PR middle-end/83517
21355         * match.pd ((t * 2) / 2) -> t): Add missing :c.
21357 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
21359         PR middle-end/81897
21360         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21361         basic blocks with a small number of successors.
21362         (convert_control_dep_chain_into_preds): Improve handling of
21363         forwarder blocks.
21364         (dump_predicates): Split apart into...
21365         (dump_pred_chain): ...here...
21366         (dump_pred_info): ...and here.
21367         (can_one_predicate_be_invalidated_p): Add debugging printfs.
21368         (can_chain_union_be_invalidated_p): Improve check for invalidation
21369         of paths.
21370         (uninit_uses_cannot_happen): Avoid unnecessary if
21371         convert_control_dep_chain_into_preds yielded nothing.
21373 2018-01-06  Martin Sebor  <msebor@redhat.com>
21375         PR tree-optimization/83640
21376         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21377         subtracting negative offset from size.
21378         (builtin_access::overlap): Adjust offset bounds of the access to fall
21379         within the size of the object if possible.
21381 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
21383         PR rtl-optimization/83699
21384         * expmed.c (extract_bit_field_1): Restrict the vector usage of
21385         extract_bit_field_as_subreg to cases in which the extracted
21386         value is also a vector.
21388         * lra-constraints.c (process_alt_operands): Test for the equivalence
21389         substitutions when detecting a possible reload cycle.
21391 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
21393         PR debug/83480
21394         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21395         by default if flag_selective_schedling{,2}.  Formatting fixes.
21397         PR rtl-optimization/83682
21398         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21399         if it has non-VECTOR_MODE element mode.
21400         (vec_duplicate_p): Likewise.
21402         PR middle-end/83694
21403         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21404         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21406 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
21408         PR target/83604
21409         * config/i386/i386-builtin.def
21410         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21411         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21412         Require also OPTION_MASK_ISA_AVX512F in addition to
21413         OPTION_MASK_ISA_GFNI.
21414         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21415         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21416         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21417         to OPTION_MASK_ISA_GFNI.
21418         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21419         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21420         OPTION_MASK_ISA_AVX512BW.
21421         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21422         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21423         addition to OPTION_MASK_ISA_GFNI.
21424         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21425         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21426         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21427         to OPTION_MASK_ISA_GFNI.
21428         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21429         a requirement for all ISAs rather than any of them with a few
21430         exceptions.
21431         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21432         processing.
21433         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21434         bitmasks to be enabled with 3 exceptions, instead of requiring any
21435         enabled ISA with lots of exceptions.
21436         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21437         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21438         Change avx512bw in isa attribute to avx512f.
21439         * config/i386/sgxintrin.h: Add license boilerplate.
21440         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
21441         to __AVX512F__ and __AVX512VL to __AVX512VL__.
21442         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21443         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21444         defined.
21445         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21446         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21447         temporarily sse2 rather than sse if not enabled already.
21449         PR target/83604
21450         * config/i386/sse.md (VI248_VLBW): Rename to ...
21451         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
21452         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21453         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21454         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21455         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21456         mode iterator instead of VI248_VLBW.
21458 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
21460         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21461         (record_modified): Skip clobbers; add debug output.
21462         (param_change_prob): Use sreal frequencies.
21464 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
21466         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21467         punt for user-aligned variables.
21469 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
21471         * tree-chrec.c (chrec_contains_symbols): Return true for
21472         POLY_INT_CST.
21474 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
21476         PR target/82439
21477         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21478         of (x|y) == x for BICS pattern.
21480 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
21482         PR tree-optimization/83605
21483         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21484         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21485         can throw.
21487 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
21489         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21490         * config/epiphany/rtems.h: New file.
21492 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
21493             Uros Bizjak  <ubizjak@gmail.com>
21495         PR target/83554
21496         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21497         QIreg_operand instead of register_operand predicate.
21498         * config/i386/i386.c (ix86_rop_should_change_byte_p,
21499         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21500         comments instead of -fmitigate[-_]rop.
21502 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21504         PR bootstrap/81926
21505         * cgraphunit.c (symbol_table::compile): Switch to text_section
21506         before calling assembly_start debug hook.
21507         * run-rtl-passes.c (run_rtl_passes): Likewise.
21508         Include output.h.
21510 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21512         * tree-vrp.c (extract_range_from_binary_expr_1): Check
21513         range_int_cst_p rather than !symbolic_range_p before calling
21514         extract_range_from_multiplicative_op_1.
21516 2018-01-04  Jeff Law  <law@redhat.com>
21518         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21519         redundant test in assertion.
21521 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21523         * doc/rtl.texi: Document machine_mode wrapper classes.
21525 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21527         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21528         using tree_to_uhwi.
21530 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21532         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21533         the VEC_PERM_EXPR fold to fail.
21535 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
21537         PR debug/83585
21538         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21539         to switched_sections.
21541 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21543         PR target/83680
21544         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21545         test for d.testing.
21547 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
21549         PR target/83387
21550         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21551         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21553 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
21555         PR debug/83666
21556         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21557         is BLKmode and bitpos not zero or mode change is needed.
21559 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
21561         PR target/83675
21562         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21563         TARGET_VIS2.
21565 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
21567         PR target/83628
21568         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21569         instead of MULT rtx.  Update all corresponding splitters.
21570         (*saddl_se): Ditto.
21571         (*ssub<modesuffix>): Ditto.
21572         (*ssubl_se): Ditto.
21573         (*cmp_sadd_di): Update split patterns.
21574         (*cmp_sadd_si): Ditto.
21575         (*cmp_sadd_sidi): Ditto.
21576         (*cmp_ssub_di): Ditto.
21577         (*cmp_ssub_si): Ditto.
21578         (*cmp_ssub_sidi): Ditto.
21579         * config/alpha/predicates.md (const23_operand): New predicate.
21580         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21581         Look for ASHIFT, not MULT inner operand.
21582         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21584 2018-01-04  Martin Liska  <mliska@suse.cz>
21586         PR gcov-profile/83669
21587         * gcov.c (output_intermediate_file): Add version to intermediate
21588         gcov file.
21589         * doc/gcov.texi: Document new field 'version' in intermediate
21590         file format. Fix location of '-k' option of gcov command.
21592 2018-01-04  Martin Liska  <mliska@suse.cz>
21594         PR ipa/82352
21595         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21597 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
21599         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21601 2018-01-03  Martin Sebor  <msebor@redhat.com>
21603         PR tree-optimization/83655
21604         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21605         checking calls with invalid arguments.
21607 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21609         * tree-vect-stmts.c (vect_get_store_rhs): New function.
21610         (vectorizable_mask_load_store): Delete.
21611         (vectorizable_call): Return false for masked loads and stores.
21612         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
21613         instead of gimple_assign_rhs1.
21614         (vectorizable_load): Handle IFN_MASK_LOAD.
21615         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21617 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21619         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21620         split out from..,
21621         (vectorizable_mask_load_store): ...here.
21622         (vectorizable_load): ...and here.
21624 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21626         * tree-vect-stmts.c (vect_build_all_ones_mask)
21627         (vect_build_zero_merge_argument): New functions, split out from...
21628         (vectorizable_load): ...here.
21630 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21632         * tree-vect-stmts.c (vect_check_store_rhs): New function,
21633         split out from...
21634         (vectorizable_mask_load_store): ...here.
21635         (vectorizable_store): ...and here.
21637 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21639         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21640         split out from...
21641         (vectorizable_mask_load_store): ...here.
21643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21645         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21646         (vect_model_store_cost): Take a vec_load_store_type instead of a
21647         vect_def_type.
21648         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21649         (vect_model_store_cost): Take a vec_load_store_type instead of a
21650         vect_def_type.
21651         (vectorizable_mask_load_store): Update accordingly.
21652         (vectorizable_store): Likewise.
21653         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21655 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21657         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21658         IFN_MASK_LOAD calls here rather than...
21659         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21661 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21662             Alan Hayward  <alan.hayward@arm.com>
21663             David Sherwood  <david.sherwood@arm.com>
21665         * expmed.c (extract_bit_field_1): For vector extracts,
21666         fall back to extract_bit_field_as_subreg if vec_extract
21667         isn't available.
21669 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21670             Alan Hayward  <alan.hayward@arm.com>
21671             David Sherwood  <david.sherwood@arm.com>
21673         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21674         they are variable or constant sized.
21675         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21676         slots for constant-sized data.
21678 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21679             Alan Hayward  <alan.hayward@arm.com>
21680             David Sherwood  <david.sherwood@arm.com>
21682         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21683         handling COND_EXPRs with boolean comparisons, try to find a better
21684         basis for the mask type than the boolean itself.
21686 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21688         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21689         is calculated and how it can be overridden.
21690         * genmodes.c (max_bitsize_mode_any_mode): New variable.
21691         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21692         if defined.
21693         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21694         if nonzero.
21696 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21697             Alan Hayward  <alan.hayward@arm.com>
21698             David Sherwood  <david.sherwood@arm.com>
21700         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21701         Remove the mode argument.
21702         (aarch64_simd_valid_immediate): Remove the mode and inverse
21703         arguments.
21704         * config/aarch64/iterators.md (bitsize): New iterator.
21705         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21706         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21707         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21708         aarch64_simd_valid_immediate.
21709         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21710         (aarch64_reg_or_bic_imm): Likewise.
21711         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21712         with an insn_type enum and msl with a modifier_type enum.
21713         Replace element_width with a scalar_mode.  Change the shift
21714         to unsigned int.  Add constructors for scalar_float_mode and
21715         scalar_int_mode elements.
21716         (aarch64_vect_float_const_representable_p): Delete.
21717         (aarch64_can_const_movi_rtx_p)
21718         (aarch64_simd_scalar_immediate_valid_for_move)
21719         (aarch64_simd_make_constant): Update call to
21720         aarch64_simd_valid_immediate.
21721         (aarch64_advsimd_valid_immediate_hs): New function.
21722         (aarch64_advsimd_valid_immediate): Likewise.
21723         (aarch64_simd_valid_immediate): Remove mode and inverse
21724         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
21725         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21726         and aarch64_float_const_representable_p on the result.
21727         (aarch64_output_simd_mov_immediate): Remove mode argument.
21728         Update call to aarch64_simd_valid_immediate and use of
21729         simd_immediate_info.
21730         (aarch64_output_scalar_simd_mov_immediate): Update call
21731         accordingly.
21733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21734             Alan Hayward  <alan.hayward@arm.com>
21735             David Sherwood  <david.sherwood@arm.com>
21737         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21738         (mode_nunits): Likewise CONST_MODE_NUNITS.
21739         * machmode.def (ADJUST_NUNITS): Document.
21740         * genmodes.c (mode_data::need_nunits_adj): New field.
21741         (blank_mode): Update accordingly.
21742         (adj_nunits): New variable.
21743         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21744         parameter.
21745         (emit_mode_size_inline): Set need_bytesize_adj for all modes
21746         listed in adj_nunits.
21747         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21748         listed in adj_nunits.  Don't emit case statements for such modes.
21749         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21750         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
21751         nothing if adj_nunits is nonnull.
21752         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21753         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21754         (emit_mode_fbit): Update use of print_maybe_const_decl.
21755         (emit_move_size): Likewise.  Treat the array as non-const
21756         if adj_nunits.
21757         (emit_mode_adjustments): Handle adj_nunits.
21759 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21761         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21762         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21763         (VECTOR_MODES): Use it.
21764         (make_vector_modes): Take the prefix as an argument.
21766 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21767             Alan Hayward  <alan.hayward@arm.com>
21768             David Sherwood  <david.sherwood@arm.com>
21770         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21771         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21772         for MODE_VECTOR_BOOL.
21773         * machmode.def (VECTOR_BOOL_MODE): Document.
21774         * genmodes.c (VECTOR_BOOL_MODE): New macro.
21775         (make_vector_bool_mode): New function.
21776         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21777         MODE_VECTOR_BOOL.
21778         * lto-streamer-in.c (lto_input_mode_table): Likewise.
21779         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21780         Likewise.
21781         * stor-layout.c (int_mode_for_mode): Likewise.
21782         * tree.c (build_vector_type_for_mode): Likewise.
21783         * varasm.c (output_constant_pool_2): Likewise.
21784         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21785         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
21786         for MODE_VECTOR_BOOL.
21787         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21788         of mode class checks.
21789         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21790         instead of a list of mode class checks.
21791         (expand_vector_scalar_condition): Likewise.
21792         (type_for_widest_vector_mode): Handle BImode as an inner mode.
21794 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21795             Alan Hayward  <alan.hayward@arm.com>
21796             David Sherwood  <david.sherwood@arm.com>
21798         * machmode.h (mode_size): Change from unsigned short to
21799         poly_uint16_pod.
21800         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21801         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21802         or if measurement_type is not polynomial.
21803         (fixed_size_mode::includes_p): Check for constant-sized modes.
21804         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21805         return a poly_uint16 rather than an unsigned short.
21806         (emit_mode_size): Change the type of mode_size from unsigned short
21807         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
21808         (emit_mode_adjustments): Cope with polynomial vector sizes.
21809         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21810         for GET_MODE_SIZE.
21811         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21812         for GET_MODE_SIZE.
21813         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21814         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21815         * caller-save.c (setup_save_areas): Likewise.
21816         (replace_reg_with_saved_mem): Likewise.
21817         * calls.c (emit_library_call_value_1): Likewise.
21818         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21819         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21820         (gen_lowpart_for_combine): Likewise.
21821         * convert.c (convert_to_integer_1): Likewise.
21822         * cse.c (equiv_constant, cse_insn): Likewise.
21823         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21824         (cselib_subst_to_values): Likewise.
21825         * dce.c (word_dce_process_block): Likewise.
21826         * df-problems.c (df_word_lr_mark_ref): Likewise.
21827         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21828         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21829         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21830         (rtl_for_decl_location): Likewise.
21831         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21832         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21833         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21834         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21835         (expand_expr_real_1): Likewise.
21836         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21837         (pad_below): Likewise.
21838         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21839         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21840         * ira.c (get_subreg_tracking_sizes): Likewise.
21841         * ira-build.c (ira_create_allocno_objects): Likewise.
21842         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21843         (ira_sort_regnos_for_alter_reg): Likewise.
21844         * ira-costs.c (record_operand_costs): Likewise.
21845         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21846         (resolve_simple_move): Likewise.
21847         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21848         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21849         (lra_constraints): Likewise.
21850         (CONST_POOL_OK_P): Reject variable-sized modes.
21851         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21852         (add_pseudo_to_slot, lra_spill): Likewise.
21853         * omp-low.c (omp_clause_aligned_alignment): Likewise.
21854         * optabs-query.c (get_best_extraction_insn): Likewise.
21855         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21856         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21857         (expand_mult_highpart, valid_multiword_target_p): Likewise.
21858         * recog.c (offsettable_address_addr_space_p): Likewise.
21859         * regcprop.c (maybe_mode_change): Likewise.
21860         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21861         * regrename.c (build_def_use): Likewise.
21862         * regstat.c (dump_reg_info): Likewise.
21863         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21864         (find_reloads, find_reloads_subreg_address): Likewise.
21865         * reload1.c (eliminate_regs_1): Likewise.
21866         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21867         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21868         (simplify_binary_operation_1, simplify_subreg): Likewise.
21869         * targhooks.c (default_function_arg_padding): Likewise.
21870         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21871         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21872         (verify_gimple_assign_ternary): Likewise.
21873         * tree-inline.c (estimate_move_cost): Likewise.
21874         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21875         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21876         (get_address_cost_ainc): Likewise.
21877         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21878         (vect_supportable_dr_alignment): Likewise.
21879         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21880         (vectorizable_reduction): Likewise.
21881         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21882         (vectorizable_operation, vectorizable_load): Likewise.
21883         * tree.c (build_same_sized_truth_vector_type): Likewise.
21884         * valtrack.c (cleanup_auto_inc_dec): Likewise.
21885         * var-tracking.c (emit_note_insn_var_location): Likewise.
21886         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21887         (ADDR_VEC_ALIGN): Likewise.
21889 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21890             Alan Hayward  <alan.hayward@arm.com>
21891             David Sherwood  <david.sherwood@arm.com>
21893         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
21894         unsigned short.
21895         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21896         or if measurement_type is polynomial.
21897         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
21898         * combine.c (make_extraction): Likewise.
21899         * dse.c (find_shift_sequence): Likewise.
21900         * dwarf2out.c (mem_loc_descriptor): Likewise.
21901         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
21902         (extract_bit_field, extract_low_bits): Likewise.
21903         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
21904         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
21905         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
21906         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
21907         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21908         * reload.c (find_reloads): Likewise.
21909         * reload1.c (alter_reg): Likewise.
21910         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21911         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21912         * tree-if-conv.c (predicate_mem_writes): Likewise.
21913         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21914         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21915         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21916         * valtrack.c (dead_debug_insert_temp): Likewise.
21917         * varasm.c (mergeable_constant_section): Likewise.
21918         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21921             Alan Hayward  <alan.hayward@arm.com>
21922             David Sherwood  <david.sherwood@arm.com>
21924         * expr.c (expand_assignment): Cope with polynomial mode sizes
21925         when assigning to a CONCAT.
21927 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21928             Alan Hayward  <alan.hayward@arm.com>
21929             David Sherwood  <david.sherwood@arm.com>
21931         * machmode.h (mode_precision): Change from unsigned short to
21932         poly_uint16_pod.
21933         (mode_to_precision): Return a poly_uint16 rather than an unsigned
21934         short.
21935         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
21936         or if measurement_type is not polynomial.
21937         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
21938         in which the mode is already known to be a scalar_int_mode.
21939         * genmodes.c (emit_mode_precision): Change the type of mode_precision
21940         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
21941         initializer.
21942         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21943         for GET_MODE_PRECISION.
21944         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21945         for GET_MODE_PRECISION.
21946         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
21947         as polynomial.
21948         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
21949         (expand_field_assignment, make_extraction): Likewise.
21950         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
21951         (get_last_value): Likewise.
21952         * convert.c (convert_to_integer_1): Likewise.
21953         * cse.c (cse_insn): Likewise.
21954         * expr.c (expand_expr_real_1): Likewise.
21955         * lra-constraints.c (simplify_operand_subreg): Likewise.
21956         * optabs-query.c (can_atomic_load_p): Likewise.
21957         * optabs.c (expand_atomic_load): Likewise.
21958         (expand_atomic_store): Likewise.
21959         * ree.c (combine_reaching_defs): Likewise.
21960         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
21961         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
21962         * tree.h (type_has_mode_precision_p): Likewise.
21963         * ubsan.c (instrument_si_overflow): Likewise.
21965 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
21966             Alan Hayward  <alan.hayward@arm.com>
21967             David Sherwood  <david.sherwood@arm.com>
21969         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
21970         polynomial numbers of units.
21971         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
21972         (valid_vector_subparts_p): New function.
21973         (build_vector_type): Remove temporary shim and take the number
21974         of units as a poly_uint64 rather than an int.
21975         (build_opaque_vector_type): Take the number of units as a
21976         poly_uint64 rather than an int.
21977         * tree.c (build_vector_from_ctor): Handle polynomial
21978         TYPE_VECTOR_SUBPARTS.
21979         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
21980         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
21981         (build_vector_from_val): If the number of units is variable,
21982         use build_vec_duplicate_cst for constant operands and
21983         VEC_DUPLICATE_EXPR otherwise.
21984         (make_vector_type): Remove temporary is_constant ().
21985         (build_vector_type, build_opaque_vector_type): Take the number of
21986         units as a poly_uint64 rather than an int.
21987         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
21988         VECTOR_CST_NELTS.
21989         * cfgexpand.c (expand_debug_expr): Likewise.
21990         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
21991         (store_constructor, expand_expr_real_1): Likewise.
21992         (const_scalar_mask_from_tree): Likewise.
21993         * fold-const-call.c (fold_const_reduction): Likewise.
21994         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
21995         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
21996         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
21997         (fold_relational_const): Likewise.
21998         (native_interpret_vector): Likewise.  Change the size from an
21999         int to an unsigned int.
22000         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22001         TYPE_VECTOR_SUBPARTS.
22002         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22003         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22004         duplicating a non-constant operand into a variable-length vector.
22005         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22006         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22007         * ipa-icf.c (sem_variable::equals): Likewise.
22008         * match.pd: Likewise.
22009         * omp-simd-clone.c (simd_clone_subparts): Likewise.
22010         * print-tree.c (print_node): Likewise.
22011         * stor-layout.c (layout_type): Likewise.
22012         * targhooks.c (default_builtin_vectorization_cost): Likewise.
22013         * tree-cfg.c (verify_gimple_comparison): Likewise.
22014         (verify_gimple_assign_binary): Likewise.
22015         (verify_gimple_assign_ternary): Likewise.
22016         (verify_gimple_assign_single): Likewise.
22017         * tree-pretty-print.c (dump_generic_node): Likewise.
22018         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22019         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22020         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22021         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22022         (vect_shift_permute_load_chain): Likewise.
22023         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22024         (expand_vector_condition, optimize_vector_constructor): Likewise.
22025         (lower_vec_perm, get_compute_type): Likewise.
22026         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22027         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22028         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22029         (vect_recog_mask_conversion_pattern): Likewise.
22030         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22031         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22032         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22033         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22034         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22035         (vectorizable_shift, vectorizable_operation, vectorizable_store)
22036         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22037         (supportable_widening_operation): Likewise.
22038         (supportable_narrowing_operation): Likewise.
22039         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22040         Likewise.
22041         * varasm.c (output_constant): 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-vect-data-refs.c (vect_permute_store_chain): Reorganize
22048         so that both the length == 3 and length != 3 cases set up their
22049         own permute vectors.  Add comments explaining why we know the
22050         number of elements is constant.
22051         (vect_permute_load_chain): Likewise.
22053 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22054             Alan Hayward  <alan.hayward@arm.com>
22055             David Sherwood  <david.sherwood@arm.com>
22057         * machmode.h (mode_nunits): Change from unsigned char to
22058         poly_uint16_pod.
22059         (ONLY_FIXED_SIZE_MODES): New macro.
22060         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22061         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22062         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22063         New typedefs.
22064         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22065         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22066         or if measurement_type is not polynomial.
22067         * genmodes.c (ZERO_COEFFS): New macro.
22068         (emit_mode_nunits_inline): Make mode_nunits_inline return a
22069         poly_uint16.
22070         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22071         Use ZERO_COEFFS when emitting initializers.
22072         * data-streamer.h (bp_pack_poly_value): New function.
22073         (bp_unpack_poly_value): Likewise.
22074         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22075         for GET_MODE_NUNITS.
22076         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22077         for GET_MODE_NUNITS.
22078         * tree.c (make_vector_type): Remove temporary shim and make
22079         the real function take the number of units as a poly_uint64
22080         rather than an int.
22081         (build_vector_type_for_mode): Handle polynomial nunits.
22082         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22083         * emit-rtl.c (const_vec_series_p_1): Likewise.
22084         (gen_rtx_CONST_VECTOR): Likewise.
22085         * fold-const.c (test_vec_duplicate_folding): Likewise.
22086         * genrecog.c (validate_pattern): Likewise.
22087         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22088         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22089         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22090         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22091         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22092         * rtlanal.c (subreg_get_info): Likewise.
22093         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22094         (vect_grouped_load_supported): Likewise.
22095         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22096         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22097         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22098         (simplify_const_unary_operation, simplify_binary_operation_1)
22099         (simplify_const_binary_operation, simplify_ternary_operation)
22100         (test_vector_ops_duplicate, test_vector_ops): Likewise.
22101         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22102         instead of CONST_VECTOR_NUNITS.
22103         * varasm.c (output_constant_pool_2): Likewise.
22104         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22105         explicit-encoded elements in the XVEC for variable-length vectors.
22107 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22109         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22112             Alan Hayward  <alan.hayward@arm.com>
22113             David Sherwood  <david.sherwood@arm.com>
22115         * coretypes.h (fixed_size_mode): Declare.
22116         (fixed_size_mode_pod): New typedef.
22117         * builtins.h (target_builtins::x_apply_args_mode)
22118         (target_builtins::x_apply_result_mode): Change type to
22119         fixed_size_mode_pod.
22120         * builtins.c (apply_args_size, apply_result_size, result_vector)
22121         (expand_builtin_apply_args_1, expand_builtin_apply)
22122         (expand_builtin_return): Update accordingly.
22124 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22126         * cse.c (hash_rtx_cb): Hash only the encoded elements.
22127         * cselib.c (cselib_hash_rtx): Likewise.
22128         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22129         CONST_VECTOR encoding.
22131 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
22132             Jeff Law  <law@redhat.com>
22134         PR target/83641
22135         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22136         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22137         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22138         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22140         PR target/83641
22141         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22142         explicitly probe *sp in a noreturn function if there were any callee
22143         register saves or frame pointer is needed.
22145 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
22147         PR debug/83621
22148         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22149         BLKmode for ternary, binary or unary expressions.
22151         PR debug/83645
22152         * var-tracking.c (delete_vta_debug_insn): New inline function.
22153         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22154         insns from get_insns () to NULL instead of each bb separately.
22155         Use delete_vta_debug_insn.  No longer static.
22156         (vt_debug_insns_local, variable_tracking_main_1): Adjust
22157         delete_vta_debug_insns callers.
22158         * rtl.h (delete_vta_debug_insns): Declare.
22159         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22160         instead of variable_tracking_main.
22162 2018-01-03  Martin Sebor  <msebor@redhat.com>
22164         PR tree-optimization/83603
22165         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22166         arguments past the endof the argument list in functions declared
22167         without a prototype.
22168         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22169         Avoid checking when arguments are null.
22171 2018-01-03  Martin Sebor  <msebor@redhat.com>
22173         PR c/83559
22174         * doc/extend.texi (attribute const): Fix a typo.
22175         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22176         issuing -Wsuggest-attribute for void functions.
22178 2018-01-03  Martin Sebor  <msebor@redhat.com>
22180         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22181         offset_int::from instead of wide_int::to_shwi.
22182         (maybe_diag_overlap): Remove assertion.
22183         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22184         * gimple-ssa-sprintf.c (format_directive): Same.
22185         (parse_directive): Same.
22186         (sprintf_dom_walker::compute_format_length): Same.
22187         (try_substitute_return_value): Same.
22189 2018-01-03  Jeff Law  <law@redhat.com>
22191         PR middle-end/83654
22192         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22193         non-constant residual for zero at runtime and avoid probing in
22194         that case.  Reorganize code for trailing problem to mirror handling
22195         of the residual.
22197 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22199         PR tree-optimization/83501
22200         * tree-ssa-strlen.c (get_string_cst): New.
22201         (handle_char_store): Call get_string_cst.
22203 2018-01-03  Martin Liska  <mliska@suse.cz>
22205         PR tree-optimization/83593
22206         * tree-ssa-strlen.c: Include tree-cfg.h.
22207         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22208         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22209         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22210         to false.
22211         (strlen_dom_walker::before_dom_children): Call
22212         gimple_purge_dead_eh_edges. Dump tranformation with details
22213         dump flags.
22214         (strlen_dom_walker::before_dom_children): Update call by adding
22215         new argument cleanup_eh.
22216         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22218 2018-01-03  Martin Liska  <mliska@suse.cz>
22220         PR ipa/83549
22221         * cif-code.def (VARIADIC_THUNK): New enum value.
22222         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22223         thunks.
22225 2018-01-03  Jan Beulich  <jbeulich@suse.com>
22227         * sse.md (mov<mode>_internal): Tighten condition for when to use
22228         vmovdqu<ssescalarsize> for TI and OI modes.
22230 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
22232         Update copyright years.
22234 2018-01-03  Martin Liska  <mliska@suse.cz>
22236         PR ipa/83594
22237         * ipa-visibility.c (function_and_variable_visibility): Skip
22238         functions with noipa attribure.
22240 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
22242         * gcc.c (process_command): Update copyright notice dates.
22243         * gcov-dump.c (print_version): Ditto.
22244         * gcov.c (print_version): Ditto.
22245         * gcov-tool.c (print_version): Ditto.
22246         * gengtype.c (create_file): Ditto.
22247         * doc/cpp.texi: Bump @copying's copyright year.
22248         * doc/cppinternals.texi: Ditto.
22249         * doc/gcc.texi: Ditto.
22250         * doc/gccint.texi: Ditto.
22251         * doc/gcov.texi: Ditto.
22252         * doc/install.texi: Ditto.
22253         * doc/invoke.texi: Ditto.
22255 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22257         * vector-builder.h (vector_builder::m_full_nelts): Change from
22258         unsigned int to poly_uint64.
22259         (vector_builder::full_nelts): Update prototype accordingly.
22260         (vector_builder::new_vector): Likewise.
22261         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22262         (vector_builder::operator ==): Likewise.
22263         (vector_builder::finalize): Likewise.
22264         * int-vector-builder.h (int_vector_builder::int_vector_builder):
22265         Take the number of elements as a poly_uint64 rather than an
22266         unsigned int.
22267         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22268         from unsigned int to poly_uint64.
22269         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22270         (vec_perm_indices::new_vector): Likewise.
22271         (vec_perm_indices::length): Likewise.
22272         (vec_perm_indices::nelts_per_input): Likewise.
22273         (vec_perm_indices::input_nelts): Likewise.
22274         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22275         number of elements per input as a poly_uint64 rather than an
22276         unsigned int.  Use the original encoding for variable-length
22277         vectors, rather than clamping each individual element.
22278         For the second and subsequent elements in each pattern,
22279         clamp the step and base before clamping their sum.
22280         (vec_perm_indices::series_p): Handle polynomial element counts.
22281         (vec_perm_indices::all_in_range_p): Likewise.
22282         (vec_perm_indices_to_tree): Likewise.
22283         (vec_perm_indices_to_rtx): Likewise.
22284         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22285         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22286         (tree_vector_builder::new_binary_operation): Handle polynomial
22287         element counts.  Return false if we need to know the number
22288         of elements at compile time.
22289         * fold-const.c (fold_vec_perm): Punt if the number of elements
22290         isn't known at compile time.
22292 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22294         * vec-perm-indices.h (vec_perm_builder): Change element type
22295         from HOST_WIDE_INT to poly_int64.
22296         (vec_perm_indices::element_type): Update accordingly.
22297         (vec_perm_indices::clamp): Handle polynomial element_types.
22298         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22299         (vec_perm_indices::all_in_range_p): Likewise.
22300         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22301         than shwi trees.
22302         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22303         polynomial vec_perm_indices element types.
22304         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22305         * fold-const.c (fold_vec_perm): Likewise.
22306         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22307         * tree-vect-generic.c (lower_vec_perm): Likewise.
22308         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22309         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22310         element type to HOST_WIDE_INT.
22312 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22313             Alan Hayward  <alan.hayward@arm.com>
22314             David Sherwood  <david.sherwood@arm.com>
22316         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22317         rather than an int.  Use plus_constant.
22318         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22319         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22321 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22322             Alan Hayward  <alan.hayward@arm.com>
22323             David Sherwood  <david.sherwood@arm.com>
22325         * calls.c (emit_call_1, expand_call): Change struct_value_size from
22326         a HOST_WIDE_INT to a poly_int64.
22328 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22329             Alan Hayward  <alan.hayward@arm.com>
22330             David Sherwood  <david.sherwood@arm.com>
22332         * calls.c (load_register_parameters): Cope with polynomial
22333         mode sizes.  Require a constant size for BLKmode parameters
22334         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
22335         forces a parameter to be padded at the lsb end in order to
22336         fill a complete number of words, require the parameter size
22337         to be ordered wrt UNITS_PER_WORD.
22339 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22340             Alan Hayward  <alan.hayward@arm.com>
22341             David Sherwood  <david.sherwood@arm.com>
22343         * reload1.c (spill_stack_slot_width): Change element type
22344         from unsigned int to poly_uint64_pod.
22345         (alter_reg): Treat mode sizes as polynomial.
22347 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22348             Alan Hayward  <alan.hayward@arm.com>
22349             David Sherwood  <david.sherwood@arm.com>
22351         * reload.c (complex_word_subreg_p): New function.
22352         (reload_inner_reg_of_subreg, push_reload): Use it.
22354 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22355             Alan Hayward  <alan.hayward@arm.com>
22356             David Sherwood  <david.sherwood@arm.com>
22358         * lra-constraints.c (process_alt_operands): Reject matched
22359         operands whose sizes aren't ordered.
22360         (match_reload): Refer to this check here.
22362 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22363             Alan Hayward  <alan.hayward@arm.com>
22364             David Sherwood  <david.sherwood@arm.com>
22366         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22367         that the mode size is in the set {1, 2, 4, 8, 16}.
22369 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22370             Alan Hayward  <alan.hayward@arm.com>
22371             David Sherwood  <david.sherwood@arm.com>
22373         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22374         Use plus_constant instead of gen_rtx_PLUS.
22376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22377             Alan Hayward  <alan.hayward@arm.com>
22378             David Sherwood  <david.sherwood@arm.com>
22380         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22381         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22382         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22383         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22384         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22385         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22386         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22387         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22388         * config/i386/i386.c (ix86_push_rounding): ...this new function.
22389         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22390         a poly_int64.
22391         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22392         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22393         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22394         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22395         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22396         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22397         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22398         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22399         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22400         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22401         function.
22402         * expr.c (emit_move_resolve_push): Treat the input and result
22403         of PUSH_ROUNDING as a poly_int64.
22404         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22405         (emit_push_insn): Likewise.
22406         * lra-eliminations.c (mark_not_eliminable): Likewise.
22407         * recog.c (push_operand): Likewise.
22408         * reload1.c (elimination_effects): Likewise.
22409         * rtlanal.c (nonzero_bits1): Likewise.
22410         * calls.c (store_one_arg): Likewise.  Require the padding to be
22411         known at compile time.
22413 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22414             Alan Hayward  <alan.hayward@arm.com>
22415             David Sherwood  <david.sherwood@arm.com>
22417         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22418         Use plus_constant instead of gen_rtx_PLUS.
22420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22421             Alan Hayward  <alan.hayward@arm.com>
22422             David Sherwood  <david.sherwood@arm.com>
22424         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22425         rather than an int.
22427 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22428             Alan Hayward  <alan.hayward@arm.com>
22429             David Sherwood  <david.sherwood@arm.com>
22431         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22432         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22433         via stack temporaries.  Treat the mode size as polynomial too.
22435 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22436             Alan Hayward  <alan.hayward@arm.com>
22437             David Sherwood  <david.sherwood@arm.com>
22439         * expr.c (expand_expr_real_2): When handling conversions involving
22440         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22441         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
22442         as a poly_uint64 too.
22444 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22445             Alan Hayward  <alan.hayward@arm.com>
22446             David Sherwood  <david.sherwood@arm.com>
22448         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22450 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22451             Alan Hayward  <alan.hayward@arm.com>
22452             David Sherwood  <david.sherwood@arm.com>
22454         * combine.c (can_change_dest_mode): Handle polynomial
22455         REGMODE_NATURAL_SIZE.
22456         * expmed.c (store_bit_field_1): Likewise.
22457         * expr.c (store_constructor): Likewise.
22458         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22459         and polynomial REGMODE_NATURAL_SIZE.
22460         (gen_lowpart_common): Likewise.
22461         * reginfo.c (record_subregs_of_mode): Likewise.
22462         * rtlanal.c (read_modify_subreg_p): Likewise.
22464 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22465             Alan Hayward  <alan.hayward@arm.com>
22466             David Sherwood  <david.sherwood@arm.com>
22468         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22469         numbers of elements.
22471 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22472             Alan Hayward  <alan.hayward@arm.com>
22473             David Sherwood  <david.sherwood@arm.com>
22475         * match.pd: Cope with polynomial numbers of vector elements.
22477 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22478             Alan Hayward  <alan.hayward@arm.com>
22479             David Sherwood  <david.sherwood@arm.com>
22481         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22482         in a POINTER_PLUS_EXPR.
22484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22485             Alan Hayward  <alan.hayward@arm.com>
22486             David Sherwood  <david.sherwood@arm.com>
22488         * omp-simd-clone.c (simd_clone_subparts): New function.
22489         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22490         (ipa_simd_modify_function_body): Likewise.
22492 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22493             Alan Hayward  <alan.hayward@arm.com>
22494             David Sherwood  <david.sherwood@arm.com>
22496         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22497         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22498         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22499         (expand_vector_condition, vector_element): Likewise.
22500         (subparts_gt): New function.
22501         (get_compute_type): Use subparts_gt.
22502         (count_type_subparts): Delete.
22503         (expand_vector_operations_1): Use subparts_gt instead of
22504         count_type_subparts.
22506 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22507             Alan Hayward  <alan.hayward@arm.com>
22508             David Sherwood  <david.sherwood@arm.com>
22510         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22511         (vect_compile_time_alias): ...this new function.  Do the calculation
22512         on poly_ints rather than trees.
22513         (vect_prune_runtime_alias_test_list): Update call accordingly.
22515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22516             Alan Hayward  <alan.hayward@arm.com>
22517             David Sherwood  <david.sherwood@arm.com>
22519         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22520         numbers of units.
22521         (vect_schedule_slp_instance): Likewise.
22523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22524             Alan Hayward  <alan.hayward@arm.com>
22525             David Sherwood  <david.sherwood@arm.com>
22527         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22528         constant and extern definitions for variable-length vectors.
22529         (vect_get_constant_vectors): Note that the number of units
22530         is known to be constant.
22532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22533             Alan Hayward  <alan.hayward@arm.com>
22534             David Sherwood  <david.sherwood@arm.com>
22536         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22537         of units as polynomial.  Choose between WIDE and NARROW based
22538         on multiple_p.
22540 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22541             Alan Hayward  <alan.hayward@arm.com>
22542             David Sherwood  <david.sherwood@arm.com>
22544         * tree-vect-stmts.c (simd_clone_subparts): New function.
22545         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22547 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22548             Alan Hayward  <alan.hayward@arm.com>
22549             David Sherwood  <david.sherwood@arm.com>
22551         * tree-vect-stmts.c (vectorizable_call): Treat the number of
22552         vectors as polynomial.  Use build_index_vector for
22553         IFN_GOMP_SIMD_LANE.
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         * tree-vect-stmts.c (get_load_store_type): Treat the number of
22560         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22561         for variable-length vectors.
22562         (vectorizable_mask_load_store): Treat the number of units as
22563         polynomial, asserting that it is constant if the condition has
22564         already been enforced.
22565         (vectorizable_store, vectorizable_load): Likewise.
22567 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22568             Alan Hayward  <alan.hayward@arm.com>
22569             David Sherwood  <david.sherwood@arm.com>
22571         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22572         of units as polynomial.  Punt if we can't tell at compile time
22573         which vector contains the final result.
22575 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22576             Alan Hayward  <alan.hayward@arm.com>
22577             David Sherwood  <david.sherwood@arm.com>
22579         * tree-vect-loop.c (vectorizable_induction): Treat the number
22580         of units as polynomial.  Punt on SLP inductions.  Use an integer
22581         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
22582         cast of such a series for variable-length floating-point
22583         reductions.
22585 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22586             Alan Hayward  <alan.hayward@arm.com>
22587             David Sherwood  <david.sherwood@arm.com>
22589         * tree.h (build_index_vector): Declare.
22590         * tree.c (build_index_vector): New function.
22591         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22592         of units as polynomial, forcibly converting it to a constant if
22593         vectorizable_reduction has already enforced the condition.
22594         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
22595         to create a {1,2,3,...} vector.
22596         (vectorizable_reduction): Treat the number of units as polynomial.
22597         Choose vectype_in based on the largest scalar element size rather
22598         than the smallest number of units.  Enforce the restrictions
22599         relied on above.
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-data-refs.c (vector_alignment_reachable_p): Treat the
22606         number of units as polynomial.
22608 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22609             Alan Hayward  <alan.hayward@arm.com>
22610             David Sherwood  <david.sherwood@arm.com>
22612         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22613         * target.def (autovectorize_vector_sizes): Return the vector sizes
22614         by pointer, using vector_sizes rather than a bitmask.
22615         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22616         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22617         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22618         Likewise.
22619         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22620         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22621         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22622         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22623         * omp-general.c (omp_max_vf): Likewise.
22624         * omp-low.c (omp_clause_aligned_alignment): Likewise.
22625         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22626         * tree-vect-loop.c (vect_analyze_loop): Likewise.
22627         * tree-vect-slp.c (vect_slp_bb): Likewise.
22628         * doc/tm.texi: Regenerate.
22629         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22630         to a poly_uint64.
22631         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22632         the vector size as a poly_uint64 rather than an unsigned int.
22633         (current_vector_size): Change from an unsigned int to a poly_uint64.
22634         (get_vectype_for_scalar_type): Update accordingly.
22635         * tree.h (build_truth_vector_type): Take the size and number of
22636         units as a poly_uint64 rather than an unsigned int.
22637         (build_vector_type): Add a temporary overload that takes
22638         the number of units as a poly_uint64 rather than an unsigned int.
22639         * tree.c (make_vector_type): Likewise.
22640         (build_truth_vector_type): Take the number of units as a poly_uint64
22641         rather than an unsigned int.
22643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22644             Alan Hayward  <alan.hayward@arm.com>
22645             David Sherwood  <david.sherwood@arm.com>
22647         * target.def (get_mask_mode): Take the number of units and length
22648         as poly_uint64s rather than unsigned ints.
22649         * targhooks.h (default_get_mask_mode): Update accordingly.
22650         * targhooks.c (default_get_mask_mode): Likewise.
22651         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22652         * doc/tm.texi: Regenerate.
22654 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22655             Alan Hayward  <alan.hayward@arm.com>
22656             David Sherwood  <david.sherwood@arm.com>
22658         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22659         * omp-general.c (omp_max_vf): Likewise.
22660         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22661         (expand_omp_simd): Handle polynomial safelen.
22662         * omp-low.c (omplow_simd_context): Add a default constructor.
22663         (omplow_simd_context::max_vf): Change from int to poly_uint64.
22664         (lower_rec_simd_input_clauses): Update accordingly.
22665         (lower_rec_input_clauses): Likewise.
22667 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22668             Alan Hayward  <alan.hayward@arm.com>
22669             David Sherwood  <david.sherwood@arm.com>
22671         * tree-vectorizer.h (vect_nunits_for_cost): New function.
22672         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22673         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22674         (vect_analyze_slp_cost): Likewise.
22675         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22676         (vect_model_load_cost): Likewise.
22678 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22679             Alan Hayward  <alan.hayward@arm.com>
22680             David Sherwood  <david.sherwood@arm.com>
22682         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22683         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22684         from an unsigned int * to a poly_uint64_pod *.
22685         (calculate_unrolling_factor): New function.
22686         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
22688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22689             Alan Hayward  <alan.hayward@arm.com>
22690             David Sherwood  <david.sherwood@arm.com>
22692         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22693         from an unsigned int to a poly_uint64.
22694         (_loop_vec_info::slp_unrolling_factor): Likewise.
22695         (_loop_vec_info::vectorization_factor): Change from an int
22696         to a poly_uint64.
22697         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22698         (vect_get_num_vectors): New function.
22699         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22700         (vect_get_num_copies): Use vect_get_num_vectors.
22701         (vect_analyze_data_ref_dependences): Change max_vf from an int *
22702         to an unsigned int *.
22703         (vect_analyze_data_refs): Change min_vf from an int * to a
22704         poly_uint64 *.
22705         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22706         than an unsigned HOST_WIDE_INT.
22707         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22708         (vect_analyze_data_ref_dependence): Change max_vf from an int *
22709         to an unsigned int *.
22710         (vect_analyze_data_ref_dependences): Likewise.
22711         (vect_compute_data_ref_alignment): Handle polynomial vf.
22712         (vect_enhance_data_refs_alignment): Likewise.
22713         (vect_prune_runtime_alias_test_list): Likewise.
22714         (vect_shift_permute_load_chain): Likewise.
22715         (vect_supportable_dr_alignment): Likewise.
22716         (dependence_distance_ge_vf): Take the vectorization factor as a
22717         poly_uint64 rather than an unsigned HOST_WIDE_INT.
22718         (vect_analyze_data_refs): Change min_vf from an int * to a
22719         poly_uint64 *.
22720         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22721         vfm1 as a poly_uint64 rather than an int.  Make the same change
22722         for the returned bound_scalar.
22723         (vect_gen_vector_loop_niters): Handle polynomial vf.
22724         (vect_do_peeling): Likewise.  Update call to
22725         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22726         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22727         be constant.
22728         * tree-vect-loop.c (vect_determine_vectorization_factor)
22729         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22730         (vect_get_known_peeling_cost): Likewise.
22731         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22732         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22733         (vect_transform_loop): Likewise.  Use the lowest possible VF when
22734         updating the upper bounds of the loop.
22735         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
22736         rather than an int.
22737         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22738         polynomial unroll factors.
22739         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22740         (vect_make_slp_decision): Likewise.
22741         (vect_supported_load_permutation_p): Likewise, and polynomial
22742         vf too.
22743         (vect_analyze_slp_cost): Handle polynomial vf.
22744         (vect_slp_analyze_node_operations): Likewise.
22745         (vect_slp_analyze_bb_1): Likewise.
22746         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22747         than an unsigned HOST_WIDE_INT.
22748         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22749         (vectorizable_load): Handle polynomial vf.
22750         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22751         a poly_uint64.
22752         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22754 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22755             Alan Hayward  <alan.hayward@arm.com>
22756             David Sherwood  <david.sherwood@arm.com>
22758         * match.pd: Handle bit operations involving three constants
22759         and try to fold one pair.
22761 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22763         * tree-vect-loop-manip.c: Include gimple-fold.h.
22764         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22765         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
22766         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22767         Add a path that uses a step of VF instead of 1, but disable it
22768         for now.
22769         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22770         and niters_no_overflow parameters.  Update calls to
22771         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22772         Create a new SSA name if the latter choses to use a ste other
22773         than zero, and return it via niters_vector_mult_vf_var.
22774         * tree-vect-loop.c (vect_transform_loop): Update calls to
22775         vect_do_peeling, vect_gen_vector_loop_niters and
22776         slpeel_make_loop_iterate_ntimes.
22777         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22778         (vect_gen_vector_loop_niters): Update declarations after above changes.
22780 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
22782         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22783         128-bit round to integer instructions.
22784         (ceil<mode>2): Likewise.
22785         (btrunc<mode>2): Likewise.
22786         (round<mode>2): Likewise.
22788 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22790         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22791         unaligned VSX load/store on P8/P9.
22792         (expand_block_clear): Allow the use of unaligned VSX
22793         load/store on P8/P9.
22795 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22797         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22798         New function.
22799         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22800         swap associated with both a load and a store.
22802 2018-01-02  Andrew Waterman  <andrew@sifive.com>
22804         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22805         * config/riscv/riscv.md (clear_cache): Use it.
22807 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
22809         * web.c: Remove out-of-date comment.
22811 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22813         * expr.c (fixup_args_size_notes): Check that any existing
22814         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22815         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22816         (emit_single_push_insn): ...here.
22818 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22820         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22821         (const_vector_encoded_nelts): New function.
22822         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22823         (const_vector_int_elt, const_vector_elt): Declare.
22824         * emit-rtl.c (const_vector_int_elt_1): New function.
22825         (const_vector_elt): Likewise.
22826         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22827         of CONST_VECTOR_ELT.
22829 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22831         * expr.c: Include rtx-vector-builder.h.
22832         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22833         directly on the tree encoding.
22834         (const_vector_from_tree): Likewise.
22835         * optabs.c: Include rtx-vector-builder.h.
22836         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22837         sequence of "u" values.
22838         * vec-perm-indices.c: Include rtx-vector-builder.h.
22839         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22840         directly on the vec_perm_indices encoding.
22842 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22844         * doc/rtl.texi (const_vector): Describe new encoding scheme.
22845         * Makefile.in (OBJS): Add rtx-vector-builder.o.
22846         * rtx-vector-builder.h: New file.
22847         * rtx-vector-builder.c: Likewise.
22848         * rtl.h (rtx_def::u2): Add a const_vector field.
22849         (CONST_VECTOR_NPATTERNS): New macro.
22850         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22851         (CONST_VECTOR_DUPLICATE_P): Likewise.
22852         (CONST_VECTOR_STEPPED_P): Likewise.
22853         (CONST_VECTOR_ENCODED_ELT): Likewise.
22854         (const_vec_duplicate_p): Check for a duplicated vector encoding.
22855         (unwrap_const_vec_duplicate): Likewise.
22856         (const_vec_series_p): Check for a non-duplicated vector encoding.
22857         Say that the function only returns true for integer vectors.
22858         * emit-rtl.c: Include rtx-vector-builder.h.
22859         (gen_const_vec_duplicate_1): Delete.
22860         (gen_const_vector): Call gen_const_vec_duplicate instead of
22861         gen_const_vec_duplicate_1.
22862         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22863         (gen_const_vec_duplicate): Use rtx_vector_builder.
22864         (gen_const_vec_series): Likewise.
22865         (gen_rtx_CONST_VECTOR): Likewise.
22866         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22867         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22868         Build a new vector rather than modifying a CONST_VECTOR in-place.
22869         (handle_special_swappables): Update call accordingly.
22870         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22871         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22872         Build a new vector rather than modifying a CONST_VECTOR in-place.
22873         (handle_special_swappables): Update call accordingly.
22875 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22877         * simplify-rtx.c (simplify_const_binary_operation): Use
22878         CONST_VECTOR_ELT instead of XVECEXP.
22880 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22882         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22883         the selector elements to be different from the data elements
22884         if the selector is a VECTOR_CST.
22885         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22886         ssizetype for the selector.
22888 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22890         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
22891         before testing each element individually.
22892         * tree-vect-generic.c (lower_vec_perm): Likewise.
22894 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22896         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
22897         * selftest-run-tests.c (selftest::run_tests): Call it.
22898         * vector-builder.h (vector_builder::operator ==): New function.
22899         (vector_builder::operator !=): Likewise.
22900         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
22901         (vec_perm_indices::all_from_input_p): New function.
22902         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22903         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
22904         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
22905         instead of reading the VECTOR_CST directly.  Detect whether both
22906         vector inputs are the same before constructing the vec_perm_indices,
22907         and update the number of inputs argument accordingly.  Use the
22908         utility functions added above.  Only construct sel2 if we need to.
22910 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22912         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22913         the broadcast of the low byte.
22914         (expand_mult_highpart): Use an explicit encoding for the permutes.
22915         * optabs-query.c (can_mult_highpart_p): Likewise.
22916         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22917         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22918         (vectorizable_bswap): Likewise.
22919         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22920         explicit encoding for the power-of-2 permutes.
22921         (vect_permute_store_chain): Likewise.
22922         (vect_grouped_load_supported): Likewise.
22923         (vect_permute_load_chain): Likewise.
22925 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22927         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
22928         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
22929         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
22930         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22931         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
22932         (vect_gen_perm_mask_any): Likewise.
22934 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
22936         * int-vector-builder.h: New file.
22937         * vec-perm-indices.h: Include int-vector-builder.h.
22938         (vec_perm_indices): Redefine as an int_vector_builder.
22939         (auto_vec_perm_indices): Delete.
22940         (vec_perm_builder): Redefine as a stand-alone class.
22941         (vec_perm_indices::vec_perm_indices): New function.
22942         (vec_perm_indices::clamp): Likewise.
22943         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
22944         (vec_perm_indices::new_vector): New function.
22945         (vec_perm_indices::new_expanded_vector): Update for new
22946         vec_perm_indices class.
22947         (vec_perm_indices::rotate_inputs): New function.
22948         (vec_perm_indices::all_in_range_p): Operate directly on the
22949         encoded form, without computing elided elements.
22950         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
22951         encoding.  Update for new vec_perm_indices class.
22952         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
22953         the given vec_perm_builder.
22954         (expand_vec_perm_var): Update vec_perm_builder constructor.
22955         (expand_mult_highpart): Use vec_perm_builder instead of
22956         auto_vec_perm_indices.
22957         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
22958         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
22959         or double series encoding as appropriate.
22960         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
22961         vec_perm_indices instead of auto_vec_perm_indices.
22962         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22963         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22964         (vect_permute_store_chain): Likewise.
22965         (vect_grouped_load_supported): Likewise.
22966         (vect_permute_load_chain): Likewise.
22967         (vect_shift_permute_load_chain): Likewise.
22968         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22969         (vect_transform_slp_perm_load): Likewise.
22970         (vect_schedule_slp_instance): Likewise.
22971         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22972         (vectorizable_mask_load_store): Likewise.
22973         (vectorizable_bswap): Likewise.
22974         (vectorizable_store): Likewise.
22975         (vectorizable_load): Likewise.
22976         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
22977         vec_perm_indices instead of auto_vec_perm_indices.  Use
22978         tree_to_vec_perm_builder to read the vector from a tree.
22979         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
22980         vec_perm_builder instead of a vec_perm_indices.
22981         (have_whole_vector_shift): Use vec_perm_builder and
22982         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
22983         truncation to calc_vec_perm_mask_for_shift.
22984         (vect_create_epilog_for_reduction): Likewise.
22985         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
22986         from auto_vec_perm_indices to vec_perm_indices.
22987         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22988         instead of changing individual elements.
22989         (aarch64_vectorize_vec_perm_const): Use new_vector to install
22990         the vector in d.perm.
22991         * config/arm/arm.c (expand_vec_perm_d::perm): Change
22992         from auto_vec_perm_indices to vec_perm_indices.
22993         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22994         instead of changing individual elements.
22995         (arm_vectorize_vec_perm_const): Use new_vector to install
22996         the vector in d.perm.
22997         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
22998         Update vec_perm_builder constructor.
22999         (rs6000_expand_interleave): Likewise.
23000         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23001         (rs6000_expand_interleave): Likewise.
23003 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23005         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23006         to qimode could truncate the indices.
23007         * optabs.c (expand_vec_perm_var): Likewise.
23009 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23011         * Makefile.in (OBJS): Add vec-perm-indices.o.
23012         * vec-perm-indices.h: New file.
23013         * vec-perm-indices.c: Likewise.
23014         * target.h (vec_perm_indices): Replace with a forward class
23015         declaration.
23016         (auto_vec_perm_indices): Move to vec-perm-indices.h.
23017         * optabs.h: Include vec-perm-indices.h.
23018         (expand_vec_perm): Delete.
23019         (selector_fits_mode_p, expand_vec_perm_var): Declare.
23020         (expand_vec_perm_const): Declare.
23021         * target.def (vec_perm_const_ok): Replace with...
23022         (vec_perm_const): ...this new hook.
23023         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23024         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23025         * doc/tm.texi: Regenerate.
23026         * optabs.def (vec_perm_const): Delete.
23027         * doc/md.texi (vec_perm_const): Likewise.
23028         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23029         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23030         expand_vec_perm for constant permutation vectors.  Assert that
23031         the mode of variable permutation vectors is the integer equivalent
23032         of the mode that is being permuted.
23033         * optabs-query.h (selector_fits_mode_p): Declare.
23034         * optabs-query.c: Include vec-perm-indices.h.
23035         (selector_fits_mode_p): New function.
23036         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23037         is defined, instead of checking whether the vec_perm_const_optab
23038         exists.  Use targetm.vectorize.vec_perm_const instead of
23039         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
23040         fit in the vector mode before using a variable permute.
23041         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23042         vec_perm_indices instead of an rtx.
23043         (expand_vec_perm): Replace with...
23044         (expand_vec_perm_const): ...this new function.  Take the selector
23045         as a vec_perm_indices rather than an rtx.  Also take the mode of
23046         the selector.  Update call to shift_amt_for_vec_perm_mask.
23047         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23048         Use vec_perm_indices::new_expanded_vector to expand the original
23049         selector into bytes.  Check whether the indices fit in the vector
23050         mode before using a variable permute.
23051         (expand_vec_perm_var): Make global.
23052         (expand_mult_highpart): Use expand_vec_perm_const.
23053         * fold-const.c: Includes vec-perm-indices.h.
23054         * tree-ssa-forwprop.c: Likewise.
23055         * tree-vect-data-refs.c: Likewise.
23056         * tree-vect-generic.c: Likewise.
23057         * tree-vect-loop.c: Likewise.
23058         * tree-vect-slp.c: Likewise.
23059         * tree-vect-stmts.c: Likewise.
23060         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23061         Delete.
23062         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23063         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23064         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23065         (aarch64_vectorize_vec_perm_const): ...this new function.
23066         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23067         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23068         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23069         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23070         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23071         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23072         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23073         into...
23074         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
23075         check for NEON modes.
23076         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23077         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23078         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23079         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23080         into...
23081         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
23082         the old VEC_PERM_CONST conditions.
23083         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23084         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23085         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23086         (ia64_vectorize_vec_perm_const_ok): Merge into...
23087         (ia64_vectorize_vec_perm_const): ...this new function.
23088         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23089         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23090         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23091         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23092         * config/mips/mips.c (mips_expand_vec_perm_const)
23093         (mips_vectorize_vec_perm_const_ok): Merge into...
23094         (mips_vectorize_vec_perm_const): ...this new function.
23095         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23096         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23097         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23098         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23099         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23100         (rs6000_expand_vec_perm_const): Delete.
23101         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23102         Delete.
23103         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23104         (altivec_expand_vec_perm_const_le): Take each operand individually.
23105         Operate on constant selectors rather than rtxes.
23106         (altivec_expand_vec_perm_const): Likewise.  Update call to
23107         altivec_expand_vec_perm_const_le.
23108         (rs6000_expand_vec_perm_const): Delete.
23109         (rs6000_vectorize_vec_perm_const_ok): Delete.
23110         (rs6000_vectorize_vec_perm_const): New function.
23111         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23112         an element count and rtx array.
23113         (rs6000_expand_extract_even): Update call accordingly.
23114         (rs6000_expand_interleave): Likewise.
23115         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23116         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23117         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23118         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23119         (rs6000_expand_vec_perm_const): Delete.
23120         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23121         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23122         (altivec_expand_vec_perm_const_le): Take each operand individually.
23123         Operate on constant selectors rather than rtxes.
23124         (altivec_expand_vec_perm_const): Likewise.  Update call to
23125         altivec_expand_vec_perm_const_le.
23126         (rs6000_expand_vec_perm_const): Delete.
23127         (rs6000_vectorize_vec_perm_const_ok): Delete.
23128         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
23129         reference to the SPE evmerge intructions.
23130         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23131         an element count and rtx array.
23132         (rs6000_expand_extract_even): Update call accordingly.
23133         (rs6000_expand_interleave): Likewise.
23134         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23135         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23136         new function.
23137         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23139 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23141         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23142         vector mode and that that mode matches the mode of the data
23143         being permuted.
23144         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23145         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
23146         directly using expand_vec_perm_1 when forcing selectors into
23147         registers.
23148         (expand_vec_perm_var): New function, split out from expand_vec_perm.
23150 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23152         * optabs-query.h (can_vec_perm_p): Delete.
23153         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23154         * optabs-query.c (can_vec_perm_p): Split into...
23155         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23156         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23157         particular selector is valid.
23158         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23159         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23160         (vect_grouped_load_supported): Likewise.
23161         (vect_shift_permute_load_chain): Likewise.
23162         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23163         (vect_transform_slp_perm_load): Likewise.
23164         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23165         (vectorizable_bswap): Likewise.
23166         (vect_gen_perm_mask_checked): Likewise.
23167         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
23168         implementations of variable permutation vectors into account
23169         when deciding which selector to use.
23170         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23171         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23172         with a false third argument.
23173         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23174         to test whether the constant selector is valid and can_vec_perm_var_p
23175         to test whether a variable selector is valid.
23177 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23179         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23180         * optabs-query.c (can_vec_perm_p): Likewise.
23181         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23182         instead of vec_perm_indices.
23183         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23184         (vect_gen_perm_mask_checked): Likewise,
23185         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23186         (vect_gen_perm_mask_checked): Likewise,
23188 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
23190         * optabs-query.h (qimode_for_vec_perm): Declare.
23191         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23192         (qimode_for_vec_perm): ...this new function.
23193         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23195 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23197         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23198         does not have a conditional at the top.
23200 2018-01-02  Richard Biener  <rguenther@suse.de>
23202         * ipa-inline.c (big_speedup_p): Fix expression.
23204 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
23206         PR target/81616
23207         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23208         for generic 4->6.
23210 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
23212         PR target/81616
23213         Generic tuning.
23214         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23215         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23216         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23217         cond_taken_branch_cost 3->4.
23219 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
23221         PR tree-optimization/83581
23222         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23223         TODO_cleanup_cfg if any changes have been made.
23225         PR middle-end/83608
23226         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23227         convert_modes if target mode has the right side, but different mode
23228         class.
23230         PR middle-end/83609
23231         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23232         last argument when extracting from CONCAT.  If either from_real or
23233         from_imag is NULL, use expansion through memory.  If result is not
23234         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23235         the parts directly to inner mode, if even that fails, use expansion
23236         through memory.
23238         PR middle-end/83623
23239         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23240         check for bswap in mode rather than HImode and use that in expand_unop
23241         too.
23243 Copyright (C) 2018 Free Software Foundation, Inc.
23245 Copying and distribution of this file, with or without modification,
23246 are permitted in any medium without royalty provided the copyright
23247 notice and this notice are preserved.