Add assember CFI directives to millicode division and remainder routines.
[official-gcc.git] / gcc / ChangeLog-2018
bloba6a4011b54c3c2782eef5d088628d0919c81846d
1 2018-12-31  Max Filippov  <jcmvbkbc@gmail.com>
3         PR target/86814
4         * config/xtensa/xtensa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5         Define to speculation_safe_value_not_needed.
7 2018-12-31  Martin Liska  <mliska@suse.cz>
9         * predict.c (set_even_probabilities): Include also
10         unlikely_count in calculation.
11         (combine_predictions_for_bb): Consider also HOT and
12         COLD labels predictions.
13         * predict.def (PRED_HOT_LABEL): Move it just after
14         __builtin_expect_with_probability predictor.
15         (PRED_COLD_LABEL): Likewise.
17 2018-12-30  Jan Hubicka  <hubicka@ucw.cz>
19         * x86-tune.def: Enable inter_unit_moves_to_vec for generic.
21 2018-12-30  Martin Sebor  <msebor@redhat.com>
22             Jeff Law  <law@redhat.com>
24         * gimple-fold.c (get_range_strlen): Minor logic cleanup.  Add comments
25         on code's intent.
27         * gimple-fold.c (strlen_range_kind): New enum.
28         (get_range_strlen): Update signature to use strlen_range_kind
29         instead of type+fuzzy.
30         (get_range_strlen_tree): Update signature to use
31         strlen_range_kind instead of type+fuzzy.  Pass rkind down to
32         get_range_strlen.  Check the rkind enum value instead of
33         type+fuzzy.
34         (get_range_strlen): Similarly.
35         (get_maxval_strlen): Update signature to make NONSTR argument have
36         a default value and make static.  Add asserts to ensure sanity of
37         arguments.  Update calls to get_range_strlen.
38         (gimple_fold_builtin_strcpy): Update calls to get_maxval_strlen.
39         (gimple_fold_builtin_strcat, gimple_fold_builtin_fputs): Similarly.
40         (gimple_fold_builtin_memory_chk): Similarly.
41         (gimple_fold_builtin_stxcpy_chk): Similarly.
42         (gimple_fold_builtin_snprintf_chk): Similarly.
43         (gimple_fold_builtin_sprintf, gimple_fold_builtin_snprintf): Similarly.
44         * gimple-fold.h (get_maxval_strlen): Delete prototype.
46 2018-12-29  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
48         * config/msp430/msp430.c (msp430_attr): Warn when the critical and
49         interrupt function attributes are used together.
50         * gcc/doc/extend.texi: Update documentation on the critical attribute.
52 2018-12-29  John David Anglin  <danglin@gcc.gnu.org>
54         * config.gcc (hppa*64*-*-linux*): Add pa/t-pa to tmake_file.  Define
55         d_target_objs.
56         (hppa*-*-openbsd*): Likewise.
57         (hppa[12]*-*-hpux10*): Likewise.
58         (hppa*64*-*-hpux11*): Likewise.
59         (hppa[12]*-*-hpux11*): Likewise.
61 2018-12-29  Jakub Jelinek  <jakub@redhat.com>
63         PR ipa/88586
64         * omp-low.c (lower_omp_task_reductions): Set DECL_CONTEXT on field
65         and ifield.  Update TYPE_ALIGN from alignment of field, ifield or
66         bfield.
68 2018-12-28  Uros Bizjak  <ubizjak@gmail.com>
70         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
71         sil, dil, bpl and spl aliases.
73 2018-12-28  Martin Sebor  <msebor@redhat.com>
75         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Avoid
76         undesirable warning.
78 2018-12-28  Thomas Schwinge  <thomas@codesourcery.com>
80         * omp-expand.c (expand_omp_target): Restructure OpenACC vs. OpenMP
81         code paths.  Update for libgomp OpenACC entry points change.
83 2018-12-28  Thomas Schwinge  <thomas@codesourcery.com>
84             Julian Brown  <julian@codesourcery.com>
86         * gimplify.c (gimplify_scan_omp_clauses): Fix known_eq typo/bug.
88 2018-12-27  Jan Hubicka  <hubicka@ucw.cz>
90         * ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types.
91         (polymorphic_call_target_hasher::hash): Hash it.
92         (polymorphic_call_target_hasher::equal): Compare it.
93         (possible_polymorphic_call_targets): Set it.
94         * tree.c (free_lang_data): Rebuild type inheritance graph even on
95         non-LTO path.
97 2018-12-27  Martin Liska  <mliska@suse.cz>
99         PR gcov-profile/88225
100         * gcov.c(source_info::get_functions_at_location):
101         Use newly added line_to_function_map.
102         (source_info::add_function): New.
103         (output_json_intermediate_file): Use a pointer return
104         type for get_functions_at_location.
105         (process_all_functions): Use add_function instead
106         of direct push to a s->functions container.
107         (release_structures): Release ident_to_fn.
108         (read_graph_file): Register function into ident_to_fn.
109         (read_count_file): Use the map.
110         (output_lines): Handle pointer return type of
111         get_functions_at_location.
113 2018-12-27  Martin Liska  <mliska@suse.cz>
115         * builtins.c (expand_movstr): Compare with RETURN_BEGIN.
116         * expr.c (move_by_pieces): Likewise.
117         (store_by_pieces): Likewise.
118         (store_expr): Fix GNU coding style.
120 2018-12-26  Andrew Pinski  <apinski@marvell.com>
122         * config/aarch64/aarch64-cores.def (octeontx): New.
123         (octeontx81): Likewise.
124         (octeontx83): Likewise.
125         * config/aarch64/aarch64-tune.md: Regenerate.
126         * doc/invoke.texi (AArch64 Options) [mtune]: Add octeontx, octeontx81
127         and octeontx83.
129 2018-12-24  Martin Sebor  <msebor@redhat.com>
131         * gimple-fold.c (get_range_strlen_tree): Factored out of
132         get_range_strlen.  Minor comment updates/additions.  Assert
133         argument is not a simple SSA_NAME.
134         (get_range_strlen): Call get_range_strlen_tree as appropriate.
135         Minor comment updates/additions.
137 2018-12-24  Jan Hubicka  <hubicka@ucw.cz>
139         * ipa-devirt.c (dump_targets): Cap number of targets printed.
140         (dump_possible_polymorphic_call_targets): Add verbose parameter.
141         (ipa_devirt): Use it.
142         * ipa-utils.h (dump_possible_polymorphic_call_targets): Add verbose
143         parametrs.
145 2018-12-24  Jan Hubicka  <hubicka@ucw.cz>
147         PR lto/88140
148         * tree.c (fld_simplified_type): Temporarily disable array
149         simplification.
151 2018-12-24  Jan Hubicka  <hubicka@ucw.cz>
153         * lto-symtab.c (lto_symtab_merge_decls_2): Do not report ODR violations
154         for method whose basetype was already reported.
155         * ipa-devirt.c (odr_type_violation_reported_p): New.
156         * ipa-utils.h (odr_type_violation_reported_p): Declare.
158 2018-12-24  Iain Sandoe  <iain@sandoe.co.uk>
160         * configure.ac (dwarf2_debug_line): Check for the debug_line
161         section using otool when there’s no objdump available.
162         * configure: Regenerate.
164 2018-12-24  Jan Hubicka  <hubicka@ucw.cz>
166         * ipa-utils.c (ipa_merge_profiles): Fix updating of fnsummary;
167         also handle rescaling of mismatched profiles.
168         * ipa-fnsummary.c (analyze_function): Handle speculative edges.
170 2018-12-23  Martin Sebor  <msebor@redhat.com>
171             Jeff Law  <law@redhat.com>
173         * builtins.h (c_strlen_data): Add new fields and comments.
174         * builtins.c (unterminated_array): Change field reference from
175         "len" to "minlen" in c_strlen_data instance.
176         * gimple-fold.c (get_range_strlen): Likewise.
177         * gimple-ssa-sprintf.c (get_string_length): Likewise.
179         * builtins.c (unterminated_array): Rename "data" to "lendata".  Fix
180         a few comments.
181         (expand_builtin_strnlen, expand_builtin_stpcpy_1): Likewise.
182         (fold_builtin_strlen): Likewise.
183         * gimple-fold.c (get_range_strlen): Likewise.  Also rename a couple
184         instances of "type" to "optype" and "arg" to "fld".
186         * builtins.c (unterminated_array): Use empty brace initialization
187         for c_strlen_data.
188         (c_strlen, expand_builtin_strnlen): Likewise.
189         (expand_builtin_stpcpy_1, fold_builtin_strlen): Likewise.
190         * gimple-fold.c (get_range_strlen): Likewise.
191         (gimple_fold_builtin_stpcpy): Likewise.
192         * gimple-ssa-sprintf.c (get_string_length): Likewise.
194 2018-12-23  Alan Modra  <amodra@gmail.com>
196         PR 88346
197         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Correct %e message.  Handle
198         -mcpu=rs64, not -mcpu=rs64a.  Handle -mcpu=powerpc64 and -mcpu=titan.
199         * config/rs6000/driver-rs6000.c (asm_names): Similarly.
200         * config/rs6000/aix71.h (ASM_CPU_SPEC): Delete %e message.  Handle
201         -mcpu=rs64, not -mcpu=rs64a.
202         * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
204 2018-12-23  Iain Sandoe  <iain@sandoe.co.uk>
206         * lto-wrapper.c (tool_cleanup): Don’t report ficticious temp files.
207         (debug_objcopy): Improve temp filenames.  Append the early
208         debug objects to the outfiles.  Remove the phase prelinking the
209         early debug. Emit the early debug filenames for ld.
211 2018-12-23  Iain Sandoe  <iain@sandoe.co.uk>
213         * config/i386/darwin.h (TARGET_ASM_OUTPUT_IDENT): New.
215 2018-12-23  Iain Sandoe  <iain@sandoe.co.uk>
217         * collect2.c (maybe_run_lto_and_relink): Don’t say we have a temp file
218         unless we actually did some LTO.
219         (has_lto_section, is_lto_object_file): New.
220         (maybe_lto_object_file): Remove.
221         (scan_prog_file): Use is_lto_object_file() instead of scanning the
222         output of nm.
224 2018-12-23  Iain Sandoe  <iain@sandoe.co.uk>
226         * collect2.c (main): Parse the output file early so we can make nicer
227         temp names.  Respond to “-save-temps” in the GCC OPTIONS.
228         (maybe_unlink): Don’t print “[Leaving…”] for files we never created
229         and don’t exist.
231 2018-12-23  Iain Sandoe  <iain@sandoe.co.uk>
233         * collect2.c (main): Combine flags from both the command line and
234         COLLECT_GCC_OPTIONS to determine the set in force
236 2018-12-21  Jan Hubicka  <hubicka@ucw.cz>
238         * ipa-utils.c (ipa_merge_profiles): Recompute summaries.
240 2018-12-21  Jakub Jelinek  <jakub@redhat.com>
242         PR middle-end/85594
243         PR middle-end/88553
244         * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition
245         if needed.
246         (expand_omp_for_generic): Don't clobber t temporary for ordered loops.
248 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
250         PR ipa/88561
251         * ipa-polymorphic-call.c
252         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle
253         arguments of thunks correctly.
254         (ipa_polymorphic_call_context::get_dynamic_context): Be ready for
255         NULL instance pinter.
256         * lto-cgraph.c (lto_output_node): Always stream thunk info.
258 2018-12-21  Andreas Krebbel  <krebbel@linux.ibm.com>
260         * config/s390/vector.md ("floatv2div2df2", "floatunsv2div2df2")
261         ("fix_truncv2dfv2di2", "fixuns_truncv2dfv2di2"): New pattern
262         definitions.
264 2018-12-21  Eric Botcazou  <ebotcazou@adacore.com>
266         PR rtl-optimization/87727
267         * combine.c (cant_combine_insn_p): On a LEAF_REGISTERS target, combine
268         again moves from leaf hard registers.
270         * final.c (final_scan_insn_1) <NOTE_INSN_INLINE_ENTRY>: Minor tweak.
272 2018-12-21  Jakub Jelinek  <jakub@redhat.com>
274         PR target/88522
275         * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask,
276         *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask,
277         *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for
278         -masm=intel.
279         (gatherq_mode): Remove mode iterator.
280         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead
281         of <xtg_mode>.
282         (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>.
283         (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>,
284         *avx512f_scatterdi<mode>): Use %X5 for -masm=intel.
286         PR rtl-optimization/88563
287         * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode
288         and mode arguments to convert_modes.  Likewise swap mode and word_mode
289         arguments.  Handle both arguments with VOIDmode before convert_modes
290         of one of them.  Formatting fixes.
292 2018-12-21  Uros Bizjak  <ubizjak@gmail.com>
294         PR target/88556
295         * internal-fn.def (COSH): New.
296         (SINH): Ditto.
297         (TANH): Ditto.
298         * optabs.def (cosh_optab): New.
299         (sinh_optab): Ditto.
300         (tanh_optab): Ditto.
301         * config/i386/i386-protos.h (ix86_emit_i387_sinh): New prototype.
302         (ix86_emit_i387_cosh): Ditto.
303         (ix86_emit_i387_tanh): Ditto.
304         * config/i386/i386.c (ix86_emit_i387_sinh): New function.
305         (ix86_emit_i387_cosh): Ditto.
306         (ix86_emit_i387_tanh): Ditto.
307         * config/i386/i386.md (sinhxf2): New expander.
308         (sinh<mode>2):  Ditto.
309         (coshxf2): Ditto.
310         (cosh<mode>2): Ditto.
311         (tanhxf2): Ditto.
312         (tanh<mode>2): Ditto.
314 2018-12-21  Jakub Jelinek  <jakub@redhat.com>
316         PR target/88547
317         * config/i386/i386.c (ix86_expand_int_sse_cmp): Optimize
318         x > y ? 0 : -1 into min (x, y) == x ? -1 : 0.
320 2018-12-21  Jan Beulich  <jbeulich@suse.com>
322         * config/i386/sse.md (vaesdec_<mode>, vaesdeclast_<mode>): Allow
323         memory input.
325 2018-12-21  Alexandre Oliva <oliva@adacore.com>
327         * config.gcc (tmake_file): Add name to tmake_file for
328         each @name in --with-multilib-list on arm-*-* targets.
329         * doc/install.texi (with-multilib-list): Document it.
331 2018-12-21  Jakub Jelinek  <jakub@redhat.com>
333         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Use
334         DR_TARGET_ALIGNMENT on dr_info rather than dr.  Spelling fixes.
336 2018-12-20  Joern Rennecke  <joern.rennecke@riscy-ip.com>
338         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Don't do
339         versioning for data accesses with misaligned step.
341 2018-12-20  H.J. Lu  <hongjiu.lu@intel.com>
343         PR c/51628
344         * doc/invoke.texi: Document -Wno-address-of-packed-member.
346 2018-12-20  Vladimir Makarov  <vmakarov@redhat.com>
348         PR target/88457
349         * ira-color.c (fast_allocation): Choose the best cost hard reg.
351 2018-12-20  Richard Sandiford  <richard.sandiford@arm.com>
353         * config/aarch64/iterators.md (SVE_INT_UNARY, fp_int_op): Add abs.
354         (SVE_FP_UNARY): Sort.
356 2018-12-20  Richard Sandiford  <richard.sandiford@arm.com>
358         * config/aarch64/aarch64-sve.md (*cond_<optab><mode>_4): Use
359         sve_fmla_op rather than sve_fmad_op for the movprfx alternative.
361 2018-12-20  Martin Jambor  <mjambor@suse.cz>
363         PR ipa/88214
364         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Assert that
365         ptr is a pointer.
367 2018-12-20  Richard Biener  <rguenther@suse.de>
369         PR tree-optimization/84362
370         * tree-ssa-loop-im.c: Include alias.h, builtins.h and tree-dfa.h.
371         (struct im_mem_ref): add ref_canonical flag.
372         (struct mem_ref_hasher): Use ao_ref as compare_type.
373         (mem_ref_hasher::equal): Adjust and add variant comparing ao_ref
374         parts.
375         (mem_ref_alloc): Take ao_ref parameter, initialize ref_canonical
376         member.
377         (gather_mem_refs_stmt): Set up ao_ref early and do the lookup
378         using it.  If we have non-equal refs canonicalize the one
379         in the hashtable used for insertion.
380         (tree_ssa_lim_initialize): Adjust.
382 2018-12-20  Jakub Jelinek  <jakub@redhat.com>
384         PR target/88547
385         * config/i386/i386.c (ix86_expand_sse_movcc): For maskcmp, try to
386         emit vpmovm2? instruction perhaps after knot?.  Reorganize code
387         so that it doesn't have to test !maskcmp in almost every conditional.
389 2018-12-20  Alan Modra  <amodra@gmail.com>
391         * config/rs6000/sysv4.h (GNU_USER_DYNAMIC_LINKER): Define.
393 2018-12-19  Andreas Tobler  <andreast@gcc.gnu.org>
395         * config.gcc: Enable TARGET_DEFAULT_ASYNC_UNWIND_TABLES on
396         aarch64*-*-freebsd*
398 2018-12-19  Uros Bizjak  <ubizjak@gmail.com>
400         * config/i386/i386.md (SWI1248_AVX512BWDQ_64): Rename from
401         SWI1248_AVX512BWDQ2_64.  Unconditionally enable HImode.
402         (*cmp<mode>_ccz_1): Emit kortest instead of ktest insn.
403         Use SWI1248_AVX512BWDQ_64 mode iterator and enable only for
404         TARGET_AVX512F.
406 2018-12-19  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
408         * config/arm/elf.h: Update comment about condition that need to
409         match with libgcc/config/arm/lib1funcs.S to also include
410         libgcc/config/arm/t-arm.
411         * doc/sourcebuild.texi (output-exists, output-exists-not): Rename
412         subsubsection these directives are in to "Check for output files".
413         Move scan-symbol to that section and add to it new scan-symbol-not
414         directive.
416 2018-12-19  Tom de Vries  <tdevries@suse.de>
418         * config/nvptx/nvptx.c (PTX_CTA_SIZE): Define.
420 2018-12-19  Tom de Vries  <tdevries@suse.de>
422         * omp-offload.c (oacc_fn_attrib_level): Remove static.
423         * omp-offload.h (oacc_fn_attrib_level): Declare.
425 2018-12-19  Tom de Vries  <tdevries@suse.de>
427         * omp-offload.c (oacc_get_default_dim): New function.
428         * omp-offload.h (oacc_get_default_dim): Declare.
430 2018-12-19  David Malcolm  <dmalcolm@redhat.com>
432         PR c++/87504
433         * gcc-rich-location.c
434         (maybe_range_label_for_tree_type_mismatch::get_text): Move here from
435         c/c-typeck.c.
436         (binary_op_rich_location::binary_op_rich_location): New ctor.
437         (binary_op_rich_location::use_operator_loc_p): New function.
438         * gcc-rich-location.h
439         (class maybe_range_label_for_tree_type_mismatch)): Move here from
440         c/c-typeck.c.
441         (struct op_location_t): New forward decl.
442         (class binary_op_rich_location): New class.
443         * tree.h (struct op_location_t): New struct.
445 2018-12-19  David Malcolm  <dmalcolm@redhat.com>
447         PR c++/43064
448         PR c++/43486
449         * convert.c: Include "selftest.h".
450         (preserve_any_location_wrapper): New function.
451         (convert_to_pointer_maybe_fold): Update to handle location
452         wrappers.
453         (convert_to_real_maybe_fold): Likewise.
454         (convert_to_integer_1): Strip expr when using TREE_OVERFLOW.
455         Handle location wrappers when checking for INTEGER_CST.
456         (convert_to_integer_maybe_fold): Update to handle location
457         wrappers.
458         (convert_to_complex_maybe_fold): Likewise.
459         (selftest::test_convert_to_integer_maybe_fold): New functions.
460         (selftest::convert_c_tests): New function.
461         * convert.h (preserve_any_location_wrapper): New decl.
462         * fold-const.c (size_binop_loc): Strip location wrappers when
463         using TREE_OVERFLOW.
464         (operand_equal_p): Strip any location wrappers.
465         (integer_valued_real_p): Strip any location wrapper.
466         * selftest-run-tests.c (selftest::run_tests): Call
467         selftest::convert_c_tests.
468         * selftest.h (selftest::convert_c_tests): New decl.
469         * tree.c (build_complex): Assert that REAL and IMAG are constants.
470         (integer_zerop): Look through location wrappers.
471         (integer_onep): Likewise.
472         (integer_each_onep): Likewise.
473         (integer_all_onesp): Likewise.
474         (integer_minus_onep): Likewise.
475         (integer_pow2p): Likewise.
476         (integer_nonzerop): Likewise.
477         (integer_truep): Likewise.
478         (fixed_zerop): Likewise.
479         (real_zerop): Likewise.
480         (real_onep): Likewise.
481         (real_minus_onep): Likewise.
482         (tree_int_cst_equal): Likewise.
483         (simple_cst_equal): Treat location wrappers with non-equal source
484         locations as being unequal.
485         (uniform_integer_cst_p): Look through location wrappers.
486         (maybe_wrap_with_location): Don't create wrappers if any
487         auto_suppress_location_wrappers are active.
488         (suppress_location_wrappers): New variable.
489         (selftest::test_predicates): New test.
490         (selftest::tree_c_tests): Call it.
491         * tree.h (CONSTANT_CLASS_OR_WRAPPER_P): New macro.
492         (suppress_location_wrappers): New decl.
493         (class auto_suppress_location_wrappers): New class.
495 2018-12-19  Paul A. Clarke  <pc@us.ibm.com>
497         * config/rs6000/tmmintrin.h (_mm_hadds_epi16): Vector lanes swapped.
498         (_mm_hsub_epi32): Likewise.
499         (_mm_shuffle_epi8): Fix reversed interpretation of parameters.
500         (_mm_shuffle_pi8): Likewise.
501         (_mm_addubs_pi16): Likewise.
503 2018-12-19  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
505         * config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT
506         targets.
507         * config/arm/arm.c (output_return_instruction): Only check
508         TARGET_HARD_FLOAT to decide whether FP instructions are available.
510 2018-12-19  Tom de Vries  <tom@codesourcery.com>
512         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
513         dump files): Add offload-rtl.
515 2018-12-19  Segher Boessenkool  <segher@kernel.crashing.org>
517         PR target/88213
518         * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
519         Require TARGET_POWERPC64.
521 2018-12-19  Richard Biener  <rguenther@suse.de>
523         PR tree-optimization/88533
524         Revert
525         2018-04-30  Richard Biener  <rguenther@suse.de>
527         PR tree-optimization/28364
528         PR tree-optimization/85275
529         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
530         copying first exit test.
532         * tree-ssa-loop-ch.c: Include tree-phinodes.h and
533         ssa-iterators.h.
534         (should_duplicate_loop_header_p): Track whether stmt compute
535         loop invariants or values based on IVs.  Apart from the
536         original loop header only duplicate blocks with exit tests
537         that are based on IVs or invariants.
539 2018-12-19  Tom de Vries  <tdevries@suse.de>
541         * config/nvptx/nvptx.c (nvptx_gen_shared_bcast, shared_prop_gen)
542         (nvptx_goacc_expand_accel_var): Use MAX and ROUND_UP.
543         (nvptx_assemble_value, nvptx_output_skip): Use MIN.
544         (nvptx_shared_propagate, nvptx_single, nvptx_expand_shared_addr): Use
545         MAX.
547 2018-12-19  Tom de Vries  <tdevries@suse.de>
549         * config/nvptx/nvptx.c (nvptx_gen_wcast): Rename as
550         nvptx_gen_warp_bcast.
551         (nvptx_gen_wcast): Rename to nvptx_gen_shared_bcast, add bool
552         vector argument, and update call to nvptx_gen_shared_bcast.
553         (propagator_fn): Add bool argument.
554         (nvptx_propagate): New bool argument, pass bool argument to fn.
555         (vprop_gen): Rename to warp_prop_gen, update call to
556         nvptx_gen_warp_bcast.
557         (nvptx_vpropagate): Rename to nvptx_warp_propagate, update call to
558         nvptx_propagate.
559         (wprop_gen): Rename to shared_prop_gen, update call to
560         nvptx_gen_shared_bcast.
561         (nvptx_wpropagate): Rename to nvptx_shared_propagate, update call
562         to nvptx_propagate.
563         (nvptx_wsync): Rename to nvptx_cta_sync.
564         (nvptx_single): Update calls to nvptx_gen_warp_bcast,
565         nvptx_gen_shared_bcast and nvptx_cta_sync.
566         (nvptx_process_pars): Likewise.
567         (write_worker_buffer): Rename as write_shared_buffer.
568         (nvptx_file_end): Update calls to write_shared_buffer.
569         (nvptx_expand_worker_addr): Rename as nvptx_expand_shared_addr.
570         (nvptx_expand_builtin): Update call to nvptx_expand_shared_addr.
571         (nvptx_get_worker_red_addr): Rename as nvptx_get_shared_red_addr.
572         (nvptx_goacc_reduction_setup): Update call to
573         nvptx_get_shared_red_addr.
574         (nvptx_goacc_reduction_fini): Likewise.
575         (nvptx_goacc_reduction_teardown): Likewise.
577 2018-12-19  Tom de Vries  <tdevries@suse.de>
579         * config/nvptx/nvptx.c (worker_bcast_size): Rename as
580         oacc_bcast_size.
581         (worker_bcast_align): Rename as oacc_bcast_align.
582         (worker_bcast_sym): Rename as oacc_bcast_sym.
583         (nvptx_option_override): Update usage of oacc_bcast_*.
584         (struct wcast_data_t): Rename as broadcast_data_t.
585         (nvptx_gen_wcast): Update type of data argument and usage of
586         oacc_bcast_align.
587         (wprop_gen): Update type of data_ and usage of oacc_bcast_align.
588         (nvptx_wpropagate): Update type of data and usage of
589         oacc_bcast_{sym,size}.
590         (nvptx_single): Update type of data and usage of oacc_bcast_size.
591         (nvptx_file_end): Update usage of oacc_bcast_{sym,align,size}.
593 2018-12-19  Tom de Vries  <tdevries@suse.de>
595         * config/nvptx/nvptx.md (nvptx_barsync): Add and handle operand.
596         * config/nvptx/nvptx.c (nvptx_wsync): Update call to gen_nvptx_barsync.
598 2018-12-19  Tom de Vries  <tdevries@suse.de>
600         * config/nvptx/nvptx.c (nvptx_single): Always pass false to
601         nvptx_wsync.
602         (nvptx_process_pars): Likewise.
604 2018-12-19  Tom de Vries  <tdevries@suse.de>
606         * config/nvptx/nvptx.c (nvptx_previous_fndecl): Declare.
607         (nvptx_set_current_function): New function.
608         (TARGET_SET_CURRENT_FUNCTION): Define.
610 2018-12-19  Shaokun Zhang  <zhangshaokun@hisilicon.com>
612         * config/aarch64/aarch64-cores.def (tsv110): Fix architecture.  This
613         part is really Armv8.2 with some permitted Armv8.4 extensions.
615 2018-12-19  Jakub Jelinek  <jakub@redhat.com>
617         PR target/88541
618         * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
619         for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
621 2018-12-19  Alexandre Oliva  <aoliva@redhat.com>
623         PR testsuite/86153
624         PR middle-end/83239
625         * vr-values.c
626         (vr_values::vrp_evaluate_conditional_warnv_with_ops): Extend
627         simplification of overflow tests to cover cases in which we
628         can determine the result of the comparison.
630 2018-12-19  Bin Cheng  <bin.cheng@linux.alibaba.com>
632         * auto-profile.c (afdo_indirect_call): Skip generating histogram
633         value if we can't find cgraph_node for then indirected callee.  Save
634         profile_id of the cgraph_node in histogram value's first counter.
635         * value-prof.c (gimple_value_profile_transformations): Don't skip
636         for flag_auto_profile.
638 2018-12-18  Vladimir Makarov  <vmakarov@redhat.com>
640         PR rtl-optimization/87759
641         * lra-assigns.c (lra_split_hard_reg_for): Recalculate
642         non_reload_pseudos.
644 2018-12-18  Jakub Jelinek  <jakub@redhat.com>
646         PR target/88464
647         * config/i386/i386-builtin-types.def
648         (VOID_FTYPE_PDOUBLE_QI_V8SI_V4DF_INT,
649         VOID_FTYPE_PFLOAT_QI_V4DI_V8SF_INT,
650         VOID_FTYPE_PLONGLONG_QI_V8SI_V4DI_INT,
651         VOID_FTYPE_PINT_QI_V4DI_V8SI_INT,
652         VOID_FTYPE_PDOUBLE_QI_V4SI_V2DF_INT,
653         VOID_FTYPE_PFLOAT_QI_V2DI_V4SF_INT,
654         VOID_FTYPE_PLONGLONG_QI_V4SI_V2DI_INT,
655         VOID_FTYPE_PINT_QI_V2DI_V4SI_INT): New builtin types.
656         * config/i386/i386.c (enum ix86_builtins): Add
657         IX86_BUILTIN_SCATTERALTSIV4DF, IX86_BUILTIN_SCATTERALTDIV8SF,
658         IX86_BUILTIN_SCATTERALTSIV4DI, IX86_BUILTIN_SCATTERALTDIV8SI,
659         IX86_BUILTIN_SCATTERALTSIV2DF, IX86_BUILTIN_SCATTERALTDIV4SF,
660         IX86_BUILTIN_SCATTERALTSIV2DI and IX86_BUILTIN_SCATTERALTDIV4SI.
661         (ix86_init_mmx_sse_builtins): Fix up names of IX86_BUILTIN_GATHERALT*,
662         IX86_BUILTIN_GATHER3ALT* and IX86_BUILTIN_SCATTERALT* builtins to
663         match the IX86_BUILTIN codes.  Build    IX86_BUILTIN_SCATTERALTSIV4DF,
664         IX86_BUILTIN_SCATTERALTDIV8SF, IX86_BUILTIN_SCATTERALTSIV4DI,
665         IX86_BUILTIN_SCATTERALTDIV8SI, IX86_BUILTIN_SCATTERALTSIV2DF,
666         IX86_BUILTIN_SCATTERALTDIV4SF, IX86_BUILTIN_SCATTERALTSIV2DI and
667         IX86_BUILTIN_SCATTERALTDIV4SI decls.
668         (ix86_vectorize_builtin_scatter): Expand those new builtins.
670 2018-12-18  Bill Schmidt  <wschmidt@linux.ibm.com>
672         * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
673         Describe when a typedef name can be used as the type specifier for
674         a vector type, and when it cannot.
676 2018-12-18  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
678         * config/msp430/msp430.h: Define TARGET_VTABLE_ENTRY_ALIGN.
680 2018-12-18  Jakub Jelinek  <jakub@redhat.com>
682         PR target/88513
683         PR target/88514
684         * optabs.def (vec_pack_sbool_trunc_optab, vec_unpacks_sbool_hi_optab,
685         vec_unpacks_sbool_lo_optab): New optabs.
686         * optabs.c (expand_widen_pattern_expr): Use vec_unpacks_sbool_*_optab
687         and pass additional argument if both input and target have the same
688         scalar mode of VECTOR_BOOLEAN_TYPE_P vectors.
689         * expr.c (expand_expr_real_2) <case VEC_PACK_TRUNC_EXPR>: Handle
690         VECTOR_BOOLEAN_TYPE_P pack where result has the same scalar mode
691         as the operands using vec_pack_sbool_trunc_optab.
692         * tree-vect-stmts.c (supportable_widening_operation): Use
693         vec_unpacks_sbool_{lo,hi}_optab for VECTOR_BOOLEAN_TYPE_P conversions
694         where both wider_vectype and vectype have the same scalar mode.
695         (supportable_narrowing_operation): Similarly use
696         vec_pack_sbool_trunc_optab if narrow_vectype and vectype have the same
697         scalar mode.
698         * config/i386/i386.c (ix86_get_builtin)
699         <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Check for VECTOR_MODE_P
700         rather than non-VOIDmode.
701         * config/i386/sse.md (vec_pack_trunc_qi, vec_pack_trunc_<mode>):
702         Remove useless ()s around "register_operand", formatting fixes.
703         (vec_pack_sbool_trunc_qi, vec_unpacks_sbool_lo_qi,
704         vec_unpacks_sbool_hi_qi): New expanders.
705         * doc/md.texi (vec_pack_sbool_trunc_M, vec_unpacks_sbool_hi_M,
706         vec_unpacks_sbool_lo_M): Document.
708 2018-12-18  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
710         * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits
711         of src in nonzero_bits_mode if the mode of src is MODE_INT and
712         HWI_COMPUTABLE.
713         (reg_nonzero_bits_for_combine): Add clarification to comment.
715 2018-12-18  Wei Xiao  <wei3.xiao@intel.com>
717         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cascadelake.
718         * config/i386/i386.c (fold_builtin_cpu): Handle cascadelake.
719         * doc/extend.texi: Add cascadelake.
720 2018-12-17  Peter Bergner  <bergner@linux.ibm.com>
722         PR target/87870
723         * config/rs6000/vsx.md (nW): New mode iterator.
724         (vsx_mov<mode>_64bit): Use it.  Remove redundant GPR 0/-1 alternative.
725         Update length attribute for (<??r>, <nW>)  alternative.
726         (vsx_mov<mode>_32bit): Likewise.
728 2018-12-17  Tom de Vries  <tdevries@suse.de>
730         * config/nvptx/nvptx.c (PTX_VECTOR_LENGTH, PTX_WORKER_LENGTH,
731         PTX_DEFAULT_RUNTIME_DIM): Move to the top of the file.
733 2018-12-17  Tom de Vries  <tdevries@suse.de>
735         * config/nvptx/nvptx.c (PTX_WARP_SIZE): Define.
736         (nvptx_simt_vf): Return PTX_WARP_SIZE instead of PTX_VECTOR_LENGTH.
738 2018-12-17  Tom de Vries  <tdevries@suse.de>
740         * config/nvptx/nvptx.c (nvptx_single): Fix whitespace.
741         (nvptx_neuter_pars): Likewise.
743 2018-12-17  Tom de Vries  <tdevries@suse.de>
745         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Work around Fortran
746         bug PR72741 by overriding dims parameter for routines.
748 2018-12-17  Tom de Vries  <tdevries@suse.de>
750         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Rewrite using
751         predicate vars.
753 2018-12-17  Steve Ellcey  <sellcey@cavium.com>
755         * config/aarch64/aarch64-protos.h (aarch64_use_simple_return_insn_p):
756         New prototype.
757         (aarch64_epilogue_uses): Ditto.
758         * config/aarch64/aarch64.c (aarch64_attribute_table): New array.
759         (aarch64_simd_decl_p): New function.
760         (aarch64_reg_save_mode): New function.
761         (aarch64_function_ok_for_sibcall): Check for simd calls.
762         (aarch64_layout_frame): Check for simd function.
763         (aarch64_gen_storewb_pair): Handle E_TFmode.
764         (aarch64_push_regs): Use aarch64_reg_save_mode to get mode.
765         (aarch64_gen_loadwb_pair): Handle E_TFmode.
766         (aarch64_pop_regs): Use aarch64_reg_save_mode to get mode.
767         (aarch64_gen_store_pair): Handle E_TFmode.
768         (aarch64_gen_load_pair): Ditto.
769         (aarch64_save_callee_saves): Handle different mode sizes.
770         (aarch64_restore_callee_saves): Ditto.
771         (aarch64_components_for_bb): Check for simd function.
772         (aarch64_epilogue_uses): New function.
773         (aarch64_process_components): Check for simd function.
774         (aarch64_expand_prologue): Ditto.
775         (aarch64_expand_epilogue): Ditto.
776         (aarch64_expand_call): Ditto.
777         (aarch64_use_simple_return_insn_p): New function.
778         (TARGET_ATTRIBUTE_TABLE): New define.
779         * config/aarch64/aarch64.h (EPILOGUE_USES): Redefine.
780         (FP_SIMD_SAVED_REGNUM_P): New macro.
781         * config/aarch64/aarch64.md (simple_return): New define_expand.
782         (load_pair_dw_tftf): New instruction.
783         (store_pair_dw_tftf): Ditto.
784         (loadwb_pair<TX:mode>_<P:mode>): Ditto.
785         (storewb_pair<TX:mode>_<P:mode>): Ditto.
787 2018-12-17  Uros Bizjak  <ubizjak@gmail.com>
789         PR target/88502
790         * internal-fn.def (ACOSH): New.
791         (ASINH): Ditto.
792         (ATANH): Ditto.
793         * optabs.def (acosh_optab): New.
794         (asinh_optab): Ditto.
795         (atanh_optab): Ditto.
796         * config/i386/i386-protos.h (ix86_emit_i387_asinh): New prototype.
797         (ix86_emit_i387_acosh): Ditto.
798         (ix86_emit_i387_atanh): Ditto.
799         * config/i386/i386.c (ix86_emit_i387_asinh): New function.
800         (ix86_emit_i387_acosh): Ditto.
801         (ix86_emit_i387_atanh): Ditto.
802         * config/i386/i386.md (asinhxf2): New expander.
803         (asinh<mode>2): Ditto.
804         (acoshxf2): Ditto.
805         (acosh<mode>2): Ditto.
806         (atanhxf2): Ditto.
807         (atanh<mode>2): Ditto.
809 2018-12-17  David Edelsohn  <dje.gcc@gmail.com>
811         * config.gcc (powerpc-ibm-aix6.*): Delete extra_headers.
812         (powerpc-ibm-aix7.1.*): Same.
813         (powerpc-ibm-aix[789].*): Same.
815 2018-12-17  H.J. Lu  <hongjiu.lu@intel.com>
817         PR debug/79342
818         * dwarf2out.c (find_AT_string_in_table): Add insert argument
819         defaulting to INSERT and replace INSERT.
820         (find_AT_string): Likewise.
821         (output_macinfo_op): Pass NO_INSERT to find_AT_string.
823 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
825         * coverage.c (struct conts_entry): Add n_counts.
826         (remap_counts_file): Record number of ocunts.
827         (get_coverage_counts): Verify that counts match.
828         * coverage.h (get_coverage_counts): Update prototype.
829         * profile.c (get_exec_counts. compute_value_histograms): Add
830         n_counts parametrs.
832 2018-12-17  Senthil Kumar Selvaraj  <senthilkumar.selvaraj@microchip.com>
834         PR rtl-optimization/88253
835         * combine.c (combine_simplify_rtx): Test for side-effects before
836         substituting by zero.
838 2018-12-17  Richard Sandiford  <richard.sandiford@arm.com>
839             Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
840             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
842         * doc/invoke.texi (-fversion-loops-for-strides): Document
843         (loop-versioning-group-size, loop-versioning-max-inner-insns)
844         (loop-versioning-max-outer-insns): Document new --params.
845         * Makefile.in (OBJS): Add gimple-loop-versioning.o.
846         * common.opt (fversion-loops-for-strides): New option.
847         * opts.c (default_options_table): Enable fversion-loops-for-strides
848         at -O3.
849         * params.def (PARAM_LOOP_VERSIONING_GROUP_SIZE)
850         (PARAM_LOOP_VERSIONING_MAX_INNER_INSNS)
851         (PARAM_LOOP_VERSIONING_MAX_OUTER_INSNS): New parameters.
852         * passes.def: Add pass_loop_versioning.
853         * timevar.def (TV_LOOP_VERSIONING): New time variable.
854         * tree-ssa-propagate.h
855         (substitute_and_fold_engine::substitute_and_fold): Add an optional
856         block parameter.
857         * tree-ssa-propagate.c
858         (substitute_and_fold_engine::substitute_and_fold): Likewise.
859         When passed, only walk blocks dominated by that block.
860         * tree-vrp.h (range_includes_p): Declare.
861         (range_includes_zero_p): Turn into an inline wrapper around
862         range_includes_p.
863         * tree-vrp.c (range_includes_p): New function, generalizing...
864         (range_includes_zero_p): ...this.
865         * tree-pass.h (make_pass_loop_versioning): Declare.
866         * gimple-loop-versioning.cc: New file.
868 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
870         * ipa-fnsummary.c (remap_edge_change_prob): Do not ICE when changes
871         are not streamed in.
873 2018-12-15  H.J. Lu  <hongjiu.lu@intel.com>
875         PR target/88483
876         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Revert
877         reversion 267133.
879 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
881         * ipa-fnsummary.c (analyze_function_body): Do not loeak conds and
882         size_time_table.
883         (ipa_fn_summary_generate): Add prevails parameter; do not allocate
884         data when symbol is not prevailing.
885         (inline_read_section): Likewise.
887 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
889         * cgraph.h (cgraph_node): Add predicate prevailing_p.
890         (cgraph_edge): Add predicate possible_call_in_translation_unit_p.
891         * ipa-prop.c (ipa_write_jump_function): Optimize streaming of ADDR_EXPR.
892         (ipa_read_jump_function): Add prevails parameter; optimize streaming.
893         (ipa_read_edge_info): Break out from ...
894         (ipa_read_node_info): ... here; optimize streaming.
895         * cgraph.c (cgraph_edge::possibly_call_in_translation_unit_p): New
896         predicate.
898 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
900         * ipa-utils.c (ipa_merge_profiles): Do no merging when source function
901         has zero count.
903 2018-12-15  Jakub Jelinek  <jakub@redhat.com>
905         PR tree-optimization/88464
906         PR target/88498
907         * tree-vect-stmts.c (vect_build_gather_load_calls): For NARROWING
908         and mask with integral masktype, don't try to permute mask vectors,
909         instead emit VEC_UNPACK_{LO,HI}_EXPR.  Fix up NOP_EXPR operand.
910         (vectorizable_store): Handle masked scatters with decl and integral
911         mask type.
912         (permute_vec_elements): Allow scalar_dest to be NULL.
913         * config/i386/i386.c (ix86_get_builtin)
914         <case IX86_BUILTIN_GATHER3ALTDIV16SF>: Use lowpart_subreg for masks.
915         <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Don't assume mask and src have
916         to be the same.
918 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
920         * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
921         parameters.
922         (cgraph_build_static_cdtor): Update.
923         (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
924         callsed.
926 2018-12-15  Alan Modra  <amodra@gmail.com>
928         * config/rs6000/rs6000.c (generate_set_vrsave, rs6000_emit_savres_rtx),
929         (rs6000_emit_prologue, rs6000_call_aix, rs6000_call_sysv),
930         (rs6000_call_darwin_1): Use gen_hard_reg_clobber.
932 2018-12-15  Jakub Jelinek  <jakub@redhat.com>
934         PR target/88489
935         * config/i386/sse.md (UNSPEC_SFIXUPIMM): New unspec enumerator.
936         (avx512f_sfixupimm<mode><mask_name><round_saeonly_name>): Use it
937         instead of UNSPEC_FIXUPIMM.
939         PR rtl-optimization/88478
940         * cselib.c (cselib_record_sets): Move sets[i].src_elt tests
941         after REG_P (dest) test.
943 2018-12-14  Martin Sebor  <msebor@redhat.com>
945         PR tree-optimization/88372
946         * calls.c (maybe_warn_alloc_args_overflow): Handle function pointers.
947         * tree-object-size.c (alloc_object_size): Same.  Simplify.
948         * doc/extend.texi (Object Size Checking): Update.
949         (Other Builtins): Add __builtin_object_size.
950         (Common Type Attributes): Add alloc_size.
951         (Common Variable Attributes): Ditto.
953 2018-12-14  Martin Sebor  <msebor@redhat.com>
955         PR rtl-optimization/87096
956         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid
957         folding calls whose bound may exceed INT_MAX.  Diagnose bound ranges
958         that exceed the limit.
960 2018-12-14  Martin Sebor  <msebor@redhat.com>
962         PR web/79738
963         * doc/extend.texi (attribute const, pure): Clarify.
965 2018-12-14  H.J. Lu  <hongjiu.lu@intel.com>
967         * config/i386/i386.c (rest_of_insert_endbranch): Insert ENDBR
968         at the function entry only when -mmanual-endbr isn't used or
969         there is cf_check function attribute.
970         (ix86_attribute_table): Add cf_check.
971         * config/i386/i386.opt: Add -mmanual-endbr.
972         * doc/extend.texi: Document cf_check attribute.
973         * doc/invoke.texi: Document -mmanual-endbr.
975 2018-12-14  Thomas Schwinge  <thomas@codesourcery.com>
976             Cesar Philippidis  <cesar@codesourcery.com>
978         * omp-offload.c (inform_oacc_loop): New function.
979         (execute_oacc_device_lower): Use it to display loop parallelism.
981 2018-12-14  Jakub Jelinek  <jakub@redhat.com>
983         PR c++/82294
984         PR c++/87436
985         * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
986         * expr.c (categorize_ctor_elements_1): Likewise.  Compute it like
987         p_nz_elts, except don't multiply it by mult.  Adjust recursive call.
988         Fix up COMPLEX_CST handling.
989         (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
990         it and pass it through to categorize_ctor_elements_1.
991         (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
992         * gimplify.c (gimplify_init_constructor): Likewise.  Don't force
993         ctor into readonly data section if num_unique_nonzero_elements is
994         smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
995         bytes.
997 2018-12-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
999         * config/arm/arm.md (arm_movdi): Split if -mslow-flash-data and
1000         source is a constant that would be loaded by literal pool.
1001         (movsf expander): Generate a no_literal_pool_sf_immediate insn if
1002         -mslow-flash-data is present, targeting hardfloat ABI and source is a
1003         float constant that cannot be loaded via vmov.
1004         (movdf expander): Likewise but generate a no_literal_pool_df_immediate
1005         insn.
1006         (arm_movsf_soft_insn): Split if -mslow-flash-data and source is a
1007         float constant that would be loaded by literal pool.
1008         (softfloat constant movsf splitter): Splitter for the above case.
1009         (movdf_soft_insn): Split if -mslow-flash-data and source is a float
1010         constant that would be loaded by literal pool.
1011         (softfloat constant movdf splitter): Splitter for the above case.
1012         * config/arm/constraints.md (Pz): Document existing constraint.
1013         (Ha): Define constraint.
1014         (Tu): Likewise.
1015         * config/arm/predicates.md (hard_sf_operand): New predicate.
1016         (hard_df_operand): Likewise.
1017         * config/arm/thumb2.md (thumb2_movsi_insn): Split if
1018         -mslow-flash-data and constant would be loaded by literal pool.
1019         * constant/arm/vfp.md (thumb2_movsi_vfp): Likewise and disable constant
1020         load in VFP register.
1021         (movdi_vfp): Likewise.
1022         (thumb2_movsf_vfp): Use hard_sf_operand as predicate for source to
1023         prevent match for a constant load if -mslow-flash-data and constant
1024         cannot be loaded via vmov.  Adapt constraint accordingly by
1025         using Ha instead of E for generic floating-point constant load.
1026         (thumb2_movdf_vfp): Likewise using hard_df_operand predicate instead.
1027         (no_literal_pool_df_immediate): Add a clobber to use as the
1028         intermediate general purpose register and also enable it after reload
1029         but disable it constant is a valid FP constant.  Add constraints and
1030         generate a DI intermediate load rather than 2 SI loads.
1031         (no_literal_pool_sf_immediate): Add a clobber to use as the
1032         intermediate general purpose register and also enable it after
1033         reload.
1035 2018-12-14  Uros Bizjak  <ubizjak@gmail.com>
1037         PR target/88474
1038         * internal-fn.def (HYPOT): New.
1039         * optabs.def (hypot_optab): New.
1040         * config/i386/i386.md (hypot<mode>3): New expander.
1042 2018-12-14  Jeff Law  <law@redhat.com>
1044         * target.def (post_cfi_startproc): Fix text.
1046 2018-12-14  Sam Tebbs  <sam.tebbs@arm.com>
1048         * doc/tm.texi (TARGET_ASM_POST_CFI_STARTPROC): Define.
1049         * doc/tm.texi.in (TARGET_ASM_POST_CFI_STARTPROC): Define.
1050         * dwarf2out.c (dwarf2out_do_cfi_startproc): Trigger the hook.
1051         * hooks.c (hook_void_FILEptr_tree): Define.
1052         * hooks.h (hook_void_FILEptr_tree): Define.
1053         * target.def (post_cfi_startproc): Define.
1055 2018-12-14  Tom de Vries  <tdevries@suse.de>
1057         * lto-cgraph.c (verify_node_partition): New function.
1058         (input_overwrite_node, input_varpool_node): Use verify_node_partition.
1060 2018-12-14  H.J. Lu  <hongjiu.lu@intel.com>
1062         PR target/88483
1063         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
1064         use get_frame_size ().
1066 2018-12-14  Andrew Stubbs  <ams@codesourcery.com>
1068         * gcc/lra-int.h (lra_register_new_scratch_op): Add third parameter.
1069         * gcc/lra-remat.c (update_scratch_ops): Pass icode to
1070         lra_register_new_scratch_op.
1071         * gcc/lra.c (struct sloc): Add icode field.
1072         (lra_register_new_scratch_op): Add icode parameter.
1073         Use icode to skip insns that have changed beyond recognition.
1075 2018-12-14  Eric Botcazou  <ebotcazou@adacore.com>
1077         * dwarf2out.c (analyze_discr_in_predicate): Simplify.
1078         (analyze_variants_discr): Deal with naked boolean discriminants.
1080 2018-12-14  Claudiu Zissulescu  <claziss@synopsys.com>
1082         * config/arc/arc.c (arc_restore_callee_milli) Don't clobber off
1083         variable.
1085 2018-12-14  Claudiu Zissulescu  <claziss@synopsys.com>
1087         * config/arc/arc.h (reg_class): Add SIBCALL_REGS.
1089 2018-12-14  Richard Biener  <rguenther@suse.de>
1091         * gimple-ssa-split-paths.c (is_feasible_trace): Remove
1092         duplicated code block.
1094 2018-12-14  Alan Modra  <amodra@gmail.com>
1096         PR rtl-optimization/88311
1097         * config/rs6000/rs6000.c (rs6000_call_sysv): Do not mask cookie.
1098         (rs6000_sibcall_sysv): Likewise.
1100 2018-12-14  Segher Boessenkool  <segher@kernel.crashing.org>
1102         PR rtl-optimization/88001
1103         * function.c (match_asm_constraints_1): Don't invalidly share RTL.
1105 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
1107         * auto-profile.c (afdo_annotate_cfg): Call update_max_bb_count even
1108         if autofdo counts are all zeros.
1110 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
1112         * auto-profile.c (afdo_indirect_call): Skip generating histogram
1113         value for internal call.
1115 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
1117         * auto-profile.c (AFDO_EINFO): New macro.
1118         (class edge_info): New class.
1119         (is_edge_annotated, set_edge_annotated): Delete.
1120         (afdo_propagate_edge, afdo_propagate_circuit, afdo_propagate): Remove
1121         parameter.  Adjust edge count computation and annotation using class
1122         edge_info.
1123         (afdo_calculate_branch_prob, afdo_annotate_cfg): Likewise.
1125 2018-12-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1127         * ipa-cp.c (print_all_lattices): Skip cp clones.
1129         * ipa-cp.c: Fix various comment typos.
1131 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
1133         PR tree-optimization/88444
1134         * tree-inline.c (fold_marked_statements): Iterate up to
1135         last_basic_block_for_fn rather than n_basic_blocks_for_fn.
1137         PR rtl-optimization/88470
1138         * cfgcleanup.c (outgoing_edges_match): If the function is
1139         shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
1140         edge to EXIT, return false.
1142 2018-12-13  Martin Sebor  <msebor@redhat.com>
1144         * doc/extend.texi: Consistently use @code for const and volatile
1145         qualifiers, the true and false constants, and asm statements.
1147 2018-12-13  Vladimir Makarov  <vmakarov@redhat.com>
1149         PR rtl-optimization/88414
1150         * lra-int.h (lra_asm_error_p): New.
1151         * lra-assigns.c (lra_assign): Check lra_asm_error_p for checking
1152         call crossed pseudo assignment correctness.
1153         (lra_split_hard_reg_for): Set up lra_asm_error_p.
1154         * lra-constraints.c (curr_insn_transform): Ditto.
1155         * lra.c (lra_asm_error_p): New.
1156         (lra): Reset lra_asm_error_p.
1158 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
1160         PR rtl-optimization/88416
1161         * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
1162         even if !AUTO_INC_DEC.
1164 2018-12-13  David Edelsohn  <dje.gcc@gmail.com>
1166         * config/rs6000/rs6000.c (rs6000_function_arg): Ensure type is
1167         non-NULL.
1168         (rs6000_arg_partial_bytes): Same.
1170 2018-12-13  Jason Merrill  <jason@redhat.com>
1172         * gdbinit.in (pp): New macro.
1173         (pbb): Remove.
1175 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
1177         PR tree-optimization/88464
1178         * tree-vect-stmts.c (vect_build_gather_load_calls): Handle INTEGER_TYPE
1179         masktype if mask is non-NULL.
1180         (vectorizable_load): Don't reject masked gather loads if masktype
1181         in the decl is INTEGER_TYPE.
1183 2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>
1185         * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
1186         s390_vec_double_s64 instead of s390_vcdgb.
1187         (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
1188         s390_vcdlgb.
1190 2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>
1192         * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
1193         ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
1194         ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
1196 2018-12-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1198         * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
1199         behavior.
1201 2018-12-13  Wei Xiao  <wei3.xiao@intel.com>
1203         * common/config/i386/i386-common.c (processor_names): Add cascadelake.
1204         (processor_alias_table): Add cascadelake.
1205         * config.gcc: Add -march=cascadelake.
1206         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1207         cascadelake.
1208         * config/i386/i386.c (Add m_CASCADELAKE): New.
1209         (processor_cost_table): Add cascadelake.
1210         (get_builtin_code_for_version): Handle cascadelake.
1211         * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New.
1212         (PTA_CASCADELAKE): Ditto.
1213         * doc/invoke.texi: Add -march=cascadelake.
1215 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
1217         PR target/88465
1218         * config/i386/i386.md (*movdi_internal, *movsi_internal,
1219         *movhi_internal, *movqi_internal): Add alternative(s) to load
1220         0 or -1 into k registers using kxor or kxnoq instructions.
1222         PR target/88461
1223         * config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New
1224         mode iterators.
1225         (<avx512>_testm<mode>3<mask_scalar_merge_name>,
1226         <avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns
1227         with VI12_AVX512VL and VI48_AVX512VL iterators into ones with
1228         VI1248_AVX512VLBW iterator.
1229         (*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext,
1230         *<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask,
1231         *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext,
1232         *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New
1233         define_insns.
1235         PR target/88461
1236         * config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2,
1237         *zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives.
1239 2018-12-12  Tom de Vries  <tdevries@suse.de>
1240             Jakub Jelinek  <jakub@redhat.com>
1242         * omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1243         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1244         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1245         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1246         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1247         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up
1248         function types - remove one argument.
1250 2018-12-12  Martin Sebor  <msebor@redhat.com>
1252         * doc/extend.texi (attribute copy): Fix typos.
1254 2018-12-12  Martin Sebor  <msebor@redhat.com>
1256         * doc/extend.texi (Function Attributes): Clarify C++ aspects.
1257         (Variable Attributes): Same.
1258         (Type Attributes): Same.
1260 2018-12-12  Olivier Hainque  <hainque@adacore.com>
1262         * config/aarch64/aarch64.c (aarch64_override_options): Once arch,
1263         cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if
1264         defined.
1266 2018-12-12  Olivier Hainque  <hainque@adacore.com>
1268         * config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as
1269         R9_REGNUM instead of 9.
1270         (PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10.
1272 2018-12-12  Eric Botcazou  <ebotcazou@adacore.com>
1274         PR target/86806
1275         * config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1276         (speculation_barrier): New instruction for V9.
1278 2018-12-12  Segher Boessenkool  <segher@kernel.crashing.org>
1279             Iain Sandoe  <iain@sandoe.co.uk>
1281         PR target/88343
1282         * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1283         unless it has been used.
1284         (first_reg_to_save): Remove dead code.
1286 2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>
1288         * config/rs6000/rs6000-protos.h (rs6000_call_darwin): New.
1289         (rs6000_sibcall_darwin): New. (macho_call_template): Remove.
1290         * config/rs6000/rs6000.c (get_prev_label): Forward declaration. 
1291         (rs6000_call_template_1): Handle Darwin.
1292         (macho_call_template): Remove.
1293         (rs6000_call_sysv): Remove handling for Darwin.
1294         (rs6000_call_darwin_1, rs6000_call_darwin): New
1295         (rs6000_sibcall_darwin): New.
1296         * config/rs6000/rs6000.md (define_expand “call”): Handle Darwin
1297         with its own expander. (define_expand “call_value”): Likewise.
1298         (define_expand “sibcall”): Likewise.
1299         (define_expand “sibcall_value”): Likewise.
1300         (call_nonlocal_sysv): Remove Darwin special-casing.
1301         (call_value_nonlocal_sysv): Likewise.
1303 2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>
1305         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove.
1306         (call_nonlocal_darwin64): Remove.
1307         (call_value_indirect_nonlocal_darwin64): Remove.
1308         (call_value_nonlocal_darwin64): Remove.
1309         * config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with
1310         the same asm output as AIX/ELFv2.  (rs6000_call_sysv): Preserve the
1311         CALL_LONG flag when needed for Darwin.
1312         * config/rs6000/rs6000.md (define expand “call”): Expand Darwin as
1313         per sysv.
1314         (define_expand “call_value”): Likewise.
1315         (define_expand “sibcall”): Likewise.
1316         (define_expand “sibcall_value”): Likewise.
1317         (call_indirect_nonlocal_sysv): Mark the clobber mode P.
1318         (call_nonlocal_sysv): Likewise.
1319         (call_nonlocal_sysv_secure): Likewise.
1320         (call_value_indirect_nonlocal_sysv): Likewise.
1321         (call_value_nonlocal_sysv): Likewise.
1322         (call_value_nonlocal_sysv_secure): Likewise.
1323         (call_local64): Mark the clobber DI.
1324         (call_value_local64): Likewise.
1326 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
1328         PR tree-optimization/88444
1329         * tree-vrp.c (register_edge_assert_for_2): Only register assertions
1330         for conversions if rhs1 is a SSA_NAME.
1332 2018-12-11  Dimitar Dimitrov  <dimitar@dinux.eu>
1334         * cfgexpand.c (asm_clobber_reg_is_valid): Also produce
1335         error when stack pointer is clobbered.
1336         (expand_asm_stmt): Refactor clobber check in separate function.
1338 2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>
1340         * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
1341         accounting for STACK_BOUNDARY 128.
1342         (STACK_DYNAMIC_OFFSET): Likewise.
1344 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
1346         PR target/88425
1347         * config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>):
1348         New define_insn_and_split.
1350 2018-12-11  Richard Biener  <rguenther@suse.de>
1352         PR middle-end/88448
1353         PR middle-end/88415
1354         * gimple.c (gimple_assign_set_rhs_with_ops): Revert previous
1355         change.
1356         * tree-complex.c (update_complex_assignment): Properly transfer
1357         or clean EH info around gimple_assign_set_rhs_with_ops.
1359 2018-12-11  Richard Earnshaw  <rearnsha@arm.com>
1361         PR target/87369
1362         * config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode.
1363         (Vbtype): Add SFmode mapping.
1364         * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
1365         (copysign<GPF:mode>3): New expand pattern.
1366         (copysign<GPF:mode>3_insn): New insn pattern.
1368 2018-12-11  Richard Biener  <rguenther@suse.de>
1370         * ccmp.c (ccmp_candidate_p): Use GIMPLE API properly.
1372 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
1374         PR lto/86004
1375         * doc/sourcebuild.texi (lto_incremental): Document new effective
1376         target.
1378 2018-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
1380         PR target/88145
1381         * config/rs6000/rs6000.c (rs6000_expand_zeroop_builtin): Use
1382         rs6000_isa_flags instead of rs6000_isa_flags_explicit to decide
1383         whether soft float is enabled.
1384         (rs6000_expand_mtfsb_builtin): Ditto.
1385         (rs6000_expand_set_fpscr_rn_builtin): Ditto.
1386         (rs6000_expand_set_fpscr_drn_builtin): Ditto.
1388 2018-12-10  Jeff Law  <law@redhat.com>
1390         PR tree-optimization/80520
1391         * gimple-ssa-split-paths.c (is_feasible_trace): Recognize half
1392         diamonds that are likely if convertable.
1394 2018-12-10  Martin Sebor  <msebor@redhat.com>
1396         PR tree-optimization/86196
1397         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
1398         base size only of arrays.
1400 2018-12-10  Segher Boessenkool  <segher@kernel.crashing.org>
1402         * config.gcc (Obsolete configurations): Delete powerpc*-*-*spe*.
1403         (Unsupported targets): Add powerpc*-*-*spe*.
1404         (powerpc*-*-*spe*): Delete.
1405         (powerpc-*-eabispe*): Delete.
1406         (powerpc-*-rtems*spe*): Delete.
1407         (powerpc*-*-linux*spe*): Delete.
1408         (powerpc*-*-linux*): Do not handle the linux*spe* targets.
1409         (powerpc-wrs-vxworks*spe): Delete.
1410         (with_cpu setting code): Delete powerpc*-*-*spe* handling.
1411         * config.host (target powerpc*-*-*spe*): Delete.
1412         * doc/invoke.texi (PowerPC SPE Options): Delete.
1413         (PowerPC SPE Options): Delete.
1414         * config/powerpcspe: Delete.
1416 2018-12-10  Uros Bizjak  <ubizjak@gmail.com>
1418         PR target/88418
1419         * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
1420         check operand 1 with vector_operand predicate.
1421         (ix86_expand_sse_movcc): For vector modes, check op_true with
1422         vector_operand, not nonimmediate_operand.
1424 2018-12-10  Richard Biener  <rguenther@suse.de>
1426         * tree-affine.c (tree_to_aff_combination): Remove unreachable
1427         MEM_REF case.
1428         (aff_combination_expand): Cache on SSA names, not possibly
1429         on conversion trees.  Avoid expanding cases we do not handle.
1431 2018-12-10  Richard Biener  <rguenther@suse.de>
1433         PR tree-optimization/88427
1434         * vr-values.c (vr_values::extract_range_from_phi_node):
1435         Handle symbolic ranges conservatively when trying to drop
1436         to Inf +- 1.
1438 2018-12-10  Martin Jambor  <mjambor@suse.cz>
1440         PR ipa/88214
1441         * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1442         we check pointers against pointers.
1444 2018-12-10  Richard Biener  <rguenther@suse.de>
1446         PR middle-end/88415
1447         * gimple.c (gimple_assign_set_rhs_with_ops): Transfer EH
1448         info to a newly allocated stmt.
1450 2018-12-10  Jerome Lambourg  <lambourg@adacore.com>
1452         * config/vxworksae.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): Define.
1454 2018-12-10  Fredrik Nyström  <fredrik@lysator.liu.se>
1456         PR bootstrap/65725
1457         * config/sol2.h: Only use libgcc-unwind.map if
1458         ENABLE_SHARED_LIBGCC.
1460 2018-12-10  Bin Cheng  <bin.cheng@linux.alibaba.com>
1462         * auto-profile.c (afdo_calculate_branch_prob): Convert profile_count
1463         to afdo precision quality.
1464         (afdo_callsite_hot_enough_for_early_inline): Likewise.
1465         (afdo_annotate_cfg): Set profile_count for exit basic block.
1467 2018-12-10  Bin Cheng  <bin.cheng@linux.alibaba.com>
1469         * auto-profile.c (string_table::get_index_by_decl)
1470         (function_instance::get_function_instance_by_decl): Avoid infinite
1471         recursion by using DECL_FROM_INLINE.
1473 2018-12-09  John David Anglin  <danglin@gcc.gnu.org>
1475         * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file.  Define
1476         d_target_objs.
1477         * config/pa/pa-protos.h (pa_d_target_versions): New prototype.
1478         * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define.
1479         * config/pa/pa-d.c: New file.
1480         * config/pa/t-pa: New file.
1482 2018-12-08  Bin Cheng  <bin.cheng@linux.alibaba.com>
1484         * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for
1485         zero profile count.
1487 2018-12-08  Jakub Jelinek  <jakub@redhat.com>
1489         PR fortran/88304
1490         * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
1491         lhs is not a decl normally, don't call use_pointer_in_frame on that
1492         lhs.
1494 2018-12-08  Eric Botcazou  <ebotcazou@adacore.com>
1496         PR rtl-optimization/88390
1497         * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
1498         (cfi_row_equal_p): Compare it.
1499         (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
1500         If FAKE is false, set window_save of the current row.
1501         (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
1502         function with FAKE parameter set to true.
1503         <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
1504         (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
1506 2018-12-07  Peter Bergner  <bergner@linux.ibm.com>
1508         PR target/87496
1509         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1510         -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
1511         Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
1512         * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
1513         require -mlong-double-128.
1515 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1517         * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
1518         sve_width entry.
1519         (aarch64_parse_sve_width_string): Define.
1521 2018-12-07  Jeff Law  <law@redhat.com>
1523         PR middle-end/87813
1524         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
1525         m_update_global_ranges member.  Add corresponding argument to ctor.
1526         * gimple-ssa-evrp-analyze.c
1527         (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
1528         initialize m_update_global_ranges.
1529         (evrp_range_analyzer::set_ssa_range_info): Assert that we are
1530         updating global ranges.
1531         (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
1532         update global ranges if explicitly requested.
1533         (evrp_range_analyzer::record_ranges_from_phis): Similarly.
1534         (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
1535         * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
1536         evrp_range_analyzer ctor.
1537         * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
1538         * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
1540 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1542         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
1543         Add SVE_NOT_IMPLEMENTED value.
1544         * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
1545         field.
1546         * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
1547         cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
1548         cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
1549         thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
1550         saphira_tunings, thunderx2t99_tunings, emag_tunings):
1551         Specify sve_width.
1552         (aarch64_estimated_poly_value): Define.
1553         (TARGET_ESTIMATED_POLY_VALUE): Define.
1555 2018-12-07  Paul A. Clarke  <pc@us.ibm.com>
1557         PR target/88408
1558         * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
1560 2018-12-07  Vladimir Makarov  <vmakarov@redhat.com>
1562         PR rtl-optimization/88349
1563         * ira-costs.c (record_operand_costs): Check bigger reg class on
1564         NO_REGS.
1566 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
1568         * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
1569         Split the patterns after reload if we don't need the predicate
1570         operand.
1571         (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
1573 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
1575         * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
1576         iterator.
1577         (sve_fp_op): Handle minus and mult.
1578         * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
1579         (*mul<mode>3): Split the patterns after reload if we don't
1580         need the predicate operand.
1581         (*post_ra_<sve_fp_op><mode>3): New pattern.
1583 2018-12-07  Bin Cheng  <bin.cheng@linux.alibaba.com>
1585         * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
1586         return condition when *this is precise zero.
1588 2018-12-07  Jakub Jelinek  <jakub@redhat.com>
1590         PR target/85593
1591         * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
1592         for functions with naked attribute.
1594         PR c/88367
1595         * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
1596         with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
1597         is non-NULL and offset is known to have most significant bit clear.
1598         * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
1599         of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
1600         most significant bit clear.  If offset does have most significant bit
1601         set and -fno-delete-null-pointer-checks, don't return true even if
1602         the base pointer is non-NULL.
1604 2018-12-06  Alexandre Oliva  <aoliva@redhat.com>
1606         * cselib.c (cselib_record_sets): Skip strict low part sets
1607         with NULL src_elt.
1609 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
1611         PR target/88316
1612         * config/rs6000/smmintrin.h: New file.
1613         * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
1615 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
1617         PR target/88316
1618         * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
1619         (_mm_unpacklo_pi8): Likewise.
1620         (_mm_mulhi_pi16): Likewise.
1621         (_mm_packs_pi16): Fix for big-endian. Use preferred API.
1622         (_mm_packs_pi32): Likewise.
1623         (_mm_packs_pu16): Likewise.
1624         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
1625         (_mm_cvtss_si64): Likewise.
1626         (_mm_cvtpi32x2_ps): Likewise.
1627         (_mm_shuffle_ps): Likewise.
1628         (_mm_movemask_pi8): Likewise.
1629         (_mm_mulhi_pu16): Likewise.
1630         (_mm_sad_pu8): Likewise.
1631         (_mm_sad_pu8): Likewise.
1632         (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
1633         (_mm_cvtpu8_ps): Likewise.
1634         (_mm_movemask_ps): Better #else case for big-endian (no functional
1635         change).
1636         (_mm_shuffle_pi16): Likewise.
1637         * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
1638         Better #else case for big-endian (no functional change).
1639         (_mm_movemask_epi8): Likewise.
1640         (_mm_shufflehi_epi16): Likewise.
1641         (_mm_shufflelo_epi16): Likewise.
1642         (_mm_shuffle_epi32): Likewise.
1643         (_mm_mul_epu32): Fix for big-endian.
1644         (_mm_bsrli_si128): Likewise.
1645         (_mm_cvtps_pd): Better #else case for big endian.
1646         (_mm_mulhi_epi16): Likewise.
1647         (_mm_mul_epu32): Likewise.
1648         (_mm_slli_si128): Likewise.
1649         (_mm_sll_epi16): Likewise.
1650         (_mm_sll_epi32): Likewise.
1651         (_mm_sra_epi16): Likewise.
1652         (_mm_sra_epi32): Likewise.
1653         (_mm_srl_epi16): Likewise.
1654         (_mm_srl_epi32): Likewise.
1655         (_mm_mulhi_epu16): Likewise.
1656         (_mm_sad_epu8): Likewise.
1657         * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
1658         (_mm_sub_ps): Likewise.
1659         * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
1660         * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
1661         macros consistently (no functional changes).
1662         (_mm_alignr_pi8): Likewise.
1664 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
1666         PR c++/87380
1667         * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
1668         default.
1669         * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
1671 2018-12-06  Vladimir Makarov  <vmakarov@redhat.com>
1673         PR target/88282
1674         * ira.c (ira_init_register_move_cost): Use info from
1675         hard_regno_mode_ok instead of contains_reg_of_mode.
1676         * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
1677         hard register class for some fixed hard registers.
1679 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
1681         * doc/extend.texi (Using Assembly Language with C): Document asm inline.
1682         (Size of an asm): Fix typo.  Document asm inline.
1683         * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
1684         * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
1685         (gimple_asm_set_volatile): Fix typo.
1686         (gimple_asm_inline_p): New.
1687         (gimple_asm_set_inline): New.
1688         * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
1689         tree to gimple.
1690         * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
1691         gimple_asm_inline_p flag, too.
1692         * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
1693         in an ASM_EXPR.
1694         * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
1695         a minimum size for an asm.
1696         * tree.h (ASM_INLINE_P): New.
1698 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
1700         PR inline-asm/55681
1701         * doc/extend.texi (Basic Asm): Update grammar.
1702         (Extended Asm): Update grammar.
1704 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
1706         PR target/78444
1707         * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
1708         * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
1709         stack alignment in non-leaf functions.
1711 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
1713         PR target/87598
1714         * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
1715         call output_operand_lossage on VOIDmode CONST_INTs.  After
1716         output_operand_lossage do return false.
1718 2018-12-06  Richard Biener  <rguenther@suse.de>
1720         * df-problems.c (df_rd_local_compute): Use bitmap_release.
1721         (df_live_free): Likewise.
1722         (df_md_local_compute): Likewise.
1723         (df_md_free): Release df_md_scratch bitmap.
1724         * loop-invariant.c (calculate_loop_reg_pressure): Use
1725         bitmap_release.
1726         * sched-deps.c (true_dependency_cache, output_dependency_cache,
1727         anti_dependency_cache, control_dependency_cache,
1728         spec_dependency_cache): Use bitmap instead of bitmap_head *.
1729         * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
1730         dont_calc_deps as bitmap allocated from obstack not GC.
1731         (schedule_ebbs_finish): Use bitmap_release.
1732         * sched-rgn.c (schedule_insns): Initialize non-GTY
1733         not_in_df as bitmap allocated from obstack not GC.
1734         Use bitmap_release.
1735         * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
1736         (sel_region_init): Allocate forced_ebb_heads.
1737         (sel_region_finish): Free forced_ebb_heads.
1739 2018-12-06  Richard Biener  <rguenther@suse.de>
1741         * bitmap.c (bitmap_head::crashme): Define.
1742         * bitmap.h (bitmap_head): Add constexpr default constructor
1743         poisoning the obstack member.
1744         (bitmap_head::crashme): Declare.
1745         (bitmap_release): New function clearing a bitmap and poisoning
1746         the obstack member.
1747         * gengtype.c (main): Make it recognize CONSTEXPR.
1748         * lra-constraints.c (lra_inheritance): Use bitmap_release
1749         instead of bitmap_clear.
1750         * ira.c (ira): Work around class-memaccess warning.
1751         * regrename.c (create_new_chain): Likewise.
1753 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
1755         PR tree-optimization/85726
1756         * generic-match-head.c (optimize_successive_divisions_p): New function.
1757         * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
1758         * match.pd: Don't combine successive divisions if they aren't exact
1759         and optimize_successive_divisions_p is false.
1761         PR fortran/88304
1762         * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
1763         for non-local automatic decls.
1765 2018-12-05  David Edelsohn  <dje.gcc@gmail.com>
1767         * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
1768         default. 
1770 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
1772         * dwarf2out.c (field_byte_offset): Remove
1773         #ifdef PCC_BITFIELD_TYPE_MATTERS check.
1775 2018-12-05  Richard Biener  <rguenther@suse.de>
1776             Martin Sebor  <msebor@redhat.com>
1778         PR c/87028
1779         * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
1780         SSA_NAME_VAR.
1781         * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
1782         * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
1784 2018-12-05  Iain Sandoe  <iain@sandoe.co.uk>
1786         * configure.ac (gcc_cv_otool): Set.
1787         * configure: Regenerate.
1789 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
1791         * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
1792         (GROUP_BITFIELDS_BY_ALIGN): Ditto.
1794 2018-12-05  Richard Sandiford  <richard.sandiford@arm.com>
1796         PR tree-optimization/88064
1797         * tree-vect-patterns.c (vect_convert_input): Convert the result of
1798         an existing cast if it has the right width but the wrong sign.
1799         Do not test the signedness of the required result when
1800         considering whether to split an existing cast; instead split to
1801         a type with the same signedness as the source of the cast, then
1802         convert it to the opposite signedness where necessary.
1803         (vect_recog_over_widening_pattern): Handle sign changes between
1804         the final PLUS_EXPR and the RSHIFT_EXPR.
1805         (vect_recog_average_pattern): Use an unsigned operation when
1806         truncating an addition, subtraction or multiplication.  Cast the
1807         result back to the "real" signedness before promoting.
1809 2018-12-05  Richard Biener  <rguenther@suse.de>
1811         PR tree-optimization/86637
1812         * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
1813         vect_location at the end.
1815 2018-12-05  Jakub Jelinek  <jakub@redhat.com>
1817         PR tree-optimization/87360
1818         * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1819         data dependencies, don't return false, just continue.  Formatting
1820         fixes.
1821         (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
1822         fuse_loops): Formatting fixes.
1824         PR sanitizer/88333
1825         * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
1826         aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
1828 2018-12-05  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1830         * config.gcc (csky-*-linux-gnu*): Force .init_array support.
1832 2018-12-05  Segher Boessenkool  <segher@kernel.crashing.org>
1834         * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
1835         aggregate handling on actual AIX, not on somewhat similar systems.
1836         (rs6000_arg_partial_bytes): Ditto.
1838 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
1840         PR rtl-optimization/88317
1841         * lra-constraints.c (split_reg): Don't set up check_only_regs if
1842         we are outside the inheritance pass.
1844 2018-12-04  Jan Hubicka  <hubicka@ucw.cz>
1846         * ipa-prop.c (jump_function_useful_p): New.
1847         (ipa_read_node_info): Do not allocated useless jump functions.
1849 2018-12-04  David Malcolm  <dmalcolm@redhat.com>
1851         PR c/82967
1852         * spellcheck.c (selftest::test_suggestions): Add another assertion.
1854 2018-12-04  Martin Sebor  <msebor@redhat.com>
1856         * doc/extend.texi (__builtin_has_attribute): Fix typo.
1858 2018-12-04  Jakub Jelinek  <jakub@redhat.com>
1860         PR target/88188
1861         * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
1862         failures.
1863         (print_operand): Use REG_P instead of GET_CODE == REG.
1864         <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
1865         <case 't'>: Likewise.  Remove GET_MODE check.
1866         * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
1867         return -1.
1869 2018-12-04  Martin Sebor  <msebor@redhat.com>
1871         * doc/extend.texi (attribute aligned): Expand.
1873 2018-12-04  David Edelsohn  <dje.gcc@gmail.com>
1875         PR target/61976
1876         * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
1877         in FPRs on AIX.
1878         (rs6000_arg_partial_bytes): Same.
1880 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
1882         PR target/88282
1883         * ira-costs.c (exec): Try bigger class to use smaller register
1884         move cost.
1886 2018-12-04  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1888         PR ipa/88297
1889         * ipa-cp.c (create_specialized_node): Track clone counters by
1890         node assembler names.
1891         (ipcp_driver): Change type of clone_num_suffixes key to const char*.
1893 2018-12-04  Richard Biener  <rguenther@suse.de>
1895         * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
1896         for epilogue vectorization.
1898 2018-12-04  Richard Biener  <rguenther@suse.de>
1900         PR tree-optimization/88301
1901         * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
1902         issues in last commit.
1904 2018-12-04  Richard Biener  <rguenther@suse.de>
1906         PR tree-optimization/88315
1907         * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
1908         and fix initialization vector for SAD and DOT_PROD SLP reductions.
1910 2018-12-03  Sandra Loosemore  <sandra@codesourcery.com>
1912         PR c/59039
1913         * doc/extend.texi (Nonlocal gotos): New section.
1915 2018-12-03  Uros Bizjak  <ubizjak@gmail.com>
1917         * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
1918         alternative to (v,vm,c).
1920 2018-12-03  Segher Boessenkool  <segher@kernel.crashing.org>
1922         * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
1923         (powerpc*-*-linux*): Don't handle it here.
1925 2018-12-03  Jeff Law  <law@redhat.com>
1927         * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
1928         warnings quiet.
1930         * config/h8300/h8300.md (low byte masking peepholes): Only accept
1931         nonimmediate_operand for the to-be-masked operand.
1933 2018-12-03  Richard Biener  <rguenther@suse.de>
1935         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1936         allow expected function-pointer cast re-instantiation.
1938 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
1940         * common/config/s390/s390-common.c (s390_option_init_struct):
1941         Use section anchors by default.
1942         * config/s390/s390.c (s390_check_symref_alignment): Handle
1943         anchors.
1944         (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
1945         * output.h (assemble_align): Pass `align' as unsigned int, so
1946         that the value 0x80000000, which corresponds to `aligned(1 <<
1947         28)', would pass the `align > BITS_PER_UNIT' check.
1948         * varasm.c (assemble_align): Likewise.
1950 2018-12-03  Julian Brown  <julian@codesourcery.com>
1952         * tree-pretty-print.c (dump_omp_clause): Make default case
1953         gcc_unreachable.
1955 2018-12-03  Richard Biener  <rguenther@suse.de>
1957         PR tree-optimization/88301
1958         * tree-vrp.c (register_edge_assert_for_2): Handle conversions
1959         that do not change the value by registering the same assert
1960         for the operand.
1962 2018-12-03  Richard Biener  <rguenther@suse.de>
1964         * tree-vect-loop.c (vect_transform_loop): Properly compute
1965         upper bound for the epilogue when doing epilogue vectorization.
1967 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
1969         PR target/80080
1970         * cfgcleanup.c (class pass_postreload_jump): New pass.
1971         (pass_postreload_jump::execute): Likewise.
1972         (make_pass_postreload_jump): Likewise.
1973         * passes.def: Add pass_postreload_jump before
1974         pass_postreload_cse.
1975         * tree-pass.h (make_pass_postreload_jump): New pass.
1977 2018-12-02  Uros Bizjak  <ubizjak@gmail.com>
1979         * config/i386/i386.c (inline_memory_move_cost): Correctly
1980         compute number of TFmode GPR moves.
1982 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
1984         PR target/88278
1985         * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
1987 2018-12-02  Jeff Law  <law@redhat.com>
1989         * config/h8300/h8300.md (call, call_value): Drop mode from
1990         operand containing number of bytes pushed.
1991         (call_insn, call_value_insn): Use mode iterator for number
1992         of bytes pushed operand.  Make pattern name vary appropriately.
1994 2018-12-02  Jan Hubicka  <jh@suse.cz>
1996         * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
1997         trust precise profiles.
1999 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
2001         * tree-nested.c (convert_nonlocal_omp_clauses,
2002         convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
2003         OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
2004         (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
2005         Convert clauses for GIMPLE_OMP_TASKGROUP.
2007         * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
2008         cancellation point with taskgroup clause inside of taskloop.  Consider
2009         a taskloop construct without nogroup clause as implicit taskgroup for
2010         diagnostics if cancel/cancellation point with taskgroup clause is
2011         closely nested inside of taskgroup region.
2013         PR sanitizer/88291
2014         * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
2015         to the start of the function.
2016         (asan_emit_stack_protection): When emitting clearing sequence for
2017         epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
2018         add last_size_aligned which is last_size padded to multiples of
2019         ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
2020         boundaries.
2022 2018-12-01  Jakub Jelinek  <jakub@redhat.com>
2024         PR sanitizer/88289
2025         * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
2026         an off-by-one for BYTES_BIG_ENDIAN.
2028         PR target/54589
2029         * combine.c (find_split_point): For invalid memory address
2030         nonobj + obj + const, if reg + obj + const is valid addressing
2031         mode, split at nonobj.  Use if rather than else if for the
2032         fallback.  Comment fixes.
2034 2018-11-30  Indu Bhagat  <indu.bhagat@oracle.com>
2036         * coverage.c (get_coverage_counts): Use from_function_decl for precise
2037         function location.
2038         * profile-count.c (profile_count::dump): Add handling for precise
2039         profile quality.
2040         * profile.c (compute_branch_probabilities): Rely on exec_counts instead
2041         of profile_info to set profile_status of function.
2042         (branch_prob): Do not set profile_status of function based on
2043         profile_info. Done above based on exec_counts.
2045 2018-11-30  Jakub Jelinek  <jakub@redhat.com>
2047         PR tree-optimization/88274
2048         * tree-ssa-reassoc.c (optimize_range_tests_xor,
2049         optimize_range_tests_diff): If type has smaller precision than
2050         corresponding mode or if it has non-standard min/max, compute
2051         everything in a standard type for the precision.
2053         PR testsuite/85368
2054         * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
2055         * tree-ssa-ifcombine.c (ifcombine_ifandif): If
2056         --param logical-op-non-short-circuit is present, override
2057         LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
2058         * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
2060 2018-11-30  Jeff Law  <law@redhat.com>
2062         * optabs.c (expand_binop): Use "machine_mode" rather than
2063         "enum machine mode" in most recent change.
2065 2018-11-30  Wilco Dijkstra  <wdijkstr@arm.com>
2067         PR middle-end/64242
2068         * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
2069         the frame pointer.
2070         (expand_builtin_nonlocal_goto): Likewise.
2072 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
2074         * diagnostic-core.h (emit_diagnostic): New decl.
2075         * diagnostic.c (emit_diagnostic): New overload, taking a
2076         rich_location *.
2078 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
2080         * pretty-print.c (class selftest::test_pretty_printer): New
2081         subclass of pretty_printer.
2082         (selftest::test_prefixes_and_wrapping): New test.
2083         (selftest::pretty_print_c_tests): Call it.
2085 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2087         Minimize clone counter memory usage in create_virtual_clone.
2088         * cgraph.h (clone_function_name): Add a variant that takes a
2089         tree decl.
2090         * cgraph.h (cgraph_node::create_virtual_clone): Add a new
2091         argument: num_suffix.
2092         * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
2093         num_suffix to clone_function_name.
2094         (clone_function_name): Add a variant that takes a tree decl.
2095         * ipa-cp.c (create_specialized_node): Keep track of clone
2096         counters in clone_num_suffixes hash map.
2097         (ipcp_driver): Free the counter hash map.
2098         * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
2099         clone per function.
2101 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2102         Make function assembly more independent.
2104         This is achieved by having clone_function_name assign unique clone
2105         numbers for each function independently.
2107         * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
2108         hash map.
2109         (clone_function_name_numbered): Use clone_fn_ids.
2111 2018-11-30  Vladimir Makarov  <vmakarov@redhat.com>
2113         PR rtl-optimization/88179
2114         * lra-constraints.c (address_eliminator::address_eliminator):
2115         Don't eleminate regs in illegitimate address.
2117 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
2119         PR preprocessor/88257
2120         * input.c (get_substring_ranges_for_loc): Fix indentation.
2121         Bulletproof against getting a different files back from
2122         linemap_resolve_location and expand_location_to_spelling_point.
2124 2018-11-30  Alexander Monakov  <amonakov@ispras.ru>
2126         PR gcov-profile/88279
2127         * tree-profile.c (init_ic_make_global_vars): Call
2128         decl_default_tls_model for variable and not it's type.
2130 2018-11-30  Sam Tebbs  <sam.tebbs@arm.com>
2132         * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
2134 2018-11-30  Martin Liska  <mliska@suse.cz>
2136         PR sanitizer/81715
2137         * asan.c (asan_shadow_cst): Remove, partially transform
2138         into flush_redzone_payload.
2139         (RZ_BUFFER_SIZE): New.
2140         (struct asan_redzone_buffer): New.
2141         (asan_redzone_buffer::emit_redzone_byte): Likewise.
2142         (asan_redzone_buffer::flush_redzone_payload): Likewise.
2143         (asan_redzone_buffer::flush_if_full): Likewise.
2144         (asan_emit_stack_protection): Use asan_redzone_buffer class
2145         that is responsible for proper aligned stores and flushing
2146         of shadow memory payload.
2147         * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
2148         (asan_var_and_redzone_size): Likewise.
2149         * cfgexpand.c (expand_stack_vars): Use smaller alignment
2150         (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
2151         for automatic variables more compact.
2153 2018-11-30  Alan Modra  <amodra@gmail.com>
2155         * config/rs6000/predicates.md (easy_fp_constant): Avoid long
2156         dependent insn sequences.
2157         * config/rs6000/rs6000.c (num_insns_constant): Support long
2158         double constants.
2159         * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
2160         length attribute.
2162 2018-11-30  Alan Modra  <amodra@gmail.com>
2164         * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
2165         num_insns_constant_wide.  Make static.  Revise comment.
2166         (num_insns_constant_multi): New function.
2167         (num_insns_constant): Formatting.  Correct CONST_WIDE_INT
2168         calculation.  Simplify and extract code common to both
2169         CONST_INT and CONST_DOUBLE.  Add gcc_unreachable for unhandled
2170         const_double modes.
2171         * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
2173 2018-11-30  Richard Biener  <rguenther@suse.de>
2175         PR tree-optimization/88274
2176         * tree-vrp.c (ranges_from_anti_range): Fix handling of
2177         TYPE_MIN/MAX_VALUE.
2179 2018-11-29  Andi Kleen  <ak@linux.intel.com>
2181         * config/i386/i386.c (current_fentry_section): Add.
2182         (x86_function_profiler): Handle fentry section.
2183         (ix86_attribute_table): Add fentry section.
2184         * config/i386/i386.opt: Add -mfentry-section.
2185         * doc/extend.texi: Document fentry_section attribute.
2186         * doc/invoke.texi: Document -mfentry-section.
2188 2018-11-29  Andi Kleen  <ak@linux.intel.com>
2190         * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
2191         (current_fentry_name): Add.
2192         (x86_function_profiler): Handle fentry_name attribute.
2193         (ix86_handle_fentry_name): Add.
2194         (ix86_attribute_table): Add fentry_name.
2195         * config/i386/i386.opt: Add -mfentry-name
2196         * doc/extend.texi: Document fentry_name.
2197         * doc/invoke.texi: Document minstrument-return.
2199 2018-11-29  Andi Kleen  <ak@linux.intel.com>
2201         * config/i386/i386-opts.h (enum instrument_return): Add.
2202         * config/i386/i386.c (output_return_instrumentation): Add.
2203         (ix86_output_function_return): Call output_return_instrumentation.
2204         (ix86_output_call_insn): Call output_return_instrumentation.
2205         * config/i386/i386.opt: Add -minstrument-return=.
2206         * doc/invoke.texi (-minstrument-return): Document.
2208 2018-11-29  Eric Botcazou  <ebotcazou@adacore.com>
2210         PR target/87807
2211         * config/sparc/sparc-modes.def: Minor tweak.
2212         * config/sparc/sparc.c: Minor reordering.
2213         (sparc_pass_by_reference): Move around.
2214         (traverse_record_type): Change offset from HOST_WIDE_INT to int.
2215         (classify_registers): Likewise for bitpos.
2216         (function_arg_slotno): Remove dead test and tweak comments.
2217         <MODE_RANDOM>: Remove useless assertion and test whether the
2218         parameter is named in order to pass it in FP registers.  Return
2219         the regno for floating-point vector types.
2220         (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
2221         (compute_fp_layout): Likewise.
2222         (count_registers): Likewise.
2223         (assign_int_registers): Likewise.
2224         (assign_fp_registers): Likewise.
2225         (assign_registers): Likewise.
2226         (function_arg_record_value): Change size from HOST_WIDE_INT to int
2227         and use CEIL_NWORDS to compute the number of registers.
2228         (function_arg_union_value): Minor tweaks.
2229         (function_arg_vector_value): Add slotno and named parameters, use
2230         CEIL_NWORDS to compute the number of registers.
2231         (sparc_function_arg_1): Rework handling of vector types.  Change
2232         size from HOST_WIDE_INT to int.
2233         (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
2234         with vector types for the 64-bt ABI.
2235         (sparc_function_arg_advance): Likewise.
2236         (sparc_return_in_memory): Add reference to -fpcc-struct-return.
2237         (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
2238         (sparc_function_value_1): Rework handling of vector types.  Change
2239         size from HOST_WIDE_INT to int.
2241 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
2243         PR target/88152
2244         * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
2245         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
2246         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
2247         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
2248         *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
2249         define_insn_and_split patterns.
2251         PR target/54700
2252         * config/i386/sse.md
2253         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
2254         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
2255         *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
2256         define_insn_and_split.
2258 2018-11-29  Peter Bergner  <bergner@linux.ibm.com>
2260         PR target/87496
2261         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2262         -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
2264 2018-11-29  Uros Bizjak  <ubizjak@gmail.com>
2266         * config/i386/i386.c (inline_memory_move_cost):
2267         Check "in" for 2 in MMX_CLASS_P case.
2268         * config/i386/mmx.md (*mov<mode>_internal): Correct
2269         TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
2270         alternatives in preferred_for_speed attribute calculation.
2272 2018-11-29  Martin Sebor  <msebor@redhat.com>
2274         PR c/88172
2275         PR testsuite/88208
2276         * doc/extend.texi (attribute constructor): Clarify.
2278 2018-11-29  Martin Liska  <mliska@suse.cz>
2280         PR middle-end/88246
2281         * builtins.c (expand_movstr): Fix thinko introduced
2282         when switching to the new enum.
2284 2018-11-29  qing zhao  <qing.zhao@oracle.com>
2286         * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
2287         * common.opt: Add -flive-patching flag.
2288         * doc/invoke.texi: Document -flive-patching.
2289         * flag-types.h (enum live_patching_level): New enum.
2290         * ipa-inline.c (can_inline_edge_p): Disable external functions from
2291         inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
2292         * opts.c (control_options_for_live_patching): New function.
2293         (finish_options): Make flag_live_patching incompatible with flag_lto.
2294         Control IPA optimizations based on different levels of
2295         flag_live_patching.
2297 2018-11-29  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2299         * match.pd (sinh (atanh (x))): New simplification rules.
2300         (cosh (atanh (x))): Likewise.
2302 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
2304         PR target/54700
2305         * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
2306         (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
2307         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
2308         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
2309         *<sse4_1_avx2>_pblendvb_lt): New define_insns.
2311         PR target/88152
2312         * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
2313         * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
2314         * match.pd (define_predicates): Add uniform_integer_cst_p.
2315         (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
2316         so that it works also for vector comparisons with uniform constants
2317         with INTEGER_CST element.
2319         PR target/88234
2320         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
2321         vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
2322         in unsigned_type_for instead of vector integral type where overflow
2323         doesn't wrap.
2325 2018-11-29  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2327         There can be at most one .resolver clone per function
2328         * config/rs6000/rs6000.c (make_resolver_func): Generate
2329         resolver symbol with clone_function_name instead of
2330         clone_function_name_numbered.
2332 2018-11-29  Richard Biener  <rguenther@suse.de>
2334         PR tree-optimization/88243
2335         * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
2336         type of all pattern-sequence stmts to vect_internal_def.
2338 2018-11-29  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2340         PR target/88224
2341         * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
2342         (cortex-r7, cortex-r8): Update default and add new configuration.
2343         * doc/invoke.texi (armv7-r): Add two new vfp options.
2344         (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
2345         support this option.
2347 2018-11-29  Alan Modra  <amodra@gmail.com>
2349         * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
2350         double split for targets other than Darwin.
2352 2018-11-29  Alan Modra  <amodra@gmail.com>
2354         * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
2356 2018-11-29  Alan Modra  <amodra@gmail.com>
2358         * config/rs6000/predicates.md (easy_fp_constant): Remove code
2359         dealing with integers in const_double.  Assert on unexpected
2360         modes.  Delete superfluous ABI_V4 test.
2361         * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
2362         Don't call easy_fp_constant for Pmode.
2363         (rs6000_hash_constant): Remove code dealing with integers in
2364         const_double.
2365         (rs6000_legitimate_constant_p): Likewise.
2366         (output_toc): Formatting, use CONST_DOUBLE_P.
2368 2018-11-29  Alan Modra  <amodra@gmail.com>
2370         * config/rs6000/constraints.md (G, H): Comment on purpose of
2371         constraint.  Correct mode comments and "G" mode test.
2372         * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
2373         alternative handling "F".  Add length attr.  Formatting.
2374         (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
2375         alternative, and move "F" to separate alternative.  Correct
2376         insn lengths.
2378 2018-11-29  Alan Modra  <amodra@gmail.com>
2380         * config/rs6000/predicates.md (easy_vector_constant): Don't call
2381         easy_fp_constant.
2383 2018-11-29  Alan Modra  <amodra@gmail.com>
2385         * config.in (HAVE_AS_PLTSEQ): Add.
2386         * config/rs6000/predicates.md (indirect_call_operand): New.
2387         * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
2388         (rs6000_sibcall_sysv): Declare.
2389         * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
2390         CALL_LONG for -fno-plt.
2391         (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
2392         (rs6000_indirect_call_template_1): Emit .reloc directives for
2393         UNSPEC_PLTSEQ calls.
2394         (rs6000_pltseq_template): New function.
2395         (rs6000_longcall_ref): Add arg parameter.  Use PLT16 insns if
2396         relocs supported by assembler.  Move SYMBOL_REF test to callers.
2397         (rs6000_call_aix): Adjust rs6000_longcall_ref call.  Package
2398         insns in UNSPEC_PLTSEQ, preserving original func_desc.
2399         (rs6000_call_sysv): Likewise.
2400         (rs6000_sibcall_sysv): New function.
2401         * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
2402         * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
2403         UNSPEC_PLT16_LO): New.
2404         (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
2405         (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
2406         cookie in constraints.  Test explicitly for flags in length attr.
2407         Handle unspec operand 1.
2408         (call_value_indirect_nonlocal_sysv): Likewise.
2409         (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
2410         (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
2411         (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
2412         (sibcall_indirect_nonlocal_sysv): New pattern.
2413         (sibcall_value_indirect_nonlocal_sysv): Likewise.
2414         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
2415         call alternatives.
2416         * configure.ac: Check for gas plt sequence marker support.
2417         * configure: Regenerate.
2419 2018-11-29  Alan Modra  <amodra@gmail.com>
2421         * config/rs6000/predicates.md (unspec_tls): New.
2422         * config/rs6000/rs6000-protos.h (rs6000_call_template),
2423         (rs6000_sibcall_template): Update prototype.
2424         (rs6000_longcall_ref): Delete.
2425         (rs6000_call_sysv): Declare.
2426         * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
2427         (global_tlsarg): New variable.
2428         (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
2429         handling.
2430         (print_operand): Extract UNSPEC_TLSGD address operand.
2431         (rs6000_call_template, rs6000_sibcall_template): Remove arg
2432         parameter, extract from second call operand instead.
2433         (rs6000_longcall_ref): Make static, localize vars.
2434         (rs6000_call_aix): Rename parameter to reflect new usage.  Take
2435         tlsarg from global_tlsarg.  Don't create unused rtl or nop insns.
2436         (rs6000_sibcall_aix): Rename parameter to reflect new usage.  Take
2437         tlsarg from global_tlsarg.
2438         (rs6000_call_sysv): New function.
2439         * config/rs6000/rs6000.md: Adjust rs6000_call_template and
2440         rs6000_sibcall_template throughout.
2441         (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
2442         (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
2443         (tls_gdld_nomark): New insn.
2444         (tls_gd): Swap operand order.  Simplify mode selection.
2445         (tls_gd_high, tls_gd_low): Swap operand order.
2446         (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
2447         Simplify mode selection.
2448         (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
2449         (call, call_value): Don't assert for second call operand.
2450         Use rs6000_call_sysv.
2452 2018-11-29  Alan Modra  <amodra@gmail.com>
2454         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
2455         (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
2456         (call_value_nonlocal_darwin64): Remove constraints from second call
2457         arg, the rounded_stack_size_rtx arg.
2458         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
2459         (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
2460         (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
2461         (call_value_local64, call_indirect_nonlocal_sysv),
2462         (call_nonlocal_sysv, call_nonlocal_sysv_secure),
2463         (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
2464         (call_value_nonlocal_sysv_secure, call_local_aix),
2465         (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
2466         (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
2467         (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
2468         (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
2469         (sibcall_value_aix): Likewise.
2471 2018-11-29  Alan Modra  <amodra@gmail.com>
2473         * config/rs6000/rs6000.md (TLSmode): Delete mode iterator.  Replace
2474         with P throughout except for call mems which should use SI.
2475         (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
2476         attributes.  Replace with bits, mode and ptrload respectively.
2478 2018-11-29  Alan Modra  <amodra@gmail.com>
2480         * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
2481         (rs6000_indirect_sibcall_template): Declare.
2482         * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
2483         (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
2484         New functions.
2485         * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
2486         (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
2487         (call_indirect_aix, call_value_indirect_aix): Use
2488         rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
2489         call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
2490         handle both speculation and non-speculation cases.
2491         (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
2492         (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
2494 2018-11-29  Alan Modra  <amodra@gmail.com>
2496         * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
2497         (rs6000_sibcall_template): Declare.
2498         (macho_call_template): Rename from output_call.
2499         * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
2500         (rs6000_call_template, rs6000_sibcall_template): Likewise.
2501         (macho_call_template): Rename from output_call.
2502         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
2503         (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
2504         (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
2505         (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
2506         (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
2507         (call_value_nonlocal_aix): Use rs6000_call_template and update
2508         occurrences of output_call to macho_call_template.
2509         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
2510         (sibcall_value_aix): Use rs6000_sibcall_template.
2512 2018-11-28  Aaron Sawdey  <acsawdey@linux.ibm.com>
2514         * config/rs6000/rs6000-string.c (expand_block_clear): Change how
2515         we determine if unaligned vsx is ok.
2517 2018-11-28  Jan Hubicka  <jh@suse.cz>
2519         * predict.c (determine_unlikely_bbs): Forward declare; also determine
2520         edges that are always known to be taken because it is only likely
2521         edge out of given BB.
2522         (tree_estimate_probability): Call before profile guessing to get bit
2523         of extra precision.
2525 2018-11-28  Jan Hubicka  <jh@suse.cz>
2527         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
2528         profile_probability::always better.
2530 2018-11-28  Jan Hubicka  <jh@suse.cz>
2532         * profile-count.h (profile_count::split): Give better result when
2533         splitting profile_probability::always.
2535 2018-11-28  Vladimir Makarov  <vmakarov@redhat.com>
2537         PR target/88207
2538         * ira-costs.c (scan_one_insn): Process subregs when updating costs
2539         for pseudos and allocnos from insn.
2541 2018-11-28  David Edelsohn  <dje.gcc@gmail.com>
2543         * config/rs6000/aix72.h: Update to match aix71.h changes.
2545 2018-11-28  Jeff Law  <law@redhat.com>
2547         * config/h8300/constraints.md: Add "C" constraint for call insns.
2548         * config/h8300/h8300.md (call, call_value): Turn into a define_expand
2549         and define_insn pair.  Move invalid call targets into a register in
2550         the expander and fix constraints in the matching pattern.
2551         * config/h8300/predicates.md (call_expander_operand): Renamed from
2552         call_insn_operand.  Reject things we shouldn't be trying to handle.
2553         (call_insn_operand): New predicate for use by the call/call_value
2554         insns.
2555         (small_call_insn_operand): Update appropriately.
2557 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
2559         * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
2560         calls to strtok with strtok_r.
2562 2018-11-28  Richard Biener  <rguenther@suse.de>
2564         PR tree-optimization/88223
2565         * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
2566         over a stored-same value may-alias store make sure to consider
2567         partial overlaps which are valid when TBAA reasonings do not
2568         apply and byte-granular overlaps are possible at all.
2570 2018-11-28  Richard Biener  <rguenther@suse.de>
2572         PR tree-optimization/88217
2573         * vr-values.c (vr_values::extract_range_from_phi_node): Make
2574         sure to handle results > +INF and < -INF correctly when
2575         trying to drop down to +INF - 1 or -INF + 1.
2577 2018-11-28  Alan Modra  <amodra@gmail.com>
2579         * xcoffout.c (do_block): Signed/unsigned warning fix.
2581 2018-11-28  Richard Biener  <rguenther@suse.de>
2583         PR tree-optimization/79351
2584         * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
2585         empty CONSTRUCTORs ensure the store is at a constant position.
2587 2018-11-28  Richard Biener  <rguenther@suse.de>
2589         PR tree-optimization/88229
2590         * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
2591         is an INTEGER_CST before accessing it so.
2593 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
2595         * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
2597 2018-11-28  Jakub Jelinek  <jakub@redhat.com>
2599         PR target/88189
2600         * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
2601         SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1.  Formatting fixes.
2602         * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
2604 2018-11-27  Martin Liska  <mliska@suse.cz>
2606         * gcov.c (generate_results): Append current_working_directory
2607         only when exists.
2609 2018-11-27  Mike Gulick  <mgulick@mathworks.com>
2611         PR preprocessor/83173
2612         * input.c (dump_location_info): Dump reason and included_from
2613         fields from line_map_ordinary struct.  Fix indentation when
2614         location > 5 digits.
2615         * diagnostic-show-locus.c (num_digits, num_digits): Move to
2616         diagnostic.c to allow it to be utilized by input.c.
2617         * diagnostic.c (num_digits, selftest::test_num_digits): Moved
2618         here.
2619         (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
2620         * diagnostic.h (num_digits): Add extern definition.
2622 2018-11-27  Fredrik Noring  <noring@nocrew.org>
2624         * config/mips/mips.c (mips_reorg_process_insns)
2625         (mips_option_override): Handle `-mfix-r5900'.
2626         * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
2627         `mno-fix-r5900'.
2628         * config/mips/mips.opt (mfix-r5900): New option.
2629         * doc/invoke.texi: Document the `r5900' processor name, and
2630         `-mfix-r5900' and `-mno-fix-r5900' options.
2632 2018-11-27  Jakub Jelinek  <jakub@redhat.com>
2634         PR target/88188
2635         * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
2636         output_operand_lossage instead of gcc_assert.
2637         <case 't'>: Likewise.
2638         <case 'z'>: Likewise.
2639         <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
2641 2018-11-27  Jeff Law  <law@redhat.com>
2643         * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
2644         in call to move_by_pieces.
2646         * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
2647         to move_by_pieces.
2649         * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
2650         call to move_by_pieces.
2652         * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
2653         in call to move_by_pieces.
2655         * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
2656         RETURN_BEGIN in call to move_by_pieces.
2657         (microblaze_expand_block_move): Likewise.
2659 2018-11-27  Tamar Christina  <tamar.christina@arm.com>
2661         * doc/md.texi (xorsign): Document it.
2663 2018-11-27  Martin Liska  <mliska@suse.cz>
2665         * config/gnu-user.h (TARGET_F951_OPTIONS): New.
2666         * gcc.c (find_fortran_preinclude_file): New function
2667         to handle Fortran pre-include.
2669 2018-11-27  Martin Liska  <mliska@suse.cz>
2671         * asan.c (asan_emit_stack_protection): Use new enum values
2672         instead of int constants.
2673         * builtins.c (expand_builtin_memory_copy_args): Replace int
2674         type with memop_ret enum type.
2675         (expand_builtin_mempcpy_args): Likewise.
2676         (expand_builtin_memcpy): Use new enum values
2677         instead of int constants. Likewise.
2678         (expand_builtin_mempcpy): Likewise.
2679         (expand_movstr): Likewise.
2680         (expand_builtin_strcpy_args): Likewise.
2681         (expand_builtin_stpcpy_1): Likewise.
2682         (expand_builtin_strncpy): Likewise.
2683         (expand_builtin_memset_args): Likewise.
2684         * expr.c (move_by_pieces_d::finish_endp): Rename to ...
2685         (move_by_pieces_d::finish_retmode): ... this.
2686         (move_by_pieces): Change last argument type to memop_ret.
2687         (store_by_pieces): Use new enum values
2688         instead of int constants.
2689         (emit_block_move_hints): Likewise.
2690         (emit_push_insn): Likewise.
2691         (store_expr): Likewise.
2692         * expr.h (store_by_pieces): Change int to newly added enum
2693         type.
2694         * rtl.h (enum memop_ret): Define.
2695         (move_by_pieces): Use the enum type.
2697 2018-11-27  Alan Modra  <amodra@gmail.com>
2699         * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
2700         cpu here.
2701         (ASM_CPU_SPEC): Do so here.  Rewrite using if .. else if .. specs
2702         form.  Error on missing -mcpu case.
2703         * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
2704         (asm_names <!_AIX>): Add missing cpus.  Update NULL case.  Apply
2705         PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
2706         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
2707         else if .. specs form.  Error on missing -mcpu case.  Don't output
2708         duplicate -maltivec.  Apply PR63177 fix for -mcpu=powerpc64le.
2710 2018-11-26  David Malcolm  <dmalcolm@redhat.com>
2712         * dump-context.h (dump_context::dump_loc): Convert 1st param from
2713         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
2714         const dump_location_t & to const dump_user_location_t &.
2715         (dump_context::dump_loc_immediate): Convert 2nd param from
2716         const dump_location_t & to const dump_user_location_t &.
2717         (dump_context::dump_gimple_stmt): Convert 1st param from
2718         dump_flags_t to const dump_metadata_t &.
2719         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2720         2nd param from const dump_location_t & to
2721         const dump_user_location_t &.
2722         (dump_context::dump_gimple_expr): Convert 1st param from
2723         dump_flags_t to const dump_metadata_t &.
2724         (dump_context::dump_gimple_expr_loc): Likewise; convert
2725         2nd param from const dump_location_t & to
2726         const dump_user_location_t &.
2727         (dump_context::dump_generic_expr): Convert 1st param from
2728         dump_flags_t to const dump_metadata_t &.
2729         (dump_context::dump_generic_expr_loc): Likewise; convert
2730         2nd param from const dump_location_t & to
2731         const dump_user_location_t &.
2732         (dump_context::dump_printf_va): Convert 1st param from
2733         dump_flags_t to const dump_metadata_t &.
2734         (dump_context::dump_printf_loc_va): Likewise; convert
2735         2nd param from const dump_location_t & to
2736         const dump_user_location_t &.
2737         (dump_context::dump_dec): Convert 1st param from
2738         dump_flags_t to const dump_metadata_t &.
2739         (dump_context::dump_symtab_node): Likewise.
2740         (dump_context::begin_scope): Split out 2nd param into
2741         user and impl locations.
2742         (dump_context::ensure_pending_optinfo): Add metadata param.
2743         (dump_context::begin_next_optinfo): Replace dump_location_t param
2744         with metadata and user location.
2745         * dumpfile.c (dump_context::dump_loc): Convert 1st param from
2746         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
2747         const dump_location_t & to const dump_user_location_t &.
2748         (dump_context::dump_loc_immediate): Convert 2nd param from
2749         const dump_location_t & to const dump_user_location_t &.
2750         (dump_context::dump_gimple_stmt): Convert 1st param from
2751         dump_flags_t to const dump_metadata_t &.
2752         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2753         2nd param from const dump_location_t & to
2754         const dump_user_location_t &.
2755         (dump_context::dump_gimple_expr): Convert 1st param from
2756         dump_flags_t to const dump_metadata_t &.
2757         (dump_context::dump_gimple_expr_loc): Likewise; convert
2758         2nd param from const dump_location_t & to
2759         const dump_user_location_t &.
2760         (dump_context::dump_generic_expr): Convert 1st param from
2761         dump_flags_t to const dump_metadata_t &.
2762         (dump_context::dump_generic_expr_loc): Likewise; convert
2763         2nd param from const dump_location_t & to
2764         const dump_user_location_t &.
2765         (dump_context::dump_printf_va): Convert 1st param from
2766         dump_flags_t to const dump_metadata_t &.
2767         (dump_context::dump_printf_loc_va): Likewise; convert
2768         2nd param from const dump_location_t & to
2769         const dump_user_location_t &.
2770         (dump_context::dump_dec): Convert 1st param from
2771         dump_flags_t to const dump_metadata_t &.
2772         (dump_context::dump_symtab_node): Likewise.
2773         (dump_context::begin_scope): Split out 2nd param into
2774         user and impl locations.
2775         (dump_context::ensure_pending_optinfo): Add metadata param.
2776         (dump_context::begin_next_optinfo): Replace dump_location_t param
2777         with metadata and user location.
2778         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2779         const dump_metadata_t &.
2780         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2781         const dump_location_t & to const dump_user_location_t &.
2782         (dump_gimple_expr): Convert 1st param from dump_flags_t to
2783         const dump_metadata_t &.
2784         (dump_gimple_expr_loc): Likewise; convert 2nd param from
2785         const dump_location_t & to const dump_user_location_t &.
2786         (dump_generic_expr): Convert 1st param from dump_flags_t to
2787         const dump_metadata_t &.
2788         (dump_generic_expr_loc): Likewise; convert 2nd param from
2789         const dump_location_t & to const dump_user_location_t &.
2790         (dump_printf): Convert 1st param from dump_flags_t to
2791         const dump_metadata_t &.
2792         (dump_printf_loc): Likewise; convert 2nd param from
2793         const dump_location_t & to const dump_user_location_t &.
2794         (dump_dec): Convert 1st param from dump_flags_t to
2795         const dump_metadata_t &.
2796         (dump_symtab_node): Likewise.
2797         (dump_begin_scope): Split out 2nd param into user and impl
2798         locations.
2799         (selftest::assert_impl_location_eq): New function.
2800         (ASSERT_IMPL_LOCATION_EQ): New macro.
2801         (selftest::test_impl_location): Update to use
2802         ASSERT_IMPL_LOCATION_EQ.
2803         (selftest::test_capture_of_dump_calls): Convert "loc" to
2804         dump_user_location_t.  Add ASSERT_IMPL_LOCATION_EQ throughout,
2805         verifying line numbers of dump emissions.
2806         * dumpfile.h (class dump_metadata_t): New class.
2807         (dump_printf): Convert 1st param from dump_flags_t to
2808         const dump_metadata_t &.
2809         (dump_printf_loc): Likewise; convert 2nd param from
2810         const dump_location_t & to const dump_user_location_t &.
2811         (dump_generic_expr_loc): Likewise.
2812         (dump_generic_expr): Convert 1st param from dump_flags_t to
2813         const dump_metadata_t &.
2814         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2815         const dump_location_t & to const dump_user_location_t &.
2816         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2817         const dump_metadata_t &.
2818         (dump_gimple_expr_loc): Likewise; convert 2nd param from
2819         const dump_location_t & to const dump_user_location_t &.
2820         (dump_gimple_expr): Convert 1st param from dump_flags_t to
2821         const dump_metadata_t &.
2822         (dump_symtab_node): Likewise.
2823         (dump_dec): Likewise.
2824         (dump_begin_scope): Split out 2nd param into user and impl
2825         locations.
2826         (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
2827         location and impl_location, and capture the impl_location.
2828         (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
2829         * loop-unroll.c (report_unroll): Update for changes to
2830         dump_printf_loc and dump_printf.
2831         * opt-problem.cc (opt_problem::opt_problem): Update for change to
2832         dump_loc.
2833         * optinfo-emit-json.cc
2834         (selftest::test_building_json_from_dump_calls): Convert "loc" from
2835         dump_location_t to dump_user_location_t.
2836         * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
2837         dump_loc_immediate.
2838         * profile.c (compute_branch_probabilities): Update for change to
2839         dump_printf_loc.
2840         * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
2841         * tree-vect-slp.c (vect_print_slp_tree): Update for change to
2842         dump_printf_loc.
2844 2018-11-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2846         * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
2847         and check_effective_target_ptr_eq_long.
2849 2018-11-27  Alan Modra  <amodra@gmail.com>
2851         * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
2852         (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
2853         (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
2854         (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
2855         (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
2856         * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
2857         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2858         (ASM_APP_ON, ASM_APP_OFF): Don't define.
2859         * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2860         * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2861         * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2862         * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
2863         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2864         * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2865         * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
2866         (GNU_USER_TARGET_CRTN): Redefine.
2867         (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
2868         (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
2869         (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
2870         (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
2871         (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
2872         (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
2873         (LINK_EH_SPEC): Don't define.
2875 2018-11-26  Martin Sebor  <msebor@redhat.com>
2877         PR 87756
2878         * expr.c (string_constant): Handle top-level decls of all character
2879         types and subobjects of narrow character type.
2881 2018-11-27  Alan Modra  <amodra@gmail.com>
2883         * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
2884         * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
2885         GNU_USER_TARGET_OS_CPP_BUILTINS.
2886         (RS6000_ABI_NAME): Define.
2887         * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
2888         GNU_USER_TARGET_OS_CPP_BUILTINS.
2889         (MUSL_DYNAMIC_LINKER32): Undef before defining.
2890         (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
2891         (CHOOSE_DYNAMIC_LINKER): Don't define.
2892         (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
2893         * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
2894         (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
2895         not already defined.
2896         (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
2897         TARGET_OS_CPP_BUILTINS.
2899 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2901         * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
2903 2018-11-26  Jakub Jelinek  <jakub@redhat.com>
2905         PR target/88195
2906         * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
2907         and !TARGET_64BIT, return NULL_TREE.
2909         PR c++/86900
2910         * dwarf2out.c (secname_for_decl): For functions with
2911         DECL_SECTION_NAME if in_cold_section_p, try to return
2912         current_function_section's name if it is a named section.
2914 2018-11-26  Maya Rashish  <coypu@sdf.org>
2916         PR target/58397
2917         * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
2918         * config/x-netbsd: New file.
2919         * config/host-netbsd: New file.
2921 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2923         * doc/sourcebuild.texi: Document check_effective_target_size20plus.
2924         Clarify documentation for check_effective_target_size32plus.
2926 2018-11-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2928         * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
2929         and modify expand_strncmp_vec_sequence.
2930         (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
2931         (generate_6432_conversion): New function.
2932         (expand_block_compare): Add support for vsx.
2933         (expand_block_compare_gpr): New function.
2934         * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
2935         default limit to 63 because of more compact vsx code.
2937 2018-11-26  Uros Bizjak  <ubizjak@gmail.com>
2939         PR target/88178
2940         * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
2941         for registers for which no debug information can be generated.
2942         Use INVALID_REGNUM for non-existent registers.
2943         (dbx64_register_map): Ditto.
2944         (svr4_dbx_register_map): Ditto.
2946 2018-11-26  Sandra Loosemore  <sandra@codesourcery.com>
2948         PR c/57182
2949         * doc/invoke.texi (Optimize Options): Document that
2950         -freorder-functions also uses "hot" and "cold" attributes.
2952 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2954         * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
2956 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2958         * config/arm/types.md (mul64): Rename to...
2959         (widen_mul64): ... This.
2960         * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
2961         * config/arm/cortex-a15.md: Likewise.
2962         * config/arm/cortex-a5.md: Likewise.
2963         * config/arm/cortex-a53.md: Likewise.
2964         * config/arm/cortex-a57.md: Likewise.
2965         * config/arm/cortex-a7.md: Likewise.
2966         * config/arm/cortex-m4.md: Likewise.
2967         * config/arm/exynos-m1.md: Likewise.
2968         * config/arm/marvell-pj4.md: Likewise.
2969         * config/arm/xgene1.md: Likewise.
2971 2018-11-26  Richard Biener  <rguenther@suse.de>
2973         PR tree-optimization/88182
2974         * tree-vect-loop.c (vectorizable_reduction): Pick up single
2975         correct reduc_def_info.
2976         * tree-vect-slp.c (vect_analyze_slp_instance): Set
2977         STMT_VINFO_REDUC_DEF of the first stmt.
2979 2018-11-26  Andreas Krebbel  <krebbel@linux.ibm.com>
2981         * doc/invoke.texi: Document z14/arch12 -march option.
2983 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
2985         PR c/57166
2986         * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
2987         (Warning Options): Likewise.
2989 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
2991         PR web/79738
2992         * doc/extend.texi (Common Function Attributes): Clarify that
2993         functions with "const" attribute can read const global variables.
2995 2018-11-26  Alan Modra  <amodra@gmail.com>
2997         * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
2999 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
3001         PR other/54265
3002         * doc/extend.texi (Common Variable Attributes): Use preferred
3003         placement of type attributes in examples, plus whitespace fixes.
3004         (Type Attributes): Clarify why placement of attributes
3005         immediately after struct/union/enum keyword is preferred.
3006         (Common Type Attributes): Use preferred placement of type
3007         attributes in examples, plus more whitespace fixes.
3009 2018-11-25  Paul Koning  <ni1d@arrl.net>
3011         * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
3013 2018-11-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3015         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
3016         attributes): Document d_runtime.
3018 2018-11-25  Vladimir Makarov  <vmakarov@redhat.com>
3020         PR bootstrap/88157
3021         * ira-costs.c (record_operand_costs): Initiate register move cost
3022         for mode.
3024 2018-11-23  Jeff Law  <law@redhat.com>
3026         PR rtl-optimization/87468
3027         * tree-ssa-threadupdate.c (create_block_for_threading): Clear
3028         EDGE_IGNORE on all outgoing edges of the duplicate block.
3030 2018-11-23  Vladimir Makarov  <vmakarov@redhat.com>
3032         PR bootstrap/88157
3033         * ira-costs.c (record_operand_costs): Use bigger hard reg class if
3034         its mode does not fit to the original class.
3036 2018-11-23  Martin Sebor  <msebor@redhat.com>
3038         PR tree-optimization/87756
3039         * expr.c (string_constant): Synthesize a string literal from
3040         the address of a constant character.
3041         * tree.c (build_string_literal): Add an argument.
3042         * tree.h (build_string_literal): Same.
3044 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
3046         * config/aarch64/aarch64-cores.def: Define emag.
3047         * config/aarch64/aarch64-tune.md: Regenerated with emag.
3048         * config/aarch64/aarch64.c (emag_tunings): New struct.
3049         * doc/invoke.texi: Document mtune value.
3051 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
3052             Philipp Tomsich  <philipp.tomsich@theobroma-systems.com>
3054         * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
3055         for GCC 9.
3056         * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
3057         prefetch tunings.
3058         * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
3059         costs.
3060         * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
3061         for XGene1.
3063 2018-11-23  Richard Biener  <rguenther@suse.de>
3065         PR tree-optimization/88149
3066         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
3067         the case where there are two different def types for the
3068         same operand at different operand position in the same stmt.
3070 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
3072         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
3073         gen_atomic_compare_swap_1.
3074         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
3075         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
3076         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
3077         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
3078         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
3079         checks.
3080         * config/arm/arm.md (neon_vext<mode>),
3081         (neon_vrev64<mode>, neon_vrev32<mode>),
3082         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
3083         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
3084         before the pattern name.
3085         * config/arm/sync.md:
3086         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
3087         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
3089 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
3091         PR tree-optimization/86614
3092         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
3093         if TREE_NO_WARNING is set on ref.ref.
3095 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
3096             Alan Coopersmith  <alan.coopersmith@oracle.com>
3098         PR c/53608
3099         * doc/extend.texi (Designated Inits): Clarify handling of multiple
3100         initializers for unions.
3102 2018-11-22  Jan Hubicka  <jh@suse.cz>
3104         PR lto/88142
3105         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
3106         parameters; do not warn here.
3107         (odr_subtypes_equivalent_p): Likewise.
3108         (warn_odr): Fix typo.
3109         (warn_types_mismatch): Do not output confused warnings on integer types.
3110         (odr_types_equivalent_p): Update.
3112 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
3114         * config/i386/i386.c (ix86_option_override_internal,
3115         ix86_can_inline_p, classify_argument, construct_container,
3116         ix86_expand_prologue, ix86_expand_split_stack_prologue,
3117         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
3118         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
3119         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
3120         Formatting fixes.
3122         * config/i386/i386.c (ix86_option_override_internal): For
3123         stack_protector_guard related options, use opts_set->x_ instead
3124         of global_options_set. and prefix options with opts->x_ .  Move
3125         defaults for offset and reg into else block.
3127 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
3129         * config/i386/i386.c (ix86_check_avx_upper_register):
3130         Return true for all SSE registers with mode bitsize > 128.
3132 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
3134         PR rtl-optimization/87718
3135         * ira-costs.c: Remove trailing white-spaces.
3136         (record_operand_costs): Add a special treatment for moves
3137         involving a hard register.
3139 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
3141         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
3142         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
3144 2018-11-22  Martin Liska  <mliska@suse.cz>
3146         * common/config/i386/i386-common.c (processor_names): Add
3147         static assert and add missing "znver2".
3148         (ix86_get_valid_option_values): Add checking assert for null
3149         values and add "native" value if feasible.
3150         * config/i386/i386.h: Do not declare size of processor_names.
3151         * common/config/i386/i386-common.c:
3152         * config/i386/i386.c: Add static assert for size
3153         of processor_cost_table.
3155 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
3157         * target-insns.def (stack_protect_combined_set): Define new standard
3158         pattern name.
3159         (stack_protect_combined_test): Likewise.
3160         * cfgexpand.c (stack_protect_prologue): Try new
3161         stack_protect_combined_set pattern first.
3162         * function.c (stack_protect_epilogue): Try new
3163         stack_protect_combined_test pattern first.
3164         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
3165         parameters to control which register to use as PIC register and force
3166         reloading PIC register respectively.  Insert in the stream of insns if
3167         possible.
3168         (legitimize_pic_address): Expose above new parameters in prototype and
3169         adapt recursive calls accordingly.  Use pic_reg if non null instead of
3170         cached one.
3171         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
3172         (arm_legitimize_address): Adapt to new legitimize_pic_address
3173         prototype.
3174         (thumb_legitimize_address): Likewise.
3175         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
3176         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
3177         (thumb1_expand_prologue): Likewise.
3178         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
3179         change.
3180         (arm_load_pic_register): Likewise.
3181         * config/arm/predicated.md (guard_addr_operand): New predicate.
3182         (guard_operand): New predicate.
3183         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
3184         prototype change.
3185         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
3186         prototype change.
3187         (stack_protect_combined_set): New expander..
3188         (stack_protect_combined_set_insn): New insn_and_split pattern.
3189         (stack_protect_set_insn): New insn pattern.
3190         (stack_protect_combined_test): New expander.
3191         (stack_protect_combined_test_insn): New insn_and_split pattern.
3192         (arm_stack_protect_test_insn): New insn pattern.
3193         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
3194         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
3195         (UNSPEC_SP_TEST): Likewise.
3196         * doc/md.texi (stack_protect_combined_set): Document new standard
3197         pattern name.
3198         (stack_protect_set): Clarify that the operand for guard's address is
3199         legal.
3200         (stack_protect_combined_test): Document new standard pattern name.
3201         (stack_protect_test): Clarify that the operand for guard's address is
3202         legal.
3204 2018-11-22  Richard Biener  <rguenther@suse.de>
3206         PR tree-optimization/88148
3207         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
3208         valueization callback parameter and handle it.
3209         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
3210         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
3212 2018-11-22  Richard Biener  <rguenther@suse.de>
3214         PR lto/87229
3215         PR lto/88112
3216         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
3217         which can appear in size expressions.
3218         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
3219         CALL_EXPR_BY_DESCRIPTOR.
3220         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
3221         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
3222         CALL_EXPR_BY_DESCRIPTOR.
3223         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
3225         Revert
3226         PR lto/87229
3227         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
3228         sizepos values.
3230 2018-11-22  Richard Biener  <rguenther@suse.de>
3232         PR tree-optimization/88069
3233         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
3234         apply to default defs.
3236 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
3238         PR target/85644
3239         PR target/86832
3240         * config/i386/i386.c (ix86_option_override_internal): Default
3241         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
3242         is defined.
3243         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
3244         stack_protect_test, stack_protect_test_<mode>): Use empty condition
3245         instead of TARGET_SSP_TLS_GUARD.
3247 2018-11-22  Martin Liska  <mliska@suse.cz>
3249         PR sanitizer/88017
3250         * config/i386/i386.c (ix86_option_override_internal): Error
3251         about usage -mabi=ms and -fsanitize=thread.
3253 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
3255         Revert the revert:
3256         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
3258         Revert:
3259         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
3261         * lra-spills.c (lra_final_code_change): Remove useless move insns.
3263 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
3265         PR middle-end/88129
3266         * function.c (expand_function_end): Do not emit extra blockage insn.
3268 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
3270         PR target/85667
3271         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
3272         of FIRST_SSE_REG for 4 or 8 byte modes.
3274 2018-11-21  Jan Hubicka  <jh@suse.cz>
3276         PR lto/87957
3277         * tree.c (fld_decl_context): Break out from ...
3278         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
3279         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
3280         (fld_incomplete_type_of): Build copy of TYP_DECL.
3281         * ipa-devirt.c (free_enum_values): Rename to ...
3282         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
3283         and TREE_TYPEs of TYPE_DECLs.
3284         (get_odr_type): Initialize odr_vtable_hash if needed.
3286 2018-11-21  Alexandre Oliva  <oliva@adacore.com>
3288         * final.c (compute_discriminator): Declare.  Renamed from...
3289         (maybe_set_discriminator): ... this.  Set and return a local.
3290         (override_discriminator): New.
3291         (final_scan_insn_1): Set it.
3292         (notice_source_line): Adjust.  Always set discriminator.
3294 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
3296         PR target/87839
3297         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
3298         rIJ constraint for aarch64_plus_operand rather than rn.
3300 2018-11-21  Renlin Li  <renlin.li@arm.com>
3302         PR middle-end/84877
3303         * explow.h (get_dynamic_stack_size): Declare it as external.
3304         * explow.c (record_new_stack_level): Remove function static attribute.
3305         * function.c (assign_stack_local_1): Dynamically align the stack slot
3306         addr for parameter copy on the stack.
3308 2018-11-21  Richard Biener  <rguenther@suse.de>
3310         PR bootstrap/88133
3311         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
3312         * Makefile.in (bitmap.o-warn): Remove again.
3314 2018-11-20  Jeff Law  <law@redhat.com>
3316         PR tree-optimization/88069
3317         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
3318         degenerate virtual PHIs.
3320 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
3322         PR target/87317
3323         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
3324         nonimmediate_operand with register_operand.
3325         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
3326         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
3327         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
3328         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
3329         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
3330         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
3331         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
3332         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
3333         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
3334         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
3335         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
3336         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
3337         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
3338         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
3339         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
3340         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
3341         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
3342         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
3343         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
3344         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
3345         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
3346         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
3347         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
3348         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
3349         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
3350         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
3351         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
3353 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
3355         * read-rtl.c (apply_subst_iterator): Handle define_split and
3356         define_insn_and_split.
3358 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
3360         PR rtl-optimization/87817
3361         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
3362         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
3363         in the pattern to avoid triggering UB when operands[2] is zero.
3364         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
3365         (*tbm_bextri_<mode>): ... this.
3367 2018-11-21  Tom de Vries  <tdevries@suse.de>
3369         PR driver/79855
3370         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
3371         period.
3373 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
3375         * configure.ac: Add check for Binutils to determine whether vector
3376         load/store alignments hints are being supported.
3377         * config.in: Regenerate.
3378         * configure: Regenerate.
3379         * config/s390/s390.c (print_operand): Support new output
3380         modifier A.
3381         * config/s390/s390.md ("movti"): Append alignment hint output
3382         using the new output modifier 'A'.
3383         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
3384         ("*vec_ti_to_v1ti"): Likewise.
3386 2018-11-20  Martin Sebor  <msebor@redhat.com>
3388         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
3390 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
3392         PR lto/84044
3393         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
3394         compare ENUM values.
3396 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
3398         PR tree-optimization/87895
3399         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
3400         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
3401         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
3402         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
3403         has any preds and always use EDGE_PRED (, 0).
3405 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
3407         PR target/88070
3408         * mode-switching.c (create_pre_exit): After reload, always split the
3409         fallthrough edge to the exit block.
3411 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
3413         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
3414         * tree.c (build_array_type_1): Forward declare.
3415         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
3416         (fld_type_variant): Likewise.
3417         (fld_simplified_types): New hash.
3418         (fld_process_array_type): New function.
3419         (fld_incomplete_type_of): Handle array and enumeration types.
3420         (fld_simplified_type): Handle simplification of arrays.
3421         (free_lang_data): Allocate and free simplified types hash.
3423 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
3425         PR lto/87957
3426         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
3427         when sybtype already violates ODR.
3428         (get_odr_type): Do not ICE when insert is false and type duplicate
3429         is not registered yet.
3430         (register_odr_type): Be sure to register subtypes first.
3432 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
3434         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
3435         DImode parts of the target operand.
3437 2018-11-20  Nathan Sidwell  <nathan@acm.org>
3439         PR 87926
3440         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
3442 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
3444         PR lto/87997
3445         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
3446         type checking to...
3447         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
3449 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
3451         PR ipa/87706
3452         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
3453         * ipa.c (possible_inline_candidate_p): Break out from ..
3454         (process_references): ... here ; drop before_inlining_p;
3455         cleanup handling of alises.
3456         (walk_polymorphic_call_targets): Likewise.
3457         (symbol_table::remove_unreachable_nodes): Likewise.
3458         * passes.c (pass_data_ipa_remove_symbols): New structure.
3459         (pass_ipa_remove_symbols): New pass.
3460         (make_pass_ipa_remove_symbols): New function.
3461         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
3462         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
3464 2018-11-20  Richard Biener  <rguenther@suse.de>
3466         * tree-vect-stmts.c (vectorizable_condition): Do not get
3467         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
3468         pointless vect_is_simple_use calls.
3470 2018-11-20  Richard Biener  <rguenther@suse.de>
3472         PR tree-optimization/88074
3473         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
3474         backedges.
3476 2018-11-20  Richard Biener  <rguenther@suse.de>
3478         PR tree-optimization/88069
3479         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
3480         virtual PHI arguments.
3482 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
3484         PR target/88083
3485         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
3486         involved.
3487         * rtl.h (contains_constant_pool_address_p): New function.
3488         * rtlanal.c (contains_constant_pool_address_p): Likewise.
3490 2018-11-20  Richard Biener  <rguenther@suse.de>
3492         PR middle-end/83215
3493         * alias.c (component_uses_parent_alias_set_from): Remove
3494         alias-set zero and TYPE_TYPELESS_STORAGE case both already
3495         handled in other ways.
3497 2018-11-20  Richard Biener  <rguenther@suse.de>
3499         PR tree-optimization/88087
3500         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
3501         call fntype.
3502         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
3503         call fntype.
3505 2018-11-20  Richard Biener  <rguenther@suse.de>
3507         PR middle-end/88089
3508         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
3510 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
3512         PR lto/87957
3513         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
3515 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
3517         PR driver/50250
3518         * doc/invoke.texi (Link Options): Mention shared libraries
3519         in documentation for the -l option.  Simplify discussion and
3520         point to the system linker documentation for details.
3522 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
3524         PR rtl-optimization/88033
3525         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
3526         to itself.  Use HARD_REGISTER_NUM_P.
3528 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3530         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
3531         of the loop if vf is non-constant.
3533 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
3535         PR tree-optimization/87025
3536         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
3537         immediately after creating the scope optinfo.
3538         (selftest::test_pr87025): New function.
3539         (selftest::dumpfile_c_tests): Call it.
3540         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
3541         that we're not popping the top-level records array.
3542         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
3543         not changing the kind of a "scope" optinfo.
3545 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
3547         PR tree-optimization/87025
3548         * dump-context.h: Include "optinfo.h".
3549         (class optrecord_json_writer): New forward decl.
3550         (dump_context::forcibly_enable_optinfo_p): Delete.
3551         (dump_context::optinfo_enabled_p): New member function.
3552         (dump_context::optimization_records_enabled_p): New member
3553         function.
3554         (dump_context::set_json_writer): New member function.
3555         (dump_context::emit_optinfo): New member function.
3556         (dump_context::m_forcibly_enable_optinfo): Delete.
3557         (dump_context::m_json_writer): New member data.
3558         * dumpfile.c (dump_context::set_json_writer): New member function.
3559         (dump_context::finish_any_json_writer): New member function.
3560         (dump_context::end_scope): Replace call to
3561         optimization_records_maybe_pop_dump_scope with call to
3562         m_json_writer->pop_scope.
3563         (dump_context::optinfo_enabled_p): New member function.
3564         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
3565         to dump_context::emit_optinfo.
3566         (dump_context::emit_optinfo): New member function.
3567         (temp_dump_context::temp_dump_context): Replace
3568         m_forcibly_enable_optinfo with call to set_json_writer.
3569         (temp_dump_context::~temp_dump_context): Clean up any json writer.
3570         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
3571         optinfo-emit-json.h
3572         (the_json_writer): Delete.
3573         (optimization_records_start): Delete.
3574         (optimization_records_finish): Delete.
3575         (optimization_records_enabled_p): Delete, in favor of
3576         dump_context::optimization_records_enabled_p.
3577         (optimization_records_maybe_record_optinfo): Delete.
3578         (optimization_records_maybe_pop_dump_scope): Delete.
3579         * optinfo-emit-json.h: Include "json.h".  Delete forward
3580         decl of opt_pass.
3581         (optimization_records_start): Delete.
3582         (optimization_records_finish): Delete.
3583         (optimization_records_enabled_p): Delete.
3584         (optimization_records_maybe_record_optinfo): Delete.
3585         (optimization_records_maybe_pop_dump_scope): Delete.
3586         (class optrecord_json_writer): Move here from
3587         optinfo-emit-json.cc.
3588         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
3589         to optinfo::emit with call to dump_context::emit_optinfo.
3590         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3591         (optinfo_enabled_p): Delete, in favor of
3592         dump_context::optinfo_enabled_p.
3593         (optinfo_wants_inlining_info_p): Update for conversion o
3594         optimization_records_enabled_p to a member function of
3595         dump_context.
3596         * optinfo.h (optinfo_enabled_p): Delete, in favor of
3597         dump_context::optinfo_enabled_p.
3598         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3599         * toplev.c: Include "dump-context.h".
3600         (compile_file): Replace call to optimization_records_finish with
3601         dump_context::finish_any_json_writer.
3602         (do_compile): Replace call to optimization_records_start with
3603         conditionally creating a optrecord_json_writer for the
3604         dump_context.
3606 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
3608         PR tree-optimization/88071
3609         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
3610         of false as last argument to gsi_remove.
3611         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
3612         false as last argument to gsi_replace.
3614         PR debug/87039
3615         * omp-expand.c: Don't include debug.h.
3616         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
3617         from innermost outer parallel, task, teams or target that has a
3618         child_fn set, or, if there is no such outer region, use
3619         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
3620         whether a suitable BLOCK is found or not.
3621         (expand_parallel_call, expand_teams_call): Don't call
3622         adjust_context_and_scope here.
3623         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
3624         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
3625         adjust_context_and_scope.
3626         * dwarf2out.c (dwarf2out_early_global_decl): For
3627         decl_function_context recurse instead of calling dwarf2out_decl.
3629 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
3631         PR rtl-optimization/85925
3632         * rtl.h (word_register_operation_p): New predicate.
3633         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
3634         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
3635         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
3636         (num_sign_bit_copies1): Likewise.
3638 2018-11-19  Richard Biener  <rguenther@suse.de>
3640         PR lto/87229
3641         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
3642         sizepos values.
3644 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
3646         * compare-elim.c (struct comparison): Add not_in_a field.
3647         (is_not): New static function.
3648         (strip_not): Likewise.
3649         (conforming_compare): Handle a NOT in the first operand.
3650         (can_eliminate_compare): Likewise.
3651         (find_comparison_dom_walker::before_dom_children): Likewise.
3652         (try_eliminate_compare): Likewise.
3653         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
3654         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
3656 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
3658         * doc/extend.texi (Common Type Attributes): Fix typos.
3660 2018-11-19  Martin Liska  <mliska@suse.cz>
3662         PR gcov-profile/88045
3663         * coverage.c (coverage_begin_function): Add assert.
3665 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
3667         PR other/40498
3668         * doc/extend.texi (Common Function Attributes): Document that
3669         no_instrument_function applies to -p and -pg, too.
3670         * doc/invoke.texi (Instrumentation Options): Add cross-references
3671         to docs for -p, -pg, and -finstrument-functions.
3673 2018-11-18  Alan Modra  <amodra@gmail.com>
3675         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
3676         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
3677         (GNU_USER_TARGET_CRTN): Define.
3678         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
3679         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
3681 2018-11-18  Alan Modra  <amodra@gmail.com>
3683         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
3684         !HAVE_LD_PIE variant.
3685         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
3687 2018-11-17  Nick Clifton  <nickc@redhat.com>
3688             Sandra Loosemore  <sandra@codesourcery.com>
3690         PR driver/31357
3691         * doc/invoke.texi (Overall Options): Document that --help and
3692         --help=<value> options cannot be combined.
3694 2018-11-17  Jim Wilson  <jimw@sifive.com>
3696         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
3697         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
3698         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
3699         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
3700         to step2 if frame_pointer_needed and step1 is 0.
3702 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
3704         PR c++/4225
3705         * doc/cpp.texi (System Headers): Add note about implicit
3706         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
3708 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
3710         PR c++/4025
3711         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
3712         -fno-implicit-templates.
3714 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
3716         PR tree-optimization/87546
3717         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
3718         min_precision variable, initially set it to orig_precision, only does
3719         something if op_type's precision is <= min_precision and update
3720         min_precision whenever calling set_op.
3722 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
3724         PR ipa/87957
3725         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
3727 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
3729         * doc/invoke.texi (Option Summary): Fix whitespace and line
3730         breaks in @gccoptlist environments.
3731         (Warning Options): Likewise.
3732         (Optimize Options): Likewise.
3733         (PowerPC SPE Options): Likewise.
3734         (RS/6000 and PowerPC Options): Likewise.
3736 2018-11-16  Jeff Law  <law@redhat.com>
3738         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
3739         gcc_assert.
3741 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
3743         PR middle-end/23197
3744         * doc/invoke.texi (Optimize Options): Update options enabled by
3745         fprofile-generate, -fprofile-use, and -fauto-profile.
3747 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3749         PR target/87927
3750         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
3751         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
3752         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
3753         the byte_op hook.
3754         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
3755         * targhooks.c (default_print_patchable_function_entry): Assert
3756         asm_int_op does not return a NULL string.
3757         * varasm.c (integer_asm_op): Return the op for a partial int type
3758         when the requested size does not correspond to an integer type.
3759         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
3760         * doc/tm.texi: Regenerate.
3762 2018-11-12  Jason Merrill  <jason@redhat.com>
3764         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
3766 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
3768         * config/rs6000/constraints.md (wF constraint): Remove power9
3769         fusion documentation.  Just document wF's use for power8 fusion.
3770         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
3771         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
3772         fusion support to require the upper 12 bits to be all 0's or all
3773         1's.
3774         (fusion_gpr_mem_load): Add comment.
3775         (fusion_addis_mem_combo_load): Remove power9 fusion support.
3776         (fusion_addis_mem_combo_store): Delete.
3777         (fusion_offsettable_mem_operand): Delete.
3778         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
3779         power8 fusion here.
3780         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
3781         (POWERPC_MASKS): Delete power9 fusion.
3782         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
3783         (fusion_p9_p): Delete.
3784         (expand_fusion_p9_load): Delete.
3785         (expand_fusion_p9_store): Delete.
3786         (emit_fusion_p9_load): Delete.
3787         (emit_fusion_p9_store): Delete.
3788         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
3789         fusion support.
3790         (rs6000_option_override_internal): Set power8 fusion based on
3791         whether we are tuning for power8.  Delete power9 fusion support.
3792         (rs6000_opt_masks): Delete -mpower9-fusion switch.
3793         (emit_fusion_load): Rename emit_fusion_load_store to
3794         emit_fusion_load, and drop fusion store support.  Update callers.
3795         (emit_fusion_load_store): Likewise.
3796         (emit_fusion_gpr_load): Likewise.
3797         (fusion_p9_p): Delete.
3798         (expand_fusion_p9_load): Delete.
3799         (expand_fusion_p9_store): Delete.
3800         (emit_fusion_p9_load): Delete.
3801         (emit_fusion_p9_store): Delete.
3802         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
3803         (GPR_FUSION): Delete.
3804         (FPR_FUSION): Delete.
3805         (power9 fusion peephole2s): Delete.
3806         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
3807         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
3808         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
3809         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
3810         (fusion_p9_<mode>_constant): Delete.
3811         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
3812         power9 fusion switch.
3813         * doc/md.texi (PowerPC constraints): Update wF constraint
3814         documentation for power8 fusion only.
3816 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
3818         PR rtl-optimization/87475
3819         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
3820         CROSSING_JUMP_P insns.
3821         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
3823 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
3825         PR target/88051
3826         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
3827         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
3828         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
3830 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
3832         PR middle-end/88032
3833         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
3834         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
3836 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
3838         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
3839         location_t parameter.
3840         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3841         (switch_decision_tree::do_jump_if_equal): Likewise.
3842         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
3843         of switch statement to emit_case_nodes.
3844         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
3845         set it on the newly built GIMPLE comparison statement.
3846         (switch_decision_tree::do_jump_if_equal): Likewise.
3847         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
3848         in calls to do_jump_if_equal as well as recursive calls.
3850 2018-11-16  Andi Kleen  <ak@linux.intel.com>
3852         * config/i386/i386.md: Allow memory operands to ptwrite.
3854 2018-11-16  Michael Matz  <matz@suse.de>
3856         * system.h (PRsa): New macro.
3857         (SIZE_AMOUNT): Cast number to uint64_t.
3858         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
3859         (pool_usage::dump_footer): Likewise and also use PRIu64.
3860         * bitmap.h (bitmap_usage::dump): Likewise.
3861         * ggc-common.c (ggc_usage::dump): Likewise.
3862         * ggc-page.c (ggc_print_statistics): Likewise.
3863         * input.c (dump_line_table_statistics): Likewise.
3864         * mem-stats.h (mem_usage::dump): Likewise.
3865         (mem_usage::dump_footer): Likewise.
3866         * rtl.c (dump_rtx_statistics): Likewise.
3867         * tree-cfg.c (dump_cfg_stats): Likewise.
3868         * tree-dfa.c (dump_dfa_stats): Likewise.
3869         * tree-phinodes.c (phinodes_print_statistics): Likewise.
3870         * tree-ssanames (ssanames_print_statistics): Likewise.
3871         * vec.c (vec_usage::dump): Likewise.
3872         (vec_usage::dump_footer): Likewise.
3874 2018-11-16  Richard Biener  <rguenther@suse.de>
3876         PR tree-optimization/88011
3877         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
3878         replacing set_value_range_to_undefined and
3879         set_value_range_to_varying with method calls.
3881 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
3883         * config/s390/s390.md
3884         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
3885         assembler.
3886         (*r<noxa>sbg_<mode>_sll): Likewise.
3887         (*r<noxa>sbg_<mode>_srl): Likewise.
3888         (*r<noxa>sbg_sidi_srl): New pattern.
3890 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
3892         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
3893         Default to TARGET_VXWORKS_RTP.
3894         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3895         of TARGET_VXWORKS_RTP.
3896         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3897         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
3899 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
3901         PR c++/25759
3902         * doc/extend.texi (Common Type Attributes): Make it explicit
3903         that attribute "packed" can apply to C++ classes.
3905 2018-11-15  Martin Sebor  <msebor@redhat.com>
3907         PR c++/87541
3908         PR c++/87542
3909         * tree.c (type_argument_type): New function.
3910         * tree.h (type_argument_type): Declare it.
3911         * gcc/doc/extend.texi (alloc_align): Update and clarify.
3912         (alloc_size, nonnull, sentinel): Same.
3914 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
3915             Kwok Cheung Yeung  <kcy@codesourcery.com>
3917         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
3918         int_mode_for_size fails.
3919         (vectorizable_load): Likewise.
3921 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3923         * doc/ux.texi (Group logically-related diagnostics): Move
3924         discussion of auto_diagnostic_group into this new subsection.
3925         Give an example of where this grouping is used.
3927 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3929         PR other/19165
3930         * Makefile.in (OBJS): Move json.o to...
3931         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
3932         * common.opt (fdiagnostics-format=): New option.
3933         (diagnostics_output_format): New enum.
3934         * diagnostic-format-json.cc: New file.
3935         * diagnostic.c (default_diagnostic_final_cb): New function, taken
3936         from start of diagnostic_finish.
3937         (diagnostic_initialize): Initialize final_cb to
3938         default_diagnostic_final_cb.
3939         (diagnostic_finish): Move "being treated as errors" messages to
3940         default_diagnostic_final_cb.  Call any final_cb.
3941         (default_diagnostic_finalizer): Add diagnostic_t param.
3942         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
3943         diagnostic_finalizer callback.
3944         * diagnostic.h (enum diagnostics_output_format): New enum.
3945         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
3946         (struct diagnostic_context): Add "final_cb".
3947         (default_diagnostic_finalizer): Add diagnostic_t param.
3948         (diagnostic_output_format_init): New decl.
3949         * doc/invoke.texi (-fdiagnostics-format): New option.
3950         * dwarf2out.c (gen_producer_string): Ignore
3951         OPT_fdiagnostics_format_.
3952         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
3953         * lto-wrapper.c (append_diag_options): Ignore it.
3954         * opts.c (common_handle_option): Handle it.
3956 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3958         PR tree-optimization/88015
3959         * graphite-isl-ast-to-gimple.c
3960         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
3961         for dump_enabled_p.
3962         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
3964 2018-11-15  Richard Biener  <rguenther@suse.de>
3966         PR middle-end/88029
3967         * gimple.c (gimple_call_flags): Union flags from decl, type
3968         and call fntype.
3969         * trans-mem.c (is_tm_pure_call): Simplify.
3971 2018-11-15  Richard Biener  <rguenther@suse.de>
3973         PR tree-optimization/88031
3974         * tree-vect-loop.c (vectorizable_reduction): Move check
3975         for multiple types earlier so we get the expected dump.
3976         Simplify calls to vectorizable_condition.
3977         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
3978         * tree-vect-stmts.c (vectorizable_condition): Instead of
3979         reduc_def and reduc_index take just a flag.  Simplify
3980         code-generation now that we can rely on the defs being set up.
3981         (vectorizable_comparison): Remove unused argument.
3983 2018-11-15  Richard Biener  <rguenther@suse.de>
3985         PR tree-optimization/88030
3986         * tree-complex.c (need_eh_cleanup): New global.
3987         (update_complex_assignment): Mark blocks that need EH update.
3988         (expand_complex_comparison): Likewise.
3989         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
3990         perform EH cleanup and schedule CFG cleanup if that did anything.
3992 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
3994         PR rtl-optimization/88018
3995         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
3996         on the presence of fallthru edge, rather than if it is a USE or not.
3998 2018-11-15  Richard Biener  <rguenther@suse.de>
4000         PR middle-end/87917
4001         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
4002         analyze_subscript_affine_affine properly.
4004 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4006         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
4007         (GLIBC_DYNAMIC_LINKER): Define.
4008         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
4010 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4012         PR tree-optimization/84648
4013         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
4014         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
4015         by calling adjust_cond_for_loop_until_wrap.
4017 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
4019         PR other/56334
4020         * doc/extend.texi (Common Function Attributes): Clarify linker
4021         restrictions on "aligned" attribute.
4022         (Common Variable Attributes): Likewise.  Mention that linker
4023         restrictions don't apply to stack-allocated variables.
4025 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
4027         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
4028         as an lvalue.
4030 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
4032         PR lto/55102
4033         PR lto/56700
4034         * doc/invoke.texi (Optimize Options): Remove bad example about
4035         interaction between -flto and -O.  Replace it with a note that
4036         you need to compile with -O and not just link.  Copy-edit -flto
4037         discussion to reduce verbiage and improve flow.
4039 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
4041         PR middle-end/59658
4042         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
4043         also suppress many optimizations.  Alphabetize option lists for
4044         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
4045         correct documentation for those options to say that.
4046         * opts.c (default_options_table): Sort table by level and option
4047         name, to make it easier to correlate to the manual.
4049 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
4051         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
4052         printing of the names of segment registers.
4054 2018-11-14  Nathan Sidwell  <nathan@acm.org>
4056         PR debug/88006
4057         PR debug/87462
4058         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
4059         type list.
4061 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
4063         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
4065 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
4067         P1236R1 - Signed integers are two's complement
4068         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
4070         PR bootstrap/86739
4071         * hash-map.h (hash_map::iterator::reference_pair): New class.
4072         (hash_map::iterator::operator*): Return it rather than std::pair.
4074 2018-11-14  Jeff Law  <law@redhat.com>
4076         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
4077         iff the operand is a constant.
4079 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
4081         * gimple-ssa-evrp-analyze.c
4082         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
4083         ignore_equivs_equal_p to equal_p.
4084         * ipa-cp.c (meet_with_1): Use equal_p instead of
4085         ignore_equivs_equal_p.
4086         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
4087         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
4088         (value_range::operator==): Remove.
4089         (value_range::operator!=): Remove.
4090         (vrp_prop::visit_stmt): Use equal_p.
4091         * tree-vrp.h (value_range): Remove operator==, operator!=,
4092         ignore_equivs_equal_p.
4093         * vr-values.c (update_value_range): Use equal_p.
4095 2018-11-14  Michael Matz  <matz@suse.de>
4097         PR middle-end/86575
4098         * gimplify.c (collect_fallthrough_labels): Add new argument,
4099         return location via that, don't modify statements.
4100         (warn_implicit_fallthrough_r): Adjust call, don't use
4101         statement location directly.
4103 2018-11-14  Richard Biener  <rguenther@suse.de>
4105         PR middle-end/87985
4106         * tree-data-ref.c (split_constant_offset): Add wrapper
4107         allocating a cache hash-map.
4108         (split_constant_offset_1): Cache results of expanding
4109         expressions from SSA def stmts.
4111 2018-11-14  Richard Biener  <rguenther@suse.de>
4113         PR middle-end/88021
4114         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
4115         to lambda_int.
4116         (lambda_vector_mult_const): Likewise.
4117         (lambda_matrix_right_hermite): Use lambda_int temporaries.
4119 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
4120             Jackson Woodruff  <jackson.woodruff@arm.com>
4122         PR 71026/tree-optimization
4123         * match.pd: Simplify floating point comparisons.
4125 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
4127         PR rtl-optimization/87817
4128         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
4129         last argument has low 8 bits clear, fold to 0.
4131 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
4133         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
4134         MAX_OFILE_ALIGNMENT as needed.
4136 2018-11-14  Richard Biener  <rguenther@suse.de>
4138         PR tree-optimization/88019
4139         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
4140         COND_EXPR arguments to non-trapping overflow.
4142 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
4144         * config/arc/arc.h (reg_class): Reorder registers classes, remove
4145         unused register classes.
4146         (REG_CLASS_NAMES): Likewise.
4147         (REG_CLASS_CONTENTS): Likewise.
4148         (FIXED_REGISTERS): Make lp_count fixed.
4149         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
4150         (PROGRAM_COUNTER_REGNO): Remove.
4151         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
4152         register classes, use constants for register numbers, remove
4153         reg_class_contents references.
4154         (arc_process_double_reg_moves): Add asserts.
4155         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
4156         lra_in_progress predicate.
4157         (arc_init_reg_tables): Remove unused register classes.
4158         (arc_register_move_cost): Likewise.
4159         (arc_preferred_reload_class): Likewise.
4160         (hwloop_optimize): Update rtx patterns involving lp_count register.
4161         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
4162         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
4163         CHEAP_CORE_REGS.  Former one will be used for LRA.
4164         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
4165         one will be used for LRA.
4166         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
4167         one will be used for LRA.
4168         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
4169         Former one will be used for LRA.
4170         ("f"): Delete constraint.
4171         ("k"): Likewise.
4172         ("e"): Likewise.
4173         ("l"): Change it from register constraint to constraint.
4174         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
4175         (movhi_insn): Likewise.
4176         (movsi_insn): Update pattern.
4177         (arc_lp): Likewise.
4178         (dbnz): Likewise.
4179         (stack_tie): Remove 'b' constraint letter.
4180         (R4_REG): Define.
4181         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
4182         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
4183         (R57_REG, R59_REG, PCL_REG): Likewise.
4184         (ILINK1_REGNUM): Renamed to ILINK1_REG.
4185         (ILINK2_REGNUM): Renamed to ILINK2_REG.
4186         (Rgp): Remove.
4187         (SP_REGS): Likewise.
4188         (Rcw): Remove unused reg classes.
4189         * config/arc/predicates.md (dest_reg_operand): Just default on
4190         register_operand predicate.
4191         (mpy_dest_reg_operand): Likewise.
4192         (move_dest_operand): Use macros instead of constants.
4193         * config/arc/arc.opt (mlra): Switch to lra as default.
4195 2018-11-14  Richard Biener  <rguenther@suse.de>
4197         PR tree-optimization/87974
4198         * tree-vect-loop.c (vectorizable_reduction): When computing
4199         the vectorized reduction PHI vector type ignore constant
4200         and external defs.
4202 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
4204         PR tree-optimization/87977
4205         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
4206         stmt, build a new one and replace the old one with it.  Formatting fix.
4207         Call release_ssa_name (x) if !has_other_use and !delete_div.
4208         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
4209         verify lhs of stmt is still def.
4211 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
4213         PR rtl-optimization/87507
4214         * lower-subreg.c (operand_for_swap_move_operator): New function.
4215         (simple_move): Strip simple operators.
4216         (find_pseudo_copy): Likewise.
4217         (resolve_operand_for_swap_move_operator): New function.
4218         (resolve_simple_move): Strip simple operators and swap operands.
4220 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
4222         PR tree-optimization/87898
4223         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
4224         (ipa_simd_modify_function_body): Remove debug stmts where the first
4225         argument was changed into a non-decl.
4227 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
4229         PR rtl-optimization/87899
4230         * lra-lives.c (start_living): Update white space in comment.
4231         (enum point_type): New.
4232         (sparseset_contains_pseudos_p): New function.
4233         (update_pseudo_point): Likewise.
4234         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
4235         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
4236         handling.  Move early exit after adding conflicts.
4237         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
4238         if regno is already live.  Remove all handling of program points.
4239         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
4240         after adding conflicts.  Remove all handling of program points and
4241         ignore_reg_for_conflicts.
4242         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
4243         and do not guard call to mark_pseudo_live.
4244         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
4245         and do not guard call to mark_pseudo_dead.
4246         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
4247         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
4248         Use new function update_pseudo_point.  Handle register copies by
4249         removing the source register from the live set.  Handle INOUT operands.
4250         Update to the next program point using the unused_set, dead_set and
4251         start_dying sets.
4252         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
4254 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
4256         * builtins.c: Replace "source_location" with "location_t".
4257         * diagnostic-show-locus.c: Likewise.
4258         * diagnostic.c: Likewise.
4259         * dumpfile.c: Likewise.
4260         * gcc-rich-location.h: Likewise.
4261         * genmatch.c: Likewise.
4262         * gimple.h: Likewise.
4263         * gimplify.c: Likewise.
4264         * input.c: Likewise.
4265         * input.h: Likewise.  Eliminate the typedef.
4266         * omp-expand.c: Likewise.
4267         * selftest.h: Likewise.
4268         * substring-locations.h (get_source_location_for_substring):
4269         Rename to...
4270         (get_location_within_string): ...this.
4271         * tree-cfg.c: Replace "source_location" with "location_t".
4272         * tree-cfgcleanup.c: Likewise.
4273         * tree-diagnostic.c: Likewise.
4274         * tree-into-ssa.c: Likewise.
4275         * tree-outof-ssa.c: Likewise.
4276         * tree-parloops.c: Likewise.
4277         * tree-phinodes.c: Likewise.
4278         * tree-phinodes.h: Likewise.
4279         * tree-ssa-loop-ivopts.c: Likewise.
4280         * tree-ssa-loop-manip.c: Likewise.
4281         * tree-ssa-phiopt.c: Likewise.
4282         * tree-ssa-phiprop.c: Likewise.
4283         * tree-ssa-threadupdate.c: Likewise.
4284         * tree-ssa.c: Likewise.
4285         * tree-ssa.h: Likewise.
4286         * tree-vect-loop-manip.c: Likewise.
4288 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
4290         * doc/extend.texi: Fix typo in the weakref description.
4292 2018-11-13  Richard Biener  <rguenther@suse.de>
4294         PR tree-optimization/86991
4295         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
4296         group building until we have successfully detected the SLP
4297         reduction.
4298         (vect_is_simple_reduction): Remove fixup code here.
4300 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
4302         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
4303         (dump_gimple_stmt): Use it.
4304         (dump_gimple_stmt_loc): Likewise.
4305         (dump_gimple_expr): Likewise.
4306         (dump_gimple_expr_loc): Likewise.
4307         (dump_generic_expr): Likewise.
4308         (dump_generic_expr_loc): Likewise.
4309         (dump_printf): Likewise.
4310         (dump_printf_loc): Likewise.
4311         (dump_dec): Likewise.
4312         (dump_dec): Likewise.
4313         (dump_hex): Likewise.
4314         (dump_symtab_node): Likewise.
4315         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4316         Guard dump call with dump_enabled_p.
4317         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
4318         * graphite-optimize-isl.c (optimize_isl): Likewise.
4319         * graphite.c (graphite_transform_loops): Likewise.
4320         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
4321         * tree-parloops.c (parallelize_loops): Likewise.
4322         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
4323         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
4324         (vect_prune_runtime_alias_test_list): Likewise.
4325         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
4326         (vect_estimate_min_profitable_iters): Likewise.
4327         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
4328         (vect_build_slp_tree_2): Likewise.
4329         (vect_supported_load_permutation_p): Likewise.
4330         (vect_slp_analyze_operations): Likewise.
4331         (vect_slp_analyze_bb_1): Likewise.
4332         (vect_slp_bb): Likewise.
4333         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
4334         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
4335         (pass_slp_vectorize::execute): Likewise.
4336         (increase_alignment): Likewise.
4338 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
4340         PR ipa/87955
4341         * ipa-inline.c (report_inline_failed_reason): Guard calls to
4342         cl_target_option_print_diff and cl_optimization_print_diff with
4343         if (dump_file).
4345 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
4347         * doc/invoke.texi (-fsave-optimization-record): Note that the
4348         output is compressed.
4349         * optinfo-emit-json.cc: Include <zlib.h>.
4350         (optrecord_json_writer::write): Compress the output.
4352 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
4354         * tree-vrp.c (value_range_base::dump): Dump type.
4355         Do not use INF nomenclature for 1-bit types.
4356         (dump_value_range): Group all variants to common dumping code.
4357         (debug): New overloaded functions for value_ranges.
4358         (value_range_base::dump): Remove no argument version.
4359         (value_range::dump): Same.
4361 2018-11-13  Richard Biener  <rguenther@suse.de>
4363         PR tree-optimization/87931
4364         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
4365         nested cycles we support to latch computations vectorizable_reduction
4366         handles.
4368 2018-11-13  Martin Liska  <mliska@suse.cz>
4370         PR tree-optimization/87885
4371         * cfghooks.c (account_profile_record): Rename
4372         to ...
4373         (profile_record_check_consistency): ... this.
4374         Calculate missing num_mismatched_freq_in.
4375         (profile_record_account_profile): New function
4376         that calculates time and size of a function.
4377         * cfghooks.h (struct profile_record): Remove
4378         all tuples.
4379         (struct cfg_hooks): Remove after_pass flag.
4380         (account_profile_record): Rename to ...
4381         (profile_record_check_consistency): ... this.
4382         (profile_record_account_profile): New.
4383         * cfgrtl.c (rtl_account_profile_record): Remove
4384         after_pass flag.
4385         * passes.c (check_profile_consistency): Do only
4386         checking.
4387         (account_profile): Calculate size and time of
4388         function only.
4389         (pass_manager::dump_profile_report): Reformat
4390         output.
4391         (execute_one_ipa_transform_pass): Call
4392         consistency check before clean upand call account_profile
4393         after a clean up is done.
4394         (execute_one_pass): Call check_profile_consistency and
4395         account_profile instead of using after_pass flag..
4396         * tree-cfg.c (gimple_account_profile_record): Likewise.
4398 2018-11-13  Martin Liska  <mliska@suse.cz>
4400         PR sanitizer/87930
4401         * config/i386/i386.c (ix86_option_override_internal): Error
4402         about usage -mabi=ms and -fsanitize={,kernel-}address.
4404 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4406         * config/aarch64/aarch64.c
4407         (aarch64_vectorize_preferred_vector_alignment): Change return type to
4408         poly_uint64.
4409         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
4410         alignment being a poly int.
4411         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
4412         return type to poly_uint64.
4413         * target.def (default_preferred_vector_alignment): Likewise.
4414         * targhooks.c (default_preferred_vector_alignment): Likewise.
4415         * targhooks.h (default_preferred_vector_alignment): Likewise.
4416         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
4417         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
4418         poly int.
4419         (vect_update_misalignment_for_peel): Likewise.
4420         (vect_enhance_data_refs_alignment): Likewise.
4421         (vect_find_same_alignment_drs): Likewise.
4422         (vect_duplicate_ssa_name_ptr_info): Likewise.
4423         (vect_setup_realignment): Likewise.
4424         (vect_can_force_dr_alignment_p): Change alignment parameter type to
4425         poly_uint64.
4426         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
4427         mask with a compile time variable vector alignment.
4428         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
4429         int.
4430         (vect_do_peeling): Exit early if vector alignment is not constant.
4431         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
4432         a poly int.
4433         (vectorizable_store): Likewise.
4434         (vectorizable_load): Likweise.
4435         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
4436         poly_uint64.
4437         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
4438         poly int.
4439         (vect_can_force_dr_alignment_p): Change alignment parameter type to
4440         poly_uint64.
4442 2018-11-13  Richard Biener  <rguenther@suse.de>
4444         PR tree-optimization/87962
4445         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
4446         detect outer reduction for disqualifying in-loop uses.
4448 2018-11-13  Richard Biener  <rguenther@suse.de>
4450         PR tree-optimization/87967
4451         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
4452         for constants for the scalar loop.
4454 2018-11-13  Alan Modra  <amodra@gmail.com>
4456         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
4457         integer constants when -mcmodel=medium.
4459 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
4461         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
4462         * config/arc/elf.h (PROFILE_HOOK): Define.
4463         * config/arc/linux.h (PROFILE_HOOK): Likewise.
4465 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
4467         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
4468         jump table data in the text section.
4470 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
4472         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
4473         to fit the eh_return pattern.
4474         * config/arc/arc.md (eh_return): Define.
4475         (VUNSPEC_ARC_EH_RETURN): Likewise.
4476         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
4477         new implementation.
4478         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
4481 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
4483         * common/config/arc/arc-common.c (arc_option_optimization_table):
4484         Millicode optimization is default on for size optimizations.
4485         * config/arc/arc-protos.h (arc_check_multi): New function.
4486         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
4487         (ENTER_LEAVE_START_REG): Define.
4488         (ENTER_LEAVE_END_REG): Likewise.
4489         (arc_override_options): Disable millicode when long calls option
4490         is on.
4491         (arc_frame_info): Change it from int to bool.
4492         (arc_compute_frame_size): Clean up.
4493         (arc_save_restore): Remove.
4494         (frame_save_reg): New function.
4495         (frame_restore_reg): Likewise.
4496         (arc_enter_leave_p): Likewise.
4497         (arc_save_callee_saves): Likewise.
4498         (arc_restore_callee_saves): Likewise.
4499         (arc_save_callee_enter): Likewise.
4500         (arc_restore_callee_leave): Likewise.
4501         (arc_save_callee_milli): Likewise.
4502         (arc_restore_callee_milli): Likewise.
4503         (arc_expand_prologue): Reimplement to emit enter/leave
4504         instructions.
4505         (arc_expand_epilogue): Likewise.
4506         (arc_check_multi): New function.
4507         * config/arc/arc.md (push_multi_fp): New pattern.
4508         (push_multi_fp_blink): Likewise.
4509         (pop_multi_fp): Likewise.
4510         (pop_multi_fp_blink): Likewise.
4511         (pop_multi_fp_ret): Likewise.
4512         (pop_multi_fp_blink_ret): Likewise.
4513         * config/arc/arc.opt (mmillicode): Update option.
4514         (mcode-density-frame): New option.
4515         * config/arc/predicates.md (push_multi_operand): New predicate.
4516         (pop_multi_operand): Likewise.
4517         * doc/invoke.texi (ARC): Update ARC options information.
4520 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
4522         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
4523         * config/arc/arc.c (operands_ok_ldd_std): New function.
4524         (mem_ok_for_ldd_std): Likewise.
4525         (gen_operands_ldd_std): Likewise.
4526         * config/arc/arc.md: Add peephole2 rules for std/ldd.
4528 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
4530         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
4531         test on stack_usage_file.
4532         (lang_dependent_init): Do not open the .su file if generating LTO.
4534 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
4536         PR rtl-optimization/87918
4537         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
4538         simplify_gen_relational rather than simplify_gen_binary.
4540 2018-11-13  Richard Biener  <rguenther@suse.de>
4542         * tree-ssanames.h (set_range_info): Use value_range_base.
4543         (get_range_info): Likewise.
4544         * tree-ssanames.c (set_range_info): Likewise.
4545         (get_range_info): Likewise.
4546         * tree-vrp.c (value_range_base::union_helper): Split
4547         out common parts of value_range[_base]::union_.
4548         (value_range_base::union_): Update.
4549         (value_range::union_): Likewise.
4550         (determine_value_range_1): Use value_range_base.
4551         (determine_value_range): Likewise.
4552         * tree-vrp.h (value_range_base::union_helper): Move ...
4553         (value_range::union_helper): ... from here.
4555 2018-11-13  Alan Modra  <amodra@gmail.com>
4557         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
4558         offset for PRE_DEC.
4559         (rs6000_secondary_reload_gpr): Don't call find_replacement.
4561 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
4563         PR middle-end/59634
4564         * doc/invoke.texi (Optimize Options): Clarify that the
4565         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
4566         apply to data cache size.
4568 2018-11-13  Alan Modra  <amodra@gmail.com>
4570         * config/rs6000/predicates.md (logical_const_operand),
4571         (logical_operand): Correct comment.
4572         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
4574 2018-11-13  Alan Modra  <amodra@gmail.com>
4576         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
4578 2018-11-13  Alan Modra  <amodra@gmail.com>
4580         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
4581         description.
4582         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
4583         covered by alternative.
4584         (movcc_internal1): Ignore h for register preference.
4585         (mov<mode>_hardfloat64): Likewise.
4586         (mov<mode>_softfloat): Ignore c, l, h for register preference.
4588 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
4590         PR preprocessor/47823
4591         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
4592         documentation to...
4593         (Pragmas): ...here.
4594         * doc/extend.texi (Pragmas): Note additional pragmas documented
4595         in the CPP manual.
4597 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
4599         PR middle-end/86677
4600         PR middle-end/87528
4601         * tree-scalar-evolution.c (expression_expensive_p): Make
4602         BUILTIN POPCOUNT as expensive when backend does not define it.
4604 2018-11-12  Fredrik Noring  <noring@nocrew.org>
4606         * config.gcc: Update with-llsc defaults for MIPS r5900.
4608 2018-11-12  Martin Liska  <mliska@suse.cz>
4610         PR gcov-profile/87442
4611         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
4612         options.
4613         * doc/invoke.texi: Document them.
4614         * tree-profile.c (parse_profile_filter): New.
4615         (parse_profile_file_filtering): Likewise.
4616         (release_profile_file_filtering): Likewise.
4617         (include_source_file_for_profile): Likewise.
4618         (tree_profiling): Filter source files based on the
4619         newly added options.
4621 2018-11-12  Martin Liska  <mliska@suse.cz>
4623         PR target/87903
4624         * doc/extend.texi: Add missing values for __builtin_cpu_is and
4625         __builtin_cpu_supports for x86 target.
4627 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
4629         PR middle-end/21110
4630         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
4631         have to be Pmode.
4632         (Arithmetic): Likewise for "lo_sum".
4634 2018-11-12  Renlin Li  <renlin.li@arm.com>
4636         PR target/87815
4637         * dse.c (get_stored_val): Add check for compile-time constantness
4638         of gap.
4640 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
4642         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
4643         (ARMv8_5a): New fgroup.
4644         (armv8.5-a): New arch.
4645         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
4646         sb and predres.
4647         * config/arm/arm-tables.opt: Regenerate.
4648         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
4649         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
4650         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
4651         Add matching rules for -march=armv8.5-a and extensions.
4652         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
4653         Add sb and predres to all armv8-a except armv8.5-a.
4655 2018-11-12  Richard Biener  <rguenther@suse.de>
4657         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
4658         overload for single value.
4659         (value_range[_base]::set_nonnull): New.
4660         (value_range[_base]::set_null): Likewise.
4661         (value_range): Document bitmap copying behavior, mark
4662         copy constructor and assignment operator deleted.
4663         (value_range::move): New.
4664         (value_range::set_and_canonicalize): Default bitmap to zero.
4665         (set_value_range_to_nonnull): Remove.
4666         (set_value_range_to_null): Likewise.
4667         (set_value_range): Likewise.
4668         (set_value_range_to_value): Likewise.
4669         (extract_range_from_unary_expr): Work on value_range_base.
4670         (extract_range_from_binary_expr_1): Likewise.  Rename to...
4671         (extract_range_from_binary_expr): ... this.
4672         * tree-vrp.c (value_range::update): Clear equiv bitmap
4673         if required.
4674         (value_range::move): New, move equiv bitmap.
4675         (value_range_base::set_undefined): Avoid assignment.
4676         (value_range::set_undefined): Likewise.
4677         (value_range_base::set_varying): Likewise.
4678         (value_range::set_varying): Likewise.
4679         (set_value_range): Remove.
4680         (value_range_base::set): New overload for value.
4681         (value_range::set): Likewise.
4682         (set_value_range_to_nonnull): Remove.
4683         (value_range_base::set_nonnull): New.
4684         (value_range::set_nonnull): Likewise.
4685         (set_value_range_to_null): Remove.
4686         (value_range_base::set_null): New.
4687         (value_range::set_null): Likewise.
4688         (range_is_null): Work on value_range_base.
4689         (range_is_nonnull): Likewise.
4690         (ranges_from_anti_range): Likewise.
4691         (extract_range_into_wide_ints): Likewise.
4692         (extract_range_from_multiplicative_op): Likewise.
4693         (extract_range_from_binary_expr): Likewise.  Update for API changes.
4694         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
4695         handling.
4696         (value_range::intersect_helper): Avoid copy and assignment.
4697         (value_range::union_helper): Likewise.
4698         (determine_value_range_1): Adjust.
4699         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
4700         Avoid assignment by using move.
4701         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
4702         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4703         Likewise.
4704         * tree-ssanames.c (get_range_info): Likewise.
4705         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
4706         * vr-values.c (vr_values::get_value_range): Adjust.
4707         (vr_values::update_value_range): Likewise.
4708         (symbolic_range_based_on_p): Work on value_range_base.
4709         (vr_values::extract_range_from_binary_expr): Use value_range_base.
4710         (vr_values::extract_range_from_unary_expr): Likewise.
4711         (vr_values::extract_range_from_cond_expr): Avoid assignment.
4712         (vr_values::extract_range_from_comparison): Adjust.
4713         (vr_values::check_for_binary_op_overflow): Use value_range_base.
4714         (vr_values::extract_range_basic): Adjust.
4715         (vr_values::adjust_range_with_scev): Likewise.
4716         (vr_values::vrp_visit_assignment_or_call): Likewise.
4717         (vr_values::get_vr_for_comparison): Change API to avoid
4718         assignment and copy construction.
4719         (vr_values::compare_name_with_value): Adjust accordingly.
4720         (vr_values::compare_names): Likewise.
4721         (vr_values::extract_range_from_phi_node): Avoid assignment and
4722         bogus in-place modify of equiv bitmap.
4723         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
4724         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
4725         for extract_range_from_unary_expr API change.
4726         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
4728 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
4730         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
4732 2018-11-12  Richard Biener  <rguenther@suse.de>
4734         * tree-vrp.h (value_range_base::symbolic_p,
4735         value_range_base::constant_p, value_range_base::zero_p,
4736         value_range_base::singleton_p): Move from value_range.
4737         (value_range::dump): Add.
4738         * gimple-ssa-evrp-analyze.c
4739         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
4740         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
4741         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4742         Use set_varying.
4743         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
4744         (value_range::constant_p): Likewise.
4745         (value_range::singleton_p): Likewise.
4746         (value_range_base::dump): Add.
4747         (set_value_range_to_undefined): Remove.
4748         (set_value_range_to_varying): Likewise.
4749         (range_int_cst_p): Take value_range_base argument.
4750         (range_int_cst_singleton_p): Likewise.
4751         (value_range_constant_singleton): Likewise.
4752         (vrp_set_zero_nonzero_bits): Likewise.
4753         (extract_range_from_multiplicative_op): Use set_varying.
4754         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
4755         (extract_range_from_unary_expr): Likewise.
4756         (dump_value_range_base): Change to overload of dump_value_range.
4757         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
4758         (vrp_prop::visit_stmt): Likewise.
4759         (value_range::intersect_helper): Likewise.
4760         (value_range::union_helper): Likewise.
4761         (determine_value_range_1): Likewise.
4763 2018-11-12  Richard Biener  <rguenther@suse.de>
4765         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
4766         (set_value_range_to_null): Likewise.
4767         * vr-values.c (vr_values::extract_range_from_comparison):
4768         Clear equiv for constant singleton ranges.
4770 2018-11-12  Wei Xiao  <wei3.xiao@intel.com>
4772         * config/i386/sse.md: Combine VFIXUPIMM* patterns
4773         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4774         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4775         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
4776         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4777         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4778         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
4780 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
4782         PR c/69502
4783         * doc/extend.texi (Common Type Attributes): For the align type
4784         attribute, copy language about decreasing alignment from the
4785         corresponding variable attribute.
4787 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4789         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
4790         -frounding-math.
4792 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
4794         PR c++/43105
4795         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
4796         -frtti and -fno-rtti code.
4798 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
4800         PR c/26366
4801         * doc/extend.texi (Other Builtins): Document probability associated
4802         with __builtin_expect.
4804 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
4806         PR target/87928
4807         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
4808         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
4809         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
4810         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
4812 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4814         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
4816 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4818         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
4820 2018-11-11  Richard Biener  <rguenther@suse.de>
4822         * tree-vrp.h (class value_range_base): New base class for
4823         value_range containing all but the m_equiv member.
4824         (dump_value_range_base): Add.
4825         (range_includes_zero_p): Work on value_range_base.
4826         * tree-vrp.c (value_range_base::set): Split out base handling
4827         from...
4828         (value_range::set): this.
4829         (value_range::set_equiv): New.
4830         (value_range_base::value_range_base): New constructors.
4831         (value_range_base::check): Split out base handling from...
4832         (value_range::check): this.
4833         (value_range::equal_p): Refactor in terms of
4834         ignore_equivs_equal_p which is now member of the base.
4835         (value_range_base::set_undefined): New.
4836         (value_range_base::set_varying): Likewise.
4837         (value_range_base::dump):Split out base handling from...
4838         (value_range::dump): this.
4839         (value_range_base::set_and_canonicalize): Split out base handling
4840         from...
4841         (value_range::set_and_canonicalize): this.
4842         (value_range_base::union_): New.
4843         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
4844         for m_vr.
4845         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
4846         instead of value_range everywhere.
4847         (ipcp_vr_lattice::print): Use dump_value_range_base.
4848         (ipcp_vr_lattice::meet_with): Adjust.
4849         (ipcp_vr_lattice::meet_with_1): Likewise.
4850         (ipa_vr_operation_and_type_effects): Likewise.
4851         (propagate_vr_across_jump_function): Likewise.
4852         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
4853         (ipa_get_value_range): Likewise.
4854         (ipa_set_jfunc_vr): Likewise.
4855         (ipa_compute_jump_functions_for_edge): Likewise.
4857 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
4859         PR middle-end/65703
4860         * doc/invoke.texi (Optimize Options): Add @opindex entries
4861         for the positive forms of -fno-xxx and -mno-xxx options
4862         that were lacking them.
4864 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
4866         * combine.c (make_more_copies): Only make an intermediate copy if the
4867         dest of a move is a pseudo.
4869 2018-11-09  Maya Rashish  <coypu@sdf.org>
4871         PR target/87221
4872         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
4873         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
4875 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
4877         PR driver/41179
4878         PR middle-end/65703
4879         * doc/invoke.texi (Optimize Options): Clarify default behavior
4880         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
4882 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4884         PR tree-optimization/87940
4885         * expr.c (string_constant): Don't strip NOPS in subexpressions.
4886         Fold PLUS_EXPR correctly.
4888 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
4890         PR target/87762
4891         * config/s390/s390.c (s390_safe_relative_long_p): New function.
4892         (annotate_constant_pool_refs): Skip insns which support
4893         relative addressing.
4894         (annotate_constant_pool_refs_1): New helper function.
4895         (find_constant_pool_ref): Skip insns which support relative
4896         addression.
4897         (find_constant_pool_ref_1): New helper function.
4898         (replace_constant_pool_ref): Skip insns which support
4899         relative addressing.
4900         (replace_constant_pool_ref_1): New helper function.
4901         (s390_mainpool_start): Adapt to the new signature.
4902         (s390_mainpool_finish): Likewise.
4903         (s390_chunkify_start): Likewise.
4904         (s390_chunkify_finish): Likewise.
4905         (pass_s390_early_mach::execute): Likewise.
4906         (s390_prologue_plus_offset): Likewise.
4907         (s390_emit_prologue): Likewise.
4908         (s390_emit_epilogue): Likewise.
4910 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
4912         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
4913         but unsupported lastprivate with conditional modifier.
4915 2018-11-09  Jeff Law  <law@redhat.com>
4917         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
4918         unused argument better.  Add gcc_unreachable to silence warning.
4920 2018-11-09  Martin Sebor  <msebor@redhat.com>
4922         PR middle-end/81824
4923         * attribs.c (has_attribute): New helper function.
4924         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
4925         * attribs.h (decls_mismatched_attributes): Declare.
4926         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
4927         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
4928         * common.opt (-Wattribute-alias): Take an argument.
4929         (-Wno-attribute-alias): New option.
4930         * doc/extend.texi (Common Function Attributes): Document copy.
4931         (Common Variable Attributes): Same.
4932         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
4933         (-Wattribute-alias): Document new option argument.
4935 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
4937         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
4938         matching criteria.  Remove unused array initializer.
4940 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
4941             Jinsong Ji  <jji@us.ibm.com>
4943         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
4944         constraints by introducing a new temporary.
4945         (_mm_cvtss_si64): Likewise.
4947 2018-11-09  Martin Liska  <mliska@suse.cz>
4949         * common.opt: Add -fipa-stack-alignment flag.
4950         * doc/invoke.texi: Document it.
4951         * final.c (rest_of_clean_state): Guard stack
4952         shrinking with flag.
4954 2018-11-09  Martin Liska  <mliska@suse.cz>
4956         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
4957         to ...
4958         (ipa_discover_variable_flags): ... this.
4959         * common.opt: Come up with new flag -fipa-reference-addressable.
4960         * doc/invoke.texi: Document it.
4961         * ipa-reference.c (propagate): Call the renamed fn.
4962         * ipa-visibility.c (whole_program_function_and_variable_visibility):
4963         Likewise.
4964         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
4965         ...
4966         (ipa_discover_variable_flags): ... this.  Discover
4967         non-addressable variables only with the newly added flag.
4968         * opts.c: Enable the newly added flag with -O1 and higher
4969         optimization level.
4971 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
4973         * json.cc (selftest::test_writing_literals): Fix comment.
4975 2018-11-09  Martin Liska  <mliska@suse.cz>
4977         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
4978         string to a stack buffer.
4979         (aarch64_parse_cpu): Likewise.
4980         (aarch64_parse_tune): Likewise.
4982 2018-11-09  Richard Biener  <rguenther@suse.de>
4984         PR tree-optimization/87953
4985         * tree-vect-loop.c (vectorizable_reduction): For analysis
4986         always pass ops[0] to vectorizable_condition.
4988 2018-11-09  Stafford Horne  <shorne@gmail.com>
4989             Richard Henderson  <rth@twiddle.net>
4990             Joel Sherrill  <joel@rtems.org>
4992         * common/config/or1k/or1k-common.c: New file.
4993         * config/or1k/*: New.
4994         * config.gcc (or1k*-*-*): New.
4995         * configure.ac (or1k*-*-*): New test for openrisc tls.
4996         * configure: Regenerated.
4997         * doc/install.texi: Document OpenRISC triplets.
4998         * doc/invoke.texi: Document OpenRISC arguments.
4999         * doc/md.texi: Document OpenRISC.
5001 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
5003         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
5004         (arm7tdmi-s): Delete CPU.
5005         (arm710t): Add aliases for arm720t and arm740t.
5006         (arm720t, arm740t): Delete CPUs.
5007         (arm920t): Add aliases for arm920, arm922t and arm940t.
5008         (arm920, arm922t, arm940t): Delete CPUs.
5009         (arm10tdmi): Add alias for arm1020t.
5010         (arm1020t): Delete CPU.
5011         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
5012         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
5013         (arm10e): Add aliases for arm1020e and arm1022e.
5014         (arm1020e, arm1022e): Delete CPU.
5015         * config/arm/arm.md (generic_sched): Remove entries that are now
5016         handled by aliases.
5017         (generic_vfp): Likewise.
5018         * config/arm/arm1020e.md: Simplify tuning selection based on alias
5019         changes.
5020         * config/arm/arm-tune.md: Regenerated.
5021         * config/arm/arm-tables.opt: Regenerated.
5023 2018-11-09  Richard Biener  <rguenther@suse.de>
5025         PR tree-optimization/87621
5026         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
5027         op with only phi inputs.
5028         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
5029         (ch_base::copy_headers): Run CSE on copied loop headers.
5030         (pass_ch_vect::process_loop_p): Simplify.
5032 2018-11-09  Alexandre Oliva  <oliva@adacore.com>
5034         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
5035         for non-w64 x86_64 biarch.
5037 2018-11-09  Alexandre Oliva  <aoliva@redhat.com>
5039         PR rtl-optimization/86438
5040         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
5041         of in_b for the compare if in_b is SET_DEST.
5043         PR target/87793
5044         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
5045         non-toplevel UNSPEC.
5047 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
5049         * tree-vrp.c (value_range::check): Do not access internals
5050         directly.
5051         (value_range::singleton_p): Same.
5052         (value_range::type): Same.
5053         (vrp_finalize): Use value_range API.
5055 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
5057         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
5059 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
5061         * vr-values.c (vr_values::get_value_range): Use value_range API
5062         instead of piecing together ranges.
5063         (vr_values::update_value_range): Same.
5065 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
5067         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
5068         of performing ad-hoc calculations.
5069         * tree-ssanames.c (set_range_info): New overloaded function
5070         accepting value_range &.
5071         (get_range_info): Same.
5072         * tree-ssanames.h (set_range_info_raw): Remove.
5073         (set_range_info): New prototype.
5074         (get_range_info): Same.
5075         * tree-vrp.h (value_range::null_p): Rename to zero_p.
5076         * tree-vrp.c (value_range::null_p): Same.
5078 2018-11-09  Jan Hubicka  <jh@suse.cz>
5080         * tree.c (fld_type_variant_equal_p): Test user align flag.
5081         (flt_type_variant): Copy user align flag.
5082         (fld_incomplete_type_of): Clear it.
5084 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5086         * config/arm/neon.md (div<mode>3): New pattern.
5088 2018-11-08  Andi Kleen  <ak@linux.intel.com>
5090         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
5091         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
5092         (ix86_handle_option): Handle OPT_mptwrite.
5093         * config/i386/cpuid.h (bit_PTWRITE): Add.
5094         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
5095         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
5096         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
5097         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
5098         (ix86_option_override_internal): Handle PTA_PTWRITE.
5099         (ix86_valid_target_attribute_inner_p): Define ptwrite.
5100         (def_builtin2): Force UINT64 to be 64bit only.
5101         * config/i386/i386.h (TARGET_PTWRITE): Add.
5102         (TARGET_PTWRITE_P): Add.
5103         (PTA_PTWRITE): Add.
5104         * config/i386/i386.md: Define ptwrite.
5105         * config/i386/i386.opt: Add -mptwrite.
5106         * config/i386/immintrin.h (_ptwrite64): Add.
5107         (_ptwrite32): Add
5108         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
5109         * doc/invoke.texi: Document -mptwrite.
5111 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
5113         PR rtl-optimization/87600
5114         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
5115         * lra-constraints.c (process_alt_operands): Skip illegal hard
5116         register usage.  Prefer reloading non hard register operands.
5118 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
5120         PR other/36572
5121         * doc/invoke.texi (Optimize Options): Clarify default behavior
5122         for -fno-sched-interblock and -fno-sched-spec.
5124 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
5126         * collect2.c (linker_select):  Add USE_LLD_LD.
5127         (ld_suffixes): Add ld.lld.
5128         (main): Handle -fuse-ld=lld.
5129         * common.opt (-fuse-ld=lld): New option.
5130         * doc/invoke.texi (-fuse-ld=lld): Document.
5131         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
5133 2018-11-08  Paul Koning  <ni1d@arrl.net>
5135         * config/pdp11/constraints.md: Add "Z" series constraints for use
5136         with pre-dec and post-inc addressing.
5137         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
5138         (pdp11_expand_operands): Add int argument (word count).
5139         (pdp11_sp_frame_offset): Delete.
5140         (pdp11_cmp_length): New function.
5141         (pushpop_regeq): New function.
5142         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
5143         Add hook.
5144         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
5145         frame layout.
5146         (pdp11_initial_elimination_offset): Ditto.
5147         (pdp11_expand_operands): Add word count argument.  Bugfixes.
5148         (output_move_multiple): Change how pointer adjustment is done.
5149         (pdp11_gen_int_label): Correct format.
5150         (output_ascii): Ditto.
5151         (pdp11_asm_output_var): Add code for DEC assembler case.
5152         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
5153         value.
5154         (legitimate_const_double_p): Ditto.
5155         (pdp11_register_move_cost): Adjust for new register classes.
5156         (pdp11_regno_reg_class): Ditto.
5157         (expand_block_move): Delete.
5158         (pushpop_regeq): New function.
5159         (pdp11_legitimate_address_p): Bugfix in check for constant
5160         offset.
5161         (pdp11_sp_frame_offset): Delete.
5162         (pdp11_reg_save_size): New helper function for new frame layout.
5163         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
5164         (pdp11_expand_shift): Bugfix in check for constant shift count.
5165         (pdp11_shift_length): Ditto.
5166         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
5167         (pdp11_cmp_length): New function.
5168         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
5169         some compile options.
5170         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
5171         (CALL_USED_REGISTERS): Ditto.
5172         (ELIMINABLE_REGS): Ditto.
5173         (REGISTER_NAMES): Ditto.
5174         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
5175         constraints.
5176         (REG_CLASS_NAMES): Ditto.
5177         (REG_CLASS_CONTENTS): Ditto.  Also remove
5178         HARD_FRAME_POINTER_REGNUM.
5179         (CPU_REG_CLASS): New macro.
5180         (CLASS_MAX_NREGS): Adjust for new register classes.
5181         (FUNCTION_PROFILER): Make no-op.
5182         (may_call_alloca): Remove unused declaration.
5183         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
5184         (ASM_OUTPUT_SKIP): Fix format.
5185         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
5186         (HARD_FRAME_POINTER_REGNUM): Remove.
5187         (return): Delete.
5188         (*rts): Rename.  Remove epilogue related checks.
5189         (cmpsi, cmpdi): New insn.
5190         (cbranch<mode>4): Change to apply to SI and DI modes as well.
5191         (mov<mode>): Change constraints to enforce that push/pop
5192         destination cannot use the same register as source.
5193         (*mov<mode><cc_cc>): Ditto.
5194         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
5195         at assembly output rather than as RTL expander.
5196         (zero_extendqihi2): Bugfix in check for same registers.
5197         (adddi3_nocc): Bugfix in check for constant operand.
5198         (addsi3_nocc): Ditto.
5199         (subdi3_nocc): Ditto.
5200         (subsi3_nocc): Ditto.
5201         (negdi2_nocc): Copy input to pdp11_expand_operands.
5202         (negsi2_nocc): Ditto.
5203         (bswap2_nocc): Ditto.
5204         * config/pdp11/pdp11.opt (mlra): Fix documentation.
5205         * config/pdp11/t-pdp11: Use -Os.
5207 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
5209         * config/arm/parsecpu.awk (/alias/): New parsing rule.
5210         (/begin cpu/): Check that the cpu name hasn't been previously defined.
5211         (gen_comm_data): Print out CPU alias tables.
5212         (check_cpu): Match aliases when checking the CPU name.
5213         * config/arm/arm-protos.h (cpu_alias): New structure.
5214         (cpu_option): Add entry for aliases.
5215         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
5216         strongarm1100 and strongarm1110.
5217         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
5218         (config/arm/arm-generic.md): Remove redundant references to
5219         strongarm110, strongarm1100 and strongarm1110.
5220         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5221         Scan aliases for additional hints.
5222         (arm_parse_cpu_option_name): Also match a cpu name against the list
5223         of aliases.
5224         * config/arm/arm-tables.opt: Regenerated.
5225         * config/arm/arm-tune.md: Regenerated.
5227 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
5229         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
5230         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
5231         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
5232         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
5233         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
5234         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
5235         * gengtype.c (open_base_files): Add omp-general.h.
5236         * gimple.c (gimple_build_omp_critical):
5237         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
5238         gimple_omp_taskgroup_set_clauses.
5239         (gimple_build_omp_atomic_load): Add mo argument, call
5240         gimple_omp_atomic_set_memory_order.
5241         (gimple_build_omp_atomic_store): Likewise.
5242         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
5243         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
5244         instead of GSS_OMP.
5245         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
5246         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
5247         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
5248         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
5249         different value for GF_OMP_ATOMIC_NEED_VALUE.
5250         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
5251         comments.
5252         (struct gimple_statement_omp_single_layout): And remove here.
5253         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
5254         than gimple_statement_omp_single_layout.
5255         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
5256         GIMPLE_OMP_TEAMS.
5257         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
5258         (gimple_omp_subcode): Formatting fix.
5259         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
5260         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
5261         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
5262         gimple_omp_teams_host, gimple_omp_teams_set_host,
5263         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
5264         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
5265         gimple_omp_taskgroup_set_clauses): New inline functions.
5266         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
5267         (gimple_build_omp_atomic_store): Likewise.
5268         (gimple_omp_atomic_seq_cst_p): Remove.
5269         (gimple_omp_atomic_memory_order): New function.
5270         (gimple_omp_atomic_set_seq_cst): Remove.
5271         (gimple_omp_atomic_set_memory_order): New function.
5272         (gimple_build_omp_taskgroup): Add clauses argument.
5273         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
5274         (dump_gimple_omp_task): Print taskwait with depend clauses.
5275         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
5276         dump_omp_atomic_memory_order.
5277         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
5278         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
5279         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
5280         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
5281         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
5282         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
5283         ORT_UNTIED_TASKLOOP enumerators.
5284         (enum gimplify_defaultmap_kind): New.
5285         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
5286         target_map_pointers_as_0len_arrays members, add defaultmap.
5287         (new_omp_context): Initialize defaultmap member.
5288         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5289         (maybe_fold_stmt): Don't fold even in host teams regions.
5290         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
5291         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
5292         ctx->omp_firstprivatize_variable.
5293         (omp_add_variable): Don't add private/firstprivate for VLAs in
5294         ORT_TASKGROUP.
5295         (omp_default_clause): Print "taskloop" rather than "task" if
5296         ORT_*TASKLOOP.
5297         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5298         Handle new defaultmap clause kinds.
5299         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
5300         iterator to be lastprivate or private.  Fix up diagnostics if linear
5301         is used on collapse>1 simd iterator.
5302         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5303         (gimplify_omp_depend): New function.
5304         (gimplify_scan_omp_clauses): Add shared clause on parallel for
5305         combined parallel master taskloop{, simd} if taskloop has
5306         firstprivate, lastprivate or reduction clause.  Handle
5307         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
5308         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
5309         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
5310         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
5311         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
5312         lastprivate.
5313         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
5314         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
5315         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
5316         OMP_CLAUSE_{TASK,IN}_REDUCTION.
5317         (gimplify_omp_task): Handle taskwait with depend clauses.
5318         (gimplify_omp_for): Add shared clause on parallel for combined
5319         parallel master taskloop{, simd} if taskloop has firstprivate,
5320         lastprivate or reduction clause.  Use ORT_TASKLOOP or
5321         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
5322         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
5323         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
5324         __for_end and __for_range temporaries on OMP_PARALLEL for
5325         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
5326         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
5327         sandwiched in between two taskloops.
5328         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
5329         instead of ctx->omp_firstprivatize_variable.
5330         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
5331         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
5332         host teams, use gimplify_and_return_first etc. for body like
5333         for target or target data constructs, and at the end call
5334         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
5335         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
5336         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
5337         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
5338         gimple_omp_atomic_set_seq_cst calls.
5339         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
5340         case, handle taskgroup clauses.
5341         * lto-streamer-out.c (hash_tree): Handle
5342         OMP_CLAUSE_{TASK,IN}_REDUCTION.
5343         * Makefile.in (GTFILES): Add omp-general.h.
5344         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
5345         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
5346         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
5347         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
5348         BUILT_IN_GOMP_LOOP_DOACROSS_START,
5349         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
5350         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5351         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
5352         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
5353         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
5354         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
5355         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
5356         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5357         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
5358         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
5359         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
5360         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
5361         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
5362         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
5363         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
5364         * omp-expand.c (workshare_safe_to_combine_p): Return false for
5365         non-worksharing loops.
5366         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
5367         (determine_parallel_type): Don't combine parallel with worksharing
5368         which has _reductemp_ clause.
5369         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
5370         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
5371         if there is nonmonotonic modifier or if there is no modifier and no
5372         ordered clause.  For dynamic and guided schedule without monotonic
5373         and nonmonotonic modifier, default to nonmonotonic.
5374         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
5375         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
5376         task reductions.
5377         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
5378         there are any reduction clauses.
5379         (expand_taskwait_call): New function.
5380         (expand_teams_call): New function.
5381         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
5382         expand_teams_call for it.  Formatting fix.  Handle taskwait with
5383         depend clauses.
5384         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
5385         of worksharing loops with task reductions.
5386         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
5387         expansion of worksharing loops with task reductions.
5388         (expand_omp_sections): Handle expansion of sections with task
5389         reductions.
5390         (expand_omp_synch): For host teams call expand_omp_taskreg.
5391         (omp_memory_order_to_memmodel): New function.
5392         (expand_omp_atomic_load, expand_omp_atomic_store,
5393         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
5394         instead of gimple_omp_atomic_seq_cst_p.
5395         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
5396         depend clauses as a standalone directive.
5397         * omp-general.c (enum omp_requires): New variable.
5398         (omp_extract_for_data): Initialize have_reductemp member.  Allow
5399         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
5400         GT_EXPR loops depending on incr sign.  Formatting fixes.
5401         * omp-general.h (struct omp_for_data): Add have_reductemp member.
5402         (enum omp_requires): New enum.
5403         (omp_requires_mask): Declare.
5404         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
5405         Fix comment typos.
5406         * omp-low.c (struct omp_context): Add task_reductions and
5407         task_reduction_map fields.
5408         (is_host_teams_ctx): New function.
5409         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
5410         (use_pointer_for_field): Use is_global_var instead of
5411         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
5412         in outer contexts.
5413         (build_outer_var_ref): Ignore taskgroup outer contexts.
5414         (delete_omp_context): Release task_reductions and task_reduction_map.
5415         (scan_sharing_clauses): Don't add any fields for reduction clause on
5416         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
5417         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
5418         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
5419         Handle OMP_CLAUSE_NONTEMPORAL.
5420         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
5421         needed.
5422         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
5423         clauses with task modifier.
5424         (scan_omp_task): Handle taskwait with depend clauses.
5425         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
5426         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
5427         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
5428         (scan_omp_for): Fix comment formatting.
5429         (scan_omp_teams): Handle host teams constructs.
5430         (check_omp_nesting_restrictions): Allow teams with no outer
5431         OpenMP context.  Adjust diagnostics for teams strictly nested into
5432         some explicit OpenMP construct other than target.  Allow OpenMP atomics
5433         inside of simd regions.
5434         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
5435         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
5436         taskreg_nesting_level while scanning host teams construct.
5437         (task_reduction_read): New function.
5438         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
5439         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
5440         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
5441         second argument create_tmp_var if it is NULL.  Don't ignore shared
5442         clauses in is_host_teams_ctx contexts.  Handle
5443         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
5444         clauses.
5445         (lower_reduction_clauses): Ignore reduction clauses with task
5446         modifier.  Remove second argument create_tmp_var if it is NULL.
5447         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
5448         (lower_send_clauses): Ignore reduction clauses with task modifier.
5449         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
5450         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
5451         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
5452         rely that it is the last stmt in body so far.  Ignore outer taskgroup
5453         contexts.
5454         (omp_task_reductions_find_first, omp_task_reduction_iterate,
5455         lower_omp_task_reductions): New functions.
5456         (lower_omp_sections): Handle reduction clauses with taskgroup
5457         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
5458         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
5459         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
5460         modifiers.
5461         (lower_omp_taskgroup): Handle taskgroup reductions.
5462         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
5463         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
5464         (lower_depend_clauses): If there are any
5465         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5466         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
5467         seen, assume lowering is done already and return early.  Set kind
5468         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
5469         (lower_omp_taskreg): Handle reduction clauses with task modifier on
5470         parallel construct.  Handle reduction clause on taskloop construct.
5471         Handle taskwait with depend clauses.
5472         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
5473         for host teams constructs.
5474         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
5475         nontemporal and _reductemp_ clause entries.
5476         (omp_clause_code_name): Likewise.
5477         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
5478         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
5479         * tree-core.h (enum omp_clause_code): Add
5480         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
5481         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
5482         (struct tree_base): Add omp_atomic_memory_order field into union.
5483         Remove OMP_ATOMIC_SEQ_CST comment.
5484         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5485         and OMP_CLAUSE_DEPEND_DEPOBJ.
5486         (struct tree_omp_clause): Add subcode.defaultmap_kind.
5487         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
5488         OpenMP constructs with body and clauses operands.
5489         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
5490         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
5491         (OMP_TASKGROUP_CLAUSES): Define.
5492         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
5493         OMP_CLAUSE__LOOPTEMP_.
5494         (OMP_ATOMIC_SEQ_CST): Remove.
5495         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
5496         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
5497         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
5498         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
5499         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
5500         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
5501         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
5502         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
5503         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
5504         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
5505         Define.
5506         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
5507         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5508         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
5509         (convert_local_omp_clauses): Likewise.  Remove useless test.
5510         * tree-parloops.c (create_call_for_reduction_1): Pass
5511         OMP_MEMORY_ORDER_RELAXED as new argument to
5512         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
5513         * tree-pretty-print.c (dump_omp_iterators): New function.
5514         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
5515         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
5516         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
5517         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
5518         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
5519         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
5520         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
5521         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
5522         (dump_omp_atomic_memory_order): New function.
5523         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
5524         taskwait with depend clauses.
5525         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
5526         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
5527         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
5528         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
5529         write_ts_omp_clause_tree_pointers): Likewise.
5531 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
5533         PR ipa/86395
5534         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
5535         "-missed", "-note", and "-all" sub-options.
5536         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
5537         API.
5538         (can_early_inline_edge_p): Likewise.
5539         (want_early_inline_function_p): Likewise.
5540         (want_inline_self_recursive_call_p): Likewise.
5541         (recursive_inlining): Likewise.
5542         (inline_small_functions): Likewise.
5543         (flatten_function): Likewise.
5544         (ipa_inline): Likewise.
5545         (inline_always_inline_functions): Likewise.
5546         (early_inline_small_functions): Likewise.
5547         (early_inliner): Likewise.
5548         * tree-inline.c (expand_call_inline): Likewise.
5550 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
5552         * pretty-print.c (pp_format): Handle %f.
5553         (selftest::test_pp_format): Add test of %f.
5554         * pretty-print.h (pp_double): New macro.
5556 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
5558         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
5559         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
5560         (dump_pretty_printer::decode_format): Implement "%C" for
5561         cgraph_node *.
5562         (selftest::test_capture_of_dump_calls): Rename "where" to
5563         "stmt_loc".  Convert test_decl to a function decl and set its
5564         location.  Add a symbol_table_test RAII instance and a
5565         cgraph_node, using it to test "%C" and dump_symtab_node.
5567 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
5569         PR middle-end/87916
5570         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
5572 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
5574         * cgraph.c: Include "selftest.h".
5575         (saved_symtab): New variable.
5576         (selftest::symbol_table_test::symbol_table_test): New ctor.
5577         (selftest::symbol_table_test::~symbol_table_test): New dtor.
5578         (selftest::test_symbol_table_test): New test.
5579         (selftest::cgraph_c_tests): New.
5580         * cgraph.h (saved_symtab): New decl.
5581         (selftest::symbol_table_test): New class.
5582         * selftest-run-tests.c (selftest::run_tests): Call
5583         selftest::cgraph_c_tests.
5584         * selftest.h (selftest::cgraph_c_tests): New decl.
5586 2018-11-08  Richard Biener  <rguenther@suse.de>
5588         * tree-data-ref.h (lambda_int): New typedef.
5589         (lambda_vector_gcd): Adjust.
5590         (lambda_vector_new): Likewise.
5591         (lambda_matrix_new): Likewise.
5592         * tree-data-ref.c  (print_lambda_vector): Adjust.
5594 2018-11-08  Richard Biener  <rguenther@suse.de>
5596         PR tree-optimization/87929
5597         * tree-complex.c (expand_complex_comparison): Clean EH.
5599 2018-11-08  Martin Liska  <mliska@suse.cz>
5601         * doc/extend.texi: Reword.
5602         * predict.c (expr_expected_value_1): Likewise.
5604 2018-11-08  Richard Biener  <rguenther@suse.de>
5606         PR tree-optimization/87913
5607         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
5608         of extreme values to ordered comparisons.
5610 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
5612         PR middle-end/42726
5613         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
5615 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
5617         * doc/invoke.texi: Remove leading dash from @opindex entries
5618         throughout the file.
5620 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
5622         PR driver/80828
5623         * doc/invoke.texi (Option Summary): Add -e and --entry.
5624         (Link Options): Likewise.
5626 2018-11-07  Nathan Sidwell  <nathan@acm.org>
5628         PR 87926
5629         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
5630         --disable-checking bootstrap.
5632 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
5634         * configure: Regenerated.
5636 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5638         PR c/87691
5639         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
5640         to the mode of the widest field iff the widest field has mode class
5641         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
5642         reference.
5644 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
5646         * common.opt: Add -Wattribute-warning.
5647         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
5648         * expr.c (expand_expr_real_1): Add new attribute to warning_at
5649         call to allow user configure behavior of "warning" attribute.
5651 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
5653         * target.def: Put @: after every vs., e.g., and i.e. where it is
5654         followed by whitespace.
5655         * doc/extend.texi: Ditto.
5656         * doc/fragments.texi: Ditto.
5657         * doc/gimple.texi: Ditto.
5658         * doc/implement-c.texi: Ditto.
5659         * doc/install.texi: Ditto.
5660         * doc/invoke.texi: Ditto.
5661         * doc/md.texi: Ditto.
5662         * doc/plugins.texi: Ditto.
5663         * doc/rtl.texi: Ditto.
5664         * doc/sourcebuild.texi: Ditto.
5665         * doc/tm.texi.in: Ditto.
5666         * doc/ux.texi: Ditto.
5667         * doc/tm.texi: Regenerate.
5669 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5671         * config/arm/arm-cpus.in (ares): New entry.
5672         * config/arm/arm-tables.opt: Regenerate.
5673         * config/arm/arm-tune.md: Likewise.
5674         * doc/invoke.texi (ARM Options): Document ares.
5676 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5678         * config/aarch64/aarch64-cores.def (ares): Define.
5679         * config/aarch64/aarch64-tune.md: Regenerate.
5680         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
5682 2018-11-07  Jan Hubicka  <jh@suse.cz>
5684         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
5685         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
5686         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
5687         TYPE_DECL.
5689 2018-11-07  Richard Biener  <rguenther@suse.de>
5691         PR tree-optimization/87914
5692         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
5693         of nested cycles.
5694         (vectorizable_reduction): Handle shifts and rotates by dispatching
5695         to vectorizable_shift.
5696         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
5697         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
5698         def cases.
5699         (vectorizable_shift): Export and handle being called as
5700         vect_nested_cycle.
5701         (vect_analyze_stmt): Call vectorizable_shift after
5702         vectorizable_reduction.
5703         * tree-vectorizer.h (vectorizable_shift): Declare.
5705 2018-11-07  Jan Hubicka  <jh@suse.cz>
5707         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
5708         than const decls in TREE_VALUE of enum.
5709         (dump_type_inheritance_graph): Improve duplicate dumping.
5710         (free_enum_values): New.
5711         (build_type_inheritance_graph): Use it.
5712         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
5713         which are not main variants or not ODR types.
5714         (verify_type_variant): Expect variants to have no TYPE_VALUES.
5716 2018-11-07  Richard Biener  <rguenther@suse.de>
5718         * ipa-inline.c (want_inline_small_function_p): Compute
5719         big_speedup_p lazily and last.
5721 2018-11-07  Jan Hubicka  <jh@suse.cz>
5723         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
5724         building incomplete variant of complete type.
5725         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
5726         variant of complete type.
5728 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5730         * config/mips/mips.c: Fix typo in documentation of
5731         mips_loongson_ext2_prefetch_cookie.
5732         (mips_option_override): fix brain twister logical.
5733         * config/mips/mips.h: Fix typo in documentation of
5734         ISA_HAS_CTZ_CTO and define pattern.
5735         * config/mips/mips.md (prefetch): Hoist EXT2 above
5736         the 2EF/EXT block.
5737         (prefetch_indexed): Hoist EXT2 above the EXT block.
5739 2018-11-07  Jan Hubicka  <jh@suse.cz>
5741         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
5742         return and parameter types of function and method types.
5743         (free_lang_data_in_cgraph): Update.
5745 2018-11-07  Martin Liska  <mliska@suse.cz>
5747         PR rtl-optimization/87868
5748         * postreload-gcse.c (eliminate_partially_redundant_load): Set
5749         threshold to max_count if we would overflow.
5750         * profile-count.h: Make max_count a public constant.
5752 2018-11-07  Martin Liska  <mliska@suse.cz>
5754         * mem-stats.h: Fix GNU coding style.
5756 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5758         * config/mips/gs264e.md: New.
5759         * config/mips/mips-cpus.def: Define gs264e.
5760         * config/mips/mips-tables.opt: Regenerate.
5761         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5762         gs264e.
5763         (mips_issue_rate): Add support for gs264e.
5764         (mips_multipass_dfa_lookahead): Likewise.
5765         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
5766         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
5767         (MIPS_ASE_MSA_SPEC): New.
5768         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
5769         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
5770         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
5771         * config/mips/mips.md: Include gs264e.md.
5772         (processor): Add gs264e.
5773         * config/mips/mips.opt (MSA): Use Mask instead of Var.
5774         * doc/invoke.texi: Add gs264e to supported architectures.
5776 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5778         * config/mips/gs464e.md: New.
5779         * config/mips/mips-cpus.def: Define gs464e.
5780         * config/mips/mips-tables.opt: Regenerate.
5781         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5782         gs464e.
5783         (mips_issue_rate): Add support for gs464e.
5784         (mips_multipass_dfa_lookahead): Likewise.
5785         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
5786         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
5787         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
5788         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
5789         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
5790         * config/mips/mips.md: Include gs464e.md.
5791         (processor): Add gs464e.
5792         * doc/invoke.texi: Add gs464e to supported architectures.
5794 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5796         * config/mips/loongson3a.md: Rename to ...
5797         * config/mips/gs464.md: ... here.
5798         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
5799         as an alias of gs464 processor.
5800         * config/mips/mips-tables.opt: Regenerate.
5801         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
5802         instead of PROCESSOR_LOONGSON_3A.
5803         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
5804         TUNE_LOONGSON_3A.
5805         (mips_option_override): Enable MMI and EXT for gs464.
5806         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
5807         Rename TUNE_LOONGSON_3A to TUNE_GS464.
5808         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
5809         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
5810         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
5811         TARGET_LOONGSON_3A.
5812         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
5813         (processor): Add gs464;
5814         * doc/invoke.texi: Add gs464 to supported architectures.
5816 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5818         * config/mips/mips-protos.h
5819         (mips_loongson_ext2_prefetch_cookie): New prototype.
5820         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
5821         (mips_option_override): Enable TARGET_LOONGSON_EXT when
5822         TARGET_LOONGSON_EXT2 is true.
5823         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
5824         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
5825         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
5826         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
5827         TARGET_LOONGSON_EXT2.
5828         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
5829         (define_insn "ctz<mode>2"): New insn pattern.
5830         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
5831         (define_insn "prefetch_indexed_<mode>"): Include
5832         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
5833         * config/mips/mips.opt (-mloongson-ext2): Add option.
5834         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
5836 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5838         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
5839         __mips_loongson_ext.
5840         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
5841         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
5842         -mloongson-ext.
5843         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
5844         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
5845         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
5846         instead of TARGET_LOONGSON_3A.
5847         * config/mips/mips.opt (-mloongson-ext): Add option.
5848         * gcc/doc/invoke.texi (-mloongson-ext): Document.
5850 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5852         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
5853         * config/mips/loongson.md: Move to ...
5854         * config/mips/loongson-mmi.md: here; Adjustment.
5855         * config/mips/loongson.h: Move to ...
5856         State as deprecated. Include loongson-mmiintrin.h for back
5857         compatibility and warning.
5858         * config/mips/loongson-mmiintrin.h: ... here.
5859         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
5860         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
5861         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5862         (mips_option_override): Make sure MMI use hard float;
5863         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
5864         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
5865         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
5866         TARGET_LOONGSON_VECTORS.
5867         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
5868         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
5869         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
5870         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
5871         -mloongson-mmi.
5872         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
5873         TARGET_LOONGSON_VECTORS.
5874         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
5875         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5876         (Loongson MMI patterns): Include loongson-mmi.md instead of
5877         loongson.md.
5878         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
5879         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
5881 2018-11-07  Richard Biener  <rguenther@suse.de>
5883         PR lto/87906
5884         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
5885         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
5887 2018-11-07  Alexandre Oliva  <aoliva@redhat.com>
5889         PR rtl-optimization/87874
5890         * lra.c (lra_substitute_pseudo): Do not create a subreg for
5891         const wide ints.
5893 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
5895         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
5896         if not in indexed or indirect form.
5897         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
5898         (bswap<mode>2_store): Ditto.
5900 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
5902         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
5903         the UNSPEC.
5905 2018-11-06  Richard Biener  <rguenther@suse.de>
5907         PR tree-optimization/86850
5908         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
5909         instead of src.m_vec.
5911 2018-11-06  Jan Hubicka  <jh@suse.cz>
5913         * tree.c (fld_simplified_type_name): Break out form ...
5914         (free_lang_data_in_type): ... here.
5915         (fld_type_variant_equal_p): Use it.
5917 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5919         * config/default-d.c: Include memmodel.h.
5921         * config/sol2-d.c: New file.
5922         * config/t-sol2 (sol2-d.o): New rule.
5923         * config.gcc <*-*-solaris2*>: Set d_target_objs,
5924         target_has_targetdm.
5926 2018-11-06  Jan Hubicka  <jh@suse.cz>
5928         * tree.c (fld_type_variant): Also copy alignment; be sure that
5929         new variant is equal.
5931 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5933         PR target/87762
5934         * config/s390/s390.md: Add relative_long attribute.
5936 2018-11-06  Jan Hubicka  <jh@suse.cz>
5938         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
5939         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
5940         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
5941         stream TYPE_NEEDS_CONSTRUCTING.
5942         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
5943         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
5945 2018-11-06  Richard Biener  <rguenther@suse.de>
5947         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
5948         dump-scope ...
5949         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
5951 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5953         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
5954         REG_CLASS_CONTENTS[GEN_REGS].
5955         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
5957 2018-11-06  Jan Hubicka  <jh@suse.cz>
5959          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
5961 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5963         PR sanitizer/80953
5964         * config/sol2.h (ASAN_CC1_SPEC): Define.
5965         (LD_WHOLE_ARCHIVE_OPTION): Define.
5966         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
5967         (ASAN_REJECT_SPEC): Provide default.
5968         (LIBASAN_EARLY_SPEC): Define.
5969         (LIBTSAN_EARLY_SPEC): Define.
5970         (LIBLSAN_EARLY_SPEC): Define.
5971         * config/i386/sol2.h (CC1_SPEC): Redefine.
5972         (ASAN_REJECT_SPEC): Define.
5974         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
5975         (TARGET_ASAN_SHADOW_OFFSET): Define.
5976         (sparc_asan_shadow_offset): New function.
5977         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
5978         (ASAN_REJECT_SPEC): Define.
5980 2018-11-06  Jan Hubicka  <jh@suse.cz>
5982         * tree.c (fld_type_variant): Copy canonical type.
5983         (fld_incomplete_type_of): Check that canonical types looks sane;
5984         copy canonical type.
5985         (verify_type): Accept when incomplete type has complete canonical type.
5987 2018-11-06  Jan Hubicka  <jh@suse.cz>
5989         * tree.c (free_lang_data): Reset overwite_assembler_name,
5990         print_xnode, print_decl, print_type and print_identifier of
5991         langhooks.
5993 2018-11-06  Richard Biener  <rguenther@suse.de>
5995         PR tree-optimization/87889
5996         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5997         Do nothing if old and new arg are the same
5999 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
6001         PR target/87723
6002         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
6003         attributes for operands 3 and 4.
6005 2018-11-06  Richard Biener  <rguenther@suse.de>
6007         PR middle-end/18041
6008         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
6009         matching bitfield insertion.
6011 2018-11-06  Alexandre Oliva  <aoliva@redhat.com>
6013         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
6014         (reg_next_debug_use): New.
6015         (attempt_change): Propagate adjusted expression into affected
6016         debug insns.
6017         (merge_in_block): Track uses in debug insns.
6018         (pass_inc_dec::execute): Allocate and release
6019         reg_next_debug_use.
6021 2018-11-06  Wei Xiao  <wei3.xiao@intel.com>
6023         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
6024         (_mm512_fixupimm_round_pd): Update parameters and builtin.
6025         (_mm512_maskz_fixupimm_round_pd): Ditto.
6026         (_mm512_fixupimm_round_ps): Ditto.
6027         (_mm512_maskz_fixupimm_round_ps): Ditto.
6028         (_mm_fixupimm_round_sd): Ditto.
6029         (_mm_maskz_fixupimm_round_sd): Ditto.
6030         (_mm_fixupimm_round_ss): Ditto.
6031         (_mm_maskz_fixupimm_round_ss): Ditto.
6032         (_mm512_fixupimm_pd): Ditto.
6033         (_mm512_maskz_fixupimm_pd): Ditto.
6034         (_mm512_fixupimm_ps): Ditto.
6035         (_mm512_maskz_fixupimm_ps): Ditto.
6036         (_mm_fixupimm_sd): Ditto.
6037         (_mm_maskz_fixupimm_sd): Ditto.
6038         (_mm_fixupimm_ss): Ditto.
6039         (_mm_maskz_fixupimm_ss): Ditto.
6040         (_mm512_mask_fixupimm_round_pd): Update builtin.
6041         (_mm512_mask_fixupimm_round_ps): Ditto.
6042         (_mm_mask_fixupimm_round_sd): Ditto.
6043         (_mm_mask_fixupimm_round_ss): Ditto.
6044         (_mm512_mask_fixupimm_pd): Ditto.
6045         (_mm512_mask_fixupimm_ps): Ditto.
6046         (_mm_mask_fixupimm_sd): Ditto.
6047         (_mm_mask_fixupimm_ss): Ditto.
6048         * config/i386/avx512vlintrin.h:
6049         (_mm256_fixupimm_pd): Update parameters and builtin.
6050         (_mm256_maskz_fixupimm_pd): Ditto.
6051         (_mm256_fixupimm_ps): Ditto.
6052         (_mm256_maskz_fixupimm_ps): Ditto.
6053         (_mm_fixupimm_pd): Ditto.
6054         (_mm_maskz_fixupimm_pd): Ditto.
6055         (_mm_fixupimm_ps): Ditto.
6056         (_mm_maskz_fixupimm_ps): Ditto.
6057         (_mm256_mask_fixupimm_pd): Update builtin.
6058         (_mm256_mask_fixupimm_ps): Ditto.
6059         (_mm_mask_fixupimm_pd): Ditto.
6060         (_mm_mask_fixupimm_ps): Ditto.
6061         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
6062         * config/i386/i386-builtin.def: Update builtin definitions.
6063         * config/i386/i386.c: Handle new builtin types and remove useless ones.
6064         * config/i386/sse.md: Update VFIXUPIMM* patterns.
6065         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
6066         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
6067         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
6068         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
6069         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
6070         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
6071         * config/i386/subst.md:
6072         (round_saeonly_sd_mask_operand4): Add new subst_attr.
6073         (round_saeonly_sd_mask_op4): Ditto.
6074         (round_saeonly_expand_operand5): Ditto.
6075         (round_saeonly_expand): Update.
6077 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
6079         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
6081 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
6083         PR rtl-optimization/87871
6084         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
6086 2018-11-05  Paul Koning  <ni1d@arrl.net>
6088         * doc/sourcebuild.texi (target attributes): Document new "inf"
6089         effective target keyword.
6091 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
6093         * config/s390/s390.c (s390_register_move_cost): Increase costs for
6094         moves involving the CC reg.
6096 2018-11-05  Richard Biener  <rguenther@suse.de>
6098         PR tree-optimization/87873
6099         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
6100         argument.
6101         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
6102         * tree-vect-loop.c (vect_transform_loop): When splitting the
6103         loop exit also create forwarder PHIs for constants.
6104         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
6105         Handle constant to_arg, add extra checking we match up the correct
6106         PHIs.
6108 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
6110         * config/s390/s390.md: QImode and HImode for load on condition.
6112 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
6114         * config/s390/predicates.md: Fix typo.
6115         * config/s390/s390.md: Allow immediates for load on condition.
6117 2018-11-05  Martin Liska  <mliska@suse.cz>
6119         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
6120         * bitmap.h (struct bitmap_usage): Likewise.
6121         * ggc-common.c (SCALE): Remove.
6122         (LABEL): Likewise.
6123         (struct ggc_usage): Use SIZE_AMOUNT. And update
6124         compare method.
6125         * ggc-page.c (SCALE): Remove.
6126         (STAT_LABEL): Remove.
6127         (ggc_print_statistics): Use SIZE_AMOUNT.
6128         * gimple.h (SCALE): Remove.
6129         (LABEL): Likewise.
6130         * input.c (ONE_K): Remove.
6131         (ONE_M): Likewise.
6132         (SCALE): Likewise.
6133         (STAT_LABEL): Likewise.
6134         (FORMAT_AMOUNT): Likewise.
6135         (dump_line_table_statistics): Use SIZE_AMOUNT.
6136         * mem-stats.h (struct mem_usage): Likewise.
6137         * rtl.c (dump_rtx_statistics): Likewise.
6138         (rtx_alloc_counts): Change type to size_t.
6139         (rtx_alloc_sizes): Likewise.
6140         (rtx_count_cmp): New.
6141         (dump_rtx_statistics): Sort first based on counts.
6142         * tree.c (tree_nodes_cmp): New.
6143         (tree_codes_cmp): New.
6144         (dump_tree_statistics): Sort first based on counts.
6145         * system.h (ONE_K): New.
6146         (ONE_M): Likewise.
6147         (SIZE_SCALE): Likewise.
6148         (SIZE_LABEL): Likewise.
6149         (SIZE_AMOUNT): Likewise.
6150         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
6151         * tree-dfa.c (dump_dfa_stats): Likewise.
6152         * tree-phinodes.c (phinodes_print_statistics): Likewise.
6153         * tree-ssanames.c (ssanames_print_statistics): Likewise.
6154         * tree.c (dump_tree_statistics): Likewise.
6155         * vec.c (struct vec_usage): Likewise.
6156         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
6157         trigger a -Werror=format-overflow with
6158         --enable-gather-detailed-stats.
6160 2018-11-05  Martin Liska  <mliska@suse.cz>
6162         * mem-stats.h (mem_alloc_description::release_instance_overhead):
6163         Return T *.
6164         * vec.c (struct vec_usage): Register m_element_size.
6165         (vec_prefix::register_overhead): New arguments: elements and
6166         element_size.
6167         (vec_prefix::release_overhead): Subtract elements.
6168         * vec.h (struct vec_prefix): Change signature.
6169         (va_heap::reserve): Pass proper arguments.
6170         (va_heap::release): Likewise.
6172 2018-11-05  Martin Liska  <mliska@suse.cz>
6174         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
6175         style.
6176         * vec.c: Likewise.
6178 2018-11-05  Richard Biener  <rguenther@suse.de>
6180         * tree-scalar-evolution.h (final_value_replacement_loop): Update
6181         prototype.
6182         * tree-scalar-evolution.c (final_value_replacement_loop): Return
6183         whether anything was done.
6184         (scev_const_prop): Remove constant propagation part, fold
6185         remains into ...
6186         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
6187         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
6188         conditionally.
6190 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
6192         PR tree-optimization/87859
6193         * gimple-ssa-store-merging.c (struct merged_store_group): Add
6194         only_constants and first_nonmergeable_order members.
6195         (merged_store_group::merged_store_group): Initialize them.
6196         (merged_store_group::do_merge): Clear only_constants member if
6197         adding something other than INTEGER_CST store.
6198         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
6199         stores with order >= first_nonmergeable_order.  Use
6200         merged_store->only_constants instead of always recomputing it.
6201         Set merged_store->first_nonmergeable_order if we've skipped any
6202         stores.  Attempt to merge overlapping INTEGER_CST stores that
6203         we would otherwise skip.
6205         PR sanitizer/87837
6206         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
6208 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
6210         PR target/87853
6211         * config/i386/emmintrin.h (__v16qs): New to cope with option
6212         -funsigned-char.
6213         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
6214         (_mm_cmplt_epi8): Likewise.
6215         (_mm_cmpgt_epi8): Likewise.
6217 2018-11-05  Richard Biener  <rguenther@suse.de>
6219         PR rtl-optimization/87852
6220         * fwprop.c (use_killed_between): Only consider single-defs of the
6221         use whose definition statement dominates the use.
6223 2018-11-05  Martin Liska  <mliska@suse.cz>
6225         PR web/87829
6226         * doc/invoke.texi: Remove options that are
6227         not disabled with -Os.
6229 2018-11-05  Martin Liska  <mliska@suse.cz>
6231         PR c/87811
6232         * doc/extend.texi: Update constrain about the last argument
6233         of __builtin_expect_with_probability.
6235 2018-11-05  Martin Liska  <mliska@suse.cz>
6237         PR c/87811
6238         * predict.c (expr_expected_value_1): Verify
6239         that last argument is a real constants and emit
6240         error.
6242 2018-11-05  Martin Liska  <mliska@suse.cz>
6244         PR gcov-profile/77698
6245         * ipa-profile.c (ipa_profile): Adjust hotness threshold
6246         only in LTO mode.
6248 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6250         PR tree-optimization/86572
6251         * builtins.c (c_strlen): Handle negative offsets in a safe way.
6253 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6255         PR tree-optimization/87672
6256         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
6257         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
6259 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
6261         PR middle-end/58372
6262         * cfgexpand.c (pass_expand::execute): Move the call to
6263         finish_eh_generation in front of the call to expand_stack_alignment.
6265 2018-11-04  Venkataramanan Kumar  <venkataramanan.kumar@amd.com>
6267         * common/config/i386/i386-common.c (processor_alias_table): Add
6268         znver2 entry.
6269         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
6270         (case ${target}): Add znver2.
6271         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
6272         -march=native recognize znver2 processors.
6273         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
6274         * config/i386/i386.c (m_znver2): New definition.
6275         (m_ZNVER): New definition.
6276         (m_AMD_MULTIPLE): Includes m_znver2.
6277         (processor_cost_table): Add znver2 entry.
6278         (processor_target_table): Add znver2 entry.
6279         (get_builtin_code_for_version): Set priority for
6280         PROCESSOR_ZNVER2.
6281         (processor_model): Add M_AMDFAM17H_ZNVER2.
6282         (arch_names_table): Ditto.
6283         (ix86_reassociation_width): Include znver2.
6284         * config/i386/i386.h (TARGET_znver2): New definition.
6285         (struct ix86_size_cost): Add TARGET_ZNVER2.
6286         (enum processor_type): Add PROCESSOR_ZNVER2.
6287         * config/i386/i386.md (define_attr "cpu"): Add znver2.
6288         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
6289         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
6290         (ix86_adjust_cost): Add znver2.
6291         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
6292         * gcc/doc/extend.texi: Add details about znver2.
6293         * gcc/doc/invoke.texi: Add details about znver2.
6295 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
6297         PR target/87079
6299         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
6300         pattern.
6302 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6304         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
6305         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
6307 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
6309         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
6310         shrink-wrapping when -mtrack-speculation.
6312 2018-11-02  Richard Biener  <rguenther@suse.de>
6314         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
6315         (pop_cost_one_pair): Do not free pair.
6316         (pop_best_coalesce): Likewise.
6317         (create_coalesce_list): Initialize obstack.
6318         (delete_coalesce_list): Free obstack.
6319         (find_coalesce_pair): Obstack-allocate coalesce pairs.
6320         (add_cost_one_coalesce): Likewise.
6321         (struct live_track): Remove bitmap pointer indirections.
6322         (new_live_track): Adjust.
6323         (delete_live_track): Likewise.
6324         (live_track_remove_partition): Likewise.
6325         (live_track_add_partition): Likewise.
6326         (live_track_live_p): Likewise.
6327         (live_track_process_def): Likewise.
6328         (live_track_clear_base_vars): Likewise.
6330 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6332         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
6333         * configure: Regenerate.
6334         * config.in: Regenerate.
6335         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
6336         HAVE_AS_REGISTER_PSEUDO_OP guard.
6337         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
6338         !HAVE_AS_REGISTER_PSEUDO_OP support.
6339         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
6340         guard.
6342 2018-11-02  Richard Biener  <rguenther@suse.de>
6344         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
6345         with is_type_die.
6347 2018-11-02  Richard Biener  <rguenther@suse.de>
6349         PR tree-optimization/87776
6350         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
6351         executable when iterating but running into rpo-vn-max-loop-depth
6352         and not eliding the iteration.
6354 2018-11-30  Jan Hubicka  <jh@suse.cz>
6356         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
6357         TYPE_DECL.
6359 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
6361         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
6362         prototype.
6363         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
6364         Combine with rs6000_address_for_fpconvert.
6365         (rs6000_address_for_fpconvert): Combine with
6366         rs6000_force_indexed_or_indirect_mem.
6367         (rs6000_expand_vector_init): Change function call from
6368         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6369         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
6370         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6371         (floatsi<mode>2_lfiwax_mem): Ditto.
6372         (floatunssi<mode>2_lfiwzx): Ditto.
6373         (floatunssi<mode>2_lfiwzx_mem): Ditto.
6374         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
6375         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
6376         (fix_trunc<mode>si2_stfiwx): Ditto.
6377         (fixuns_trunc<mode>si2_stfiwx): Ditto.
6378         (float_<mode>si2_hw): Ditto.
6379         (floatuns_<mode>si2_hw): Ditto.
6380         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
6381         (vsx_splat_<mode>): Ditto.
6383 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
6385         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
6386         at end of assembler input text.
6387         * configure: Regenerate.
6389 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
6391         PR tree-optimization/87826
6392         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
6393         negative or larger or equal to type's precision.
6395 2018-10-31  Alexandre Oliva  <aoliva@redhat.com>
6397         * opts.c (default_options_table): Do not enable
6398         OPT_fdelayed_branch at -Og.
6399         * doc/invoke.texi (-fdelayed-branch): Document it.
6401 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
6403         * optabs-libfuncs.c (build_libfunc_function_visibility):
6404         New, split out from...
6405         (build_libfunc_function): ... here.
6406         (init_one_libfunc_visibility): New, split out from ...
6407         (init_one_libfunc): ... here.
6409         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
6410         scratch register need not be early-clobber.  Document the reason
6411         why we cannot use ST<OP>.
6413 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
6415         PR bootstrap/82856
6416         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
6417         line for second argument of AC_DEFINE_UNQUOTED.
6418         * doc/install.texi (Tools/packages necessary for modifying GCC):
6419         Update to autoconf 2.69 and automake 1.15.1.
6420         * aclocal.m4, config.in, configure: Regenerate.
6422 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
6424         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
6425         initialization.
6427 2018-10-31  Martin Liska  <mliska@suse.cz>
6429         PR driver/83193
6430         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
6431         Add new argument invalid_extension.
6432         (aarch64_get_all_extension_candidates): New function.
6433         (aarch64_rewrite_selected_cpu): Add NULL to function call.
6434         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
6435         new argument.
6436         (aarch64_get_all_extension_candidates): New function.
6437         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
6438         argument invalid_extension.
6439         (aarch64_parse_cpu): Likewise.
6440         (aarch64_print_hint_for_extensions): New function.
6441         (aarch64_validate_mcpu): Provide hint about invalid extension.
6442         (aarch64_validate_march): Likewise.
6443         (aarch64_handle_attr_arch): Pass new argument.
6444         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
6445         (aarch64_handle_attr_isa_flags): Likewise.
6447 2018-10-31  Richard Biener  <rguenther@suse.de>
6449         PR middle-end/70359
6450         PR middle-end/86270
6451         * tree-outof-ssa.c (insert_backedge_copies): Restrict
6452         copy generation to useful cases.  Place the copy before
6453         the definition of the backedge value when possible.
6455 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
6457         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
6458         * config/arc/arc.c (arc_active_insn): New function.
6459         (check_store_cacheline_hazard): Likewise.
6460         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
6461         (arc_override_options): Disable delay slot scheduler for older
6462         A7.
6463         (arc_store_addr_hazard_p): New implementation, old one renamed to
6464         ...
6465         (arc_store_addr_hazard_internal_p): Renamed.
6466         (arc_reorg): Don't combine into brcc instructions which are part
6467         of hardware hazard solution.
6468         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
6469         (tune_arc700): Likewise.
6470         * config/arc/arc.opt (arc7xx): New tune value.
6471         * config/arc/arc700.md: Improve A7 scheduler.
6473 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
6475         * config/arc/arc.c (arc_override_options): Remove
6476         TARGET_COMPACT_CASESI.
6477         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
6478         (CASE_VECTOR_MODE): Likewise.
6479         (CASE_VECTOR_PC_RELATIVE): Likewise.
6480         (CASE_VECTOR_SHORTEN_MODE): Likewise.
6481         (CASE_VECTOR_SHORTEN_MODE1): Delete.
6482         (ADDR_VEC_ALIGN): Update.
6483         (ASM_OUTPUT_CASE_LABEL): Undefine.
6484         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
6485         (TARGET_BI_BIH): Define.
6486         (DEFAULT_BRANCH_INDEX): Likewise.
6487         * config/arc/arc.md (casesi): Rework to accept BI/BIH
6488         instructions, remove compact_casesi use case.
6489         (casesi_compact_jump): Remove.
6490         (casesi_dispatch): New pattern.
6491         * config/arc/arc.opt: Add mbranch-index option. Deprecate
6492         compact_casesi option.
6493         * doc/invoke.texi: Document mbranch-index option.
6495 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
6497         * config/arc/arc.c (arc_get_tp): Remove function.
6498         (arc_emit_call_tls_get_addr): Likewise.
6499         (arc_call_tls_get_addr): New function.
6500         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
6501         * config/arc/arc.md (tls_load_tp_soft): Remove.
6502         (tls_gd_get_addr): Likewise.
6504 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
6506         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
6507         (mulsi3_600_lib): Remove pattern.
6508         (umulsi3_highpart_600_lib_le): Likewise.
6509         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
6510         (umulsidi3): Remove call to umulsidi3_600_lib.
6511         (umulsidi3_600_lib): Remove pattern.
6512         (peephole2): Remove peephole using the above deprecated patterns.
6514 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6516         PR target/87374
6517         * config/arm/arm.c (arm_option_check_internal): Disable the combined
6518         use of -mslow-flash-data and -mword-relocations.
6519         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
6520         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
6521         flag_pic.
6522         * doc/invoke.texi (-mword-relocations): Mention conflict with
6523         -mslow-flash-data.
6524         (-mslow-flash-data): Reciprocally.
6526 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
6528         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
6529         16-byte modes held in GP registers to use an even regno.
6531         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
6532         (aarch64_atomic_ldop_supported_p): Remove.
6533         (aarch64_gen_atomic_ldop): Remove.
6534         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
6535         Fully expand LSE operations here.
6536         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
6537         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
6538         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
6539         and use ATOMIC_LDOP instead; use register_operand for the input;
6540         drop the split and emit insns directly.
6541         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
6542         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
6543         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
6545         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
6546         (aarch64_gen_atomic_ldop): Don't call it.
6547         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
6548         Use aarch64_reg_or_zero.
6549         (aarch64_atomic_exchange<ALLI>): Likewise.
6550         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
6551         operand 0; use aarch64_reg_or_zero for input; merge ...
6552         (@aarch64_atomic_swp<ALLI>): ... this and remove.
6554         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
6555         (aarch64_split_compare_and_swap): Use it.
6556         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
6557         test oldval against the proper predicate.
6558         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
6559         Use nonmemory_operand for expected.
6560         (cas_short_expected_pred): New.
6561         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
6562         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
6563         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
6564         (aarch64_plushi_operand): New.
6566         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
6567         Force oldval into the rval register for TARGET_LSE; emit the compare
6568         during initial expansion so that it may be deleted if unused.
6569         (aarch64_gen_atomic_cas): Remove.
6570         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
6571         Change =&r to +r for operand 0; use match_dup for operand 2;
6572         remove is_weak and mod_f operands as unused.  Drop the split
6573         and merge with...
6574         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
6575         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
6576         (@aarch64_atomic_cas<GPI>): Similarly.
6578 2018-10-31  Richard Biener  <rguenther@suse.de>
6580         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
6581         using ABSU_EXPR.
6583 2018-10-31  Sameera Deshpande  <sameera.deshpande@linaro.org>
6585         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
6586         * config/aarch64/aarch64.md: Include saphira.md
6587         * config/aarch64/saphira.md: New file for pipeline description.
6589 2018-10-30  Martin Sebor  <msebor@redhat.com>
6591         PR middle-end/87041
6592         * gimple-ssa-sprintf.c (format_directive): Use %G to include
6593         inlining context.
6594         (sprintf_dom_walker::compute_format_length):
6595         Avoid setting POSUNDER4K here.
6596         (get_destination_size): Handle null argument values.
6597         (get_user_idx_format): New function.
6598         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
6599         functions, including user-defined with attribute format printf.
6600         Use %G to include inlining context.
6601         Set POSUNDER4K here.
6603 2018-10-30  Jan Hubicka  <jh@suse.cz>
6605         * params.def (lto-partitions): Bump from 32 to 128.
6607 2018-10-30  Jan Hubicka  <jh@suse.cz>
6609         * tree.c
6610         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
6611         head in file.
6612         (free_lang_data_in_type): Forward declare.
6613         (fld_type_variant_equal_p): New function.
6614         (fld_type_variant): New function
6615         (fld_incomplete_types): New hash.
6616         (fld_incomplete_type_of): New function
6617         (fld_simplfied-type): New function.
6618         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
6619         (free_lang_data): Allocate and free fld_incomplete_type; update call
6620         of free_lang_data_in_decl.
6622 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
6624         * gcov.c (output_lines): Remove duplicate line.
6626 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
6628         * config/rs6000/rs6000.md (bswapdi2): Force address into register
6629         if not in indexed or indirect form.
6630         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
6631         (bswapdi2_store): Ditto.
6632         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
6633         helper function.
6634         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
6635         Prototype for helper function.
6637 2018-10-30  Martin Sebor  <msebor@redhat.com>
6639         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
6640         (target, pragma GCC optimize, pragma GCC target): Ditto.
6642 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
6644         * doc/extend.texi: Fix prototype and description of
6645         __builtin_expect_with_probability.
6647 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
6649         * cgraph.h (clone_function_name_1): Replaced by new
6650           clone_function_name_numbered that takes name as string; for
6651           privatize_symbol_name_1 use only.
6652           (clone_function_name): Renamed to
6653           clone_function_name_numbered to be explicit about numbering.
6654           (clone_function_name): New two-argument function that does
6655           not number its output.
6656           (clone_function_name): New three-argument function that
6657           takes a number to append to its output.
6658         * cgraphclones.c (duplicate_thunk_for_node):
6659           (clone_function_name_1): Renamed.
6660           (clone_function_name_numbered): Two new functions.
6661           (clone_function_name): Improved documentation.
6662           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
6663         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
6664         * final.c (final_scan_insn_1): Use the new clone_function_name
6665           without numbering.
6666         * multiple_target.c (create_dispatcher_calls): Ditto.
6667           (create_target_clone): Ditto.
6668         * omp-expand.c (grid_expand_target_grid_body): Ditto.
6669         * omp-low.c (create_omp_child_function_name): Ditto.
6670         * omp-simd-clone.c (simd_clone_create): Ditto.
6671         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
6672           new clone_function_name without numbering.
6674 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
6676         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
6677         Assert that the allocation size is not zero.
6679 2018-10-30  Richard Biener  <rguenther@suse.de>
6681         PR tree-optimization/87800
6682         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
6683         non-induction or reduction PHIs.
6685 2018-10-30  Sameera Deshpande  <sameera.deshpande@linaro.org>
6687         * config/aarch64/falkor-tag-collision-avoidance.c
6688         (execute_tag_collision_avoidance): Call df_note_add_problem.
6690 2018-10-30  Martin Liska  <mliska@suse.cz>
6692         * doc/extend.texi: Fix typo in documentation
6693         of __builtin_expect_with_probability.
6695 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
6697         PR c++/87721
6698         * input.c (get_substring_ranges_for_loc): Detect if
6699         linemap_resolve_location gives us a NULL map, and reject
6700         this case.
6702 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
6704         * config.gcc (xstormy16-*-elf): Set tm_d_file.
6706 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
6707             Martin Sebor  <msebor@redhat.com>
6708             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
6710         * doc/ux.texi (Quoting): New subsection, adapted from material at
6711         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
6712         MartinSebor and ManuelLopezIbanez.
6713         (Fix-it hints): Note that fix-it hints shouldn't be marked for
6714         translation.
6716 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
6718         PR middle-end/87469
6719         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
6720         max value.
6722 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
6724         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
6726 2018-10-29  Paul Koning  <ni1d@arrl.net>
6728         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
6730 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
6732         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
6733         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
6734         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
6735         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
6736         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
6737         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
6738         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
6739         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
6740         Change 'vector' to '__vector'.
6741         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
6742         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
6743         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
6744         _mm_avg_pu16): Likewise.  And, whitespace corrections.
6746 2018-10-29  Richard Biener  <rguenther@suse.de>
6748         PR tree-optimization/87785
6749         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
6750         internal defs.
6752 2018-10-29  Olivier Hainque  <hainque@adacore.com>
6754         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
6756 2018-10-29  Olivier Hainque  <hainque@adacore.com>
6758         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
6759         ports configuration macro, defaults to "gnu".
6760         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
6761         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
6763 2018-10-29  Olivier Hainque  <hainque@adacore.com>
6765         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
6767 2018-10-29  Richard Biener  <rguenther@suse.de>
6769         PR tree-optimization/87790
6770         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
6771         (vect_make_slp_decision): Adjust.
6772         (vect_slp_analyze_bb_1): Likewise.
6773         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
6774         edges.
6776 2018-10-29  Richard Biener  <rguenther@suse.de>
6778         PR tree-optimization/87785
6779         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
6780         and processing.
6781         (vect_build_slp_tree): Likewise.
6782         (vect_gather_slp_loads): New function.
6783         (vect_analyze_slp_instance): Gather loads separately from the
6784         SLP tree build.
6786 2018-10-29  Martin Liska  <mliska@suse.cz>
6788         * Makefile.in: Make dependency to json.o.
6789         * doc/gcov.texi: Document new JSON format, remove
6790         old intermediate format documentation.
6791         * gcov.c (struct function_info): Come up with m_name and
6792         m_demangled_name.
6793         (function_info::function_info): Initialize it.
6794         (function_info::~function_info): Release it.
6795         (main): Rename flag_intermediate_format to flag_json_format.
6796         (print_usage): Describe --json-format.
6797         (process_args): Set flag_json_format.
6798         (output_intermediate_line): Remove.
6799         (output_intermediate_json_line): Likewise.
6800         (get_gcov_intermediate_filename): Return new extension
6801         ".gcov.json.gz".
6802         (output_intermediate_file): Implement JSON emission.
6803         (output_json_intermediate_file): Implement JSON emission.
6804         (generate_results): Use ::get_name for function name.
6805         Handle JSON output file.
6806         (read_graph_file): Use ::get_name instead of cplus_demangle.
6807         (read_count_file): Likewise.
6808         (solve_flow_graph): Likewise.
6809         (add_line_counts): Likewise.
6810         (accumulate_line_counts): Use new flag_json_format.
6811         (output_function_details): Use ::get_name instead of cplus_demangle.
6812         (output_lines): Likewise.
6813         * json.cc (test_writing_literals): Add new tests.
6814         * json.h (class literal): Add new boolean constructor.
6816 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
6818         PR rtl-optimization/87701
6819         PR rtl-optimization/87780
6820         * combine.c (make_more_copies): Rewrite.
6822 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
6824         * doc/generic.texi (ABSU_EXPR): Document.
6825         * match.pd (absu(x)*absu(x) -> x*x): Handle.
6826         (absu(absu(X)) -> absu(X)): Likewise.
6827         (absu(-X) -> absu(X)): Likewise.
6828         (absu(X)  where X is nonnegative -> X): Likewise.
6830 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
6832         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
6833         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
6834         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
6835         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
6836         (s-tm-texi): Also check timestamp on d-target.def.
6837         (generated_files): Add TM_D_H and d-target-hooks-def.h.
6838         (build/genhooks.o): Also depend on D_TARGET_DEF.
6839         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
6840         variables.
6841         * config/aarch64/aarch64-d.c: New file.
6842         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
6843         Define.
6844         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
6845         prototype.
6846         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
6847         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
6848         * config/arm/arm-d.c: New file.
6849         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
6850         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
6851         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6852         * config/arm/t-arm (arm-d.o): New rule.
6853         * config/default-d.c: New file.
6854         * config/glibc-d.c: New file.
6855         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6856         * config/i386/i386-d.c: New file.
6857         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
6858         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
6859         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6860         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
6861         * config/i386/t-i386 (i386-d.o): New rule.
6862         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6863         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6864         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
6865         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6866         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6867         * config/mips/mips-d.c: New file.
6868         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
6869         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
6870         * config/mips/t-mips (mips-d.o): New rule.
6871         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6872         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6873         * config/powerpcspe/powerpcspe-d.c: New file.
6874         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
6875         New prototype.
6876         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6877         Support GNU D by using 0 as the language type.
6878         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
6879         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
6880         * config/riscv/riscv-d.c: New file.
6881         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
6882         prototype.
6883         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
6884         * config/riscv/t-riscv (riscv-d.o): New rule.
6885         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6886         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6887         * config/rs6000/rs6000-d.c: New file.
6888         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
6889         prototype.
6890         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6891         Support GNU D by using 0 as the language type.
6892         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
6893         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
6894         * config/s390/s390-d.c: New file.
6895         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
6896         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
6897         * config/s390/t-s390 (s390-d.o): New rule.
6898         * config/sparc/sparc-d.c: New file.
6899         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
6900         prototype.
6901         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
6902         * config/sparc/t-sparc (sparc-d.o): New rule.
6903         * config/t-glibc (glibc-d.o): New rule.
6904         * configure: Regenerated.
6905         * configure.ac (tm_d_file): New variable.
6906         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
6907         * doc/contrib.texi (Contributors): Add self for the D frontend.
6908         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
6909         * doc/install.texi (Configuration): Mention libphobos as an option for
6910         --enable-shared.  Mention d as an option for --enable-languages.
6911         (Testing): Mention check-d as a target.
6912         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
6913         name suffixes.  Mention d as a -x option.
6914         * doc/sourcebuild.texi (Top Level): Mention libphobos.
6915         * doc/standards.texi (Standards): Add section on D language.
6916         * doc/tm.texi: Regenerated.
6917         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
6918         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
6919         * dwarf2out.c (is_dlang): New function.
6920         (gen_compile_unit_die): Use DW_LANG_D for D.
6921         (declare_in_namespace): Return module die for D, instead of adding
6922         extra declarations into the namespace.
6923         (gen_namespace_die): Generate DW_TAG_module for D.
6924         (gen_decl_die): Handle CONST_DECLSs for D.
6925         (dwarf2out_decl): Likewise.
6926         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
6927         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
6928         kinds of aggregates.
6929         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
6930         * genhooks.c: Include d/d-target.def.
6932 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
6934         PR target/85669
6935         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
6936         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
6937         (STACK_DYNAMIC_OFFSET): Likewise.
6939 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
6941         PR target/80024
6942         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
6943         error message.
6945 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
6947         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
6948         return type and other typos.
6950 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
6952         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
6953         a shorter sequence with fewer branches.
6954         (emit_final_str_compare_gpr): Ditto.
6956 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6958         * config/rs6000/tmmintrin.h: New file.
6959         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
6961 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6963         * config/rs6000/mmintrin.h: Enable 32bit compilation.
6964         * config/rs6000/xmmintrin.h: Likewise.
6966 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6968         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
6970 2018-10-26  Richard Biener  <rguenther@suse.de>
6972         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
6973         and wrapper.
6974         (vect_mark_slp_stmts_relevant): Likewise.
6975         (vect_detect_hybrid_slp_stmts): Likewise.
6976         (vect_bb_slp_scalar_cost): Likewise.
6977         (vect_remove_slp_scalar_calls): Likewise.
6979 2018-10-26  Jan Hubicka  <jh@suse.cz>
6981         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
6982         (warn_types_mismatch): Fix walk of DECL_NAME.
6983         (odr_types_equivalent_p): Fix overactive assert.
6985 2018-10-26  Richard Biener  <rguenther@suse.de>
6987         PR tree-optimization/87105
6988         * tree-vectorizer.h (_slp_tree::refcnt): New member.
6989         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
6990         refcnt.
6991         (vect_create_new_slp_node): Initialize refcnt to one.
6992         (bst_traits): Move.
6993         (scalar_stmts_set_t, bst_fail): Remove.
6994         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
6995         (vect_build_slp_tree): Add bst_map argument and lookup
6996         already created SLP nodes.
6997         (vect_print_slp_tree): Handle a SLP graph, print SLP node
6998         addresses.
6999         (vect_slp_rearrange_stmts): Handle a SLP graph.
7000         (vect_analyze_slp_instance): Adjust and free SLP nodes from
7001         the CSE map.  Fix indenting.
7002         (vect_schedule_slp_instance): Add short-cut.
7004 2018-10-26  Martin Liska  <mliska@suse.cz>
7006         PR testsuite/86158
7007         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
7008         addr_expr and not with pointers.
7010 2018-10-26  Jan Hubicka  <jh@suse.cz>
7012         * tree.c (free_lang_data_in_type): Only check main variants.
7013         * ipa-devirt.c (warn_odr): Make static.
7014         (types_same_for_odr): Drop strict variant.
7015         (types_odr_comparable): Likewise.
7016         (odr_or_derived_type_p): Look for main variants.
7017         (odr_name_hasher::equal): Cleanup comment.
7018         (odr_subtypes_equivalent): Add warn and warned arguments; check main
7019         variants.
7020         (type_variants_equivalent_p): break out from ...
7021         (odr_types_equivalent): ... here; go for main variants where needed.
7022         (warn_odr): ... here; turn static.
7023         (warn_types_mismatch): Compare mangled names of main variants.
7024         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
7025         (type_with_linkage_p): Sanity check that we look at main variant.
7026         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
7027         * tree.h (types_same_for_odr): Drop strict argument.
7029 2018-10-26  Richard Biener  <rguenther@suse.de>
7031         PR tree-optimization/87746
7032         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
7033         Simplify and fix WRT strided store groups with size not
7034         equal to step in element count.
7035         (vect_analyze_group_access_1): Dump the whole group.
7037 2018-10-25  Carl Love  <cel@us.ibm.com>
7039         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
7040         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
7041         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
7042         precicion entry for each overloaded builtin.
7043         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
7044         VSCEDPUO): Rename overloaded name.
7045         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
7046         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
7047         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
7048         define_expand for xscmpexqp instruction.
7049         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
7051 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
7052             Jinsong Ji  <jji@us.ibm.com>
7054         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
7055         function with vec_sl.
7056         (_mm_slli_epi32): Likewise.
7057         (_mm_slli_epi64): Likewise.
7058         (_mm_srai_epi16): Replace deprecated function with vec_sra.
7059         (_mm_srai_epi32): Likewise.
7060         (_mm_srli_epi16): Replace deprecated function with vec_sr.
7061         (_mm_srli_epi32): Likewise.
7062         (_mm_srli_epi64): Likewise.
7063         (_mm_sll_epi16): Replace deprecated function with vec_sl.
7064         (_mm_sll_epi32): Likewise.
7065         (_mm_sll_epi64): Likewise.
7066         (_mm_sra_epi16): Replace deprecated function with vec_sra.
7067         (_mm_sra_epi32): Likewise.
7068         (_mm_srl_epi16): Replace deprecated function with vec_sr.
7069         (_mm_srl_epi32): Likewise.
7070         (_mm_srl_epi64): Likewise.
7072 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
7073             Jinsong Ji  <jji@us.ibm.com>
7075         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
7076         comparison operators with vec_cmp* for compatibility due to
7077         unfortunate history; clean up formatting and use types more
7078         appropriately.
7079         (_mm_sll_epi32): Likewise.
7080         (_mm_sll_epi64): Likewise.
7081         (_mm_srl_epi16): Likewise.
7082         (_mm_srl_epi32): Likewise.
7083         (_mm_srl_epi64): Likewise.
7085 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
7086             Jinsong Ji  <jji@us.ibm.com>
7088         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
7089         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
7090         __vector __bool int.  Use vec_cmpgt in preference to deprecated
7091         function vec_vcmpgtfp.
7092         (_mm_max_ps): Likewise.
7094 2018-10-25  Jeff Law  <law@redhat.com>
7096         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
7097         if returning false.
7099 2018-10-25  Martin Sebor  <msebor@redhat.com>
7101         * doc/extend.texi (aligned): Expand attribute description.
7102         (Alignment): Rename section.  Discuss function arguments.
7104 2018-10-25  Jan Hubicka  <jh@suse.cz>
7106         * ipa-devirt.c (main_odr_variant): Remove.
7107         (hash_odr_name, types_same_for_odr, types_odr_comparable,
7108         odr_name_hasher::equal, odr_subtypes_equivalent_p):
7109         Drop use of main_odr_variant.
7110         (add_type_duplicate): Silence confused warnings on integer types.
7111         (get_odr_type): Always look for main variant.
7112         (register_odr_type): Simplify.
7114 2018-10-25  Richard Biener  <rguenther@suse.de>
7116         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
7117         Initialize ng to silence error with release checking bootstrap.
7119 2018-10-25  Richard Biener  <rguenther@suse.de>
7121         * tree-if-conv.c: Include tree-ssa-sccvn.h.
7122         (tree_if_conversion): Run CSE on the if-converted loop body.
7124 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
7126         * config/s390/constraints.md (ZL): New constraint.
7127         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
7128         operands.
7129         * config/s390/s390.md (movdi_larl): Remove.
7130         (movdi_64): Add the LARL alternative.
7132 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
7134         PR bootstrap/87747
7135         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
7136         (RTX_CODE_HWINT_P): New macro.
7137         (rtx_code_size): Use RTX_CODE_HWINT_P ().
7139 2018-10-25  Jan Hubicka  <jh@suse.cz>
7141         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
7142         is anonymous.
7144 2018-10-25  Richard Biener  <rguenther@suse.de>
7146         PR tree-optimization/87665
7147         PR tree-optimization/87745
7148         * tree-vectorizer.h (get_earlier_stmt): Remove.
7149         (get_later_stmt): Pick up UID from the original non-pattern stmt.
7151 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
7153         * options.texi (Deprecated): Move list to Var section.
7155 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
7156             Jinsong Ji  <jji@us.ibm.com>
7158         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
7159         __vector long to __vector long long.
7160         (_mm_cvtpd_ps): Likewise.
7161         (_mm_cvttpd_epi32): Likewise.
7162         (_mm_cvtpi32_pd): Likewise.
7163         (_mm_unpackhi_epi64): Likewise.
7164         (_mm_unpacklo_epi64): Likewise.
7166 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
7168         PR rtl-optimization/87720
7169         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
7171 2018-10-24  Alexandre Oliva  <aoliva@redhat.com>
7173         * gimple-ssa-isolate-paths.c
7174         (find_implicit_erroneous_behavior): Do not change code if the
7175         pass is running for warnings only.
7176         (find_explicit_erroneous_behavior): Likewise.
7178 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
7180         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
7181         Define as rs6000_mangle_decl_assembler_name.
7182         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
7183         long double to IEEE long double, switch the names of the long
7184         double built-in functions to be <func>f128 instead of <func>l.
7186 2018-10-24  Martin Sebor  <msebor@redhat.com>
7188         * doc/extend.texi (nonnull): List no-argument form.  Reference
7189         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
7191 2018-10-24  Richard Biener  <rguenther@suse.de>
7193         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
7195 2018-10-24  Martin Liska  <mliska@suse.cz>
7197         PR tree-optimization/84436
7198         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
7199         Remove.
7200         (switch_conversion::contains_linear_function_p): New.
7201         (switch_conversion::build_one_array): Support linear
7202         transformation on input.
7203         * tree-switch-conversion.h (struct switch_conversion): Add
7204         contains_linear_function_p declaration.
7206 2018-10-24  Richard Biener  <rguenther@suse.de>
7208         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
7209         if its argument is CONSTANT_CLASS_P.
7211 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
7213         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
7214         it is wrong for forward declarations.
7216 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
7218         * config/s390/s390.c (s390_check_qrst_address): Add the missing
7219         SYMBOL_REF_P () check.
7221 2018-10-24  Richard Biener  <rguenther@suse.de>
7223         PR tree-optimization/87105
7224         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
7225         dump classification.
7226         (vect_analyze_data_ref_accesses): Handle duplicate loads and
7227         stores by splitting the affected group after the fact.
7228         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
7229         fail the SLP build because of size constraints.
7231 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7233         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
7234         * configure: Regenerate.
7235         * config.in: Regenerate.
7236         * varasm.c (mergeable_string_section): Use readonly_data_section
7237         if linker doesn't support SHF_MERGE with alignment > 8.
7238         (mergeable_constant_section): Likewise.
7240 2018-10-24  Richard Biener  <rguenther@suse.de>
7242         PR tree-optimization/84013
7243         * tree-ssa-structalias.c (struct msdi_data): New struct for
7244         marshalling data to walk_stmt_load_store_ops.
7245         (maybe_set_dependence_info): Refactor as callback for
7246         walk_stmt_load_store_ops.
7247         (compute_dependence_clique): Set restrict info on all stmt kinds.
7249 2018-10-24  Martin Liska  <mliska@suse.cz>
7251         * cgraph.c (cgraph_node::dump):
7252         Remove reduntant dumps and make tp_first_run dump more compact.
7254 2018-10-24  Richard Biener  <rguenther@suse.de>
7256         PR tree-optimization/87665
7257         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
7258         to reflect reality.
7260 2018-10-12  Jeff Law  <law@redhat.com>
7262         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
7263         for H8/S.
7265 2018-10-23  Richard Biener  <rguenther@suse.de>
7267         * tree-vrp.c (add_assert_info): Guard dump_printf with
7268         dump_enabled_p.
7269         * gimple-ssa-evrp-analyze.c
7270         (evrp_range_analyzer::record_ranges_from_incoming_edge):
7271         Use value_range::ignore_equivs_equal_p.
7273 2018-10-23  Richard Biener  <rguenther@suse.de>
7275         PR tree-optimization/87105
7276         PR tree-optimization/87608
7277         * passes.def (pass_all_early_optimizations): Add early phi-opt
7278         after dce.
7279         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
7280         addition to debug stmts.
7281         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
7282         and abs replacement early.
7283         * tree-cfg.c (gimple_empty_block_p): Likewise.
7285 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
7287         PR target/86383
7288         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
7289         specified to configure.
7290         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
7292 2018-10-23  Richard Biener  <rguenther@suse.de>
7294         PR tree-optimization/87700
7295         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
7297 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
7299         PR target/87674
7300         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
7301         second argument from __mmask16 to __mmask8.
7302         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
7303         _mm_mask_packs_epi32): Likewise.
7304         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
7305         Likewise.
7306         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
7308 2018-10-23  Richard Biener  <rguenther@suse.de>
7310         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
7312 2018-10-23  Richard Biener  <rguenther@suse.de>
7314         PR tree-optimization/86144
7315         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
7316         over simd attribute.
7318 2018-10-23  Richard Biener  <rguenther@suse.de>
7320         PR tree-optimization/87693
7321         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
7322         the case we do not find the taken edge.
7324 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
7325             Jinsong Ji  <jji@us.ibm.com>
7327         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
7328         (_mm_store_pd): Use unaligned vector type for pointer cast.
7329         (_mm_maskmoveu_si128): Likewise.
7330         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
7331         (_mm_store_ps): Use unaligned vector type for pointer cast.
7333 2018-10-22  Paul Koning  <ni1d@arrl.net>
7335         * symtab.c (symtab_node::increase_alignment): Correct max
7336         alignment check.
7338 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
7340         PR tree-optimization/87633
7341         * match.pd: Do not generate unordered integer comparisons.
7343 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
7345         PR rtl-optimization/87600
7346         * combine.c: Add include of expr.h.
7347         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
7348         register to a pseudo.
7349         (make_more_copies): New function, add a copy to a new pseudo after
7350         the moves from hard registers into pseudos.
7351         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
7352         later.  Call make_more_copies.
7354 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
7356         * lra-constraints.c (process_alt_operands): New local array,
7357         matching_early_clobber.  Check matching_early_clobber before
7358         decrementing reject, and set matching_early_clobber after.
7360 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
7362         PR target/87598
7363         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
7364         call output_addr_const and hope for the best.
7366 2018-10-22  Richard Biener  <rguenther@suse.de>
7368         * gimple-ssa-evrp-analyze.c
7369         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
7370         smarter about what ranges to use.
7371         * tree-vrp.c (add_assert_info): Dump here.
7372         (register_edge_assert_for_2): Instead of here at multiple but
7373         not all places.
7375         * gcc.dg/tree-ssa/evrp12.c: New testcase.
7376         * gcc.dg/predict-6.c: Adjust.
7377         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
7378         * gcc.dg/tree-ssa/vrp02.c: Likewise.
7379         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
7381 2018-10-22  Steven Bosscher  <steven@gcc.gnu.org>
7382         Richard Biener  <rguenther@suse.de>
7384         * bitmap.h: Update data structure documentation, including a
7385         description of bitmap views as either linked-lists or splay trees.
7386         (struct bitmap_element_def): Update comments for splay tree bitmaps.
7387         (struct bitmap_head_def): Likewise.
7388         (bitmap_list_view, bitmap_tree_view): New prototypes.
7389         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
7390         tree_form fields.
7391         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
7392         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
7393         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
7394         released bitmap element here.
7395         (bitmap_element_free): Remove.
7396         (bitmap_elt_clear_from): Work on splay tree bitmaps.
7397         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
7398         this function similar ones such that linked-list bitmap implementation
7399         functions are grouped.
7400         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
7401         and moved for grouping.
7402         (bitmap_list_insert_element_after): Renamed from
7403         bitmap_elt_insert_after, and moved for grouping.
7404         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
7405         (bitmap_tree_link_left, bitmap_tree_link_right,
7406         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
7407         bitmap_tree_link_element, bitmap_tree_unlink_element,
7408         bitmap_tree_find_element): New functions for splay-tree bitmap
7409         implementation.
7410         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
7411         Renamed and moved, see above entries.
7412         (bitmap_tree_listify_from): New function to convert part of a splay
7413         tree bitmap to a linked-list bitmap.
7414         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
7415         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
7416         (bitmap_find_bit): Remove.
7417         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
7418         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
7419         Handle splay tree bitmaps.
7420         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
7421         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
7422         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
7423         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
7424         bitmap_intersect_compl_p, bitmap_ior_and_compl,
7425         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
7426         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
7427         corresponding changes to use linked-list specific bitmap_element
7428         manipulation functions as applicable for efficiency.
7429         (bitmap_tree_to_vec): New function.
7430         (debug_bitmap_elt_file): New function split out from ...
7431         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
7432         (bitmap_print): Likewise.
7434         PR tree-optimization/63155
7435         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
7436         SSA edge worklists.
7437         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
7438         in tree-view.
7440 2018-10-22  Martin Liska  <mliska@suse.cz>
7442         PR tree-optimization/87686
7443         Revert
7444         2018-08-29  Martin Liska  <mliska@suse.cz>
7446         * tree-switch-conversion.c (switch_conversion::expand):
7447         Strenghten assumption about gswitch statements.
7449 2018-10-22  Martin Liska  <mliska@suse.cz>
7451         * ipa-icf.c (sem_item::compare_attributes): Remove.
7452         (sem_item::compare_referenced_symbol_properties): Use
7453         attribute_list_equal instead.
7454         (sem_function::equals_wpa): Likewise.
7455         * ipa-icf.h: Remove compare_attributes.
7457 2018-10-22  Richard Biener  <rguenther@suse.de>
7459         PR middle-end/87682
7460         * mem-stats.h (mem_usage::operator==): Fix pasto.
7462 2018-10-22  Richard Biener  <rguenther@suse.de>
7464         PR tree-optimization/87640
7465         * tree-vrp.c (set_value_range_with_overflow): Decompose
7466         incomplete result.
7467         (extract_range_from_binary_expr_1): Adjust.
7469 2018-10-22  Martin Jambor  <mjambor@suse.cz>
7471         * tree-eh.h (stmt_could_throw_p): Add function parameter.
7472         (stmt_can_throw_external): Likewise.
7473         (stmt_can_throw_internal): Likewise.
7474         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
7475         (lower_eh_constructs_2): Likewise.
7476         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
7477         (stmt_can_throw_external): Likewise.
7478         (stmt_can_throw_internal): Likewise.
7479         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
7480         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
7481         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
7482         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
7483         (pass_lower_eh_dispatch::execute): Pass cfun to
7484         stmt_can_throw_external.
7485         (cleanup_empty_eh): Likewise.
7486         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
7487         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
7488         stmt_can_throw_external instead of pushing it to cfun.
7489         (symbol_table::create_edge): Likewise.
7490         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
7491         stmt_can_throw_internal.
7492         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
7493         to stmt_could_throw_p.
7494         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
7495         stmt_can_throw_internal.
7496         (pass_store_merging::execute): Likewise.
7497         * gimple-ssa-strength-reduction.c
7498         (find_candidates_dom_walker::before_dom_children): Pass cfun to
7499         stmt_could_throw_p.
7500         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
7501         stmt_can_throw_internal.
7502         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
7503         to stmt_can_throw_external.
7504         (check_stmt): Pass cfun to stmt_could_throw_p.
7505         (check_stmt): Pass cfun to stmt_can_throw_external.
7506         (pass_nothrow::execute): Likewise.
7507         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
7508         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
7509         stmt_can_throw_internal.
7510         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
7511         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
7512         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
7513         * tree-complex.c (expand_complex_libcall): Pass cfun to
7514         stmt_could_throw_p and to stmt_can_throw_internal.
7515         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
7516         * tree-inline.c (copy_edges_for_bb): Likewise.
7517         (maybe_move_debug_stmts_to_successors): Likewise.
7518         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
7519         stmt_could_throw_p.
7520         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
7521         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
7522         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
7523         stmt_can_throw_internal.
7524         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
7525         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
7526         stmt_could_throw_p.
7527         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
7528         stmt_can_throw_internal.
7529         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
7530         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
7531         stmt_could_throw_p.
7532         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
7533         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
7534         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7535         (convert_mult_to_fma_1): Likewise.
7536         (convert_to_divmod): Likewise.
7537         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
7538         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
7539         * tree-ssa-propagate.c
7540         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
7541         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
7542         (maybe_optimize_range_tests): Likewise.
7543         (linearize_expr_tree): Likewise.
7544         (reassociate_bb): Likewise.
7545         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7546         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
7547         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
7548         (handle_char_store): Likewise.
7549         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
7550         stmt_can_throw_internal.
7551         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
7552         stmt_could_throw_p.
7553         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7554         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
7555         (vectorizable_simd_clone_call): Likewise.
7556         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
7557         (gimple_stringop_fixed_value): Likewise.
7559 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
7561         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
7562         literal pool references.
7563         (s390_check_qrst_address): Adapt to the new behavior of
7564         s390_loadrelative_operand_p ().
7566 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
7568         PR target/72782
7569         * config/i386/sse.md (*andnot<mode>3_bcst): New.
7571 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
7573         PR target/72782
7574         * config/i386/sse.md (*<code><mode>3_bcst): New.
7576 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
7578         PR target/72782
7579         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
7580         V4DI, V16SI and V8DI.
7581         (*sub<mode>3<mask_name>_bcst): New.
7582         (*add<mode>3<mask_name>_bcst): Likewise.
7584 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
7585             Jinsong Ji  <jji@us.ibm.com>
7587         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
7588         __m64 with __vector unsigned long long for compatibility.
7589         (_mm_movemask_epi8): Likewise.
7590         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
7591         (_mm_cvttps_pi32): Likewise.
7592         (_mm_cvtpi32_ps): Likewise.
7593         (_mm_cvtps_pi16): Likewise.
7594         (_mm_loadh_pi): Likewise.
7595         (_mm_storeh_pi): Likewise.
7596         (_mm_movehl_ps): Likewise.
7597         (_mm_movelh_ps): Likewise.
7598         (_mm_loadl_pi): Likewise.
7599         (_mm_storel_pi): Likewise.
7600         (_mm_movemask_ps): Likewise.
7601         (_mm_shuffle_pi16): Likewise.
7603 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7605         PR target/72782
7606         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
7607         __builtin_ia32_vfnmsubpd512_mask.
7608         (_mm512_mask_fnmsub_round_pd): Likewise.
7609         (_mm512_fnmsub_pd): Likewise.
7610         (_mm512_mask_fnmsub_pd): Likewise.
7611         (_mm512_maskz_fnmsub_round_pd): Use
7612         __builtin_ia32_vfnmsubpd512_maskz.
7613         (_mm512_maskz_fnmsub_pd): Likewise.
7614         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
7615         (_mm512_mask_fnmsub_round_ps): Likewise.
7616         (_mm512_fnmsub_ps): Likewise.
7617         (_mm512_mask_fnmsub_ps): Likewise.
7618         (_mm512_maskz_fnmsub_round_ps): Use
7619         __builtin_ia32_vfnmsubps512_maskz.
7620         (_mm512_maskz_fnmsub_ps): Likewise.
7621         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
7622         __builtin_ia32_vfnmsubpd256_mask.
7623         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
7624         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7625         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
7626         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7627         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7628         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
7629         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
7630         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
7631         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
7632         __builtin_ia32_vfnmsubpd.
7633         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
7634         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
7635         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
7636         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
7637         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
7638         * config/i386/i386-builtin.def: Add
7639         __builtin_ia32_vfnmsubpd256_mask,
7640         __builtin_ia32_vfnmsubpd256_maskz,
7641         __builtin_ia32_vfnmsubpd128_mask,
7642         __builtin_ia32_vfnmsubpd128_maskz,
7643         __builtin_ia32_vfnmsubps256_mask,
7644         __builtin_ia32_vfnmsubps256_maskz,
7645         __builtin_ia32_vfnmsubps128_mask,
7646         __builtin_ia32_vfnmsubps128_maskz,
7647         __builtin_ia32_vfnmsubpd512_mask,
7648         __builtin_ia32_vfnmsubpd512_maskz,
7649         __builtin_ia32_vfnmsubps512_mask,
7650         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
7651         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
7652         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
7653         __builtin_ia32_vfnmsubpd256.
7654         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
7655         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
7656         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
7657         Likewise.
7658         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
7659         Likewise.
7660         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
7661         Likewise.
7662         (fmai_vmfnmsub_<mode><round_name>): Likewise.
7664 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7666         PR target/72782
7667         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
7668         __builtin_ia32_vfnmaddpd512_mask.
7669         (_mm512_mask_fnmadd_round_pd): Likewise.
7670         (_mm512_fnmadd_pd): Likewise.
7671         (_mm512_mask_fnmadd_pd): Likewise.
7672         (_mm512_maskz_fnmadd_round_pd): Use
7673         __builtin_ia32_vfnmaddpd512_maskz.
7674         (_mm512_maskz_fnmadd_pd): Likewise.
7675         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
7676         (_mm512_mask_fnmadd_round_ps): Likewise.
7677         (_mm512_fnmadd_ps): Likewise.
7678         (_mm512_mask_fnmadd_ps): Likewise.
7679         (_mm512_maskz_fnmadd_round_ps): Use
7680         __builtin_ia32_vfnmaddps512_maskz.
7681         (_mm512_maskz_fnmadd_ps): Likewise.
7682         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
7683         __builtin_ia32_vfnmaddpd256_mask.
7684         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
7685         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
7686         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
7687         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7688         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7689         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
7690         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
7691         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
7692         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
7693         __builtin_ia32_vfnmaddpd.
7694         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
7695         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
7696         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
7697         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
7698         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
7699         * config/i386/i386-builtin.def: Add
7700         __builtin_ia32_vfnmaddpd256_mask,
7701         __builtin_ia32_vfnmaddpd256_maskz,
7702         __builtin_ia32_vfnmaddpd128_mask,
7703         __builtin_ia32_vfnmaddpd128_maskz,
7704         __builtin_ia32_vfnmaddps256_mask,
7705         __builtin_ia32_vfnmaddps256_maskz,
7706         __builtin_ia32_vfnmaddps128_mask,
7707         __builtin_ia32_vfnmaddps128_maskz,
7708         __builtin_ia32_vfnmaddpd512_mask,
7709         __builtin_ia32_vfnmaddpd512_maskz,
7710         __builtin_ia32_vfnmaddps512_mask,
7711         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
7712         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
7713         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
7714         __builtin_ia32_vfnmaddpd256.
7715         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
7716         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
7717         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
7718         Likewise.
7719         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
7720         Likewise.
7721         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
7722         Likewise.
7723         (fmai_vmfnmadd_<mode><round_name>): Likewise.
7725 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7727         PR target/72782
7728         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
7729         __builtin_ia32_vfmsubpd512_mask.
7730         (_mm512_mask_fmsub_round_pd): Likewise.
7731         (_mm512_fmsub_pd): Likewise.
7732         (_mm512_mask_fmsub_pd): Likewise.
7733         (_mm512_maskz_fmsub_round_pd): Use
7734         __builtin_ia32_vfmsubpd512_maskz.
7735         (_mm512_maskz_fmsub_pd): Likewise.
7736         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
7737         (_mm512_mask_fmsub_round_ps): Likewise.
7738         (_mm512_fmsub_ps): Likewise.
7739         (_mm512_mask_fmsub_ps): Likewise.
7740         (_mm512_maskz_fmsub_round_ps): Use
7741         __builtin_ia32_vfmsubps512_maskz.
7742         (_mm512_maskz_fmsub_ps): Likewise.
7743         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
7744         __builtin_ia32_vfmsubpd256_mask.
7745         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
7746         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7747         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
7748         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7749         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7750         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
7751         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
7752         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
7753         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
7754         __builtin_ia32_vfmsubpd.
7755         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
7756         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
7757         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
7758         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
7759         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
7760         * config/i386/i386-builtin.def: Add
7761         __builtin_ia32_vfmsubpd256_mask,
7762         __builtin_ia32_vfmsubpd256_maskz,
7763         __builtin_ia32_vfmsubpd128_mask,
7764         __builtin_ia32_vfmsubpd128_maskz,
7765         __builtin_ia32_vfmsubps256_mask,
7766         __builtin_ia32_vfmsubps256_maskz,
7767         __builtin_ia32_vfmsubps128_mask,
7768         __builtin_ia32_vfmsubps128_maskz,
7769         __builtin_ia32_vfmsubpd512_mask,
7770         __builtin_ia32_vfmsubpd512_maskz,
7771         __builtin_ia32_vfmsubps512_mask,
7772         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
7773         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
7774         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
7775         __builtin_ia32_vfmsubpd256.
7776         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
7777         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
7778         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
7779         Likewise.
7780         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
7781         Likewise.
7782         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
7783         Likewise.
7784         (fmai_vmfmsub_<mode><round_name>): Likewise.
7786 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7788         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
7789         Remove plus.  Renamed to ...
7790         (*sub<mode>3<mask_name>_bcst): This.
7791         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
7792         (*add<mode>3<mask_name>_bcst): This.
7794 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7796         PR target/72782
7797         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
7799 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
7801         PR target/87662
7802         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
7803         (_mm_or_epi32): Likewise.
7804         (_mm256_xor_epi32): Likewise.
7805         (_mm_xor_epi32): Likewise.
7806         (_mm256_or_epi64): Likewise.
7807         (_mm_or_epi64): Likewise.
7808         (_mm256_xor_epi64): Likewise.
7809         (_mm_xor_epi64): Likewise.
7811 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
7813         PR target/72782
7814         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
7816 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
7818         PR middle-end/87647
7819         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
7821 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
7823         * doc/ux.texi: Move @section directly after @node.
7825 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
7827         PR middle-end/85488
7828         PR middle-end/87649
7829         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
7830         depend closely nested inside of loop with ordered clause with
7831         a parameter.
7833 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
7835         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
7836         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
7837         * doc/ux.texi: New file.
7839 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
7841         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
7842         be the first CR field allocated.
7844 2018-10-19  Richard Biener  <rguenther@suse.de>
7846         PR target/87657
7847         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
7848         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
7850 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
7852         PR target/72782
7853         * config/i386/sse.md
7854         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
7855         (*add<mode>3<mask_name>_bcst_2): Likewise.
7857 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
7859         * config/i386/sse.md
7860         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7861         Replace nonimmediate_operand with register_operand.
7862         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7863         Likewise.
7864         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7865         Likewise.
7867 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7869         PR rtl-optimization/87596
7870         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
7871         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
7872         for instructions in FROM..TO range.
7874 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
7876         * cfgexpand.c (expand_one_var): Use specific wording in error message
7877         for non-local frame variables.
7878         * stor-layout.c (layout_decl): Do not issue a warning for them.
7880 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
7882         * haifa-sched.c (priority): Add force_recompute parameter.
7883         (apply_replacement): Call priority () with force_recompute = true.
7884         (restore_pattern): Likewise.
7886 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7888         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
7889         HOST_BITS_PER_WIDE_INT.
7890         (test_vector_ops_duplicate): Likewise.
7892 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7894         PR target/72782
7895         * config/i386/sse.md (VF_AVX512): New.
7896         (avx512bcst): Likewise.
7897         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7898         Likewise.
7899         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7900         Likewise.
7901         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7902         Likewise.
7904 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
7906         * doc/invoke.texi (-dumpversion): Improve grammar.
7907         (-dumpfullversion): Make more consistent with -dumpversion.
7909 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
7911         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
7912         Set JUMP_LABEL to the jump insn.
7913         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
7914         Predict emitted jump and add label to jump insn.
7916 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
7918         PR tree-optimization/87562
7919         * input.c (get_substring_ranges_for_loc): Use
7920         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
7921         getting the linemap for the endpoint.  Verify that it's either
7922         in the same linemap as the start point's spelling location, or
7923         at least in the same file.
7925 2018-10-18  Richard Biener  <rguenther@suse.de>
7927         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
7928         feed width-specific load/store costs through ix86_vec_cost.
7929         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
7930         (k8_cost): Likewise.
7931         (bdver_cost): Likewise.
7932         (znver1_cost): Likewise.
7933         (btver1_cost): Likewise.
7934         (btver2_cost): Likewise.
7936 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7938         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
7939         to simplify subreg of vec_merge.
7941 2018-10-18  Richard Biener  <rguenther@suse.de>
7943         * config/i386/i386.c: Fix costing of vector FMA.
7945 2018-10-18  Richard Biener  <rguenther@suse.de>
7947         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
7948         and argument.
7949         (ix86_builtin_vectorization_cost): For vec_construct properly
7950         cost insertion into SSE regs.
7951         (...): Adjust calls to ix86_vec_cost.
7953 2018-10-18  Richard Biener  <rguenther@suse.de>
7955         PR middle-end/87087
7956         Revert
7957         2018-02-07  Richard Biener  <rguenther@suse.de>
7959         PR tree-optimization/84204
7960         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7961         this place.
7963 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7965         PR target/87537
7966         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
7967         of vec_duplicate.
7968         (test_vector_ops_duplicate): Add test for a scalar subreg of a
7969         VEC_MERGE of a VEC_DUPLICATE.
7971 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
7973         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
7974         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
7975         * doc/standards.texi (C Language): Document C2X.
7976         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
7977         (rl78_option_override): Handle "GNU C2X" language name.
7979 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
7981         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
7982         Document C17 as published in 2018.
7984 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
7986         PR middle-end/87623
7987         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
7988         bail out if both sides do not have the same storage order.
7990 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
7992         * bitmap.c (bitmap_head::dump): New.
7993         * bitmap.h (bitmap_head): Add dump().
7994         * gimple-ssa-evrp-analyze.c
7995         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
7996         (evrp_range_analyzer::set_ssa_range_info): Same.
7997         (evrp_range_analyzer::record_ranges_from_phis): Same.
7998         (evrp_range_analyzer::record_ranges_from_stmt): Same.
7999         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
8000         * gimple-ssa-sprintf.c (get_int_range): Same.
8001         (format_integer): Same.
8002         (sprintf_dom_walker::handle_gimple_call): Same.
8003         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
8004         (ipcp_vr_lattice::top_p): Same.
8005         (ipcp_vr_lattice::bottom_p): Same.
8006         (ipcp_vr_lattice::set_to_bottom): Same.
8007         (ipa_vr_operation_and_type_effects): Same.
8008         (propagate_vr_across_jump_function): Same.
8009         (ipcp_store_vr_results): Same.
8010         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
8011         (ipa_print_node_jump_functions_for_edge): Same.
8012         (ipa_get_value_range): Same.
8013         (ipa_compute_jump_functions_for_edge): Same.
8014         (ipa_write_jump_function): Same.
8015         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
8016         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
8017         Same.
8018         * vr-values.c (set_value_range_to_nonnegative): Same.
8019         (set_value_range_to_truthvalue): Same.
8020         (vr_values::get_value_range): Same.
8021         (vr_values::set_defs_to_varying): Same.
8022         (vr_values::update_value_range): Same.
8023         (symbolic_range_based_on_p): Same.
8024         (vr_values::op_with_boolean_value_range_p): Same.
8025         (vr_values::extract_range_for_var_from_comparison_expr): Same.
8026         (vr_values::extract_range_from_ssa_name): Same.
8027         (vr_values::extract_range_from_binary_expr): Same.
8028         (vr_values::extract_range_from_unary_expr): Same.
8029         (vr_values::extract_range_from_cond_expr): Same.
8030         (vr_values::extract_range_from_comparison): Same.
8031         (vr_values::check_for_binary_op_overflow): Same.
8032         (vr_values::extract_range_basic): Same.
8033         (vr_values::extract_range_from_assignment): Same.
8034         (compare_ranges): Same.
8035         (compare_range_with_value): Same.
8036         (vr_values::adjust_range_with_scev): Same.
8037         (vrp_valueize): Same.
8038         (vrp_valueize_1): Same.
8039         (vr_values::get_vr_for_comparison): Same.
8040         (vr_values::compare_name_with_value): Same.
8041         (vr_values::compare_names): Same.
8042         (vr_values::vrp_evaluate_conditional): Same.
8043         (find_case_label_ranges): Same.
8044         (vr_values::vrp_visit_switch_stmt): Same.
8045         (vr_values::extract_range_from_phi_node): Same.
8046         (vr_values::simplify_div_or_mod_using_ranges): Same.
8047         (vr_values::simplify_bit_ops_using_ranges): Same.
8048         (test_for_singularity): Same.
8049         (range_fits_type_p): Same.
8050         (vr_values::simplify_cond_using_ranges_1): Same.
8051         (vr_values::simplify_switch_using_ranges): Same.
8052         (vr_values::simplify_float_conversion_using_ranges): Same.
8053         (vr_values::two_valued_val_range_p): Same.
8054         (vr_values::add_equivalence): Move to value_range::equiv_add.
8055         * vr-values.h (vr_values::add_equivalence): Remove.
8056         (VR_INITIALIZER): Remove.
8057         * tree-vrp.c (value_range::set): New.
8058         (value_range::equiv_add): New.
8059         (value_range::value_range): New.
8060         (value_range::deep_copy): New.
8061         (value_range::check): New.
8062         (value_range::equal_p): New.
8063         (value_range::ignore_equivs_equal_p): New.
8064         (value_range::operator==): New.
8065         (value_range::operator!=): New.
8066         (value_range::symbolic_p): New.
8067         (value_range::numeric_p): New.
8068         (value_range::set_undefined): New.
8069         (value_range::set_varying): New.
8070         (value_range::may_contain_p): New.
8071         (value_range::equiv_clear): New.
8072         (value_range::singleton_p): New.
8073         (value_range::intersect): New.
8074         (value_range::dump): New.
8075         (value_range::set_and_canonicalize): New.
8076         (set_value_range): Adjust for value_range API.
8077         (set_value_range_to_undefined): Same.
8078         (set_value_range_to_varying): Same.
8079         (set_and_canonicalize_value_range): Same.
8080         (set_value_range_to_nonnull): Same.
8081         (set_value_range_to_null): Same.
8082         (range_is_null): Same.
8083         (range_is_nonnull): Same.
8084         (range_int_cst_p): Same.
8085         (range_int_cst_singleton_p): Same.
8086         (symbolic_range_p): Same.
8087         (range_includes_zero_p): Same.
8088         (value_range_constant_singleton): Same.
8089         (vrp_set_zero_nonzero_bits): Same.
8090         (ranges_from_anti_range): Same.
8091         (extract_range_into_wide_ints): Same.
8092         (extract_range_from_multiplicative_op): Same.
8093         (set_value_range_with_overflow): Same.
8094         (extract_range_from_binary_expr_1): Same.
8095         (extract_range_from_unary_expr): Same.
8096         (dump_value_range): Same.
8097         (debug_value_range): Same.
8098         (vrp_prop::check_array_ref): Same.
8099         (vrp_prop::check_mem_ref): Same.
8100         (vrp_prop::vrp_initialize): Same.
8101         (vrp_prop::visit_stmt): Same.
8102         (intersect_ranges): Same.
8103         (vrp_prop::visit_phi): Same.
8104         (vrp_prop::vrp_finalize): Same.
8105         (determine_value_range_1): Same.
8106         (determine_value_range): Same.
8107         (vrp_intersect_ranges_1): Rename to...
8108         (vrp_intersect_1): this.
8109         (vrp_intersect_ranges): Rename to...
8110         (value_range::intersect_helper): ...this.
8111         (vrp_meet_1): Rename to...
8112         (value_range::union_helper): ...this.
8113         (vrp_meet): Rename to...
8114         (value_range::union_): ...this.
8115         (copy_value_range): Remove.
8116         * tree-vrp.h (struct value_range): Rewrite into a proper class.
8117         (value_range::vrtype): New.
8118         (value_range::type): New.
8119         (value_range::equiv): New.
8120         (value_range::min): New.
8121         (value_range::max): New.
8122         (value_range::varying_p): New.
8123         (value_range::undefined_p): New.
8124         (value_range::null_p): New.
8125         (value_range::equiv_add): New.
8126         (copy_value_range): Remove.
8128 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
8130         * Makefile.in (SELFTEST_TARGETS): New.
8131         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
8132         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
8133         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
8134         c/Make-lang.in.
8135         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
8136         (selftest-c++-gdb, selftest-c++-valgrind): Move to
8137         cp/Make-lang.in.
8138         * configure: Regenerate.
8139         * configure.ac (selftest_languages): New.
8141 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
8143         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
8144         overflow wraps argument.
8145         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
8146         wide_int_range_multiplicative_op.
8147         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
8148         overflow wraps argument.
8149         (wide_int_range_multiplicative_op): Same.
8150         (wide_int_range_lshift): Same.
8151         (wide_int_range_div): Same.
8152         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
8153         (wide_int_range_lshift): Same.
8154         (wide_int_range_div): Same.
8156 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
8158         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
8159         use sign as argument.
8160         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
8161         wide_int_range_shift_undefined_p.
8163 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8165         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
8166         Rename to...
8167         (@despeculate_copy<ALLI_TI:mode>): ... This.
8168         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
8169         switch statement.
8171 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8173         * config.gcc: Obsolete *-*-solaris2.10*.
8174         * doc/install.texi (Specific, *-*-solaris2*): Document it.
8176 2018-10-12  Jeff Law  <law@redhat.com>
8178         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
8179         reg + sym +- const_int addressing modes.
8181 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
8183         * common.opt (fdiagnostics-minimum-margin-width=): New option.
8184         * diagnostic-show-locus.c (layout::layout): Apply the minimum
8185         margin width.
8186         (layout::start_annotation_line): Only print up to 3 of the
8187         margin character, to avoid touching the left-hand side.
8188         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
8189         minimum margin width, as set by test_diagnostic_context's ctor.
8190         (selftest::test_fixit_insert_containing_newline): Likewise.
8191         (selftest::test_fixit_insert_containing_newline_2): Likewise.
8192         (selftest::test_line_numbers_multiline_range): Clear
8193         dc.min_margin_width.
8194         * diagnostic.c (diagnostic_initialize): Initialize
8195         min_margin_width.
8196         * diagnostic.h (struct diagnostic_context): Add field
8197         "min_margin_width".
8198         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
8199         * opts.c (common_handle_option): Handle
8200         OPT_fdiagnostics_minimum_margin_width_.
8201         * selftest-diagnostic.c
8202         (selftest::test_diagnostic_context::test_diagnostic_context):
8203         Initialize min_margin_width to 6.
8204         * toplev.c (general_init): Initialize global_dc->min_margin_width.
8206 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
8208         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
8209         Fix usage of "error_at_rich_loc" in the comment.
8211 2018-10-15  Renlin Li  <renlin.li@arm.com>
8213         PR target/87563
8214         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
8215         if-conversioned loop when it contains ifn with types not
8216         supported by backend.
8217         * internal-fn.c (expand_direct_optab_fn): Add an assert.
8218         (direct_internal_fn_supported_p): New helper function.
8219         * internal-fn.h (direct_internal_fn_supported_p): Declare.
8221 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
8223         PR target/87572
8224         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
8225         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
8226         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
8228 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
8230         PR tree-optimization/87022
8231         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
8232         bits in dist vector rather than the first one.
8234 2018-10-15  Richard Biener  <rguenther@suse.de>
8236         PR middle-end/87610
8237         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
8238         (visit_loadstore): When a used restrict tag escaped verify that
8239         the points-to solution of "other" pointers do not include
8240         escaped.
8241         (compute_dependence_clique): If a used restrict tag escaped
8242         communicated that down to visit_loadstore.
8244 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
8246         * config/s390/s390.c (s390_expand_vec_init): Force vector element
8247         into reg if it isn't a general operand.
8249 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
8251         PR target/87599
8252         * config/i386/sse.md (*vec_dupv2di): Add register source to
8253         movddup.
8255 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
8257         PR target/87572
8258         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
8259         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
8260         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
8261         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
8263 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
8265         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
8266         (notice_args_size): Set it in the current trace if no insn that can
8267         throw internally has been seen yet.
8268         (connect_traces): When connecting args_size between traces, allow the
8269         incoming values not to match if there is an insn setting it before the
8270         first insn that can throw internally; in that case, force the creation
8271         of a CFI note on this latter insn.
8273 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
8275         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
8276         the base class, because using the injected-class-name was not clearly
8277         specified until DR 176.
8279 2018-10-12  Paul Koning  <ni1d@arrl.net>
8281         * config/pdp11/pdp11.md (doloop_end): New expander.
8282         (doloop_end_insn): renamed from "doloop_end".
8283         (addqi3): New pattern.
8284         (subqi3): New pattern.
8285         * config/pdp11/predicates.md (incdec_operand): New predicate.
8287 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
8289         PR middle-end/81376
8290         * real.c (format_helper::can_represent_integral_type_p): New function
8291         * real.h (format_helper::can_represent_integral_type_p): Ditto.
8292         * match.pd: New pattern.
8294 2018-10-12  Alexandre Oliva  <oliva@adacore.com>
8296         * configure.ac: Introduce --enable-large-address-aware
8297         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
8298         * doc/install.texi: Document it.
8299         * configure, config.in: Rebuilt.
8300         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
8301         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
8302         (LINK_SPEC): Insert it.
8303         * config/i386/mingw-w64.h: Likewise.
8305         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
8307 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
8309         PR rtl-optimization/87600
8310         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
8312 2018-10-12  Paul Koning  <ni1d@arrl.net>
8314         * doc/md.texi (doloop_end): Document that the pattern code may
8315         need to check operand mode.
8317 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
8319         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
8320         to zero-extend between int and floating-point registers.
8321         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
8322         ldp into floating-point registers.  Add type and arch attributes.
8323         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
8324         Use f_loads for type attribute.
8326 2018-10-11  Martin Sebor  <msebor@redhat.com>
8328         * doc/extend.texi (attribute packed): Correct typos.
8330 2018-10-11  Martin Sebor  <msebor@redhat.com>
8332         * doc/extend.texi (attribute flatten): Mention interaction with
8333         noinline.
8335 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
8337         PR target/87156
8338         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8339         Set new_decl virtual flag to zero.
8341 2018-10-11  Martin Sebor  <msebor@redhat.com>
8343         PR middle-end/87593
8344         * doc/extend.texi (attribute format_arg): Discuss using multiple
8345         attributes on a single function.
8347 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8349         PR tree-optimization/86829
8350         * match.pd (sin (atan (x))): New simplification rules.
8351         (cos (atan (x))): Likewise.
8352         * real.c (build_sinatan_real): New function.
8353         * real.h (build_sinatan_real): Prototype.
8355 2018-10-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
8357         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
8358         function.
8359         (fold_mergeeo_helper): New helper function.
8360         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
8361         intrinsics.  Correct some whitespace indentation issues.
8363 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
8365         PR target/87511
8366         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
8367         Use HOST_WIDE_INT_1U for shift.
8369 2018-10-11  Doug Rupp  <rupp@adacore.com>
8370             Olivier Hainque  <hainque@adacore.com>
8372         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
8373         Pass --relax to the linker for RTPs.
8374         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
8376 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
8377             Jan Hubicka  <jh@suse.cz>
8378             Martin Jambor  <mjambor@suse.cz>
8380         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
8381         the same elements are repeated rather than printing all of them.
8382         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
8383         "repeated" elements.
8384         * read-rtl-function.c (test_loading_repeat): New function.
8385         (read_rtl_function_c_tests): Call test_loading_repeat.
8386         * rtl-tests.c (test_dumping_repeat): New function.
8387         (rtl_tests_c_tests): Call test_dumping_repeat.
8389 2018-10-11  Richard Biener  <rguenther@suse.de>
8391         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
8392         bdver?_cost): Unify to ...
8393         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
8394         * config/i386/i386.c (processor_cost_table): Adjust.
8396 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
8398         PR middle-end/87574
8399         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
8400         the thunk when expanding to GIMPLE.
8402 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8404         PR bootstrap/87551
8405         * varasm.c (mergeable_string_section): Don't try to move zero-length
8406         strings to the merge section.
8408 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
8410         PR target/87573
8411         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
8413 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
8415         PR target/87550
8416         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
8417         to special_args set.
8419 2018-10-10  Richard Biener  <rguenther@suse.de>
8421         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
8422         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
8423         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
8424         and recursing and pattern terminating the recursion on SSE
8425         vector width using ix86_expand_reduc.
8426         (reduc_sminmax_scal_<mode>): Split into part reducing to half
8427         width and recursing and SSE2 vector variant doing the final
8428         reduction with ix86_expand_reduc.
8429         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
8430         with terminating the recursion at AVX level, splitting that
8431         to SSE there.
8433 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
8435         * genmatch.c (error_cb): Rename to...
8436         (diagnostic_cb): ...this, converting int params to enums.
8437         (fatal_at): Update for renaming.
8438         (warning_at): Likewise.
8439         (main): Likewise.
8440         * input.c (selftest::ebcdic_execution_charset::apply):
8441         Update for renaming of...
8442         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
8443         to...
8444         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
8445         converting level and reason to enums.
8446         (class selftest::lexer_error_sink): Rename to...
8447         (class selftest::lexer_test_options): ...this, renaming field
8448         "m_errors" to "m_diagnostics".
8449         (selftest::lexer_test_options::apply): Update for renaming of...
8450         (selftest::lexer_test_options::on_error): ...this, renaming to...
8451         (selftest::lexer_test_options::on_diagnostic): ...this
8452         converting level and reason to enums.
8453         (selftest::test_lexer_string_locations_raw_string_unterminated):
8454         Update for renamings.
8455         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
8456         "reason".
8458 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
8460         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
8461         * config/rs6000/pmmintrin.h: New file.
8463 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
8465         PR tree-optimization/86659
8466         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
8468 2018-10-09  Richard Biener  <rguenther@suse.de>
8470         PR tree-optimization/63155
8471         * tree-ssa-structalias.c: Include tree-ssa.h.
8472         (get_constraint_for_ssa_var): For undefs return nothing_id.
8473         (find_func_aliases): Cleanup PHI handling.
8475 2018-10-09  Richard Biener  <rguenther@suse.de>
8477         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
8478         replacements.
8480 2018-10-09  Martin Liska  <mliska@suse.cz>
8482         * asan.c (asan_emit_stack_protection): If a stack variable
8483         is located in a same file as current function, then emit
8484         line info into variable definition string.
8486 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
8488         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
8489         information.
8491 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
8493         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
8494         on the thunk.
8496 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
8498         PR tree-optimization/86659
8499         * gimple-match.h (struct gimple_match_op): Add reverse field.
8500         (gimple_match_op::set_op): New overloaded method.
8501         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
8502         the REF_REVERSE_STORAGE_ORDER flag on the value.
8503         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
8504         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
8506 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
8508         PR middle-end/63155
8509         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
8510         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
8512 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
8514         PR target/87517
8515         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
8516         Defined with __builtin_ia32_vfmaddsubpd512_mask.
8518 2018-10-08  Richard Biener  <rguenther@suse.de>
8520         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
8521         cost the same as AVX128 ones.
8523 2018-10-08  Paul Koning  <ni1d@arrl.net>
8525         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
8526         (expand_block_move): New function.
8527         * config/pdp11/pdp11.c (output_block_move): Remove.
8528         (expand_block_move): New function.
8529         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
8530         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
8531         (*movmemhi1): Remove.
8533 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
8535         * config/s390/2827.md: Increase latencies for some FP instructions.
8537 2018-10-08  Richard Biener  <rguenther@suse.de>
8539         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8540         Open a dump scope.
8541         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
8542         * tree-vectorizer.h (dump_stmt_cost): Adjust.
8543         (add_stmt_cost): Dump return value of the hook.
8545 2018-10-08  Richard Biener  <rguenther@suse.de>
8547         PR tree-optimization/63155
8548         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
8549         (ssa_propagation_engine::ssa_propagate): Remove redundant
8550         bitmap bit clearing.
8552 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
8554         PR rtl-optimization/86939
8555         PR rtl-optimization/87479
8556         * ira.h (non_conflicting_reg_copy_p): New prototype.
8557         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
8558         (make_hard_regno_dead): Don't add conflicts for register
8559         ignore_reg_for_conflicts.
8560         (make_object_dead): Likewise.
8561         (non_conflicting_reg_copy_p): New function.
8562         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
8563         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
8564         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
8565         (make_hard_regno_dead): Don't add conflicts for register
8566         ignore_reg_for_conflicts.  Remove special conflict handling of
8567         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
8568         check_pic_pseudo_p and update callers.
8569         (mark_pseudo_dead): Don't add conflicts for register
8570         ignore_reg_for_conflicts.
8571         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
8573 2018-10-05  Andrew Waterman  <andrew@sifive.com>
8574             Jim Wilson  <jimw@sifive.com>
8576         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
8577         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
8578         new pattern using HONOR_SNANS that emits one extra instruction.
8580 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
8582         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
8583         patterns): Merge SI and DI patterns to a GPR pattern.
8584         (unnamed define_insn and define_split for record form of that): Merge
8585         to a single define_insn_and_split pattern.
8587 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
8589         PR c++/56856
8590         * input.c (expand_location_to_spelling_point): Add param "aspect"
8591         and use rather than hardcoding LOCATION_ASPECT_CARET.
8592         (get_substring_ranges_for_loc): Handle the case of a single token
8593         within a macro expansion.
8594         * input.h (expand_location_to_spelling_point): Add "aspect" param,
8595         defaulting to LOCATION_ASPECT_CARET.
8597 2018-10-05  Paul Koning  <ni1d@arrl.net>
8599         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
8600         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
8601         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
8602         (pdp11_guard_type): New function.
8604 2018-10-05  Paul Koning  <ni1d@arrl.net>
8606         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
8607         * config/pdp11/pdp11.opt (mfloat32): Remove.
8608         (mfloat64): Remove.
8609         * doc/invoke.texi (pdp11 -mfloat32): Remove:
8610         (pdp11 -mfloat64): Remove.
8612 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
8614         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
8615         (*cmp<mode>_cc_i387): Ditto.
8616         (*cmpu<mode>_cc_i387): Ditto.
8617         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8618         * config/i386/i386.c (ix86_expand_fp_compare): Remove
8619         "scratch" argument.
8620         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
8621         Emit x86_sahf_1 pattern.
8622         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
8623         (ix86_expand_carry_flag_compare): Ditto.
8625 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
8627         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
8628         to reg_or_0_operand.  Add "C" constraint.
8629         (*cmpxf_cc_i387): Ditto.
8630         (*cmp<mode>_i387): Change operand 2 predicate
8631         to nonimm_or_0_operand.  Add "C" constraint.
8632         (*cmp<mode>_cc_i387): Ditto.
8633         (*cmp<mode>_0_i387): Remove insn pattern.
8634         (*cmp<mode>_0_cc_i387): Ditto.
8636 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
8638         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
8639         * config/i386/predicates.md (nonimm_or_0_operand): Rename
8640         from vector_move_operand.  Update all uses.
8642 2018-10-05  Martin Sebor  <msebor@redhat.com>
8644         PR tree-optimization/87490
8645         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
8646         consistently.
8648 2018-10-05  Richard Biener  <rguenther@suse.de>
8650         PR tree-optimization/63155
8651         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
8652         vertical space in dumpfiles.
8653         * tree-ssa-propagate.h
8654         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
8655         * tree-ssa-propagate.c (cfg_blocks_back): New global.
8656         (ssa_edge_worklist_back): Likewise.
8657         (curr_order): Likewise.
8658         (cfg_blocks_get): Remove abstraction.
8659         (cfg_blocks_add): Likewise.
8660         (cfg_blocks_empty_p): Likewise.
8661         (add_ssa_edge): Add to current or next worklist based on
8662         RPO index.
8663         (add_control_edge): Likewise.
8664         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
8665         into ...
8666         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
8667         iteration from CFG and SSA edge worklist so we process
8668         everything in RPO order, prioritizing forward progress
8669         over iteration.
8670         (ssa_prop_init): Allocate new worklists, do not dump
8671         immediate uses.
8672         (ssa_prop_fini): Free new worklists.
8674 2018-10-05  Richard Biener  <rguenther@suse.de>
8676         * tree-core.h (tree_block::abstract_flag): Remove.
8677         (tree_block::block_num): Make full 32bits.
8678         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
8679         * tree.h (BLOCK_ABSTRACT): Remove.
8680         * dwarf2out.c (gen_lexical_block_die): Remove dead code
8681         resulting from BLOCK_ABSTRACT being always false.
8682         (gen_inlined_subroutine_die): Likewise.
8683         (gen_block_die): Likewise.
8684         * tree.c (block_ultimate_origin): Likewise.
8685         * tree-pretty-print.c (dump_block_node): Remove code dealing
8686         with BLOCK_ABSTRACT.
8687         * tree-ssa-live.c (dump_scope_block): Likewise.
8688         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
8689         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
8691 2018-10-05   Richard Biener  <rguenther@suse.de>
8693         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
8694         is asked for initialize mode to the component mode of the
8695         vector type.
8697 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
8699         PR target/87522
8700         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
8701         assembler for -mavx.
8702         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
8704 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
8706         PR target/87509
8707         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
8708         RS6000_BTM_DFP.
8709         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
8710         to be DImode.  When using mffscrn, force the operand to a register.
8712 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
8714         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
8715         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
8716         X87MODEF mode iterator.
8717         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
8718         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
8719         X87MODEF mode iterator.
8721 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
8723         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
8724         -Wno-prio-ctor-dtor.
8726 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
8728         * Makefile.in (OBJS): Add opt-problem.o.
8729         * dump-context.h: Include "selftest.h.
8730         (selftest::temp_dump_context): New forward decl.
8731         (class dump_context): Make friend of class
8732         selftest::temp_dump_context.
8733         (dump_context::dump_loc_immediate): New decl.
8734         (class dump_pretty_printer): Move here from dumpfile.c.
8735         (class temp_dump_context): Move to namespace selftest.
8736         (temp_dump_context::temp_dump_context): Add param
8737         "forcibly_enable_dumping".
8738         (selftest::verify_dumped_text):
8739         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
8740         (selftest::verify_item):
8741         (ASSERT_IS_TEXT): Move here from dumpfile.c.
8742         (ASSERT_IS_TREE): Likewise.
8743         (ASSERT_IS_GIMPLE): Likewise.
8744         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
8745         to...
8746         (dump_context::dump_loc_immediate): ...this new function.
8747         (class dump_pretty_printer): Move to dump-context.h.
8748         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
8749         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
8750         (temp_dump_context::temp_dump_context): Move to "selftest"
8751         namespace.  Add param "forcibly_enable_dumping", and use it to
8752         conditionalize the use of m_pp;
8753         (selftest::verify_dumped_text): Make non-static.
8754         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
8755         (selftest::verify_item): Make non-static.
8756         (ASSERT_IS_TEXT): Move to dump-context.h.
8757         (ASSERT_IS_TREE): Likewise.
8758         (ASSERT_IS_GIMPLE): Likewise.
8759         (selftest::test_capture_of_dump_calls): Pass "true" for new
8760         param of temp_dump_context.
8761         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
8762         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
8763         TDF_COMPARE_DEBUG.
8764         * opt-problem.cc: New file.
8765         * opt-problem.h: New file.
8766         * optinfo-emit-json.cc
8767         (selftest::test_building_json_from_dump_calls): Pass "true" for
8768         new param of temp_dump_context.
8769         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
8770         (optinfo::emit_for_opt_problem): New function.
8771         (optinfo::emit): Clarity which emit_item is used.
8772         * optinfo.h (optinfo::get_dump_location): New accessor.
8773         (optinfo::emit_for_opt_problem): New decl.
8774         (optinfo::emit): Make const.
8775         * selftest-run-tests.c (selftest::run_tests): Call
8776         selftest::opt_problem_cc_tests.
8777         * selftest.h (selftest::opt_problem_cc_tests): New decl.
8778         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
8779         bool to opt_result, converting fprintf messages to
8780         opt_result::failure_at calls.  Add "stmt" param for use by the
8781         failure_at calls.
8782         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
8783         (runtime_alias_check_p): Convert return type from bool to
8784         opt_result, converting dump_printf calls to
8785         opt_result::failure_at, using the statement DDR_A for their
8786         location.
8787         (find_data_references_in_stmt): Convert return type from bool to
8788         opt_result, converting "return false" to opt_result::failure_at
8789         with a new message.
8790         * tree-data-ref.h: Include "opt-problem.h".
8791         (dr_analyze_innermost): Convert return type from bool to opt_result,
8792         and add a const gimple * param.
8793         (find_data_references_in_stmt): Convert return type from bool to
8794         opt_result.
8795         (runtime_alias_check_p): Likewise.
8796         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
8797         dr_analyze_innermost.
8798         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
8799         Convert return type from bool to opt_result, adding a message for
8800         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
8801         (vect_analyze_data_ref_dependence): Convert return type from bool
8802         to opt_result.  Change sense of return type from "false"
8803         effectively meaning "no problems" to "false" meaning a problem,
8804         so that "return false" becomes "return opt_result::success".
8805         Convert "return true" calls to opt_result::failure_at, using
8806         the location of statement A rather than vect_location.
8807         (vect_analyze_data_ref_dependences): Convert return type from bool
8808         to opt_result.
8809         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
8810         calls to opt_result::failure_at, using the stmt location rather
8811         than vect_location.
8812         (vect_verify_datarefs_alignment): Convert return type from bool
8813         to opt_result.
8814         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
8815         into multiple more-tightly-scoped copies.
8816         (vect_analyze_data_refs_alignment): Convert return type from bool
8817         to opt_result.
8818         (vect_analyze_data_ref_accesses): Likewise, converting a
8819         "return false" to a "return opt_result::failure_at", adding a
8820         new message.
8821         (vect_prune_runtime_alias_test_list): Convert return type from
8822         bool to opt_result, converting dump_printf_loc to
8823         opt_result::failure_at.  Add a %G to show the pertinent statement,
8824         and use the stmt's location rather than vect_location.
8825         (vect_find_stmt_data_reference): Convert return type from
8826         bool to opt_result, converting dump_printf_loc to
8827         opt_result::failure_at, using stmt's location.
8828         (vect_analyze_data_refs):  Convert return type from bool to
8829         opt_result.  Convert "return false" to "return
8830         opt_result::failure_at", adding messages as needed.
8831         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
8832         type from bool to opt_result.
8833         (vect_determine_vf_for_stmt): Likewise.
8834         (vect_determine_vectorization_factor): Likewise, converting
8835         dump_printf_loc to opt_result::failure_at, using location of phi
8836         rather than vect_location.
8837         (vect_analyze_loop_form_1): Convert return type from bool to
8838         opt_result, converting dump_printf_loc calls, retaining the use of
8839         vect_location.
8840         (vect_analyze_loop_form): Convert return type from loop_vec_info
8841         to opt_loop_vec_info.
8842         (vect_analyze_loop_operations): Convert return type from bool to
8843         opt_result, converting dump_printf_loc calls, using the location
8844         of phi/stmt rather than vect_location where available.  Convert
8845         various "return false" to "return opt_result::failure_at" with
8846         "unsupported phi" messages.
8847         (vect_get_datarefs_in_loop): Convert return type from bool to
8848         opt_result.  Add a message for the
8849         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
8850         (vect_analyze_loop_2): Convert return type from bool to
8851         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
8852         each "goto again;", adding new messages where needed.
8853         Add "unsupported grouped {store|load}" messages.
8854         (vect_analyze_loop): Convert return type from loop_vec_info to
8855         opt_loop_vec_info.
8856         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
8857         bool to opt_result.
8858         * tree-vect-stmts.c (process_use): Likewise, converting
8859         dump_printf_loc call and using stmt location, rather than
8860         vect_location.
8861         (vect_mark_stmts_to_be_vectorized): Likeise.
8862         (vect_analyze_stmt): Likewise, adding a %G.
8863         (vect_get_vector_types_for_stmt): Convert return type from bool to
8864         opt_result, converting dump_printf_loc calls and using stmt
8865         location, rather than vect_location.
8866         (vect_get_mask_type_for_stmt): Convert return type from tree to
8867         opt_tree, converting dump_printf_loc calls and using stmt location.
8868         * tree-vectorizer.c: Include "opt-problem.h.
8869         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
8870         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
8871         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
8872         enabled, use it to report at the top level "couldn't vectorize
8873         loop" followed by the problem.
8874         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
8875         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
8876         to opt_result.
8877         (vect_analyze_stmt): Likewise.
8878         (vect_get_vector_types_for_stmt): Likewise.
8879         (tree vect_get_mask_type_for_stmt): Likewise.
8880         (vect_analyze_data_ref_dependences): Likewise.
8881         (vect_enhance_data_refs_alignment): Likewise.
8882         (vect_analyze_data_refs_alignment): Likewise.
8883         (vect_verify_datarefs_alignment): Likewise.
8884         (vect_analyze_data_ref_accesses): Likewise.
8885         (vect_prune_runtime_alias_test_list): Likewise.
8886         (vect_find_stmt_data_reference): Likewise.
8887         (vect_analyze_data_refs): Likewise.
8888         (vect_analyze_loop): Convert return type from loop_vec_info to
8889         opt_loop_vec_info.
8890         (vect_analyze_loop_form): Likewise.
8891         (vect_analyze_slp): Convert return type from bool to opt_result.
8893 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
8895         * doc/invoke.texi (-fopt-info): Document new "internals"
8896         sub-option.
8897         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
8898         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
8899         MSG_ALL_KINDS.
8900         (optinfo_verbosity_options): Add "internals".
8901         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
8902         (dump_context::apply_dump_filter_p): New member function.
8903         (dump_context::dump_loc): Use apply_dump_filter_p rather than
8904         explicitly masking the dump_kind.
8905         (dump_context::begin_scope): Increment the scope depth first.  Use
8906         apply_dump_filter_p rather than explicitly masking the dump_kind.
8907         (dump_context::emit_item): Use apply_dump_filter_p rather than
8908         explicitly masking the dump_kind.
8909         (dump_dec): Likewise.
8910         (dump_hex): Likewise.
8911         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
8912         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
8913         (opt_info_switch_p): Update handling of default
8914         MSG_OPTIMIZED_LOCATIONS to cope with default of
8915         MSG_PRIORITY_USER_FACING.
8916         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
8917         masking the dump_kind.
8918         (selftest::test_capture_of_dump_calls): Update test_dump_context
8919         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
8920         than MSG_ALL.  Generalize scope test to be run at all four
8921         combinations of with/without MSG_PRIORITY_USER_FACING and
8922         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
8923         for each of the two values.
8924         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
8925         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
8926         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
8927         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
8928         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
8929         with MSG_PRIORITY_*.
8930         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
8931         dump messages as MSG_PRIORITY_USER_FACING.
8932         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
8933         about the interaction with MSG_PRIORITY_*.
8935 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8937         * varasm.c (output_constant): Add new parameter merge_strings.
8938         Make strings properly zero terminated in merge string sections.
8939         (mergeable_string_section): Don't fail if the last char is non-zero.
8940         (assemble_variable_contents): Handle merge string sections.
8941         (assemble_variable): Likewise.
8942         (assemble_constant_contents): Likewise.
8943         (output_constant_def_contents): Likewise.
8944         (output_constructor_array_range,
8945         output_constructor_regular_field): Adjust call to output_constant.
8946         (output_object_block): Adjust call to assemble_constant_contents
8947         and assemble_variable_contents.
8949 2018-10-04  Martin Liska  <mliska@suse.cz>
8951         PR c/87483
8952         * cgraphunit.c (process_function_and_variable_attributes):
8953         Warn about a function with alias attribute and a body.
8955 2018-10-04  Martin Liska  <mliska@suse.cz>
8957         PR ipa/82625
8958         * multiple_target.c (redirect_to_specific_clone): New function.
8959         (ipa_target_clone): Use it.
8960         * tree-inline.c: Fix comment.
8962 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
8964         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
8965         fields.
8966         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
8967         (gcc::dump_manager::register_pass): New member function, adapted
8968         from loop body in gcc::pass_manager::register_pass, adding a
8969         call to update_dfi_for_opt_info.
8970         (gcc::dump_manager::opt_info_enable_passes): Store the
8971         -fopt-info options into the new fields.  Move the loop
8972         bodies into...
8973         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
8974         function.
8975         * dumpfile.h (struct opt_pass): New forward decl.
8976         (gcc::dump_manager::register_pass): New decl.
8977         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
8978         (class gcc::dump_manager): Add fields "m_optgroup_flags",
8979         "m_optinfo_flags", and "m_optinfo_filename".
8980         * passes.c (gcc::pass_manager::register_pass): Move all of the
8981         dump-handling code to gcc::dump_manager::register_pass.
8983 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
8985         PR rtl-optimization/87466
8986         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
8987         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
8988         * doc/tm.texi: Regenerate.
8989         * ira-lives.c (process_bb_node_lives): Use the new target hook.
8990         * lra-lives.c (process_bb_lives): Likewise.
8991         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
8992         Define.
8994 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
8996         * params.c (add_params): Fix initialization.
8998 2018-10-04  Martin Liska  <mliska@suse.cz>
9000         PR gcov-profile/84107
9001         * tree-profile.c (init_ic_make_global_vars):
9002         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
9003         Come up with new ic_tuple* variables.  Emit
9004         __gcov_indirect_call{,_topn} variables.
9005         (gimple_gen_ic_profiler): Access the variable
9006         and emit gimple.
9007         (gimple_gen_ic_func_profiler): Access
9008         __gcov_indirect_call.callee field.
9009         (gimple_init_gcov_profiler): Use ptr_type_node.
9010         * value-prof.c (gimple_ic): Use ptr_type_node.
9012 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9014         PR tree-optimization/85787
9015         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
9016         malloc_candidate_p into this function and add support for
9017         detecting multiple phis.
9018         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
9020 2018-10-04  Martin Liska  <mliska@suse.cz>
9022         PR ipa/87491
9023         * ipa-inline.c (inline_to_all_callers_1):
9024         Call ultimate_alias_target for node being inlined.
9026 2018-10-03  Jeff Law  <law@redhat.com>
9028         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
9029         target's wchar_t.
9030         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
9031         * tree.h (get_typenode_from_name): Prototype.
9033 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
9035         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
9036         Change operand 2 predicate to nonimmediate_operand.
9037         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
9039 2018-10-03  Martin Sebor  <msebor@redhat.com>
9040             Jeff Law  <law@redhat.com>
9042         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
9043         initialize it.
9044         (get_string_length): Detect unterminated arrays.
9045         (format_string): Same.
9046         (format_directive): Warn about unterminated arrays.
9047         (handle_gimple_call): Mark statements with no_warning as needed.
9049 2018-10-03  Jim Wilson  <jimw@sifive.com>
9051         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
9052         also define __riscv_abi_rve.  Delete trailing white space.
9054 2018-10-03  Paul Koning  <ni1d@arrl.net>
9056         Enable LRA register allocator for PDP11.
9057         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
9058         (R): Likewise.
9059         (D): Likewise.
9060         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
9061         * config/pdp11/pdp11.opt (-mlra): New option.
9062         * doc/invoke.texi (PDP-11 Options): Document -mlra.
9064 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
9066         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
9067         (*<absneg:code>extend<mode>xf2): Ditto.
9069 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
9071         PR tree-optimization/87415
9072         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
9073         precision fields.
9075 2018-10-02  Jeff Law  <law@redhat.com>
9077         * gimple-fold.c (get_range_strlen): Only set *nonstr when
9078         an unterminated string is discovered.  Bubble up range
9079         even for unterminated strings.
9080         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
9081         indicates the string was not terminated via NONSTR.
9083 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
9085         * tree-vrp.c (extract_range_from_unary_expr): Special case all
9086         pointer conversions.
9087         Do not do anything special for anti-ranges.
9089 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
9091         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
9092         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
9094 2018-10-03  Martin Liska  <mliska@suse.cz>
9096         PR gcov-profile/86109
9097         * coverage.c (coverage_begin_function): Do not
9098         mark lambdas as artificial.
9099         * tree-core.h (struct GTY): Remove tm_clone_flag
9100         and introduce new lambda_function.
9101         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
9103 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9105         PR target/87474
9106         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
9107         P8_VECTOR and VSX are enabled.
9109 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
9111         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
9112         0x3907 as CPU model number.
9114 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
9116         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
9117         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
9118         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
9119         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
9120         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
9121         * config/s390/s390.md: Likewise. Rename also the cpu attribute
9122         value from arch12 to z14.
9124 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
9126         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
9127         (isinfxf2): Ditto.
9128         (isinf<mode>2): Ditto.
9130 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
9132         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
9133         before emitting fxam.  Perform calculations in XFmode.
9135 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
9137         * match.pd (((X /[ex] A) +- B) * A): New transformation.
9139 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
9141         PR middle-end/87319
9142         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
9143         * tree.c (signed_or_unsigned_type_for): Handle complex.
9145 2018-10-02  Jeff Law  <law@redhat.com>
9147         * gimple-fold.c (get_range_strlen): Remove dead code.
9149 2018-10-02  Martin Sebor  <msebor@redhat.com>
9150             Jeff Law  <law@redhat.com>
9152         * builtins.c (unterminated_array): Add new arguments.
9153         If argument is not terminated, bubble up size and exact
9154         state to callers.
9155         (expand_builtin_strnlen): Detect, avoid expanding
9156         and diagnose unterminated arrays.
9157         (c_strlen): Fill in offset of start of unterminated strings.
9158         * builtins.h (unterminated_array): Update prototype.
9160 2018-10-02  Richard Biener  <rguenther@suse.de>
9162         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
9163         of haddv4df, first reduce to SSE width and exploit the fact
9164         that we only need element zero with the reduction result.
9165         (reduc_plus_scal_v2df): Likewise.
9167 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
9169         * dojump.h (do_jump): Delete.
9170         (do_jump_1): Likewise.
9171         (split_comparison): Move around.
9172         * dojump.c (do_jump): Make static.
9173         (do_jump_1): Likewise.
9174         (jumpifnot): Move around.
9175         (jumpifnot_1): Likewise.
9176         (jumpif): Likewise.
9177         (jumpif_1): Likewise.
9178         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
9180 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
9182         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
9183         insns in the delay slot and add_insn_after for the jump insn.
9185 2018-10-02  Richard Biener  <rguenther@suse.de>
9187         * tree-inline.c (expand_call_inline): Use the location of
9188         the callee declaration for the inline-entry marker.
9189         * final.c (notice_source_line): Remove special-casing of
9190         NOTE_INSN_INLINE_ENTRY.
9192 2018-10-01  Carl Love  <cel@us.ibm.com>
9194         PR 69431
9195         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
9196         (__builtin_mtfsb0): New.
9197         (__builtin_mtfsb1): New.
9198         ( __builtin_set_fpscr_rn): New.
9199         (__builtin_set_fpscr_drn): New.
9200         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
9201         (rs6000_expand_set_fpscr_rn_builtin): Add.
9202         (rs6000_expand_set_fpscr_drn_builtin): Add.
9203         (rs6000_expand_builtin): Add case statement entries for
9204         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
9205         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
9206         RS6000_BUILTIN_MFFSL.
9207         (rs6000_init_builtins): Add ftype initialization and def_builtin
9208         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
9209         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
9210         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
9211         rs6000_mffscdrn): Add define_insn.
9212         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
9213         * doc/extend.texi: Add documentation for the builtins.
9215 2018-10-01  Richard Biener  <rguenther@suse.de>
9217         PR tree-optimization/87465
9218         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
9219         causing branch miscounts.
9221 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9223         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
9224         aarch64_option_default_param):  New.
9225         (params.h): Include.
9226         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
9227         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
9228         stack-clash protection validation code.
9230 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9232         * params.c (validate_param): New.
9233         (add_params): Use it.
9234         (set_param_value): Refactor param validation into validate_param.
9235         (diagnostic.h): Include.
9236         * diagnostic.h (diagnostic_ready_p): New.
9238 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9240         * params.c (set_param_value):
9241         Add index of parameter being validated.
9242         * common/common-target.def (option_validate_param): New.
9243         * common/common-targhooks.h (default_option_validate_param): New.
9244         * common/common-targhooks.c (default_option_validate_param): New.
9245         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
9246         * doc/tm.texi: Regenerate.
9248 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9250         PR target/86486
9251         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9252         Add validation for stack-clash parameters and set defaults.
9254 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9256         PR target/86486
9257         * configure.ac: Add stack-clash-protection-guard-size.
9258         * doc/install.texi: Document it.
9259         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
9260         * params.def: Update comment for guard-size.
9261         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
9262         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
9263         * configure: Regenerate.
9265 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9267         PR target/86486
9268         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
9269         STACK_DYNAMIC_OFFSET): New.
9270         * config/aarch64/aarch64.c (aarch64_layout_frame):
9271         Update outgoing args size.
9272         (aarch64_stack_clash_protection_alloca_probe_range,
9273         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
9275 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9277         PR target/86486
9278         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
9279         probe ranges.
9280         * target.def (stack_clash_protection_alloca_probe_range): New.
9281         (stack_clash_protection_final_dynamic_probe): Remove.
9282         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
9283         (default_stack_clash_protection_final_dynamic_probe): Remove.
9284         * targhooks.c: Likewise.
9285         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
9286         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
9287         * doc/tm.texi: Regenerate.
9289 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9291         PR target/86486
9292         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
9293         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
9294         aarch64_clamp_to_uimm12_shift): New.
9295         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
9296         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
9298 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
9300         PR target/86486
9301         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
9303 2018-10-01  Jeff Law  <law@redhat.com>
9304             Richard Sandiford  <richard.sandiford@linaro.org>
9305             Tamar Christina  <tamar.christina@arm.com>
9307         PR target/86486
9308         * config/aarch64/aarch64.md
9309         (probe_stack_range): Add k (SP) constraint.
9310         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
9311         STACK_CLASH_MAX_UNROLL_PAGES): New.
9312         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
9313         stack probes for stack clash.
9314         (aarch64_allocate_and_probe_stack_space): New.
9315         (aarch64_expand_prologue): Use it.
9316         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
9317         (aarch64_sub_sp): Add emit_move_imm optional param.
9319 2018-10-01  MCC CS  <deswurstes@users.noreply.github.com>
9321         PR tree-optimization/87261
9322         * match.pd: Remove trailing whitespace.
9323         Add (x & y) | ~(x | y) -> ~(x ^ y),
9324         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
9326 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
9328         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
9329         constraints.
9330         (ashlsi3_insn): Update instruction constraints.
9331         (ashrsi3_insn): Likewise.
9332         (rotrsi3): Likewise.
9333         (add_shift): Likewise.
9334         * config/arc/constraints.md (Csz): New 32 bit constraint. It
9335         avoids placing in the limm field small constants which, otherwise,
9336         could end into a small instruction.
9338 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
9340         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
9341         destination register is not odd-even.
9342         (umaddsidi4_split): Likewise.
9344 2018-10-01  Richard Biener  <rguenther@suse.de>
9346         * tree-inline.c (expand_call_inline): Store origin of fn
9347         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
9348         * tree.c (block_ultimate_origin): Simplify and do some
9349         checking.
9351 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
9353         * config/i386/mmx.md (EMMS): New int iterator.
9354         (emms): New int attribute.
9355         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
9356         EMMS int iterator.  Explicitly declare clobbers.
9357         (mmx_emms): Remove expander.
9358         (mmx_femms): Ditto.
9359         * config/i386/predicates.md (emms_operation): Remove predicate.
9360         (vzeroall_pattern): New predicate.
9361         (vzeroupper_pattern): Rename from vzeroupper_operation.
9362         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
9363         vzeroupper_pattern and vzeroall_pattern predicates.
9365 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
9367         PR rtl-optimization/86939
9368         * ira-lives.c (make_hard_regno_born): Rename from this...
9369         (make_hard_regno_live): ... to this.  Remove update to conflict
9370         information.  Update function comment.
9371         (make_hard_regno_dead): Add conflict information update.  Update
9372         function comment.
9373         (make_object_born): Rename from this...
9374         (make_object_live): ... to this.  Remove update to conflict information.
9375         Update function comment.
9376         (make_object_dead):  Add conflict information update.  Update function
9377         comment.
9378         (mark_pseudo_regno_live): Call make_object_live.
9379         (mark_pseudo_regno_subword_live): Likewise.
9380         (mark_hard_reg_dead): Update function comment.
9381         (mark_hard_reg_live): Call make_hard_regno_live.
9382         (process_bb_node_lives): Likewise.
9383         * lra-lives.c (make_hard_regno_born): Rename from this...
9384         (make_hard_regno_live): ... to this.  Remove update to conflict
9385         information.  Remove now uneeded check_pic_pseudo_p argument.
9386         Update function comment.
9387         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
9388         to conflict information.  Update function comment.
9389         (mark_pseudo_live): Remove update to conflict information.  Update
9390         function comment.
9391         (mark_pseudo_dead): Add conflict information update.
9392         (mark_regno_live): Call make_hard_regno_live.
9393         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
9394         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
9396 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
9398         PR target/87370
9399         * config/i386/i386.c (construct_container): Use TImode for
9400         BLKmode values in 2 integer registers.
9402 2018-09-29  Jeff Law  <law@redhat.com>
9404         * builtins.c (unterminated_array): Pass in c_strlen_data * to
9405         c_strlen rather than just a tree *.
9406         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
9407         Update recursive calls appropriately.  If caller did not provide a
9408         suitable data pointer, create a local one.  When a non-terminated
9409         string is discovered, bubble up information about the string via the
9410         c_strlen_data object.
9411         * builtins.h (c_strlen): Update prototype.
9412         (c_strlen_data): New structure.
9413         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
9414         For a type 2 call, if c_strlen indicates a non-terminated string
9415         use the length of the non-terminated string.
9416         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
9418 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
9420         PR target/87467
9421         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
9422         __m512d type for __A argument rather than __m512.
9424 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
9426         * match.pd (simple_comparison): Don't optimize if either operand is
9427         a function pointer when target needs function pointer canonicalization.
9429 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
9431         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
9432         power5 .. power9 to remove indirection.
9433         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
9434         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
9435         ASM_CPU_476_SPEC): Delete.
9436         (ASM_CPU_SPEC): Adjust.
9437         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
9438         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
9440 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
9442         * config.in: Delete HAVE_AS_DCI.
9443         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
9444         * config/rs6000/rs6000.h: Ditto.
9445         * configure.ac: Delete HAVE_AS_DCI.
9446         * configure: Regenerate.
9448 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
9450         * config.in (HAVE_AS_LWSYNC): Delete.
9451         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9452         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
9453         do it as a .long .
9454         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9455         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
9456         as a .long .
9457         * configure.ac: Delete HAVE_AS_LWSYNC.
9458         * configure: Regenerate.
9460 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
9461             Pierre-Marie de Rodat  <derodat@adacore.com>
9463         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
9464         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
9465         (cgraph_node::create_thunk): Add indirect_offset parameter.
9466         (thunk_adjust): Likewise.
9467         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
9468         and initialize the corresponding field with it.
9469         (cgraph_node::dump): Dump indirect_offset field.
9470         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
9471         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
9472         (thunk_adjust): Add indirect_offset parameter and deal with it.
9473         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
9474         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
9475         or if the thunk is external or local.  Fix formatting.  Do not chain
9476         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
9477         if any, in the GIMPLE representation.
9478         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
9479         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
9480         (input_node): Read indirect_offset field.
9481         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
9482         call to thunk_adjust.
9483         * tree-nested.c (struct nesting_info): Add thunk_p field.
9484         (create_nesting_tree): Set it.
9485         (convert_all_function_calls): Copy static chain from targets to thunks.
9486         (finalize_nesting_tree_1): Return early for thunks.
9487         (unnest_nesting_tree_1): Do not finalize thunks.
9488         (gimplify_all_functions): Do not gimplify thunks.
9490 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
9492         * opt-suggestions.c (option_proposer::build_option_suggestions):
9493         Release "option_values".
9495 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
9497         * coverage.c (get_coverage_counts): Convert problem-reporting dump
9498         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
9499         * dumpfile.c (kind_as_string): New function.
9500         (dump_loc): Rather than a hardcoded prefix of "note: ", use
9501         kind_as_string to vary the prefix based on dump_kind.
9502         (selftest::test_capture_of_dump_calls): Update for above.
9504 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
9506         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
9507         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
9509 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
9511         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
9512         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
9513         INVALID_REGNUM instead of FPSR_REG.
9514         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
9515         * config/i386/i386.md: Update comment of FP compares.
9516         (fldenv): Do not clobber FPSR_REG.
9518 2018-09-28  Richard Biener  <rguenther@suse.de>
9520         * tree.h (BLOCK_ORIGIN): New.
9521         * omp-expand.c (grid_expand_target_grid_body): Assign
9522         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
9523         * tree-inline.c (remap_block): Likewise.
9524         * auto-profile.c (get_function_decl_from_block): Simplify
9525         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
9526         * langhooks.c (lhd_print_error_function): Likewise.
9527         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
9528         Likewise.
9529         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9530         * tree.c (block_nonartificial_location): Likewise.
9531         (block_ultimate_origin): Likewise.
9532         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
9533         no longer needed LTO case.
9535 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
9536             Jan Hubicka  <jh@suse.cz>
9537             Martin Jambor  <mjambor@suse.cz>
9539         * simplify-rtx.c (simplify_merge_mask): New function.
9540         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
9541         same masks are used in op1 or op2.
9542         (test_vec_merge): New function.
9543         (test_vector_ops): Call test_vec_merge.
9545 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
9547         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
9548         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
9549         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
9550         (cypress_costs): Set it.
9551         (supersparc_costs): Likewise.
9552         (hypersparc_costs): Likewise.
9553         (leon_cost): Likewise.
9554         (leon3_costs): Likewise.
9555         (sparclet_costs): Likewise.
9556         (ultrasparc_costs): Likewise.
9557         (ultrasparc_costs): Likewise.
9558         (niagara_costs): Likewise.
9559         (niagara2_costs): Likewise.
9560         (niagara3_costs): Likewise.
9561         (niagara4_costs): Likewise.
9562         (niagara7_costs): Likewise.
9563         (m8_costs): Likewise.
9564         (TARGET_CAN_FOLLOW_JUMP): Define.
9565         (pass_work_around_errata::gate): Minor tweak.
9566         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
9567         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
9568         Automaitcally clear MASK_FSMULD mask for V7 processors.
9569         (sparc_can_follow_jump): New static function.
9570         (output_ubranch): Deal with CROSSING_JUMP_P.
9571         (sparc_use_sched_lookahead): Rewrite using switch statement.
9572         (sparc_issue_rate): Reorder.
9573         (sparc_branch_cost): New function.
9575 2018-09-27  Martin Sebor  <msebor@redhat.com>
9577         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
9578         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
9579         (int_fits_type_p): Same.
9581 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
9583         * config/i386/i386.md (FPCR_REG): Remove.
9584         (UNSPEC_FLDCW): Remove.
9585         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
9586         (x86_fldcw_1): Remove insn pattern.
9587         (fnstenv): Do not clobber FPCR_REG.
9588         (fldenv): Ditto.
9589         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
9590         (CALL_USED_REGISTERS): Ditto.
9591         (REG_ALLOC_ORDER): Ditto.
9592         (REG_CLASS_CONTENTS): Ditto.
9593         (HI_REGISTER_NAMES): Ditto.
9594         (ADDITIONAL_REGISTER_NAMES): Use defines instead
9595         of numerical constants.
9596         * config/i386/i386.c (regclass_map): Remove fpsr register.
9597         (dbx_register_map): Ditto.
9598         (dbx64_register_map): Ditto.
9599         (svr4_dbx_register_map): Ditto.
9600         (print_reg): Do not handle FPCR_REG.
9602 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
9604         PR target/87149
9605         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
9606         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
9607         Delete, always treat as true.
9608         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9609         Ditto.  Simplify remaining code.
9610         * config/powerpcspe/powerpcspe.h: Ditto.
9611         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
9612         Simplify remaining code.
9613         (rs6000_expand_builtin): Ditto.
9614         * config/rs6000/rs6000.h: Ditto.
9615         * configure.ac: Ditto.
9616         * configure: Regenerate.
9618 2018-09-27  Martin Liska  <mliska@suse.cz>
9620         * coverage.c (get_coverage_counts): Revert the formatting
9621         of missing profile opt info.
9623 2018-09-27  Richard Biener  <rguenther@suse.de>
9625         PR debug/37801
9626         PR debug/87440
9627         * dwarf2out.c (set_block_origin_self): Do not mark outermost
9628         block as we do not output that.
9629         (gen_inlined_subroutine_die): Elide the originally outermost
9630         block, matching what we do for concrete instances.
9631         (decls_for_scope): Add parameter specifying whether to recurse
9632         to subblocks.
9634 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
9635             Tom de Vries  <tom@codesourcery.com>
9637         PR 82089
9639         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
9640         STORE_FLAG_VALUE == 1.
9642 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
9644         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
9645         constant definitions.
9646         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
9647         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
9648         ("speculation_barrier"): New expander definition.
9650 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
9652         PR gcov-profile/86957
9653         * common.opt: New warning option -Wmissing-profile.
9654         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
9655         * doc/invoke.texi: Document -Wmissing-profile.
9657 2018-09-26  Jim Wilson  <jimw@sifive.com>
9659         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
9660         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
9662 2018-09-26  Martin Sebor  <msebor@redhat.com>
9664         * tree.c (zerop): Change return type to bool.
9665         (integer_zerop, integer_onep, integer_each_onep): Same.
9666         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9667         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9668         (real_onep, real_minus_onep, chain_index): Same.
9669         (print_type_hash_statistics, type_list_equal): Same.
9670         * tree.h (zerop): Same.
9671         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
9672         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9673         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9674         (real_onep, real_minus_onep, chain_index): Same.
9675         (print_type_hash_statistics, type_list_equal): Same.
9677 2018-09-26  Jim Wilson  <jimw@sifive.com>
9679         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
9681 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
9683         PR target/87414
9684         * config/i386/i386.c: Include debug.h and dwarf2out.h.
9685         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
9686         call.
9688 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
9690         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
9692 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
9694         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
9695         and FP_SECOND_SSE_REGS.
9696         (REG_CLASS_NAMES): Ditto.
9697         (REG_CLASS_CONTENTS): Ditto.
9698         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
9699         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
9700         (ix86_preferred_output_reload_class): Ditto.
9701         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
9702         clobber constraint to "=&f".
9703         (fix_truncdi_i387): Ditto.
9704         (lrintxfdi2): Ditto.
9705         (fistdi2_<rounding>): Ditto.
9706         (fpremxf4_i387): Change "=u" constraint to "=f".
9707         (fprem1xf4_i387): Ditto.
9708         (sincosxf3): Ditto.
9709         (fptanxf4_i387): Ditto.
9710         (fxtractxf3_i387): Ditto.
9711         (fscalexf4_i387): Ditto.
9712         (atan2xf3): Change "u" constraint to "f".
9713         (fyl2xxf3_i387): Ditto.
9714         (fyl2xp1xf3_i387): Ditto.
9716 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
9718         PR target/87439
9719         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
9720         for removed I387_MASK_PM entity.
9723 2018-09-26  Jeff Law  <law@redhat.com>
9724         Revert
9725         2018-09-26  Alexey Neyman  <stilor@att.net>
9727         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9728         headers are no longer pulled in by <isl/val.h>.
9730 2018-09-26  Richard Biener  <rguenther@suse.de>
9732         PR debug/87443
9733         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
9734         or concrete instance DIE to the tree.  Create abstract origin
9735         attributes also for concrete instances.
9737 2018-09-26  Alexey Neyman  <stilor@att.net>
9739         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9740         headers are no longer pulled in by <isl/val.h>.
9742 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
9744         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
9745         Use new helper functions.
9746         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
9747         Use new helper functions.
9748         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
9749         aarch_mm_needs_release): New declarations.
9750         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
9751         aarch_mm_needs_release): New.
9753 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
9755         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
9756         (arm32_output_mi_thunk): Deal with long calls.
9758 2018-09-26  Richard Biener  <rguenther@suse.de>
9760         PR debug/87428
9761         PR debug/87362
9762         * tree-inline.c (expand_call_inline): When the location
9763         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
9764         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
9765         the inserted BLOCK to make inlined_function_outer_scope_p
9766         recognize it.
9767         * dwarf2out.c (add_call_src_coords_attributes): Do not add
9768         coords for reserved locations.
9770 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
9772         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
9773         (*call_indirect_nonlocal_sysv<mode>): Ditto.
9774         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
9775         (*sibcall_nonlocal_sysv<mode>): Ditto.
9776         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
9777         (<bd>_<mode>): Ditto.
9778         (<bd>tf_<mode>): Ditto.
9780 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
9782         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
9783         control string as a list of templates instead of as C code.
9784         (*altivec_movti): Ditto.
9785         * config/rs6000/darwin.md (movdf_low_di): Ditto.
9787 2018-09-25  Jim Wilson  <jimw@sifive.com>
9789         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
9790         when target symbol is weak.
9792 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9794         PR c/87387
9795         * builtins.c (unterminated_array): Simplify.
9796         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
9797         where pointer arithmetic is safe.
9799 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
9801         PR target/86987
9802         * config/rs6000/altivec.md (altivec_vspltb): Use
9803         const_0_to_15_operand instead of u5bit_cint_operand.
9804         (*altivec_vspltb_internal): Ditto.
9805         (altivec_vspltb_direct): Ditto.
9806         (altivec_vsplth): Use const_0_to_7_operand instead of
9807         u5bit_cint_operand.
9808         (*altivec_vsplth_internal): Ditto.
9809         (altivec_vsplth_direct): Ditto.
9810         (altivec_vspltw): Use const_0_to_3_operand instead of
9811         u5bit_cint_operand.
9812         (*altivec_vspltw_internal): Ditto.
9813         (altivec_vspltw_direct): Ditto.
9814         (altivec_vspltsf): Ditto.
9815         (*altivec_vspltsf_internal): Ditto.
9816         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
9817         various splats with the proper size immediate.  Reorder the various
9818         cases by ascending size of immediate, and put all such together.
9820 2018-09-25  Richard Biener  <rguenther@suse.de>
9822         PR debug/83941
9823         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
9824         GC-ification.
9825         (maybe_create_die_with_external_ref): Do not create
9826         DW_TAG_imported_unit here.
9827         (add_abstract_origin_attribute): Handle external BLOCK refs.
9828         (dwarf2out_abstract_function): Simplify LTO case.
9829         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
9830         rather than using maybe_create_die_with_external_ref.
9832 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
9834         PR target/71278
9835         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
9836         (frndintxf2_mask_pm_i387): Ditto.
9837         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
9838         Enable for !flag_trapping_math.
9839         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
9840         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
9841         Change operand 1 predicate to nonimmediate_operand.
9842         (attr "i387_cw"): Remove mask_pm.
9843         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
9844         (enum ix86_entity): Remove I387_MASK_PM.
9845         * config/i386/i386.c (ix86_i387_mode_needed): Do not
9846         handle I387_MASK_PM.
9847         (ix86_mode_needed): Ditto.
9848         (ix86_mode_after): Ditto.
9849         (ix86_mode_entry): Ditto.
9850         (ix86_mode_exit): Ditto.
9851         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
9853 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
9855         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
9856         to_update_switch_stmts to vNULL instead of calling create on them
9857         immediately.
9859 2018-09-25  Richard Biener  <rguenther@suse.de>
9861         PR tree-optimization/87402
9862         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
9863         (visit_phi): Re-instantiate handling of supposed to be VARYING
9864         but non-VARYING backedge value.
9866 2018-09-25  Richard Biener  <rguenther@suse.de>
9868         PR debug/83941
9869         * dwarf2out.c (struct sym_off_pair): New.
9870         (external_die_map): New global.
9871         (lookup_decl_die): When in LTO create DIEs lazily from the
9872         external_die_map.
9873         (lookup_block_die): New function, create DIEs lazily in LTO.
9874         (equate_block_to_die): New function.
9875         (dwarf2out_die_ref_for_decl): During WPA get the association
9876         from the external DIE map.
9877         (dwarf2out_register_external_die): Record mapping into the
9878         external DIE map.
9879         (maybe_create_die_with_external_ref): New function split out from
9880         DIE generation part of old dwarf2out_register_external_die.
9881         (add_abstract_origin_attribute): Do not return the DIE.  When
9882         in LTO reference externals directly.
9883         (dwarf2out_abstract_function): When in LTO ignore calls for
9884         decls with external DIEs (already present abstract instances).
9885         (gen_call_site_die): Adjust.
9886         (add_high_low_attributes): Likewise.
9887         (gen_lexical_block_die): Likewise.
9888         (gen_inlined_subroutine_die): Likewie.
9889         (gen_block_die): Likewise.
9890         (dwarf2out_inline_entry): Likewise.
9891         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
9892         DIEs.
9894 2018-09-25  Martin Liska  <mliska@suse.cz>
9896         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
9897         integers and not by a float value.
9899 2018-09-25  Martin Liska  <mliska@suse.cz>
9901         PR fortran/87394
9902         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
9903         instead of NULL.
9904         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
9905         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
9906         * godump.c (go_format_type): Remove extra parenthesis.
9908 2018-09-25  Martin Liska  <mliska@suse.cz>
9910         * alias.c (set_dest_equal_p): Remove unused function.
9911         * config/i386/i386.c (def_builtin_pure2): Likewise.
9912         * diagnostic-show-locus.c (class layout): Remove
9913         unused field.
9914         (layout::layout): Likewise here.
9915         * dump-context.h (class temp_dump_context): Likewise.
9916         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
9917         (add_AT_loclistsptr): Likewise.
9918         (add_AT_offset): Likewise.
9919         (get_AT_hi_pc): Likewise.
9920         (is_comdat_die): Likewise.
9921         (type_is_enum): Likewise.
9922         (ceiling): Likewise.
9923         (add_AT_vms_delta): Likewise.
9924         (is_class_die): Likewise.
9925         * edit-context.c (class line_event): Remove unused field.
9926         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
9927         unused function.
9928         * ipa-cp.c (ipa_get_vr_lat): Likewise.
9929         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
9930         (ok_for_base_p_nonstrict): Likewise.
9931         * tree-chrec.c (is_not_constant_evolution): Likewise.
9932         (chrec_fold_poly_cst): Likewise.
9933         * tree-if-conv.c (has_pred_critical_p): Likewise.
9934         * tree-ssa-coalesce.c (print_exprs): Likewise.
9935         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
9936         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
9937         * tree-vrp.c (value_ranges_intersect_p): Likewise.
9938         (value_range_nonnegative_p): Likewise.
9940 2018-09-25  Martin Liska  <mliska@suse.cz>
9942         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
9943         Do not handle "GNU Pascal".
9944         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
9945         Likewise.
9946         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
9947         from documentation. Likewise.
9948         * dbxout.c (dbxout_range_type): Likewise.
9949         * doc/cpp.texi: Likewise.
9950         * doc/extend.texi: Likewise.
9951         * doc/frontends.texi: Likewise.
9952         * doc/invoke.texi: Remove Pascal entry.
9953         * tree.def (CLEANUP_POINT_EXPR): Likewise.
9954         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
9956 2018-09-25  Martin Liska  <mliska@suse.cz>
9958         PR middle-end/86078
9959         * doc/invoke.texi: Document all parameters and remove default
9960         of the parameters.
9962 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
9964         PR bootstrap/87417
9965         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
9966         contains HOST_WIDE_INTs when computing its size.
9968 2018-09-24  Jim Wilson  <jimw@sifive.com>
9970         PR target/87391
9971         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
9972         not TARGET_RVE.
9973         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
9975 2018-09-24  Andrew Pinski  <apinski@marvell.com>
9977         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
9978         access prev before checking it for NULLness in the
9979         AARCH64_FUSE_CMP_BRANCH case.
9981 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
9983         PR target/82699
9984         * config/i386/i386.c (rest_of_insert_endbranch): Set
9985         endbr_queued_at_entrance to true and don't insert ENDBR if
9986         x86_function_profiler will be called.
9987         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
9988         is true.
9989         * config/i386/i386.h (machine_function): Add
9990         endbr_queued_at_entrance.
9992 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
9994         * genattrtab.c (mk_attr_alt): Use alternative_mask.
9995         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
9996         types.
9997         (check_attr_test): Use alternative_mask.
9998         (get_attr_value): Likewise.
9999         (compute_alternative_mask): Use alternative_mask and XWINT.
10000         (make_alternative_compare): Use alternative_mask.
10001         (attr_alt_subset_p): Use XWINT.
10002         (attr_alt_subset_of_compl_p): Likewise.
10003         (attr_alt_intersection): Use alternative_mask and XWINT.
10004         (attr_alt_union): Likewise.
10005         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
10006         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
10007         (simplify_test_exp): Use alternative_mask and XWINT.
10008         (write_test_expr): Use alternative_mask and XWINT, adjust bit
10009         number calculation to support 64 bits.  Generate code that
10010         checks 64-bit masks.
10011         (main): Use alternative_mask.
10012         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
10014 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
10016         PR target/80080
10017         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
10018         RETURN+USE when returning via %r14.
10020 2018-09-24  Martin Liska  <mliska@suse.cz>
10022         * gcov.c (output_lines): Print colorization legend
10023         for both flag_use_colors and flag_use_hotness_colors.
10024         Reword the help.
10026 2018-09-24  Martin Liska  <mliska@suse.cz>
10028         * coverage.c (get_coverage_counts): Use warning_at
10029         with current_function_decl location. Use %qD in warning
10030         message.
10032 2018-09-24  Martin Liska  <mliska@suse.cz>
10034         * memory-block.h (memory_block_pool::release): Annotate with
10035         valgrind that the memory is not accessible.
10037 2018-09-24  Martin Liska  <mliska@suse.cz>
10039         PR sanitizer/85774
10040         * asan.c: Make asan_handled_variables extern.
10041         * asan.h: Likewise.
10042         * cfgexpand.c (expand_stack_vars): Make sure
10043         a representative is unpoison if another
10044         variable in the partition is handled by
10045         use-after-scope sanitization.
10047 2018-09-24  Richard Biener  <rguenther@suse.de>
10049         PR tree-optimization/63155
10050         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
10051         the worklist when the edge of the respective argument isn't
10052         executable.
10054 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
10056         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
10057         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
10058         (MASK_CLASS_P): Update for rename.
10059         (MAYBE_MASK_CLASS_P): Ditto.
10060         (REG_CLASS_NAMES): Update.
10061         (REG_CLASS_CONTENT): Update.
10062         * config/i386/i386.c (regclass_map): Update for MASK_REG
10063         and ALL_MASK_REGS rename.
10064         * config/i386/constraints.md (Yk): Update for rename.
10065         (k): Ditto.
10067 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
10069         * config/i386/i386.h (enum reg_class): Remove
10070         EVEX_SSE_REGS and MOD4_SSE_REGS.
10071         (REG_CLASS_NAMES): Update.
10072         (REG_CLASS_CONTENT): Update.
10073         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
10074         registers as ALL_SSE_REGS.
10075         (ix86_additional_allocno_class_p): Remove.
10076         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
10077         (ix86_register_priority): Lower priority of EVEX SSE registers.
10078         Use IN_RANGE macro where appropriate.
10079         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
10080         AVX-5124VNNIW checks.
10081         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
10082         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
10083         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
10084         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
10085         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
10086         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
10087         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
10088         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
10089         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
10090         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
10091         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
10092         * config/i386/constraints.md (Yh): Remove.
10094 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
10096         * config/i386/i386.c (regclass_map): Declare integer REX registers
10097         as GENERAL_REGS.
10099 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
10101         * doc/service.texi (Service): Switch the fsf.org link to https.
10103 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
10105         PR target/86798
10106         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10107         Define to speculation_safe_value_not_needed.
10109 2018-09-21  Florian Weimer  <fweimer@redhat.com>
10111         PR middle-end/81035
10112         * doc/extend.texi (Common Function Attributes): Mention that
10113         noreturn suppresses tail call optimization.
10115 2018-09-21  Jeff Law  <law@redhat.com>
10117         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
10118         vr_values::cleanup_edges_and_switches.
10119         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
10120         vr_values class.
10121         (identify_jump_threads): Remove EDGE_IGNORE handling.
10122         (execute_vrp): Move handling of to_remove_edges and
10123         to_update_switch_stmts into vr_values class member functions.
10124         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
10125         (to_update_switch_stmts): Likewise.
10126         * vr-values.c: Include cfghooks.h.
10127         (vr_values::vr_values): Initialize to_remove_edges and
10128         to_update_switch_stmts.
10129         (vr_values::~vr_values): Verify to_remove_edges and
10130         to_update_switch_stmts are empty.
10131         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
10132         (vr_values::cleanup_edges_and_switches): New member function.
10133         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
10134         function.  Add new data members.
10136 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
10138         PR tree-optimization/87309
10139         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
10140         calls with pflags and alt_flags.
10141         (selftest::test_capture_of_dump_calls): Add test of interaction of
10142         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
10144 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10146         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
10148 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10150         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
10152 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10154         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
10155         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
10156         (VXWORKS_LIBS_RTP): Minor reordering.
10158 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10160         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
10161         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
10163 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10165         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
10166         (PTRDIFF_TYPE): Likewise.
10168 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10170         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
10171         triplet, similar to support for VxWorks7.
10172         * config/vxworks-dummy.h: Provide a default definition
10173         of TARGET_VXWORKS64 to 0.
10175 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10177         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
10178         * config/vxworks-dummy.h: here.
10180 2018-09-21  Olivier Hainque  <hainque@adacore.com>
10182         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
10184 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
10185             Bo Zhou  <zbo.zhou@hisilicon.com>
10187         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
10188         * config/aarch64/aarch64-tune.md: Regenerated.
10189         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
10190         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
10191         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
10193 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
10194             Julian Brown  <julian@codesourcery.com>
10196         * builtins.c (get_builtin_sync_mem): Handle address spaces.
10198 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
10200         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
10201         if the call takes a static chain.
10203 2018-09-21  Martin Liska  <mliska@suse.cz>
10205         * auto-profile.c (autofdo_source_profile::read): Do not
10206         set sum_all.
10207         (read_profile): Do not add working sets.
10208         (read_autofdo_file): Remove sum_all.
10209         (afdo_callsite_hot_enough_for_early_inline): Remove const
10210         qualifier.
10211         * coverage.c (struct counts_entry): Remove gcov_summary.
10212         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
10213         do not support GCOV_TAG_PROGRAM_SUMMARY.
10214         (get_coverage_counts): Remove summary and expected
10215         arguments.
10216         * coverage.h (get_coverage_counts): Likewise.
10217         * doc/gcov-dump.texi: Remove -w option.
10218         * gcov-dump.c (dump_working_sets): Remove.
10219         (main): Do not support '-w' option.
10220         (print_usage): Likewise.
10221         (tag_summary): Likewise.
10222         * gcov-io.c (gcov_write_summary): Do not dump
10223         histogram.
10224         (gcov_read_summary): Likewise.
10225         (gcov_histo_index): Remove.
10226         (gcov_histogram_merge): Likewise.
10227         (compute_working_sets): Likewise.
10228         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
10229         it not obsolete.
10230         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
10231         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
10232         (GCOV_HISTOGRAM_SIZE): Remove.
10233         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
10234         (struct gcov_summary): Simplify rapidly just
10235         to runs and sum_max fields.
10236         (gcov_histo_index): Remove.
10237         (NUM_GCOV_WORKING_SETS): Likewise.
10238         (compute_working_sets): Likewise.
10239         * gcov-tool.c (print_overlap_usage_message): Remove
10240         trailing empty line.
10241         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
10242         (output_lines): Remove program related line.
10243         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
10244         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
10245         histogram.
10246         (input_profile_summary): Do not read it.
10247         (merge_profile_summaries): And do not merge it.
10248         (input_symtab): Do not call removed function.
10249         * modulo-sched.c (sms_schedule): Do not print sum_max.
10250         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
10251         removed when histogram method was invented.
10252         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
10253         mode.
10254         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
10255         GCOV coding style.
10256         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
10257         and dump selected value.
10258         * profile.c (add_working_set): Remove.
10259         (get_working_sets): Likewise.
10260         (find_working_set): Likewise.
10261         (get_exec_counts): Do not work with working sets.
10262         (read_profile_edge_counts): Do not inform as sum_max is removed.
10263         (compute_branch_probabilities): Likewise.
10264         (compute_value_histograms): Remove argument for call of
10265         get_coverage_counts.
10266         * profile.h: Do not make gcov_summary const.
10268 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
10270         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
10272 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
10274         PR tree-optimization/86990
10275         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
10276         Check that the entire merged store group is made of constants only for
10277         overlapping stores.
10279 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
10281         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
10282         (VTABLE_VERIFICATION_SPEC): Likewise.
10283         (SANITIZER_EARLY_SPEC): Likewise.
10284         (SANITIZER_SPEC): Likewise.
10285         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
10286         * doc/invoke.texi (Link Options): Document -r.
10288 2018-09-20  Richard Biener  <rguenther@suse.de>
10290         PR middle-end/87054
10291         * gimplify.c (gimplify_expr): Retain alignment of
10292         addressable lvalue in dereference.
10294 2018-09-20  Alexandre Oliva  <aoliva@redhat.com>
10296         PR bootstrap/87013
10297         * configure.ac: Check for .loc is_stmt support.
10298         * configure, config.in: Rebuilt.
10299         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
10300         if not supported.
10302 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
10304         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
10305         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
10306         -misel=no.
10308 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
10310         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
10311         VECTOR_OTHER.
10312         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
10313         case VECTOR_OTHER.
10315 2018-09-20  Marek Polacek  <polacek@redhat.com>
10317         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
10319 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
10321         PR tree-optimization/87288
10322         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
10323         into account when determining PEELING_FOR_NITERS.
10325 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
10327         PR tree-optimization/86877
10328         * tree-vect-loop.c (vect_analyze_loop_2): Call
10329         vect_verify_datarefs_alignment.
10331 2018-09-19  Marek Polacek  <polacek@redhat.com>
10333         * doc/invoke.texi: Document -Wclass-conversion.
10335 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
10337         * config/pa/pa.c (pa_adjust_priority): Delete.
10338         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
10340         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
10341         (atomic_storehi): Likewise.
10342         (atomic_storesi): Likewise.
10343         (atomic_loaddi): Restore compare and swap exchange loop code.
10345 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
10347         PR rtl-optimization/86902
10348         * combine.c (try_combine): When changing the CC mode used, don't change
10349         an unrelated mode in other_insn to that new CC mode.
10351 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
10353         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
10354         with %T in place of calls to dump_generic_expr.
10355         (prune_runtime_alias_test_list): Likewise.
10356         (create_runtime_alias_checks): Likewise.
10357         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
10358         (vect_analyze_data_ref_dependence): Likewise.
10359         (vect_slp_analyze_data_ref_dependence): Likewise.
10360         (vect_record_base_alignment): Likewise.  Use %G in place of call
10361         to dump_gimple_stmt.
10362         (vect_compute_data_ref_alignment): Likewise.
10363         (verify_data_ref_alignment): Likewise.
10364         (vect_find_same_alignment_drs): Likewise.
10365         (vect_analyze_group_access_1): Likewise.
10366         (vect_analyze_data_ref_accesses): Likewise.
10367         (dependence_distance_ge_vf): Likewise.
10368         (dump_lower_bound): Likewise.
10369         (vect_prune_runtime_alias_test_list): Likewise.
10370         (vect_find_stmt_data_reference): Likewise.
10371         (vect_analyze_data_refs): Likewise.
10372         (vect_create_addr_base_for_vector_ref): Likewise.
10373         (vect_create_data_ref_ptr): Likewise.
10374         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
10375         (vect_can_advance_ivs_p): Likewise.
10376         (vect_update_ivs_after_vectorizer): Likewise.
10377         (vect_gen_prolog_loop_niters): Likewise.
10378         (vect_prepare_for_masked_peels): Likewise.
10379         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10380         (vect_determine_vectorization_factor): Likewise.
10381         (vect_is_simple_iv_evolution): Likewise.
10382         (vect_analyze_scalar_cycles_1): Likewise.
10383         (vect_analyze_loop_operations): Likewise.
10384         (report_vect_op): Likewise.
10385         (vect_is_slp_reduction): Likewise.
10386         (check_reduction_path): Likewise.
10387         (vect_is_simple_reduction): Likewise.
10388         (vect_create_epilog_for_reduction): Likewise.
10389         (vect_finalize_reduction:): Likewise.
10390         (vectorizable_induction): Likewise.
10391         (vect_transform_loop_stmt): Likewise.
10392         (vect_transform_loop): Likewise.
10393         (optimize_mask_stores): Likewise.
10394         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
10395         (vect_split_statement): Likewise.
10396         (vect_recog_over_widening_pattern): Likewise.
10397         (vect_recog_average_pattern): Likewise.
10398         (vect_determine_min_output_precision_1): Likewise.
10399         (vect_determine_precisions_from_range): Likewise.
10400         (vect_determine_precisions_from_users): Likewise.
10401         (vect_mark_pattern_stmts): Likewise.
10402         (vect_pattern_recog_1): Likewise.
10403         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10404         (vect_record_max_nunits): Likewise.
10405         (vect_build_slp_tree_1): Likewise.
10406         (vect_build_slp_tree_2): Likewise.
10407         (vect_print_slp_tree): Likewise.
10408         (vect_analyze_slp_instance): Likewise.
10409         (vect_detect_hybrid_slp_stmts): Likewise.
10410         (vect_detect_hybrid_slp_1): Likewise.
10411         (vect_slp_analyze_operations): Likewise.
10412         (vect_slp_analyze_bb_1): Likewise.
10413         (vect_transform_slp_perm_load): Likewise.
10414         (vect_schedule_slp_instance): Likewise.
10415         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10416         (vect_mark_stmts_to_be_vectorized): Likewise.
10417         (vect_init_vector_1): Likewise.
10418         (vect_get_vec_def_for_operand): Likewise.
10419         (vect_finish_stmt_generation_1): Likewise.
10420         (vect_check_load_store_mask): Likewise.
10421         (vectorizable_call): Likewise.
10422         (vectorizable_conversion): Likewise.
10423         (vectorizable_operation): Likewise.
10424         (vectorizable_load): Likewise.
10425         (vect_analyze_stmt): Likewise.
10426         (vect_is_simple_use): Likewise.
10427         (vect_get_vector_types_for_stmt): Likewise.
10428         (vect_get_mask_type_for_stmt): Likewise.
10429         * tree-vectorizer.c (increase_alignment): Likewise.
10431 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
10433         * doc/rtl.texi: Adjust vec_select description.
10434         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
10435         non-constant selectors.
10437 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
10439         * config/aarch64/aarch64-protos.h
10440         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
10441         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
10442         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
10443         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
10444         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
10445         (AARCH64_FL_PROFILE): Move index so flags are ordered.
10446         (AARCH64_ISA_RCPC8_4): New flag.
10447         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
10448         to aarch64_offset_9bit_signed_unscaled_p.
10449         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
10450         and use stlur.
10451         * config/aarch64/constraints.md (Ust): New constraint.
10452         * config/aarch64/predicates.md.
10453         (aarch64_9bit_offset_memory_operand): New predicate.
10454         (aarch64_rcpc_memory_operand): New predicate.
10456 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
10458         PR rtl-optimization/87361
10459         * rtlanal.c (nonzero_bits1): Revert accidental change.
10461 2018-09-19  Richard Biener  <rguenther@suse.de>
10463         PR tree-optimization/87349
10464         PR tree-optimization/87342
10465         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
10467 2018-09-18  Marek Polacek  <polacek@redhat.com>
10469         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
10470         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
10472 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
10474         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
10476 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
10478         PR rtl-optimization/86882
10479         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
10481 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
10483         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
10484         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
10486 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
10488         PR other/87353
10489         * doc/invoke.texi (Link Options): Fix formatting and grammar.
10491 2018-09-18  Richard Biener  <rguenther@suse.de>
10493         PR middle-end/63155
10494         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
10495         (compute_samebase_partition_bases): Likewise.
10496         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
10497         (gimple_can_coalesce_p): Simplify.
10499 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
10501         Handle a library implementation of ffs calling __builtin_ffs.
10502         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
10503         (mmix_init_libfuncs): New function: make __builtin_ffs expand
10504         to __ffsdi2.
10506 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
10508         * diagnostic-show-locus.c (class layout_range): Add field
10509         "m_original_idx".
10510         (layout_range::layout_range): Add "original_idx" param and use it
10511         to initialize new field.
10512         (make_range): Use 0 for original_idx.
10513         (layout::layout): Pass in index to calls to
10514         maybe_add_location_range.
10515         (layout::maybe_add_location_range): Add param "original_idx" and
10516         pass it on to layout_range.
10517         (layout::print_any_labels): Pass on range->m_original_idx to
10518         get_text call.
10519         (gcc_rich_location::add_location_if_nearby): Use 0 for
10520         original_idx.
10521         * gcc-rich-location.h (text_range_label::get_text): Update for new
10522         param.
10523         (range_label_for_type_mismatch::get_text): Likewise.
10525 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
10527         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
10529 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
10531         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
10532         format_string_diagnostic_t.
10533         (fmtwarn_n): Likewise.
10534         * substring-locations.c
10535         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
10536         (format_warning_n_va): Convert to...
10537         (format_string_diagnostic_t::emit_warning_n_va): ...this.
10538         (format_warning_va): Convert to...
10539         (format_string_diagnostic_t::emit_warning_va): ...this.
10540         (format_warning_at_substring): Convert to...
10541         (format_string_diagnostic_t::emit_warning): ...this.
10542         (format_warning_at_substring_n): Convert to...
10543         (format_string_diagnostic_t::emit_warning_n): ...this.
10544         * substring-locations.h (class format_string_diagnostic_t): New
10545         class.
10546         (format_warning_va): Convert to
10547         format_string_diagnostic_t::emit_warning_va.
10548         (format_warning_n_va): Convert to
10549         format_string_diagnostic_t::emit_warning_n_va.
10550         (format_warning_at_substring): Convert to
10551         format_string_diagnostic_t::emit_warning.
10552         (format_warning_at_substring_n): Convert to
10553         format_string_diagnostic_t::emit_warning_n.
10555 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
10556             Bernd Schmidt  <bernds_cb1@t-online.de>
10558         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
10559         SImode args.
10561 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
10563         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
10564         operand 0 predicate to nonimmediate operand.
10565         (rint<mode>2_frndint): Remove insn pattern.
10566         (rint<mode>2): Change operand 1 predicate to general_operand.
10567         Extend operand 1 to XFmode and generate rintxf2 insn.
10568         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
10569         Do not use X87MODEF mode macro.
10570         (frndintxf2_<rounding>_i387): Rename from
10571         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
10572         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
10573         to XFmode and generate significandxf3 insn.
10575 2018-09-17  Richard Biener  <rguenther@suse.de>
10577         PR tree-optimization/87328
10578         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
10579         visiting unexecutable backedges when not iterating.
10580         (do_rpo_vn): Mark all edges not executable even when not
10581         iterating.
10583 2018-09-17  Martin Jambor  <mjambor@suse.cz>
10585         PR c/63886
10586         * doc/invoke.texi (Warning Options): Likewise.
10588 2018-09-17  Richard Biener  <rguenther@suse.de>
10590         PR tree-optimization/87301
10591         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
10592         clean EH info from leftover copy assignments.
10594 2018-09-17  Martin Liska  <mliska@suse.cz>
10596         PR gcov-profile/85871
10597         * gcov.c (output_intermediate_file): Fix out of bounds
10598         access.
10600 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
10602         * config/arc/arc.c: Object attributes for core4 not reflected
10603         correctly.
10604         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
10605         core3).
10607 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
10609         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
10611 2018-09-17  Martin Liska  <mliska@suse.cz>
10613         * doc/gcov.texi: Document new option --use-hotness-colors.
10614         * gcov.c (struct source_info): Declare new field.
10615         (source_info::source_info): Set default for maximum_count.
10616         (print_usage): Add new -q option.
10617         (process_args): Process it.
10618         (accumulate_line_info): Save src->maximum_count.
10619         (output_line_beginning): Make color line number if
10620         flag_use_hotness_colors is set.
10621         (output_line_details): Pass default argument value.
10622         (output_lines): Pass src->maximum_count.
10624 2018-09-17  Martin Liska  <mliska@suse.cz>
10626         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
10627         Use processor_names table.
10628         * config/i386/i386.c (ix86_default_align): Use
10629         processor_cost_table for alignment values.
10630         (ix86_option_override_internal): Use processor_names.
10631         (ix86_function_specific_print): Likewise.
10632         * config/i386/i386.h (struct processor_costs):
10633         Add alignment values.
10634         (struct ptt): Remove and replace with const char *.
10635         * config/i386/x86-tune-costs.h (struct processor_costs):
10636         Declare default alignments for all costs.
10638 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
10640         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
10641         symbolics or VR_VARYING ranges for ABS_EXPR.
10642         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
10643         when range will wrap.
10645 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
10647         PR middle-end/86864
10648         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
10649         before and after a JUMP_TABLE_DATA.
10651 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
10653         PR middle-end/87188
10654         * dojump.c (do_compare_and_jump): Canonicalize function pointers
10655         when one operand is a function pointer.  Use POINTER_TYPE_P and
10656         FUNC_OR_METHOD_TYPE_P.
10657         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
10658         * fold-const.c (build_range_check): Likewise.
10659         * match.pd (simple_comparison): Likewise.
10661 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
10663         PR c/82967
10664         * spellcheck.c (get_edit_distance_cutoff): New function.
10665         (selftest::test_edit_distance_unit_test_oneway): Rename to...
10666         (selftest::test_get_edit_distance_one_way): ...this.
10667         (selftest::test_get_edit_distance_unit): Rename to...
10668         (selftest::test_get_edit_distance_both_ways): ...this.
10669         (selftest::test_edit_distances): Move tests to this new function,
10670         and test some more pairs of strings.  Update for above renaming.
10671         (selftest::get_old_cutoff): New function.
10672         (selftest::test_get_edit_distance_cutoff): New function.
10673         (selftest::assert_suggested_for): New function.
10674         (ASSERT_SUGGESTED_FOR): New macro.
10675         (selftest::assert_not_suggested_for): New function.
10676         (ASSERT_NOT_SUGGESTED_FOR): New macro.
10677         (selftest::test_suggestions): New function.
10678         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
10679         tests to selftest::test_edit_distances and call it.  Add calls to
10680         selftest::test_get_edit_distance_cutoff and
10681         selftest::test_suggestions.
10682         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
10683         (best_match::consider): Replace hard-coded cutoff calculation with
10684         a call to...
10685         (best_match::get_cutoff): New declaration.
10686         (best_match::get_best_meaningful_candidate): Likewise.
10688 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10690         * builtins.c (fold_builtin_strlen): Remove TODO comment.
10692 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10694         revert:
10695         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10697         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10698         terminated string literal.
10700 2018-09-14  Martin Sebor  <msebor@redhat.com>
10702         * builtins.c (unterminated_array): Handle ARRAY_REF.
10703         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
10704         * builtins.h (unterminated_array): Declare extern.
10705         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
10706         arrays.
10707         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
10708         calls.
10710 2018-09-14  Martin Sebor  <msebor@redhat.com>
10711             Jeff Law  <law@redhat.com>
10713         * builtins.c (unterminated_array): New.
10714         (expand_builtin_strcpy): Adjust.
10715         (expand_builtin_strcpy_args): Detect unterminated arrays.
10716         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
10717         unterminated arrays.
10718         * gimple-fold.h (get_maxval_strlen): Add argument.
10719         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
10721         * gimple-fold.c (get_range_strlen): Add argument.
10722         (get_maxval_strlen): Adjust.
10723         * gimple-fold.h (get_range_strlen): Add argument.
10725 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
10727         * config/i386/movdirintrin.h: Fix copyright year.
10729 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
10731         * reg-stack.c: Include regs.h.
10732         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
10733         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
10734         FIRST_STACK_REG, not DFmode.
10735         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
10736         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
10737         (change stack): Default register mode to the reg_raw_mode of
10738         FIRST_STACK_REG, not DFmode.
10739         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
10740         (*swapxf): Rename from swapxf.
10742 2018-09-14  Carl Love  <cel@us.ibm.com>
10744         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
10745         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
10747 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
10749         PR target/87224
10750         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
10751         alternatives.
10753 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
10755         PR target/85628
10756         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
10758 2018-09-14  Jason Merrill  <jason@redhat.com>
10760         Fix --enable-gather-detailed-mem-stats.
10761         * hash-table.c (hash_table_usage): Change from variable to function.
10762         * hash-table.h: Adjust.
10763         * Makefile.in: Add missing dependencies on hash-table.h.
10765 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10767         PR tree-optimization/87259
10768         PR lto/87283
10769         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
10770         execute_cse_reciprocals_1 has tried transforming.
10772 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
10774         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
10775         VR_VARYING for PLUS/MINUS_EXPR.
10777 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
10779         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
10780         formatting.
10782 2018-09-14  Richard Biener  <rguenther@suse.de>
10784         PR middle-end/63155
10785         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
10786         bits for the merged partition.
10788 2018-09-13  Martin Sebor  <msebor@redhat.com>
10789             Bernd Edlinger  <bernd.edlinger@hotmail.de>
10791         * builtins.h (c_srlen): Add argument.
10792         * builtins.c (warn_string_no_nul): New function.
10793         (c_strlen): Add argument and use it.  Update recursive calls.
10794         Pass DECL argument to string_constant to get info on non
10795         terminated strings.  Update *NONSTR as needed.
10796         (fold_builtin_strlen): Add argument to calls to c_strlen.
10797         Warn for unterminated arrays.
10798         (warn_string_no_null): Add prototype.
10799         * expr.c (string_constant): Update arguments.  Update recursive
10800         calls appropriately.  Detect missing NUL terminator and outermost
10801         declaration its missing in.
10802         Improve checks for arrays with nonzero lower bound or elements
10803         that are not a single byte.  Simplify offset computation.
10804         Simplify checks for non-NUL terminated strings.
10805         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
10806         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
10808 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10810         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
10811         correctly.
10812         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
10813         argument.  Fix range checks.
10814         * fold-const.h (c_getstr): Adjust protoype.
10815         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
10816         string is constant but contains no NUL byte.
10818         * expr.c (string_constant): Adjust function comment.
10819         Remove bogus check for zero termination.
10821         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
10823         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
10824         (get_constant_size): Don't make STRING_CSTs larger than they are.
10825         (check_string_literal): New check function for STRING_CSTs.
10826         (output_constant): Use it.
10828 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
10830         PR target/86812
10831         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
10833 2018-09-13  Richard Biener  <rguenther@suse.de>
10835         PR tree-optimization/87263
10836         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
10837         (struct unwind_state): Add max_rpo field.
10838         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
10839         Compute max_rpo, the max RPO number a block can be backwards reached
10840         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
10841         separating it from the iterating mode.
10843 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
10845         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
10846         (rfs_decision): New scheduling decision.
10848 2018-09-13  Richard Biener  <rguenther@suse.de>
10850         PR bootstrap/87134
10851         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
10852         (vn_nary_op_insert_pieces_predicated): Do not write useless
10853         valid_dominated_by_p entry outside of the allocated storage.
10855 2018-09-13  Omar Sandoval  <osandov@osandov.com>
10856             Tom de Vries  <tdevries@suse.de>
10858         PR debug/86985
10859         * dwarf2out.c (is_c): New function.
10860         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
10862 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
10864         PR target/85628
10865         * config/aarch64/aarch64.md (*aarch64_bfxil):
10866         Define.
10867         * config/aarch64/constraints.md (Ulc): Define.
10868         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
10869         Define.
10870         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
10871         New function.
10873 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
10875         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
10876         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
10877         aarch64_layout_frame call.
10878         (aarch64_expand_epilogue): Likewise.
10879         (aarch64_initial_elimination_offset): Likewise.
10880         (aarch64_get_separate_components): Likewise.
10881         (aarch64_use_return_insn_p): Likewise.
10882         (aarch64_layout_frame): Remove unneeded check.
10884 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10886         * configure.ac: Only append
10887         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
10888         gcc_config_arguments if it was never reconfigured or last reconfigure
10889         was with different arguments.
10890         * configure: Regenerated.
10892 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10893             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10895         PR middle-end/87290
10896         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
10897         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
10899 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10901         PR tree-optimization/87287
10902         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
10903         X % C == 0 to X % (unsigned) C == 0 optimization to ...
10904         * match.pd (X % C == 0): ... here.  New optimization.
10906 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10908         PR middle-end/82853
10909         * expr.h (maybe_optimize_mod_cmp): Declare.
10910         * expr.c (mod_inv): New function.
10911         (maybe_optimize_mod_cmp): New function.
10912         (do_store_flag): Use it.
10913         * cfgexpand.c (expand_gimple_cond): Likewise.
10915 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
10916             Julian Brown  <julian@codesourcery.com>
10918         PR middle-end/86336
10919         * gimplify.c (gimplify_scan_omp_clauses): Set
10920         target_firstprivatize_array_bases in OpenACC parallel and kernels
10921         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
10922         OpenACC data regions.
10924 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
10926         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
10927         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
10929 2018-09-12  Richard Biener  <rguenther@suse.de>
10931         PR tree-optimization/87280
10932         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
10933         edge but unreachable target.
10934         (do_rpo_vn): For conservatively handling a PHI only mark
10935         the backedge executable but not the block reachable.
10937 2018-09-12  Richard Biener  <rguenther@suse.de>
10939         PR tree-optimization/87266
10940         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
10941         visited blocks.
10943 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
10945         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
10946         constants.
10947         ("trunc<BFP:mode><DFP_ALL:mode>2")
10948         ("trunc<DFP_ALL:mode><BFP:mode>2")
10949         ("extend<BFP:mode><DFP_ALL:mode>2")
10950         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
10951         according to the target operand type.
10953 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10954             Andreas Krebbel  <krebbel@linux.ibm.com>
10956         PR tree-optimization/86844
10957         * gimple-ssa-store-merging.c
10958         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
10959         there are any overlapping stores in between them, make sure they are
10960         also coalesced or we give up completely.
10962 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10964         PR middle-end/87248
10965         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
10966         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
10968 2018-09-12  Tom de Vries  <tdevries@suse.de>
10970         * common.opt (gdescribe-dies): Add option.
10971         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
10972         attribute for artifical and nameless decls.
10973         (dwarf2out_register_external_die): Add description attribute to
10974         external reference die.
10975         (add_desc_attribute): New functions.
10976         (gen_subprogram_die): Add description attribute to
10977         DW_TAG_call_site_parameter.
10978         * tree-pretty-print.c (print_generic_expr_to_str): New function.
10979         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
10980         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
10981         -gno-describe-dies.
10982         (@item -gdescribe-dies): Add.
10984 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
10986         * tree-vrp.c (vrp_shift_undefined_p): Remove.
10987         (extract_range_from_binary_expr_1: Call
10988         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
10989         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
10990         depend on sign.
10992 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
10994         * gimple-ssa-warn-alloca.c
10995         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
10996         field for ALLOCA_BOUND_*_LARGE.
10998 2018-09-11  Nathan Sidwell  <nathan@acm.org>
11000         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
11002 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
11004         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
11005         for clobbers.  Remove obsolete comment.
11007 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
11009         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
11010         mpxchk, mpxld and mpxst types.
11011         (define_attr length_immediate): Remove all processing of mpx types.
11012         (define_attr prefix_0f): Ditto.
11013         (define_attr memory): Ditto.
11015 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
11017         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
11018         (log<mode>2): Change operand 1 predicate to general_operand.
11019         Extend operand 1 to XFmode and generate logxf3 insn.
11020         (log10<mode>2): Change operand 1 predicate to general_operand.
11021         Extend operand 1 to XFmode and generate log10xf3 insn.
11022         (log2<mode>2): Change operand 1 predicate to general_operand.
11023         Extend operand 1 to XFmode and generate log2xf3 insn.
11024         (fyl2xp1_extend<mode>xf3_i387): Remove.
11025         (log1p<mode>2): Change operand 1 predicate to general_operand.
11026         Extend operand 1 to XFmode and generate log1pxf3 insn.
11027         (fxtract_extend<mode>xf3_i387): Remove.
11028         (logb<mode>2): Change operand 1 predicate to general_operand.
11029         Extend operand 1 to XFmode and generate logbxf3 insn.
11030         (ilogb<mode>2): Change operand 1 predicate to general_operand.
11031         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
11032         (significand<mode>2): Change operand 1 predicate to general_operand.
11033         Extend operand 1 to XFmode and generate significandxf3 insn.
11035 2018-09-11  Nathan Sidwell  <nathan@acm.org>
11037         * gcc.c (perror_with_name, pfatal_with_name): Delete.
11038         (load_specs): Use fatal_error.
11039         (DELETE_IF_ORDINARY, process_command): Use error.
11040         (execute, run_attempt): Use fatal_error.
11042 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
11044         * diagnostic-core.h (sorry_at): New prototype.
11045         * diagnostic.c (sorry_at): New function.
11047 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
11049         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
11050         by zero as VR_UNDEFINED.
11052 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
11054         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
11055         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
11056         (<sincos>mode2): New expander.
11057         (sincos_extend<mode>xf3_i387): Remove insn pattern.
11058         (sincos -> sin, cos splitters): Remove splitter patterns.
11059         (sincos<mode>3): Change operand 2 predicate to general_operand.
11060         Extend operand 2 to XFmode and generate sincosxf3 insn.
11061         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
11062         Change operand 3 predicate to const1_operand.
11063         (fptan_extend<mode>xf4_i387): Remove insn pattern.
11064         (tanxf2): Update operands in the call to fptanxf4_i387.
11065         (tan<mode>2): Change operand 1 predicate to general_operand.
11066         Extend operand 1 to XFmode and generate tanxf3 insn.
11067         (atan2xf3): Rename from *fpatanxf3_i387.
11068         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
11069         (atan2xf3): Remove expander.
11070         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
11071         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
11072         (atan<mode>2): Change operand 1 predicate to general_operand.
11073         Extend operand 1 to XFmode and generate atanxf3 insn.
11075 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
11077         * config/i386/i386.md (x87/SSE constant load splitter): Use
11078         memory_operand instead of nonimmediate_operand for input operand
11079         predicate.
11081 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
11083         * config/i386/i386.md (float partial SSE register stall splitter): Move
11084         splitter near its instruction pattern.
11085         (float_extend partial SSE register stall splitter): Ditto.
11086         (float_truncate partial SSE register stall splitter): Ditto.
11088 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
11090         PR target/86794
11091         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
11092         to speculation_safe_value_not_needed.
11094         PR target/85666
11095         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
11096         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
11097         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
11098         leaf_function_p, instead use has_hard_reg_initial_val.
11100 2018-09-09  Nathan Sidwell  <nathan@acm.org>
11102         * gcc.h (pfatal_with_name): Don't declare here.
11103         * gcc.c (pfatal_with_name): Make static.
11105 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
11107         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
11108         earlyclobber.
11110 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
11112         PR rtl-optimization/85458
11113         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
11114         priority hook to reduce the priority of EXPR.
11116 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
11118         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
11119         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
11120         with X87_ENABLE_FLOAT.  Remove preparation code.
11121         (*float<SWI48:mode><MODEF:mode>2): Rename from
11122         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
11123         math using "enabled" attribute.
11124         (*floatdi<MODEF:mode>2_i387): Rename from
11125         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
11126         enable for 32bit targets only.
11127         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
11128         splitter.
11129         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
11130         as operand 1 predicate.  Rewrite as define_insn_and_split.
11131         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
11133 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
11135         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
11136         to fallthru to FLOAT case.
11138 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
11140         PR target/86731
11141         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
11142         around folding of vec_sl to handle out of range shift values.
11144 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
11146         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
11147         Update callers to gen_fix_trunc<mode>_i387_fisttp
11148         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
11149         nonimmediate_operand.
11150         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
11151         and corresponding splitters.
11152         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
11153         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
11154         (fix_truncdi_i387_with_temp): Remove insn pattern
11155         and corresponding splitters.
11156         (fix_trunc<mode>_i387): Change operand 0 predicate to
11157         nonimmediate_operand.
11158         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
11159         and corresponding splitters.
11160         (*fistdi2_1): Remove.
11161         (fistdi2): Ditto.
11162         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
11163         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
11164         (*fist<mode>2_1): Remove.
11165         (fist<mode>2): Ditto.
11166         (fist<mode>2_with_temp): Remove insn pattern and corresponding
11167         splitters.
11168         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
11169         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
11170         (fistdi2_<rounding>): Change operand 0 predicate to
11171         nonimmediate_operand.
11172         (fistdi2_<rounding>_with_temp): Remove insn pattern
11173         and corresponding splitters.
11174         (fist<mode>2_<rounding>): Change operand 0 predicate to
11175         nonimmediate_operand.
11176         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
11177         and corresponding splitters.
11179         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
11181 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11183         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
11184         the init value.
11186 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
11188         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
11189         early gimple folding of vec_splat().
11190         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
11191         * gimple-fold.h: Add an extern define for tree_vec_extract().
11193 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
11195         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
11196         wrappers around TREE_TYPE comparisons.
11198 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
11200         PR target/80080
11201         * config/s390/predicates.md: Add nonsym_memory_operand.
11202         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
11203         contains a SYMBOL_REF, load it into an intermediate pseudo.
11204         (s390_emit_compare_and_swap): Legitimize operand.
11205         * config/s390/s390.md: Use the new nonsym_memory_operand
11206         with UNSPECV_CAS patterns.
11208 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
11210         PR target/80080
11211         * config/s390/s390-passes.def: New file.
11212         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
11213         declaration.
11214         (make_pass_s390_early_mach): Add declaration.
11215         * config/s390/s390.c (make_pass_s390_early_mach):
11216         (s390_option_override): Remove dynamic registration.
11217         * config/s390/t-s390: Add s390-passes.def.
11219 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
11221         * config/s390/s390.c (s390_decompose_constant_pool_ref):
11222         Remove UNSPEC_LTREL_BASE check.
11223         (annotate_constant_pool_refs): Likewise.
11224         (find_constant_pool_ref): Likewise.
11225         (find_ltrel_base): Removed.
11226         (replace_ltrel_base): Removed.
11227         (s390_mainpool_finish): Remove replace_ltrel_base call.
11228         (s390_chunkify_start): Remove pending LTREL_BASE logic.
11229         (s390_chunkify_finish): Remove replace_ltrel_base call.
11230         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
11232 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
11234         PR target/86779
11235         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
11236         to speculation_safe_value_not_needed.
11238 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
11239             Bernd Schmidt  <bernds_cb1@t-online.de>
11241         * config/nvptx/nvptx-opts.h: New file.
11242         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
11243         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
11244         (ASM_SPEC): Define.
11245         (TARGET_SM35): New macro.
11246         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
11247         correct predicate.
11248         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
11249         values.
11250         (misa=): New option.
11251         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
11253 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
11255         * config/i386/i386.md (truncdfsf2): Remove expander.
11256         (truncdfsf2_with_temp): Ditto.
11257         (truncxf<mode>2): Ditto.
11258         (*truncdfsf_fast_mixed): Remove insn pattern.
11259         (*truncdfsf_fast_i387): Ditto.
11260         (*truncdfsf_mixed): Ditto.
11261         (*truncdfsf_i387): Ditto.
11262         (*truncdfsf2_i387_1): Ditto.
11263         (*truncxfsf2_mixed): Ditto.
11264         (*truncxfdf2_mixed): Ditto.
11265         (*truncxf<mode>2_i387_noop): Ditto. Update callers
11266         to call gen_truncxf<mode>2 instead.
11267         (*truncxf<mode>2_i387): Remove.
11268         (reg->reg splitters): Remove splitter pattern.
11269         (reg->mem splitters): Ditto.
11271         (truncdfsf2): New insn pattern.
11272         (truncxf<mode>2): Ditto.
11274 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11276         * tree-ssa-math-opts.c (is_mult_by): New function.
11277         (is_square_of): Use the above.
11278         (optimize_recip_sqrt): New function.
11279         (pass_cse_reciprocals::execute): Use the above.
11281 2018-09-05  Richard Biener  <rguenther@suse.de>
11283         PR bootstrap/87134
11284         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
11285         to zero-init the emplaced vec.
11287 2018-09-05  Martin Liska  <mliska@suse.cz>
11289         PR tree-optimization/87205
11290         * tree-switch-conversion.c (pass_lower_switch::execute):
11291         Group cases for switch statements.
11293 2018-09-05  Richard Biener  <rguenther@suse.de>
11295         PR tree-optimization/87217
11296         * tree-ssa-sccvn.c (vuse_valueize): New.
11297         (vn_reference_lookup_pieces): Use it.
11298         (vn_reference_lookup): Likewise.
11300 2018-09-05  Nathan Sidwell  <nathan@acm.org>
11302         PR c++/87137
11303         * stor-layout.c (place_field): Scan forwards to check last
11304         bitfield when ms_bitfield_placement is in effect.
11306 2018-09-05  Richard Biener  <rguenther@suse.de>
11308         PR bootstrap/87225
11309         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
11310         return.
11312 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11313             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
11315         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
11316         * config.gcc (extra_objs): Build it.
11317         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
11318         Likewise.
11319         * config/aarch64/aarch64-passes.def
11320         (pass_tag_collision_avoidance): New pass.
11321         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
11322         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
11323         (aarch64_classify_address): Remove static qualifier.
11324         (aarch64_address_info, aarch64_address_type): Move to...
11325         * config/aarch64/aarch64-protos.h: ... here.
11326         (make_pass_tag_collision_avoidance): New function.
11327         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
11328         New tuning flag.
11330 2018-09-05  Martin Liska  <mliska@suse.cz>
11332         * doc/gcov.texi: Update documentation of humar
11333         readable mode.
11334         * gcov.c (format_count): Print one decimal place, it provides
11335         more fine number of situations like '1G' vs. '1.4G'.
11337 2018-09-05  Martin Liska  <mliska@suse.cz>
11339         PR target/87164
11340         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
11341         * optc-gen.awk: Allow 'Var' for Deprecated options in order
11342         to generate a MASK value.
11344 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
11346         PR debug/86593
11347         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
11348         if frame pointer isn't used.
11349         (compute_frame_pointer_to_fb_displacement): Likewise.
11351 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
11353         PR target/87198
11354         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
11355         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
11356         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
11357         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
11358         and OPTION_MASK_ISA_XSAVEC_UNSET.
11360 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
11362         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
11363         XOR operations in NAND case.
11365 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
11367         * wide-int-range.cc (wide_int_range_convert): New.
11368         * wide-int-range.h (wide_int_range_convert): New.
11369         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
11370         code into wide_int_range_convert.
11371         (extract_range_into_wide_ints): Do not munge anti range constants
11372         into the entire domain.  Just return the range back.
11374 2018-09-04  Martin Liska  <mliska@suse.cz>
11376         * genmatch.c (output_line_directive): Add new argument
11377         fnargs.
11378         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
11380 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
11382         * doc/invoke.texi (Option Summary): Add whitespace.
11384         * doc/invoke.texi (Option Summary): Add -Waligned-new.
11386 2018-09-04  Richard Biener  <rguenther@suse.de>
11388         PR tree-optimization/87211
11389         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
11390         backedge value we're supposed to treat as VARYING also number
11391         the PHI to VARYING in case it got a different value-number already.
11393 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
11395         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
11396         (extract_range_from_binary_expr_1): Do not call
11397         vrp_can_optimize_bit_op.
11398         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
11399         static.
11400         (wide_int_range_get_mask_and_bounds): New.
11401         (wide_int_range_optimize_bit_op): New.
11402         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
11403         (wide_int_range_bit_and): Same.
11404         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
11405         (wide_int_range_optimize_bit_op): New.
11406         (wide_int_range_get_mask_and_bounds): New.
11408 2018-09-04  Richard Biener  <rguenther@suse.de>
11410         PR tree-optimization/87176
11411         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
11412         variable.  When value-numbering a virtual PHI node make sure
11413         to not value-number to the backedge value.
11415 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
11417         * doc/extend.texi (Long Long, Hex Floats): Document support for
11418         long long and hex floats in more recent versions of ISO C++.
11420 2018-09-03  Richard Biener  <rguenther@suse.de>
11422         PR tree-optimization/87177
11423         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
11424         cleanup.
11426 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
11428         * bb-reorder.c (edge_order): Convert to C-qsort-style
11429         tri-state comparator.
11430         (reorder_basic_blocks_simple): Change std::stable_sort to
11431         gcc_stablesort.
11433 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
11435         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
11436         tri-state comparator.
11437         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
11439 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
11441         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
11442         (mergesort): ... here as maximum count for using netsort.
11443         (gcc_qsort): Set nlim to 3 if stable sort is requested.
11444         (gcc_stablesort): New.
11445         * system.h (gcc_stablesort): Declare.
11447 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
11449         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
11450         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
11451         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
11453 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
11455         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
11456         lxsdx and stxsdx alternatives.
11457         (*mov<mode>_hardfloat64): Ditto.
11458         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
11460 2018-09-03  Richard Biener  <rguenther@suse.de>
11462         PR tree-optimization/87200
11463         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
11464         simplify result.
11466 2018-09-03  Martin Liska  <mliska@suse.cz>
11468         PR tree-optimization/87201
11469         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11470         Fix parenthesis in an expression.
11472 2018-09-03  Richard Biener  <rguenther@suse.de>
11474         PR tree-optimization/87197
11475         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
11476         visited.  CSE the VN_INFO hashtable lookup.
11478         PR tree-optimization/87169
11479         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
11480         iterating make sure there's no extra backedges from irreducible
11481         regions feeding the header.  Mark the destination block
11482         executable.
11484 2018-09-03  Martin Liska  <mliska@suse.cz>
11486         PR driver/83193
11487         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
11488         * common/common-targhooks.c (default_get_valid_option_values):
11489         New function.
11490         * common/common-targhooks.h (default_get_valid_option_values):
11491         Likewise.
11492         * common/config/i386/i386-common.c: Move processor_target_table
11493         from i386.c.
11494         (ix86_get_valid_option_values): New function.
11495         (TARGET_GET_VALID_OPTION_VALUES): New macro.
11496         * config/i386/i386.c (struct ptt): Move to i386-common.c.
11497         (PTA_*): Move all defined masks into i386-common.c.
11498         (ix86_function_specific_restore): Use new processor_cost_table.
11499         * config/i386/i386.h (struct ptt): Moved from i386.c.
11500         (struct pta): Likewise.
11501         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
11502         * doc/tm.texi.in: Likewise.
11503         * opt-suggestions.c (option_proposer::suggest_option):
11504         Pass prefix to build_option_suggestions.
11505         (option_proposer::get_completions): Likewise.
11506         (option_proposer::build_option_suggestions): Use the new target
11507         hook.
11508         * opts.c (struct option_help_tuple): New struct.
11509         (print_filtered_help): Use the new target hook.
11511 2018-09-03  Martin Liska  <mliska@suse.cz>
11513         PR middle-end/59521
11514         * predict.c (set_even_probabilities): Add likely_edges
11515         argument and handle cases where we have precisely one
11516         likely edge.
11517         (combine_predictions_for_bb): Catch also likely_edges.
11518         (tree_predict_by_opcode): Handle gswitch statements.
11519         * tree-cfg.h (find_case_label_for_value): New declaration.
11520         (find_taken_edge_switch_expr): Likewise.
11521         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11522         Find pivot in decision tree based on probabily, not by number of
11523         nodes.
11525 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
11527         * doc/standards.texi (Standards): Update Objective-C reference.
11529 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
11531         * doc/install.texi (Prerequisites): Update link for MPC.
11533 2018-09-01  Michael Matz  <matz@suse.de>
11535         PR tree-optimization/87074
11536         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
11537         PHIs for outer-loop uses.
11539 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
11541         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
11542         * doc/invoke.texi (C Dialect Options): Ditto.
11544 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
11546         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
11548 2018-08-31  Richard Biener  <rguenther@suse.de>
11550         PR tree-optimization/87168
11551         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
11552         (rpo_elim::eliminate_avail): When OP was not visited it must
11553         be available.
11555 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
11557         * tree-vrp.c (copy_value_range): Convert param "from" from
11558         "value_range *" to "const value_range *".
11559         (range_is_null): Likewise for param "vr".
11560         (range_int_cst_p): Likewise.
11561         (range_int_cst_singleton_p): Likewise.
11562         (symbolic_range_p): Likewise.
11563         (value_ranges_intersect_p): Likewise for both params.
11564         (value_range_nonnegative_p): Likewise for param "vr".
11565         (value_range_constant_singleton): Likewise.
11566         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
11567         (extract_range_into_wide_ints): Likewise for param "vr".
11568         (extract_range_from_multiplicative_op): Likewise for params "vr0"
11569         and "vr1".
11570         (vrp_can_optimize_bit_op): Likewise.
11571         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
11572         "vr1_".
11573         (extract_range_from_unary_expr): Likewise.
11574         (debug_value_range): Likewise for param "vr".
11575         (value_range::dump): Add "const" qualifier.
11576         (vrp_prop::check_array_ref): Convert local "vr" from
11577         "value_range *" to "const value_range *".
11578         (vrp_prop::check_mem_ref): Likewise.
11579         (vrp_prop::visit_stmt): Likewise for local "old_vr".
11580         (vrp_intersect_ranges_1): Likewise for param "vr_1".
11581         (vrp_intersect_ranges): Likewise.
11582         (simplify_stmt_for_jump_threading): Likewise for local "vr".
11583         (vrp_prop::vrp_finalize): Likewise.
11584         * tree-vrp.h (value_range::dump): Add "const" qualifier.
11585         (vrp_intersect_ranges): Add "const" qualifier to params as above.
11586         (extract_range_from_unary_expr): Likewise.
11587         (value_range_constant_singleton): Likewise.
11588         (symbolic_range_p): Likewise.
11589         (copy_value_range): Likewise.
11590         (extract_range_from_binary_expr_1): Likewise.
11591         (range_int_cst_p): Likewise.
11592         (vrp_set_zero_nonzero_bits): Likewise.
11593         (range_int_cst_singleton_p): Likewise.
11595 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
11597         * config/aarch64/arm_neon.h (vabsd_s64): New.
11598         (vnegd_s64): Likewise.
11600 2018-08-31  Martin Jambor  <mjambor@suse.cz>
11602         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
11604 2018-08-31  Martin Liska  <mliska@suse.cz>
11606         * ipa-icf.c (sem_item::add_type): Use
11607         sem_item::m_type_hash_cache.
11608         * ipa-icf.h: Move the cache from sem_item_optimizer
11609         to sem_item.
11611 2018-08-31  Nathan Sidwell  <nathan@acm.org>
11613         * doc/extend.texi (Backwards Compatibility): Remove implicit
11614         extern C leeway of () being (...).
11616 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11618         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
11620 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
11622         PR target/86684
11623         PR target/87149
11624         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
11626 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
11628         PR middle-end/87138
11629         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
11630         gen_int_mode.  Formatting fixes.
11632 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
11634         * target.def (custom_function_descriptors): Improve documentation.
11635         * doc/tm.texi.in (Trampolines): Expand discussion of function
11636         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
11637         beginning of the section.
11638         * doc/tm.texi: Regenerated.
11640 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
11642         * cfg.h (class auto_edge_flag): Spell out the template-id of the
11643         base class in the initializer list.  This is a workaround for
11644         building with older GCC.
11645         (class auto_bb_flag): Likewise.
11647 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
11649         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
11650         (altivec_vcmpequ<VI_char>_p): Remove star.
11651         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
11652         vector load modes.
11653         (expand_strncmp_vec_sequence): New function.
11654         (emit_final_str_compare_vec): New function.
11655         (expand_strn_compare): Add support for vector strncmp.
11656         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
11657         length specification to bytes.
11658         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
11659         (vcmpnezb_p): New pattern.
11660         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
11661         for option -mstring-compare-inline-limit.
11663 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
11665         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
11666         (PTA_SKYLAKE): Add PTA_AES.
11667         (PTA_GOLDMONT): Likewise.
11669 2018-08-29  Jan Hubicka  <jh@suse.cz>
11671         PR lto/86517
11672         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
11673         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
11675 2018-08-29  Jan Hubicka  <jh@suse.cz>
11677         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
11678         TYPE_STUB_DECL.
11679         (hash_tree): Do not visit TYPE_STUB_DECL.
11680         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
11681         stream TYPE_STUB_DECL.
11682         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
11683         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
11684         after free_lang_data.
11685         (type_in_anonymous_namespace_p): Likewise.
11687 2018-08-29  Jan Hubicka  <jh@suse.cz>
11689         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11690         comment that it has to be even number.
11691         (class sreal): Change m_sig type to int32_t.
11692         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11693         int64_t for temporary calculations.
11694         (sreal_verify_basics): Drop one bit from minimum and maximum.
11696 2018-08-30  Richard Biener  <rguenther@suse.de>
11698         PR tree-optimization/87147
11699         * tree-ssa-sccvn.c (SSA_VISITED): New function.
11700         (visit_phi): When the degenerate result is from the backedge and
11701         we didn't visit its definition yet drop to VARYING.
11702         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
11704 2018-08-29  Jan Hubicka  <jh@suse.cz>
11706         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
11707         DECL_VINDEX.
11708         (hash_tree): Likewise.
11710 2018-08-29  Jan Hubicka  <jh@suse.cz>
11712         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
11713         and TYPE_NEXT_REF_TO.
11715 2018-08-29  Jan Hubicka  <jh@suse.cz>
11717         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11718         comment that it has to be even number.
11719         (class sreal): Change m_sig type to int32_t.
11720         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11721         int64_t for temporary calculations.
11722         (sreal_verify_basics): Drop one bit from minimum and maximum.
11724 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
11726         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
11728 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
11730         PR middle-end/86995
11731         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
11732         if to_add is negative.
11734 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11736         PR middle-end/87053
11737         * builtins.c (c_strlen): Improve range checks.
11739 2018-08-29  Martin Sebor  <msebor@redhat.com>
11740             Jeff Law  <law@redhat.com>
11742         PR tree-optimization/86714
11743         PR tree-optimization/86711
11744         * builtins.c (c_strlen): Add arguments to call to string_constant.
11745         * expr.c (string_constant): Add argument.  Detect missing nul
11746         terminator and outermost declaration it's missing in.
11747         * expr.h (string_constant): Add argument.
11748         * fold-const.c (read_from_constant_string): Add arguments to call to
11749         string_constant.
11750         (c_getstr): Likewise.
11751         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11752         to string_constant.
11753         * tree-ssa-strlen.c (get_stridx): Likewise.
11755 2018-08-29  Jan Hubicka  <jh@suse.cz>
11757         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
11758         Do not stream DECL_VINDEX.
11759         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
11760         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
11761         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
11763 2018-08-29  Richard Biener  <rguenther@suse.de>
11765         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
11766         virtual operands that are not default defs to honor region
11767         boundaries.
11768         (rpo_vn_valueize): Remove ineffective code here.
11770 2018-08-29  Richard Biener  <rguenther@suse.de>
11772         PR tree-optimization/87132
11773         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
11774         when skipping defs reachable over backedges.
11776 2018-08-29  Richard Biener  <rguenther@suse.de>
11778         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
11779         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
11780         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
11781         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
11782         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
11783         (vn_reference_lookup_3): Remove use of const_parms.
11784         (free_rpo_vn): Do not free const_parms.
11785         (do_rpo_vn): Do not call init_const_parms.
11786         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
11787         SSA_NAME_POINTS_TO_READONLY_MEMORY.
11788         (call_may_clobber_ref_p_1): Likewise.
11790 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
11792         PR other/86726
11793         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
11794         (-O): Ditto.
11795         (-ftree-scev-cprop): Document.
11797 2018-08-29  Jan Hubicka  <jh@suse.cz>
11799         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
11800         parameters.
11801         (sreal constructor): Update.
11802         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
11803         sreal:operator/): Update.
11805 2018-08-29  Martin Liska  <mliska@suse.cz>
11807         * tree-switch-conversion.c (switch_conversion::expand):
11808         Strenghten assumption about gswitch statements.
11810 2018-08-29  Richard Biener  <rguenther@suse.de>
11812         PR tree-optimization/87117
11813         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
11814         re-value-number released SSA VDEFs.
11816 2018-08-29  Richard Biener  <rguenther@suse.de>
11818         PR tree-optimization/87126
11819         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
11821 2018-08-28  Jim Wilson  <jimw@sifive.com>
11823         * config/riscv/pic.md: Rewrite.
11824         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
11825         invalid address.
11826         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
11827         (SOFTF, default_load, softload, softstore): New.
11829 2018-08-28  Jeff Law  <law@redhat.com>
11831         * fold-const.c (fold_binary_loc): Remove recently added assert.
11833 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
11835         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
11836         to OP parmeter of generated function.
11838 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
11840         PR tree-optimization/87009
11841         * match.pd: Add boolean optimizations.
11843 2018-08-28  Martin Sebor  <msebor@redhat.com>
11845         PR middle-end/86631
11846         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
11847         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
11848         (pass_walloca::gate): Use it.
11849         (alloca_call_type): Same.
11850         (pass_walloca::execute): Same.
11851         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
11853 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
11855         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
11856         GCC_VERSION for usage of "__gcc_dump_printf__" format from
11857         >= 3005 to >= 9000.
11859 2018-08-28  Richard Biener  <rguenther@suse.de>
11861         PR tree-optimization/87124
11862         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
11863         constants before looking up avail.
11865 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
11867         PR middle-end/87099
11868         * calls.c (maybe_warn_nonstring_arg): Punt early if
11869         warn_stringop_overflow is zero.  Don't call get_range_strlen
11870         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
11871         Swap comparison operands to have constants on rhs.  Only use
11872         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
11873         increment lenrng[0].
11875 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
11877         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
11878         use of tree_to_shwi.  Remove duplicated test for the size being
11879         a whole number of bytes.
11881 2018-08-28  Richard Biener  <rguenther@suse.de>
11883         PR tree-optimization/87117
11884         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
11885         Handle removed stmt without LHS (GIMPLE_NOP).
11887 2018-08-28  Richard Biener  <rguenther@suse.de>
11889         PR tree-optimization/87117
11890         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
11891         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
11893 2018-08-28  Richard Biener  <rguenther@suse.de>
11895         PR tree-optimization/87117
11896         * tree-ssa-pre.c (compute_avail): Do not make expressions
11897         with predicated values available.
11898         (get_expr_value_id): Assert we do not run into predicated value
11899         expressions.
11901 2018-08-28  Richard Biener  <rguenther@suse.de>
11903         PR tree-optimization/87117
11904         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
11905         get virtual operands.
11906         (get_expr_operands): Handle STRING_CST like other decls.
11908 2018-08-28  Martin Liska  <mliska@suse.cz>
11910         * tree.h: Update documentation of fndecl_built_in_p
11911         functions.
11914 2018-08-27  Jeff Law  <law@redhat.com>
11915         PR tree-optimization/87110
11916         * tree-ssa-dse.c (compute_trims): Handle non-constant
11917         TYPE_SIZE_UNIT.
11919 2018-08-27  Martin Sebor  <msebor@redhat.com>
11921         PR tree-optimization/86914
11922         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
11924 2018-08-27  Martin Sebor  <msebor@redhat.com>
11926         PR tree-optimization/87112
11927         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
11928         the type of the bound argument.
11930 2018-08-27  Jeff Law  <law@redhat.com>
11932         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
11933         type does not have a TYPE_SIZE_UNIT.
11935 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
11937         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
11938         with include of backend.h.
11940 2018-08-27  Richard Biener  <rguenther@suse.de>
11942         PR tree-optimization/86927
11943         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
11944         use const cond reduction code.
11946 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
11948         PR tree-optimization/85758
11949         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
11951 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
11953         PR c++/87091
11954         * diagnostic-show-locus.c (class layout_range): Update for
11955         conversion of show_caret_p to a tri-state.
11956         (layout_range::layout_range): Likewise.
11957         (make_range): Likewise.
11958         (layout::maybe_add_location_range): Likewise.
11959         (layout::should_print_annotation_line_p): Don't show annotation
11960         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
11961         (layout::get_state_at_point): Update for conversion of
11962         show_caret_p to a tri-state.  Bail out early for
11963         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
11964         underlining or source colorization.
11965         (gcc_rich_location::add_location_if_nearby): Update for conversion
11966         of show_caret_p to a tri-state.
11967         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
11968         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
11969         Likewise.
11970         (selftest::test_one_liner_labels): Likewise.
11971         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
11972         conversion of show_caret_p to a tri-state.
11973         * pretty-print.c (text_info::set_location): Likewise.
11974         * pretty-print.h (text_info::set_location): Likewise.
11975         * substring-locations.c (format_warning_n_va): Likewise.
11976         * tree-diagnostic.c (default_tree_printer): Likewise.
11977         * tree-pretty-print.c (newline_and_indent): Likewise.
11979 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
11981         PR c++/87091
11982         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
11983         line above for line-insertion fix-it hints.
11984         (selftest::test_fixit_insert_containing_newline): Update the
11985         expected results, and add a test with line-numbering enabled.
11987 2018-08-27  Martin Liska  <mliska@suse.cz>
11989         PR sanitizer/86962
11990         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
11991         params with DECL_HAS_VALUE_EXPR_P.
11993 2018-08-27  Martin Liska  <mliska@suse.cz>
11995         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
11996         selected expansion strategy.
11998 2018-08-27  Martin Liska  <mliska@suse.cz>
12000         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
12001         * builtins.c (is_builtin_fn): Likewise.
12002         * attribs.c (diag_attr_exclusions): Use new function
12003         fndecl_built_in_p and remove check for FUNCTION_DECL if
12004         possible.
12005         (builtin_mathfn_code): Likewise.
12006         (fold_builtin_expect): Likewise.
12007         (fold_call_expr): Likewise.
12008         (fold_builtin_call_array): Likewise.
12009         (fold_call_stmt): Likewise.
12010         (set_builtin_user_assembler_name): Likewise.
12011         (is_simple_builtin): Likewise.
12012         * calls.c (gimple_alloca_call_p): Likewise.
12013         (maybe_warn_nonstring_arg): Likewise.
12014         * cfgexpand.c (expand_call_stmt): Likewise.
12015         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
12016         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
12017         (cgraph_node::verify_node): Likewise.
12018         * cgraphclones.c (build_function_decl_skip_args): Likewise.
12019         (cgraph_node::create_clone): Likewise.
12020         * config/arm/arm.c (arm_insert_attributes): Likewise.
12021         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
12022         * dse.c (scan_insn): Likewise.
12023         * expr.c (expand_expr_real_1): Likewise.
12024         * fold-const.c (operand_equal_p): Likewise.
12025         (fold_binary_loc): Likewise.
12026         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
12027         * gimple-low.c (lower_stmt): Likewise.
12028         * gimple-pretty-print.c (dump_gimple_call): Likewise.
12029         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12030         Likewise.
12031         * gimple.c (gimple_build_call_from_tree): Likewise.
12032         (gimple_call_builtin_p): Likewise.
12033         (gimple_call_combined_fn): Likewise.
12034         * gimplify.c (gimplify_call_expr): Likewise.
12035         (gimple_boolify): Likewise.
12036         (gimplify_modify_expr): Likewise.
12037         (gimplify_addr_expr): Likewise.
12038         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
12039         * ipa-cp.c (determine_versionability): Likewise.
12040         * ipa-fnsummary.c (compute_fn_summary): Likewise.
12041         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
12042         * ipa-split.c (visit_bb): Likewise.
12043         (split_function): Likewise.
12044         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12045         * lto-cgraph.c (input_node): Likewise.
12046         * lto-streamer-out.c (write_symbol): Likewise.
12047         * omp-low.c (setjmp_or_longjmp_p): Likewise.
12048         (lower_omp_1): Likewise.
12049         * predict.c (strip_predict_hints): Likewise.
12050         * print-tree.c (print_node): Likewise.
12051         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
12052         * trans-mem.c (is_tm_irrevocable): Likewise.
12053         (is_tm_load): Likewise.
12054         (is_tm_simple_load): Likewise.
12055         (is_tm_store): Likewise.
12056         (is_tm_simple_store): Likewise.
12057         (is_tm_abort): Likewise.
12058         (tm_region_init_1): Likewise.
12059         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
12060         * tree-cfg.c (verify_gimple_call): Likewise.
12061         (move_stmt_r): Likewise.
12062         (stmt_can_terminate_bb_p): Likewise.
12063         * tree-eh.c (lower_eh_constructs_2): Likewise.
12064         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
12065         * tree-inline.c (remap_gimple_stmt): Likewise.
12066         (copy_bb): Likewise.
12067         (estimate_num_insns): Likewise.
12068         (fold_marked_statements): Likewise.
12069         * tree-sra.c (scan_function): Likewise.
12070         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
12071         (optimize_stack_restore): Likewise.
12072         (pass_fold_builtins::execute): Likewise.
12073         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
12074         (mark_all_reaching_defs_necessary_1): Likewise.
12075         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
12076         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
12077         (pass_forwprop::execute): Likewise.
12078         * tree-ssa-loop-im.c (stmt_cost): Likewise.
12079         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
12080         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
12081         * tree-ssa-strlen.c (get_string_length): Likewise.
12082         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
12083         (find_func_aliases_for_call): Likewise.
12084         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
12085         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
12086         * tree-tailcall.c (find_tail_calls): Likewise.
12087         * tree.c (need_assembler_name_p): Likewise.
12088         (free_lang_data_in_decl): Likewise.
12089         (get_call_combined_fn): Likewise.
12090         * ubsan.c (is_ubsan_builtin_p): Likewise.
12091         * varasm.c (incorporeal_function_p): Likewise.
12092         * tree.h (DECL_BUILT_IN): Remove and replace with
12093         fndecl_built_in_p.
12094         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
12095         (fndecl_built_in_p): New.
12097 2018-08-27  Martin Liska  <mliska@suse.cz>
12099         PR tree-optimization/86847
12100         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
12101         Dump also subtree probability.
12102         (switch_decision_tree::do_jump_if_equal): New function.
12103         (switch_decision_tree::emit_case_nodes): Handle special
12104         situations in balanced tree that can be emitted much simpler.
12105         Fix calculation of probabilities that happen in tree expansion.
12106         * tree-switch-conversion.h (struct cluster): Add
12107         is_single_value_p.
12108         (struct simple_cluster): Likewise.
12109         (struct case_tree_node): Add new function has_child.
12110         (do_jump_if_equal): New.
12112 2018-08-27  Martin Liska  <mliska@suse.cz>
12114         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
12115         Add new argument to bit_test_cluster constructor.
12116         (bit_test_cluster::emit): Set bits really number of values
12117         handlel by a test.
12118         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
12119         probability argument.
12120         * tree-switch-conversion.h (struct bit_test_cluster):
12121         Add m_handles_entire_switch.
12123 2018-08-27  Martin Liska  <mliska@suse.cz>
12125         PR tree-optimization/86702
12126         * tree-switch-conversion.c (jump_table_cluster::emit):
12127         Make probabilities even for values in jump table
12128         according to number of cases handled.
12129         (switch_decision_tree::compute_cases_per_edge): Pass
12130         argument to reset_out_edges_aux function.
12131         (switch_decision_tree::analyze_switch_statement): Likewise.
12132         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
12133         Make it static.
12135 2018-08-27  Martin Liska  <mliska@suse.cz>
12137         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
12138         cfun argument explicitly.
12139         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
12140         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
12141         function gimple_switch_default_bb.
12142         (convert_switch_statements):
12143         (expand_builtins):
12144         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
12145         * stmt.c (label_to_block_fn): Use label_to_block and pass
12146         cfun argument explicitly and use gimple_switch_label_bb.
12147         (expand_case): Likewise.
12148         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
12149         cfun argument explicitly. Likewise.
12150         (make_edges_bb): Likewise.
12151         (make_cond_expr_edges): Likewise.
12152         (get_cases_for_edge): Likewise.
12153         (make_gimple_switch_edges): Likewise.
12154         (label_to_block_fn): Likewise.
12155         (label_to_block): Likewise.
12156         (make_goto_expr_edges): Likewise.
12157         (make_gimple_asm_edges): Likewise.
12158         (main_block_label): Likewise.
12159         (group_case_labels_stmt): Likewise.
12160         (find_taken_edge_computed_goto): Likewise.
12161         (find_taken_edge_switch_expr): Likewise.
12162         (gimple_verify_flow_info): Likewise.
12163         (gimple_redirect_edge_and_branch): Likewise.
12164         (gimple_switch_label_bb): New function.
12165         (gimple_switch_default_bb): Likewise.
12166         (gimple_switch_edge): Likewise.
12167         (gimple_switch_default_edge): Likewise.
12168         * tree-cfg.h (label_to_block_fn): Remove and replace ...
12169         (label_to_block): ... with this.
12170         (gimple_switch_label_bb): New.
12171         (gimple_switch_default_bb): Likewise.
12172         (gimple_switch_edge): Likewise.
12173         (gimple_switch_default_edge): Likewise.
12174         * tree-cfgcleanup.c (convert_single_case_switch): Use
12175         new gimple functions and pass new argument to label_to_block.
12176         (cleanup_control_flow_bb):
12177         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
12178         cfun argument explicitly.
12179         (make_eh_edges): Likewise.
12180         (redirect_eh_dispatch_edge): Likewise.
12181         (lower_resx): Likewise.
12182         (lower_eh_dispatch): Likewise.
12183         (maybe_remove_unreachable_handlers): Likewise.
12184         (unsplit_eh): Likewise.
12185         (cleanup_empty_eh): Likewise.
12186         (verify_eh_edges): Likewise.
12187         (verify_eh_dispatch_edge): Likewise.
12188         * tree-ssa-dom.c (record_edge_info): Likewise.
12189         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
12190         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
12191         (thread_through_normal_block): Likewise.
12192         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
12193         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
12194         * tree-switch-conversion.c (switch_conversion::collect): Use new
12195         gimple functions.
12196         (switch_conversion::check_final_bb): Likewise.
12197         (switch_conversion::gather_default_values): Pass new argument
12198         to label_to_block.
12199         (switch_conversion::build_constructors): Likewise.
12200         (switch_decision_tree::compute_cases_per_edge): Use new
12201         gimple_switch_edge function.
12202         (switch_decision_tree::analyze_switch_statement): Pass new argument
12203         to label_to_block.
12204         (switch_decision_tree::try_switch_expansion): Use
12205         gimple_switch_default_edge.
12206         * tree-vrp.c (find_switch_asserts): Pass new argument
12207         to label_to_block.
12208         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
12209         (vr_values::simplify_switch_using_ranges): Likewise.
12211 2018-08-27  Richard Biener  <rguenther@suse.de>
12213         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
12214         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
12216         * tree-ssa-sccvn.h (struct vn_pval): New structure.
12217         (struct vn_nary_op_s): Add unwind_to member.  Add
12218         predicated_values flag and put result into a union together
12219         with a linked list of vn_pval.
12220         (struct vn_ssa_aux): Add name member to make maintaining
12221         a map of SSA name to vn_ssa_aux possible.  Remove no longer
12222         needed info, dfsnum, low, visited, on_sccstack, use_processed
12223         and range_info_anti_range_p members.
12224         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
12225         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
12226         New functions.
12227         (vn_valueize): New global.
12228         (vn_context_bb): Likewise.
12229         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
12230         VN_INFO_PTR_INFO): Remove.
12231         * tree-ssa-sccvn.c: ... (rewrite)
12232         (pass_fre::execute): For -O2+ initialize loops and run
12233         RPO VN in optimistic mode (iterating).  For -O1 and -Og
12234         run RPO VN in non-optimistic mode.
12235         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
12236         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
12237         * doc/invoke.texi (sccvn-max-scc-size): Remove.
12238         (rpo-vn-max-loop-depth): Document.
12239         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
12240         when valuezing the VUSE signals we walked out of the region.
12241         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
12242         (phi_translate): Set VN context block to use for availability
12243         lookup.
12244         (compute_avail): Likewise.
12245         (pre_valueize): New function.
12246         (pass_pre::execute): Adjust to the RPO VN API.
12248         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
12249         (propagate_constants_for_unrolling): Remove.
12250         (tree_unroll_loops_completely): Perform value-numbering
12251         on the unrolled bodies loop parent.
12253 2018-08-27  Richard Biener  <rguenther@suse.de>
12255         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
12256         for partial antic compute.
12258 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
12260         PR rtl-optimization/87065
12261         * combine.c (simplify_if_then_else): Formatting fix.
12262         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
12263         check.
12264         (known_cond): Don't return const_true_rtx for vector modes.  Use
12265         CONST0_RTX instead of const0_rtx.  Formatting fixes.
12267 2018-08-27  Martin Liska  <mliska@suse.cz>
12269         PR gcov-profile/87069
12270         * gcov.c (process_file): Record files already processed
12271         and warn about a file being processed multiple times.
12273 2018-08-27  Martin Liska  <mliska@suse.cz>
12275         PR driver/83193
12276         * config/aarch64/aarch64.c (aarch64_override_options_internal):
12277         Set default values for x_aarch64_*_string strings.
12278         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
12279         prefix.  For -mabi do not print '=ABI' in help and use
12280         <option_value> format for -msve-vector-bits and -moverride
12281         options.
12283 2018-08-26  Jeff Law  <law@redhat.com>
12285         * config/mips/frame-header-opt.c: Include "backend.h" rather than
12286         "cfg.h"
12288 2018-08-26  Marek Polacek  <polacek@redhat.com>
12290         PR c++/87029, Implement -Wredundant-move.
12291         * doc/invoke.texi: Document -Wredundant-move.
12293 2018-08-25  Martin Sebor  <msebor@redhat.com>
12295         PR tree-optimization/87059
12296         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
12297         to the same type as the other.
12298         * fold-const.c (fold_binary_loc): Assert expectation.
12300 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
12302         * config/darwin.c (machopic_legitimize_pic_address): Clean up
12303         extraneous parentheses, dead code section and formatting.
12305 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
12307         PR c++/87091
12308         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
12309         wide enough for jumps in the line-numbering to be visible.
12310         (layout::print_gap_in_line_numbering): New member function.
12311         (layout::calculate_line_spans): When using line numbering, merge
12312         line spans that are only 1 line apart.
12313         (diagnostic_show_locus): When printing line numbers, show gaps in
12314         line numbering directly, rather than printing headers.
12315         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
12316         line-numbering with multiple line spans.
12317         (selftest::test_fixit_insert_containing_newline_2): Add test of
12318         line-numbering, in which the spans are close enough to be merged.
12320 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
12322         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
12323         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
12324         * tree-vrp.c (range_is_nonnull): Remove.
12325         (range_includes_zero_p): Accept value_range instead of min/max.
12326         (extract_range_from_binary_expr_1): Do not early bail on
12327         POINTER_PLUS_EXPR.
12328         Use range_includes_zero_p instead of range_is_nonnull.
12329         (extract_range_from_unary_expr): Use range_includes_zero_p instead
12330         of range_is_nonnull.
12331         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
12332         special case VR_ANTI_RANGE.
12333         (vrp_finalize): Same.
12334         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
12335         instead of min/max.
12336         (range_is_nonnull): Remove.
12337         * vr-values.c (vrp_stmt_computes_nonzero): Use
12338         range_includes_zero_p instead of range_is_nonnull.
12339         (extract_range_basic): Pass value_range to range_includes_zero_p
12340         instead of range_is_nonnull.
12342 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
12344         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
12345         * emit-rtl.h (rtl_data): Remove return_bnd.
12346         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
12347         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
12348         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
12349         (POINTER_BOUNDS_MODE): Remove definition.
12350         (make_pointer_bounds_mode): Remove.
12351         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
12352         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
12353         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
12354         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
12355         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
12356         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
12357         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
12359         * config/i386/i386-modes.def (BND32, BND64): Remove.
12360         * config/i386/i386.c (dbx_register_map): Remove bound registers.
12361         (dbx64_register_map): Ditto.
12362         (svr4_dbx_register_map): Ditto.
12363         (indirect_thunk_bnd_needed): Remove.
12364         (indirect_thunks_bnd_used): Ditto.
12365         (indirect_return_bnd_needed): Ditto.
12366         (indirect_return_via_cx_bnd): Ditto.
12367         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
12368         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
12369         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
12370         (output_indirect_thunk_function): Remove handling of
12371         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
12372         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
12373         (ix86_save_reg): Remove handling of crtl->return_bnd.
12374         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
12375         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
12376         and UNSPEC_BNDLX_ADDR.
12377         (ix86_output_indirect_branch_via_reg): Remove handling of
12378         indirect_thunk_prefix_bnd.
12379         (ix86_output_indirect_branch_via_push): Ditto.
12380         (ix86_output_function_return): Ditto.
12381         (ix86_output_indirect_function_return): Ditto.
12382         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
12383         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
12384         (CALL_USED_REGISTERS): Ditto.
12385         (REG_ALLOC_ORDER): Update for removal of bound registers.
12386         (HI_REGISTER_NAMES): Ditto.
12387         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
12388         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
12389         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
12390         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
12391         (FIRST_PSEUDO_REG): Update.
12392         (BND): Remove mode iterator.
12393         * config/i386/predicates.md (bnd_mem_operator): Remove.
12395 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
12397         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
12398         vectors.
12400 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
12402         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
12403         the case in which the permute needs only a single element and
12404         repeats for every vector of the result.  Extend that case to
12405         handle variable-length vectors.
12406         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
12408 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
12410         PR debug/79342
12411         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
12412         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
12414 2018-08-24  Richard Biener  <rguenther@suse.de>
12416         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
12417         bb_flags_allocated members.
12418         (auto_flag): New RAII class for allocating flags.
12419         (auto_edge_flag): New RAII class for allocating edge flags.
12420         (auto_bb_flag): New RAII class for allocating bb flags.
12421         * cfgloop.c (verify_loop_structure): Allocate temporary edge
12422         flag dynamically.
12423         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
12424         in favor of temporarily allocated BB flag.
12425         * hsa-brig.c: Re-order includes.
12426         * hsa-dump.c: Likewise.
12427         * hsa-regalloc.c: Likewise.
12428         * print-rtl.c: Likewise.
12429         * profile-count.c: Likewise.
12431 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
12433         PR target/86989
12434         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
12435         the TOC register.
12437 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
12439         PR 87073/bootstrap
12440         * wide-int-range.cc (wide_int_range_div): Do not ignore result
12441         from wide_int_range_multiplicative_op.
12443 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12445         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
12446         "permutaion".
12448 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
12450         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
12451         to 'expanded'.
12453 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
12455         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
12456         full GENERIC expression used for replacement.
12458 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
12460         * tree-vrp.c (abs_extent_range): Remove.
12461         (extract_range_into_wide_ints): Pass wide ints by reference.
12462         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
12463         Pass wide ints by reference in all calls to
12464         extract_range_into_wide_ints.
12465         * wide-int-range.cc (wide_int_range_div): New.
12466         * wide-int-range.h (wide_int_range_div): New.
12467         (wide_int_range_includes_zero_p): New.
12468         (wide_int_range_zero_p): New.
12470 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
12472         * config/aarch64/aarch64.md (arches): New enum.
12473         (arch): New enum attr.
12474         (arch_enabled): New attr.
12475         (enabled): Now uses arch_enabled only.
12476         (simd, sve, fp16): Removed attribute.
12477         (fp): Attr now defined in terms of 'arch'.
12478         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
12479         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
12480         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
12481         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
12482         attributes into 'arch'.
12483         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
12484         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
12485         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
12486         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
12487         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
12488         'simd' attribute into 'arch'.
12489         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
12490         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
12491         Convert use of 'fp' attribute to 'arch'.
12492         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
12493         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
12494         into 'arch'.
12495         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
12496         (different modes) Merge 'fp' and 'simd' into 'arch'.
12497         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
12498         'simd' into 'arch'.
12500 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
12502         PR rtl-optimization/87026
12503         * expmed.c (canonicalize_comparison): If we can no longer create
12504         pseudoregisters, don't.
12506 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
12508         PR target/86951
12509         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
12510         prototype.
12511         * config/arm/arm.c (speculation_barrier_libfunc): New static
12512         variable.
12513         (arm_init_libfuncs): Initialize it.
12514         (arm_emit_speculation_barrier): New function.
12515         * config/arm/arm.md (speculation_barrier): Call
12516         arm_emit_speculation_barrier for architectures that do not have
12517         DSB or ISB.
12518         (speculation_barrier_insn): Only match on Armv7 or later.
12520 2018-08-23  Richard Biener  <rguenther@suse.de>
12522         PR middle-end/87024
12523         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
12524         calls.
12526 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
12528         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
12529         of single-vector TBLs.
12530         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
12531         one input is given.
12533 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
12535         PR target/85910
12536         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
12537         aarch64_evpc_tbl guard.
12539 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12541         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
12542         behaviour.
12544 2018-08-22  Martin Sebor  <msebor@redhat.com>
12546         PR middle-end/87052
12547         * tree-pretty-print.c (pretty_print_string): Add argument.
12548         (dump_generic_node): Call to pretty_print_string with string size.
12550 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
12552         PR rtl-optimization/86771
12553         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
12554         of two SETs into those two SETs, one to be placed at i2, if that SETs
12555         destination is modified between i2 and i3.
12557 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
12559         PR tree-optimization/86725
12560         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
12561         function.
12562         (vect_analyze_scalar_cycles_1): Check it.
12564 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
12566         PR tree-optimization/86725
12567         * tree-vect-loop.c (vect_is_simple_reduction): When treating
12568         an outer loop phi as a double reduction, make sure that the
12569         single user of the phi result is an inner loop phi.
12571 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
12573         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
12574         grouped stores with gaps to a strided group.
12576 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
12578         * tree-vect-stmts.c (get_group_load_store_type)
12579         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
12580         first statement in a group.
12582 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
12584         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
12585         the sequence used in gcc/gcc.c.
12587 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
12589         PR other/704
12590         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
12591         building it.
12593 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
12595         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
12596         Darwin10-specific unwinder-shim.
12597         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
12598         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
12599         New to cater for Darwin10 Rosetta.
12601 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
12603         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
12604         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
12606 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
12608         PR bootstrap/81033
12609         PR target/81733
12610         PR target/52795
12611         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
12612         (dwarf2out_switch_text_section): Generate a local label for the
12613         second function sub-section and apply it as the second FDE start
12614         label.
12615         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
12616         second sub-section start.
12618 2018-08-22  Richard Biener  <rguenther@suse.de>
12620         PR tree-optimization/86988
12621         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
12623 2018-08-22  Richard Biener  <rguenther@suse.de>
12625         PR tree-optimization/86945
12626         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
12628 2018-08-22  Alexandre Oliva  <oliva@adacore.com>
12630         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
12631         a comment about how uses of r2 for .sdata2 come about.
12633 2018-08-22  Alexandre Oliva  <aoliva@redhat.com>
12635         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
12637 2018-08-21  Marek Polacek  <polacek@redhat.com>
12639         PR c++/86981, Implement -Wpessimizing-move.
12640         * doc/invoke.texi: Document -Wpessimizing-move.
12642 2018-08-21  Jan Hubicka  <jh@suse.cz>
12644         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
12645         * tree.h (is_redundant_typedef): Remove.
12646         * dwarf2out.c (is_redundant_typedef): Turn into static function.
12648 2018-08-21  Jan Hubicka  <jh@suse.cz>
12650         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
12651         when possible.
12653 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
12655         * expmed.c (extract_low_bits): Reject invalid subregs early.
12657 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12659         PR middle-end/86121
12660         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
12661         behaviour.
12663 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12665         * config/vxworks.h: Guard vxworks_asm_out_constructor and
12666         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
12667         * config/vxworks.c: Likewise.
12669 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12671         * config/vxworks.c: Set targetm.have_ctors_dtors
12672         if HAVE_INITFINI_ARRAY_SUPPORT.
12673         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
12674         if HAVE_INITFINI_ARRAY_SUPPORT.
12676 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12678         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
12679         default search path for VxWorks < 7.
12681 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12683         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
12684         (get_format_string): Refer to c_getstr.
12686 2018-08-21  Tom de Vries  <tdevries@suse.de>
12688         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12689         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12690         (debuginfo_early_stop): Declare.
12691         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12692         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12693         (debuginfo_early_stop): New function.
12694         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
12695         and debuginfo_early_stop.
12696         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
12697         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
12698         (general_init): Call debuginfo_early_init.
12699         (finalize): Call debuginfo_fini.
12700         (do_compile): Call debuginfo_init.
12701         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
12702         -fdump-early-debug.
12703         (@item -fdump-debug, @item -fdump-earlydebug): Add.
12705 2018-08-21  Tom de Vries  <tdevries@suse.de>
12707         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
12708         flag_dump_noaddr and flag_dump_unnumbered.
12710 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
12712         * wide-int-range.cc (wide_int_range_abs): New.
12713         (wide_int_range_order_set): Rename from wide_int_range_min_max.
12714         * wide-int-range.h (wide_int_range_abs): New.
12715         (wide_int_range_min_max): New.
12716         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
12717         case to call wide_int_range_abs.
12718         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
12719         (extract_range_from_abs_expr): Delete.
12721 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
12723         PR target/87033
12724         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
12725         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
12726         for indexed loads.
12728 2018-08-20  Nathan Sidwell  <nathan@acm.org>
12729             Jeff Law  <law@redhat.com>
12731         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
12732         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
12734 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
12736         PR other/84889
12737         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
12738         (decl_attributes): Likewise.
12739         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
12740         instance.
12741         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12742         * diagnostic-core.h (class auto_diagnostic_group): New class.
12743         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
12744         (diagnostic_report_diagnostic): Handle the first diagnostics within
12745         a group.
12746         (emit_diagnostic): Add auto_diagnostic_group instance.
12747         (inform): Likewise.
12748         (inform_n): Likewise.
12749         (warning): Likewise.
12750         (warning_at): Likewise.
12751         (warning_n): Likewise.
12752         (pedwarn): Likewise.
12753         (permerror): Likewise.
12754         (error): Likewise.
12755         (error_n): Likewise.
12756         (error_at): Likewise.
12757         (sorry): Likewise.
12758         (fatal_error): Likewise.
12759         (internal_error): Likewise.
12760         (internal_error_no_backtrace): Likewise.
12761         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
12762         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
12763         * diagnostic.h (struct diagnostic_context): Add fields
12764         "diagnostic_group_nesting_depth",
12765         "diagnostic_group_emission_count", "begin_group_cb",
12766         "end_group_cb".
12767         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
12768         Add auto_diagnostic_group instance(s).
12769         (find_explicit_erroneous_behavior): Likewise.
12770         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
12771         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
12772         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
12773         (gimplify_va_arg_expr): Likewise.
12774         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
12775         (HSA_SORRY_AT): Likewise.
12776         * ipa-devirt.c (compare_virtual_tables): Likewise.
12777         (warn_odr): Likewise.
12778         * multiple_target.c (expand_target_clones): Likewise.
12779         * opts-common.c (cmdline_handle_error): Likewise.
12780         * reginfo.c (globalize_reg): Likewise.
12781         * substring-locations.c (format_warning_n_va): Likewise.
12782         * tree-inline.c (expand_call_inline): Likewise.
12783         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
12784         * tree-ssa-loop-niter.c
12785         (do_warn_aggressive_loop_optimizations): Likewise.
12786         * tree-ssa-uninit.c (warn_uninit): Likewise.
12787         * tree.c (warn_deprecated_use): Likewise.
12789 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
12791         PR target/87014
12792         * config/i386/i386.md (eh_return): Always update EH return
12793         address in word_mode.
12795 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
12797         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
12798         TARGET_SPLIT_COMPLEX_ARG is defined.
12800 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12802         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
12804 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12806         PR target/86984
12807         * expr.c (expand_assignment): Assert that bitpos is positive.
12808         (store_field): Likewise
12809         (expand_expr_real_1): Make sure that bitpos is positive.
12810         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
12811         integer overflow.
12813 2018-08-20  Nathan Sidwell  <nathan@acm.org>
12815         * Makefile.in (CPP_ID_DATA_H): Delete.
12816         (CPP_INTERNAL_H): Don't add it.
12817         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
12818         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
12820 2018-08-20  Richard Biener  <rguenther@suse.de>
12822         PR tree-optimization/78655
12823         * tree-vrp.c (extract_range_from_binary_expr_1): Make
12824         pointer + offset nonnull if either operand is nonnull work.
12826 2018-08-20  Tom de Vries  <tdevries@suse.de>
12828         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
12829         unless the referenced die describes the added property using
12830         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
12831         Otherwise, add a DW_AT_location to the referenced die.
12833 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
12835         PR target/86994
12836         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
12837         register_operand when calling ix86_set_reg_reg_cost.
12838         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
12839         Set *total to 0 for operands that satisfy x86_64_immediate_operand
12840         predicate and to 1 otherwise.
12842 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
12844         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
12845         emit a diagnostic that it is not supported and reset the option.
12846         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
12847         supported and consume the option.  (ASM_FINAL_SPEC): New.
12849 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
12851         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
12852         a sentence.
12854 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
12856         C-SKY port: Documentation
12858         * doc/extend.texi (C-SKY Function Attributes): New section.
12859         * doc/invoke.texi (Option Summary): Add C-SKY options.
12860         (C-SKY Options): New section.
12861         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
12863 2018-08-17  Jojo  <jijie_rong@c-sky.com>
12864             Huibin Wang  <huibin_wang@c-sky.com>
12865             Sandra Loosemore  <sandra@codesourcery.com>
12866             Chung-Lin Tang  <cltang@codesourcery.com>
12868         C-SKY port: Backend implementation
12870         * config/csky/*: New.
12871         * common/config/csky/*: New.
12873 2018-08-17  Jojo  <jijie_rong@c-sky.com>
12874             Huibin Wang  <huibin_wang@c-sky.com>
12875             Sandra Loosemore  <sandra@codesourcery.com>
12876             Chung-Lin Tang  <cltang@codesourcery.com>
12877             Andrew Jenner  <andrew@codesourcery.com>
12879         C-SKY port: Configury
12881         * config.gcc (csky-*-*): New.
12882         * configure.ac: Add csky to targets for dwarf2 debug_line support.
12883         * configure: Regenerated.
12885 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
12887         * dump-context.h: Include "dumpfile.h".
12888         (dump_context::dump_printf_va): Convert final param from va_list
12889         to va_list *.  Convert from ATTRIBUTE_PRINTF to
12890         ATTRIBUTE_GCC_DUMP_PRINTF.
12891         (dump_context::dump_printf_loc_va): Likewise.
12892         * dumpfile.c: Include "stringpool.h".
12893         (make_item_for_dump_printf_va): Delete.
12894         (make_item_for_dump_printf): Delete.
12895         (class dump_pretty_printer): New class.
12896         (dump_pretty_printer::dump_pretty_printer): New ctor.
12897         (dump_pretty_printer::emit_items): New member function.
12898         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
12899         function.
12900         (dump_pretty_printer::emit_item): New member function.
12901         (dump_pretty_printer::stash_item): New member function.
12902         (dump_pretty_printer::format_decoder_cb): New member function.
12903         (dump_pretty_printer::decode_format): New member function.
12904         (dump_context::dump_printf_va): Reimplement in terms of
12905         dump_pretty_printer.
12906         (dump_context::dump_printf_loc_va): Convert final param from va_list
12907         to va_list *.
12908         (dump_context::begin_scope): Reimplement call to
12909         make_item_for_dump_printf.
12910         (dump_printf): Update for change to dump_printf_va.
12911         (dump_printf_loc): Likewise.
12912         (selftest::test_capture_of_dump_calls): Convert "stmt" from
12913         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
12914         with %T, %E, and %G.
12915         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
12916         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
12917         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
12918         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
12919         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
12920         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
12921         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
12922         within a dump_printf_loc call to "%wu".
12923         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
12924         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
12925         missing space after "=".
12926         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
12927         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
12928         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
12929         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
12930         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
12931         duplicate "vectorized" from message.
12933 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
12935         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
12936         polyNxK_t element's TYPE_STRING_FLAG.
12938 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
12940         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
12941         (they were unnamed before).  Fix comments.
12943 2018-08-17  Nathan Sidwell  <nathan@acm.org>
12945         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
12947 2018-08-17  Richard Biener  <rguenther@suse.de>
12949         PR tree-optimization/86841
12950         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
12952 2018-08-17  Martin Liska  <mliska@suse.cz>
12954         * common.opt: Remove Warn, Init and Report for options with
12955         Ignore/Deprecated flag. Warning is done automatically for
12956         Deprecated flags.
12957         * config/i386/i386.opt: Likewise.
12958         * config/ia64/ia64.opt: Likewise.
12959         * config/rs6000/rs6000.opt: Likewise.
12960         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
12961         Remove usage of flag_check_pointer_bounds.
12962         * lto-wrapper.c (merge_and_complain): Do not handle
12963         OPT_fcheck_pointer_bounds.
12964         (append_compiler_options): Likewise.
12965         * opt-functions.awk: Do not handle Deprecated.
12966         * optc-gen.awk: Check that Var, Report and Init are not
12967         used for an option with Ignore/Deprecated flag.
12968         * opts-common.c (decode_cmdline_option): Do not report
12969         CL_ERR_DEPRECATED.
12970         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
12971         options.
12972         * opts.h (struct cl_option): Remove cl_deprecated flag.
12973         (CL_ERR_DEPRECATED): Remove error enum value.
12975 2018-08-17  Richard Biener  <rguenther@suse.de>
12977         PR middle-end/86505
12978         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
12979         across a va-arg-pack using call adjust its return value accordingly.
12981 2018-08-16  Martin Sebor  <msebor@redhat.com>
12983         PR tree-optimization/86853
12984         * gimple-ssa-sprintf.c (struct format_result): Rename member.
12985         (struct fmtresult): Add member and initialize it in ctors.
12986         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
12987         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
12988         (format_directive): Set POSUNDER4K when MAYFAIL is set.
12989         (parse_directive): Handle %C same as %c.
12990         (sprintf_dom_walker::compute_format_length): Adjust.
12991         (is_call_safe): Adjust.
12993 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12995         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
12996         for determining how to count the elements.
12997         * builtins.h (c_strlen): Adjust prototype.
12998         * expr.c (string_constant): Add new parameter mem_size.
12999         Set *mem_size appropriately.
13000         * expr.h (string_constant): Adjust protoype.
13001         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
13002         * gimple-fold.h (get_range_strlen): Adjust prototype.
13003         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
13004         (format_string): Call get_string_length with eltsize.
13006 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
13008         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
13009         to emit the span, rather than setting it as the prefix.
13011 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
13013         * diagnostic-show-locus.c (layout::start_annotation_line): Add
13014         "margin_char" parameter, defaulting to space.  Use it in place
13015         of pp_space for the initial part of the margin.
13016         (layout::print_leading_fixits): Use '+' when filling the margin
13017         of line-insertion fix-it hints.
13019 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
13021         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
13022         Delete.
13024 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
13026         * config/rs6000/altivec.md: Don't set length attribute to the default
13027         value.
13028         * config/rs6000/darwin.md: Ditto.
13029         * config/rs6000/dfp.md: Ditto.
13030         * config/rs6000/htm.md: Ditto.
13031         * config/rs6000/rs6000.md: Ditto.
13032         * config/rs6000/sync.md: Ditto.
13033         * config/rs6000/vsx.md: Ditto.
13035 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
13037         * config/rs6000/altivec.md: Don't set length attribute to the default
13038         value, for branch instructions.
13039         * config/rs6000/darwin.md: Ditto.
13040         * config/rs6000/rs6000.md: Ditto.
13042 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
13044         * config/rs6000/rs6000.md (length): Always define as const_int 4.
13045         (unnamed conditional branch define_insn): Set length to 4 or 8
13046         depending on offset.
13047         (<bd>_<mode>): Similar, for alternative 0.
13048         (<bd>tf_<mode>): Ditto.
13050 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
13052         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
13054 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
13056         * doc/rtl.texi: Replace old RTX class names with new names.
13059 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
13061         * expmed.h (canonicalize_comparison): New declaration.
13062         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
13063         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
13064         * optabs.c (prepare_cmp_insn): Likewise.
13065         * rtl.h (unsigned_condition_p): New function which checks if a
13066         comparison operator is unsigned.
13068 2018-08-16  Nathan Sidwell  <nathan@acm.org>
13070         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
13071         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
13073 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
13075         PR target/84711
13076         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
13077         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
13078         (mov<mov>): ..this and enable unconditionally.
13080 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
13082         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
13084 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
13086         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
13087         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
13088         "Common" with "Target".
13090 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
13092         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
13093         * doc/invoke.texi (mmitigate-rop): Remove.
13094         * config/i386/i386.c: Do not include "regrename.h".
13095         (ix86_rop_should_change_byte_p, reg_encoded_number)
13096         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
13097         Remove.
13098         (ix86_reorg): Remove call to ix86_mitigate_rop.
13099         * config/i386/i386.md (attr "modrm_class"): Remove.
13100         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
13101         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
13102         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
13104 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
13106         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
13107         allow folding of mergeh() and mergel() for the float and double types.
13108         (fold_mergehl_helper): Rework to handle building a permute tree
13109         for float vectors.
13111 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
13113         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
13114         for TARGET_SSE.
13116 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
13118         * common.opt (fdiagnostics-show-labels): New option.
13119         * diagnostic-show-locus.c (class layout_range): Add field
13120         "m_label".
13121         (class layout): Add field "m_show_labels_p".
13122         (layout_range::layout_range): Add param "label" and use it to
13123         initialize m_label.
13124         (make_range): Pass in NULL for new "label" param of layout_range's
13125         ctor.
13126         (layout::layout): Initialize m_show_labels_p.
13127         (layout::maybe_add_location_range): Pass in loc_range->m_label
13128         when constructing layout_range instances.
13129         (struct line_label): New struct.
13130         (layout::print_any_labels): New member function.
13131         (layout::print_line): Call it if label-printing is enabled.
13132         (selftest::test_one_liner_labels): New test.
13133         (selftest::test_diagnostic_show_locus_one_liner): Call it.
13134         * diagnostic.c (diagnostic_initialize): Initialize
13135         context->show_labels_p.
13136         * diagnostic.h (struct diagnostic_context): Add field
13137         "show_labels_p".
13138         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
13139         -fno-diagnostics-show-labels.
13140         * dwarf2out.c (gen_producer_string): Add
13141         OPT_fdiagnostics_show_labels to the ignored options.
13142         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
13143         param.
13144         (gcc_rich_location::maybe_add_expr): Likewise.
13145         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
13146         label" param, defaulting to NULL.
13147         (gcc_rich_location::add_expr): Add "label" param.
13148         (gcc_rich_location::maybe_add_expr): Likewise.
13149         (class text_range_label): New class.
13150         (class range_label_for_type_mismatch): New class.
13151         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
13152         of format_warning_va.
13153         (fmtwarn_n): Likewise for new params of format_warning_n_va.
13154         * lto-wrapper.c (merge_and_complain): Add
13155         OPT_fdiagnostics_show_labels to the "pick one setting" options.
13156         (append_compiler_options): Likewise to the dropped options.
13157         (append_diag_options): Likewise to the passed-on options.
13158         * opts.c (common_handle_option): Handle the new option.
13159         * selftest-diagnostic.c
13160         (test_diagnostic_context::test_diagnostic_context): Enable
13161         show_labels_p.
13162         * substring-locations.c: Include "gcc-rich-location.h".
13163         (format_warning_n_va): Add "fmt_label" and "param_label" params
13164         and use them as appropriate.
13165         (format_warning_va): Add "fmt_label" and "param_label" params,
13166         passing them on to format_warning_n_va.
13167         (format_warning_at_substring): Likewise.
13168         (format_warning_at_substring_n): Likewise.
13169         * substring-locations.h (format_warning_va): Add "fmt_label" and
13170         "param_label" params.
13171         (format_warning_n_va): Likewise.
13172         (format_warning_at_substring): Likewise.
13173         (format_warning_at_substring_n): Likewise.
13174         * toplev.c (general_init): Initialize global_dc->show_labels_p.
13176 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
13178         PR testsuite/86519
13179         * builtins.c (expand_builtin_memcmp): Do not expand the call
13180         when overflow is detected.
13182 2018-08-15  Martin Sebor  <msebor@redhat.com>
13184         PR tree-optimization/71625
13185         * config/aarch64/aarch64-builtins.c
13186         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
13188 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
13190         * config/s390/s390.c (s390_reorg): Remove loop.
13192 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
13194         * config/darwin.c
13195          (darwin_function_switched_text_sections): Delete.
13196         * gcc/config/darwin.h
13197          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
13199 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
13201         PR target/81685
13202         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
13203         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
13204         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
13206 2018-08-15  Martin Liska  <mliska@suse.cz>
13208         PR tree-optimization/86925
13209         * predict.c (expr_expected_value_1): When taking
13210         later predictor, assign also probability.
13211         Use fold_build2_initializer_loc in order to fold
13212         the expression in -frounding-math.
13214 2018-08-14  Allan Sandfeld Jensen  <allan.jensen@qt.io>
13216         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
13217         patterns.
13218         (expand_vec_perm_1): Try the new method.
13220 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
13222         PR target/86547
13223         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
13224         Check whether lra_live_max_point is 0 before dividing.
13226 2018-08-14  Martin Sebor  <msebor@redhat.com>
13228         PR tree-optimization/86650
13229         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
13230         (vrp_prop::check_mem_ref): Same.
13232 2018-08-13  Liu Hao  <lh_mouse@126.com>
13234         * pretty-print.c (eat_esc_sequence): Swap the foreground and
13235         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
13236         and clear it thereafter, as it only works for DBCS.
13238 2018-08-13  Liu Hao  <lh_mouse@126.com>
13240         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
13241         handle returned by _get_osf_handle().
13243 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
13245         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
13246         for folding vec_perm.
13248 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
13250         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
13251         Add support for gimple-folding of vec_pack() and vec_unpack()
13252         intrinsics.
13254 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
13256         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
13257         vec_xst variants to the list.
13258         (rs6000_gimple_fold_builtin): Add support for folding unaligned
13259         vector loads and stores.
13261 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
13263         * config.gcc (rs6000-ibm-aix4.x): Delete.
13264         (rs6000-ibm-aix5.1): Delete.
13265         (rs6000-ibm-aix5.2): Delete.
13266         (rs6000-ibm-aix5.3): Delete.
13267         * config/rs6000/aix43.h: Delete.
13268         * config/rs6000/aix51.h: Delete.
13269         * config/rs6000/aix52.h: Delete.
13270         * config/rs6000/t-aix43: Delete.
13272 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
13274         * config/s390/s390.c (s390_decompose_constant_pool_ref):
13275         New function.
13276         (s390_decompose_address): Factor out constant pool ref
13277         decomposition.
13279 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13281         * config/nds32/nds32-predicates.c
13282         (nds32_can_use_bclr_p): Change return type as bool.
13283         (nds32_can_use_bset_p): Ditto.
13284         (nds32_can_use_btgl_p): Ditto.
13285         (nds32_can_use_bitci_p): Ditto.
13286         * config/nds32/nds32-protos.h
13287         (nds32_can_use_bclr_p): Change declaration.
13288         (nds32_can_use_bset_p): Ditto.
13289         (nds32_can_use_btgl_p): Ditto.
13290         (nds32_can_use_bitci_p): Ditto.
13292 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13294         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
13295         Support -msched-prolog-epilog option.
13296         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
13298 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13300         * common/config/nds32/nds32-common.c
13301         (nds32_option_optimization_table): Enalbe -malways-align.
13303 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
13305         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
13306         extra_headers.
13307         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
13308         OPT_misr_secure_ case.
13309         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
13310         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
13311         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
13312         secure attribute.
13313         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
13314         (nds32_isr_info): New field security_level.
13315         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
13316         * config/nds32/nds32.md (return_internal): Consider critical attribute.
13317         * config/nds32/nds32.opt (misr-secure): New option.
13318         * config/nds32/nds32_init.inc: New file.
13319         * config/nds32/nds32_isr.h: New file.
13321 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
13323         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
13324         Update comment for atomic instructions.
13325         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
13326         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
13327         Remove.
13328         (atomic_loaddi): Revise fence expansion to only emit fence prior to
13329         load for __ATOMIC_SEQ_CST model.
13330         (atomic_loaddi_1): Remove float register target.
13331         (atomic_storedi): Handle CONST_INT values.
13332         (atomic_storedi_1): Remove float register source.  Add special case
13333         for zero value.
13334         (memory_barrier): New expander and insn.
13336 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
13338         PR tree-optimization/86835
13339         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
13340         new_stmt after def_gsi, make sure to insert new_square_stmt after
13341         that stmt, not 2 stmts before it.
13343 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
13345         PR target/82418
13346         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
13347         instead of SWI48.
13349 2018-08-10  Martin Liska  <mliska@suse.cz>
13351         PR target/83610
13352         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
13353         function type.
13354         * builtins.c (expand_builtin_expect_with_probability):
13355         New function.
13356         (expand_builtin_expect_with_probability): New function.
13357         (build_builtin_expect_predicate): Add new argumnet probability
13358         for BUILT_IN_EXPECT_WITH_PROBABILITY.
13359         (fold_builtin_expect):
13360         (fold_builtin_2):
13361         (fold_builtin_3):
13362         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
13363         * builtins.h (fold_builtin_expect): Set new argument.
13364         * doc/extend.texi: Document __builtin_expect_with_probability.
13365         * doc/invoke.texi: Likewise.
13366         * gimple-fold.c (gimple_fold_call): Pass new argument.
13367         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
13368         also BUILT_IN_EXPECT_WITH_PROBABILITY.
13369         * predict.c (get_predictor_value): New function.
13370         (expr_expected_value): Add new argument probability. Assume
13371         that predictor and probability are always non-null.
13372         (expr_expected_value_1): Likewise.  For __builtin_expect and
13373         __builtin_expect_with_probability set probability.  Handle
13374         combination in binary expressions.
13375         (tree_predict_by_opcode): Simplify code by simply calling
13376         get_predictor_value.
13377         (pass_strip_predict_hints::execute): Add handling of
13378         BUILT_IN_EXPECT_WITH_PROBABILITY.
13379         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
13380         new predictor.
13381         * tree.h (DECL_BUILT_IN_P): New function.
13383 2018-08-10  Martin Liska  <mliska@suse.cz>
13385         PR tree-optimization/85799
13386         * passes.def: Add argument for pass_strip_predict_hints.
13387         * predict.c (class pass_strip_predict_hints): Add new argument
13388         early_p.
13389         (strip_predictor_early): New function.
13390         (pass_strip_predict_hints::execute): Call the function to
13391         strip predictors.
13392         (strip_predict_hints): New function.
13393         * predict.def: Fix comment.
13395 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
13397         * Makefile.in: Clarify which tm.texi to copy over to assert the
13398         right to grant a GFDL license for all.
13400 2018-08-09  Jeff Law  <law@redhat.com>
13402         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
13403         unused variable.
13405 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
13407         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
13408         prototype.
13410 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
13412         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
13413         reductions for variable-length vectors.
13415 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
13417         PR other/84889
13418         * common.opt (fdiagnostics-show-line-numbers): New option.
13419         * diagnostic-show-locus.c (class layout): Add fields
13420         "m_show_line_numbers_p" and "m_linenum_width";
13421         (num_digits): New function.
13422         (test_num_digits): New function.
13423         (layout::layout): Initialize new fields.  Update m_x_offset
13424         logic to handle any left margin.
13425         (layout::print_source_line): Print line number when requested.
13426         (layout::start_annotation_line): New member function.
13427         (layout::print_annotation_line): Call it.
13428         (layout::print_leading_fixits): Likewise.
13429         (layout::print_trailing_fixits): Likewise.  Update calls to
13430         move_to_column for new parameter.
13431         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
13432         it to potentially call start_annotation_line.
13433         (layout::show_ruler): Call start_annotation_line.
13434         (selftest::test_line_numbers_multiline_range): New selftest.
13435         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
13436         and selftest::test_line_numbers_multiline_range.
13437         * diagnostic.c (diagnostic_initialize): Initialize
13438         show_line_numbers_p.
13439         * diagnostic.h (struct diagnostic_context): Add field
13440         "show_line_numbers_p".
13441         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
13442         -fno-diagnostics-show-line-numbers.
13443         * dwarf2out.c (gen_producer_string): Add
13444         OPT_fdiagnostics_show_line_numbers to the ignored options.
13445         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
13446         one setting" options.
13447         (append_compiler_options): Likewise to the dropped options.
13448         (append_diag_options): Likewise to the passed-on options.
13449         * opts.c (common_handle_option): Handle the new option.
13450         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
13452 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13454         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
13455         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
13456         third argument of type "const signed char" to descriptions of
13457         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
13458         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
13459         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
13460         __builtin_bcdsub_ov functions.
13462 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
13464         PR tree-optimization/86858
13465         * tree-vect-loop.c (vect_is_simple_reduction): Restore
13466         flow_bb_inside_loop_p calls.
13468 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
13470         PR tree-optimization/86871
13471         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
13472         instead of gimple_assign_lhs.
13474 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
13476         PR target/86887
13477         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
13478         register constraint to operand 0.
13479         (add<mode>3_carryinC): Likewise.
13480         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
13482 2018-08-09  Martin Liska  <mliska@suse.cz>
13484         PR c/86895
13485         * common.opt: Remove extra line.
13487 2018-08-09  Martin Liska  <mliska@suse.cz>
13489         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
13490         at the end of a line, make first letter capital and end up
13491         a sentence with a dot.
13492         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
13493         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
13494         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
13495         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
13496         (PARAM_MAX_ISL_OPERATIONS): Likewise.
13497         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
13498         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
13499         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
13500         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
13501         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
13502         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
13503         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
13504         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
13505         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
13506         (PARAM_TREE_REASSOC_WIDTH): Likewise.
13507         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
13508         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
13509         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
13511 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
13513         PR target/84332
13514         * config/s390/s390.c (s390_option_override_internal): Reduce the
13515         stack-clash-protection-probe-interval param if it would be too big
13516         for z900.
13518 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
13520         PR target/46179
13521         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
13522         * config/m68k/m68k.c (handle_move_double): Don't call
13523         m68k_final_prescan_insn.
13524         (m68k_adjust_decorated_operand): Renamed from
13525         m68k_final_prescan_insn, remove first and third operand and
13526         simplify.
13527         (print_operand): Call it.
13528         (print_operand_address): Call it.
13530 2018-08-08  Nathan Sidwell  <nathan@acm.org>
13532         * diagnostic.c (diagnostic_report_current_module): Use
13533         linemap_included_from & linemap_included_from_linemap.
13535 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
13537         * config/aarch64/aarch64-cores.def: Add phecda core.
13538         * config/aarch64/aarch64-tune.md: Regenerate.
13539         * doc/invoke.texi: Add phecda core.
13541 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
13543         PR target/85295
13544         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
13545         definitions.
13546         * config/s390/s390.md ("movti"): Add more alternatives for
13547         constant to GPR copies.
13549 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
13551         * config/s390/s390.c: Fix whitespace damage throughout the file.
13552         * config/s390/s390.h: Likewise.
13553         * config/s390/tpf.h: Likewise.
13555 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
13557         * config/s390/s390.c (s390_loadrelative_operand_p):
13558         Remove TARGET_CPU_ZARCH usages.
13559         (s390_rtx_costs): Likewise.
13560         (s390_legitimate_constant_p): Likewise.
13561         (s390_cannot_force_const_mem): Likewise.
13562         (legitimate_reload_constant_p): Likewise.
13563         (s390_preferred_reload_class): Likewise.
13564         (legitimize_pic_address): Likewise.
13565         (legitimize_tls_address): Likewise.
13566         (s390_split_branches): Removed.
13567         (s390_add_execute): Removed.
13568         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
13569         (s390_mainpool_start): Likewise.
13570         (s390_mainpool_finish): Likewise.
13571         (s390_mainpool_cancel): Removed.
13572         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
13573         (s390_chunkify_cancel): Likewise.
13574         (s390_return_addr_rtx): Likewise.
13575         (s390_register_info): Remove split_branches_pending_p uages.
13576         (s390_optimize_register_info): Likewise.
13577         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
13578         split_branches_pending_p usages.
13579         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
13580         (s390_load_got): Likewise.
13581         (s390_expand_split_stack_prologue): Likewise.
13582         (output_asm_nops): Likewise.
13583         (s390_function_profiler): Likewise.
13584         (s390_emit_call): Likewise.
13585         (s390_conditional_register_usage): Likewise.
13586         (s390_optimize_prologue): Likewise.
13587         (s390_reorg): Remove TARGET_CPU_ZARCH and
13588         split_branches_pending_p usages.
13589         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
13590         usages.
13591         (s390_output_indirect_thunk_function): Likewise.
13592         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
13593         (TARGET_CPU_ZARCH_P): Removed.
13594         (struct machine_function): Remove split_branches_pending_p.
13595         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
13597 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
13599         * common/config/s390/s390-common.c (processor_flags_table):
13600         Remove flags.
13601         * config.gcc: Remove with_arch/with_tune support.
13602         * config/s390/2064.md: Remove cpu attribute comparisons.
13603         * config/s390/driver-native.c (s390_host_detect_local_cpu):
13604         Remove MTN.
13605         * config/s390/linux.h (ASM_SPEC):
13606         Remove -march support.
13607         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
13608         Use a table to get an arch level.
13609         * config/s390/s390-opts.h (enum processor_type):
13610         Remove enum values.
13611         * config/s390/s390.c
13612         (processor_table): Remove entries, add arch_level values.
13613         (s390_issue_rate): Remove cases.
13614         (s390_option_override): Adjust
13615         s390_option_override_internal() call.
13616         (s390_option_override_internal): Remove deprecation warning.
13617         (s390_valid_target_attribute_tree): Adjust
13618         s390_option_override_internal() call.
13619         * config/s390/s390.h (struct s390_processor):
13620         Share with s390-c.c, add arch_level field.
13621         * config/s390/s390.md:
13622         Remove occurrences in cpu attribute.
13623         * config/s390/s390.opt: Remove -march/-mtune support.
13624         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
13625         * doc/invoke.texi: Remove deprecation warning.
13627 2018-08-08  Luis Machado  <luis.machado@linaro.org>
13629         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
13630         global.
13631         (qdf24xx_tunings): Set vector cost structure to
13632         qdf24xx_vector_cost.
13634         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
13635         <register_sextend>: Set to 3.
13637 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
13639         PR target/86838
13640         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
13641         * config/aarch64/aarch64-simd.md
13642         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
13643         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
13644         (aarch64_frecpx<mode>): ...this new pattern.
13645         * config/aarch64/aarch64-simd-builtins.def: Remove comment
13646         about aarch64_frecp<FRECP:frecp_suffix><mode>.
13648 2018-08-07  Martin Liska  <mliska@suse.cz>
13650         PR middle-end/83023
13651         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
13652         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
13653         * predict.def (PRED_MALLOC_NONNULL): New predictor.
13654         * doc/extend.texi: Document that malloc attribute adds
13655         hit to compiler.
13657 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
13659         PR target/86785
13660         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13661         Define to speculation_safe_value_not_needed.
13663 2018-08-06  Jeff Law  <law@redhat.com>
13665         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
13666         the vr_values instance to cprop_into_stmt.
13667         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
13668         (cprop_operand): Also query EVRP to determine if OP is a constant.
13670 2018-08-06  Nathan Sidwell  <nathan@acm.org>
13672         * diagnostic.c (diagnostic_report_current_module): Reroll
13673         included-at loop.  Translate text.
13675 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
13677         * function-tests.c (selftest::test_expansion_to_rtl): Call
13678         free_after_compilation.
13680 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13682         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
13684 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
13686         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
13687         loops with memory block operations from getting unrolled.
13689 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
13691         PR target/86807
13692         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13693         Define to speculation_safe_value_not_needed.
13695 2018-08-06  Jeff Law  <law@redhat.com>
13697         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
13698         assert.
13700 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13702         PR target/86662
13703         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
13704         with all enabled __intN types.
13706         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
13708 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13710         * alias.c (record_set): Check for clobber high.
13711         * cfgexpand.c (expand_gimple_stmt): Likewise.
13712         * combine-stack-adj.c (single_set_for_csa): Likewise.
13713         * combine.c (find_single_use_1): Likewise.
13714         (set_nonzero_bits_and_sign_copies): Likewise.
13715         (get_combine_src_dest): Likewise.
13716         (is_parallel_of_n_reg_sets): Likewise.
13717         (try_combine): Likewise.
13718         (record_dead_and_set_regs_1): Likewise.
13719         (reg_dead_at_p_1): Likewise.
13720         (reg_dead_at_p): Likewise.
13721         * dce.c (deletable_insn_p): Likewise.
13722         (mark_nonreg_stores_1): Likewise.
13723         (mark_nonreg_stores_2): Likewise.
13724         * df-scan.c (df_find_hard_reg_defs): Likewise.
13725         (df_uses_record): Likewise.
13726         (df_get_call_refs): Likewise.
13727         * dwarf2out.c (mem_loc_descriptor): Likewise.
13728         * haifa-sched.c (haifa_classify_rtx): Likewise.
13729         * ira-build.c (create_insn_allocnos): Likewise.
13730         * ira-costs.c (scan_one_insn): Likewise.
13731         * ira.c (equiv_init_movable_p): Likewise.
13732         (rtx_moveable_p): Likewise.
13733         (interesting_dest_for_shprep): Likewise.
13734         * jump.c (mark_jump_label_1): Likewise.
13735         * postreload-gcse.c (record_opr_changes): Likewise.
13736         * postreload.c (reload_cse_simplify): Likewise.
13737         (struct reg_use): Add source expr.
13738         (reload_combine): Check for clobber high.
13739         (reload_combine_note_use): Likewise.
13740         (reload_cse_move2add): Likewise.
13741         (move2add_note_store): Likewise.
13742         * print-rtl.c (print_pattern): Likewise.
13743         * recog.c (decode_asm_operands): Likewise.
13744         (store_data_bypass_p): Likewise.
13745         (if_test_bypass_p): Likewise.
13746         * regcprop.c (kill_clobbered_value): Likewise.
13747         (kill_set_value): Likewise.
13748         * reginfo.c (reg_scan_mark_refs): Likewise.
13749         * reload1.c (maybe_fix_stack_asms): Likewise.
13750         (eliminate_regs_1): Likewise.
13751         (elimination_effects): Likewise.
13752         (mark_not_eliminable): Likewise.
13753         (scan_paradoxical_subregs): Likewise.
13754         (forget_old_reloads_1): Likewise.
13755         * reorg.c (find_end_label): Likewise.
13756         (try_merge_delay_insns): Likewise.
13757         (redundant_insn): Likewise.
13758         (own_thread_p): Likewise.
13759         (fill_simple_delay_slots): Likewise.
13760         (fill_slots_from_thread): Likewise.
13761         (dbr_schedule): Likewise.
13762         * resource.c (update_live_status): Likewise.
13763         (mark_referenced_resources): Likewise.
13764         (mark_set_resources): Likewise.
13765         * rtl.c (copy_rtx): Likewise.
13766         * rtlanal.c (reg_referenced_p): Likewise.
13767         (single_set_2): Likewise.
13768         (noop_move_p): Likewise.
13769         (note_stores): Likewise.
13770         * sched-deps.c (sched_analyze_reg): Likewise.
13771         (sched_analyze_insn): Likewise.
13773 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13775         * cse.c (invalidate_reg): New function extracted from...
13776         (invalidate): ...here.
13777         (canonicalize_insn): Check for clobber high.
13778         (invalidate_from_clobbers): invalidate clobber highs.
13779         (invalidate_from_sets_and_clobbers): Likewise.
13780         (count_reg_usage): Check for clobber high.
13781         (insn_live_p): Likewise.
13782         * cselib.c (cselib_expand_value_rtx_1):Likewise.
13783         (cselib_invalidate_regno): Check for clobber in setter.
13784         (cselib_invalidate_rtx): Pass through setter.
13785         (cselib_invalidate_rtx_note_stores):
13786         (cselib_process_insn): Check for clobber high.
13787         * cselib.h (cselib_invalidate_rtx): Add operand.
13789 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13791         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
13792         (mark_not_eliminable): Likewise.
13793         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
13794         * lra-lives.c (process_bb_lives): Check for clobber high.
13795         * lra.c (new_insn_reg): Remember clobber highs.
13796         (collect_non_operand_hard_regs): Check for clobber high.
13797         (lra_set_insn_recog_data): Likewise.
13798         (add_regs_to_insn_regno_info): Likewise.
13799         (lra_update_insn_regno_info): Likewise.
13801 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13803         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
13804         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
13806 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13808         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
13809         (copy_insn_1): Likewise.
13810         (gen_hard_reg_clobber_high): New gen function.
13811         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
13812         * genemit.c (gen_exp): Likewise.
13813         (gen_emit_seq): Pass through info.
13814         (gen_insn): Check for CLOBBER_HIGH.
13815         (gen_expand): Pass through info.
13816         (gen_split): Likewise.
13817         (output_add_clobbers): Likewise.
13818         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
13819         (remove_clobbers): Likewise.
13820         * rtl.h (gen_hard_reg_clobber_high): New declaration.
13822 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13824         * doc/rtl.texi (clobber_high): Add.
13825         (parallel): Add in clobber high
13826         * rtl.c (rtl_check_failed_code3): Add function.
13827         * rtl.def (CLOBBER_HIGH): Add expression.
13828         * rtl.h (RTL_CHECKC3): Add macro.
13829         (rtl_check_failed_code3): Add declaration.
13830         (XC3EXP): Add macro.
13832 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
13834         PR target/86386
13835         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
13836         cfun->machine->max_used_stack_alignment if needed.
13838 2018-08-04  Martin Sebor  <msebor@redhat.com>
13840         PR tree-optimization/86571
13841         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
13842         NaN output to 4.
13844 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
13846         PR target/86799
13847         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13848         Define.
13850 2018-08-03  Jeff Law  <law@redhat.com>
13852         PR target/86795
13853         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13854         Define to speculation_safe_value_not_needed.
13856 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
13858         * doc/gcov.texi (-x): Remove duplicate "to".
13859         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
13860         (-Wif-not-aligned): Remove duplicate "is".
13861         (-flto): Remove duplicate "the".
13862         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
13863         duplicate "v5.00.b".
13864         (MSP430 Options): Remove duplicate "and" from the description
13865         of "-mgprel-sec=regexp".
13866         (x86 Options): Remove duplicate copies of "vmldLog102" and
13867         vmlsLog104 from description of "-mveclibabi=type".
13869 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
13871         * internal-fn.h (first_commutative_argument): Declare.
13872         * internal-fn.c (first_commutative_argument): New function.
13873         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
13874         restrictions for pattern statements.  Use first_commutative_argument
13875         to look for commutative operands in calls to internal functions.
13877 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
13879         * Makefile.in (wide-int-range.o): New.
13880         * tree-vrp.c: Move all the wide_int_* functions to...
13881         * wide-int-range.cc: ...here.
13882         * tree-vrp.h: Move all the wide_int_* prototypes to...
13883         * wide-int-range.h: ...here.
13885 2018-08-03  Tom de Vries  <tdevries@suse.de>
13887         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
13888         UI_NONE.
13889         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
13890         * except.c (output_function_exception_table): Do early exit if
13891         targetm_common.except_unwind_info (&global_options) == UI_NONE.
13893 2018-08-03  Martin Liska  <mliska@suse.cz>
13895         * predict.c (dump_prediction): Change to 2 digits
13896         in fraction part.
13898 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
13900         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
13901         neon_dup_q to...
13902         (falkor_am_1_gtov_gtov): ... a new insn reservation.
13904 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
13906         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
13907         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
13908         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
13909         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
13910         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
13912 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
13914         * diagnostic-show-locus.c (diagnostic_show_locus): Use
13915         pp_take_prefix when saving the existing prefix.
13916         * diagnostic.c (diagnostic_append_note): Likewise.
13917         * langhooks.c (lhd_print_error_function): Likewise.
13918         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
13919         param's type.  Free the existing prefix.
13920         (pp_take_prefix): New function.
13921         (pretty_printer::pretty_printer): Drop the prefix parameter.
13922         Rename the length parameter to match the comment.
13923         (pretty_printer::~pretty_printer): Free the prefix.
13924         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
13925         parameter.
13926         (struct pretty_printer): Drop the "const" from "prefix" field's
13927         type and clarify memory management.
13928         (pp_set_prefix): Drop the "const" from the 2nd param.
13929         (pp_take_prefix): New decl.
13931 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
13933         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
13934         for word_mode_ok here instead of passing as argument.
13935         (expand_block_compare): Change select_block_compare_mode() call.
13936         (expand_strncmp_gpr_sequence): New function.
13937         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
13939 2018-08-02  Jeff Law  <law@redhat.com>
13941         PR target/86790
13942         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13943         Define to speculation_safe_value_not_needed.
13945         PR target/86784
13946         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13947         Define to speculation_safe_value_not_needed.
13949 2018-08-02  Tom de Vries  <tdevries@suse.de>
13951         PR target/86660
13952         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
13953         function.  Return UI_TARGET unconditionally.
13954         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
13955         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
13957 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
13959         * genemit.c (print_overload_test): Fix typo.
13961 2018-08-02  Richard Biener  <rguenther@suse.de>
13963         PR tree-optimization/86816
13964         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
13965         which checks for value availability before querying it.
13966         (gvn_uses_equal): Use it.
13967         (same_succ_hash): Likewise.
13968         (gimple_equal_p): Likewise.
13970 2018-08-02  Nick Clifton  <nickc@redhat.com>
13972         PR target/86813
13973         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13974         Define to speculation_safe_value_not_needed.
13976         PR target/86810
13977         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13978         Define to speculation_safe_value_not_needed.
13980         PR target/86810
13981         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13982         Define to speculation_safe_value_not_needed.
13984         PR target/86803
13985         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13986         Define to speculation_safe_value_not_needed.
13988         PR target/86797
13989         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13990         Define to speculation_safe_value_not_needed.
13992         PR target/86791
13993         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13994         Define to speculation_safe_value_not_needed.
13996         PR target/86789
13997         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13998         Define to speculation_safe_value_not_needed.
14000         PR target/86787
14001         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
14002         Define to speculation_safe_value_not_needed.
14004         PR target/86782
14005         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
14006         speculation_safe_value_not_needed.
14008         PR target/86781
14009         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
14010         to speculation_safe_value_not_needed.
14012 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
14014         * doc/md.texi: Expand the documentation of instruction names
14015         to mention port-local uses.  Document '@' in pattern names.
14016         * read-md.h (overloaded_instance, overloaded_name): New structs.
14017         (mapping): Declare.
14018         (md_reader::handle_overloaded_name): New member function.
14019         (md_reader::get_overloads): Likewise.
14020         (md_reader::m_first_overload): New member variable.
14021         (md_reader::m_next_overload_ptr): Likewise.
14022         (md_reader::m_overloads_htab): Likewise.
14023         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
14024         m_next_overload_ptr and m_overloads_htab.
14025         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
14026         (get_mode_token, get_code_token, get_int_token): New functions.
14027         (map_attr_string): Add an optional argument that passes back
14028         the associated iterator.
14029         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
14030         (md_reader::handle_overloaded_name, add_overload_instance): New
14031         functions.
14032         (apply_iterators): Handle '@' names.  Report an error if '@'
14033         is used without iterators.
14034         (initialize_iterators): Initialize the new iterator_group fields.
14035         * genopinit.c (handle_overloaded_code_for)
14036         (handle_overloaded_gen): New functions.
14037         (main): Use them to print declarations of maybe_code_for_* and
14038         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
14039         * genemit.c (print_overload_arguments, print_overload_test)
14040         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
14041         (main): Use it to print definitions of maybe_code_for_* and
14042         maybe_gen_* functions.
14043         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
14044         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
14045         instead of explicit mode checks.
14046         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
14047         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
14048         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
14049         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
14050         (aarch64_expand_compare_and_swap): Likewise
14051         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
14052         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
14053         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
14054         (aarch64_constant_pool_reload_icode): Delete.
14055         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
14056         instead of aarch64_constant_pool_reload_icode.  Use
14057         code_for_aarch64_reload_mov instead of explicit mode checks.
14058         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
14059         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
14060         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
14061         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
14062         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
14063         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
14064         (aarch64_atomic_load_op_code): Delete.
14065         (aarch64_emit_atomic_load_op): Likewise.
14066         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
14067         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
14068         instead of aarch64_emit_atomic_load_op.
14069         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
14070         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
14071         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
14072         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
14073         character before the pattern name.
14074         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
14075         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
14076         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
14077         (aarch64_frecps<mode>): Likewise.
14078         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
14079         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
14080         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
14081         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
14082         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
14084 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
14086         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
14087         Allow HFmode constants if TARGET_FP_F16INST.
14089 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
14091         PR target/86014
14092         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
14093         No longer check last store for clobber of address register.
14095 2018-08-02  Martin Liska  <mliska@suse.cz>
14097         PR gcov-profile/86817
14098         * gcov.c (process_all_functions): New function.
14099         (main): Call it.
14100         (process_file): Move functions processing to
14101         process_all_functions.
14103 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
14105         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
14106         "const" to the "gimple *" and "rtx_insn *" parameters.
14107         * dumpfile.h (dump_user_location_t::dump_user_location_t):
14108         Likewise.
14109         (dump_location_t::dump_location_t): Likewise.
14111 2018-08-01  Martin Sebor  <msebor@redhat.com>
14113         PR tree-optimization/86650
14114         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
14115         rather than a "gcall *".  Directly pass the data of interest
14116         to percent_K_format, rather than building a temporary CALL_EXPR
14117         to hold it.
14118         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
14119         (gimple_fold_builtin_strncat): Adjust.
14120         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
14121         gcall* argument with gimple*.
14122         * gimple-ssa-warn-restrict.c (check_call): Same.
14123         (wrestrict_dom_walker::before_dom_children): Same.
14124         (builtin_access::builtin_access): Same.
14125         (check_bounds_or_overlap): Same
14126         (maybe_diag_overlap): Same.
14127         (maybe_diag_offset_bounds): Same.
14128         * tree-diagnostic.c (default_tree_printer): Move usage of
14129         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
14130         to this callsite.
14131         * tree-pretty-print.c (percent_K_format): Add argument.
14132         * tree-pretty-print.h: Add argument.
14133         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
14134         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
14135         (maybe_diag_stxncpy_trunc): Same.
14136         (handle_builtin_stxncpy): Same.
14137         (handle_builtin_strcat): Same.
14139 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14141         * match.pd: Optimise pointer range checks.
14143 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14145         PR tree-optimization/86758
14146         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
14147         to remove pattern statements.
14149 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14151         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
14152         result of dfs_enumerate_from when constructing stmt_vec_infos,
14153         instead of additionally calling get_loop_body.
14155 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14157         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
14158         parameter.
14159         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
14160         When creating an iv, assert that the step is not known to be zero.
14161         (vect_setup_realignment): Update call accordingly.
14162         * tree-vect-stmts.c (vectorizable_store): Likewise.
14163         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
14165 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14167         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
14168         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
14169         (vectorizable_reduction): Likewise.
14170         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14171         (vect_detect_hybrid_slp_stmts): Likewise.
14172         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
14174 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
14176         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
14177         (wide_int_set_zero_nonzero_bits): ...this.
14178         (zero_nonzero_bits_from_vr): Rename to...
14179         (vrp_set_zero_nonzero_bits): ...this.
14180         (extract_range_from_multiplicative_op_1): Abstract wide int
14181         code...
14182         (wide_int_range_multiplicative_op): ...here.
14183         (extract_range_from_binary_expr_1): Extract wide int binary
14184         operations into their own functions.
14185         (wide_int_range_lshift): New.
14186         (wide_int_range_can_optimize_bit_op): New.
14187         (wide_int_range_shift_undefined_p): New.
14188         (wide_int_range_bit_xor): New.
14189         (wide_int_range_bit_ior): New.
14190         (wide_int_range_bit_and): New.
14191         (wide_int_range_trunc_mod): New.
14192         (extract_range_into_wide_ints): New.
14193         (vrp_shift_undefined_p): New.
14194         (extract_range_from_multiplicative_op): New.
14195         (vrp_can_optimize_bit_op): New.
14196         * tree-vrp.h (value_range::dump): New.
14197         (wide_int_range_multiplicative_op): New.
14198         (wide_int_range_lshift):New.
14199         (wide_int_range_shift_undefined_p): New.
14200         (wide_int_range_bit_xor): New.
14201         (wide_int_range_bit_ior): New.
14202         (wide_int_range_bit_and): New.
14203         (wide_int_range_trunc_mod): New.
14204         (zero_nonzero_bits_from_bounds): Rename to...
14205         (wide_int_set_zero_nonzero_bits): ...this.
14206         (zero_nonzero_bits_from_vr): Rename to...
14207         (vrp_set_zero_nonzero_bits): ...this.
14208         (range_easy_mask_min_max): Rename to...
14209         (wide_int_range_can_optimize_bit_op): this.
14210         * vr-values.c (simplify_bit_ops_using_ranges): Rename
14211         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
14213 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14215         * tree-vectorizer.h (vect_orig_stmt): New function.
14216         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
14217         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
14218         (vect_create_epilog_for_reduction): Likewise.
14219         (vectorizable_live_operation): Likewise.
14220         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
14221         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
14222         * tree-vect-stmts.c (vectorizable_call): Likewise.
14223         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
14225 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14227         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
14228         argument.
14229         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
14230         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
14231         (vect_transform_loop): Likewise.
14232         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
14234 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14236         * tree-vectorizer.h (vect_schedule_slp): Return void.
14237         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
14238         (vect_schedule_slp): Likewise.
14240 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14242         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
14243         argument.
14244         (vect_transform_loop): Update calls accordingly.  Schedule SLP
14245         instances before the main loop, if any exist.
14247 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
14249         PR tree-optimization/86749
14250         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
14251         If the lhs is used in a COND_EXPR, check that it is being used
14252         as the "then" or "else" value.
14254 2018-08-01  Tom de Vries  <tdevries@suse.de>
14256         PR target/86800
14257         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
14258         speculation_safe_value_not_needed.
14260 2018-08-01  Richard Biener  <rguenther@suse.de>
14262         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
14263         as base and offset.
14265 2018-08-01  Martin Liska  <mliska@suse.cz>
14267         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
14268         format how successful transformation is dumped.
14269         (gimple_mod_pow2_value_transform): Likewise.
14270         (gimple_mod_subtract_transform): Likewise.
14271         (gimple_stringops_transform): Likewise.
14273 2018-08-01  Martin Liska  <mliska@suse.cz>
14275         PR value-prof/35543
14276         * value-prof.c (interesting_stringop_to_profile_p):
14277         Simplify the code and add BUILT_IN_MEMMOVE.
14278         (gimple_stringops_transform): Likewise.
14280 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
14282         * config/aarch64/aarch64-simd.md
14283         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
14284         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
14285         use GPI iterator instead of SI mode.
14287 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
14289         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
14290         rs6000_speculation_barrier.
14291         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
14292         new barrier pattern name.
14294 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
14296         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
14297         (speculation_barrier): New insn.
14299 2018-08-01  Richard Biener  <rguenther@suse.de>
14301         PR bootstrap/86724
14302         * graphite.h: Include isl/id.h and isl/space.h to allow build
14303         with ISL 0.20.
14305 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
14307         PR target/86651
14308         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14309         mode for COFF targets.
14310         * defaults.h (TARGET_COFF): Define.
14311         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
14312         TARGET_COFF): Define.
14313         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
14314         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
14315         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
14317 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
14319         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
14320         * gimple-streamer-out.c (output_bb): Save it.
14321         * lto-streamer-in.c (input_struct_function_base): Restore
14322         instance discriminator if available.  Create map on demand.
14323         * lto-streamer-out.c (output_struct_function_base): Save it if
14324         available.
14325         * final.c (decl_to_instance_map): Document LTO strategy.
14327 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
14328             Olivier Hainque  <hainque@adacore.com>
14330         * debug.h (decl_to_instance_map_t): New type.
14331         (decl_to_instance_map): Declare.
14332         (maybe_create_decl_to_instance_map): New inline function.
14333         * final.c (bb_discriminator, last_bb_discriminator): New statics,
14334         to track basic block discriminators.
14335         (final_start_function_1): Initialize them.
14336         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
14337         bb_discriminator.
14338         (decl_to_instance_map): New variable.
14339         (map_decl_to_instance, maybe_set_discriminator): New functions.
14340         (notice_source_line): Set discriminator.
14342 2018-07-31  Ian Lance Taylor  <iant@golang.org>
14344         * targhooks.c (default_have_speculation_safe_value): Add
14345         ATTRIBUTE_UNUSED.
14347 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
14349         * dump-context.h: Include "pretty-print.h".
14350         (dump_context::refresh_dumps_are_enabled): New decl.
14351         (dump_context::emit_item): New decl.
14352         (class dump_context): Add fields "m_test_pp" and
14353         "m_test_pp_flags".
14354         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14355         (temp_dump_context::get_dumped_text): New decl.
14356         (class temp_dump_context): Add field "m_pp".
14357         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
14358         (dump_context::refresh_dumps_are_enabled): ...and add a test for
14359         m_test_pp.
14360         (set_dump_file): Update for above change.
14361         (set_alt_dump_file): Likewise.
14362         (dump_loc): New overload, taking a pretty_printer *.
14363         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
14364         to any test pretty-printer.
14365         (make_item_for_dump_gimple_stmt): New function, adapted from
14366         optinfo::add_gimple_stmt.
14367         (dump_context::dump_gimple_stmt): Call it, and use the result,
14368         eliminating the direct usage of dump_file and alt_dump_file in
14369         favor of indirectly using them via emit_item.
14370         (make_item_for_dump_gimple_expr): New function, adapted from
14371         optinfo::add_gimple_expr.
14372         (dump_context::dump_gimple_expr): Call it, and use the result,
14373         eliminating the direct usage of dump_file and alt_dump_file in
14374         favor of indirectly using them via emit_item.
14375         (make_item_for_dump_generic_expr): New function, adapted from
14376         optinfo::add_tree.
14377         (dump_context::dump_generic_expr): Call it, and use the result,
14378         eliminating the direct usage of dump_file and alt_dump_file in
14379         favor of indirectly using them via emit_item.
14380         (make_item_for_dump_printf_va): New function, adapted from
14381         optinfo::add_printf_va.
14382         (make_item_for_dump_printf): New function.
14383         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
14384         and use the result, eliminating the direct usage of dump_file and
14385         alt_dump_file in favor of indirectly using them via emit_item.
14386         (make_item_for_dump_dec): New function.
14387         (dump_context::dump_dec): Call it, and use the result,
14388         eliminating the direct usage of dump_file and alt_dump_file in
14389         favor of indirectly using them via emit_item.
14390         (make_item_for_dump_symtab_node): New function, adapted from
14391         optinfo::add_symtab_node.
14392         (dump_context::dump_symtab_node): Call it, and use the result,
14393         eliminating the direct usage of dump_file and alt_dump_file in
14394         favor of indirectly using them via emit_item.
14395         (dump_context::begin_scope): Reimplement, avoiding direct usage
14396         of dump_file and alt_dump_file in favor of indirectly using them
14397         via emit_item.
14398         (dump_context::emit_item): New member function.
14399         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14400         Set up test pretty-printer on the underlying context.  Call
14401         refresh_dumps_are_enabled.
14402         (temp_dump_context::~temp_dump_context): Call
14403         refresh_dumps_are_enabled.
14404         (temp_dump_context::get_dumped_text): New member function.
14405         (selftest::verify_dumped_text): New function.
14406         (ASSERT_DUMPED_TEXT_EQ): New macro.
14407         (selftest::test_capture_of_dump_calls): Run all tests twice, with
14408         and then without optinfo enabled.  Add uses of
14409         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
14410         * dumpfile.h: Update comment for the dump_* API.
14411         * optinfo-emit-json.cc
14412         (selftest::test_building_json_from_dump_calls): Update for new
14413         param for temp_dump_context ctor.
14414         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
14415         and "m_owned" field.
14416         (optinfo_item::~optinfo_item): Likewise.
14417         (optinfo::add_item): New member function.
14418         (optinfo::emit): Update comment.
14419         (optinfo::add_string): Delete.
14420         (optinfo::add_printf): Delete.
14421         (optinfo::add_printf_va): Delete.
14422         (optinfo::add_gimple_stmt): Delete.
14423         (optinfo::add_gimple_expr): Delete.
14424         (optinfo::add_tree): Delete.
14425         (optinfo::add_symtab_node): Delete.
14426         (optinfo::add_dec): Delete.
14427         * optinfo.h (class dump_context): New forward decl.
14428         (optinfo::add_item): New decl.
14429         (optinfo::add_string): Delete.
14430         (optinfo::add_printf): Delete.
14431         (optinfo::add_printf_va): Delete.
14432         (optinfo::add_gimple_stmt): Delete.
14433         (optinfo::add_gimple_expr): Delete.
14434         (optinfo::add_tree): Delete.
14435         (optinfo::add_symtab_node): Delete.
14436         (optinfo::add_dec): Delete.
14437         (optinfo::add_poly_int): Delete.
14438         (optinfo_item::optinfo_item): Remove "owned" param.
14439         (class optinfo_item): Remove field "m_owned".
14441 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14443         PR middle-end/86705
14444         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
14445         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
14446         requested variable alignment.
14447         (expand_one_ssa_partition): Likewise.
14448         (expand_one_var): Likewise.
14450 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14452         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
14453         to speculation_safe_value_not_needed.
14455 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14457         * targhooks.h (speculation_safe_value_not_needed): New prototype.
14458         * targhooks.c (speculation_safe_value_not_needed): New function.
14459         * target.def (have_speculation_safe_value): Update documentation.
14460         * doc/tm.texi: Regenerated.
14462 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14464         * config/aarch64/iterators.md (ALLI_TI): New iterator.
14465         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
14466         expand.
14467         (despeculate_copy<ALLI:mode>_insn): New insn.
14468         (despeculate_copyti_insn): New insn.
14469         (despeculate_simple<ALLI:mode>): New insn
14470         (despeculate_simpleti): New insn.
14471         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
14472         function.
14473         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
14474         aarch64_speculation_safe_value.
14475         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
14477 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14479         * config/aarch64/aarch64-speculation.cc: New file.
14480         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
14481         before pass_reorder_blocks.
14482         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
14483         prototype.
14484         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
14485         X14 and X15 when tracking speculation.
14486         * config/aarch64/aarch64.md (register name constants): Add
14487         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
14488         (unspec): Add UNSPEC_SPECULATION_TRACKER.
14489         (speculation_barrier): New insn attribute.
14490         (cmp<mode>): Allow SP in comparisons.
14491         (speculation_tracker): New insn.
14492         (speculation_barrier): Add speculation_barrier attribute.
14493         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
14494         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
14495         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
14497 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14499         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
14500         aarch64_track_speculation is true.
14501         (tb<optab><mode>1): Likewise.
14502         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
14503         generate CB[N]Z when tracking speculation.
14504         (aarch64_split_compare_and_swap): Likewise.
14505         (aarch64_split_atomic_op): Likewise.
14507 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14509         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
14511 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14513         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
14514         (speculation_barrier): New insn.
14516 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14518         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
14519         * config/arm/arm.md (speculation_barrier): New expand.
14520         (speculation_barrier_insn): New pattern.
14522 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
14524         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
14525         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
14526         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
14527         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
14528         list.
14529         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
14530         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
14531         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
14532         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
14533         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
14534         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
14535         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
14536         * builtins.c (expand_speculation_safe_value): New function.
14537         (expand_builtin): Call it.
14538         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
14539         * doc/extend.texi: Document __builtin_speculation_safe_value.
14540         * doc/md.texi: Document "speculation_barrier" pattern.
14541         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
14542         TARGET_HAVE_SPECULATION_SAFE_VALUE.
14543         * doc/tm.texi: Regenerated.
14544         * target.def (have_speculation_safe_value, speculation_safe_value): New
14545         hooks.
14546         * targhooks.c (default_have_speculation_safe_value): New function.
14547         (default_speculation_safe_value): New function.
14548         * targhooks.h (default_have_speculation_safe_value): Add prototype.
14549         (default_speculation_safe_value): Add prototype.
14551 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
14553         * dump-context.h (dump_context::dump_loc): New decl.
14554         * dumpfile.c (dump_context::dump_loc): New member function.
14555         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
14556         and dump_gimple_stmt.
14557         (dump_context::dump_gimple_expr_loc): Likewise, using
14558         dump_gimple_expr.
14559         (dump_context::dump_generic_expr_loc): Likewise, using
14560         dump_generic_expr.
14561         (dump_context::dump_printf_loc_va): Likewise, using
14562         dump_printf_va.
14563         (dump_context::begin_scope): Explicitly using the global function
14564         "dump_loc", rather than the member function.
14566 2018-07-31  Martin Sebor  <msebor@redhat.com>
14568         PR tree-optimization/86741
14569         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
14571 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
14573         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
14574         depend on whether prefetch instructions will be emitted or not.
14575         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
14576         will be emitted or not.
14577         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
14578         (TARGET_SETMEM_PFD): New macros.
14580 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14582         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
14583         (NULL_STMT_VEC_INFO): Delete.
14584         (stmt_vec_info::operator*): Likewise.
14585         (stmt_vec_info::operator gimple *): Likewise.
14586         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
14587         of NULL_STMT_VEC_INFO.
14588         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14589         (vect_reassociating_reduction_p): Likewise.
14590         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
14591         (vectorizable_store): Likewise.
14592         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
14593         (vec_info::free_stmt_vec_infos): Likewise.
14595 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14597         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
14598         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
14600 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14602         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
14603         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14604         (vec_info::free_stmt_vec_info): New private member functions.
14605         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
14606         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
14607         * tree-parloops.c (gather_scalar_reductions): Remove calls to
14608         set_stmt_vec_info_vec and free_stmt_vec_infos.
14609         * tree-vect-loop.c (_loop_vec_info): Remove call to
14610         set_stmt_vec_info_vec.
14611         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
14612         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
14613         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
14614         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14615         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
14616         assignments in {vec_info::,}new_stmt_vec_info that are redundant
14617         with the clearing in the xcalloc.
14618         (stmt_vec_info_vec): Delete.
14619         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
14620         (vectorize_loops): Likewise.
14621         (vec_info::~vec_info): Remove argument from call to
14622         free_stmt_vec_infos.
14623         (vec_info::add_stmt): Remove vinfo argument from call to
14624         new_stmt_vec_info.
14626 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14628         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
14629         rather than a gimple stmt.
14630         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
14631         information for pattern statements when passed the original
14632         statement; instead wait to be passed the pattern statement itself.
14633         Don't call set_vinfo_for_stmt here.
14634         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
14635         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
14636         stmt_vec_infos here.
14637         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
14638         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
14639         stmt_vec_infos entry.
14641 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14643         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
14644         * tree-vectorizer.c (vec_info::replace_stmt): New function.
14645         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
14646         * tree-vect-stmts.c (vectorizable_call): Likewise.
14647         (vectorizable_simd_clone_call): Likewise.
14649 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14651         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
14652         * tree-vectorizer.c (vec_info::remove_stmt): New function.
14653         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
14654         * tree-vect-loop.c (vect_transform_loop): Likewise.
14655         * tree-vect-slp.c (vect_schedule_slp): Likewise.
14656         * tree-vect-stmts.c (vect_remove_stores): Likewise.
14658 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14660         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
14661         (vect_dr_stmt): Delete.
14662         * tree-vectorizer.c (vec_info::lookup_dr): New function.
14663         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
14664         of DR_VECT_AUX.
14665         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14666         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
14667         (vect_verify_datarefs_alignment, vect_peeling_supportable)
14668         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14669         (vect_analyze_data_refs): Likewise.
14670         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
14671         argument.
14672         (vect_find_same_alignment_drs): Likewise.
14673         (vect_slp_analyze_node_dependences): Update calls accordingly.
14674         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
14675         instead of DR_VECT_AUX.
14676         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
14677         of a vector data references.  Use vec_info::lookup_dr instead of
14678         DR_VECT_AUX.
14679         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
14680         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
14681         instead of DR_VECT_AUX.
14683 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14685         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
14686         dr_vec_info.
14687         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14688         accordingly.
14689         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14690         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14691         (vect_gen_prolog_loop_niters): Likewise.
14693 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14695         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
14696         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
14697         (vect_known_alignment_in_bytes, vect_dr_behavior)
14698         (vect_get_scalar_dr_size): Take references as dr_vec_infos
14699         instead of data_references.  Update calls to other routines for
14700         which the same change has been made.
14701         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
14702         dr_vec_infos instead of stmt_vec_infos.
14703         (vect_analyze_data_ref_dependence): Update call accordingly.
14704         (vect_slp_analyze_data_ref_dependence)
14705         (vect_record_base_alignments): Use DR_VECT_AUX.
14706         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
14707         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
14708         (vector_alignment_reachable_p, vect_get_data_access_cost)
14709         (vect_peeling_supportable, vect_analyze_group_access_1)
14710         (vect_analyze_group_access, vect_analyze_data_ref_access)
14711         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
14712         (vect_compile_time_alias, vect_small_gap_p)
14713         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
14714         (vect_supportable_dr_alignment): Take references as dr_vec_infos
14715         instead of data_references.  Update calls to other routines for
14716         which the same change has been made.
14717         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
14718         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14719         (vect_slp_analyze_and_verify_node_alignment)
14720         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14721         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14722         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
14723         above changes.
14724         (_vect_peel_info::dr): Replace with...
14725         (_vect_peel_info::dr_info): ...this new field.
14726         (vect_peeling_hash_get_most_frequent)
14727         (vect_peeling_hash_choose_best_peeling): Update accordingly.
14728         (vect_peeling_hash_get_lowest_cost):
14729         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
14730         routines for which the same change has been made.
14731         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
14732         data_reference.
14733         * tree-vect-loop-manip.c (get_misalign_in_elems)
14734         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
14735         above changes.
14736         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14737         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
14738         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
14739         (get_group_load_store_type, get_negative_load_store_type)
14740         (vect_get_data_ptr_increment, vectorizable_store)
14741         (vectorizable_load): Likewise.
14742         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
14743         Update calls to other routines for which the same change has been made.
14745 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14747         * tree-vectorizer.h (vec_info::move_dr): New member function.
14748         (dataref_aux): Rename to...
14749         (dr_vec_info): ...this and add "dr" and "stmt" fields.
14750         (_stmt_vec_info::dr_aux): Update accordingly.
14751         (_stmt_vec_info::data_ref_info): Delete.
14752         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
14753         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
14754         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
14755         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
14756         of data_ref.
14757         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
14758         (STMT_VINFO_DR_INFO): New macro.
14759         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
14760         (set_dr_misalignment): Update after rename of dataref_aux.
14761         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
14762         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
14763         initialization of STMT_VINFO_DATA_REF.
14764         * tree-vectorizer.c (vec_info::move_dr): New function.
14765         * tree-vect-patterns.c (vect_recog_bool_pattern)
14766         (vect_recog_mask_conversion_pattern)
14767         (vect_recog_gather_scatter_pattern): Use it.
14768         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
14769         the "dr" and "stmt" fields of dr_vec_info instead of
14770         STMT_VINFO_DATA_REF.
14772 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14774         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
14775         (is_pattern_stmt_p): Use it.
14776         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
14777         on pattern statements.
14779 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14781         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
14782         original stmt as a stmt_vec_info rather than a gimple stmt.
14783         (vect_pattern_recog_1): Take the statement directly as a
14784         stmt_vec_info, rather than via a gimple_stmt_iterator.
14785         Update call to vect_mark_pattern_stmts.
14786         (vect_pattern_recog): Update calls accordingly.
14788 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14790         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
14791         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
14792         a vect_def_type for the first argument.
14793         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
14794         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
14795         operand if it isn't defined by a vectorized statement.
14796         (vect_build_gather_load_calls): Remove the mask_dt argument and
14797         update calls to vect_get_vec_def_for_stmt_copy.
14798         (vectorizable_bswap): Likewise the dt argument.
14799         (vectorizable_call): Update calls to vectorizable_bswap and
14800         vect_get_vec_def_for_stmt_copy.
14801         (vectorizable_simd_clone_call, vectorizable_assignment)
14802         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
14803         (vectorizable_comparison): Update calls to
14804         vect_get_vec_def_for_stmt_copy.
14805         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
14806         vect_is_simple_use.
14807         (vect_get_loop_based_defs): Remove dt argument and update call
14808         to vect_get_vec_def_for_stmt_copy.
14809         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
14810         and vect_get_vec_def_for_stmt_copy.
14811         (vectorizable_load): Update calls to vect_build_gather_load_calls
14812         and vect_get_vec_def_for_stmt_copy.
14813         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14814         (vectorizable_reduction, vectorizable_live_operation): Update calls
14815         to vect_get_vec_def_for_stmt_copy.
14817 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14819         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
14820         and gimple stmt arguments with a stmt_vec_info.
14821         (vect_record_base_alignments): Update calls accordingly.
14822         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
14823         and gimple stmt arguments with a stmt_vec_info.
14824         (vect_build_slp_tree_1): Remove vinfo argument and update call
14825         to vect_record_max_nunits.
14826         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
14827         and vect_record_max_nunits.
14829 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14831         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
14832         file and take a stmt_vec_info instead of a gimple stmt.
14833         (supportable_widening_operation, vect_finish_replace_stmt)
14834         (vect_finish_stmt_generation, vect_get_store_rhs)
14835         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
14836         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
14837         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
14838         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
14839         (vect_create_data_ref_ptr, bump_vector_ptr)
14840         (vect_permute_store_chain, vect_setup_realignment)
14841         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
14842         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
14843         (vectorizable_reduction, vectorizable_induction)
14844         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
14845         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
14846         than gimple stmts as arguments.
14847         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
14848         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
14849         (can_group_stmts_p, vect_check_gather_scatter)
14850         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14851         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
14852         (vect_permute_load_chain, vect_shift_permute_load_chain)
14853         (vect_transform_grouped_load)
14854         (vect_record_grouped_load_vectors): Likewise.
14855         * tree-vect-loop.c (vect_fixup_reduc_chain)
14856         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
14857         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
14858         (vectorizable_reduction, vectorizable_induction)
14859         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
14860         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
14861         (vect_get_load_store_mask): Likewise.
14862         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
14863         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
14864         * tree-vect-stmts.c (vect_mark_relevant)
14865         (is_simple_and_all_uses_invariant)
14866         (exist_non_indexing_operands_for_use_p, process_use)
14867         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
14868         (vect_get_vec_def_for_operand, vect_get_vec_defs)
14869         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
14870         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
14871         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
14872         (get_negative_load_store_type, get_load_store_type)
14873         (vect_check_load_store_mask, vect_check_store_rhs)
14874         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
14875         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14876         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14877         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14878         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
14879         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
14880         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
14881         (supportable_widening_operation): Likewise.
14883 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14885         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
14886         a stmt_vec_info instead of a gcall.
14887         (vect_check_gather_scatter): Update call accordingly.
14888         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
14889         of a gphi.
14890         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
14891         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
14892         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
14893         instead of a gimple stmt.
14894         (vect_transform_loop): Update calls accordingly.
14895         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
14896         stmt_vec_infos instead of gimple stmts.
14897         (vect_analyze_slp_instance): Update use accordingly.
14898         * tree-vect-stmts.c (read_vector_array, write_vector_array)
14899         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
14900         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
14901         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
14902         (vect_gen_widened_results_half, vect_get_loop_based_defs)
14903         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
14904         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
14905         down to subroutines.
14907 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14909         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
14910         of the worklist from a vector of gimple stmts to a vector of
14911         stmt_vec_infos.
14912         * tree-vect-stmts.c (vect_mark_relevant, process_use)
14913         (vect_mark_stmts_to_be_vectorized): Likewise
14915 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14917         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
14918         statement before passing it to vect_analyze_stmt.
14919         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
14920         the chain of phi vector definitions.  Track the exit phi via its
14921         stmt_vec_info.
14922         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
14923         STMT_VINFO_REDUC_DEF.
14924         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
14925         stmt_vec_infos to handle the statement chains.
14926         (vect_get_slp_defs): Record the first statement in the node
14927         using a stmt_vec_info.
14928         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
14929         statements here and pass their stmt_vec_info down to subroutines.
14930         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
14931         down to vect_finish_stmt_generation.
14932         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
14933         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
14934         stmt_vec_infos to subroutines.
14935         (vect_remove_stores): Use stmt_vec_infos to handle the statement
14936         chains.
14938 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14940         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
14941         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
14942         (vect_permute_store_chain, vect_setup_realignment)
14943         (vect_permute_load_chain, vect_shift_permute_load_chain)
14944         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
14945         stmts internally, and when passing values to other vectorizer routines.
14946         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14947         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
14948         (vect_analyze_loop_operations, get_initial_def_for_reduction)
14949         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
14950         (vectorizable_reduction, vectorizable_induction)
14951         (vectorizable_live_operation, vect_transform_loop_stmt)
14952         (vect_transform_loop): Likewise.
14953         * tree-vect-patterns.c (vect_reassociating_reduction_p)
14954         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
14955         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
14956         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14957         (vect_slp_analyze_node_operations_1): Likewise.
14958         * tree-vect-stmts.c (vect_mark_relevant, process_use)
14959         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
14960         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
14961         (vect_finish_stmt_generation_1, get_group_load_store_type)
14962         (get_load_store_type, vect_build_gather_load_calls)
14963         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14964         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14965         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14966         (vectorizable_store, vectorizable_load, vectorizable_condition)
14967         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
14968         (supportable_widening_operation): Likewise.
14969         (vect_get_vector_types_for_stmt): Likewise.
14970         * tree-vectorizer.h (vect_dr_behavior): Likewise.
14972 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14974         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14975         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
14976         (vect_permute_store_chain, vect_permute_load_chain)
14977         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
14978         repeated stmt_vec_info lookups.
14979         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14980         (vect_update_ivs_after_vectorizer): Likewise.
14981         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
14982         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14983         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
14984         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14985         (vect_bb_slp_scalar_cost): Likewise.
14986         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
14988 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14990         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
14991         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
14992         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
14993         to get gassigns and gcalls, rather than operating on generc gimple
14994         stmts.
14995         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
14996         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
14997         (vectorizable_load, vect_analyze_stmt): Likewise.
14998         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
15000 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15002         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
15003         return stmt_vec_infos rather than gimple stmts.  Do not accept
15004         null arguments.
15005         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
15006         of a gimple stmt.
15007         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
15008         Update use of get_later_stmt.
15009         (vect_get_constant_vectors): Update call accordingly.
15010         (vect_schedule_slp_instance): Likewise
15011         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
15012         (vect_slp_analyze_instance_dependence): Likewise.
15013         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
15015 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15017         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
15018         (stmt_info_for_cost::stmt_info): ...this new field.
15019         (add_stmt_costs): Update accordingly.
15020         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
15021         (vect_get_known_peeling_cost): Likewise.
15022         (vect_estimate_min_profitable_iters): Likewise.
15023         * tree-vect-stmts.c (record_stmt_cost): Likewise.
15025 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15027         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
15028         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
15029         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
15030         accordingly.
15031         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
15033 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15035         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
15036         a gimple stmt to a stmt_vec_info.
15037         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
15039 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15041         * tree-vectorizer.h (vec_info::grouped_stores): Change from
15042         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
15043         (_loop_vec_info::reduction_chains): Likewise.
15044         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
15045         accordingly.
15046         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15048 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15050         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
15051         a gimple stmt to a stmt_vec_info.
15052         (_stmt_vec_info::next_element): Likewise.
15053         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
15054         (vect_slp_analyze_and_verify_node_alignment)
15055         (vect_analyze_group_access_1, vect_analyze_group_access)
15056         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
15057         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
15058         (vect_supportable_dr_alignment): Update accordingly.
15059         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
15060         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
15061         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
15062         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15063         * tree-vect-slp.c (vect_build_slp_tree_1)
15064         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
15065         (vect_split_slp_store_group, vect_analyze_slp_instance)
15066         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
15067         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
15068         (get_group_load_store_type, get_load_store_type)
15069         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
15070         (vect_transform_stmt, vect_remove_stores): Likewise.
15072 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15074         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
15075         than a gimple stmt.
15076         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15077         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
15078         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
15079         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
15080         (vector_alignment_reachable_p, vect_get_data_access_cost)
15081         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
15082         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
15083         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
15084         (vect_analyze_group_access_1, vect_analyze_group_access)
15085         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
15086         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
15087         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
15088         result of vect_dr_stmt and use the stmt_vec_info instead of
15089         the associated gimple stmt.
15090         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15091         (vect_gen_prolog_loop_niters): Likewise.
15092         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15094 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15096         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
15097         to a vec<stmt_vec_info>.
15098         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
15099         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
15100         vec<stmt_vec_info>.
15101         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
15102         to a vec<stmt_vec_info>.
15103         (bst_traits::value_type, bst_traits::value_type): Likewise.
15104         (bst_traits::hash): Update accordingly.
15105         (vect_get_and_check_slp_defs): Change the stmts parameter from
15106         a vec<gimple *> to a vec<stmt_vec_info>.
15107         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
15108         (vect_build_slp_tree): Likewise.
15109         (vect_build_slp_tree_2): Likewise.  Update uses of
15110         SLP_TREE_SCALAR_STMTS.
15111         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
15112         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
15113         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
15114         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
15115         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
15116         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
15117         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
15118         (vect_get_constant_vectors, vect_get_slp_defs)
15119         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
15120         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
15121         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
15122         instead of gimple stmts.
15123         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
15124         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
15125         (vect_slp_analyze_instance_dependence): Update uses of
15126         SLP_TREE_SCALAR_STMTS.
15127         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15128         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
15129         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
15130         (get_initial_defs_for_reduction): Likewise.
15131         (vect_create_epilog_for_reduction): Likewise.
15132         (vectorize_fold_left_reduction): Likewise.
15133         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
15134         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
15135         (can_vectorize_live_stmts): Likewise.
15137 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15139         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
15140         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
15141         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
15142         than gimple stmts.
15143         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
15144         of a gimple stmt.
15145         (gather_scalar_reductions): Update after above interface changes.
15146         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
15147         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
15148         than gimple stmts.
15149         (vect_force_simple_reduction): Likewise.
15150         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
15151         LOOP_VINFO_REDUCTIONS.
15152         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15154 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15156         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
15157         a gimple stmt to a stmt_vec_info.
15158         * tree-vect-loop.c (vect_active_double_reduction_p)
15159         (vect_force_simple_reduction, vectorizable_reduction): Update
15160         accordingly.
15162 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15164         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
15165         vec<gimple *> to a vec<stmt_vec_info>.
15166         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
15167         the reduction_phis argument from a vec<gimple *> to a
15168         vec<stmt_vec_info>.
15169         (vectorizable_reduction): Likewise the phis local variable that
15170         is passed to vect_create_epilog_for_reduction.  Update for new type
15171         of SLP_TREE_VEC_STMTS.
15172         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
15173         (vectorizable_live_operation): Likewise.
15174         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
15175         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
15177 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15179         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
15180         a gimple stmt to a stmt_vec_info.
15181         (vectorizable_condition, vectorizable_live_operation)
15182         (vectorizable_reduction, vectorizable_induction): Pass back the
15183         vectorized statement as a stmt_vec_info.
15184         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
15185         use of STMT_VINFO_VEC_STMT.
15186         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
15187         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
15188         as stmt_vec_infos rather than gimple stmts.
15189         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
15190         to a stmt_vec_info.
15191         (vectorizable_live_operation): Likewise.
15192         (vectorizable_reduction, vectorizable_induction): Likewise,
15193         updating use of STMT_VINFO_VEC_STMT.
15194         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
15195         of STMT_VINFO_VEC_STMT.
15196         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
15197         (vectorizable_simd_clone_call, vectorizable_conversion)
15198         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
15199         (vectorizable_store, vectorizable_load, vectorizable_condition)
15200         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
15201         from a gimple stmt to a stmt_vec_info.
15202         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
15203         pointer to a stmt_vec_info to the vectorizable_* routines.
15205 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15207         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
15208         a gimple stmt to a stmt_vec_info.
15209         (is_pattern_stmt_p): Update accordingly.
15210         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
15211         (vect_record_grouped_load_vectors): Likewise.
15212         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
15213         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
15214         (vect_model_reduction_cost): Likewise.
15215         (vect_create_epilog_for_reduction): Likewise.
15216         (vectorizable_reduction, vectorizable_induction): Likewise.
15217         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15218         Return the stmt_vec_info for the pattern statement.
15219         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
15220         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
15221         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
15222         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
15223         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
15224         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
15225         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
15226         (free_stmt_vec_info, vect_is_simple_use): Likewise.
15228 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15230         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
15231         (vect_finish_stmt_generation): Likewise.
15232         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
15233         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
15234         (vect_build_gather_load_calls): Use the return value of the above
15235         functions instead of a separate call to vinfo_for_stmt.  Use narrow
15236         scopes for the input gimple stmt and wider scopes for the associated
15237         stmt_vec_info.  Use vec_info::lookup_def when setting these
15238         stmt_vec_infos from an SSA_NAME definition.
15239         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
15240         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
15241         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
15242         (vectorizable_store, vectorizable_load, vectorizable_condition)
15243         (vectorizable_comparison): Likewise.
15244         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
15245         (vectorizable_reduction): Likewise.
15247 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15249         * tree-vectorizer.h (vect_is_simple_use): Add an optional
15250         stmt_vec_info * parameter before the optional gimple **.
15251         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
15252         (process_use, vect_get_vec_def_for_operand_1): Update callers.
15253         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
15254         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15255         (vectorizable_live_operation): Likewise.
15256         * tree-vect-patterns.c (type_conversion_p): Likewise.
15257         (vect_look_through_possible_promotion): Likewise.
15258         (vect_recog_rotate_pattern): Likewise.
15259         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
15261 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15263         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
15264         a typedef to a wrapper class.
15265         (NULL_STMT_VEC_INFO): New macro.
15266         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
15267         (stmt_vec_info::operator*): New function.
15268         (stmt_vec_info::operator gimple *): Likewise.
15269         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
15270         (add_stmt_costs): Likewise.
15271         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
15272         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
15273         (vect_get_known_peeling_cost): Likewise.
15274         (vect_estimate_min_profitable_iters): Likewise.
15275         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15276         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
15277         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
15278         (vectorizable_store, free_stmt_vec_infos): Likewise.
15279         (new_stmt_vec_info): Change return type of xcalloc to
15280         _stmt_vec_info *.
15282 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15284         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
15285         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
15286         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
15287         a single_imm_use-based sequence.
15288         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
15290 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15292         * tree-vectorizer.h (vec_info::lookup_def): Declare.
15293         * tree-vectorizer.c (vec_info::lookup_def): New function.
15294         * tree-vect-patterns.c (vect_get_internal_def): Use it.
15295         (vect_widened_op_tree): Likewise.
15296         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
15297         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
15298         (vectorizable_reduction): Likewise.
15299         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
15300         of a gimple *.
15301         (vect_is_slp_reduction): Update calls accordingly.  Use
15302         vec_info::lookup_def.
15303         (vect_is_simple_reduction): Likewise
15304         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
15306 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15308         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
15309         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
15310         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
15311         of vinfo_for_stmt.
15312         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
15313         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
15314         (vect_update_vf_for_slp, vect_analyze_loop_operations)
15315         (vect_is_slp_reduction, vectorizable_induction)
15316         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
15317         * tree-vect-patterns.c (vect_init_pattern_stmt):
15318         (vect_determine_min_output_precision_1, vect_determine_precisions)
15319         (vect_pattern_recog): Likewise.
15320         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
15321         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
15322         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
15323         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
15324         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
15325         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
15326         info field from a loop to a loop_vec_info.
15328 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15330         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
15331         (vec_info::add_stmt): Declare.
15332         * tree-vectorizer.c (vec_info::add_stmt): New function.
15333         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
15334         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
15335         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
15336         (vectorizable_induction): Likewise.
15337         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
15338         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
15339         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
15340         (vectorizable_load): Likewise.
15341         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15342         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
15343         (vect_recog_gather_scatter_pattern): Likewise.
15344         (append_pattern_def_seq): Likewise.  Remove a check that is
15345         performed by add_stmt itself.
15347 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15349         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
15350         which make_ssa_name was called with new_stmt before new_stmt
15351         had been created.
15353 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15355         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
15356         split out from...
15357         (vect_is_slp_reduction): ...here...
15358         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
15359         that are already known to be false.
15361 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15363         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
15364         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
15365         STMT_VINFO_NUM_SLP_USES when it's true.
15366         (vect_free_slp_instance): Add a final_p parameter and pass it to
15367         vect_free_slp_tree.
15368         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
15369         (vect_analyze_slp_instance): Likewise.
15370         (vect_slp_analyze_operations): Likewise.
15371         (vect_slp_analyze_bb_1): Likewise.
15372         * tree-vectorizer.c (vec_info): Likewise.
15373         * tree-vect-loop.c (vect_transform_loop): Likewise.
15375 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15377         * tree-vect-loop.c (vectorizable_reduction): Assert that the
15378         function is not called for second and subsequent members of
15379         a reduction group.
15381 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
15383         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
15384         cases for nested loops from here to ...
15385         (vect_create_epilog_for_reduction): ...here.  Only call
15386         vect_is_simple_use for inner-loop reductions.
15388 2018-07-31  Martin Liska  <mliska@suse.cz>
15390         PR gcov-profile/85338
15391         PR gcov-profile/85350
15392         PR gcov-profile/85372
15393         * profile.c (struct location_triplet): New.
15394         (struct location_triplet_hash): Likewise.
15395         (output_location): Do not output a BB that
15396         is already recorded for a line.
15397         (branch_prob): Use streamed_locations.
15399 2018-07-31  Martin Liska  <mliska@suse.cz>
15401         PR gcov-profile/85370
15402         * coverage.c (coverage_begin_function): Do not mark target
15403         clones as artificial functions.
15405 2018-07-31  Martin Liska  <mliska@suse.cz>
15407         PR gcov-profile/83813
15408         PR gcov-profile/84758
15409         PR gcov-profile/85217
15410         PR gcov-profile/85332
15411         * profile.c (branch_prob): Do not record GOTO expressions
15412         for GIMPLE statements which locations are already streamed.
15414 2018-07-31  Olivier Hainque  <hainque@adacore.com>
15416         * gcc.c (handle_spec_function): Accept a soft_matched_part
15417         argument, as do_spec_1.  Pass it down to ...
15418         (eval_spec_function): Accept a soft_matched_part argument,
15419         and pass it down to ...
15420         (do_spec_2): Accept a soft_matched_part argument, and pass
15421         it down to do_spec_1.
15422         (do_spec_1): Pass soft_matched_part to handle_spec_function.
15423         (handle_braces): Update call to handle_spec_function.
15424         (driver::set_up_specs): Update calls to do_spec_2.
15425         (compare_debug_dump_opt_spec_function): Likewise.
15426         (compare_debug_self_opt_spec_function): Likewise.
15428 2018-07-31  Olivier Hainque  <hainque@adacore.com>
15430         * common.opt (nolibc): New option.
15431         * doc/invoke.texi (Link Options): Document it.
15432         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
15433         * config/alpha/linux.h: Likewise.
15434         * config/arc/elf.h: Likewise.
15435         * config/arm/uclinux-elf.h: Likewise.
15436         * config/arm/unknown-elf.h: Likewise.
15437         * config/avr/avrlibc.h: Likewise.
15438         * config/bfin/bfin.h: Likewise.
15439         * config/bfin/linux.h: Likewise.
15440         * config/bfin/uclinux.h: Likewise.
15441         * config/darwin.h: Likewise.
15442         * config/darwin10.h: Likewise.
15443         * config/darwin12.h: Likewise.
15444         * config/gnu-user.h: Likewise.
15445         * config/lm32/uclinux-elf.h: Likewise.
15446         * config/pa/pa-hpux11.h: Likewise.
15447         * config/pa/pa64-hpux.h: Likewise.
15448         * config/sparc/sparc.h: Likewise.
15450 2018-07-31  Olivier Hainque  <hainque@adacore.com>
15452         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
15453         undefined variables.
15455 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
15457         PR target/86640
15458         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
15459         instead of GEN_INT.
15461 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15463         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
15464         terminated string literal.
15466 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
15468         PR rtl-optimization/85160
15469         * combine.c (is_just_move): New function.
15470         (try_combine): Allow combining two instructions into two if neither of
15471         the original instructions was a move.
15473 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
15475         PR target/86673
15476         * doc/extend.texi (Global Register Variables): Discourage use of type
15477         qualifiers.
15478         (Local Register Variables): Likewise.
15480 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
15482         PR tree-optimization/86506
15483         * hwint.h (ceil_log2): Resync with hwint.c implementation.
15485 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
15487         PR target/86547
15488         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
15489         hard_regno, make sure no insn between `from` and `to` clobbers it.
15491 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
15492             Tom de Vries  <tdevries@suse.de>
15494         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
15495         (PTX_DEFAULT_RUNTIME_DIM): ... this.
15496         (nvptx_goacc_validate_dims): Set default worker and gang dims to
15497         PTX_DEFAULT_RUNTIME_DIM.
15498         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
15500 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
15502         * config/pa/pa.c (pa_output_addr_vec): Align address table.
15503         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
15504         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
15506 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
15508         * config/rs6000/constraints.md (wG constraint): Delete, no longer
15509         used.
15510         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
15511         predicate to reflect toc fusion has been deleted.
15512         (toc_fusion_mem_raw): Delete, no longer used.
15513         (toc_fusion_mem_wrapped): Likewise.
15514         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
15515         fusion mask bit.
15516         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
15517         Delete, no longer used.
15518         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
15519         meant to be used for toc fusion.
15520         (rs6000_debug_print_mode): Delete toc fusion debugging.
15521         (rs6000_debug_reg_global): Likewise.
15522         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
15523         fusion and secondary reload support that were never used.
15524         (rs6000_option_override_internal): Delete TOC fusion, that was only
15525         partially defined, and it did not work unless you also used the
15526         -mcmodel= switch.
15527         (rs6000_legitimate_address_p): Delete TOC fusion support.
15528         (rs6000_opt_masks): Likewise.
15529         (fusion_wrap_memory_address): Delete function, no longer used.
15530         (fusion_split_address); Delete TOC fusion support.
15531         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
15532         longer used with toc fusion being deleted.
15533         (TARGET_TOC_FUSION_FP): Likewise.
15534         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
15535         UNSPEC.
15536         (toc fusion spliter): Delete TOC fusion support.
15537         (toc_fusionload_<mode>): Likewise.
15538         (toc_fusionload_di): Likewise.
15539         (fusion_gpr_load_<mode>): Delete generator function, this insn no
15540         longer needs to be named.  Rename predicate to delete TOC fusion.
15541         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15542         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15543         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15544         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15545         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
15547 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15549         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
15550         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
15551         __int128 in built-in function prototypes.
15552         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
15553         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
15555 2018-07-27  Martin Sebor  <msebor@redhat.com>
15557         PR tree-optimization/86696
15558         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
15559         types, including enums.
15560         (handle_char_store): Be prepared for the above function to fail.
15562 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
15564         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
15565         when optimization level is lower than 2 or optimize for size.
15567 2018-07-26  Martin Sebor  <msebor@redhat.com>
15569         PR tree-optimization/86043
15570         PR tree-optimization/86042
15571         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
15572         (get_string_cst_length): Rename...
15573         (get_min_string_length): ...to this.  Add argument.
15574         (handle_char_store): Extend to handle multi-character stores by
15575         MEM_REF.
15576         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
15577         * tree.h (initializer_zerop): Add argument.
15579 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
15581         PR middle-end/86660
15582         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
15583         declare target to variables if they have always,{to,from,tofrom} map
15584         kinds.
15586 2018-07-26  Martin Liska  <mliska@suse.cz>
15588         PR lto/86548
15589         * lto-wrapper.c: Add linker_output as prefix
15590         for ltrans_output_file.
15592 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
15594         PR rtl-optimization/85805
15595         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
15596         value for hard registers if that was written in the same mode.
15598 2018-07-26  Martin Liska  <mliska@suse.cz>
15600         PR gcov-profile/86536
15601         * gcov.c (format_gcov): Use printf format %.*f directly
15602         and do not handle special values.
15604 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
15606         * common/config/arc/arc-common.c (arc_option_optimization_table):
15607         Update default optimizations for size.
15609 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
15611         * config/arc/arc.md (movsf_insn): Add short instruction selection.
15612         * config/arc/constraints.md (CfZ): New constraint.
15613         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
15614         (subsf3_fpu): Likewise.
15615         (cmpsf_fpu): Likewise.
15616         (cmpsf_fpu_uneq): Likewise.
15618 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
15620         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
15621         accesses as well.
15622         (arc_is_uncached_mem_p): uncached applies to both the variable and
15623         the pointer.
15625 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
15627         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
15628         register names.
15630 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
15632         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
15633         field "m_scopes" from vec to auto_vec.
15635 2018-07-25  Martin Liska  <mliska@suse.cz>
15637         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
15638         return type.
15640 2018-07-25  Richard Biener  <rguenther@suse.de>
15642         PR debug/86654
15643         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
15644         special wrt context_die late.
15645         (gen_subprogram_die): Re-use DIEs in local scope.
15647 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
15649         PR tree-optimization/86644
15650         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
15652 2018-07-25  Martin Liska  <mliska@suse.cz>
15654         PR middle-end/86645
15655         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
15656         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
15658 2018-07-25  Martin Liska  <mliska@suse.cz>
15660         PR sanitizer/79635
15661         * params.def: Explain ASan abbreviation and provide
15662         a documentation link.
15664 2018-07-24  Martin Sebor  <msebor@redhat.com>
15666         PR tree-optimization/86622
15667         PR tree-optimization/86532
15668         * builtins.h (string_length): Declare.
15669         * builtins.c (c_strlen): Correct handling of non-constant offsets.
15670         (check_access): Be prepared for non-constant length ranges.
15671         (string_length): Make extern.
15672         * expr.c (string_constant): Only handle the minor non-constant
15673         array index.  Use string_constant to compute the length of
15674         a generic string constant.
15676 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
15678         PR tree-optimization/86618
15679         * tree-vect-stmts.c (vectorizable_call): Don't take the address
15680         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
15682 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
15684         PR tree-optimization/86636
15685         * json.cc (json::object::set): Fix comment.  Add assertions.
15686         (json::array::append): Move here from json.h.  Add comment and an
15687         assertion.
15688         (json::string::string): Likewise.
15689         * json.h (json::array::append): Move to json.cc.
15690         (json::string::string): Likewise.
15691         * optinfo-emit-json.cc
15692         (optrecord_json_writer::impl_location_to_json): Assert that we
15693         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
15694         wrapper around it.  Expand the location once, rather than three
15695         times.
15696         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
15697         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
15698         wrappers.
15699         (optrecord_json_writer::optinfo_to_json): Likewise, in four
15700         places.  Fix some overlong lines.
15702 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
15704         * config/aarch64/aarch64-simd.md
15705         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
15706         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
15707         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
15708         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
15709         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
15710         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
15711         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
15712         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
15713         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
15715 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
15717         PR middle-end/86627
15718         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
15719         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
15720         and abs_d == d, do the power of two handling if profitable.
15722 2018-07-24  Richard Biener  <rguenther@suse.de>
15724         * match.pd: Add BIT_FIELD_REF canonicalizations.
15726 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15728         PR c/86617
15729         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
15731 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15733         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
15734         terminated STRING_CST object.
15736 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15738         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
15740 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
15742         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
15743         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
15744         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
15745         the elements into a register.
15746         (rs6000_split_v4si_init_di_reg): Delete.
15747         (rs6000_split_v4si_init): Delete.
15748         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
15749         (vsx_init_v4si): Rewrite as a define_expand.
15751 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
15753         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
15754         zero_extend argument from memory): New.
15756 2018-07-22  Martin Sebor  <msebor@redhat.com>
15758         PR bootstrap/86621
15759         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
15760         diagnosing calls with unknown arguments unless -Walloca-larger-than
15761         is restricted to less than PTRDIFF_MAX bytes.
15763 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
15765         * doc/gcov.texi (Invoking Gcov): Editorial changes.
15767 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
15769         * pretty-print.c (text_info::set_location): Remove redundant
15770         "line_table" parameter from call to rich_location::set_range.
15772 2018-07-20  Martin Sebor  <msebor@redhat.com>
15774         PR middle-end/82063
15775         * builtins.c (expand_builtin_alloca): Adjust.
15776         * calls.c (alloc_max_size): Simplify.
15777         * cgraphunit.c (cgraph_node::expand): Adjust.
15778         * common.opt (larger_than_size, warn_frame_larger_than): Remove
15779         variables.
15780         (frame_larger_than_size): Same.
15781         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
15782         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
15783         Initialize.
15784         * doc/invoke.texi (GCC Command Options): Document option arguments.
15785         Explain byte-size arguments and suffixes.
15786         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
15787         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
15788         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
15789         * doc/options.texi (UInteger): Expand.
15790         (Host_Wide_Int, ByteSize): Document new properties.
15791         * final.c (final_start_function_1): Include sizes in an error message.
15792         * function.c (frame_offset_overflow): Same.
15793         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
15794         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
15795         Diagnose unbounded alloca calls only for limits of less than
15796         PTRDIFF_MAX.
15797         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
15798         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
15799         for alloca(0).
15800         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
15801         only for limits of less than PTRDIFF_MAX.
15802         * langhooks-def.h (lhd_handle_option): Change function argument
15803         to HOST_WIDE_INT.
15804         * langhooks.c (lhd_handle_option): Same.
15805         * langhooks.h (handle_option): Same.
15806         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
15807         ByteSize flags.
15808         (var_type, var_type_struct): Same.
15809         (var_set): Handle ByteSize flag.
15810         * optc-gen.awk: Add comments to output to ease debugging.  Make
15811         use of HOST_WIDE_INT where appropriate.
15812         * opts-gen-save.awk:  Use %lx to format unsigned long.
15813         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
15814         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
15815         arguments.  Parse bytes-size suffixes.
15816         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
15817         (enum_value_to_arg): Same.
15818         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
15819         (handle_option): Adjust.
15820         (generate_option): Change function argument to HOST_WIDE_INT.
15821         (cmdline_handle_error): Adjust.
15822         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
15823         (set_option): Change function argument to HOST_WIDE_INT.
15824         (option_enabled): Handle cl_host_wide_int.
15825         (get_option_state): Handle CLVC_SIZE.
15826         (control_warning_option): Same.
15827         * opts.c (common_handle_option): Change function argument to
15828         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
15829         OPT_Wvla_larger_than_.
15830         * opts.h (enum cl_var_type): Add an enumerator.
15831         * stor-layout.c (layout_decl): Print a more meaningful warning.
15832         * toplev.c (output_stack_usage): Adjust.
15834 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
15836         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
15837         call to inline_expand_builtin_string_cmp.
15838         (expand_builtin_strcmp): Likewise.
15839         (expand_builtin_strncmp): Likewise.
15840         (inline_string_cmp): Delete the last parameter, change char_type_node
15841         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
15842         two operands.
15843         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
15844         the inlining expansion on target where the type of the call has same or
15845         narrower precision than unsigned char.
15847 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
15849         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
15850         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
15851         * common.opt (fsave-optimization-record): New option.
15852         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
15853         * doc/invoke.texi (-fsave-optimization-record): New option.
15854         * dumpfile.c: Include "optinfo-emit-json.h".
15855         (struct kv_pair): Move to coretypes.h.
15856         (optgroup_options): Make non-static.
15857         (dump_context::end_scope): Call
15858         optimization_records_maybe_pop_dump_scope.
15859         * dumpfile.h (optgroup_options): New decl.
15860         * json.cc: New file.
15861         * json.h: New file.
15862         * optinfo-emit-json.cc: New file.
15863         * optinfo-emit-json.h: New file.
15864         * optinfo.cc: Include "optinfo-emit-json.h".
15865         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
15866         (optinfo_enabled_p): Check optimization_records_enabled_p.
15867         (optinfo_wants_inlining_info_p): Likewise.
15868         * optinfo.h: Update comment.
15869         * profile-count.c (profile_quality_as_string): New function.
15870         * profile-count.h (profile_quality_as_string): New decl.
15871         (profile_count::quality): New accessor.
15872         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
15873         and optinfo_emit_json_cc_tests.
15874         * selftest.h (selftest::json_cc_tests): New decl.
15875         (selftest::optinfo_emit_json_cc_tests): New decl.
15876         * toplev.c: Include "optinfo-emit-json.h".
15877         (compile_file): Call optimization_records_finish.
15878         (do_compile): Call optimization_records_start.
15879         * tree-ssa-live.c: Include optinfo.h.
15880         (remove_unused_scope_block_p): Retain inlining information if
15881         optinfo_wants_inlining_info_p returns true.
15883 2018-07-20  Richard Biener  <rguenther@suse.de>
15885         PR debug/86585
15886         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
15887         to cover -flto-partition=none.
15889 2018-07-20  Martin Liska  <mliska@suse.cz>
15891         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
15892         (get_decl_source_range): Remove unused function.
15894 2018-07-20  Richard Biener  <rguenther@suse.de>
15896         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
15897         (struct vn_phi_s): Likewise.
15898         (struct vn_reference_s): Likewise.
15899         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
15900         for searching the slot of an entry known to be in the hash itself.
15901         (vn_phi_hasher::equal): Likewise.
15902         (vn_reference_hasher::equal): Likewise.
15903         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
15904         globals.
15905         (optimistic_info, current_info): Remove, keeping only valid_info.
15906         (vn_reference_lookup_1): Remove fallback lookup.
15907         (vn_reference_lookup_2): Likewise.
15908         (vn_nary_op_lookup_1): Likewise.
15909         (vn_phi_lookup): Likewise.
15910         (vn_nary_build_or_lookup_1): Make sure to not chain the built
15911         hash element.
15912         (vn_reference_insert): Adjust, chain the inserted hash element
15913         at last_inserted_ref.
15914         (vn_reference_insert_pieces): Likewise.
15915         (visit_reference_op_call): Likewise.
15916         (vn_nary_op_insert_into): Chain the inserted hash element at
15917         last_inserted_nary.
15918         (vn_nary_op_insert_pieces): Adjust.
15919         (vn_nary_op_insert): Likewise.
15920         (vn_nary_op_insert_stmt): Likewise.
15921         (vn_phi_insert): Adjust, chain the inserted hash element at
15922         last_inserted_phi.
15923         (process_scc): Remove clearing and copying the optimistic
15924         table.  Instead remove elements inserted during an optimistic
15925         iteration from the single table we maintain.
15926         (init_scc_vn): Adjust.
15927         (free_scc_vn): Likewise.
15928         (sccvn_dom_walker::record_cond): Likewise.
15929         (sccvn_dom_walker::after_dom_children): Likewise.
15931 2018-07-19  Martin Sebor  <msebor@redhat.com>
15933         PR tree-optimization/84047
15934         PR tree-optimization/83776
15935         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
15936         (check_array_bounds): Call it.
15938 2018-07-19  Martin Sebor  <msebor@redhat.com>
15940         * align.h (align_flags): Use member initialization.
15942 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
15944         * Makefile.in (OBJS): Add optinfo.o.
15945         * coretypes.h (class symtab_node): New forward decl.
15946         (struct cgraph_node): New forward decl.
15947         (class varpool_node): New forward decl.
15948         * dump-context.h: New file.
15949         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
15950         "tree-pass.h".
15951         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
15952         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
15953         (set_alt_dump_file): Likewise.
15954         (dump_context::~dump_context): New dtor.
15955         (dump_gimple_stmt): Move implementation to...
15956         (dump_context::dump_gimple_stmt): ...this new member function.
15957         Add the stmt to any pending optinfo, creating one if need be.
15958         (dump_gimple_stmt_loc): Move implementation to...
15959         (dump_context::dump_gimple_stmt_loc): ...this new member function.
15960         Start a new optinfo and add the stmt to it.
15961         (dump_gimple_expr): Move implementation to...
15962         (dump_context::dump_gimple_expr): ...this new member function.
15963         Add the stmt to any pending optinfo, creating one if need be.
15964         (dump_gimple_expr_loc): Move implementation to...
15965         (dump_context::dump_gimple_expr_loc): ...this new member function.
15966         Start a new optinfo and add the stmt to it.
15967         (dump_generic_expr): Move implementation to...
15968         (dump_context::dump_generic_expr): ...this new member function.
15969         Add the tree to any pending optinfo, creating one if need be.
15970         (dump_generic_expr_loc): Move implementation to...
15971         (dump_context::dump_generic_expr_loc): ...this new member
15972         function.  Add the tree to any pending optinfo, creating one if
15973         need be.
15974         (dump_printf): Move implementation to...
15975         (dump_context::dump_printf_va): ...this new member function.  Add
15976         the text to any pending optinfo, creating one if need be.
15977         (dump_printf_loc): Move implementation to...
15978         (dump_context::dump_printf_loc_va): ...this new member function.
15979         Start a new optinfo and add the stmt to it.
15980         (dump_dec): Move implementation to...
15981         (dump_context::dump_dec): ...this new member function.  Add the
15982         value to any pending optinfo, creating one if need be.
15983         (dump_context::dump_symtab_node): New member function.
15984         (dump_context::get_scope_depth): New member function.
15985         (dump_context::begin_scope): New member function.
15986         (dump_context::end_scope): New member function.
15987         (dump_context::ensure_pending_optinfo): New member function.
15988         (dump_context::begin_next_optinfo): New member function.
15989         (dump_context::end_any_optinfo): New member function.
15990         (dump_context::s_current): New global.
15991         (dump_context::s_default): New global.
15992         (dump_scope_depth): Delete global.
15993         (dumpfile_ensure_any_optinfo_are_flushed): New function.
15994         (dump_symtab_node): New function.
15995         (get_dump_scope_depth): Reimplement in terms of dump_context.
15996         (dump_begin_scope): Likewise.
15997         (dump_end_scope): Likewise.
15998         (selftest::temp_dump_context::temp_dump_context): New ctor.
15999         (selftest::temp_dump_context::~temp_dump_context): New dtor.
16000         (selftest::verify_item): New function.
16001         (ASSERT_IS_TEXT): New macro.
16002         (ASSERT_IS_TREE): New macro.
16003         (ASSERT_IS_GIMPLE): New macro.
16004         (selftest::test_capture_of_dump_calls): New test.
16005         (selftest::dumpfile_c_tests): Call it.
16006         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
16007         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
16008         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
16009         descriptive comment.
16010         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
16011         (dump_node, dump_bb): Move these unrelated decls.
16012         (class dump_manager): Add leading comment.
16013         * optinfo.cc: New file.
16014         * optinfo.h: New file.
16016 2018-07-19  Michael Collison  <michael.collison@arm.com>
16017             Richard Henderson  <rth@redhat.com>
16019         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
16020         (subti): Handle op1 zero.
16021         (subvti4, usub4ti4): New.
16022         (*sub<GPI>3_compare1_imm): New.
16023         (sub<GPI>3_carryinCV): New.
16024         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
16025         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
16027 2018-07-19  Michael Collison  <michael.collison@arm.com>
16028             Richard Henderson  <rth@redhat.com>
16030         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
16031         (addti3): Create simpler code if low part is already known to be 0.
16032         (addvti4, uaddvti4): New.
16033         (*add<GPI>3_compareC_cconly_imm): New.
16034         (*add<GPI>3_compareC_cconly): New.
16035         (*add<GPI>3_compareC_imm): New.
16036         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
16037         handle constants within this pattern..
16038         (*add<GPI>3_compareV_cconly_imm): New.
16039         (*add<GPI>3_compareV_cconly): New.
16040         (*add<GPI>3_compareV_imm): New.
16041         (add<GPI>3_compareV): New.
16042         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
16043         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
16044         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
16045         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
16046         with 'comparison' operator.
16047         (*add<GPI>3_compareV_cconly_imm): Ditto.
16048         (*add<GPI>3_compareV_cconly): Ditto.
16049         (*add<GPI>3_compareV_imm): Ditto.
16050         (add<GPI>3_compareV): Ditto.
16051         (add<mode>3_carryinC): Ditto.
16052         (*add<mode>3_carryinC_zero): Ditto.
16053         (*add<mode>3_carryinC): Ditto.
16054         (add<mode>3_carryinV): Ditto.
16055         (*add<mode>3_carryinV_zero): Ditto.
16056         (*add<mode>3_carryinV): Ditto.
16058 2018-07-19  Michael Collison  <michael.collison@arm.com>
16059             Richard Henderson  <rth@redhat.com>
16061         * config/aarch64/aarch64-modes.def (CC_V): New.
16062         * config/aarch64/aarch64-protos.h
16063         (aarch64_addti_scratch_regs): Declare
16064         (aarch64_subvti_scratch_regs): Declare.
16065         (aarch64_expand_subvti): Declare.
16066         (aarch64_gen_unlikely_cbranch): Declare
16067         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
16068         for signed overflow using CC_Vmode.
16069         (aarch64_get_condition_code_1): Handle CC_Vmode.
16070         (aarch64_gen_unlikely_cbranch): New function.
16071         (aarch64_addti_scratch_regs): New function.
16072         (aarch64_subvti_scratch_regs): New function.
16073         (aarch64_expand_subvti): New function.
16075 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16077         * config/aarch64/aarch64-option-extensions.def: New entry for profile
16078         extension.
16079         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
16080         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
16081         extension.
16083 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16085         PR target/83009
16086         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16087         address check not strict.
16089 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16091         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
16092         Umq with Umn.
16093         (store_pair_lanes<mode>): Likewise.
16094         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
16095         enum value 'ADDR_QUERY_LDP_STP_N'.
16096         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
16097         (aarch64_print_address_internal): Add declaration.
16098         (aarch64_print_ldpstp_address): Remove.
16099         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
16100         (aarch64_print_operand): Change printing of 'y'.
16101         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
16102         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
16103         'true' rather than '1'.
16104         * config/aarch64/constraints.md (Uml): Likewise.
16105         (Uml): Rename to Umn.
16106         (Umq): Remove.
16108 2018-07-19  Richard Biener  <rguenther@suse.de>
16110         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
16111         a trailing array.
16112         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
16113         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
16114         (vn_reference_hasher): Likewise.
16115         (struct vn_tables_s): Remove obstack and alloc-pool members.
16116         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
16117         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
16118         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
16119         (vn_reference_insert_pieces): Likewise.
16120         (alloc_vn_nary_op_noinit): Adjust.
16121         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
16122         (vn_phi_eq): Adjust.
16123         (shared_lookup_phiargs): Remove.
16124         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
16125         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
16126         (visit_reference_op_call): Likewise.
16127         (copy_nary, copy_phi, copy_reference): Remove.
16128         (process_scc): Rewind the obstack when iterating.  Do not
16129         copy the elements to valid_info but just move them from one
16130         hashtable to the other.
16131         (allocate_vn_table): Adjust.
16132         (free_vn_table): Likewise.
16133         (init_scc_vn): Likewise.
16134         (free_scc_vn): Likewise.
16136 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
16138         PR target/86560
16139         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
16140         indirect_return as function type attribute.
16141         (ix86_attribute_table): Change indirect_return to function
16142         type attribute.
16143         * doc/extend.texi: Update indirect_return attribute.
16145 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
16147         * wide-int.h (widest2_int): New.
16148         * gimple-fold.c (arith_overflowed_p): Use it.
16149         * tree.h (widest2_int_cst): New.
16150         * tree-vrp.c (wide_int_binop_overflow): Rename from
16151         vrp_int_const_binop.
16152         Rewrite to work on trees.
16153         (extract_range_from_multiplicative_op_1): Abstract code to...
16154         (wide_int_range_min_max): ...here.
16155         (wide_int_range_cross_product): ...and here.
16156         (extract_range_from_binary_expr_1): Abstract overflow code to...
16157         (wide_int_range_mult_wrapping): ...here.
16158         * tree-vrp.h (wide_int_range_cross_product): New.
16159         (wide_int_range_mult_wrapping): New.
16161 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
16162             Julia Koval  <julia.koval@intel.com>
16164         * config/i386/x86-tune-costs.h (skylake_memcpy,
16165         skylake_memset): Replace rep_prefix with unrolling for size 512.
16167 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
16169         PR middle-end/86544
16170         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
16171         comparision with EQ_EXPR in last stmt.
16173 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16175         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
16176         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
16177         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
16178         previously known as "PowerPC AltiVec Built-in Functions".  Move
16179         some material to new subsubsections "PowerPC AltiVec Built-in
16180         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
16181         ISA 2.07".
16182         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
16183         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
16184         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
16185         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
16187 2018-07-18  Richard Biener  <rguenther@suse.de>
16189         PR tree-optimization/86557
16190         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
16191         EXACT_DIV_EXPR.
16193 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
16195         * config/s390/s390.c (s390_function_profiler): Generate CFI.
16197 2018-07-17  Jeff Law  <law@redhat.com>
16199         * config/arm/arm.c (get_label_padding): Update for recent
16200         changes to label_to_alignment.
16202         PR tree-optimization/86010
16203         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
16205         * config/mips/mips.c (vr4130_align_insns): Update for recent
16206         changes to label_to_alignment.
16208         * config/frv/frv.c (frv_label_align): Update for recent changes
16209         to label_to_alignment.
16211         * config/nios2/nios2.c (nios2_label_align): Update for recent
16212         changes which dropped ALIGN_LABELS_LOG.
16214 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
16216         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
16217         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
16219 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
16221         * config/arc/arc.c (arc_label_align): Use align_labels instead of
16222         deprecated align_labels_log.
16224 2018-07-17  Richard Biener  <rguenther@suse.de>
16226         PR lto/86456
16227         * dwarf2out.c (init_sections_and_labels): Always generate
16228         a debug_line_str_section for early LTO debug.
16229         (dwarf2out_finish): Reset debug_line_str_hash output early.
16230         Bump counter for extra dwarf5 .debug_loc labels to not conflict
16231         with fat LTO part.
16232         (dwarf2out_early_finish): Output debug_line_str.
16234 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
16236         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
16237         index register on z196 or later.
16239 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
16241         * config/s390/s390.c (s390_default_align): Set default function
16242         alignment to 16.
16243         (s390_override_options_after_change): Call s390_default align.
16244         (s390_option_override_internal): Call s390_default_align.
16245         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
16247 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
16249         PR middle-end/86542
16250         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
16251         to _looptemp_ clauses, other than the first two.
16253 2018-07-17  Martin Liska  <mliska@suse.cz>
16255         * opts.c: Do not enable OPT_falign_* for -Os.
16257 2018-07-17  Martin Liska  <mliska@suse.cz>
16259         * align.h (MAX_CODE_ALIGN): New.
16260         (MAX_CODE_ALIGN_VALUE): New.
16261         * common/config/i386/i386-common.c (ix86_handle_option):
16262         (MAX_CODE_ALIGN): Moved to align.h.
16263         * final.c (MAX_CODE_ALIGN): Likewise.
16264         * opts.c (parse_and_check_align_values):
16265         (MAX_CODE_ALIGN): Likewise.
16266         (MAX_CODE_ALIGN_VALUE): Likewise.
16268 2018-07-17  Martin Liska  <mliska@suse.cz>
16270         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
16271         in order to fulfil coding style.
16272         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
16273         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16274         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
16275         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
16276         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
16277         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
16278         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
16279         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
16280         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16282 2018-07-17  Martin Liska  <mliska@suse.cz>
16284         * align.h: New file.
16285         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
16286         directly.
16287         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
16288         align_flags of label_to_alignment.
16289         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
16290         align_flags class.
16291         * config/m68k/m68k.c: Do not use removed align_labels_value and
16292         align_loops_value.
16293         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
16294         (LOOP_ALIGN): Likewise.
16295         (LABEL_ALIGN): Likewise.
16296         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
16297         Remove not used macro.
16298         (rs6000_loop_align): Change return type to align_flags.
16299         (rs6000_loop_align_max_skip): Remove.
16300         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
16301         Change return type to align_flags.
16302         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
16303         Remove not used macro.
16304         (rs6000_loop_align):  Change return type to align_flags.
16305         (rs6000_loop_align_max_skip): Remove.
16306         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
16307         * config/rx/rx-protos.h (rx_align_for_label): Make it
16308         static function.
16309         * config/rx/rx.c (rx_align_for_label): Change return type
16310         to align_flags.
16311         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
16312         macro definitions.
16313         into align_flags class.
16314         (LABEL_ALIGN): Likewise.
16315         (LOOP_ALIGN): Likewise.
16316         * config/s390/s390.c (s390_label_align): Use align_flags
16317         class member.
16318         (s390_asm_output_function_label): Likewise.
16319         * config/sh/sh.c (sh_override_options_after_change):
16320         Use align_flags class directly without macros.
16321         (find_barrier): Likewise.
16322         (barrier_align): Likewise.
16323         (sh_loop_align): Likewise.
16324         * config/spu/spu.c (spu_option_override):
16325         Use align_flags_tuple::get_value instead of removed macros.
16326         (spu_sched_init): Likewise.
16327         * config/spu/spu.h (GTY): Likewise.
16328         * config/visium/visium.c (visium_option_override):
16329         Set "8" as default secondary alignment.
16330         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
16331         in order to guarantee secondary alignment of 8.
16332         * coretypes.h: Include align.h header file.
16333         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
16334         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
16335         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
16336         * doc/tm.texi.in: Likewise.
16337         * final.c (struct label_alignment): Remove not used structure.
16338         (LABEL_ALIGN): Change type to align_flags.
16339         (LOOP_ALIGN): Likewise.
16340         (JUMP_ALIGN): Likewise.
16341         (default_loop_align_max_skip): Remove.
16342         (default_label_align_max_skip): Likewise.
16343         (default_jump_align_max_skip): Likewise.
16344         (default_label_align_after_barrier_max_skip):
16345         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
16346         (LABEL_TO_MAX_SKIP): Remove.
16347         (label_to_alignment): Return align_flags type instead of integer.
16348         (label_to_max_skip): Remove.
16349         (align_fuzz): Use align_flags type.
16350         (compute_alignments): Use align_flags type and use align_flags::max
16351         to combine multiple alignments.
16352         (grow_label_align): Grow vec instead of C array.
16353         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
16354         (shorten_branches):  Use align_flags type and use align_flags::max
16355         to combine multiple alignments.
16356         (final_scan_insn_1): Remove usage of secondary alignment that comes
16357         from label alignment, but instead use proper secondary alignment
16358         which is computed in grow_label_align.
16359         * flags.h (struct align_flags_tuple): Move to align.h.
16360         (struct align_flags): Likewise.
16361         (state_align_loops): Rename to align_loops.
16362         (state_align_jumps): Rename to align_jumps.
16363         (state_align_labels): Rename to align_labels.
16364         (state_align_functions): Rename to align_functions.
16365         (align_loops_log): Remove.
16366         (align_jumps_log): Remove.
16367         (align_labels_log): Remove.
16368         (align_functions_log): Remove.
16369         (align_loops_max_skip): Remove.
16370         (align_jumps_max_skip): Remove.
16371         (align_labels_max_skip): Remove.
16372         (align_functions_max_skip): Remove.
16373         (align_loops_value): Remove.
16374         (align_jumps_value): Remove.
16375         (align_labels_value): Remove.
16376         (align_functions_value): Remove.
16377         * output.h (label_to_alignment): Change return type to align_flags.
16378         (label_to_max_skip): Remove.
16379         * target.def: Remove loop_align_max_skip, label_align_max_skip,
16380         jump_align_max_skip macros.
16381         * targhooks.h (default_loop_align_max_skip): Remove.
16382         (default_label_align_max_skip): Likewise.
16383         (default_jump_align_max_skip): Likewise.
16384         (default_label_align_after_barrier_max_skip): Remove.
16385         * toplev.c (read_log_maxskip): Use ::normalize function.
16386         (parse_N_M): Remove not used argument and also call ::normalize.
16387         (parse_alignment_opts): Do not pass unused arguments.
16388         * varasm.c (assemble_start_function): Use directly align_functions
16389         instead of removed macros.
16390         * system.h: Do not poison removed macros.
16392 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
16394         PR middle-end/86539
16395         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
16396         and cond temporaries don't have reference type if iterator has
16397         pointer type.  For init use &for_pre_body instead of pre_p if
16398         for_pre_body is non-empty.
16400 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
16402         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
16403         double-double modes to SFmode directly directly.
16404         (trunc<mode>sf2_fprs): Delete.
16406 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
16408         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
16409         for conversions between IFmode and the decimal floating point modes.
16410         (init_float128_ieee): Use the correct names for conversions between
16411         KFmode and the decimal floating point modes.
16413 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
16415         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
16416         for the conversions between TDmode and IFmode.
16417         (init_float128_ieee): Use more correct names for the conversions
16418         between TDmode and KFmode.
16420 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
16422         PR tree-optimization/86526
16423         * builtins.c (expand_builtin_memcmp): Formatting fixes.
16424         (inline_expand_builtin_string_cmp): Likewise.
16425         (inline_string_cmp): Likewise.  Use c_readstr instead of
16426         builtin_memcpy_read_str.  Add unit_mode temporary.
16428 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16430         PR middle-end/86528
16431         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
16432         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
16434 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16436         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
16437         Alphabetize prototypes of built-in functions, separating out
16438         built-in functions that are listed in this section but should be
16439         described elsewhere.
16441 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
16443         PR target/86511
16444         * expmed.c (emit_store_flag): Do not emit setcc followed by a
16445         conditional move when trapping comparison was split to a
16446         non-trapping one (and vice versa).
16448 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
16450         * config/s390/s390.c (s390_function_profiler): Generate nops
16451         instead of profiler call sequences.
16452         * config/s390/s390.opt: Add the new option.
16454 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
16456         * config/s390/s390.c (s390_function_profiler): Generate
16457         __mcount_loc section.
16458         * config/s390/s390.opt: Add the new option.
16460 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
16462         * common.opt: Add the new warning.
16463         * config/s390/s390.c (s390_function_profiler): Emit "brasl
16464         %r0,__fentry__" when -mfentry is specified.
16465         (s390_option_override_internal): Disallow -mfentry for 31-bit
16466         CPUs.
16467         * config/s390/s390.opt: Add the new option.
16469 2018-07-16  Richard Biener  <rguenther@suse.de>
16471         PR lto/86523
16472         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
16473         for function-local FUNCTION_DECL and RESULT_DECL immediately.
16475 2018-07-16  Martin Liska  <mliska@suse.cz>
16477         PR ipa/86529
16478         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
16479         to ::get_create.
16481 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
16483         * config/arc/arcHS.md: Update ARCHS scheduling rules.
16485 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
16487         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
16488         for ARCHS4x.
16489         * config/arc/arc-cpus.def (hs4x): New cpu.
16490         (hs4xd): Likewise.
16491         * config/arc/arc-tables.opt: Regenerate.
16492         * config/arc/arc.c (arc_sched_issue_rate): New function.
16493         (TARGET_SCHED_ISSUE_RATE): Define.
16494         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
16495         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
16496         fpu_cvt.
16497         (attr tune): Add ARCHS4x tune values.
16498         (attr tune_dspmpy): Define.
16499         (*tst): Correct instruction type.
16500         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
16501         * config/arc/arcHS4x.md: New file.
16502         * config/arc/fpu.md: Update instruction type attributes.
16503         * config/arc/t-multilib: Regenerate.
16505 2018-07-16  Tom de Vries  <tdevries@suse.de>
16507         PR debug/86455
16508         * var-tracking.c (vt_initialize): Fix pre_dec handling.
16510 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16512         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
16513         early clobber.
16515 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
16517         PR tree-optimization/86514
16518         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
16519         conversion to a boolean type from a type with greater precision.
16521 2018-07-16  Tom de Vries  <tdevries@suse.de>
16523         * var-tracking.c (vt_initialize): Print adjusted insn slim if
16524         dump_flags request TDF_SLIM.
16526 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
16528         * fold-const.c (int_const_binop_1): Abstract...
16529         (wide_int_binop): ...wide int code here.
16530         (poly_int_binop): ...poly int code here.
16531         Abstract the rest of int_const_binop_1 into int_const_binop.
16532         * fold-const.h (wide_int_binop): New.
16533         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
16534         Remove useless PLUS/MINUS_EXPR case.
16535         (zero_nonzero_bits_from_vr): Move wide int code...
16536         (zero_nonzero_bits_from_bounds): ...here.
16537         (extract_range_from_binary_expr_1): Move mask optimization code...
16538         (range_easy_mask_min_max): ...here.
16539         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
16540         (range_easy_mask_min_max): New.
16542 2018-07-15  Jeff Law  <law@redhat.com>
16544         PR target/85993
16545         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
16546         block.
16548 2018-07-14  Jim Wilson  <jimw@sifive.com>
16550         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
16552 2018-07-14  Paul Koning  <ni1d@arrl.net>
16554         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
16556 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
16558         * lto-streamer-out.c (copy_function_or_variable): Dump info about
16559         copying section.
16561 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
16562             Steve Munroe  <munroesj52@gmail.com>
16564         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
16565         (_mm_andnot_si128): Likewise.
16566         (_mm_or_si128): Likewise.
16567         (_mm_xor_si128): Likewise.
16569 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
16571         PR middle-end/78809
16572         * builtins.c (expand_builtin_memcmp): Inline the calls first
16573         when result_eq is false.
16574         (expand_builtin_strcmp): Inline the calls first.
16575         (expand_builtin_strncmp): Likewise.
16576         (inline_string_cmp): New routine. Expand a string compare
16577         call by using a sequence of char comparison.
16578         (inline_expand_builtin_string_cmp): New routine. Inline expansion
16579         a call to str(n)cmp/memcmp.
16580         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
16581         New option.
16582         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
16584 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
16586         * config/arm/driver-arm.c: Include arm-native.h.
16587         (host_detect_local_cpu): Use auto-generated data tables.
16588         (vendors, arm_cpu_table): Delete.  Move part information to ...
16589         * config/arm/arm-cpus.in: ... here.
16590         * config/arm/parsecpu.awk (gen_native): New function.
16591         (vendor, part): New CPU fields.
16592         (END): Add support for building the native CPU detection tables.
16593         * config/arm/t-arm (arm-native.h): Add build rule.
16594         (driver-arm.o): Add dependency on arm-native.h.
16596 2018-07-13  Richard Biener  <rguenther@suse.de>
16598         PR middle-end/85974
16599         * match.pd (addr1 - addr2): Allow either of the operand to
16600         have a conversion.
16602 2018-07-13  Tom de Vries  <tdevries@suse.de>
16604         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
16605         in remap_ssa_name.
16607 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16609         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
16610         arrays instead of numbered variables.
16612 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
16614         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
16615         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
16617 2018-07-13  Richard Biener  <rguenther@suse.de>
16619         PR debug/86452
16620         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
16621         instead of get_context_die.
16623 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16624             Richard Biener  <rguenther@suse.de>
16626         PR middle-end/86489
16627         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
16628         that the loop latch destination where phi is defined.
16630 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
16632         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
16633         (riscv_expand_epilogue): Add assertion to check interrupt mode.
16634         (riscv_set_current_function): Extract getting interrupt type to new
16635         function.
16636         (riscv_get_interrupt_type): New function.
16637         (riscv_merge_decl_attributes): New function, checking interrupt type is
16638         same.
16639         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
16641 2018-07-12  Paul Koning  <ni1d@arrl.net>
16643         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
16644         directive.
16646 2018-07-12  Paul Koning  <ni1d@arrl.net>
16648         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
16649         zero reference, add doloop_end instead.
16650         * doc/md.texi (decrement_and_branch_until_zero): Remove.
16651         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
16652         detail for doloop_end.
16654 2018-07-12  Martin Sebor  <msebor@redhat.com>
16656         PR c/86453
16657         * attribs.c (decl_attributes): Reject conflicting attributes before
16658         calling attribute handlers.
16660 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
16662         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
16663          parameter.
16664         (gcc::dump_manager::get_dump_file_name): likewise.
16665         (dump_begin): Likewise.
16666         * dumpfile.h (dump_begin): Update prototype.
16667         (gcc::dump_manager::get_dump_file_name,
16668         gcc::dump_manager::get_dump_file_name): Update prototype.
16670 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
16672         * internal-fn.h (vectorizable_internal_fn_p): New function.
16673         * tree-vect-slp.c (compatible_calls_p): Likewise.
16674         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
16675         to internal functions.
16676         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
16678 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
16680         * fold-const.h (inverse_conditions_p): Declare.
16681         * fold-const.c (inverse_conditions_p): New function.
16682         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
16683         that test the inverse condition of a conditional internal function.
16684         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
16685         * internal-fn.c (internal_fn_mask_index): Handle conditional
16686         internal functions.
16687         (vectorized_internal_fn_supported_p): New function.
16688         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
16689         (any_pred_load_store): Replace with...
16690         (need_to_predicate): ...this new variable.
16691         (redundant_ssa_names): New variable.
16692         (ifcvt_can_use_mask_load_store): Move initial checks to...
16693         (ifcvt_can_predicate): ...this new function.  Handle tree codes
16694         for which a conditional internal function exists.
16695         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
16696         instead of ifcvt_can_use_mask_load_store.  Update after variable
16697         name change.
16698         (predicate_load_or_store): New function, split out from
16699         predicate_mem_writes.
16700         (check_redundant_cond_expr): New function.
16701         (value_available_p): Likewise.
16702         (predicate_rhs_code): Likewise.
16703         (predicate_mem_writes): Rename to...
16704         (predicate_statements): ...this.  Use predicate_load_or_store
16705         and predicate_rhs_code.
16706         (combine_blocks, tree_if_conversion): Update after above name changes.
16707         (ifcvt_local_dce): Handle redundant_ssa_names.
16708         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
16709         general conditional functions.
16710         * tree-vect-stmts.c (vectorizable_call): Likewise.
16712 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
16713             Alan Hayward  <alan.hayward@arm.com>
16714             David Sherwood  <david.sherwood@arm.com>
16716         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
16717         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
16718         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
16719         plus and minus and convert them into IFN_COND_FMA-based sequences.
16720         (convert_mult_to_fma): Handle conditional plus and minus.
16722 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
16724         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
16725         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
16726         (cond_fnms_optab): New optabs.
16727         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
16728         internal functions.
16729         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
16730         * internal-fn.h (get_conditional_internal_fn): Declare.
16731         (get_unconditional_internal_fn): Likewise.
16732         * internal-fn.c (cond_ternary_direct): New macro.
16733         (expand_cond_ternary_optab_fn): Likewise.
16734         (direct_cond_ternary_optab_supported_p): Likewise.
16735         (FOR_EACH_COND_FN_PAIR): Likewise.
16736         (get_conditional_internal_fn): New function.
16737         (get_unconditional_internal_fn): Likewise.
16738         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
16739         (gimple_match_op::gimple_match_op): Add a new overload for 5
16740         operands.
16741         (gimple_match_op::set_op): Likewise.
16742         (gimple_resimplify5): Declare.
16743         * genmatch.c (decision_tree::gen): Generate simplifications for
16744         5 operands.
16745         * gimple-match-head.c (gimple_simplify): Define an overload for
16746         5 operands.  Handle calls with 5 arguments in the top-level overload.
16747         (convert_conditional_op): Handle conversions from unconditional
16748         internal functions to conditional ones.
16749         (gimple_resimplify5): New function.
16750         (build_call_internal): Pass a fifth operand.
16751         (maybe_push_res_to_seq): Likewise.
16752         (try_conditional_simplification): Try converting conditional
16753         internal functions to unconditional internal functions.
16754         Handle 3-operand unconditional forms.
16755         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
16756         Define ternary equivalents of the current rules for binary conditional
16757         internal functions.
16758         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
16759         ternary operations.
16760         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
16761         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
16762         (optab): Handle them.
16763         (SVE_COND_FP_TERNARY): New int iterator.
16764         (sve_fmla_op, sve_fmad_op): New int attributes.
16765         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
16766         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
16767         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
16769 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
16771         * target.def (preferred_else_value): New target hook.
16772         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
16773         * doc/tm.texi: Regenerate.
16774         * targhooks.h (default_preferred_else_value): Declare.
16775         * targhooks.c (default_preferred_else_value): New function.
16776         * internal-fn.h (conditional_internal_fn_code): Declare.
16777         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
16778         (get_conditional_internal_fn): Use it.
16779         (conditional_internal_fn_code): New function.
16780         * gimple-match.h (gimple_match_cond): New struct.
16781         (gimple_match_op): Add a cond member function.
16782         (gimple_match_op::gimple_match_op): Update all forms to take a
16783         gimple_match_cond.
16784         * genmatch.c (expr::gen_transform): Use the same condition as res_op
16785         for the suboperation, but don't specify a particular else_value.
16786         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
16787         (visit_nary_op, visit_reference_op_load): Pass
16788         gimple_match_cond::UNCOND to the gimple_match_op constructor.
16789         * gimple-match-head.c: Include tree-eh.h
16790         (convert_conditional_op): New function.
16791         (maybe_resimplify_conditional_op): Likewise.
16792         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
16793         (gimple_resimplify2): Likewise.
16794         (gimple_resimplify3): Likewise.
16795         (gimple_resimplify4): Likewise.
16796         (maybe_push_res_to_seq): Return null for conditional operations.
16797         (try_conditional_simplification): New function.
16798         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
16799         constructor.
16800         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
16801         IFN_COND_* call.
16802         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
16803         function.
16804         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
16806 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
16808         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
16809         DECL_FCONTEXT
16810         (hash_tree): Do not hash DECL_FCONTEXT
16811         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
16812         Do not stream DECL_FCONTEXT.
16813         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
16814         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
16816 2018-07-12  Richard Biener  <rguenther@suse.de>
16818         PR debug/86462
16819         * dwarf2out.c (gen_block_die): Only output blocks when they have
16820         at least one !DECL_IGNORED_P variable.
16822 2018-07-12  Richard Biener  <rguenther@suse.de>
16824         PR target/84829
16825         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
16826         Remove -mieee-fp handling.
16828 2018-07-12  Richard Biener  <rguenther@suse.de>
16830         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
16831         left-over from last patch.
16833 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
16835         PR tree-optimization/86492
16836         * gimple-ssa-store-merging.c
16837         (imm_store_chain_info::coalesce_immediate_stores): Call
16838         check_no_overlap even for the merge_overlapping case.  Formatting fix.
16840 2018-07-12  Richard Biener  <rguenther@suse.de>
16842         PR middle-end/86479
16843         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
16844         move possibly trapping operations into the conditional.
16846 2018-07-12  Richard Biener  <rguenther@suse.de>
16848         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
16849         (vn_lookup_simplify_result): Remove recursion limit applied
16850         here.
16851         (vn_nary_build_or_lookup_1): Adjust.
16852         (try_to_simplify): Likewise.
16853         * gimple-match-head.c (gimple_resimplify1): Instead apply one
16854         here.
16855         (gimple_resimplify2): Likewise.
16856         (gimple_resimplify3): Likewise.
16857         (gimple_resimplify4): Likewise.
16859 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
16861         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
16862         Use __mmask64 type instead of __mmask8 for __M argument.
16863         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
16864         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
16865         __U argument.
16866         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
16867         __mmask16 for __M argument.
16868         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
16869         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
16870         to __mmask16 instead of __mmask8.
16871         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
16872         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
16873         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
16874         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
16875         instead of __mmask16 for __U argument.
16876         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
16877         __mmask16 instead of __mmask8 for __U argument.
16878         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
16879         __U argument.
16880         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
16881         __mmask16.
16882         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
16883         argument.
16884         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
16885         __U argument.
16886         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
16887         __mmask16.
16888         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
16889         of __mmask16.
16890         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
16891         __U argument.
16892         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
16893         __U argument.
16894         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
16895         __U argument.
16896         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
16897         __U argument.
16898         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
16899         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
16900         return type as well as __M argument type and all casts from __mmask8
16901         to __mmask32.
16902         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
16903         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
16904         return type as well as __M argument type and all casts from __mmask8
16905         to __mmask16.
16906         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
16907         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
16908         return type as well as __M argument type and all casts from __mmask8
16909         to __mmask32.
16910         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
16911         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
16912         return type as well as __M argument type and all casts from __mmask8
16913         to __mmask16.
16914         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
16915         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
16916         __mmask16.
16918 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
16920         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
16921         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
16922         for __U argument.
16924 2018-07-11  Paul Koning  <ni1d@arrl.net>
16926         * doc/md.texi (define_subst): Document how multiple occurrences of
16927         the same argument in the replacement pattern are handled.
16929 2018-07-11  Paul Koning  <ni1d@arrl.net>
16931         * doc/extend.texi (Common Variable Attributes): Move "mode" into
16932         alphabetical order.
16933         (Common Type Attributes): Add "mode" attribute.
16935 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
16937         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
16938         stream DECL_ORIGINAL_TYPE.
16939         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
16940         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
16941         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
16942         Do not walk original type.
16943         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
16944         external decls.
16945         (write_ts_decl_non_common_tree_pointers): Do not stream
16946         DECL_ORIGINAL_TYPE
16947         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
16948         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
16950 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
16952         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
16953         thread twice from the same starting edge.
16955 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
16957         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
16958         * gimple.c (gimple_call_nonnull_result_p): ...here...
16959         (gimple_call_nonnull_arg): ...and here.
16960         * gimple.h (gimple_call_nonnull_result_p): New.
16961         (gimple_call_nonnull_arg): New.
16963 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
16965         * config/arm/arm-cpus.in: Move information from fpu field of each
16966         cpu definition to the isa field.
16967         * config/arm/parsecpu.awk (fpu): Delete match rule.
16968         (gen_comm_data): Don't add bits from the CPU's FPU entry.
16970 2018-07-11  Richard Biener  <rguenther@suse.de>
16972         PR debug/86457
16973         * dwarf2out.c (init_sections_and_labels): Use
16974         output_asm_line_debug_info consistently.
16975         (dwarf2out_early_finish): Likewise.
16976         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
16977         type units.
16979 2018-07-11  Richard Biener  <rguenther@suse.de>
16981         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
16982         Rework father_bb setting in a way to avoid propagating constants
16983         multiple times on a loop body.
16985 2018-07-10  Mark Wielaard  <mark@klomp.org>
16987         PR debug/86459
16988         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
16990 2018-07-10  Richard Biener  <rguenther@suse.de>
16992         * hash-map.h (hash_map::iterator::operator*): Return
16993         references to key and value.
16995 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
16997         PR c++/86443
16998         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
16999         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
17000         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
17001         (gimplify_omp_for): For composite loops, move outer
17002         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
17003         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
17004         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
17005         TREE_LIST for both the original class iterator and the "last" helper
17006         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
17007         loop, remember has_decl_expr from outer composite loops for the
17008         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
17010 2018-07-09  Martin Sebor  <msebor@redhat.com>
17012         PR middle-end/77357
17013         PR middle-end/86428
17014         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
17015         accessing implicitly initialized array elements.
17016         * expr.c (string_constant): Handle string initializers of
17017         character arrays within aggregates.
17018         * gimple-fold.c (fold_array_ctor_reference): Add argument.
17019         Store element offset.  As a special case, handle zero size.
17020         (fold_nonarray_ctor_reference): Same.
17021         (fold_ctor_reference): Add argument.  Store subobject offset.
17022         * gimple-fold.h (fold_ctor_reference): Add argument.
17024 2018-07-09  Paul Koning  <ni1d@arrl.net>
17026         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
17027         (pdp11_insn_cost): New function.
17028         (pdp11_md_asm_adjust): New function.
17029         (TARGET_INVALID_WITHIN_DOLOOP): Define.
17030         (pdp11_rtx_costs): Update to match machine better.
17031         (output_addr_const_pdp11): Correct format mismatch warnings.
17032         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
17033         * config/pdp11/pdp11.md: General change to add base_cost and/or
17034         length attributes for use by new pdp11_insn_cost function.
17035         (MIN_BRANCH): Correct definition.
17036         (MIN_SOB): Ditto.
17037         (doloop_end): Use standard pattern name for looping pattern.
17038         (doloop_end_nocc): New.
17039         (movsf): Add another constraint alternative.
17040         (zero_extendqihi2): Add constraint alternatives for not in place
17041         extend.
17042         (zero_extendhisi2): Remove.
17043         (shift patterns): Add CC handling variants.
17044         (bswaphi2): New.
17045         (bswapsi2): New.
17046         (rothi3): New.
17047         (define_peephole2): New peephole to recognize mov that sets CC for
17048         subsequent test.
17050 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17052         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
17053         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
17054         wi::add.
17056 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
17058         PR c/86420
17059         * real.c (real_nextafter): Return true if result is denormal.
17061 2018-07-09  Martin Liska  <mliska@suse.cz>
17063         * common.opt: Add back wrongly removed attribute.
17065 2018-07-09  Richard Biener  <rguenther@suse.de>
17067         PR debug/86413
17068         * dwarf2out.c (gen_block_die): For an early generated DIE
17069         always output high/low PC attributes.
17071 2018-07-09  Tom de Vries  <tdevries@suse.de>
17073         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
17074         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
17075         onto VAR_DECL with abstract origin.
17077 2018-07-07  Jim Wilson  <jimw@sifive.com>
17079         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
17081 2018-07-07  Tom de Vries  <tdevries@suse.de>
17083         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
17085 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
17087         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
17088         overflow_type.
17089         (combine_bound): Use wide-int overflow calculation instead of
17090         rolling our own.
17091         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
17092         overflow_type.
17093         * fold-const.c (int_const_binop_2): Same.
17094         (extract_muldiv_1): Same.
17095         (fold_div_compare): Same.
17096         (fold_abs_const): Same.
17097         * match.pd: Same.
17098         * poly-int.h (add): Same.
17099         (sub): Same.
17100         (neg): Same.
17101         (mul): Same.
17102         * predict.c (predict_iv_comparison): Same.
17103         * profile-count.c (slow_safe_scale_64bit): Same.
17104         * simplify-rtx.c (simplify_const_binary_operation): Same.
17105         * tree-chrec.c (tree_fold_binomial): Same.
17106         * tree-data-ref.c (split_constant_offset_1): Same.
17107         * tree-if-conv.c (idx_within_array_bound): Same.
17108         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
17109         * tree-ssa-phiopt.c (minmax_replacement): Same.
17110         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
17111         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
17112         * vr-values.c (vr_values::adjust_range_with_scev): Same.
17113         * wide-int.cc (wi::add_large): Same.
17114         (wi::mul_internal): Same.
17115         (wi::sub_large): Same.
17116         (wi::divmod_internal): Same.
17117         * wide-int.h: Change overflow type to overflow_type for neg, add,
17118         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
17119         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
17120         mul_internal, divmod_internal.
17121         (overflow_type): New enum.
17122         (accumulate_overflow): New.
17124 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
17126         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
17127         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
17129 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
17131         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
17132         argument is checked for zero before entering loop, avoid checking again.
17134 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
17136         * gimplify.h (generic_expr_could_trap_p): Set as global function.
17137         * gimplify.h (generic_expr_could_trap_p): Likwise.
17138         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
17140 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
17142         PR tree-optimization/86401
17143         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
17144         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
17145         (fold_bit_and_mask): ... here.  New helper function for match.pd.
17146         * fold-const.h (fold_bit_and_mask): Declare.
17147         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
17149 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
17151         PR target/86324
17152         * target.def (translate_mode_attribute): New hook.
17153         * targhooks.h (default_translate_mode_attribute): Declare.
17154         * targhooks.c (default_translate_mode_attribute): New function.
17155         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
17156         * doc/tm.texi: Regenerate.
17157         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
17158         (rs6000_translate_mode_attribute): New function.
17160 2018-07-06  Paul Koning  <ni1d@arrl.net>
17162         * doc/md.texi (define_split): Document DONE and FAIL.
17163         (define_peephole2): Ditto.
17165 2018-07-05  Jeff Law  <law@redhat.com>
17167         PR tree-optimization/86010
17168         * tree-ssa-dse.c (compute_trims): More aggressively trim at
17169         both the head and tail of mem* and str* calls.
17171 2018-07-05  Jim Wilson  <jimw@sifive.com>
17173         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
17175 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
17177         * config/aarch64/aarch64-simd.md: correct flags text for
17178         MIN_EXPR replacement.
17180 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
17182         * configure: Regenerated.
17184 2018-07-05  Carl Love  <cel@us.ibm.com>
17186         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
17187         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
17188         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
17189         VSX_BUILTIN_DOUBLEL_V4SF.
17191 2018-07-05  Martin Sebor  <msebor@redhat.com>
17193         PR c++/86400
17194         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
17195         than its domain to compute its the upper bound of a char array.
17197 2018-07-05  Nathan Sidwell  <nathan@acm.org>
17199         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
17200         * doc/cpp.texi: Update comment.
17201         * doc/tm.texi: Rebuilt.
17202         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
17203         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
17204         * doc/extend.texi (Backwards Compatibility): Clarify it is system
17205         headers affected by extern "C".
17206         * system.h: Poison NO_IMPLICIT_EXTERN_C.
17207         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
17208         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
17209         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
17210         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
17211         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
17212         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
17213         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
17214         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
17215         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
17216         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
17217         config/sparc/sp64-elf.h, config/spu/spu.h,
17218         config/stormy16/stormy16.h, config/v850/v850.h,
17219         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
17220         define NO_IMPLICIT_EXTERN_C.
17221         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
17223 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
17225         PR target/84711
17226         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17227         instead of GET_MODE_SIZE when comparing Units.
17229 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
17231         PR target/84711
17232         * rtlanal.c (set_noop_p): Constrain on mode change,
17233         include hard-reg-set.h
17235 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
17237         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
17239 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
17241         Revert
17242         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
17244         PR sanitizer/84250
17245         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
17246         libasan.
17247         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
17249 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
17251         PR sanitizer/84250
17252         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
17253         libasan.
17254         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
17256 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
17258         PR middle-end/86380
17259         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
17261 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
17263         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
17264         neg_*_op* variables.
17266 2018-07-04  Martin Liska  <mliska@suse.cz>
17268         * tree-switch-conversion.c: Define
17269         max_ratio_for_speed and max_ratio_for_size constants.
17271 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
17272             Martin Liska  <mliska@suse.cz>
17274         PR middle-end/66240
17275         PR target/45996
17276         PR c/84100
17277         * common.opt: Rename align options with 'str_' prefix.
17278         * common/config/i386/i386-common.c (set_malign_value): New
17279         function.
17280         (ix86_handle_option): Use it to set -falign-* options/
17281         * config/aarch64/aarch64-protos.h (struct tune_params): Change
17282         type from int to string.
17283         * config/aarch64/aarch64.c: Update default values from int
17284         to string.
17285         * config/alpha/alpha.c (alpha_override_options_after_change):
17286         Likewise.
17287         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
17288         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17289         max skip conditionally.
17290         * config/i386/freebsd.h (SUBALIGN_LOG): New.
17291         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17292         max skip conditionally.
17293         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17294         max skip conditionally.
17295         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
17296         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17297         max skip conditionally.
17298         * config/i386/i386.c (struct ptt): Change type from int to
17299         string.
17300         (ix86_default_align): Set default values.
17301         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
17302         max skip conditionally.
17303         * config/i386/iamcu.h (SUBALIGN_LOG): New.
17304         (ASM_OUTPUT_MAX_SKIP_ALIGN):
17305         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
17306         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17307         max skip conditionally.
17308         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
17309         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
17310         * config/i386/x86-64.h (SUBALIGN_LOG): New.
17311         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17312         max skip conditionally.
17313         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17314         * config/ia64/ia64.c (ia64_option_override): Set default values
17315         for alignment options.
17316         * config/m68k/m68k.c: Handle new str_align_* options.
17317         * config/mips/mips.c (mips_set_compression_mode): Change
17318         type of constants.
17319         (mips_option_override): Set default values for options.
17320         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
17321         Likewise.
17322         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17323         Likewise.
17324         * config/rx/rx.c (rx_option_override): Likewise.
17325         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
17326         (LABEL_ALIGN): Use align_labels_log.
17327         (LOOP_ALIGN): Use align_loops_align.
17328         * config/s390/s390.c (s390_asm_output_function_label): Use new
17329         macros.
17330         * config/sh/sh.c (sh_override_options_after_change):
17331         Change type of constants.
17332         * config/spu/spu.c (spu_sched_init): Likewise.
17333         * config/sparc/sparc.c (sparc_option_override): Set default
17334         values for options.
17335         * config/visium/visium.c (visium_option_override): Likewise.
17336         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
17337         emit p2align format with last argument if it's not needed.
17338         * doc/invoke.texi: Document extended format of -falign-*.
17339         * final.c: Use align_labels alignment.
17340         * flags.h (struct target_flag_state): Change type to use
17341         align_flags.
17342         (struct align_flags_tuple): New.
17343         (struct align_flags): Likewise.
17344         (align_loops_log): Redefine macro to use new types.
17345         (align_loops_max_skip): Redefine macro to use new types.
17346         (align_jumps_log): Redefine macro to use new types.
17347         (align_jumps_max_skip): Redefine macro to use new types.
17348         (align_labels_log): Redefine macro to use new types.
17349         (align_labels_max_skip): Redefine macro to use new types.
17350         (align_functions_log): Redefine macro to use new types.
17351         (align_loops): Redefine macro to use new types.
17352         (align_jumps): Redefine macro to use new types.
17353         (align_labels): Redefine macro to use new types.
17354         (align_functions): Redefine macro to use new types.
17355         (align_functions_max_skip): Redefine macro to use new types.
17356         (align_loops_value): New macro.
17357         (align_jumps_value): New macro.
17358         (align_labels_value): New macro.
17359         (align_functions_value): New macro.
17360         * function.c (invoke_set_current_function_hook): Propagate
17361         alignment values from flags to global variables default in
17362         topleev.h.
17363         * ipa-icf.c (sem_function::equals_wpa): Use
17364         cl_optimization_option_eq instead of memcmp.
17365         * lto-streamer.h (cl_optimization_stream_out): Support streaming
17366         of string types.
17367         (cl_optimization_stream_in): Likewise.
17368         * optc-save-gen.awk: Support strings in cl_optimization.
17369         * opth-gen.awk: Likewise.
17370         * opts.c (finish_options): Remove error checking of invalid
17371         value ranges.
17372         (MAX_CODE_ALIGN): Remove.
17373         (MAX_CODE_ALIGN_VALUE): Likewise.
17374         (parse_and_check_align_values): New function.
17375         (check_alignment_argument): Likewise.
17376         (common_handle_option): Use check_alignment_argument.
17377         * opts.h (parse_and_check_align_values): Declare.
17378         * toplev.c (init_alignments): Remove.
17379         (read_log_maxskip): New.
17380         (parse_N_M): Likewise.
17381         (parse_alignment_opts): Likewise.
17382         (backend_init_target): Remove usage of init_alignments.
17383         * toplev.h (parse_alignment_opts): Declare.
17384         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
17385         argument.
17386         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
17387         * tree.c (cl_option_hasher::equal): New.
17388         * varasm.c: Use new global macros.
17390 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
17392         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
17393         Use a simpler align directive also if MAXSKIP = ALIGN-1.
17394         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17395         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17396         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17397         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17398         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
17399         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
17400         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
17401         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17402         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17403         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17404         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17406 2018-07-04  Martin Liska  <mliska@suse.cz>
17407             Jonathan Wakely  <jwakely@redhat.com>
17409         * coverage.c: Use correct type.
17410         * doc/invoke.texi: Language correction.
17412 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
17414         PR target/85620
17415         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
17416         ENDBRANCH for non-tail call which may return via indirect branch.
17417         * doc/extend.texi: Document indirect_return attribute.
17419 2018-07-03  Martin Sebor  <msebor@redhat.com>
17421         PR tree-optimization/86274
17422         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
17423         precondition.
17424         (format_floating): Correct handling of infinities and NaNs.
17426 2018-07-03  Martin Sebor  <msebor@redhat.com>
17428         * print-tree.c (print_real_cst): New function.
17429         (print_node_brief): Call it.
17430         (print_node): Ditto.
17432 2018-07-03  Jeff Law  <law@redhat.com>
17434         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
17435         into a single pattern.
17437         * config/h8300/h8300.md (ors code_iterator): New.
17438         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
17439         a single pattern and single splitter.
17440         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
17441         (iorqi3_1, xorqi3_1): Likewise.
17442         (iorqi3, xorqi3 expanders): Similarly.
17444         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
17445         (movmd_internal) into a single pattern using the P mode iterator.
17446         (movmd splitters): Similarly.
17447         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
17448         (movsd splitters): Similarly.
17450         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
17451         ADDB, ADDW and ADDL into a single ADD attribute which selects the
17452         right table based on the size of the operand.
17453         * config/h8300/h8300.md (length_table): Corresponding changes. All
17454         references to "addb", "addw" and "addl" changed to "add".
17455         (btst patterns): Merge two variants into a single pattern.
17456         (tstqi, tsthi): Likewise.
17457         (addhi3_incdec, addsi3_incdec): Likewise.
17458         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
17459         (mulhi3, mulsi3): Likewise.
17460         (udivhi3, udivsi3): Likewise.
17461         (divhi3, divsi3): Likewise.
17462         (andorqi3, andorhi3, andorsi3): Likewise.
17464 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
17466         PR target/85694
17467         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
17468         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
17470 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17472         PR tree-optimization/85694
17473         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
17474         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
17475         UNSPEC_URHADD.
17476         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
17477         (<u>avg<mode>3_ceil): New patterns.
17479 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
17481         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
17482         scan-tree-dump directive.
17483         * gcc.dg/vect/slp-perm-2.c: Likewise.
17484         * gcc.dg/vect/slp-perm-3.c: Likewise.
17485         * gcc.dg/vect/slp-perm-5.c: Likewise.
17486         * gcc.dg/vect/slp-perm-6.c: Likewise.
17487         * gcc.dg/vect/slp-perm-7.c: Likewise.
17488         * gcc.dg/vect/slp-perm-8.c: Likewise.
17490 2018-07-03  Marek Polacek  <polacek@redhat.com>
17492         PR middle-end/86202
17493         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
17495 2018-07-03  Richard Biener  <rguenther@suse.de>
17497         PR ipa/86389
17498         * tree-ssa-structalias.c (find_func_clobbers): Properly
17499         handle indirect calls.
17501 2018-07-03  Jeff Law  <law@redhat.com>
17503         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
17504         (shifts): New code iterator.
17505         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
17506         expander.  Fix HImode handling on H8/SX.
17507         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
17508         (subqi3, subhi3, subsi3 expanders): Likewise.
17509         (andqi3, andhi3, andsi3 expanders): Likewise.
17510         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
17511         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
17512         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
17513         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
17514         (zero_extendqihi2, zero_extendqisi2): Likewise.
17515         (extendqihi2, extendqisi2): Likewise.
17516         (rotlqi3, rotlhi3, rotlsi3): Likewise.
17517         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
17518         (rotlqi3_1, rotlhi3_1): Likewise.
17519         (logicalhi3_sn, logicalsi3_sn): Likewise.
17520         (logicalhi3, logicalsi3): Likewise.
17522 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17524         * tree-vect-patterns.c (vect_recog_rotate_pattern)
17525         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17526         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
17527         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
17528         type to append_pattern_def_seq instead of creating a stmt_vec_info
17529         directly.
17530         (build_mask_conversion): Likewise.  Remove vinfo argument.
17531         (vect_add_conversion_to_patterm): Likewise, renaming to...
17532         (vect_add_conversion_to_pattern): ...this.
17533         (vect_recog_mask_conversion_pattern): Update call to
17534         build_mask_conversion.  Pass the vector type to
17535         append_pattern_def_seq here too.
17536         (vect_recog_gather_scatter_pattern): Update call to
17537         vect_add_conversion_to_pattern.
17539 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17541         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
17542         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17543         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
17544         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
17545         STMT_VINFO_PATTERN_DEF_SEQ to null here.
17546         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
17547         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
17548         append_pattern_def_seq instead of new_pattern_def_seq.
17549         (vect_recog_divmod_pattern): Do both of the above.
17550         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
17551         is null.
17553 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17555         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
17556         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
17557         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
17558         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
17559         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
17560         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17561         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
17562         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
17563         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
17564         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
17565         parameter with a single stmt_vec_info.
17566         (vect_recog_func_ptr): Likewise.
17567         (vect_recog_gather_scatter_pattern): Likewise, folding in...
17568         (vect_try_gather_scatter_pattern): ...this.
17569         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
17570         the stmt_vec_info of the statement to be matched.  Don't clear
17571         STMT_VINFO_RELATED_STMT.
17572         (vect_pattern_recog): Update call accordingly.
17574 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17576         PR tree-optimization/85694
17577         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
17578         (uavgM3_ceil): Document new optabs.
17579         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
17580         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
17581         functions.
17582         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
17583         (savg_ceil_optab): New optabs.
17584         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
17585         (vect_vect_recog_func_ptrs): Add it.
17586         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
17587         constant directly from the associated lhs.
17589 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17591         * tree-vect-patterns.c (vect_split_statement): New function.
17592         (vect_convert_input): Use it to try to split an existing cast.
17594 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17596         * poly-int.h (print_hex): New function.
17597         * dumpfile.h (dump_dec, dump_hex): Declare.
17598         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
17599         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
17600         min_input_precision, operation_precision and operation_sign.
17601         * tree-vect-patterns.c (vect_get_range_info): New function.
17602         (vect_same_loop_or_bb_p, vect_single_imm_use)
17603         (vect_operation_fits_smaller_type): Delete.
17604         (vect_look_through_possible_promotion): Add an optional
17605         single_use_p parameter.
17606         (vect_recog_over_widening_pattern): Rewrite to use new
17607         stmt_vec_info infomration.  Handle one operation at a time.
17608         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
17609         (vect_truncatable_operation_p, vect_set_operation_type)
17610         (vect_set_min_input_precision): New functions.
17611         (vect_determine_min_output_precision_1): Likewise.
17612         (vect_determine_min_output_precision): Likewise.
17613         (vect_determine_precisions_from_range): Likewise.
17614         (vect_determine_precisions_from_users): Likewise.
17615         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
17616         (vect_vect_recog_func_ptrs): Put over_widening first.
17617         Add cast_forwprop.
17618         (vect_pattern_recog): Call vect_determine_precisions.
17620 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
17622         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
17623         statements that have been replaced by further pattern statements.
17624         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
17626 2018-07-03  Richard Biener  <rguenther@suse.de>
17628         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
17629         always set *dt.  Dump vectype in vectype overload.
17630         * dumpfile.h (dump_gimple_expr): New function.
17631         (dump_gimple_expr_loc): Likewise.
17632         * dumpfile.c (dump_gimple_expr): New function.
17633         (dump_gimple_expr_loc): Likewise.
17635 2018-07-02  Jeff Law  <law@redhat.com>
17637         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
17638         the H8/300, H8/300H and H8/S variants into a single pattern.
17639         (movhi_h8300, movqi_h8300hs): Similarly.
17640         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
17641         (QHI mode iterator): New.
17643         * config/h8300/h8300.md: Remove trailing whitespace.
17645 2018-07-02  Jim Wilson  <jimw@sifive.com>
17647         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
17648         instead of emit_insn for interrupt returns.
17649         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
17650         (riscv_sret, riscv_uret): Likewise.
17652 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
17654         * pretty-print.c (selftest::test_pp_format): Move save and restore
17655         of quotes to class auto_fix_quotes, and add an instance.
17656         * selftest.c: Include "intl.h".
17657         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
17658         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
17659         * selftest.h (selftest::auto_fix_quotes): New class.
17661 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
17663         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
17664         (aarch64_sve_prepare_conditional_op): Remove.
17665         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
17666         Allow aarch64_simd_reg_or_zero as select operand; remove
17667         the aarch64_sve_prepare_conditional_op call.
17668         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
17669         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17670         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
17671         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
17672         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
17673         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
17674         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
17675         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
17676         and a splitters to match all of the *_any patterns.
17677         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
17679         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
17680         (SVE_COND_FP_BINARY_REV): Remove.
17681         (sve_int_op_rev, sve_fp_op_rev): New.
17682         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
17683         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
17684         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
17685         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
17686         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
17687         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
17688         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
17689         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
17690         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
17692         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
17693         Remove match_dup 1 from the inner unspec.
17694         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17696         * config/aarch64/aarch64.md (movprfx): New attr.
17697         (length): Default movprfx to 8.
17698         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
17699         (*madd<SVE_I>, *msub<SVE_I): Likewise.
17700         (*<su>mul<SVE_I>3_highpart): Likewise.
17701         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
17702         (*v<ASHIFT><SVE_I>3): Likewise.
17703         (*<su><MAXMIN><SVE_I>3): Likewise.
17704         (*<su><MAXMIN><SVE_F>3): Likewise.
17705         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
17706         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
17707         (*div<SVE_F>4): Likewise.
17709 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
17711         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
17712         in dump string.
17714 2018-07-02  Richard Biener  <rguenther@suse.de>
17716         PR tree-optimization/86363
17717         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
17718         memset argument refers to a non-variable address.
17720 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
17722         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
17723         {PLUS,MINUS}_EXPR code to...
17724         (adjust_symbolic_bound): ...here,
17725         (combine_bound): ...here,
17726         (set_value_range_with_overflow): ...and here.
17728 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
17730         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
17731         code...
17732         (extract_range_from_abs_expr): ...here.
17734 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
17736         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
17737         -fno-omit-frame-pointer when not optimizing.
17739 2018-07-02  Martin Liska  <mliska@suse.cz>
17741         PR ipa/86279
17742         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
17743         (propagate_nothrow): Likewise.
17745 2018-07-02  Martin Liska  <mliska@suse.cz>
17747         PR ipa/86323
17748         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
17750 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
17752         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
17753         function in r262149, changing "loc" param from source_location to
17754         const dump_location_t &.
17755         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
17756         declaration, as above.
17758 2018-07-01  Paul Koning  <ni1d@arrl.net>
17760         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17761         -munit-asm, -mgnu-asm, -mdec-asm.
17762         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
17763         (pdp11_output_labelref): New.
17764         (pdp11_output_def): New.
17765         (pdp11_output_addr_vec_elt): New.
17766         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
17767         %# and %@ format codes.
17768         (pdp11_option_override): New.
17769         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
17770         (pdp11_output_ident): New.
17771         (pdp11_asm_named_section): New.
17772         (pdp11_asm_init_sections): New.
17773         (pdp11_file_start): New.
17774         (pdp11_file_end): New.
17775         (output_ascii): Use .ascii/.asciz for -mdec-asm.
17776         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
17777         %o, like %c but octal.
17778         (pdp11_option_override): New.
17779         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
17780         -mdec-asm.
17781         (DATA_SECTION_ASM_OP): Ditto.
17782         (READONLY_DATA_SECTION_ASM_OP): New.
17783         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
17784         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
17785         (ASM_OUTPUT_LABELREF): Ditto.
17786         (ASM_OUTPUT_DEF): Ditto.
17787         (ASM_OUTPUT_EXTERNAL): New.
17788         (ASM_OUTPUT_SOURCE_FILENAME): New.
17789         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
17790         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
17791         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
17792         %# and %@ format codes.
17793         * config/pdp11/pdp11.opt (mgnu-asm): New.
17794         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
17795         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
17796         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
17798 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
17800         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
17801         dereferencing path[] beyond its length.
17802         (debug_path): New.
17803         (debug_all_paths): New.
17804         (rewire_first_differing_edge): New.
17805         (adjust_paths_after_duplication): New.
17806         (duplicate_thread_path): Call adjust_paths_after_duplication.
17807         Add new argument.
17808         (thread_through_all_blocks): Add new argument to
17809         duplicate_thread_path.
17811 2018-06-30  Jim Wilson  <jimw@sifive.com>
17813         * config/riscv/predicates.md (p2m1_shift_operand): New.
17814         (high_mask_shift_operand): New.
17815         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
17816         pattern using p2m1_shift_operand.
17817         (lshsi3_zero_extend_3+2): New combiner pattern using
17818         high_mask_shift_operand.
17820 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17822         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
17823         split out from...
17824         (vect_recog_rotate_pattern): ...here.
17825         (vect_convert_input): Try to insert casts of invariants in the
17826         preheader.
17827         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
17828         preheader to be empty.
17830 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17832         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
17833         vector type.  If given, install it in the new statement's
17834         STMT_VINFO_VECTYPE.
17835         (vect_element_precision): New function.
17836         (vect_unpromoted_value): New struct.
17837         (vect_unpromoted_value::vect_unpromoted_value): New function.
17838         (vect_unpromoted_value::set_op): Likewise.
17839         (vect_look_through_possible_promotion): Likewise.
17840         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
17841         (vect_widened_op_tree, vect_convert_input): Likewise.
17842         (vect_convert_inputs, vect_convert_output): Likewise.
17843         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
17844         to handle the optional cast of the multiplication result and
17845         vect_widened_op_tree to detect the widened multiplication itself.
17846         Do not require the input and output of promotion casts to have
17847         the same sign, but base the signedness of the operation on the
17848         input rather than the result.  If the pattern includes two
17849         promotions, check that those promotions have the same sign.
17850         Do not restrict the MULT_EXPR handling to a double-width result;
17851         handle quadruple-width results and wider.  Use vect_convert_inputs
17852         to convert the inputs to the common type.
17853         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
17854         to handle the optional cast of the ABS result.  Also allow a sign
17855         change or a sign extension between the ABS and MINUS.
17856         Use vect_widened_op_tree to detect the widened subtraction and use
17857         vect_convert_inputs to convert the inputs to the common type.
17858         (vect_handle_widen_op_by_const): Delete.
17859         (vect_recog_widen_op_pattern): New function.
17860         (vect_recog_widen_mult_pattern): Use it.
17861         (vect_recog_widen_shift_pattern): Likewise.
17862         (vect_recog_widen_sum_pattern): Use
17863         vect_look_through_possible_promotion to handle the promoted
17864         PLUS_EXPR operand.
17866 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17868         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
17869         the containing gimple_seq *.
17870         * gimple-iterator.h (gsi_for_stmt): Declare it.
17871         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
17872         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
17873         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17874         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17875         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
17876         checks.
17877         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
17878         split out from...
17879         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
17880         statement being replaced is part of an existing pattern
17881         definition sequence, inserting the new pattern statements before
17882         the original one.
17883         (vect_pattern_recog_1): Don't return a bool.  If the statement
17884         is already part of a pattern, instead apply pattern matching
17885         to the pattern definition statements.  Don't clear the
17886         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
17887         (vect_pattern_recog): Don't break after the first match;
17888         continue processing the pattern definition statements instead.
17889         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
17891 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17893         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17894         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17895         (vect_recog_widen_sum_pattern): Use it.
17897 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17899         * tree-vect-loop.c (vectorizable_reduction): Assert that the
17900         phi is not a pattern statement and has not been replaced by
17901         a pattern statement.
17902         * tree-vect-patterns.c (type_conversion_p): Don't check
17903         STMT_VINFO_IN_PATTERN_P.
17904         (vect_recog_vector_vector_shift_pattern): Likewise.
17905         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
17906         the pattern statement rather than the original statement; check
17907         directly for a WIDEN_MULT_EXPR here.
17908         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
17909         vect_is_simple_use to return the pattern statement rather
17910         than the original statement; use is_pattern_stmt_p to check
17911         for such a pattern statement.
17912         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
17913         to return the pattern statement rather than the original statement;
17914         don't do the same transformation here.
17915         (vect_is_simple_use): If the defining statement has been replaced
17916         by a pattern statement, return the pattern statement instead.
17917         Remove the corresponding (local) transformation from the vectype
17918         overload.
17920 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17922         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
17923         end and default to null.
17924         * tree-vect-loop.c (vect_create_epilog_for_reduction)
17925         (vectorizable_reduction): Update calls accordingly, dropping the
17926         gimple ** argument if the passed-back statement isn't needed.
17927         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
17928         (vect_recog_rotate_pattern): Likewise.
17929         (vect_recog_mask_conversion_pattern): Likewise.
17930         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
17931         (vect_mask_constant_operand_p): Likewise.
17932         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
17933         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
17934         (get_group_load_store_type, get_load_store_type): Likewise.
17935         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
17936         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
17937         (vectorizable_conversion, vectorizable_assignment): Likewise.
17938         (vectorizable_shift, vectorizable_operation): Likewise.
17939         (vectorizable_store, vect_is_simple_cond): Likewise.
17940         (vectorizable_condition, vectorizable_comparison): Likewise.
17941         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
17942         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
17943         and move it to the end.  Cope with null def_stmt_outs.
17945 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17947         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
17949 2018-06-29  Jeff Law  <law@redhat.com>
17951         * config/v850/v850.c (v850_legitimate_address_p): Handle large
17952         displacements for TARGET_V850E2V3 and newer.
17953         (TARGET_LRA_P): Remove.  Defaults to LRA now.
17954         * config/v850/v850.md (sign23byte_load): Remove.
17955         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
17956         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
17958 2018-06-29  Martin Liska  <mliska@suse.cz>
17960         PR lto/85759
17961         * coverage.c (coverage_init): Mangle full path name.
17962         * doc/invoke.texi: Document the change.
17963         * gcov-io.c (mangle_path): New.
17964         * gcov-io.h (mangle_path): Likewise.
17965         * gcov.c (mangle_name): Use mangle_path for path mangling.
17967 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17969         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
17970         if starting source register is not even.
17972 2018-06-29  Martin Liska  <mliska@suse.cz>
17974         PR tree-optimization/86263
17975         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
17976         Make edge redirection.
17978 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
17980         * dumpfile.c (dump_loc): Add indentation based on scope depth.
17981         (dump_scope_depth): New variable.
17982         (get_dump_scope_depth): New function.
17983         (dump_begin_scope): New function.
17984         (dump_end_scope): New function.
17985         * dumpfile.h (get_dump_scope_depth): New declaration.
17986         (dump_begin_scope): New declaration.
17987         (dump_end_scope): New declaration.
17988         (class auto_dump_scope): New class.
17989         (AUTO_DUMP_SCOPE): New macro.
17990         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
17991         AUTO_DUMP_SCOPE.
17993 2018-06-29  Richard Biener  <rguenther@suse.de>
17995         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
17996         compute_all_dependences succeeds.
17997         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
17998         exceed --param loop-max-datarefs-for-datadeps.
18000 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
18002         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
18004 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
18006         PR target/86348
18007         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
18008         alternative 0 in preferred_for_speed attribute.
18010 2018-06-28  Paul Koning  <ni1d@arrl.net>
18012         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
18013         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
18014         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
18015         * config/pdp11/pdp11.md: Correct "length" attribute calculation
18016         for shift insn patterns.
18018 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
18020         * cgraph.c (cgraph_node::get_body): Replace assignments to
18021         "dump_file" with calls to set_dump_file.
18022         * dumpfile.c (alt_dump_file): Make static, and group with...
18023         (alt_flags): ...this definition.
18024         (dumps_are_enabled): New variable.
18025         (refresh_dumps_are_enabled): New function.
18026         (set_dump_file): New function.
18027         (set_alt_dump_file): New function.
18028         (gcc::dump_manager::dump_start): Replace assignments to
18029         "dump_file" and "alt_dump_file" with calls to set_dump_file and
18030         set_alt_dump_file.
18031         (gcc::dump_manager::dump_finish): Likewise.
18032         * dumpfile.h (alt_dump_file): Delete decl.
18033         (dumps_are_enabled): New variable decl.
18034         (set_dump_file): New function decl.
18035         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
18036         global.
18037         * tree-nested.c (lower_nested_functions): Replace assignments to
18038         "dump_file" with calls to set_dump_file.
18040 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
18042         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
18043         goto_locus of each outgoing edge of each basic block.
18045 2018-06-28  Richard Biener  <rguenther@suse.de>
18047         * dwarf2out.c (decl_scope_table): Remove.
18048         (push_decl_scope): Likewise.
18049         (pop_decl_scope): Likewise.
18050         (gen_type_die_for_member): Do not call push/pop_decl_scope.
18051         (gen_struct_or_union_type_die): Likewise.
18052         (gen_tagged_type_die): Likewise.
18053         (dwarf2out_init): Do not initialize decl_scope_table.
18054         (dwarf2out_c_finalize): Do not free it.
18056 2018-06-28  Richard Biener  <rguenther@suse.de>
18058         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
18059         deciding whether to not re-use a DIE.
18061 2018-06-28  Richard Biener  <rguenther@suse.de>
18063         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
18064         DW_AT_abstract_origin attribute.
18066 2018-06-28  Martin Liska  <mliska@suse.cz>
18068         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
18069         Use newly introduced constants.
18070         * tree-switch-conversion.h (struct jump_table_cluster):
18071         Define max_ratio_for_size and max_ratio_for_speed.
18073 2018-06-28  Martin Liska  <mliska@suse.cz>
18075         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18076         Add new checking assert to catch invalid state.
18077         (jump_table_cluster::can_be_handled): Handle single case
18078         clusters.
18079         (jump_table_cluster::is_beneficial): Bail out for such case.
18080         (bit_test_cluster::find_bit_tests):
18081         Add new checking assert to catch invalid state.
18082         (bit_test_cluster::can_be_handled): Handle single case
18083         clusters.
18084         (bit_test_cluster::is_beneficial): Bail out for such case.
18085         (switch_decision_tree::analyze_switch_statement):
18086         Fix comment.
18088 2018-06-28  Martin Liska  <mliska@suse.cz>
18090         * common.opt: Introduce -completion option.
18091         * gcc.c (driver_handle_option): Handle it.
18092         (driver::main): Print completions if completion
18093         is set.
18094         * opt-suggestions.c (option_proposer::get_completions):
18095         New function.
18096         (option_proposer::suggest_completion): Likewise.
18097         (option_proposer::find_param_completions): Likewise.
18098         (verify_autocompletions): Likewise.
18099         (test_completion_valid_options): Likewise.
18100         (test_completion_valid_params): Likewise.
18101         (in_completion_p): Likewise.
18102         (empty_completion_p): Likewise.
18103         (test_completion_partial_match): Likewise.
18104         (test_completion_garbage): Likewise.
18105         (opt_proposer_c_tests): Likewise.
18106         * opt-suggestions.h: Declare new functions.
18107         * opts.c (common_handle_option): Handle OPT__completion_.
18108         * selftest-run-tests.c (selftest::run_tests): Add
18109         opt_proposer_c_tests.
18110         * selftest.c (assert_str_startswith): New.
18111         * selftest.h (assert_str_startswith): Likewise.
18112         (opt_proposer_c_tests): New.
18113         (ASSERT_STR_STARTSWITH): Likewise.
18115 2018-06-28  Martin Liska  <mliska@suse.cz>
18117         * Makefile.in: Add opt-suggestions.o.
18118         * gcc-main.c: Include opt-suggestions.h.
18119         * gcc.c (driver::driver): Likewise.
18120         (driver::~driver): Remove m_option_suggestions.
18121         (driver::build_option_suggestions): Moved to option_proposer.
18122         (driver::suggest_option): Likewise.
18123         (driver::handle_unrecognized_options): Use option_proposer.
18124         * gcc.h (class driver): Add new memver m_option_proposer.
18125         * opt-suggestions.c: New file.
18126         * opt-suggestions.h: New file.
18128 2018-06-28  Martin Liska  <mliska@suse.cz>
18130         * vec.h (class auto_string_vec): New (moved from auto_argvec).
18131         (auto_string_vec::~auto_string_vec): Likewise.
18133 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
18135         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
18136         prevent_decl_creation_for_types fields up and add reset_location field.
18137         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
18138         statement if id->reset_location is true.
18139         (copy_edges_for_bb): Do not set goto_locus on the new edges if
18140         id->reset_location is true.
18141         (copy_phis_for_bb): Force input_location on the arguments if
18142         id->reset_location is true.
18143         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
18144         is set on the function to be inlined.
18146 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
18148         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
18150 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
18152         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
18153         registers for Pmode.
18154         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
18155         hard registers for the clobbered pseudo.
18157 2018-06-27  Paul Koning  <ni1d@arrl.net>
18159         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
18160         mutually exclusive options.
18161         * config/pdp11/constraints.md (h): New constraint.
18162         (O): Update definition to match shift code generation.
18163         (D): New constraint.
18164         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
18165         (CCFP): Remove.
18166         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
18167         function.
18168         (output_jump): Change arguments.
18169         (pdp11_fixed_cc_regs): New function.
18170         (pdp11_cc_mode): Ditto.
18171         (pdp11_expand_shift): Ditto.
18172         (pdp11_assemble_shift): Ditto.
18173         (pdp11_small_shift): Ditto.
18174         (pdp11_branch_cost): Remove.
18175         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
18176         from output.
18177         (pdp11_register_move_cost): Update for CC registers.
18178         (pdp11_rtx_costs): Add case for LSHIFTRT.
18179         (pdp11_output_jump): Add CCNZ mode conditional branches.
18180         (notice_update_cc_on_set): Remove.
18181         (pdp11_cc_mode): New function.
18182         (simple_memory_operand): Correct pre/post decrement case.
18183         (no_side_effect_operand): New function.
18184         (pdp11_regno_reg_class): Add CC_REGS class.
18185         (pdp11_fixed_cc_regs): New function.
18186         (pdp11_small_shift): New function.
18187         (pdp11_expand_shift): New function to expand shift insns.
18188         (pdp11_assemble_shift): New function to output shifts.
18189         (pdp11_branch_cost): Remove.
18190         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
18191         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
18192         (WCHAR_TYPE): Ditto.
18193         (PTRDIFF_TYPE): Ditto.
18194         (ADJUST_INSN_LENGTH): New macro.
18195         (FIXED_REGISTERS): Add CC registers.
18196         (CALL_USED_REGISTERS): Ditto.
18197         (reg_class): Ditto.
18198         (REG_CLASS_NAMES): Ditto.
18199         (REG_CLASS_CONTENTS): Ditto.
18200         (SELECT_CC_MODE): Use new function.
18201         (TARGET_FLAGS_REGNUM): New macro.
18202         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
18203         (cc0_reg_rtx): Remove.
18204         (CC_STATUS_MDEP): Remove.
18205         (CC_STATUS_MDEFP_INIT): Remove.
18206         (CC_IN_FPU): Remove.
18207         (NOTICE_UPDATE_CC): Remove.
18208         (REGISTER_NAMES): Add CC registers.
18209         (BRANCH_COST): Change to constant 1.
18210         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
18211         handling.
18212         * config/pdp11/pdp11.opt (mbcopy): Remove.
18213         (mbcopy-builtin): Remove.
18214         (mbranch-cheap): Remove.
18215         (mbranch-expensive): Remove.
18216         * config/pdp11/predicates.md (expand_shift_operand): Update to
18217         match shift code generation.
18218         (ccnz_operator): New predicate.
18219         * doc/invoke.texi (PDP-11 Options): Remove deleted options
18220         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
18221         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
18222         exclusive options.
18223         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
18224         description of O constraint.
18226 2018-06-27  Jeff Law  <law@redhat.com>
18227             Austin Law  <austinklaw@gmail.com>
18229         * config/v850/v850.md (addsi3_set_flags): New pattern.
18230         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
18231         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
18232         (zero_extendhisi2_v850_set_flags): Likewise.
18233         (zero_extendqisi2_v850_set_flags): Likewise.
18234         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
18235         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
18236         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
18238         * config/v850/v850-protos.h (notice_update_cc): Remove.
18239         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
18240         (v850_print_operand): Handle 'D' and "d".
18241         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
18242         Add handling of arithmetic/logical operations compared against zero.
18243         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
18244         Do not look at v850_compare_op, instead get mode from last argument.
18245         (v850_gen_compare): Remove
18246         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
18247         after reload for prologue insns.
18248         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
18249         patterns.
18250         (construct_save_jarl): Likewise.
18251         (TARGET_FLAGS_REGNUM): Define.
18252         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
18253         (NOTICE_UPDATE_CC): Remove.
18254         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
18255         than cc0.  Conditionalize on reload_completed.
18256         (cmpsi_insn, setfcc_insn): Likewise.
18257         (tst1 splitter): Turn into define_and_split which sets the flags
18258         after reload.
18259         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
18260         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
18261         (cstoresf4, cstoredf4): Clobber the flags.
18262         (cmpsi, cmpsf, cmpdf): Remove expanders.
18263         (setf_insn): Remove pattern.
18264         (addsi3): Turn into define_and_split which clobbers the flags after
18265         reload and a suitable pattern (addsi3_clobber_flags) for use after
18266         reload.
18267         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
18268         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
18269         (ashrsi3, ashrsi3_v850e2): Likewise.
18270         (bins): Clobber the flags.
18271         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
18272         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
18273         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
18274         (call_value_internal_short, call_value_internal_long): Likewise.
18275         (callt_save_interrupt, callt_return_interrupt): Likewise.
18276         (save_interrupt, return_interrupt): Likewise.
18277         (callt_save_all_interrupt, save_all_interrupt): Likewise.
18278         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
18279         (restore_all_interrupt, _restore_all_interrupt): Likewise.
18280         (All FP comparisons): Only allow after reload has completed.
18281         (trfsr): Likewise.
18282         (divh, divhu): Tweak output template.
18283         (branch_z_normal, branch_z_invert): Remove
18284         (branch_nz_normal, branch_nz_invert): Likewise.
18285         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
18287         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
18288         * config/v850/v850.c (notice_update_cc): Remove.
18289         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
18290         (CC_NO_CARRY): Likewise.
18291         (NOTICE_UPDATE_CC): Define to nothing.
18292         * config/v850/v850.md: Remove block comment on cc0 handling
18293         Remove "cc" attribute from all patterns.  Remove cc_status handling
18294         from all patterns.  Minor formatting fixes.
18296 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18298         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
18299         (cortex-a76.cortex-a55): Likewise.
18300         * config/aarch64/aarch64-tune.md: Regenerate.
18301         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
18302         cortex-a76.cortex-a55.
18304 2018-06-27  Jeff Law  <law@redhat.com>
18306         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
18307         (MULTILIB_DIRNAMES): Similarly.
18309 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
18311         * gimple.h (gimple_return_retbnd): Delete.
18312         (gimple_return_set_retbnd): Likewise.
18313         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
18314         gimple_return_set_retbnd.
18315         * gimple-pretty-print.c (dump_gimple_return): Remove call to
18316         gimple_return_retbnd and adjust.
18317         * tree-inline.h (struct copy_body_data): Remove retbnd field.
18318         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
18319         Explicitly return NULL in a couple more cases.  Move assertion
18320         on debug statements and remove unreachable code.
18321         (reset_debug_binding): Do not test id->retbnd.
18322         (expand_call_inline): Do not set it.
18324 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
18326         * configure.ac: Add --disable-gcov option.
18327         * configure: Regenerate.
18328         * Makefile.in: Honour @enable_gcov@.
18329         * doc/install.texi: Document --disable-gcov.
18331 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18333         * config/arm/arm-cpus.in (cortex-a76): New entry.
18334         (cortex-a76.cortex-a55): Likewise.
18335         * config/arm/arm-tables.opt: Regenerate.
18336         * config/arm/arm-tune.md: Likewise.
18337         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
18338         * doc/invoke.texi (ARM Options): Document cortex-a76 and
18339         cortex-a76.cortex-a55.
18341 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
18343         PR target/85769
18344         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
18346 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
18348         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
18349         comment.
18350         (EPILOGUE_USES): Likewise.
18352 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
18354         * tree-inline.c (remap_location): New function extracted from...
18355         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
18356         (copy_phis_for_bb): ...here.  Call remap_location.
18357         (copy_cfg_body): Adjust call to copy_edges_for_bb.
18359 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
18361         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
18362         unaligned vsx for 16B memset.
18364 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
18366         PR target/86285
18367         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
18368         ieee128_float_type_node to long_double_type_node unless
18369         TARGET_LONG_DOUBLE_128 is set.
18371 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
18373         * cfgloop.c (get_loop_location): Convert return type from
18374         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
18375         by implicit construction from rtx_insn *, and using
18376         dump_user_location_t::from_function_decl for the fallback case.
18377         * cfgloop.h (get_loop_location): Convert return type from
18378         location_t to dump_user_location_t.
18379         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
18380         dump_printf_loc to pass in a dump_location_t rather than a
18381         location_t, via the gimple stmt.
18382         * coverage.c (get_coverage_counts): Update calls to
18383         dump_printf_loc to pass in dump_location_t rather than a
18384         location_t.
18385         * doc/optinfo.texi (Dump types): Convert example of
18386         dump_printf_loc from taking "locus" to taking "insn".  Update
18387         description of the "_loc" calls to cover dump_location_t.
18388         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
18389         "selftest.h".
18390         (dump_user_location_t::dump_user_location_t): New constructors,
18391         from gimple *stmt and rtx_insn *.
18392         (dump_user_location_t::from_function_decl): New function.
18393         (dump_loc): Make static.
18394         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
18395         const dump_location_t &.
18396         (dump_generic_expr_loc): Delete.
18397         (dump_printf_loc): Convert param "loc" from location_t to
18398         const dump_location_t &.
18399         (selftest::test_impl_location): New function.
18400         (selftest::dumpfile_c_tests): New function.
18401         * dumpfile.h: Include "profile-count.h".
18402         (class dump_user_location_t): New class.
18403         (struct dump_impl_location_t): New struct.
18404         (class dump_location_t): New class.
18405         (dump_printf_loc): Convert 2nd param from source_location to
18406         const dump_location_t &.
18407         (dump_generic_expr_loc): Delete.
18408         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
18409         const dump_location_t &.
18410         * gimple-fold.c (fold_gimple_assign): Update call to
18411         dump_printf_loc to pass in a dump_location_t rather than a
18412         location_t, via the gimple stmt.
18413         (gimple_fold_call): Likewise.
18414         * gimple-loop-interchange.cc
18415         (loop_cand::analyze_iloop_reduction_var): Update for change to
18416         check_reduction_path.
18417         (tree_loop_interchange::interchange): Update for change to
18418         find_loop_location.
18419         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
18420         change in return-type of find_loop_location.
18421         (graphite_regenerate_ast_isl): Likewise.
18422         * graphite-optimize-isl.c (optimize_isl): Likewise.
18423         * graphite.c (graphite_transform_loops): Likewise.
18424         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
18425         pass in a dump_location_t rather than a location_t, via the
18426         gimple stmt.
18427         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18428         * ipa.c (walk_polymorphic_call_targets): Likewise.
18429         * loop-unroll.c (report_unroll): Convert "locus" param from
18430         location_t to dump_location_t.
18431         (decide_unrolling): Update for change to get_loop_location's
18432         return type.
18433         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
18434         location_t to dump_user_location_t.
18435         (grid_find_single_omp_among_assignments_1): Updates calls to
18436         dump_printf_loc to pass in a dump_location_t rather than a
18437         location_t, via the gimple stmt.
18438         (grid_parallel_clauses_gridifiable): Convert "tloc" from
18439         location_t to dump_location_t.  Updates calls to dump_printf_loc
18440         to pass in a dump_location_t rather than a location_t, via the
18441         gimple stmt.
18442         (grid_inner_loop_gridifiable_p): Likewise.
18443         (grid_dist_follows_simple_pattern): Likewise.
18444         (grid_gfor_follows_tiling_pattern): Likewise.
18445         (grid_target_follows_gridifiable_pattern): Likewise.
18446         (grid_attempt_target_gridification): Convert initialization
18447         of local "grid" from memset to zero-initialization; FIXME: does
18448         this require C++11?  Update call to dump_printf_loc to pass in a
18449         optinfo_location rather than a location_t, via the gimple stmt.
18450         * profile.c (read_profile_edge_counts): Updates call to
18451         dump_printf_loc to pass in a dump_location_t rather than a
18452         location_t
18453         (compute_branch_probabilities): Likewise.
18454         * selftest-run-tests.c (selftest::run_tests): Call
18455         dumpfile_c_tests.
18456         * selftest.h (dumpfile_c_tests): New decl.
18457         * tree-loop-distribution.c (pass_loop_distribution::execute):
18458         Update for change in return type of find_loop_location.
18459         * tree-parloops.c (parallelize_loops): Likewise.
18460         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
18461         "locus" from location_t to dump_user_location_t.
18462         (canonicalize_loop_induction_variables): Likewise.
18463         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
18464         for change in return type of find_loop_location.
18465         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
18466         to dump_printf_loc to pass in a dump_location_t rather than a
18467         location_t, via the stmt.
18468         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
18469         Likewise.
18470         * tree-vect-loop-manip.c (find_loop_location): Convert return
18471         type from source_location to dump_user_location_t.
18472         (vect_do_peeling): Update for above change.
18473         (vect_loop_versioning): Update for change in type of
18474         vect_location.
18475         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
18476         from location_t to dump_user_location_t.
18477         (vect_estimate_min_profitable_iters): Update for change in type
18478         of vect_location.
18479         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
18480         location_t to dump_location_t.
18481         (vect_slp_bb): Update for change in type of vect_location.
18482         * tree-vectorizer.c (vect_location): Convert from source_location
18483         to dump_user_location_t.
18484         (try_vectorize_loop_1): Update for change in vect_location's type.
18485         (vectorize_loops): Likewise.
18486         (increase_alignment): Likewise.
18487         * tree-vectorizer.h (vect_location): Convert from source_location
18488         to dump_user_location_t.
18489         (find_loop_location): Convert return type from source_location to
18490         dump_user_location_t.
18491         (check_reduction_path): Convert 1st param from location_t to
18492         dump_user_location_t.
18493         * value-prof.c (check_counter): Update call to dump_printf_loc to
18494         pass in a dump_user_location_t rather than a location_t; update
18495         call to error_at for change in type of "locus".
18496         (check_ic_target): Update call to dump_printf_loc to
18497         pass in a dump_user_location_t rather than a location_t, via the
18498         call_stmt.
18500 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
18502         * config/s390/s390.h (enum processor_flags): Do not use
18503         default tune parameter when -march was specified.
18505 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
18507         PR target/86314
18508         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
18509         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
18510         operands.
18512 2018-06-26  Richard Biener  <rguenther@suse.de>
18514         PR tree-optimization/86287
18515         PR bootstrap/86316
18516         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
18517         (vect_analyze_loop): Initialize n_stmts.
18519 2018-06-26  Richard Biener  <rguenther@suse.de>
18521         PR middle-end/86271
18522         * fold-const.c (fold_convertible_p): Pointer extension
18523         isn't valid.
18525 2018-06-26  Alexandre Oliva  <aoliva@redhat.com>
18527         PR debug/86064
18528         * dwarf2out.c (loc_list_has_views): Adjust comments.
18529         (dw_loc_list): Split single cross-partition range with
18530         nonzero locview.
18532 2018-06-25  Jeff Law  <law@redhat.com>
18534         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
18535         on -mbig-switch by default.
18537         * config/v850/predicates.md (const_float_1_operand): Fix match_code
18538         test.
18539         (const_float_0_operand): Remove unused predicate.
18540         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
18541         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
18542         (recipsf2): New expander.  Original pattern now called
18543         (recipsf2_insn).
18544         (recipdf2, recipdf2_insn): Similarly.
18545         (rsqrtsf2, rsqrtsf2_insn): Similarly
18546         (rsqrtdf2, rsqrtdf2_insn): Similarly
18548 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
18550         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
18551         Simplify logic for FreeBSD (twice).
18553 2018-06-25  Martin Sebor  <msebor@redhat.com>
18555         PR tree-optimization/86204
18556         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
18557         a strnlen result if it's less than the length of the string.
18559 2018-06-25  Martin Sebor  <msebor@redhat.com>
18561         PR tree-optimization/85700
18562         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
18563         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
18564         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
18566 2018-06-25  Martin Sebor  <msebor@redhat.com>
18568         * doc/extend.texi (Zero-length arrays): Update and clarify.
18570 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
18572         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
18573         added IEEE/IBM long double multilib support on PowerPC little
18574         endian Linux systems.
18575         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
18576         (MULTILIB_DEFAULTS): Likewise.
18577         * config/rs6000/rs6000.c (rs6000_option_override_internal):
18578         Likewise.
18579         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
18580         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
18581         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
18583 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
18585         PR middle-end/86311
18586         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
18587         (REORDER_45): Likewise.
18589 2018-06-25  Jeff Law  <law@redhat.com>
18591         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
18592         dividend to 32 bits.  Adjust length.
18593         (udivmodhi4): Cleanup output template.  Fix length.
18595 2018-06-25  Carl Love  <cel@us.ibm.com>
18597         * config/rs6000/vsx.md: Change word selector to prefered location.
18599 2018-06-25  Richard Biener  <rguenther@suse.de>
18601         PR tree-optimization/86304
18602         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
18603         epilogue-if-converted loops as well.
18605 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
18607         * lto-section-out.c (lto_begin_section): Do not print section
18608         name for noaddr and unnumbered dumps.
18610 2018-06-25  Richard Biener  <rguenther@suse.de>
18612         * tree-vectorizer.h (struct vec_info_shared): New structure
18613         with parts split out from struct vec_info and loop_nest from
18614         struct _loop_vec_info.
18615         (struct vec_info): Adjust accordingly.
18616         (struct _loop_vec_info): Likewise.
18617         (LOOP_VINFO_LOOP_NEST): Adjust.
18618         (LOOP_VINFO_DATAREFS): Likewise.
18619         (LOOP_VINFO_DDRS): Likewise.
18620         (struct _bb_vec_info): Likewise.
18621         (BB_VINFO_DATAREFS): Likewise.
18622         (BB_VINFO_DDRS): Likewise.
18623         (struct _stmt_vec_info): Add dr_aux member.
18624         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
18625         (DR_MISALIGNMENT_UNINITIALIZED): New.
18626         (set_dr_misalignment): Adjust.
18627         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
18628         (vect_analyze_loop): Adjust prototype.
18629         (vect_analyze_loop_form): Likewise.
18630         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
18631         Compute dependences lazily.
18632         (vect_record_base_alignments): Use shared datarefs/ddrs.
18633         (vect_verify_datarefs_alignment): Likewise.
18634         (vect_analyze_data_refs_alignment): Likewise.
18635         (vect_analyze_data_ref_accesses): Likewise.
18636         (vect_analyze_data_refs): Likewise.
18637         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
18638         constructor parameter for shared part.
18639         (vect_analyze_loop_form): Pass in shared part and adjust.
18640         (vect_analyze_loop_2): Pass in storage for the number of
18641         stmts.  Move loop nest finding to the caller.  Compute
18642         datarefs lazily.
18643         (vect_analyze_loop): Pass in shared part.
18644         (vect_transform_loop): Verify shared datarefs are unchanged.
18645         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
18646         constructor parameter for shared part.
18647         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
18648         (vect_slp_bb): Verify shared datarefs are unchanged before
18649         transform.
18650         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
18651         change.
18652         (new_stmt_vec_info): Initialize DR_AUX misalignment to
18653         DR_MISALIGNMENT_UNINITIALIZED.
18654         * tree-vectorizer.c (vec_info::vec_info): Add constructor
18655         parameter for shared part.
18656         (vec_info::~vec_info): Adjust.
18657         (vec_info_shared::vec_info_shared): New.
18658         (vec_info_shared::~vec_info_shared): Likewise.
18659         (vec_info_shared::save_datarefs): Likewise.
18660         (vec_info_shared::check_datarefs): Likewise.
18661         (try_vectorize_loop_1): Construct shared part live for analyses
18662         of a single loop for multiple vector sizes.
18663         * tree-parloops.c (gather_scalar_reductions): Adjust.
18665 2018-06-25  Richard Biener  <rguenther@suse.de>
18667         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
18668         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
18669         (vect_analyze_data_refs): Remove similar code from here and
18670         simplify accordingly.
18672 2018-06-25  Richard Biener  <rguenther@suse.de>
18674         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
18675         for reverse storage order accesses rather than asserting
18676         they cannot happen here.
18678 2018-06-25  Tom de Vries  <tdevries@suse.de>
18680         PR debug/86257
18681         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
18682         Use data16 instead of .byte for insn prefix.
18684 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
18686         PR C++/86082
18687         * parser.c (make_char_string_pack): Pass this literal chars
18688         through cpp_interpret_string.
18689         (cp_parser_userdef_numeric_literal): Check the result of
18690         make_char_string_pack.
18692 2018-06-24  Maya Rashish  <coypu@sdf.org>
18694         * ginclude/stddef.h: Simplify conditions around avoiding
18695         re-definition of __size_t.
18697 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
18699         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
18700         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
18702 2018-06-22  Maya Rashish  <coypu@sdf.org>
18704         * doc/invoke.texi (mno-fancy-math-387): Update for changes
18705         made to OpenBSD and NetBSD through the years.
18707 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18709         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18710         behavior of vec_pack (vector double, vector double) to match
18711         behavior of vec_float2 (vector double, vector double).
18713 2018-06-22  Olivier Hainque  <hainque@adacore.com>
18715         * gimplify.c (gimplify_function_tree): Prevent creation
18716         of a trampoline for the address of the current function
18717         passed to entry/exit instrumentation hooks.
18719 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
18721         PR target/86222
18722         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
18723         correctly.
18725 2018-06-22  Martin Liska  <mliska@suse.cz>
18727         PR tree-optimization/86263
18728         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18729         Bail out if is_enabled is false.
18730         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
18731         New declaration.
18732         (jump_table_cluster::is_enabled): New function.
18734 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
18736         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
18737         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
18738         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
18739         (lto_input_ts_binfo_tree_pointers): Likewise.
18740         * tree-streamer-out.c (streamer_write_tree_bitfields,
18741         write_ts_binfo_tree_pointers): Likewise.
18742         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
18744 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
18746         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
18748 2018-06-22  Martin Liska  <mliska@suse.cz>
18750         * symbol-summary.h (get): Make it pure and inline move
18751         functionality from ::get function.
18752         (get): Remove and inline into ::get and ::get_create.
18753         (get_create): Move code from ::get function.
18755 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18757         PR target/85994
18758         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
18759         -x assembler-with-cpp.
18761 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18763         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
18764         _FILE_OFFSET_BITS=64 for C++.
18766 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
18768         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
18769         conversion insn that shows up when pr85657-3.c is compiled using
18770         IEEE 128-bit long double.
18771         (neg<mode>2_internal): Use the correct mode to check whether the
18772         mode is IBM extended.
18773         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
18774         multiply and divide external functions from being created more
18775         than once.
18777 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
18779         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
18780         functions.
18781         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
18782         the edge can be forwarded.
18783         (cfg_layout_merge_blocks): Likewise.
18785 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
18787         * except.c (finish_eh_generation): Commit edge insertions only after
18788         the EH edges have been redirected from post-landing to landing pads.
18790 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
18792         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
18793         create_tmp_var_for to create the FRAME decl.
18794         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
18796 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
18798         * tree-inline.c (copy_edges_for_bb): Minor tweak.
18799         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
18800         debug statement when resetting its value.
18801         (expand_call_inline): Copy the locus of the call onto the assignment
18802         of the return value, if any.  Use local variable in more cases.
18804 2018-06-21  Martin Liska  <mliska@suse.cz>
18806         * ipa-pure-const.c (propagate_nothrow): Use
18807         funct_state_summaries->get.
18808         (dump_malloc_lattice): Likewise.
18809         (propagate_malloc): Likewise.
18811 2018-06-21  Richard Biener  <rguenther@suse.de>
18813         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
18814         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
18815         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
18816         comment.
18817         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
18818         BLOCK_ABSTRACT_ORIGIN unconditionally.
18820 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
18822         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
18823         deleting it.
18824         * ipa-reference.c (ipa_reference_c_finalize): Delete
18825         ipa_ref_opt_sum_summaries and set it to NULL.
18827 2018-06-21  Tom de Vries  <tdevries@suse.de>
18829         PR tree-optimization/85859
18830         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
18831         test with comment from bb_no_side_effects_p.
18833 2018-06-21  Richard Biener  <rguenther@suse.de>
18835         PR tree-optimization/86232
18836         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
18837         max for constant niter.
18839 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18841         * config/aarch64/aarch64-simd.md
18842         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
18844 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18846         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
18847         Make opernads of the unspec commutative.
18849 2018-06-21  Richard Biener  <rguenther@suse.de>
18851         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
18852         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
18853         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
18854         (vect_analyze_data_ref_dependence): Re-order checks to deal with
18855         NULL DR_STEP.
18856         (vect_record_base_alignments): Do not record base alignment
18857         for gathers or scatters.
18858         (vect_compute_data_ref_alignment): Drop return value that is always
18859         true.  Bail out early for gathers or scatters.
18860         (vect_enhance_data_refs_alignment): Bail out early for gathers
18861         or scatters.
18862         (vect_find_same_alignment_drs): Likewise.
18863         (vect_analyze_data_refs_alignment): Remove dead code.
18864         (vect_slp_analyze_and_verify_node_alignment): Likewise.
18865         (vect_analyze_data_refs): For possible gathers or scatters do
18866         not create an alternate DR, just check their possible validity
18867         and mark them.  Adjust DECL_NONALIASED handling to not rely
18868         on DR_BASE_ADDRESS.
18869         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
18870         update inits of gathers or scatters.
18871         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
18872         Also copy gather/scatter flag to pattern vinfo.
18874 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18876         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18877         behavior of vec_packsu (vector unsigned long long, vector unsigned
18878         long long) to match behavior of vec_packs with same signature.
18880 2018-06-20  Chung-Lin Tang  <cltang@codesourcery.com>
18881             Thomas Schwinge  <thomas@codesourcery.com>
18882             Cesar Philippidis  <cesar@codesourcery.com>
18884         * gimplify.c (gimplify_scan_omp_clauses): Add support for
18885         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18886         (gimplify_adjust_omp_clauses): Likewise.
18887         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
18888         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
18889         (gimplify_omp_target_update): Update handling of acc update and
18890         enter/exit data.
18891         * omp-low.c (install_var_field): Remove unused parameter
18892         base_pointers_restrict.
18893         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
18894         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
18895         FINALIZE}
18896         (omp_target_base_pointers_restrict_p): Delete.
18897         (scan_omp_target): Update call to scan_sharing_clauses.
18898         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
18899         FINALIZE}.
18900         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
18901         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18902         (convert_local_omp_clauses): Likewise.
18903         * tree-pretty-print.c (dump_omp_clause): Likewise.
18904         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
18905         FINALIZE}.
18906         (omp_clause_code_name): Likewise.
18908 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
18910         PR debug/86194
18911         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
18912         be narrowed.
18914         PR tree-optimization/86231
18915         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
18916         anti-range don't overwrite *vr0min before using it to compute *vr0max.
18918 2018-06-20  Tom de Vries  <tdevries@suse.de>
18920         PR tree-optimization/86097
18921         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
18922         iv type if signedness of iv type is not the same as that of *nit.
18924 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
18926         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
18927         EDGE_EH edges, verify they are all EDGE_EH.
18929 2018-06-20  Maya Rashish  <coypu@sdf.org>
18931         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
18933 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18935         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
18936         * config/aarch64/aarch64.c (xgene1_tunings): Add
18937         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
18938         (aarch64_mode_valid_for_sched_fusion_p):
18939         Allow 16-byte modes.
18940         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
18941         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
18942         128-bit modes.
18943         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
18944         New pattern.
18945         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
18946         * config/aarch64/iterators.md (VQ2): New mode iterator.
18948 2018-06-20  Martin Liska  <mliska@suse.cz>
18950         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
18951         Change default ratio from 10 to 8.
18953 2018-06-20  Martin Liska  <mliska@suse.cz>
18955         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18956         New.
18957         (bit_test_cluster::find_bit_tests): Likewise.
18958         (switch_decision_tree::analyze_switch_statement): Find clusters.
18959         * tree-switch-conversion.h (struct jump_table_cluster): Document
18960         hierarchy.
18962 2018-06-20  Martin Liska  <mliska@suse.cz>
18964         * tree-switch-conversion.c (switch_conversion::collect):
18965         Record m_uniq property.
18966         (switch_conversion::expand): Bail out for special conditions.
18967         (group_cluster::~group_cluster): New.
18968         (group_cluster::group_cluster): Likewise.
18969         (group_cluster::dump): Likewise.
18970         (jump_table_cluster::emit): New.
18971         (switch_decision_tree::fix_phi_operands_for_edges): New.
18972         (struct case_node): Remove struct.
18973         (jump_table_cluster::can_be_handled): New.
18974         (case_values_threshold): Moved to header.
18975         (reset_out_edges_aux): Likewise.
18976         (jump_table_cluster::is_beneficial): New.
18977         (bit_test_cluster::can_be_handled): Likewise.
18978         (add_case_node): Remove.
18979         (bit_test_cluster::is_beneficial): New.
18980         (case_bit_test::cmp): New.
18981         (bit_test_cluster::emit): New.
18982         (expand_switch_as_decision_tree_p): Remove.
18983         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
18984         (fix_phi_operands_for_edge): Likewise.
18985         (switch_decision_tree::analyze_switch_statement): New.
18986         (compute_cases_per_edge): Move ...
18987         (switch_decision_tree::compute_cases_per_edge): ... here.
18988         (try_switch_expansion): Likewise.
18989         (switch_decision_tree::try_switch_expansion): Likewise.
18990         (record_phi_operand_mapping): Likewise.
18991         (switch_decision_tree::record_phi_operand_mapping): Likewise.
18992         (emit_case_decision_tree): Likewise.
18993         (switch_decision_tree::emit): Likewise.
18994         (balance_case_nodes): Likewise.
18995         (switch_decision_tree::balance_case_nodes): Likewise.
18996         (dump_case_nodes): Likewise.
18997         (switch_decision_tree::dump_case_nodes): Likewise.
18998         (emit_jump): Likewise.
18999         (switch_decision_tree::emit_jump): Likewise.
19000         (emit_cmp_and_jump_insns): Likewise.
19001         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
19002         (emit_case_nodes): Likewise.
19003         (switch_decision_tree::emit_case_nodes): Likewise.
19004         (conditional_probability): Remove.
19005         * tree-switch-conversion.h (enum cluster_type): New.
19006         (PRINT_CASE): New.
19007         (struct cluster): Likewise.
19008         (cluster::cluster): Likewise.
19009         (struct simple_cluster): Likewise.
19010         (simple_cluster::simple_cluster): Likewise.
19011         (struct group_cluster): Likewise.
19012         (struct jump_table_cluster): Likewise.
19013         (struct bit_test_cluster): Likewise.
19014         (struct min_cluster_item): Likewise.
19015         (struct case_tree_node): Likewise.
19016         (case_tree_node::case_tree_node): Likewise.
19017         (jump_table_cluster::case_values_threshold): Likewise.
19018         (struct case_bit_test): Likewise.
19019         (struct switch_decision_tree): Likewise.
19020         (struct switch_conversion): Likewise.
19021         (switch_decision_tree::reset_out_edges_aux): Likewise.
19023 2018-06-20  Martin Liska  <mliska@suse.cz>
19025         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
19026         (hoist_edge_and_branch_if_true): Likewise.
19027         (expand_switch_using_bit_tests_p): Likewise.
19028         (struct case_bit_test): Likewise.
19029         (case_bit_test_cmp): Likewise.
19030         (emit_case_bit_tests): Likewise.
19031         (switch_conversion::switch_conversion): New class.
19032         (struct switch_conv_info): Remove old struct.
19033         (collect_switch_conv_info): More to ...
19034         (switch_conversion::collect): ... this.
19035         (check_range): Likewise.
19036         (switch_conversion::check_range): Likewise.
19037         (check_all_empty_except_final): Likewise.
19038         (switch_conversion::check_all_empty_except_final): Likewise.
19039         (check_final_bb): Likewise.
19040         (switch_conversion::check_final_bb): Likewise.
19041         (create_temp_arrays): Likewise.
19042         (switch_conversion::create_temp_arrays): Likewise.
19043         (free_temp_arrays): Likewise.
19044         (gather_default_values): Likewise.
19045         (switch_conversion::gather_default_values): Likewise.
19046         (build_constructors): Likewise.
19047         (switch_conversion::build_constructors): Likewise.
19048         (constructor_contains_same_values_p): Likewise.
19049         (switch_conversion::contains_same_values_p): Likewise.
19050         (array_value_type): Likewise.
19051         (switch_conversion::array_value_type): Likewise.
19052         (build_one_array): Likewise.
19053         (switch_conversion::build_one_array): Likewise.
19054         (build_arrays): Likewise.
19055         (switch_conversion::build_arrays): Likewise.
19056         (gen_def_assigns): Likewise.
19057         (switch_conversion::gen_def_assigns): Likewise.
19058         (prune_bbs): Likewise.
19059         (switch_conversion::prune_bbs): Likewise.
19060         (fix_phi_nodes): Likewise.
19061         (switch_conversion::fix_phi_nodes): Likewise.
19062         (gen_inbound_check): Likewise.
19063         (switch_conversion::gen_inbound_check): Likewise.
19064         (process_switch): Use the newly created class.
19065         (switch_conversion::expand): New.
19066         (switch_conversion::~switch_conversion): New.
19067         * tree-switch-conversion.h: New file.
19069 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19071         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
19072         tree-vect-patterns.c.
19073         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
19074         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
19075         (vect_recog_sad_pattern): Likewise.
19076         (vect_recog_widen_sum_pattern): Likewise.
19077         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
19078         (vect_recog_widen_shift_pattern): Remove the type_in argument.
19079         (vect_recog_rotate_pattern): Likewise.
19080         (vect_recog_mult_pattern): Likewise.
19081         (vect_recog_vector_vector_shift_pattern): Likewise.
19082         (vect_recog_divmod_pattern): Likewise.
19083         (vect_recog_mixed_size_cond_pattern): Likewise.
19084         (vect_recog_bool_pattern): Likewise.
19085         (vect_recog_mask_conversion_pattern): Likewise.
19086         (vect_try_gather_scatter_pattern): Likewise.
19087         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
19088         (vect_recog_over_widening_pattern): Likewise.
19089         (vect_recog_gather_scatter_pattern): Likewise.
19090         (vect_recog_func_ptr): Move from tree-vectorizer.h
19091         (vect_vect_recog_func_ptrs): Move further down the file.
19092         (vect_recog_func): Likewise.  Remove the third argument.
19093         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
19094         (vect_pattern_recog_1): Expect the pattern function to do any
19095         necessary target tests.  Also expect it to provide a vector type.
19096         Remove the type_in handling.
19098 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19100         * tree-vect-patterns.c (vect_pattern_detected): New function.
19101         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
19102         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
19103         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
19104         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
19105         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
19106         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
19107         (vect_recog_mask_conversion_pattern)
19108         (vect_try_gather_scatter_pattern): Likewise.
19110 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19112         * tree-vect-patterns.c (vect_get_internal_def): New function.
19113         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
19114         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
19115         (search_type_for_mask_1): Use it.
19117 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19119         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
19120         redundant WIDEN_SUM_EXPR handling.
19121         (vect_recog_sad_pattern): Likewise.
19123 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19125         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
19126         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
19127         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
19128         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
19129         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
19131 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19133         * tree-vect-stmts.c (vectorizable_call): Make sure that we
19134         use the stmt_vec_info of the original bb statement for the
19135         new zero assignment, even if the call is part of a pattern.
19137 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19139         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
19140         that the sequence is attached to the original statement rather
19141         than the pattern statement.
19142         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
19143         PATTERN_DEF_SEQ from the original statement rather than
19144         the main pattern statement.
19145         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
19146         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
19147         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
19149 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
19151         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
19152         definition statements before the early exit for statements that aren't
19153         live or relevant.
19154         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
19155         split out from...
19156         (vect_transform_loop): ...here.  Process pattern definition
19157         statements without first checking whether the main pattern
19158         statement is live or relevant.
19160 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
19162         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
19163         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
19165 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
19167         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
19168         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
19169         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
19170         (expand_block_compare): Change select_block_compare_mode call.
19171         (expand_strncmp_align_check): Use new functions, fix comment.
19172         (emit_final_str_compare_gpr): New function.
19173         (expand_strn_compare): Refactor and clean up code.
19174         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
19176 2018-06-19  Tony Reix  <tony.reix@atos.com>
19177             Damien Bergamini  <damien.bergamini@atos.com>
19178             David Edelsohn  <dje.gcc@gmail.com>
19180         * collect2.c (static_obj): New variable.
19181         (static_libs): New variable.
19182         (is_in_list): Uncomment declaration.
19183         (main): Track AIX libraries linked statically.
19184         (is_in_list): Uncomment definition.
19185         (scan_prog_file): Don't add AIX shared libraries initializer
19186         to constructor list if linking statically.
19188 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
19190         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
19191         constant.
19192         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
19194 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
19196         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
19197         blocks.
19199 2018-06-19  Martin Liska  <mliska@suse.cz>
19201         * config/i386/i386.c (ix86_can_inline_p): Do not use
19202         ipa_fn_summaries::get_create.
19203         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
19204         get.
19205         (devirtualization_time_bonus): Likewise.
19206         (ipcp_propagate_stage): Likewise.
19207         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
19208         (edge_set_predicate): Likewise.
19209         (evaluate_conditions_for_known_args): Likewise.
19210         (evaluate_properties_for_edge): Likewise.
19211         (ipa_call_summary::reset): Tranform to ...
19212         (ipa_call_summary::~ipa_call_summary): ... this.
19213         (ipa_fn_summary::reset): Transform to ...
19214         (ipa_fn_summary::~ipa_fn_summary): ... this.
19215         (ipa_fn_summary_t::remove): Rename to ...
19216         (ipa_fn_summary_t::remove_callees): ... this.
19217         (ipa_fn_summary_t::duplicate): Use placement new
19218         instead of memory copy.
19219         (ipa_call_summary_t::duplicate): Likewise.
19220         (ipa_call_summary_t::remove): Remove.
19221         (dump_ipa_call_summary): Change get_create to get.
19222         (ipa_dump_fn_summary): Dump only when summary exists.
19223         (analyze_function_body): Use symbol_summary::get instead
19224         of get_create.
19225         (compute_fn_summary): Likewise.
19226         (estimate_edge_devirt_benefit): Likewise.
19227         (estimate_edge_size_and_time): Likewise.
19228         (inline_update_callee_summaries): Likewise.
19229         (remap_edge_change_prob): Likewise.
19230         (remap_edge_summaries): Likewise.
19231         (ipa_merge_fn_summary_after_inlining): Likewise.
19232         (write_ipa_call_summary): Likewise.
19233         (ipa_fn_summary_write): Likewise.
19234         (ipa_free_fn_summary): Likewise.
19235         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
19236         (struct ipa_call_summary): Likewise.
19237         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
19238         of get_create.
19239         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
19240         (estimate_size_after_inlining): Likewise.
19241         (estimate_growth): Likewise.
19242         (growth_likely_positive): Likewise.
19243         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
19244         (inline_call): Likewise.
19245         * ipa-inline.c (caller_growth_limits): Likewise.
19246         (can_inline_edge_p): Likewise.
19247         (can_inline_edge_by_limits_p): Likewise.
19248         (compute_uninlined_call_time): Likewise.
19249         (compute_inlined_call_time): Likewise.
19250         (want_inline_small_function_p): Likewise.
19251         (edge_badness): Likewise.
19252         (update_caller_keys): Likewise.
19253         (update_callee_keys): Likewise.
19254         (inline_small_functions): Likewise.
19255         (inline_to_all_callers_1): Likewise.
19256         (dump_overall_stats): Likewise.
19257         (early_inline_small_functions): Likewise.
19258         (early_inliner): Likewise.
19259         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
19260         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
19261         * ipa-pure-const.c (malloc_candidate_p): Likewise.
19262         * ipa-split.c (execute_split_functions): Likewise.
19263         * symbol-summary.h: Likewise.
19264         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
19266 2018-06-19  Richard Biener  <rguenther@suse.de>
19268         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
19269         (vectorize_loops): ... here.  Fix dbgcnt handling.
19270         (try_vectorize_loop): Wrap try_vectorize_loop_1.
19272 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
19274         PR target/86197
19275         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
19276         ieee128 argument takes up only one (vector) register, not two (floating
19277         point) registers.
19279 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
19281         * gimplify.c (gimplify_init_constructor): Really never clear for an
19282         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
19284 2018-06-19  Richard Biener  <rguenther@suse.de>
19286         PR tree-optimization/86179
19287         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
19288         after failed recognition.
19290 2018-06-18  Martin Sebor  <msebor@redhat.com>
19292         PR middle-end/85602
19293         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
19294         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
19295         Handle integer subtraction.
19296         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
19297         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
19299 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
19301         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
19302         param from rtx to rtx_insn *.
19303         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
19304         param.
19305         (frv_ifcvt_modify_insn): Likwise.
19306         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
19307         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
19308         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
19309         as_a <rtx_insn *> cast to local "unprotected_region" once
19310         it's been established that it's not NULL or pc_rtx.
19311         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
19312         param "sethi" from rtx to rtx_insn *.
19313         (nds32_group_float_insns): Likewise for param "insn".
19314         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
19315         param.
19316         (vax_output_int_subtract): Likewise.
19317         * config/vax/vax.c (vax_output_int_add): Likewise for param
19318         "insn".
19319         (vax_output_int_subtract): Likewise.
19320         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
19321         (emit_pattern_after): Likewise for param "after".
19322         (emit_insn_after): Likewise.
19323         (emit_jump_insn_after): Likewise.
19324         (emit_call_insn_after): Likewise.
19325         (emit_debug_insn_after): Likewise.
19326         (emit_pattern_before): Likewise for param "before".
19327         (emit_insn_before): Likewise.
19328         (emit_jump_insn_before): Likewise.
19329         * final.c (get_insn_template): Likewise for param "insn", removing
19330         a cast.
19331         * output.h (get_insn_template): Likewise for 2nd param.
19332         * rtl.h (emit_insn_before): Likewise.
19333         (emit_jump_insn_before): Likewise.
19334         (emit_debug_insn_before_noloc): Likewise.
19335         (emit_insn_after): Likewise.
19336         (emit_jump_insn_after): Likewise.
19337         (emit_call_insn_after): Likewise.
19338         (emit_debug_insn_after): Likewise.
19339         (set_insn_deleted): Likewise for param.
19341 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
19343         PR target/85358
19344         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
19345         floating point modes, so that IFmode is numerically greater than
19346         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
19347         to declare the ordering.  This prevents IFmode from being
19348         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
19349         machine.  Include rs6000-modes.h to share the fractional values
19350         between genmodes* and the rest of the compiler.
19351         (IFmode): Likewise.
19352         (KFmode): Likewise.
19353         (TFmode): Likewise.
19354         * config/rs6000/rs6000-modes.h: New file.
19355         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
19356         meaning of rs6000_long_double_size so that 126..128 selects an
19357         appropriate 128-bit floating point type.
19358         (rs6000_option_override_internal): Likewise.
19359         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
19360         (TARGET_LONG_DOUBLE_128): Change the meaning of
19361         rs6000_long_double_size so that 126..128 selects an appropriate
19362         128-bit floating point type.
19363         (LONG_DOUBLE_TYPE_SIZE): Update comment.
19364         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
19365         source and destination to match the standard usage.
19366         (truncifkf2): Likewise.
19367         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
19368         ISA 2.07 to use an explicit clobber, instead of passing in a
19369         temporary.
19370         (copysign<mode>3_soft): Likewise.
19372 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
19374         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
19375         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
19376         (vect_slp_analyze_instance_dependence): Likewise.
19377         (vect_enhance_data_refs_alignment): Likewise.
19378         (vect_analyze_data_refs_alignment): Likewise.
19379         (vect_slp_analyze_and_verify_instance_alignment
19380         (vect_analyze_data_ref_accesses): Likewise.
19381         (vect_prune_runtime_alias_test_list): Likewise.
19382         (vect_analyze_data_refs): Likewise.
19383         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
19384         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19385         (vect_analyze_scalar_cycles_1): Likewise.
19386         (vect_get_loop_niters): Likewise.
19387         (vect_analyze_loop_form_1): Likewise.
19388         (vect_update_vf_for_slp): Likewise.
19389         (vect_analyze_loop_operations): Likewise.
19390         (vect_analyze_loop): Likewise.
19391         (vectorizable_induction): Likewise.
19392         (vect_transform_loop): Likewise.
19393         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
19394         * tree-vect-slp.c (vect_analyze_slp): Likewise.
19395         (vect_make_slp_decision): Likewise.
19396         (vect_detect_hybrid_slp): Likewise.
19397         (vect_slp_analyze_operations): Likewise.
19398         (vect_slp_bb): Likewise.
19399         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
19400         (vectorizable_bswap): Likewise.
19401         (vectorizable_call): Likewise.
19402         (vectorizable_simd_clone_call): Likewise.
19403         (vectorizable_conversion): Likewise.
19404         (vectorizable_assignment): Likewise.
19405         (vectorizable_shift): Likewise.
19406         (vectorizable_operation): Likewise.
19407         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
19409 2018-06-18  Martin Sebor  <msebor@redhat.com>
19411         PR tree-optimization/81384
19412         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
19413         * builtins.c (expand_builtin_strnlen): New function.
19414         (expand_builtin): Call it.
19415         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
19416         * builtins.def (BUILT_IN_STRNLEN): New.
19417         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
19418         Warn for bounds in excess of maximum object size.
19419         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
19420         single-value ranges.  Handle strnlen.
19421         (handle_builtin_strlen): Handle strnlen.
19422         (strlen_check_and_optimize_stmt): Same.
19423         * doc/extend.texi (Other Builtins): Document strnlen.
19425 2018-06-18  Maya Rashish  <coypu@sdf.org>
19427         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
19428         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
19429         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
19431         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
19432         here to ...
19433         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
19435 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19437         * tree.c (escaped_string::escape): Replace cast to char * by
19438         const_cast<char *> (unescaped).
19440 2018-06-18  Nick Clifton  <nickc@redhat.com>
19442         PR 84195
19443         * tree.c (escaped_string): New class.  Converts an unescaped
19444         string into its escaped equivalent.
19445         (warn_deprecated_use): Use the new class to convert the
19446         deprecation message, if present.
19447         (test_escaped_strings): New self test.
19448         (test_c_tests): Add test_escaped_strings.
19449         * doc/extend.texi (deprecated): Add a note that the
19450         deprecation message is affected by the -fmessage-length
19451         option, and that control characters will be escaped.
19452         (#pragma GCC error): Document this pragma.
19453         (#pragma GCC warning): Likewise.
19454         * doc/invoke.texi (-fmessage-length): Document this option's
19455         effect on the #warning and #error preprocessor directives and
19456         the deprecated attribute.
19458 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
19460         * tree.c (decl_value_expr_lookup): Revert latest change.
19461         (decl_value_expr_insert): Likewise.
19463 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
19465         * gimplify.c (nonlocal_vlas): Delete.
19466         (nonlocal_vla_vars): Likewise.
19467         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
19468         referenced VLAs.
19469         (gimplify_body): Do not create and destroy nonlocal_vlas.
19470         * tree-nested.c: Include diagnostic.h.
19471         (use_pointer_in_frame): Tweak.
19472         (lookup_field_for_decl): Add assertion and declare the transformation.
19473         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
19474         internal error when the reference is in a wrong context.  Do not
19475         create a debug decl by default.
19476         (note_nonlocal_block_vlas): Delete.
19477         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
19478         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
19479         create a debug decl by default.
19480         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
19481         call is in a wrong context.
19482         (fixup_vla_decls): New function.
19483         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
19484         debug variables were created.
19485         * tree.c (decl_value_expr_lookup): Add checking assertion.
19486         (decl_value_expr_insert): Likewise.
19488 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
19490         PR middle-end/82479
19491         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
19492         * tree-scalar-evolution.c (interpret_expr): Likewise.
19493         (expression_expensive_p): Likewise.
19494         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
19495         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
19496         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
19497         (ssa_defined_by_minus_one_stmt_p): New.
19499 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
19501         PR middle-end/64946
19502         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
19503         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19504         * dojump.c (do_jump): Likewise.
19505         * expr.c (expand_expr_real_2): Check operand type's sign.
19506         * fold-const.c (const_unop): Handle ABSU_EXPR.
19507         (fold_abs_const): Likewise.
19508         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
19509         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
19510         (strip_sign_op_1): Likesise.
19511         * match.pd: Add new pattern to generate ABSU_EXPR.
19512         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
19513         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
19514         * tree-eh.c (operation_could_trap_helper_p): Likewise.
19515         * tree-inline.c (estimate_operator_cost): Likewise.
19516         * tree-pretty-print.c (dump_generic_node): Likewise.
19517         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
19518         * tree.def (ABSU_EXPR): New.
19520 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
19522         PR middle-end/86095
19523         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
19524         documented as preserved for backward compatibility only.
19525         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
19527         PR rtl-optimization/86108
19528         * bb-reorder.c (create_forwarder_block): Renamed to ...
19529         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
19530         jump from new landing pad to the second part.
19531         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
19532         Adjust callers.
19534 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
19536         PR middle-end/85878
19537         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
19538         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
19539         Only call store_expr for halves if the mode is the same.
19541         PR middle-end/86123
19542         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
19543         Fix up comment formatting.
19545 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
19547         * typed-splay-tree.h (typed_splay_tree::remove): New function.
19548         (typed_splay_tree::closure,
19549         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
19550         (typed_splay_tree::typed_splay_tree,
19551         typed_splay_tree::operator =): Declared private.
19552         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
19553         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
19554         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
19555         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
19556         typed_splay_tree::splay_tree_splay,
19557         typed_splay_tree::splay_tree_foreach_helper,
19558         typed_splay_tree::splay_tree_insert,
19559         typed_splay_tree::splay_tree_remove,
19560         typed_splay_tree::splay_tree_lookup,
19561         typed_splay_tree::splay_tree_predecessor,
19562         typed_splay_tree::splay_tree_successor,
19563         typed_splay_tree::splay_tree_min,
19564         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
19565         (typed_splay_tree::root, typed_splay_tree::comp,
19566         typed_splay_tree::delete_key,
19567         typed_splay_tree::delete_value): New data members.
19568         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
19569         typed_splay_tree::remove.
19571 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
19573         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
19574         -mginv and -mno-ginv to the assembler.
19575         * config/mips/mips.opt (-mcrc): New option.
19576         (-mginv): Likewise.
19577         * doc/invoke.text (-mcrc): Document.
19578         (-mginv): Likewise.
19580 2018-06-15  Nick Clifton  <nickc@redhat.com>
19582         PR 84195
19583         * tree.c (escaped_string): New class.  Converts an unescaped
19584         string into its escaped equivalent.
19585         (warn_deprecated_use): Use the new class to convert the
19586         deprecation message, if present.
19587         (test_escaped_strings): New self test.
19588         (test_c_tests): Add test_escaped_strings.
19589         * doc/extend.texi (deprecated): Add a note that the
19590         deprecation message is affected by the -fmessage-length
19591         option, and that control characters will be escaped.
19592         (#pragma GCC error): Document this pragma.
19593         (#pragma GCC warning): Likewise.
19594         * doc/invoke.texi (-fmessage-length): Document this option's
19595         effect on the #warning and #error preprocessor directives and
19596         the deprecated attribute.
19598 2018-06-15  Richard Biener  <rguenther@suse.de>
19600         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
19601         here, also noting vector size used.
19602         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
19603         size used in MSG_OPTIMIZED_LOCATIONS dump.
19604         (pass_slp_vectorize::execute): Adjust.
19606 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
19608         PR target/85968
19609         * config/arc/arc.c (arc_return_address_register): Fix
19610         if-condition.
19612 2018-06-15  Richard Biener  <rguenther@suse.de>
19614         PR middle-end/86159
19615         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
19616         leave useless conversion stripping to force_gimple_operand_gsi.
19617         (gimplify_build2): Likewise.
19618         (gimplify_build1): Likewise.
19620 2018-06-15  Richard Biener  <rguenther@suse.de>
19622         PR middle-end/86076
19623         * tree-cfg.c (move_stmt_op): unshare invariant addresses
19624         before adjusting their block.
19626 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
19628         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
19629         multilibs for *-*-rtems*.
19630         * config/riscv/t-rtems: New file.
19632 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
19634         PR middle-end/86122
19635         * match.pd ((A +- CST1) +- CST2): Punt if last resort
19636         unsigned_type_for returns NULL.
19638         PR target/85945
19639         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
19640         subregs of multi-word pseudos unless the float mode has word size.
19642 2018-06-14  Richard Biener  <rguenther@suse.de>
19644         PR middle-end/86139
19645         * tree-vect-generic.c (build_word_mode_vector_type): Remove
19646         duplicate and harmful type_hash_canon.
19647         * tree.c (type_hash_canon): Assert we didn't find ourselves.
19649 2018-06-14  Richard Biener  <rguenther@suse.de>
19651         PR ipa/86124
19652         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
19653         NULL cgraph_node.
19655 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
19657         * config/rtems.h (STDINT_LONG32): Define.
19659 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
19660             Prachi Godbole  <prachi.godbole@imgtec.com>
19662         * config/mips/mips-cpus.def: Define P6600.
19663         * config/mips/mips-tables.opt: Regenerate.
19664         * config/mips/mips.c (mips_ucbranch_type): New enum.
19665         (mips_rtx_cost_data): Add support for P6600.
19666         (mips_issue_rate): Likewise.
19667         (mips_multipass_dfa_lookahead): Likewise.
19668         (mips_avoid_hazard): Likewise.
19669         (mips_reorg_process_insns): Likewise.
19670         (mips_classify_branch_p6600): New function.
19671         * config/mips/mips.h (TUNE_P6600): New define.
19672         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
19673         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
19674         * config/mips/mips.md: Include p6600.md.
19675         (processor): Add p6600.
19676         * config/mips/p6600.md: New file.
19677         * doc/invoke.texi: Add p6600 to supported architectures.
19679 2018-06-13  Martin Sebor  <msebor@redhat.com>
19681         PR tree-optimization/86114
19682         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
19683         of integer types.
19684         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
19686 2018-06-13  Richard Biener  <rguenther@suse.de>
19688         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
19689         Properly set vector type of the intermediate stmt.
19690         * tree-vect-stmts.c (vectorizable_operation): The destination
19691         var always has vectype_out type.
19693 2018-06-13  Jeff Law  <law@redhat.com>
19695         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
19696         integer 0 for argument to print_rtl_with_bb.
19697         (rl78_reorg): Likewise.
19699 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
19701         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
19702         from rtx to rtx_insn *.
19703         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
19704         "label".
19705         (add_sched_insns_for_speculation): Likewise for local "target",
19706         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19707         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
19708         from rtx_insn ** to rtx_code_label **.
19709         (reorg_emit_nops): Likewise.
19710         (c6x_reorg): Likewise for local "call_labels".
19711         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
19712         rtx to rtx_insn *.
19713         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
19714         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
19715         the loops over LABEL_REFS.
19716         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
19717         braf_label.
19718         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19719         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
19720         (split_branches): Strengthen local "olabel" from rtx to
19721         rtx_insn *, adding a safe_as_a cast.
19722         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
19723         to "rtx_insn *".
19724         (add_insn_after): Likewise for first two params.
19725         (add_insn_before): Likewise.
19726         (remove_insn): Likewise for param.
19727         (emit_pattern_before_noloc): Likewise for second and third params.
19728         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
19729         (emit_call_insn_before_noloc): Likewise.
19730         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
19731         to "rtx_insn *".
19732         (emit_barrier_before): Likewise.
19733         (emit_label_before): Strengthen "label" param from "rtx" to
19734         "rtx_code_label *".  Strengthen "before" param from "rtx" to
19735         "rtx_insn *".
19736         (emit_insn_after_1): Strengthen "after" param from "rtx" to
19737         "rtx_insn *".
19738         (emit_pattern_after_noloc): Likewise.
19739         (emit_insn_after_noloc): Likewise.
19740         (emit_jump_insn_after_noloc): Likewise.
19741         (emit_call_insn_after_noloc): Likewise.
19742         (emit_debug_insn_after_noloc): Likewise.
19743         (emit_barrier_after): Likewise.
19744         (emit_label_after): Likewise for both params.
19745         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
19746         "loc" param from "int" to "location_t".
19747         (emit_insn_after_setloc): Likewise.
19748         (emit_jump_insn_after_setloc): Likewise.
19749         (emit_call_insn_after_setloc): Likewise.
19750         (emit_debug_insn_after_setloc): Likewise.
19751         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
19752         "loc" param from "int" to "location_t".
19753         (emit_pattern_before): Convert NULL_RTX to NULL.
19754         (emit_insn_before_setloc): Convert "loc" param from "int" to
19755         "location_t".
19756         (emit_jump_insn_before_setloc): Likewise.
19757         (emit_call_insn_before_setloc): Likewise.
19758         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
19759         rtx_insn *.  Convert "loc" param from "int" to "location_t".
19760         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
19761         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
19762         Convert 3rd param from "int" to "location_t".
19763         (emit_barrier_before, emit_barrier_after, next_real_insn):
19764         Strengthen param from rtx to rtx_insn *.
19765         (emit_label_before): Strengthen 1st param from "rtx" to
19766         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
19767         "rtx_insn *".
19768         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
19769         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
19770         Strengthen 2nd param from "rtx" to "rtx_insn *".
19771         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
19772         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
19773         Likewise. Convert 3rd param from "int" to "location_t".
19774         (emit_label_after): Strengthen 1st param from "rtx" to
19775         "rtx_code_label *".
19776         (next_real_insn, remove_insn): Strengthen param from "rtx" to
19777         "rtx_insn *".
19778         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
19779         from "rtx" to "rtx_insn *".
19781 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
19783         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
19784         bodies streamed in with -Q.
19785         * dumpfile.c (dump_files): Add lto-stream-out dump file.
19786         * dumpfile.h (tree_dump_index): Add lto_stream_out.
19787         * gimple-streamer-out.c: Include gimple-pretty-print.h
19788         (output_bb): Dump stmts streamed.
19789         * lto-section-out.c: Include print-tree.h
19790         (lto_begin_section): Dump sections created.
19791         (lto_output_decl_index): Dump decl encoded.
19792         * lto-streamer-out.c: Include print-tree.h
19793         (create_output_block): Dump output block created.
19794         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
19795         (output_function): Dump function output.
19796         (output_constructor): Dump constructor streamed.
19797         (write_global_stream): Output indexes encoded.
19798         (produce_asm_for_decls): Dump streams encoded.
19799         * lto-streamer.c (streamer_dump_file): New global var.
19800         * lto-streamer.h (streamer_dump_file): Declare.
19801         * passes.c (ipa_write_summaries): Initialize streamer dump.
19802         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
19803         in.
19805 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
19807         PR target/86048
19808         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
19809         offsets for register save directives.  Emit a second batch of save
19810         directives, if need be, when the function accesses prior frames.
19812 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19814         * config/arc/fpu.md (fmasf4): Force operand to register.
19815         (fnmasf4): Likewise.
19817 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19819         * config/arc/arc-protos.h (arc_pad_return): Remove.
19820         * config/arc/arc.c (machine_function): Remove force_short_suffix
19821         and size_reason.
19822         (arc_print_operand): Adjust printing of '&'.
19823         (arc_verify_short): Remove conditional printing of short suffix.
19824         (arc_final_prescan_insn): Remove reference to size_reason.
19825         (pad_return): New function.
19826         (arc_reorg): Call pad_return.
19827         (arc_pad_return): Remove.
19828         (arc_init_machine_status): Remove reference to force_short_suffix.
19829         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
19830         (attr length): When attribute iscompact is true force to 2
19831         regardless; in the case of maybe check if we want to force the
19832         instruction to have 4 bytes length.
19833         (nopv): Change it to generate 4 byte long nop as well.
19834         (blockage): New pattern.
19835         (simple_return): Remove call to arc_pad_return.
19836         (p_return_i): Likewise.
19838 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19840         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
19842 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19844         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
19845         ARC cores.
19847 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19849         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
19850         for ARC700 and ARCv2.
19852 2018-06-13  Chenghua Xu  <paul.hua.gm@gmail.com>
19854         PR target/86076
19855         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
19856         operands[2] instead of operands[1].
19859 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
19861         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
19862         case, check whether the outer register overlaps an unallocatable
19863         register, not just whether it fits the required class.
19865 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
19867         * poly-int.h (can_div_trunc_p): Add new overload in which all values
19868         are poly_ints.
19869         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
19870         (memrefs_conflict_p): Likewise.
19871         (init_alias_analysis): Likewise.
19872         * cfgexpand.c (expand_debug_expr): Likewise.
19873         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
19874         * cse.c (fold_rtx): Likewise.
19875         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
19876         * expr.c (emit_block_move_hints): Likewise.
19877         (clear_storage_hints, push_block, emit_push_insn): Likewise.
19878         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
19879         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
19880         (emit_group_store): Likewise.
19881         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
19882         to read the PRE/POST_MODIFY increment.
19883         * calls.c (store_one_arg): Use strip_offset.
19884         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
19885         poly_int_rtx_p.
19886         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
19887         by a VEC_SELECT.
19888         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
19889         (simplify_binary_operation_1): Extend CONST_INT handling to
19890         poly_int_rtx_p.
19891         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
19892         than a HOST_WIDE_INT.
19893         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
19894         poly_int64.
19895         (adjust_mems, add_stores): Update accodingly.
19896         (vt_canonicalize_addr): Track polynomial offsets.
19897         (emit_note_insn_var_location): Likewise.
19898         (vt_add_function_parameter): Likewise.
19899         (vt_initialize): Likewise.
19901 2018-06-12  Jeff Law  <law@redhat.com>
19903         * config.gcc (alpha*-*-freebsd*): Remove.
19904         * config/alpha/freebsd.h: Remove.
19906 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
19908         PR other/69968
19909         * spellcheck-tree.c (levenshtein_distance): Rename to...
19910         (get_edit_distance): ...this, and update for underlying renaming.
19911         * spellcheck-tree.h (levenshtein_distance): Rename to...
19912         (get_edit_distance): ...this.
19913         * spellcheck.c (levenshtein_distance): Rename to...
19914         (get_edit_distance): ...this.  Convert from Levenshtein distance
19915         to Damerau-Levenshtein distance by supporting transpositions of
19916         adjacent characters.  Rename "v1" to "v_next" and "v0" to
19917         "v_one_ago".
19918         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
19919         (selftest::test_edit_distance_unit_test_oneway): ...this, and
19920         update for underlying renaming.
19921         (selftest::levenshtein_distance_unit_test): Rename to...
19922         (selftest::test_get_edit_distance_unit): ...this, and update for
19923         underlying renaming.
19924         (selftest::test_find_closest_string): Add example from PR 69968
19925         where transposition helps
19926         (selftest::test_metric_conditions): Update for renaming.
19927         (selftest::test_metric_conditions): Likewise.
19928         (selftest::spellcheck_c_tests): Likewise.
19929         * spellcheck.h (levenshtein_distance): Rename both overloads to...
19930         (get_edit_distance): ...this.
19931         (best_match::consider): Update for renaming.
19933 2018-06-12  Martin Sebor  <msebor@redhat.com>
19935         PR tree-optimization/85259
19936         * builtins.c (compute_objsize): Handle constant offsets.
19937         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
19938         true iff a warning has been issued.
19939         * gimple.h (gimple_nonartificial_location): New function.
19940         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
19941         gimple_nonartificial_location and handle -Wno-system-headers.
19942         (handle_builtin_stxncpy): Same.
19944 2018-06-12  Martin Sebor  <msebor@redhat.com>
19946         PR c/85931
19947         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
19949 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
19951         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
19952         BUILTIN_VEC_XST entries for pointer to double and long long.
19954 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
19956         PR target/85990
19957         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19958         Update comments.
19959         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19960         Likewise.
19962 2018-06-12  Martin Liska  <mliska@suse.cz>
19964         * doc/options.texi: Document IntegerRange.
19966 2018-06-12  Martin Liska  <mliska@suse.cz>
19968         * config/i386/i386.opt: Make MPX-related options as Deprecated.
19969         * opt-functions.awk: Handle Deprecated flag.
19970         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
19971         and report error.
19972         (read_cmdline_option): Report warning for a deprecated option.
19973         * opts.h (struct cl_option): Add new field cl_deprecated.
19974         (CL_ERR_DEPRECATED): New.
19976 2018-06-12  Martin Liska  <mliska@suse.cz>
19978         * doc/options.texi: Document Deprecated option flag.
19980 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19982         * config/arc/arc-arch.h (arc_extras): New enum.
19983         (arc_cpu_t):Add field extra.
19984         (arc_cpu_types): Consider the extras.
19985         * config/arc/arc-cpus.def: Add extras info.
19986         * config/arc/arc-opts.h (processor_type): Consider extra field.
19987         * config/arc/arc.c (arc_override_options): Handle extra field.
19989 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19991         * config/arc/arc-arch.h: Update ARC_OPTX macro.
19992         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
19993         field.
19994         * config/arc/arc.c (arc_init): Update pic warning.
19995         (irq_range): Update irq range parsing warnings.
19996         (arc_override_options): Update various warning messages.
19997         (arc_handle_aux_attribute): Likewise.
19999 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
20001         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
20003 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20005         * doc/sourcebuild.texi: Document usage of line number 0 in verify
20006         compiler messages directives.
20008 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
20010         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
20011         * config/mips/mips-tables.opt: Regenerate.
20012         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
20013         mips64r6.
20014         * doc/invoke.texi: Document -march=i6500.
20016 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
20018         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
20019         (i6400_gpmul): Add cpu_unit.
20020         (i6400_gpdiv): Likewise.
20021         (i6400_msa_add_d): Update reservations.
20022         (i6400_msa_int_add) Likewise.
20023         (i6400_msa_short_logic3) Likewise.
20024         (i6400_msa_short_logic2) Likewise.
20025         (i6400_msa_short_logic) Likewise.
20026         (i6400_msa_move) Likewise.
20027         (i6400_msa_cmp) Likewise.
20028         (i6400_msa_short_float2) Likewise.
20029         (i6400_msa_div_d) Likewise.
20030         (i6400_msa_long_logic1) Likewise.
20031         (i6400_msa_long_logic2) Likewise.
20032         (i6400_msa_mult) Likewise.
20033         (i6400_msa_long_float2) Likewise.
20034         (i6400_msa_long_float4) Likewise.
20035         (i6400_msa_long_float5) Likewise.
20036         (i6400_msa_long_float8) Likewise.
20037         (i6400_fpu_fadd): Include frint type.
20038         (i6400_fpu_store): New define_insn_reservation.
20039         (i6400_fpu_load): Likewise.
20040         (i6400_fpu_move): Likewise.
20041         (i6400_fpu_fcmp): Likewise.
20042         (i6400_fpu_fmadd): Likewise.
20043         (i6400_int_mult): Include imul3nc type and update reservation.
20044         (i6400_int_div): Include idiv3 type and update reservation.
20045         (i6400_int_load): Update to check type not move_type.
20046         (i6400_int_store): Likewise.
20047         (i6400_int_prefetch): Set zero latency.
20049 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
20051         * gcc.c: Document new %@{...} sequence.
20052         (LINK_COMMAND_SPEC): Use it for the -L switches.
20053         (cpp_unique_options): Use it for the -I switches.
20054         (at_file_argbuf): New global variable.
20055         (in_at_file): Likewise.
20056         (alloc_args): Create at_file_argbuf.
20057         (clear_args): Truncate at_file_argbuf.
20058         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
20059         (open_at_file): New function.
20060         (close_at_file): Likewise.
20061         (create_at_file): Delete.
20062         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
20063         <'o'>: Likewise.
20064         <'@'>: New case.
20065         (validate_switches_from_spec): Deal with %@{...} sequence.
20066         (validate_switches): Likewise.
20067         (driver::finalize): Call clear_args.
20069 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
20071         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
20073 2018-06-11  Martin Sebor  <msebor@redhat.com>
20075         * doc/invoke.texi (-Wall): List -Wc++17-compat.
20076         (Wno-class-memaccess): Add @opindex.
20077         (Wno-templates, Wno-multiple-inheritance): Same.
20078         (Wno-virtual-inheritance, Wno-namespaces): Same.
20079         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
20080         (Wno-format-overflow, Wno-format-truncation): Same.
20081         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
20082         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
20083         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
20084         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
20085         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
20086         (Wno-misspelled-isr): Same.
20088 2018-06-11  Martin Sebor  <msebor@redhat.com>
20090         * PR tree-optimization/86083
20091         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
20093 2018-06-11  Zhouyi Zhou  <zhouzhouyi@gmail.com>
20095         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
20097 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
20099         PR target/85755
20100         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
20101         on the correct operand.
20102         (*movdi_internal64): Ditto.
20104 2018-06-11  Martin Liska  <mliska@suse.cz>
20106         PR tree-optimization/86089
20107         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
20109 2018-06-11  Julia Koval  <julia.koval@intel.com>
20111         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
20112         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
20113         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
20115 2018-06-11  Olivier Hainque  <hainque@adacore.com>
20117         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
20118         for Ada with strict dwarf2.
20120 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
20122         PR target/85755
20123         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
20124         addresses.
20126 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
20128         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
20130 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
20132         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
20133         TARGET_ELF.
20135 2018-06-08  Martin Liska  <mliska@suse.cz>
20137         * tree-cfg.h (debug_function): Fix argument type to match
20138         implementation.
20140 2018-06-08  Martin Liska  <mliska@suse.cz>
20142         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
20143         Remove usage of MPX-related (and removed) fields.
20144         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
20146 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
20148         * cfg.c (debug): Use TDF_NONE rather than 0.
20149         * cfghooks.c (debug): Likewise.
20150         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
20151         (struct dump_option_value_info): Convert to...
20152         (struct kv_pair): ...this template type.
20153         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
20154         rather than 0.
20155         (optinfo_verbosity_options): Likewise.
20156         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
20157         OPTGROUP_NONE.
20158         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
20159         than int for "optgroup_flags" param.
20160         (dump_generic_expr_loc): Use dump_flags_t rather than int for
20161         "dump_kind" param.
20162         (dump_dec): Likewise.
20163         (dump_finish): Use TDF_NONE rather than 0.
20164         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
20165         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
20166         than 0.  Update for change to option_ptr.
20167         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
20168         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
20169         0.  Update for changes to optinfo_verbosity_options and
20170         optgroup_options.
20171         (opt_info_switch_p): Convert optgroup_flags from int to
20172         optgroup_flags_t.
20173         (dump_basic_block): Use dump_flags_t rather than int
20174         for "dump_kind" param.
20175         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
20176         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
20177         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
20178         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
20179         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
20180         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
20181         TDF_NONE): Convert from macros to...
20182         (enum dump_flag): ...this new enum.
20183         (dump_flags_t): Update to use enum.
20184         (operator|, operator&, operator~, operator|=, operator&=):
20185         Implement for dump_flags_t.
20186         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
20187         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
20188         Convert from macros to...
20189         (enum optgroup_flag): ...this new enum.
20190         (optgroup_flags_t): New typedef.
20191         (operator|, operator|=): Implement for optgroup_flags_t.
20192         (struct dump_file_info): Convert field "alt_flags" to
20193         dump_flags_t.  Convert field "optgroup_flags" to
20194         optgroup_flags_t.
20195         (dump_basic_block): Use dump_flags_t rather than int for param.
20196         (dump_generic_expr_loc): Likewise.
20197         (dump_dec): Likewise.
20198         (dump_register): Convert param "optgroup_flags" to
20199         optgroup_flags_t.
20200         (opt_info_enable_passes): Likewise.
20201         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
20202         than 0.
20203         * gimple-pretty-print.c (debug): Likewise.
20204         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
20205         (merged_store_group::apply_stores): Likewise.
20206         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
20207         * gimple.c (verify_gimple_pp): Likewise.
20208         * graphite-poly.c (print_pbb_body): Likewise.
20209         * passes.c (pass_manager::register_one_dump_file): Convert
20210         local "optgroup_flags" to optgroup_flags_t.
20211         * print-tree.c (print_node): Use TDF_NONE rather than 0.
20212         (debug): Likewise.
20213         (debug_body): Likewise.
20214         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
20215         to optgroup_flags_t.
20216         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
20217         than 0.
20218         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
20219         (convert_mult_to_fma): Likewise.
20220         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
20221         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
20222         * tree-vect-data-refs.c (dump_lower_bound): Convert param
20223         "dump_kind" to dump_flags_t.
20225 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
20227         * config/rs6000/rs6000.c (min, max): Delete.
20229 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
20231         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
20232         -mabi=spe and -mabi=no-spe.
20234 2018-06-08  Martin Liska  <mliska@suse.cz>
20236         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
20237         where we expect an existing summary.
20239 2018-06-08  Martin Liska  <mliska@suse.cz>
20241         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
20242         * ipa-inline.h (estimate_edge_growth): Likewise.
20244 2018-06-08  Martin Liska  <mliska@suse.cz>
20246         * cgraph.c (function_version_hasher::hash): Use
20247         cgraph_node::get_uid ().
20248         (function_version_hasher::equal):
20249         * cgraph.h (cgraph_node::get_uid): New method.
20250         * ipa-inline.c (update_caller_keys): Use
20251         cgraph_node::get_uid ().
20252         (update_callee_keys): Likewise.
20253         * ipa-utils.c (searchc): Likewise.
20254         (ipa_reduced_postorder): Likewise.
20255         * lto-cgraph.c (input_node): Likewise.
20256         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
20257         * symbol-summary.h (symtab_insertion): Likewise.
20258         (symtab_removal): Likewise.
20259         (symtab_duplication): Likewise.
20260         * tree-pretty-print.c (dump_function_header): Likewise.
20261         * tree-sra.c (convert_callers_for_node): Likewise.
20263 2018-06-08  Martin Liska  <mliska@suse.cz>
20265         * cgraph.c (symbol_table::create_edge): Always assign a new
20266         unique number.
20267         (symbol_table::free_edge): Do not recycle numbers.
20268         * cgraph.h (cgraph_edge::get): New method.
20269         * symbol-summary.h (symtab_removal): Use it.
20270         (symtab_duplication): Likewise.
20271         (call_summary::hashable_uid): Remove.
20273 2018-06-08  Martin Liska  <mliska@suse.cz>
20275         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
20276         (initialize_growth_caches): Remove.
20277         (free_growth_caches): Likewise.
20278         (do_estimate_edge_time): Use edge_growth_cache.
20279         (do_estimate_edge_size): Likewise.
20280         (do_estimate_edge_hints): Likewise.
20281         * ipa-inline.c (reset_edge_caches): Likewise.
20282         (recursive_inlining): Likewise.
20283         (inline_small_functions): Likewise.
20284         * ipa-inline.h (initialize_growth_caches): Remove.
20285         (estimate_edge_size): Likewise.
20286         (estimate_edge_time): Likewise.
20287         (estimate_edge_hints): Likewise.
20288         (reset_edge_growth_cache): Likewise.
20289         * symbol-summary.h (call_summary::remove): New method.
20291 2018-06-08  Martin Liska  <mliska@suse.cz>
20293         * ipa-cp.c (class edge_clone_summary): New summary.
20294         (grow_edge_clone_vectors): Remove.
20295         (ipcp_edge_duplication_hook): Remove.
20296         (class edge_clone_summary_t): New call_summary class.
20297         (ipcp_edge_removal_hook): Remove.
20298         (edge_clone_summary_t::duplicate): New function.
20299         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
20300         (create_specialized_node): Likewise.
20301         (ipcp_driver): Initialize edge_clone_summaries and do not
20302         register hooks.
20304 2018-06-08  Martin Liska  <mliska@suse.cz>
20306         * symbol-summary.h (get): New function.
20307         (call_summary::m_initialize_when_cloning): New class member.
20309 2018-06-08  Martin Liska  <mliska@suse.cz>
20311         * cgraph.c (cgraph_node::remove): Do not recycle uid.
20312         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
20313         (symbol_table::allocate_cgraph_symbol): Do not set uid.
20314         * passes.c (uid_hash_t): Record removed_nodes by their uids.
20315         (remove_cgraph_node_from_order): Use the removed_nodes set.
20316         (do_per_function_toporder): Likwise.
20317         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
20318         instead of summary_uid.
20319         (symtab_removal): Likewise.
20320         (symtab_duplication): Likewise.
20322 2018-06-08  Martin Liska  <mliska@suse.cz>
20324         * ipa-cp.c (ipcp_store_bits_results): Use
20325         ipcp_transformation_sum.
20326         (ipcp_store_vr_results): Likewise.
20327         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
20328         to ...
20329         (ipcp_transformation_initialize): ... this.
20330         (ipa_set_node_agg_value_chain):
20331         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
20332         (write_ipcp_transformation_info): Likewise.
20333         (read_ipcp_transformation_info): Likewise.
20334         (ipcp_update_bits): Likewise.
20335         (ipcp_update_vr): Likewise.
20336         (ipcp_transform_function): Likewise.
20337         * ipa-prop.h: Rename ipcp_transformation_summary to
20338         ipcp_transformation.
20339         (class ipcp_transformation_t): New function summary.
20340         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
20341         (ipa_get_agg_replacements_for_node): Likewise.
20343 2018-06-08  Martin Liska  <mliska@suse.cz>
20345         * ipa-pure-const.c (struct funct_state_d): Do it class instead
20346         of struct.
20347         (class funct_state_summary_t): New function_summary class.
20348         (has_function_state): Remove.
20349         (get_function_state): Likewise.
20350         (set_function_state): Likewise.
20351         (add_new_function): Likewise.
20352         (funct_state_summary_t::insert): New function.
20353         (duplicate_node_data): Remove.
20354         (remove_node_data): Remove.
20355         (funct_state_summary_t::duplicate): New function.
20356         (register_hooks): Create new funct_state_summaries.
20357         (pure_const_generate_summary): Use it.
20358         (pure_const_write_summary): Likewise.
20359         (pure_const_read_summary): Likewise.
20360         (propagate_pure_const): Likewise.
20361         (propagate_nothrow): Likewise.
20362         (dump_malloc_lattice): Likewise.
20363         (propagate_malloc): Likewise.
20364         (execute): Do not register hooks, just remove summary
20365         instead.
20366         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
20367         constructor.
20369 2018-06-08  Martin Liska  <mliska@suse.cz>
20371         * ipa-reference.c (remove_node_data): Remove.
20372         (duplicate_node_data): Likewise.
20373         (class ipa_ref_var_info_summary_t): New class.
20374         (class ipa_ref_opt_summary_t): Likewise.
20375         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
20376         (get_reference_optimization_summary): Use
20377         ipa_ref_opt_sum_summaries.
20378         (set_reference_vars_info): Remove.
20379         (set_reference_optimization_summary): Likewise.
20380         (ipa_init): Create summaries.
20381         (init_function_info): Use function summary.
20382         (ipa_ref_opt_summary_t::duplicate): New function.
20383         (ipa_ref_opt_summary_t::remove): New function.
20384         (get_read_write_all_from_node): Fix GNU coding style.
20385         (propagate): Use function summary.
20386         (write_node_summary_p): Fix GNU coding style.
20387         (stream_out_bitmap): Likewise.
20388         (ipa_reference_read_optimization_summary): Use function summary.
20389         (ipa_reference_c_finalize): Do not release hooks.
20391 2018-06-08  Martin Liska  <mliska@suse.cz>
20393         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
20394         (analyze_function_body): Extract multiple calls of get_create.
20395         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20396         * ipa-inline.c (recursive_inlining): Use ::get method.
20397         * ipa-inline.h (estimate_edge_growth): Likewise.
20399 2018-06-08  Martin Liska  <mliska@suse.cz>
20401         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
20402         HSA_INVALID.
20403         (hsa_function_summary::hsa_function_summary): Use the new enum
20404         value.
20405         (hsa_gpu_implementation_p): Use hsa_summaries::get.
20406         * hsa-gen.c (hsa_get_host_function): Likewise.
20407         (get_brig_function_name): Likewise.
20408         * ipa-hsa.c (process_hsa_functions): Likewise.
20409         (ipa_hsa_write_summary): Likewise.
20410         * symbol-summary.h (symtab_duplication): Use ::get function/
20411         (get): New function.
20413 2018-06-08  Martin Liska  <mliska@suse.cz>
20415         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
20416         of get.
20417         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
20418         (hsa_register_kernel): Likewise.
20419         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
20420         * hsa-gen.c (hsa_get_host_function): Likewise.
20421         (get_brig_function_name): Likewise.
20422         (generate_hsa): Likewise.
20423         (pass_gen_hsail::execute): Likewise.
20424         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
20425         (devirtualization_time_bonus): Likewise.
20426         (ipcp_propagate_stage): Likewise.
20427         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
20428         (edge_set_predicate): Likewise.
20429         (evaluate_conditions_for_known_args): Likewise.
20430         (evaluate_properties_for_edge): Likewise.
20431         (ipa_fn_summary::reset): Likewise.
20432         (ipa_fn_summary_t::duplicate): Likewise.
20433         (dump_ipa_call_summary): Likewise.
20434         (ipa_dump_fn_summary): Likewise.
20435         (analyze_function_body): Likewise.
20436         (compute_fn_summary): Likewise.
20437         (estimate_edge_devirt_benefit): Likewise.
20438         (estimate_edge_size_and_time): Likewise.
20439         (estimate_calls_size_and_time): Likewise.
20440         (estimate_node_size_and_time): Likewise.
20441         (inline_update_callee_summaries): Likewise.
20442         (remap_edge_change_prob): Likewise.
20443         (remap_edge_summaries): Likewise.
20444         (ipa_merge_fn_summary_after_inlining): Likewise.
20445         (ipa_update_overall_fn_summary): Likewise.
20446         (read_ipa_call_summary): Likewise.
20447         (inline_read_section): Likewise.
20448         (write_ipa_call_summary): Likewise.
20449         (ipa_fn_summary_write): Likewise.
20450         (ipa_free_fn_summary): Likewise.
20451         * ipa-hsa.c (process_hsa_functions): Likewise.
20452         (ipa_hsa_write_summary): Likewise.
20453         (ipa_hsa_read_section): Likewise.
20454         * ipa-icf.c (sem_function::merge): Likewise.
20455         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20456         (do_estimate_edge_time): Likewise.
20457         (estimate_size_after_inlining): Likewise.
20458         (estimate_growth): Likewise.
20459         (growth_likely_positive): Likewise.
20460         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
20461         (inline_call): Likewise.
20462         * ipa-inline.c (caller_growth_limits): Likewise.
20463         (can_inline_edge_p): Likewise.
20464         (can_inline_edge_by_limits_p): Likewise.
20465         (compute_uninlined_call_time): Likewise.
20466         (compute_inlined_call_time): Likewise.
20467         (want_inline_small_function_p): Likewise.
20468         (edge_badness): Likewise.
20469         (update_caller_keys): Likewise.
20470         (update_callee_keys): Likewise.
20471         (recursive_inlining): Likewise.
20472         (inline_small_functions): Likewise.
20473         (inline_to_all_callers_1): Likewise.
20474         (dump_overall_stats): Likewise.
20475         (early_inline_small_functions): Likewise.
20476         (early_inliner): Likewise.
20477         * ipa-inline.h (estimate_edge_growth): Likewise.
20478         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
20479         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
20480         * ipa-prop.h (IPA_NODE_REF): Likewise.
20481         (IPA_EDGE_REF): Likewise.
20482         * ipa-pure-const.c (malloc_candidate_p): Likewise.
20483         (propagate_malloc): Likewise.
20484         * ipa-split.c (execute_split_functions): Likewise.
20485         * symbol-summary.h: Rename get to get_create.
20486         (get): Likewise.
20487         (get_create): Likewise.
20488         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
20490 2018-06-08  Martin Liska  <mliska@suse.cz>
20492         * symbol-summary.h (release): Move definition out of class
20493         declaration.
20494         (symtab_removal): Likewise.
20495         (symtab_duplication): Likewise.
20497 2018-06-08  Martin Liska  <mliska@suse.cz>
20499         * symbol-summary.h (function_summary): Move constructor
20500         implementation out of class declaration.
20501         (release): Likewise.
20502         (symtab_insertion): Likewise.
20503         (symtab_removal): Likewise.
20504         (symtab_duplication): Likewise.
20505         (get): Likewise.
20507 2018-06-08  Martin Liska  <mliska@suse.cz>
20509         * Makefile.in: Remove support for MPX (macros, related functions,
20510         fields in cgraph_node, ...).
20511         * builtin-types.def (BT_BND): Likewise.
20512         (BT_FN_BND_CONST_PTR): Likewise.
20513         (BT_FN_CONST_PTR_BND): Likewise.
20514         (BT_FN_VOID_PTR_BND): Likewise.
20515         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
20516         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
20517         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
20518         (expand_builtin_mempcpy_with_bounds): Likewise.
20519         (expand_builtin_memset_with_bounds): Likewise.
20520         (expand_builtin_memset_args): Likewise.
20521         (std_expand_builtin_va_start): Likewise.
20522         (expand_builtin): Likewise.
20523         (expand_builtin_with_bounds): Likewise.
20524         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
20525         (DEF_LIB_BUILTIN_CHKP): Likewise.
20526         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
20527         (DEF_CHKP_BUILTIN): Likewise.
20528         (BUILT_IN_MEMCPY): Likewise.
20529         (BUILT_IN_MEMMOVE): Likewise.
20530         (BUILT_IN_MEMPCPY): Likewise.
20531         (BUILT_IN_MEMSET): Likewise.
20532         (BUILT_IN_STPCPY): Likewise.
20533         (BUILT_IN_STRCAT): Likewise.
20534         (BUILT_IN_STRCHR): Likewise.
20535         (BUILT_IN_STRCPY): Likewise.
20536         (BUILT_IN_STRLEN): Likewise.
20537         (BUILT_IN_MEMCPY_CHK): Likewise.
20538         (BUILT_IN_MEMMOVE_CHK): Likewise.
20539         (BUILT_IN_MEMPCPY_CHK): Likewise.
20540         (BUILT_IN_MEMSET_CHK): Likewise.
20541         (BUILT_IN_STPCPY_CHK): Likewise.
20542         (BUILT_IN_STRCAT_CHK): Likewise.
20543         (BUILT_IN_STRCPY_CHK): Likewise.
20544         * calls.c (store_bounds): Likewise.
20545         (emit_call_1): Likewise.
20546         (special_function_p): Likewise.
20547         (maybe_warn_nonstring_arg): Likewise.
20548         (initialize_argument_information): Likewise.
20549         (finalize_must_preallocate): Likewise.
20550         (compute_argument_addresses): Likewise.
20551         (expand_call): Likewise.
20552         * cfgexpand.c (expand_call_stmt): Likewise.
20553         (expand_return): Likewise.
20554         (expand_gimple_stmt_1): Likewise.
20555         (pass_expand::execute): Likewise.
20556         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
20557         (cgraph_node::remove): Likewise.
20558         (cgraph_node::dump): Likewise.
20559         (cgraph_node::verify_node): Likewise.
20560         * cgraph.h (chkp_function_instrumented_p): Likewise.
20561         (symtab_node::get_alias_target): Likewise.
20562         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
20563         (cgraph_local_p): Likewise.
20564         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
20565         (cgraph_edge::rebuild_references): Likewise.
20566         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
20567         (walk_polymorphic_call_targets): Likewise.
20568         (cgraph_node::expand_thunk): Likewise.
20569         (symbol_table::output_weakrefs): Likewise.
20570         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
20571         (ix86_handle_option): Likewise.
20572         * config/i386/constraints.md: Likewise.
20573         * config/i386/i386-builtin-types.def (BND): Likewise.
20574         (VOID): Likewise.
20575         (PVOID): Likewise.
20576         (ULONG): Likewise.
20577         * config/i386/i386-builtin.def (BDESC_END): Likewise.
20578         (BDESC_FIRST): Likewise.
20579         (BDESC): Likewise.
20580         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
20581         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
20582         * config/i386/i386.c (enum reg_class): Likewise.
20583         (ix86_target_string): Likewise.
20584         (ix86_option_override_internal): Likewise.
20585         (ix86_conditional_register_usage): Likewise.
20586         (ix86_valid_target_attribute_inner_p): Likewise.
20587         (ix86_set_indirect_branch_type): Likewise.
20588         (ix86_set_current_function): Likewise.
20589         (ix86_function_arg_regno_p): Likewise.
20590         (init_cumulative_args): Likewise.
20591         (ix86_function_arg_advance): Likewise.
20592         (ix86_function_arg): Likewise.
20593         (ix86_pass_by_reference): Likewise.
20594         (ix86_function_value_regno_p): Likewise.
20595         (ix86_function_value_1): Likewise.
20596         (ix86_function_value_bounds): Likewise.
20597         (ix86_return_in_memory): Likewise.
20598         (ix86_setup_incoming_vararg_bounds): Likewise.
20599         (ix86_va_start): Likewise.
20600         (indirect_thunk_need_prefix): Likewise.
20601         (print_reg): Likewise.
20602         (ix86_print_operand): Likewise.
20603         (ix86_expand_call): Likewise.
20604         (ix86_output_function_return): Likewise.
20605         (reg_encoded_number): Likewise.
20606         (BDESC_VERIFYS): Likewise.
20607         (ix86_init_mpx_builtins): Likewise.
20608         (ix86_init_builtins): Likewise.
20609         (ix86_emit_cmove): Likewise.
20610         (ix86_emit_move_max): Likewise.
20611         (ix86_expand_builtin): Likewise.
20612         (ix86_builtin_mpx_function): Likewise.
20613         (ix86_get_arg_address_for_bt): Likewise.
20614         (ix86_load_bounds): Likewise.
20615         (ix86_store_bounds): Likewise.
20616         (ix86_load_returned_bounds): Likewise.
20617         (ix86_store_returned_bounds): Likewise.
20618         (ix86_class_likely_spilled_p): Likewise.
20619         (ix86_hard_regno_mode_ok): Likewise.
20620         (x86_order_regs_for_local_alloc): Likewise.
20621         (ix86_mitigate_rop): Likewise.
20622         (ix86_bnd_prefixed_insn_p): Likewise.
20623         (ix86_mpx_bound_mode): Likewise.
20624         (ix86_make_bounds_constant): Likewise.
20625         (ix86_initialize_bounds): Likewise.
20626         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
20627         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
20628         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
20629         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
20630         (TARGET_CHKP_BOUND_MODE): Likewise.
20631         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
20632         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
20633         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
20634         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
20635         * config/i386/i386.h (TARGET_MPX): Likewise.
20636         (TARGET_MPX_P): Likewise.
20637         (VALID_BND_REG_MODE): Likewise.
20638         (FIRST_BND_REG): Likewise.
20639         (LAST_BND_REG): Likewise.
20640         (enum reg_class): Likewise.
20641         (BND_REG_P): Likewise.
20642         (BND_REGNO_P): Likewise.
20643         (BNDmode): Likewise.
20644         (ADJUST_INSN_LENGTH): Likewise.
20645         * config/i386/i386.md: Likewise.
20646         * config/i386/i386.opt: Likewise.
20647         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
20648         (defined): Likewise.
20649         (LINK_MPX): Likewise.
20650         (MPX_SPEC): Likewise.
20651         (LIBMPX_SPEC): Likewise.
20652         (LIBMPXWRAPPERS_SPEC): Likewise.
20653         (CHKP_SPEC): Likewise.
20654         * config/i386/predicates.md: Likewise.
20655         * dbxout.c (dbxout_type): Likewise.
20656         * doc/extend.texi: Likewise.
20657         * doc/invoke.texi: Likewise.
20658         * doc/md.texi: Likewise.
20659         * doc/tm.texi: Likewise.
20660         * doc/tm.texi.in: Likewise.
20661         * dwarf2out.c (is_base_type): Likewise.
20662         (gen_formal_types_die): Likewise.
20663         (gen_subprogram_die): Likewise.
20664         (gen_type_die_with_usage): Likewise.
20665         (gen_decl_die): Likewise.
20666         (dwarf2out_late_global_decl): Likewise.
20667         * expr.c (expand_assignment): Likewise.
20668         (emit_storent_insn): Likewise.
20669         (store_expr_with_bounds): Likewise.
20670         (store_expr): Likewise.
20671         (expand_expr_real_1): Likewise.
20672         * expr.h (store_expr_with_bounds): Likewise.
20673         * function.c (use_register_for_decl): Likewise.
20674         (struct bounds_parm_data): Likewise.
20675         (assign_parms_augmented_arg_list): Likewise.
20676         (assign_parm_find_entry_rtl): Likewise.
20677         (assign_parm_is_stack_parm): Likewise.
20678         (assign_parm_load_bounds): Likewise.
20679         (assign_bounds): Likewise.
20680         (assign_parms): Likewise.
20681         (expand_function_start): Likewise.
20682         * gcc.c (CHKP_SPEC): Likewise.
20683         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
20684         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
20685         (wrestrict_dom_walker::check_call): Likewise.
20686         * gimple.c (gimple_build_call_from_tree): Likewise.
20687         * gimple.h (enum gf_mask): Likewise.
20688         (gimple_call_with_bounds_p): Likewise.
20689         (gimple_call_set_with_bounds): Likewise.
20690         * gimplify.c (gimplify_init_constructor): Likewise.
20691         * ipa-cp.c (initialize_node_lattices): Likewise.
20692         (propagate_constants_across_call): Likewise.
20693         (find_more_scalar_values_for_callers_subset): Likewise.
20694         * ipa-hsa.c (process_hsa_functions): Likewise.
20695         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
20696         * ipa-icf.c (sem_function::merge): Likewise.
20697         * ipa-inline.c (early_inliner): Likewise.
20698         * ipa-pure-const.c (warn_function_noreturn): Likewise.
20699         (warn_function_cold): Likewise.
20700         (propagate_pure_const): Likewise.
20701         * ipa-ref.h (enum GTY): Likewise.
20702         * ipa-split.c (find_retbnd): Likewise.
20703         (consider_split): Likewise.
20704         (split_function): Likewise.
20705         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
20706         * ipa.c (walk_polymorphic_call_targets): Likewise.
20707         (symbol_table::remove_unreachable_nodes): Likewise.
20708         (process_references): Likewise.
20709         (cgraph_build_static_cdtor_1): Likewise.
20710         * lto-cgraph.c (lto_output_node): Likewise.
20711         (output_refs): Likewise.
20712         (compute_ltrans_boundary): Likewise.
20713         (input_overwrite_node): Likewise.
20714         (input_node): Likewise.
20715         (input_cgraph_1): Likewise.
20716         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
20717         * passes.c (pass_manager::execute_early_local_passes): Likewise.
20718         (class pass_chkp_instrumentation_passes): Likewise.
20719         (make_pass_chkp_instrumentation_passes): Likewise.
20720         * passes.def: Likewise.
20721         * rtl.h (struct GTY): Likewise.
20722         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
20723         * stor-layout.c (layout_type): Likewise.
20724         * symtab.c: Likewise.
20725         * target.def: Likewise.
20726         * targhooks.c (default_chkp_bound_type): Likewise.
20727         (default_chkp_bound_mode): Likewise.
20728         (default_builtin_chkp_function): Likewise.
20729         (default_chkp_function_value_bounds): Likewise.
20730         (default_chkp_make_bounds_constant): Likewise.
20731         (default_chkp_initialize_bounds): Likewise.
20732         * targhooks.h (default_chkp_bound_type): Likewise.
20733         (default_chkp_bound_mode): Likewise.
20734         (default_builtin_chkp_function): Likewise.
20735         (default_chkp_function_value_bounds): Likewise.
20736         (default_chkp_make_bounds_constant): Likewise.
20737         (default_chkp_initialize_bounds): Likewise.
20738         * toplev.c (compile_file): Likewise.
20739         (process_options): Likewise.
20740         * tree-core.h (DEF_BUILTIN): Likewise.
20741         (DEF_BUILTIN_CHKP): Likewise.
20742         * tree-inline.c (declare_return_variable): Likewise.
20743         (remap_gimple_stmt): Likewise.
20744         (copy_bb): Likewise.
20745         (initialize_inlined_parameters): Likewise.
20746         (expand_call_inline): Likewise.
20747         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
20748         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
20749         (make_pass_ipa_chkp_produce_thunks): Likewise.
20750         (make_pass_chkp): Likewise.
20751         (make_pass_chkp_opt): Likewise.
20752         (make_pass_chkp_instrumentation_passes): Likewise.
20753         * tree-pretty-print.c (dump_generic_node): Likewise.
20754         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
20755         * tree-ssa-dce.c (propagate_necessity): Likewise.
20756         (eliminate_unnecessary_stmts): Likewise.
20757         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
20758         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
20759         * tree-ssa-sccvn.h: Likewise.
20760         * tree-ssa-strlen.c (get_string_length): Likewise.
20761         (valid_builtin_call): Likewise.
20762         (adjust_last_stmt): Likewise.
20763         (handle_builtin_strchr): Likewise.
20764         (handle_builtin_strcpy): Likewise.
20765         (handle_builtin_stxncpy): Likewise.
20766         (handle_builtin_memcpy): Likewise.
20767         (handle_builtin_strcat): Likewise.
20768         (strlen_check_and_optimize_stmt): Likewise.
20769         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
20770         * tree-streamer-in.c: Likewise.
20771         * tree-streamer.c (record_common_node): Likewise.
20772         * tree.c (tree_code_size): Likewise.
20773         (wide_int_to_tree_1): Likewise.
20774         (type_contains_placeholder_1): Likewise.
20775         (build_common_tree_nodes): Likewise.
20776         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
20777         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
20778         (POINTER_BOUNDS_P): Likewise.
20779         (BOUNDED_TYPE_P): Likewise.
20780         (BOUNDED_P): Likewise.
20781         (CALL_WITH_BOUNDS_P): Likewise.
20782         (pointer_bounds_type_node): Likewise.
20783         * value-prof.c (gimple_ic): Likewise.
20784         * var-tracking.c (vt_add_function_parameters): Likewise.
20785         * varasm.c (make_decl_rtl): Likewise.
20786         (assemble_start_function): Likewise.
20787         (output_constant): Likewise.
20788         (maybe_assemble_visibility): Likewise.
20789         * varpool.c (ctor_for_folding): Likewise.
20790         * chkp-builtins.def: Remove.
20791         * ipa-chkp.c: Remove.
20792         * ipa-chkp.h: Remove.
20793         * rtl-chkp.c: Remove.
20794         * rtl-chkp.h: Remove.
20795         * tree-chkp-opt.c: Remove.
20796         * tree-chkp.c: Remove.
20797         * tree-chkp.h: Remove.
20799 2018-06-07  Carl Love  <cel@us.ibm.com>
20801         * config/rs6000/vsx.md (vextract_fp_from_shorth,
20802         vextract_fp_from_shortl): Add BE support.
20804 2018-06-07  Paul Koning  <ni1d@arrl.net>
20806         * compare-elim.c (try_merge_compare): Don't merge compare if
20807         address contains a side effect.
20808         (try_eliminate_compare): Likewise.
20810 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
20812         * config.gcc: Support "tremont".
20813         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
20814         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20815         PROCESSOR_TREMONT.
20816         * config/i386/i386.c (m_TREMONT): Define.
20817         (processor_target_table): Add "tremont".
20818         (PTA_TREMONT): Define.
20819         (ix86_lea_outperforms): Add TARGET_TREMONT.
20820         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
20821         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
20822         and M_INTEL_GOLDMONT_PLUS.
20823         (fold_builtin_cpu): Add "tremont".
20824         (ix86_add_stmt_cost): Add TARGET_TREMONT.
20825         (ix86_option_override_internal): Add "tremont".
20826         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
20827         (processor_type): Add PROCESSOR_TREMONT.
20828         * config/i386/x86-tune.def: Add m_TREMONT.
20829         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
20831 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20833         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
20834         symbol defined for msp430i* devices to be lower case.
20836 2018-06-07  Richard Biener  <rguenther@suse.de>
20838         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
20839         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
20840         Properly wrap signed arithmetic if overflow wraps.
20842 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
20844         PR tree-optimization/69615
20845         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
20846         of a cast from a same precision integral SSA_NAME in a bb dominated
20847         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
20848         cast to utype if rhs2 has already a compatible type.
20850 2018-06-07  Richard Biener  <rguenther@suse.de>
20852         PR tree-optimization/85935
20853         * graphite-scop-detection.c (find_params_in_bb): Analyze
20854         condition operands with respect to the correct loop.  Assert
20855         the analysis doesn't fail.
20857 2018-06-04  Carl Love  <cel@us.ibm.com>
20859         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
20860         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
20861         as it is slightly cheaper.
20862         (first_match_or_eos_index_<mode>):
20863         Calculate index using natural element order.
20864         (first_match_index_<mode>):
20865         Calculate index using natural element order.
20866         (first_match_or_eos_index_<mode>):
20867         Calculate index using natural order.
20868         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
20869         for BE and LE modes.
20870         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
20871         P9V_BUILTIN_VCLZLSBB_V16QI.
20872         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
20873         specific.
20875 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20877         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
20878         indentation and line wrap for many prototypes.  Add missing
20879         @smallexample directives around block of prototypes for vec_xl and
20880         vec_xst.
20882 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
20884         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
20885         track if we pass or return IEEE 128-bit floating point.
20886         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
20887         C++ mangling that is compatible with GCC 8.1.
20888         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
20889         (init_cumulative_args): Note if we pass or return IEEE 128-bit
20890         floating point types.
20891         (rs6000_function_arg_advance_1): Likewise.
20892         (rs6000_mangle_type): Optionally generate mangled names that match
20893         what GCC 8.1 generated for IEEE 128-bit floating point types.
20894         (rs6000_globalize_decl_name): If we have an external function that
20895         passes or returns IEEE 128-bit types, generate a weak reference
20896         from the mangled name used in GCC 8.1 to the current mangled
20897         name.
20898         (rs6000_init_builtins): Make __ibm128 use the long double type if
20899         long double is IBM extended double.  Make __float128 use the long
20900         double type if long double is IEEE 128-bit.
20902         PR target/85657
20903         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
20904         macro for __ibm128 built-in functions.
20905         (PACK_IF): Add __ibm128 pack/unpack functions.
20906         (UNPACK_IF): Likewise.
20907         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
20908         enable long double built-in functions if long double is IEEE
20909         128-bit floating point.
20910         (rs6000_invalid_builtin): Update long double built-in function
20911         error message.
20912         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
20913         functions, adjust the built-in function to use the long double
20914         built-in function if __ibm128 and long double are the same type.
20915         * doc/extend.texi (PowerPC builtins): Update documention for
20916         __builtin_{,un}pack_longdouble.  Add documentation for
20917         __builtin_{,un}pack_ibm128.
20919 2018-06-06  Jim Wilson  <jimw@sifive.com>
20921         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
20922         (struct machine_function): New field interrupt_mode.
20923         (riscv_handle_type_attribute): New function.  Add forward declaration.
20924         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
20925         (riscv_expand_epilogue): Check interrupt_mode field.
20926         (riscv_set_current_function): Check interrupt attribute args and
20927         set interrupt_mode field.
20928         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
20929         (riscv_sret, riscv_uret): New.
20930         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
20931         new arguments to interrupt attribute.
20933 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
20935         PR target/63177
20936         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
20937         Don't handle -mcpu=power8 if -mpower9-vector is also used.
20939 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20941         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
20942         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
20943         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20944         several redundant entries.
20946 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
20948         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
20949         type from "rtx" to "rtx_insn *".
20950         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
20951         for local "call_insn", removing cast.
20952         (ix86_expand_call): Likewise, introducing a "call_insn" local.
20954 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
20956         PR tree-optimization/86066
20957         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
20958         for BIT_INSERT_EXPR stores.
20960 2018-06-06  Richard Biener  <rguenther@suse.de>
20962         PR tree-optimization/86062
20963         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
20964         component refs ontop
20965         of to be offsetted base.
20967 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20969         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
20970         to be static and remove check on interrupt attribute name.
20972 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20974         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20975         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
20977 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
20979         PR target/79924
20980         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
20981         second argument.
20982         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
20983         Remove second argument, change how error is called.
20984         (aarch64_layout_arg): Remove second argument from
20985         aarch64_err_no_fpadvsimd call.
20986         (aarch64_init_cumulative_args): Ditto.
20987         (aarch64_gimplify_va_arg_expr): Ditto.
20988         * config/aarch64/aarch64.md (mov<mode>): Ditto.
20990 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
20992         * config/i386/i386.md (simple_return_indirect_internal): New expander.
20993         (*simple_return_indirect_internal<mode>): Rename from
20994         simple_return_indirect_internal.  Use W mode iterator.
20995         (rstorssp): New expander.
20996         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
20997         (clrssbsy): New expander.
20998         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
21000 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
21002         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
21003         __typeof__.
21004         (cmse_check_pointed_object): Likewise.
21006 2018-06-05  Martin Liska  <mliska@suse.cz>
21008         PR gcov-profile/47618
21009         * doc/invoke.texi: Document how -fprofile-dir format
21010         is extended.
21012 2018-06-05  Richard Biener  <rguenther@suse.de>
21014         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
21015         removal pretend DOM info isn't available so we do not update
21016         it and only remove edges, not dominated blocks.  Actually free
21017         DOM info in case we removed something.  Remove unreachable blocks.
21018         (mfb_keep_latches): Work with either DOM info or marked backedges.
21019         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
21020         first.  Mark backedges if DOM info isn't available.
21021         (Re-)compute DOM info after cleanup_control_flow_pre.
21023 2018-06-05  Richard Biener  <rguenther@suse.de>
21025         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
21026         (locus_discrim_hasher::hash): Adjust.
21027         (locus_discrim_hasher::equal): Likewise.
21028         (next_discriminator_for_locus): Work on line directly.
21029         (same_line_p): Pass in expanded locus1 as well.
21030         (assign_discriminators): Avoid redundant location expansions.
21032 2018-06-05  Richard Biener  <rguenther@suse.de>
21034         PR tree-optimization/86046
21035         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
21036         if required after clearing TREE_ADDRESSABLE.
21038 2018-06-05  Richard Biener  <rguenther@suse.de>
21040         PR tree-optimization/86047
21041         * tree-ssa-loop.c (for_each_index): Glob handling of all
21042         decls and constants and really handle all of them.
21044 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21046         PR target/81497
21047         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
21048         qualifier_void_pointer and qualifier_const_void_pointer.
21049         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
21050         (arm_init_builtins): Handle the above.
21051         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
21052         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
21053         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
21054         void intrinsics.
21056 2018-06-05  Martin Liska  <mliska@suse.cz>
21058         * auto-profile.c (read_autofdo_file): Do not use
21059         gcov_ctr_summary struct.
21060         (afdo_callsite_hot_enough_for_early_inline): Likewise.
21061         * coverage.c (struct counts_entry): Likewise.
21062         (read_counts_file): Read just single summary entry.
21063         (get_coverage_counts): Use gcov_summary struct.
21064         * coverage.h (get_coverage_counts): Likewise.
21065         * gcov-dump.c (dump_working_sets): Likewise.
21066         (tag_summary): Dump just single summary.
21067         * gcov-io.c (gcov_write_summary): Write just histogram
21068         summary.
21069         (gcov_read_summary): Read just single summary.
21070         (compute_working_sets): Use gcov_summary struct.
21071         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
21072         of GCOV_COUNTERS_SUMMABLE.
21073         (GCOV_COUNTERS_SUMMABLE): Remove.
21074         (GCOV_FIRST_VALUE_COUNTER): Replace with
21075         GCOV_COUNTER_V_INTERVAL.
21076         (struct gcov_ctr_summary): Remove.
21077         (struct gcov_summary): Directly use fields of former
21078         gcov_ctr_summary.
21079         (compute_working_sets): Use gcov_summary struct.
21080         * gcov.c (read_count_file): Do not use ctrs fields.
21081         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
21082         struct.
21083         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
21084         struct.
21085         * profile.c: Likewise.
21086         * profile.h: Likewise.
21088 2018-06-05  Martin Liska  <mliska@suse.cz>
21090         PR gcov-profile/84846
21091         * gcov.c (output_lines): Print working directory only
21092         in intermediate format.
21094 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
21096         * config/s390/s390-builtin-types.def: Add void function type.
21097         * config/s390/s390-builtins.def: Use the function type for the
21098         tbeginc builtin.
21100 2018-06-04  Jim Wilson  <jimw@sifive.com>
21102         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
21103         to int.
21104         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
21105         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
21106         handle EH_RETURN_DATA_REGNO registers properly.
21107         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
21108         (riscv_expand_epilogue): Update comment.  Change argument name and
21109         type.  Update code to use new name and type.  Pass new args to
21110         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
21111         EXCEPTION_RETURN.
21112         * config/riscv/riscv.md (NORMAL_RETURN): New.
21113         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
21114         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
21115         (eh_return): Call gen_eh_return_internal and emit barrier.
21116         (eh_return_internal): Call riscv_expand_epilogue.
21118 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
21120         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
21121         bit_insertion field and declare can_be_merged_into method.
21122         (merged_store_group::can_be_merged_into): New method.
21123         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
21124         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
21125         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
21127 2018-06-04  Richard Biener  <rguenther@suse.de>
21129         PR tree-optimization/85955
21130         * builtins.c (fold_builtin_sincos): Convert pointers to
21131         destination to appropriate type before dereferencing.
21133 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
21135         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
21137 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
21139         * expr.c (expand_expr_real_1): Force the operand into memory if
21140         its TYPE_MODE is BLKmode and if there is no integer mode for
21141         the number of bits being extracted.
21143 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
21145         PR target/85832
21146         PR target/86036
21147         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
21148         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
21150 2018-06-04  Richard Biener  <rguenther@suse.de>
21152         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
21153         (cleanup_tree_cfg_noloop): ... single caller.  Do
21154         start_recording_case_labels later.
21156 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
21158         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
21159         to _IMMINTRIN_H_INCLUDED.
21160         * config/i386/pconfigintrin.h: Ditto.
21161         * config/i386/waitpkgintrin.h: Ditto.
21162         * config/i386/immintrin.h: Add includes for sgxintrin.h,
21163         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
21164         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
21165         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
21166         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
21167         waitpkgintrin.h and cldemoteintrin.h.
21169 2018-06-04  Richard Biener  <rguenther@suse.de>
21171         PR tree-optimization/86038
21172         * tracer.c (find_best_successor): Check probability for
21173         being initialized, bail out if not.
21175 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
21177         PR target/86003
21178         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
21179         of bits to ignore when comparing architectures.
21181 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
21183         PR tree-optimization/69615
21184         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
21185         maximum or minimum of the type, try to merge it also as if
21186         range1 is + [-, x - 1] or + [x + 1, -].
21188         PR c++/86025
21189         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
21191 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
21193         PR tree-optimization/86034
21194         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
21195         the unsigned bitfield type in a bit insertion sequence if it does not
21196         have a larger precision than the bitfield size.
21197         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
21199 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
21201         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
21203 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21205         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
21206         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
21207         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
21208         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
21210 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21212         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
21213         Disable -fdelete-null-pointer-checks for ELF toolchain.
21215 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
21216             Kito Cheng  <kito.cheng@gmail.com>
21218         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
21219         (nds32le-*-*, nds32be-*-*): Integrate checking process.
21220         (nds32*-*-*): Add glibc and uclibc conditions.
21221         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
21222         (TARGET_EXCEPT_UNWIND_INFO): Define.
21223         * config/nds32/elf.h: New file.
21224         * config/nds32/linux.h: New file.
21225         * config/nds32/nds32-elf.opt: New file.
21226         * config/nds32/nds32-linux.opt: New file.
21227         * config/nds32/nds32-fp-as-gp.c
21228         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
21229         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
21230         TARGET_LINUX_ABI.
21231         (nds32_asm_file_end): Ditto.
21232         (nds32_print_operand): Ditto.
21233         (nds32_insert_attributes): Ditto.
21234         (nds32_init_libfuncs): New function.
21235         (TARGET_HAVE_TLS): Define.
21236         (TARGET_INIT_LIBFUNCS): Define.
21237         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
21238         spec content.
21239         (TARGET_ELF): Apply different mcmodel setting.
21240         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
21241         been migrated into elf.h and linux.h files.
21242         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
21243         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
21244         (mcmodel): The content has been migrated into nds32-elf.opt and
21245         nds32-linux.opt files.
21246         * config/nds32/t-elf: New file.
21247         * config/nds32/t-linux: New file.
21249 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
21250             Shiva Chen  <shiva0217@gmail.com>
21252         * config/nds32/constants.md (unspec_volatile_element): Add
21253         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
21254         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
21255         optimization.
21256         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
21257         (make_pass_nds32_fp_as_gp): Declare.
21258         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
21259         optmization pass.
21260         (nds32_asm_function_end_prologue): Remove unused asm output.
21261         (nds32_asm_function_begin_epilogue): Remove unused asm output.
21262         (nds32_asm_file_start): Output necessary fp_as_gp information.
21263         (nds32_option_override): Adjust register usage.
21264         (nds32_expand_prologue): Consider fp_as_gp situation.
21265         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
21266         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
21267         (epilogue): Ditto.
21268         (return): Ditto.
21269         (simple_return): Ditto.
21270         (omit_fp_begin): Output special directive for fp_as_gp.
21271         (omit_fp_end): Output special directive for fp_as_gp.
21272         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
21273         mforbid-fp-as-gp): New options.
21275 2018-06-01  Mark Wielaard  <mark@klomp.org>
21277         * dwarf2out.c (dwarf2out_finish): Remove generation of
21278         DW_AT_loclists_base.
21280 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
21282         * gimple-ssa-store-merging.c: Include gimple-fold.h.
21283         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
21284         (struct merged_store_group): Add bit_insertion field.
21285         (dump_char_array): Use standard hexadecimal format.
21286         (merged_store_group::merged_store_group): Set bit_insertion to false.
21287         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
21288         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
21289         also print the mask in the dump file.
21290         (pass_store_merging::gate): Minor tweak.
21291         (imm_store_chain_info::coalesce_immediate): Fix wrong association
21292         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
21293         stores with INTEGER_CST stores.
21294         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
21295         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
21296         and use it throughout.  Generate bit insertion sequences if need be.
21297         (pass_store_merging::process_store): Remove redundant condition.
21298         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
21300 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
21302         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
21303         the 128-bit floating point types.  Fix function comment.
21305 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21307         * config/aarch64/aarch64-simd.md
21308         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
21309         mnemonics.
21310         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
21311         mnemonics.
21313 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
21315         PR tree-optimization/85989
21316         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
21317         variable.
21318         (backprop::intersect_uses): Check it when deciding whether this
21319         is a backedge reference.
21320         (backprop::process_block): Add each phi to m_visited_phis
21321         after visiting it, then clear it at the end.
21323 2018-06-01  Richard Biener  <rguenther@suse.de>
21325         * tree-vectorizer.h (vect_dr_stmt): New function.
21326         (vect_get_load_cost): Adjust.
21327         (vect_get_store_cost): Likewise.
21328         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
21329         Use vect_dr_stmt instead of DR_SMTT.
21330         (vect_record_base_alignments): Likewise.
21331         (vect_calculate_target_alignment): Likewise.
21332         (vect_compute_data_ref_alignment): Likewise and make static.
21333         (vect_update_misalignment_for_peel): Likewise.
21334         (vect_verify_datarefs_alignment): Likewise.
21335         (vector_alignment_reachable_p): Likewise.
21336         (vect_get_data_access_cost): Likewise.  Pass down
21337         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
21338         (vect_get_peeling_costs_all_drs): Likewise.
21339         (vect_peeling_hash_get_lowest_cost): Likewise.
21340         (vect_enhance_data_refs_alignment): Likewise.
21341         (vect_find_same_alignment_drs): Likewise.
21342         (vect_analyze_data_refs_alignment): Likewise.
21343         (vect_analyze_group_access_1): Likewise.
21344         (vect_analyze_group_access): Likewise.
21345         (vect_analyze_data_ref_access): Likewise.
21346         (vect_analyze_data_ref_accesses): Likewise.
21347         (vect_vfa_segment_size): Likewise.
21348         (vect_small_gap_p): Likewise.
21349         (vectorizable_with_step_bound_p): Likewise.
21350         (vect_prune_runtime_alias_test_list): Likewise.
21351         (vect_analyze_data_refs): Likewise.
21352         (vect_supportable_dr_alignment): Likewise.
21353         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
21354         (vect_gen_prolog_loop_niters): Likewise.
21355         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
21356         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
21357         modify DR_STMT.
21358         (vect_recog_mask_conversion_pattern): Likewise.
21359         (vect_try_gather_scatter_pattern): Likewise.
21360         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
21361         to vect_get_store_cost.
21362         (vect_get_store_cost): Get stmt_info instead of DR.
21363         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
21364         (vect_get_load_cost): Get stmt_info instead of DR.
21366 2018-06-01  Richard Biener  <rguenther@suse.de>
21368         PR middle-end/86017
21369         * gimple-fold.c (var_decl_component_p): Also allow offsetted
21370         vars wrapped in MEM_REFs.
21372 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
21374         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21375         Fix subreg tests so that we only return a choice between
21376         GENERAL_REGS and FP_REGS if the original classes included both.
21378 2018-06-01  Richard Biener  <rguenther@suse.de>
21380         PR ipa/85960
21381         * tree-ssa-structalias.c (get_function_part_constraint):
21382         Handle NULL fi->decl.
21383         (find_func_aliases_for_call): Properly handle indirect
21384         fi from direct call.
21385         (find_func_clobbers): Likewise.
21386         (ipa_pta_execute): Likewise.
21387         (create_variable_info_for): For functions that are ifunc_resolver
21388         resolve to a varinfo that contains the result of the resolver call.
21389         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
21390         aliases.
21392 2018-05-31  Michael Collison  <michael.collison@arm.com>
21394         * config/aarch64/aarch64.md:
21395         (*fix_to_zero_extenddfdi2): New pattern.
21396         * gcc.target/aarch64/fix_extend1.c: New testcase.
21398 2018-05-31  Qing Zhao  <qing.zhao@oracle.com>
21400         PR middle-end/78809
21401         PR middle-end/83026
21402         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
21403         and BUILT_IN_STRNCMP_EQ.
21404         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
21405         BUILT_IN_STRNCMP_EQ.
21406         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
21407         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21408         (gimple_fold_builtin): Likewise.
21409         * tree-ssa-strlen.c (compute_string_length): New function.
21410         (determine_min_obsize): New function.
21411         (handle_builtin_string_cmp): New function to handle calls to
21412         string compare functions.
21413         (strlen_optimize_stmt): Add handling to builtin string compare
21414         calls.
21415         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
21416         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21417         * tree.c (build_common_builtin_nodes): Add new defines of
21418         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
21420 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
21422         PR target/85984
21423         * bb-reorder.c (pass_partition_blocks::gate): Return false for
21424         functions with naked attribute.
21426 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
21428         * config/i386/sse.md (avx_vec_concat<mode>):
21429         Substitute concat_tg_mode mode attribute with xtg_mode.
21430         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
21431         (concat_tg_mode): Remove mode attribute.
21433 2018-05-31  Martin Sebor  <msebor@redhat.com>
21435         PR c/82063
21436         * calls.c (alloc_max_size): Correct a logic error/typo.
21437         Treat excessive arguments as infinite.  Warn for invalid arguments.
21438         * doc/invoke.texi (-Walloc-size-larger-than): Update.
21440 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
21442         PR target/85829
21443         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
21444         and movx for Haswell.
21446 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
21447             Cesar Philippidis  <cesar@codesourcery.com>
21449         PR middle-end/85879
21450         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
21451         when emitting error on private/firstprivate reductions.
21452         * omp-low.c (lower_omp_target): Avoid reference-type processing
21453         on pointers for firstprivate clause.
21455 2018-05-31  Sameera Deshpande  <sameera.deshpande@linaro.org>
21457         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
21458         (st1x2): Likewise.
21459         (st1x3): Likewise.
21460         * config/aarch64/aarch64-simd.md
21461         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
21462         (aarch64_ld1_x3_<mode>): Likewise
21463         (aarch64_st1x2<VALLDIF:mode>): Likewise
21464         (aarch64_st1_x2_<mode>): Likewise
21465         (aarch64_st1x3<VALLDIF:mode>): Likewise
21466         (aarch64_st1_x3_<mode>): Likewise
21467         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
21468         (vld1_s8_x3): Likewise.
21469         (vld1_u16_x3): Likewise.
21470         (vld1_s16_x3): Likewise.
21471         (vld1_u32_x3): Likewise.
21472         (vld1_s32_x3): Likewise.
21473         (vld1_u64_x3): Likewise.
21474         (vld1_s64_x3): Likewise.
21475         (vld1_f16_x3): Likewise.
21476         (vld1_f32_x3): Likewise.
21477         (vld1_f64_x3): Likewise.
21478         (vld1_p8_x3): Likewise.
21479         (vld1_p16_x3): Likewise.
21480         (vld1_p64_x3): Likewise.
21481         (vld1q_u8_x3): Likewise.
21482         (vld1q_s8_x3): Likewise.
21483         (vld1q_u16_x3): Likewise.
21484         (vld1q_s16_x3): Likewise.
21485         (vld1q_u32_x3): Likewise.
21486         (vld1q_s32_x3): Likewise.
21487         (vld1q_u64_x3): Likewise.
21488         (vld1q_s64_x3): Likewise.
21489         (vld1q_f16_x3): Likewise.
21490         (vld1q_f32_x3): Likewise.
21491         (vld1q_f64_x3): Likewise.
21492         (vld1q_p8_x3): Likewise.
21493         (vld1q_p16_x3): Likewise.
21494         (vld1q_p64_x3): Likewise.
21495         (vst1_s64_x2): Likewise.
21496         (vst1_u64_x2): Likewise.
21497         (vst1_f64_x2): Likewise.
21498         (vst1_s8_x2): Likewise.
21499         (vst1_p8_x2): Likewise.
21500         (vst1_s16_x2): Likewise.
21501         (vst1_p16_x2): Likewise.
21502         (vst1_s32_x2): Likewise.
21503         (vst1_u8_x2): Likewise.
21504         (vst1_u16_x2): Likewise.
21505         (vst1_u32_x2): Likewise.
21506         (vst1_f16_x2): Likewise.
21507         (vst1_f32_x2): Likewise.
21508         (vst1_p64_x2): Likewise.
21509         (vst1q_s8_x2): Likewise.
21510         (vst1q_p8_x2): Likewise.
21511         (vst1q_s16_x2): Likewise.
21512         (vst1q_p16_x2): Likewise.
21513         (vst1q_s32_x2): Likewise.
21514         (vst1q_s64_x2): Likewise.
21515         (vst1q_u8_x2): Likewise.
21516         (vst1q_u16_x2): Likewise.
21517         (vst1q_u32_x2): Likewise.
21518         (vst1q_u64_x2): Likewise.
21519         (vst1q_f16_x2): Likewise.
21520         (vst1q_f32_x2): Likewise.
21521         (vst1q_f64_x2): Likewise.
21522         (vst1q_p64_x2): Likewise.
21523         (vst1_s64_x3): Likewise.
21524         (vst1_u64_x3): Likewise.
21525         (vst1_f64_x3): Likewise.
21526         (vst1_s8_x3): Likewise.
21527         (vst1_p8_x3): Likewise.
21528         (vst1_s16_x3): Likewise.
21529         (vst1_p16_x3): Likewise.
21530         (vst1_s32_x3): Likewise.
21531         (vst1_u8_x3): Likewise.
21532         (vst1_u16_x3): Likewise.
21533         (vst1_u32_x3): Likewise.
21534         (vst1_f16_x3): Likewise.
21535         (vst1_f32_x3): Likewise.
21536         (vst1_p64_x3): Likewise.
21537         (vst1q_s8_x3): Likewise.
21538         (vst1q_p8_x3): Likewise.
21539         (vst1q_s16_x3): Likewise.
21540         (vst1q_p16_x3): Likewise.
21541         (vst1q_s32_x3): Likewise.
21542         (vst1q_s64_x3): Likewise.
21543         (vst1q_u8_x3): Likewise.
21544         (vst1q_u16_x3): Likewise.
21545         (vst1q_u32_x3): Likewise.
21546         (vst1q_u64_x3): Likewise.
21547         (vst1q_f16_x3): Likewise.
21548         (vst1q_f32_x3): Likewise.
21549         (vst1q_f64_x3): Likewise.
21550         (vst1q_p64_x3): Likewise.
21552 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
21554         * config/msp430/msp430.c (msp430_output_labelref): Prepend
21555         user_label_prefix to name.
21557         * tree-core.h: Update comment about the format of NAME string
21558         passed to handler in attribute_spec.
21560         * config/msp430/msp430.md: Remove erroneous subreg expression from
21561         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
21562         zero_extend{q,h}isi2.
21564 2018-05-30  Borislav Petkov  <bp@suse.de>
21566         * doc/extend.texi: Document some architecture specific
21567         constraints and sort entries.
21569 2018-05-30  Martin Sebor  <msebor@redhat.com>
21571         PR middle-end/85369
21572         * builtins.c (expand_builtin_stpcpy_1): New function.
21573         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
21574         only if the former succeeds.
21576 2018-05-31  Sameera Deshpande  <sameera.deshpande@linaro.org>
21578         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
21579         in saphira.
21581 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
21583         * doc/invoke.texi (-flinker-output): Document
21585 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
21587         * passes.c (ipa_write_summaries): Only modify statements if body
21588         is in memory.
21589         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
21590         incrementally linking.
21591         (ipa_passes): Likewise.
21592         * lto-cgraph.c (lto_output_node): When incrementally linking do not
21593         pass down resolution info.
21594         * common.opt (flag_incremental_link): Update info.
21595         * gcc.c (plugin specs): Turn flinker-output=* to
21596         -plugin-opt=-linker-output-known
21597         * toplev.c (compile_file): Also cut compilation when doing incremental
21598         link.
21599         * flag-types. (enum lto_partition_model): Add
21600         LTO_LINKER_OUTPUT_NOLTOREL.
21601         (invoke.texi): Add -flinker-output docs.
21602         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
21603         link same way as WPA; do not stream in dead initializers.
21605         * dwarf2out.c (dwarf2out_die_ref_for_decl,
21606         darf2out_register_external_decl): Support incremental link.
21608 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
21610         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
21612 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
21614         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
21615         it down to simple_object_copy_lto_debug_sections.
21616         (run_gcc): Determine incremental LTO link time and configure
21617         lto1 into non-wpa mode, disable renaming of debug sections.
21619 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21621         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
21622         descriptions of various incorrectly documented functions.
21624 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
21626         Revert:
21627         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21628         address check not strict.
21630 2018-05-30  Richard Biener  <rguenther@suse.de>
21632         PR tree-optimization/85964
21633         * tracer.c (better_p): Drop initialized count check, we only
21634         call the function with initialized counts now.
21635         (find_best_successor): Do find a best edge if one
21636         has uninitialized count.
21637         (find_best_predecessor): Likewise.  Do BB frequency check only
21638         if count is initialized.
21640 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
21642         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
21643         (aarch64_ldrstr_offset_compare): New.
21644         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
21645         load/store orderings.
21646         (aarch64_gen_adjusted_ldpstp): Likewise.
21648 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
21650         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21651         Check for subset of GENERAL_REGS and FP_REGS.
21652         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
21653         r=w alternative.
21655 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
21657         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
21658         and wi::to_poly_offset.  Add the current offset and then check
21659         whether the sum fits, rather than using an unchecked addition of
21660         a checked term.  Check for a shwi rather than a uhwi.
21661         * expr.c (get_bit_range): Use tree_to_poly_uint64.
21662         (store_constructor): Use poly_int_tree_p.
21663         (expand_expr_real_1): Likewise.
21664         * function.c (assign_temp): Likewise.
21665         * fold-const.c (const_binop): Use poly_int_tree_p and
21666         wi::to_poly_offset.
21667         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
21668         division.
21669         * ipa-icf-gimple.c (func_checker::compare_operand): Use
21670         to_poly_offset for MEM offsets.
21671         * ipa-icf.c (sem_variable::equals): Likewise.
21672         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
21673         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
21674         wi::to_poly_offset for BIT_FIELD_REF offsets.
21675         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
21676         wi::to_poly_offset.
21677         * var-tracking.c (emit_note_insn_var_location): Use
21678         tree_to_poly_uint64.
21680 2018-05-29  Jim Wilson  <jimw@sifive.com>
21682         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
21684 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
21686         PR target/85950
21687         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
21688         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
21689         sequence.
21690         (sse4_1_round<mode>2): Use nonimmediate_operand
21691         for operand 1 predicate.
21693 2018-05-29  Martin Sebor  <msebor@redhat.com>
21694             Richard Biener  <rguenther@suse.de>
21696         PR testsuite/85888
21697         * calls.c (get_size_range): Call determine_value_range instead
21698         of get_value_range..
21699         * tree-vrp.h (determine_value_range): Declared new function.
21700         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
21702 2018-05-29  Richard Biener  <rguenther@suse.de>
21704         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
21705         sure to use non-pattern stmts for get_earlier_stmt arguments.
21706         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
21707         called on pattern stmts.
21708         (get_later_stmt): Likewise.
21710 2018-05-29  Martin Liska  <mliska@suse.cz>
21712         PR gcov-profile/85759
21713         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
21714         env variables.
21716 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
21718         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
21719         VEC_UNPACK_*_EXPR.
21720         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
21721         VEC_PACK_*_EXPR.
21723         PR target/85918
21724         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21725         VEC_PACK_FLOAT_EXPR): New tree codes.
21726         * tree-pretty-print.c (op_code_prio): Handle
21727         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21728         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21729         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21730         * tree-inline.c (estimate_operator_cost): Likewise.
21731         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
21732         * fold-const.c (const_binop): Likewise.
21733         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
21734         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21735         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
21736         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
21737         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21738         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21739         * expr.c (expand_expr_real_2): Likewise.
21740         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
21741         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
21742         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
21743         optabs.
21744         * optabs.c (expand_widen_pattern_expr): For
21745         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
21746         sign from result type rather than operand's type.
21747         (expand_binop_directly): For vec_packu_float_optab and
21748         vec_packs_float_optab allow result type to be different from operand's
21749         type.
21750         * optabs-tree.c (optab_for_tree_code): Handle
21751         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21752         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
21753         * tree-vect-generic.c (expand_vector_operations_1):  Handle
21754         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21755         VEC_PACK_FLOAT_EXPR.
21756         * tree-vect-stmts.c (supportable_widening_operation): Handle
21757         FIX_TRUNC_EXPR.
21758         (supportable_narrowing_operation): Handle FLOAT_EXPR.
21759         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
21760         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
21761         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
21762         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
21763         mode attributes.
21764         (vec_pack<floatprefix>_float_<mode>): New expander.
21765         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
21766         attributes.
21767         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
21768         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
21769         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
21770         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
21771         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
21772         Document.
21773         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
21774         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
21775         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21776         VEC_PACK_FLOAT_EXPR): Document.
21778 2018-05-29  Richard Biener  <rguenther@suse.de>
21780         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
21781         member.
21782         (stmt_vec_info_vec): Make pointer.
21783         (init_stmt_vec_info_vec): Remove.
21784         (free_stmt_vec_info_vec): Likewise.
21785         (set_stmt_vec_info_vec): New function.
21786         (free_stmt_vec_infos): Likewise.
21787         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
21788         (set_vinfo_for_stmt): Likewise.
21789         (get_earlier_stmt): Likewise.
21790         (get_later_stmt): Likewise.
21791         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
21792         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
21793         (vec_info::~vec_info): Free stmt_vec_infos.
21794         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
21795         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
21796         (pass_slp_vectorize::execute): Likewise.
21797         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
21798         (free_stmt_vec_info_vec): Likewise.
21799         (set_stmt_vec_info_vec): New function.
21800         (free_stmt_vec_infos): Likewise.
21801         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
21802         the global stmt_vec_info_vec.
21803         * tree-parloops.c (gather_scalar_reductions): Use
21804         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
21805         vector.
21807 2018-05-29  Richard Biener  <rguenther@suse.de>
21809         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
21811 2018-05-29  Martin Liska  <mliska@suse.cz>
21812             David Malcolm  <dmalcolm@redhat.com>
21814         * vec.c (test_reverse): New.
21815         (vec_c_tests): Add new test.
21816         * vec.h (vl_ptr>::reverse): New function.
21818 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
21820         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
21822         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
21823         and later.
21825 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
21827         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
21829 2018-05-28  Richard Biener  <rguenther@suse.de>
21831         PR tree-optimization/85933
21832         * tree-vect-data-refs.c (vect_record_base_alignments): Only
21833         look at stmts marked as vectorizable.
21835 2018-05-28  Richard Biener  <rguenther@suse.de>
21837         PR tree-optimization/85934
21838         * tree-vect-generic.c (expand_vector_operations_1): Hoist
21839         vector boolean check before scalar optimization.
21841 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
21843         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
21844         for armv5te.
21846 2018-05-28  Mark Wielaard  <mark@klomp.org>
21848         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
21849         if it is an expression containing a minus sign.
21851 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
21853         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
21855 2018-05-27  Paul Koning  <ni1d@arrl.net>
21857         * config/pdp11/pdp11.md (truncsihi2): Remove.
21859 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
21860             Chung-Ju Wu  <jasonwucj@gmail.com>
21862         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
21863         implementation.
21864         (unaligned_store_dw): Ditto.
21865         * config/nds32/nds32-memory-manipulation.c
21866         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
21867         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
21868         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
21869         (emit_setmem_word_loop): Rename to ...
21870         (emit_setmem_doubleword_loop): ... this.
21871         (nds32_gen_dup_4_byte_to_word_value): New function.
21872         (nds32_gen_dup_8_byte_to_double_word_value): New function.
21873         (nds32_expand_setmem_loop): Refine implementation.
21874         (nds32_expand_setmem_loop_v3m): Ditto.
21875         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
21876         pattern.
21878 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21880         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
21882 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21884         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
21885         (nds32_init_machine_status): Initialize machine->attr_naked_p and
21886         machine->attr_no_prologue_p.
21887         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
21888         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
21889         (nds32_expand_epilogue): Consider attr_naked_p.
21890         (nds32_expand_epilogue_v3pop): Likewise.
21891         (nds32_can_use_return_insn): Likewise.
21892         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
21893         attr_no_prologue_p fields.
21894         * config/nds32/nds32.opt (mret-in-naked-func): New option.
21896 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
21898         PR target/85918
21899         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
21900         attributes.
21901         * config/i386/sse.md
21902         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
21903         Rename to ...
21904         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
21905         ... this.
21906         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
21907         Rename to ...
21908         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
21909         ... this.
21910         (*<floatsuffix>floatv2div2sf2): Rename to ...
21911         (*float<floatunssuffix>v2div2sf2): ... this.
21912         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
21913         (float<floatunssuffix>v2div2sf2_mask): ... this.
21914         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
21915         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
21916         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
21917         to ...
21918         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
21919         ... this.
21920         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21921         Rename to ...
21922         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21923         ... this.
21924         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21925         Rename to ...
21926         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21927         ... this.
21928         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
21929         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
21930         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
21931         gen_ufix_truncv8dfv8si2.
21932         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
21933         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
21934         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
21935         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
21936         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
21937         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
21938         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
21939         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
21941 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
21943         PR target/85900
21944         PR target/85345
21945         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
21947 2018-05-25  Jim Wilson  <jimw@sifive.com>
21949         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
21950         * config/riscv/riscv.c (struct machine_function): Add
21951         interrupt_handler_p and attribute_checked_p fields.
21952         (riscv_attribute_table): Add interrupt.
21953         (riscv_interrupt_type_p): New.
21954         (riscv_save_reg_p): Save extra regs for interrupt handler.
21955         (riscv_use_save_libcall): Return false  for interrupt handler.
21956         (riscv_first_stack_step): Add forward declaration.
21957         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
21958         for interrupt handler with large frame.  Use it for saved reg list.
21959         (riscv_expand_prologue): Move flag_stack_usage_info support to
21960         eliminate duplication.
21961         (riscv_expand_epilogue): Generate mret for interrupt handler.
21962         (riscv_epilogue_uses): New.
21963         (riscv_can_use_return_insn): Return false for interrupt handler.
21964         (riscv_function_ok_for_sibcall): Likewise.
21965         (riscv_set_current_function): Add interrupt handler support.
21966         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
21967         * config/riscv/riscv.md (UNSPECV_MRET): New.
21968         (GP_REGNUM): New.
21969         (riscv_frflags, riscv_fsflags): Use tab after opcode.
21970         (riscv_mret): New.
21971         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
21973 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
21975         PR tree-optimization/85712
21976         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
21977         this candidate has already been replaced in-situ by a copy.
21979 2018-05-25  Jason Merrill  <jason@redhat.com>
21981         PR c++/80485 - inline function non-zero address.
21982         * symtab.c (nonzero_address): Check DECL_COMDAT.
21984 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
21986         PR target/83628
21987         * config/alpha/alpha.md (ashlsi3): New insn pattern.
21988         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
21989         extension of SImode operation.  Use const123_operand predicate.
21990         (*saddsi_1): Remove.
21991         (*saddl_se_1): Ditto.
21992         (*ssubsi_1): Ditto.
21993         (*ssubl_se_1): Ditto.
21994         * config/alpha/predicates.md (const123_operand): New predicate.
21995         * config/alpha/constraints.md (P): Use IN_RANGE.
21997 2018-05-25  Richard Biener  <rguenther@suse.de>
21999         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
22000         defaulted to true.
22001         (ref_maybe_used_by_stmt_p): Likewise.
22002         (stmt_may_clobber_ref_p): Likewise.
22003         (stmt_may_clobber_ref_p_1): Likewise.
22004         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
22005         and pass it along.
22006         (ref_maybe_used_by_stmt_p): Likewise.
22007         (stmt_may_clobber_ref_p): Likewise.
22008         (stmt_may_clobber_ref_p_1): Likewise.
22009         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
22010         the alias oracle to disambiguate DRs with stmts DR analysis
22011         couldn't handle.
22012         (vect_analyze_data_refs): Do not give up on not analyzable
22013         DRs for BB vectorization.  Remove code truncating the dataref
22014         vector.
22016 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
22018         PR target/85832
22019         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
22020         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
22021         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
22023 2018-05-25  Richard Biener  <rguenther@suse.de>
22025         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
22026         function, combining stmt data ref gathering and fatal analysis
22027         parts.
22028         (vect_analyze_data_refs): Remove now redudnant code and simplify.
22029         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
22030         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
22031         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
22032         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
22034 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
22036         PR tree-optimization/85720
22037         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
22038         SCC if all partitions are builtins.
22039         (version_loop_by_alias_check): New parameter.  Generate cancelable
22040         runtime alias check if all partitions are builtins.
22041         (distribute_loop): Update call to above function.
22043 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
22045         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
22046         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
22047         (parm_default_def_partition_arg): Ditto.
22048         (set_parm_default_def_partition): Ditto.
22049         (get_parm_default_def_partitions): Ditto and make it static.
22050         (get_undefined_value_partitions): Ditto and make it static.
22051         (remove_ssa_form): Refactor call to init_var_map here.
22052         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
22053         computation for loop region.
22054         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
22055         (register_default_def): Delete.
22056         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
22057         (parm_default_def_partition_arg): Ditto.
22058         (set_parm_default_def_partition): Ditto.
22059         (get_parm_default_def_partitions): Ditto and make it static.
22060         (get_undefined_value_partitions): Ditto and make it static.
22061         (coalesce_with_default, coalesce_with_default): Update comment.
22062         (create_coalesce_list_for_region): New func factored out from
22063         create_outofssa_var_map.
22064         (populate_coalesce_list_for_outofssa): New func factored out from
22065         create_outofssa_var_map and coalesce_ssa_name.
22066         (create_outofssa_var_map): Delete.
22067         (coalesce_ssa_name): Refactor to support live range computation.
22068         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
22069         (get_parm_default_def_partitions): Delete.
22070         (get_undefined_value_partitions): Ditto.
22071         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
22072         computation for loop region.
22073         (new_tree_live_info, loe_visit_block): Ditto.
22074         (live_worklist, set_var_live_on_entry): Ditto.
22075         (calculate_live_on_exit, verify_live_on_entry): Ditto.
22076         * tree-ssa-live.h (struct _var_map): New fields.
22077         (init_var_map): Change decl.
22078         (region_contains_p): New.
22080 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
22082         * tree-ssa-live.h (live_merge_and_clear): Delete.
22084 2018-05-25  Richard Biener  <rguenther@suse.de>
22086         PR c++/85912
22087         * tree-dump.c (dequeue_and_dump): Remove access to removed
22088         operand 2 of a SWITCH_EXPR.
22090 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22092         * doc/sourcebuild.texi (vect_double_cond_arith): Include
22093         multiplication and division.
22094         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
22095         (cond_udiv@var{m}, cond_umod@var{m}): Document.
22096         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
22097         (cond_udiv_optab, cond_umod_optab): New optabs.
22098         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
22099         (IFN_COND_RDIV): New internal functions.
22100         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
22101         TRUNC_MOD_EXPR and RDIV_EXPR.
22102         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
22103         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
22104         New unspecs.
22105         (SVE_INT_BINARY): Include mult.
22106         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
22107         (optab, sve_int_op): Handle mult.
22108         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
22109         UNSPEC_COND_DIV.
22110         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
22111         for SVE_INT_BINARY_SD.
22113 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22115         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
22116         (optab, sve_int_op): Handle div and udiv.
22117         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
22118         for SVE_INT_BINARY_SD.
22119         (*<optab><mode>3): New insn for the same.
22121 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22123         * tree-vect-patterns.c: Include predict.h.
22124         (vect_recog_divmod_pattern): Restrict check for division support
22125         to when optimizing for size.
22127 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22129         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
22130         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
22131         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
22132         (gimple_match_op::set_op): Likewise.
22133         (gimple_resimplify4): Declare.
22134         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
22135         (expr::gen_transform): Likewise.
22136         (decision_tree::gen): Generate a simplification routine for 4 operands.
22137         * gimple-match-head.c (gimple_simplify): Add an overload for
22138         4 operands.  In the top-level function, handle up to 4 call
22139         arguments and call gimple_resimplify4.
22140         (gimple_resimplify4): New function.
22141         (build_call_internal): Pass a fourth operand.
22142         (maybe_push_to_seq): Likewise.
22143         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
22144         Fold VEC_COND_EXPRs of an operation and a default value into
22145         an IFN_COND_* function if possible.
22146         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
22147         New unspecs.
22148         (SVE_COND_FP_BINARY): Include them.
22149         (optab, sve_fp_op): Handle them.
22150         (SVE_INT_BINARY_REV): New code iterator.
22151         (SVE_COND_FP_BINARY_REV): New int iterator.
22152         (commutative): New int attribute.
22153         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
22154         Declare.
22155         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
22156         function.
22157         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
22158         (*cond_<optab><mode>): New patterns for reversed operands.
22160 2018-05-25  Richard Biener  <rguenther@suse.de>
22162         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
22163         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
22164         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
22165         (STMT_VINFO_GROUPED_ACCESS): Adjust.
22166         * tree-vect-data-refs.c (everywhere): Adjust users.
22167         * tree-vect-loop.c (everywhere): Likewise.
22168         * tree-vect-slp.c (everywhere): Likewise.
22169         * tree-vect-stmts.c (everywhere): Likewise.
22170         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
22172 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22174         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
22175         Rename to...
22176         (gcc_cv_as_section_exclude): ... this.
22177         Try Solaris as #exclude syntax.
22178         * configure: Regenerate.
22179         * config.in: Regenerate.
22180         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
22181         SECTION_EXCLUDE.
22182         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
22183         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
22185         * varasm.c (default_elf_asm_named_section): Don't check if
22186         HAVE_GAS_SECTION_EXCLUDE is defined.
22188 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22190         * doc/md.texi: Update the documentation of the cond_* optabs
22191         to mention the new final operand.  Fix GET_MODE_NUNITS call.
22192         Describe the scalar case too.
22193         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
22194         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
22195         instead of 2.
22196         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
22197         (get_conditional_internal_fn): Update comment.
22198         * tree-vect-loop.c (vectorizable_reduction): Pass the original
22199         accumulator value as a final argument to conditional functions.
22200         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
22201         a define_expand and add an "else" operand.  Assert for now that
22202         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
22203         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
22204         (*cond_<optab><mode>): New patterns.
22205         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
22206         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22207         (UNSPEC_COND_EOR): Delete.
22208         (optab): Remove associated mappings.
22209         (SVE_INT_BINARY): New code iterator.
22210         (sve_int_op): Remove int attribute and add "minus" to the code
22211         attribute.
22212         (SVE_COND_INT_OP): Delete.
22213         (SVE_COND_FP_OP): Rename to...
22214         (SVE_COND_FP_BINARY): ...this.
22216 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
22218         * optabs.c (can_reuse_operands_p): New function.
22219         (maybe_legitimize_operands): Try to reuse the results for
22220         earlier operands.
22222 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
22224         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
22225         Add {q} suffix to insn mnemonic.
22227 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
22229         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
22230         (msp430_warn_func_return): New.
22232 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
22234         * fold-const.c (tree_nonzero_bits): New function.
22235         * fold-const.h (tree_nonzero_bits): Likewise.
22236         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
22237         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
22239 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
22241         PR target/85900
22242         PR target/85345
22243         * varasm.c (assemble_alias): Check ifunc_resolver only on
22244         FUNCTION_DECL.
22246 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
22248         PR target/85903
22249         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
22250         when memory input operand is handled.
22252 2018-05-24  Luis Machado  <luis.machado@linaro.org>
22254         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
22255         global.
22256         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
22258 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
22260         * match.pd: Delay FMA folds until after vectorization.
22262 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
22264         PR target/83009
22265         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
22266         address check not strict.
22268 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
22270         * gimple-match.h (gimple_match_op): New class.
22271         (mprts_hook): Replace parameters with a gimple_match_op *.
22272         (maybe_build_generic_op): Likewise.
22273         (gimple_simplified_result_is_gimple_val): Replace parameters with
22274         a const gimple_match_op *.
22275         (gimple_simplify): Replace code_helper * and tree * parameters with
22276         a gimple_match_op * parameter.
22277         (gimple_resimplify1): Replace code_helper *, tree and tree *
22278         parameters with a gimple_match_op * parameter.
22279         (gimple_resimplify2): Likewise.
22280         (gimple_resimplify3): Likewise.
22281         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
22282         parameters with a gimple_match_op * parameter.
22283         * gimple-match-head.c (gimple_simplify): Change prototypes of
22284         auto-generated functions to take a gimple_match_op * instead of
22285         separate code_helper * and tree * parameters.  Make the same
22286         change in the top-level overload and update calls to the
22287         gimple_resimplify routines.  Update calls to the auto-generated
22288         functions and to maybe_push_res_to_seq in the publicly-facing
22289         operation-specific gimple_simplify overloads.
22290         (gimple_match_op::MAX_NUM_OPS): Define.
22291         (gimple_resimplify1): Replace rcode and ops with a single res_op
22292         parameter.  Update call to gimple_simplify.
22293         (gimple_resimplify2): Likewise.
22294         (gimple_resimplify3): Likewise.
22295         (mprts_hook): Replace parameters with a gimple_match_op *.
22296         (maybe_build_generic_op): Likewise.
22297         (build_call_internal): Replace type, nargs and ops with
22298         a gimple_match_op *.
22299         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
22300         with a single gimple_match_op *.  Update calls to mprts_hook,
22301         build_call_internal and gimple_simplified_result_is_gimple_val.
22302         Factor out code that is common to the tree_code and combined_fn cases.
22303         * genmatch.c (expr::gen_transform): Replace tem_code and
22304         tem_ops with a gimple_match_op called tem_op.  Update calls
22305         to the gimple_resimplify functions and maybe_push_res_to_seq.
22306         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
22307         res_ops.  Update call to the gimple_resimplify functions.
22308         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
22309         (decision_tree::gen): Make the functions take a gimple_match_op *
22310         called res_op instead of separate res_code and res_ops parameters.
22311         Update call accordingly.
22312         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
22313         and ops with a single res_op parameter.  Update calls to
22314         maybe_build_generic_op and maybe_push_res_to_seq.
22315         (fold_stmt_1): Update calls to gimple_simplify and
22316         replace_stmt_with_simplification.
22317         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
22318         and gimple_simplified_result_is_gimple_val.
22319         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
22320         gimple_simplify.
22321         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
22322         with a gimple_match_op *.
22323         (vn_nary_build_or_lookup): Likewise.  Update call to
22324         vn_nary_build_or_lookup_1.
22325         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
22326         gimple_match_op *.  Update calls to the gimple_resimplify routines
22327         and to gimple_simplified_result_is_gimple_val.
22328         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
22329         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
22330         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
22331         (visit_nary_op): Likewise.
22332         (visit_reference_op_load): Likewise.
22334 2018-05-23  Luis Machado  <luis.machado@linaro.org>
22336         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
22337         modifier for printing the step amount.
22339 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
22341         PR target/78849
22342         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
22343         types.
22345 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
22347         * doc/sourcebuild.texi (Endianness): New subsubsection.
22349 2018-05-23  Luis Machado  <luis.machado@linaro.org>
22351         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22352         <prefetch_dynamic_strides>: New const bool field.
22353         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22354         prefetch_dynamic_strides.
22355         (exynosm1_prefetch_tune): Likewise.
22356         (thunderxt88_prefetch_tune): Likewise.
22357         (thunderx_prefetch_tune): Likewise.
22358         (thunderx2t99_prefetch_tune): Likewise.
22359         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
22360         false.
22361         (aarch64_override_options_internal): Update to set
22362         PARAM_PREFETCH_DYNAMIC_STRIDES.
22363         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22364         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22365         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22366         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22367         prefetch-dynamic-strides setting.
22369 2018-05-23  Luis Machado  <luis.machado@linaro.org>
22371         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22372         <minimum_stride>: New const int field.
22373         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22374         minimum_stride field defaulting to -1.
22375         (exynosm1_prefetch_tune): Likewise.
22376         (thunderxt88_prefetch_tune): Likewise.
22377         (thunderx_prefetch_tune): Likewise.
22378         (thunderx2t99_prefetch_tune): Likewise.
22379         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
22380         <default_opt_level>: Set to 3.
22381         (aarch64_override_options_internal): Update to set
22382         PARAM_PREFETCH_MINIMUM_STRIDE.
22383         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22384         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22385         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22386         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22387         stride is constant and is below the minimum stride threshold.
22389 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22391         * config/arm/arm-cpus.in (mode26): Delete.
22392         (armv4): Delete mode26 reference.
22393         * config/arm/arm.c (arm_configure_build_target): Delete use of
22394         isa_bit_mode26.
22396 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
22398         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
22399         New insn pattern.
22400         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
22401         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
22402         for non-SSE modes.
22403         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
22404         (floatunsdidf2): Ditto.
22406 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
22408         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
22409         (fixuns_trunc<mode>si2_avx512f): Ditto.
22410         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
22411         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
22412         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
22414 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
22416         PR rtl-optimization/79985
22417         * df-scan.c (df_insn_refs_collect): Remove special case for
22418         global registers and asm statements.
22420 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
22422         * extend.texi (Global Register Variables): Rewrite the bullet list.
22423         Note that the register is available for allocation. Note that access
22424         via inline asm must use constraints. Add note about async-signal
22425         handlers. Remove paragraph about automagic register selection.
22427 2018-05-23  Richard Biener  <rguenther@suse.de>
22429         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
22430         of fixed offset from memset VN.
22432 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
22434         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
22435         first_interp field.
22436         (alloc_cand_and_find_basis): Initialize first_interp field.
22437         (slsr_process_mul): Modify first_interp field.
22438         (slsr_process_add): Likewise.
22439         (slsr_process_cast): Modify first_interp field for each new
22440         interpretation.
22441         (slsr_process_copy): Likewise.
22442         (dump_candidate): Dump first_interp field.
22443         (replace_mult_candidate): Process all interpretations, not just
22444         subsequent ones.
22445         (replace_rhs_if_not_dup): Likewise.
22446         (replace_one_candidate): Likewise.
22448 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
22450         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
22451         Add new boolean.
22452         (aarch64_needs_frame_chain): New function.
22453         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
22455 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
22457         PR target/84882
22458         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
22459         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
22460         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
22461         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
22462         as true for strict-align.
22463         (aarch64_can_inline_p): Perform checks even when callee has no
22464         attributes to check for strict alignment.
22465         * doc/extend.texi (AArch64 Function Attributes): Document
22466         no-strict-align.
22467         * doc/invoke.texi: (AArch64 Options): Likewise.
22469 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
22471         PR tree-optimization/85853
22472         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
22473         the handling of the root of the node to...
22474         (vect_slp_analyze_node_operations_1): ...this new function,
22475         and run the whole thing with the child nodes' def types
22476         set according to their SLP node's def type.
22478 2018-05-23  Richard Biener  <rguenther@suse.de>
22480         PR middle-end/85874
22481         * tree-data-ref.c (create_runtime_alias_checks): Defer
22482         and ignore overflow warnings.
22484 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
22486         PR tree-optimization/85822
22487         * tree-vrp.c (is_masked_range_test): Fix handling of negative
22488         constants.
22490 2018-05-23  Richard Biener  <rguenther@suse.de>
22492         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
22493         memset constants via native_interpret_expr.
22495 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
22497         PR target/85345
22498         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
22499         attribute.
22500         (cgraph_node::create_alias): Likewise.
22501         (cgraph_node::get_availability): Check ifunc_resolver instead
22502         of looking up ifunc attribute.
22503         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
22504         * varasm.c (do_assemble_alias): Likewise.
22505         (assemble_alias): Likewise.
22506         (default_binds_local_p_3): Likewise.
22507         * cgraph.h (cgraph_node): Add ifunc_resolver.
22508         (cgraph_node::only_called_directly_or_aliased_p): Return false
22509         for IFUNC resolver.
22510         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
22511         attribute.
22512         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
22513         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
22514         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
22515         instead of looking up ifunc attribute.
22517 2018-05-22  Luis Machado  <luis.machado@linaro.org>
22519         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
22521 2018-05-22  Martin Sebor  <msebor@redhat.com>
22523         PR middle-end/85359
22524         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
22525         only when expasion succeeds.
22526         (expand_builtin_strcmp): Same.
22527         (expand_builtin_strncmp): Same.
22529 2018-05-22  Martin Sebor  <msebor@redhat.com>
22531         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
22533 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
22534             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22536         * config/aarch64/aarch64-ldpstp.md: Replace uses of
22537         aarch64_mem_pair_operand with memory_operand and delete operand swapping
22538         code.
22539         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22540         Add check for legitimate_address.
22541         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
22542         (aarch64_swap_ldrstr_operands): New.
22543         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
22544         Define prototype.
22546 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
22547             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22549         * config/aarch64/aarch64.md: New patterns to generate stp
22550         and ldp.
22551         (store_pair_sw, store_pair_dw): New patterns to generate stp for
22552         single words and double words.
22553         (load_pair_sw, load_pair_dw): Likewise.
22554         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
22555         Delete.
22556         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
22557         Delete.
22558         * config/aarch64/aarch64-ldpstp.md: Modify peephole
22559         for different mode ldpstp and add peephole for merged zero stores.
22560         Likewise for loads.
22561         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22562         Add size check.
22563         (aarch64_gen_store_pair): Rename calls to match new patterns.
22564         (aarch64_gen_load_pair): Rename calls to match new patterns.
22565         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
22566         (load_pair<DREG:mode><DREG2:mode>): ... This.
22567         (store_pair<mode>): Rename to...
22568         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
22569         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
22570         New mode iterators.
22571         (V_INT_EQUIV): Handle SImode.
22572         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
22573         New predicate.
22575 2018-05-22  Martin Sebor  <msebor@redhat.com>
22577         PR c/85623
22578         * calls.c (maybe_warn_nonstring_arg): Use string length to set
22579         or ajust the presumed bound on an operation to avoid unnecessary
22580         warnings.
22582 2018-05-22  Martin Sebor  <msebor@redhat.com>
22584         PR tree-optimization/85826
22585         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
22586         assuming that a DECL necesarily has a constant size.
22588 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
22590         PR middle-end/85862
22591         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
22593 2018-05-22  Richard Biener  <rguenther@suse.de>
22595         PR tree-optimization/85834
22596         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
22597         non-constant and non-zero memset arguments.
22599 2018-05-22  Martin Liska  <mliska@suse.cz>
22601         PR ipa/85607
22602         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
22604 2018-05-22  Richard Biener  <rguenther@suse.de>
22606         PR tree-optimization/85863
22607         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
22608         comparisons when vectype is specified.
22609         (vectorizable_condition): Do not specify vectype for
22610         vect_is_simple_cond when SLP vectorizing.
22612 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
22614         PR target/85657
22615         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
22616         define __ibm128 as long double.
22617         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
22618         as a distinct type when IEEE 128-bit support is enabled.
22619         (init_float128_ieee): Fix up conversions between IFmode and IEEE
22620         128-bit types to use the correct functions.
22621         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
22622         convert between 128-bit floating point types that have different
22623         modes but the same representation, instead of using gen_lowpart to
22624         makean alias.
22625         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
22626         KFmode.
22627         (IFKF_reg): New attributes to give the register constraints for
22628         IFmode and KFmode.
22629         (extend<mode>tf2_internal): New insns to mark an explicit
22630         conversion between 128-bit floating point types that have a
22631         different mode but share the same representation.
22633 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
22635         PR tree-optimization/85814
22636         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
22637         a null return from get_strinfo when unsharing the next
22638         strinfo in the chain.
22640 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
22642         PR gcc/84923
22643         * varasm.c (weak_finish): Clean up weak_decls.
22645 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22647         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
22648         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
22649         UNSPEC_UADALP values.
22650         * config/aarch64/iterators.md (ABAL): New int iterator.
22651         (ABDL2): Likewise.
22652         (ADALP): Likewise.
22653         (sur): Add mappings for the above.
22654         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22655         New define_insn.
22656         (aarch64_<sur>abal<mode>_4): Likewise.
22657         (aarch64_<sur>adalp<mode>_3): Likewise.
22658         (<sur>sadv16qi): New define_expand.
22660 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
22662         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
22663         (*movdf_internal): Ditto.
22664         (*rcpsf2_sse): Ditto.
22665         (*rsqrtsf2_sse): Ditto.
22666         (*sqrt<mode>2_sse): Ditto.
22668 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
22670         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
22671         eor3q<mode>4.
22672         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
22673         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
22674         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
22675         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
22676         vbcaxq_s64): New.
22677         * config/aarch64/arm_neon.h: Likewise.
22678         * config/aarch64/iterators.md (VQ_I): New.
22680 2018-05-21  Alexey Brodkin  <abrodkin@synopsys.com>
22682         * config.gcc: Add arc/t-multilib-linux to tmake_file for
22683         arc*-*-linux*.
22684         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
22685         MULTILIB_DIRNAMES
22687 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
22689         * config/nds32/constraints.md (S): New constraint.
22690         * config/nds32/nds32.md (call_internal): Use constraint S.
22691         (call_value_internal): Likewise.
22692         (sibcall_internal): Likewise.
22693         (sibcall_value_internal): Likewise.
22695 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
22696             Chung-Ju Wu  <jasonwucj@gmail.com>
22698         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
22699         into consideration.
22701 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
22702             Chung-Ju Wu  <jasonwucj@gmail.com>
22704         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
22705         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
22706         (nds32_rtx_costs_impl): Simplify.
22707         (nds32_address_cost_impl): Simplify.
22708         (nds32_init_rtx_costs): New function.
22709         (nds32_rtx_costs_speed_prefer): Likewise.
22710         (nds32_rtx_costs_size_prefer): Likewise.
22711         (nds32_address_cost_speed_prefer): Likewise.
22712         (nds32_address_cost_speed_fwprop): Likewise.
22713         (nds32_address_cost_size_prefer): Likewise.
22714         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
22715         * config/nds32/nds32.c (nds32_option_override): Use
22716         nds32_init_rtx_costs function.
22718 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
22720         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
22721         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
22722         (TARGET_PIPELINE_N8): Likewise.
22723         (TARGET_PIPELINE_N10): Likewise.
22724         (TARGET_PIPELINE_N13): Likewise.
22725         (TARGET_PIPELINE_GRAYWOLF): Likewise.
22727 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
22729         * config/nds32/nds32-fpu.md: Update copyright year.
22731 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22733         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
22735 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22737         * config/nds32/nds32.c
22738         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
22739         * config/nds32/nds32.opt (minline-asm-r15): New option.
22741 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22743         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
22744         MASK_HW_ABS.
22745         * config/nds32/nds32.md (abssi2): New pattern.
22747 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
22749         * config/i386/i386.md (rex64namesuffix): New mode attribute.
22750         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
22751         Merge insn pattern from sse_cvtsi2ss<round_name> and
22752         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
22753         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
22754         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
22755         using SWI48 mode iterator.
22756         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
22757         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
22758         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
22759         pattern from sse_cvttss2si<round_saeonly_name>
22760         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
22761         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
22762         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
22763         using SWI48 mode iterator.
22764         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
22765         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
22766         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
22767         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
22768         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
22769         using SWI48 mode iterator.
22770         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
22771         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
22772         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
22773         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
22774         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
22775         SWI48 mode iterator.
22776         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
22777         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
22778         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
22779         pattern from sse_cvttsd2si<round_saeonly_name>
22780         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
22782 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22784         * config/nds32/nds32-md-auxiliary.c
22785         (nds32_valid_smw_lwm_base_p): Refine.
22786         (nds32_output_smw_single_word): Refine.
22787         (nds32_output_smw_double_word): New.
22788         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
22790 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22792         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
22793         (nds32_output_stack_pop): Refine.
22794         (nds32_expand_unaligned_load): Refine.
22795         (nds32_expand_unaligned_store): Refine.
22797 2018-05-19  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22798             Chung-Ju Wu  <jasonwucj@gmail.com>
22800         * config/nds32/constants.md: Add TP_REGNUM constant.
22801         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
22802         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
22803         UNSPEC_ADD32.
22804         * config/nds32/nds32-doubleword.md: Consider flag_pic.
22805         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
22806         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
22807         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
22808         and PIC code generation.
22809         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
22810         code generation.
22811         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
22812         optimization.
22813         * config/nds32/nds32.md: Support TLS and PIC.
22814         * config/nds32/nds32.c: Support TLS and PIC.
22815         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
22816         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
22817         predicate.
22819 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22821         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
22822         mode with E_ prefix.
22824 2018-05-19  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22825             Chung-Ju Wu  <jasonwucj@gmail.com>
22827         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
22828         * config/nds32/nds32-md-auxiliary.c
22829         (symbolic_reference_mentioned_p): New.
22830         (nds32_legitimize_ict_address): New.
22831         (nds32_expand_ict_move): New.
22832         (nds32_indirect_call_referenced_p): New.
22833         (nds32_symbol_binds_local_p): Delete.
22834         (nds32_long_call_p): Modify.
22835         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
22836         * config/nds32/nds32-protos.h
22837         (symbolic_reference_mentioned_p): Declare.
22838         (nds32_legitimize_ict_address): Declare.
22839         (nds32_expand_ict_move): Declare.
22840         (nds32_indirect_call_referenced_p): Declare.
22841         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
22842         (nds32_relax_group): Use nds32_ict_const_p as condition.
22843         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
22844         (nds32_asm_file_start): Output ict_model directive in asm code.
22845         (nds32_legitimate_address_p): Consider indirect call.
22846         (nds32_print_operand): Consider indirect call.
22847         (nds32_print_operand_address): Consider indirect call.
22848         (nds32_insert_attributes): Handle "indirect_call" attribute.
22849         (TARGET_LEGITIMATE_ADDRESS_P): Define.
22850         (TARGET_LEGITIMATE_CONSTANT_P): Define.
22851         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
22852         (TARGET_DELEGITIMIZE_ADDRESS): Define.
22853         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
22854         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
22855         (TARGET_ICT_MODEL_SMALL): Define.
22856         (TARGET_ICT_MODEL_LARGE): Define.
22857         * config/nds32/nds32.md (movsi): Consider ict model.
22858         (call, call_value): Consider ict model.
22859         (sibcall, sibcall_value): Consider ict model.
22860         * config/nds32/nds32.opt (mict-model): New option.
22861         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
22862         model.
22864 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
22865             Monk Chiang  <sh.chiang04@gmail.com>
22866             Jim Wilson  <jimw@sifive.com>
22868         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
22869         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
22870         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
22871         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
22872         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
22873         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
22874         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
22875         compute save_libcall_adjustment properly.
22876         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
22877         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
22878         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
22879         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
22880         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
22881         (ABI_SPEC): Handle mabi=ilp32e.
22882         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
22883         (RVE): Add RVE mask.
22884         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
22885         <-march>: Add rv32e as an example.
22887 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
22889         PR c++/82899
22890         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
22891         (intra_create_variable_infos): Handle C++ constructors.
22893 2018-05-18  Martin Liska  <mliska@suse.cz>
22895         * passes.def: Remove a redundant pass.
22897 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
22899         PR bootstrap/85838
22900         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
22902 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22904         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
22905         (ARMv4): Update.
22906         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
22907         (ARMv6m): Update.
22908         (armv2, armv2a, armv3, armv3m): Delete architectures.
22909         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
22910         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
22911         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
22912         Delete cpus.
22913         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
22914         (*mulsidi3adddi): Likewise.
22915         (mulsidi3): Likewise.
22916         (*mulsidi3_nov6): Likewise.
22917         (umulsidi3): Likewise.
22918         (umulsidi3_nov6): Likewise.
22919         (umaddsidi4): Likewise.
22920         (*umulsidi3adddi): Likewise.
22921         (smulsi3_highpart): Likewise.
22922         (*smulsi3_highpart_nov6): Likewise.
22923         (umulsi3_highpart): Likewise.
22924         (*umulsi3_highpart_nov6): Likewise.
22925         * config/arm/arm.h (arm_arch3m): Delete.
22926         * config/arm/arm.c (arm_arch3m): Delete.
22927         (arm_option_override_internal): Update armv3-related comment.
22928         (arm_configure_build_target): Delete use of isa_bit_mode32.
22929         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
22930         (arm_rtx_costs_internal): Delete check of arm_arch3m.
22931         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
22932         (mulsa3): Likewise.
22933         (mulusa3): Likewise.
22934         * config/arm/arm-protos.h (arm_arch3m): Delete.
22935         * config/arm/arm-tables.opt: Regenerate.
22936         * config/arm/arm-tune.md: Likewise.
22937         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
22938         deleted architectures.
22940 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22942         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
22943         (armv5t, armv5te): New features.
22944         (ARMv5, ARMv5e): Delete fgroups.
22945         (ARMv5t, ARMv5te): Adjust for above changes.
22946         (ARMv6m): Likewise.
22947         (armv5, armv5e): Delete arches.
22948         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
22949         arm_arch5.
22950         (*call_reg_arm): Likewise.
22951         (*call_value_reg_armv5): Likewise.
22952         (*call_value_reg_arm): Likewise.
22953         (*call_symbol): Likewise.
22954         (*call_value_symbol): Likewise.
22955         (*sibcall_insn): Likewise.
22956         (*sibcall_value_insn): Likewise.
22957         (clzsi2): Likewise.
22958         (prefetch): Likewise.
22959         (define_split and define_peephole2 dependent on arm_arch5):
22960         Likewise.
22961         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
22962         arm_arch5e.
22963         (TARGET_ARM_QBIT): Likewise.
22964         (TARGET_DSP_MULTIPLY): Likewise.
22965         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
22966         (arm_arch5, arm_arch5e): Delete.
22967         (arm_arch5t, arm_arch5te): Declare.
22968         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
22969         (arm_arch5t): Declare.
22970         (arm_option_reconfigure_globals): Update for the above.
22971         (arm_options_perform_arch_sanity_checks): Update comment, replace
22972         use of arm_arch5 with arm_arch5t.
22973         (use_return_insn): Likewise.
22974         (arm_emit_call_insn): Likewise.
22975         (output_return_instruction): Likewise.
22976         (arm_final_prescan_insn): Likewise.
22977         (arm_coproc_builtin_available): Likewise.
22978         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
22979         arm_arch5e with arm_arch5t and arm_arch5te.
22980         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
22981         (arm_arch5t, arm_arch5te): Declare.
22982         * config/arm/arm-tables.opt: Regenerate.
22983         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
22984         * config/arm/t-multilib: Likewise.
22985         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
22986         instead of arm_arch5.
22987         (*call_reg_thumb1): Likewise.
22988         (*call_value_reg_thumb1_v5): Likewise.
22989         (*call_value_reg_thumb1): Likewise.
22990         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
22991         unreachable path.
22992         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
22994 2018-05-18  Martin Liska  <mliska@suse.cz>
22996         PR gcov-profile/84846
22997         * doc/gcov.texi: Document -t option of gcov tool.
22999 2018-05-18  Martin Liska  <mliska@suse.cz>
23001         PR gcov-profile/84846
23002         * gcov.c (print_usage): Add new -t option.
23003         (process_args): Handle the option.
23004         (generate_results): Use stdout as output when requested by
23005         the option.
23007 2018-05-18  Martin Liska  <mliska@suse.cz>
23009         PR gcov-profile/84846
23010         * coverage.c (coverage_init): Write PWD to .gcno file.
23011         * doc/gcov.texi: Document how working directory is printed.
23012         * gcov-dump.c (dump_gcov_file): Print PWD.
23013         * gcov.c (output_intermediate_file): Likewise.
23014         (read_graph_file): Read PWD string.
23015         (output_lines): Print PWD.
23017 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23019         PR middle-end/85817
23020         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
23021         for retval and return false if all args to phi are zero.
23023 2018-05-18  Richard Biener  <rguenther@suse.de>
23025         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
23026         method.
23027         (evrp_dom_walker::before_dom_children): Call it.
23029 2018-05-18  Richard Biener  <rguenther@suse.de>
23031         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
23032         results when processing array refs with variable index.
23034 2018-05-18  Toon Moene  <toon@moene.org>
23036         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
23037         directly after that of -floop-interchange. Indicate that both
23038         options are enabled by default when specifying -O3.
23040 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23042         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
23043         iterator.  Delete separate integer-mode vec_set<mode> expander.
23044         (aarch64_simd_vec_setv2di): Delete.
23045         (vec_setv2di): Delete.
23046         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
23047         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
23048         the "w, r" alternative.
23050 2018-05-18  Martin Liska  <mliska@suse.cz>
23052         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
23053         * tree-pass.h (make_pass_lower_switch_O0): New function.
23054         * tree-switch-conversion.c (node_has_low_bound): Remove.
23055         (node_has_high_bound): Likewise.
23056         (node_is_bounded): Likewise.
23057         (class pass_lower_switch): Make it a template type and create
23058         two instances.
23059         (pass_lower_switch::execute): Add template argument.
23060         (make_pass_lower_switch): New function.
23061         (make_pass_lower_switch_O0): New function.
23062         (do_jump_if_equal): Remove.
23063         (emit_case_nodes): Simplify to just handle all 3 cases and leave
23064         all the hard work to tree optimization passes.
23066 2018-05-18  Martin Liska  <mliska@suse.cz>
23068         * dbgcnt.c (limit_low): Renamed from limit.
23069         (limit_high): New variable.
23070         (dbg_cnt_is_enabled): Check for upper limit.
23071         (dbg_cnt): Adjust dumping.
23072         (dbg_cnt_set_limit_by_index): Add new argument for high
23073         value.
23074         (dbg_cnt_set_limit_by_name): Likewise.
23075         (dbg_cnt_process_single_pair): Parse new format.
23076         (dbg_cnt_process_opt): Use strtok.
23077         (dbg_cnt_list_all_counters): Remove 'value' and add
23078         'limit_high'.
23079         * doc/invoke.texi: Document changes.
23081 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
23083         * doc/sourcebuild.texi (scalar_all_fma): Document.
23084         * tree.def (FMA_EXPR): Delete.
23085         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
23086         * internal-fn.c (ternary_direct): New macro.
23087         (expand_ternary_optab_fn): Likewise.
23088         (direct_ternary_optab_supported_p): Likewise.
23089         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
23090         * builtins.c (fold_builtin_fma): Delete.
23091         (fold_builtin_3): Don't call it.
23092         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
23093         * expr.c (expand_expr_real_2): Likewise.
23094         * fold-const.c (operand_equal_p): Likewise.
23095         (fold_ternary_loc): Likewise.
23096         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
23097         * gimple.c (DEFTREECODE): Likewise.
23098         * gimplify.c (gimplify_expr): Likewise.
23099         * optabs-tree.c (optab_for_tree_code): Likewise.
23100         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
23101         * tree-eh.c (operation_could_trap_p): Likewise.
23102         (stmt_could_throw_1_p): Likewise.
23103         * tree-inline.c (estimate_operator_cost): Likewise.
23104         * tree-pretty-print.c (dump_generic_node): Likewise.
23105         (op_code_prio): Likewise.
23106         * tree-ssa-loop-im.c (stmt_cost): Likewise.
23107         * tree-ssa-operands.c (get_expr_operands): Likewise.
23108         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
23109         * fold-const-call.h (fold_fma): Delete.
23110         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
23111         CFN_FNMA and CFN_FNMS.
23112         (fold_fma): Delete.
23113         * genmatch.c (combined_fn): New enum.
23114         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
23115         (commutative_op): New function.
23116         (commutate): Use it.  Handle more than 2 operands.
23117         (dt_operand::gen_gimple_expr): Use commutative_op.
23118         (parser::parse_expr): Allow :c to be used with non-binary
23119         operators if the commutative operand is known.
23120         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
23121         CFN_FMS, CFN_FNMA and CFN_FNMS.
23122         (backprop::process_assign_use): Remove FMA_EXPR handling.
23123         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
23124         (gen_hsa_fma): New function.
23125         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
23126         IFN_FNMA and IFN_FNMS.
23127         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
23128         * gimple-fold.h (follow_all_ssa_edges): Declare.
23129         * gimple-fold.c (follow_all_ssa_edges): New function.
23130         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
23131         gimple_build interface and use follow_all_ssa_edges to fold the result.
23132         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
23133         instead of checking for optabs directly.
23134         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
23135         rather than FMA_EXPRs.
23136         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
23137         call to IFN_FMA instead of an FMA_EXPR.
23139 2018-05-17  Jim Wilson  <jimw@sifive.com>
23141         * expr.c (do_tablejump): When converting index to Pmode, if we have a
23142         sign extended promoted subreg, and the range does not have the sign bit
23143         set, then do a sign extend.
23145         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
23146         test, check for sign extended subreg and/or constant operands, and
23147         do a sign extend in that case.
23149 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
23151         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
23152         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
23153         Add untyped.
23154         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
23155         Change logics_shift_reg to logics_shift_imm.
23156         (thunderx2t99_fp_loadpair_basic): Delete.
23157         (thunderx2t99_fp_storepair_basic): Delete.
23158         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
23159         (thunderx2t99_asimd_polynomial): Delete.
23160         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
23161         and neon_fp_mul_d_scalar_q.
23162         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
23163         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
23164         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
23165         (thunderx2t99_asimd_lut): Add missing tbl types.
23166         (thunderx2t99_asimd_ext): Delete.
23167         (thunderx2t99_asimd_load1_1_mult): Delete.
23168         (thunderx2t99_asimd_load1_2_mult): Delete.
23169         (thunderx2t99_asimd_load1_ldp): New.
23170         (thunderx2t99_asimd_load1): New.
23171         (thunderx2t99_asimd_load2): Add missing *load2* types.
23172         (thunderx2t99_asimd_load3): New.
23173         (thunderx2t99_asimd_load4): New.
23174         (thunderx2t99_asimd_store1_1_mult): Delete.
23175         (thunderx2t99_asimd_store1_2_mult): Delete.
23176         (thunderx2t99_asimd_store2_mult): Delete.
23177         (thunderx2t99_asimd_store2_onelane): Delete.
23178         (thunderx2t99_asimd_store_stp): New.
23179         (thunderx2t99_asimd_store1): New.
23180         (thunderx2t99_asimd_store2): New.
23181         (thunderx2t99_asimd_store3): New.
23182         (thunderx2t99_asimd_store4): New.
23184 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
23186         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
23187         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
23189 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
23190             Segher Boessenkool  <segher@kernel.crashing.org>
23192         PR target/85698
23193         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
23194         operand.
23196 2018-05-17  Richard Biener  <rguenther@suse.de>
23198         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
23199         for pruning loop and prune defs feeding only already visited PHIs.
23201 2018-05-17  Richard Biener  <rguenther@suse.de>
23203         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
23205 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
23206             Richard Biener  <rguenther@suse.de>
23208         PR tree-optimization/85793
23209         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
23210         for VMAT_ELEMENTWISE.
23212 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
23214         * internal-fn.h (lookup_internal_fn): Declare
23215         * internal-fn.c (lookup_internal_fn): New function.
23216         * gimple.c (gimple_build_call_from_tree): Handle calls to
23217         internal functions.
23218         * gimple-pretty-print.c (dump_gimple_call): Print "." before
23219         internal function names.
23220         * tree-pretty-print.c (dump_generic_node): Likewise.
23221         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
23223 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
23225         * gimple-fold.h (gimple_build): Make the function forms take
23226         combined_fn rather than built_in_function.
23227         (gimple_simplify): Likewise.
23228         * gimple-match-head.c (gimple_simplify): Likewise.
23229         * gimple-fold.c (gimple_build): Likewise.
23230         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
23231         rather than gimple_build_call_internal.
23232         (get_initial_defs_for_reduction): Likewise.
23233         (vect_create_epilog_for_reduction): Likewise.
23234         (vectorizable_live_operation): Likewise.
23236 2018-05-17  Martin Liska  <mliska@suse.cz>
23238         * gimple-ssa-sprintf.c (format_directive): Do not use
23239         space in between 'G_' and '('.
23241 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
23243         PR target/85323
23244         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
23245         even if the mask is not all ones.
23247         PR target/85323
23248         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
23249         vector.
23250         (ix86_gimple_fold_builtin): Likewise.
23252         PR target/85323
23253         * config/i386/i386.c: Include tree-vector-builder.h.
23254         (ix86_vector_shift_count): New function.
23255         (ix86_fold_builtin): Fold shift builtins by scalar count.
23256         (ix86_gimple_fold_builtin): Likewise.
23258         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
23259         _mm512_setzero): New intrinsics.
23261 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
23262             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23264         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
23265         code generation for cases where splatting a value is not useful.
23266         * simplify-rtx.c (simplify_ternary_operation): Simplify
23267         vec_merge across a vec_duplicate and a paradoxical subreg forming
23268         a vector mode to a vec_concat.
23270 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
23272         * config.gcc: Support "goldmont-plus".
23273         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
23274         "goldmont-plus".
23275         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23276         PROCESSOR_GOLDMONT_PLUS.
23277         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
23278         (processor_target_table): Add "goldmont-plus".
23279         (PTA_GOLDMONT_PLUS): Define.
23280         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
23281         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
23282         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
23283         (fold_builtin_cpu): Add "goldmont-plus".
23284         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
23285         (ix86_option_override_internal): Add "goldmont-plus".
23286         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
23287         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
23288         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
23289         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
23291 2018-05-17  Richard Biener  <rguenther@suse.de>
23293         PR tree-optimization/85757
23294         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
23295         remove defs that only feed that PHI from further processing.
23297 2018-05-16  Jim Wilson  <jimw@sifive.com>
23299         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
23300         asterisk to name.
23301         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
23303 2018-05-16  Mark Wielaard  <mark@klomp.org>
23305         * dwarf2out.c (count_index_strings): New function.
23306         (output_indirect_strings): Call count_index_strings and generate
23307         header for dwarf_version >= 5.
23309 2018-05-16  Mark Wielaard  <mark@klomp.org>
23311         * dwarf2out.c (dwarf_FORM): New function.
23312         (set_indirect_string): Use dwarf_FORM.
23313         (reset_indirect_string): Likewise.
23314         (size_of_die): Likewise.
23315         (value_format): Likewise.
23316         (output_die): Likewise.
23317         (add_skeleton_AT_string): Likewise.
23318         (output_macinfo_op): Likewise.
23319         (index_string): Likewise.
23320         (output_index_string_offset): Likewise.
23321         (output_index_string): Likewise.
23322         (count_index_strings): Likewise.
23324 2018-05-16  Carl Love  <cel@us.ibm.com>
23326         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23327         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
23329 2018-05-16  Martin Jambor  <mjambor@suse.cz>
23331         * ipa-prop.c (ipa_free_all_edge_args): Remove.
23332         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
23334 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
23336         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
23337         (fnma<mode>4): Likewise.
23338         (fms<mode>4): Likewise.
23339         (fnms<mode>4): Likewise.
23340         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
23341         (aarch64_fnma<mode>4): Likewise.
23342         (aarch64_fms<mode>4): Likewise.
23343         (aarch64_fnms<mode>4): Likewise.
23344         (aarch64_fnmadd<mode>4): Likewise.
23346 2018-05-16  Jason Merrill  <jason@redhat.com>
23348         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
23350 2018-05-16  Richard Biener  <rguenther@suse.de>
23352         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
23353         (dump_stmt_cost): Declare.
23354         (add_stmt_cost): Dump cost we add.
23355         (add_stmt_costs): New function.
23356         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
23357         No longer exported.
23358         (vect_analyze_stmt): Adjust prototype.
23359         (vectorizable_condition): Likewise.
23360         (vectorizable_live_operation): Likewise.
23361         (vectorizable_reduction): Likewise.
23362         (vectorizable_induction): Likewise.
23363         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
23364         cost vector to pass to vectorizable_ and record afterwards.
23365         (vect_model_reduction_cost): Take cost vector argument and adjust.
23366         (vect_model_induction_cost): Likewise.
23367         (vectorizable_reduction): Likewise.
23368         (vectorizable_induction): Likewise.
23369         (vectorizable_live_operation): Likewise.
23370         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
23371         SLP_TREE_NUMBER_OF_VEC_STMTS.
23372         (vect_analyze_slp_cost_1): Remove.
23373         (vect_analyze_slp_cost): Likewise.
23374         (vect_slp_analyze_node_operations): Take visited args and
23375         a target cost vector.  Avoid processing already visited stmt sets.
23376         (vect_slp_analyze_operations): Use a local cost vector to gather
23377         costs and register those of non-discarded instances.
23378         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
23379         (vect_schedule_slp_instance): Remove copying of
23380         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
23381         zero.
23382         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
23383         adding cost.  Record cost entry location.
23384         (vect_prologue_cost_for_slp_op): Function to compute cost of
23385         a constant or invariant generated for SLP vect in the prologue,
23386         split out from vect_analyze_slp_cost_1.
23387         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
23388         (vect_model_promotion_demotion_cost): Likewise.
23389         (vect_model_store_cost): Likewise, make static.
23390         (vect_model_load_cost): Likewise.
23391         (vectorizable_bswap): Add cost vector arg and adjust.
23392         (vectorizable_call): Likewise.
23393         (vectorizable_simd_clone_call): Likewise.
23394         (vectorizable_conversion): Likewise.
23395         (vectorizable_assignment): Likewise.
23396         (vectorizable_shift): Likewise.
23397         (vectorizable_operation): Likewise.
23398         (vectorizable_store): Likewise.
23399         (vectorizable_load): Likewise.
23400         (vectorizable_condition): Likewise.
23401         (vectorizable_comparison): Likewise.
23402         (can_vectorize_live_stmts): Likewise.
23403         (vect_analyze_stmt): Likewise.
23404         (vect_transform_stmt): Adjust calls to vectorizable_*.
23405         * tree-vectorizer.c: Include gimple-pretty-print.h.
23406         (dump_stmt_cost): New function.
23408 2018-05-16  Richard Biener  <rguenther@suse.de>
23410         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
23411         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
23412         * tree-ssa-dse.c: Include tree-ssa-loop.h.
23413         (check_name): New callback.
23414         (dse_classify_store): Track cycles via a visited bitmap of PHI
23415         defs and simplify handling of in-loop and across loop dead stores
23416         and properly fail for loop-variant refs.  Handle byte-tracking with
23417         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
23418         limiting the walk.
23420 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
23422         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
23423         (vect_get_mask_type_for_stmt): Likewise.
23424         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
23425         split out from...
23426         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
23427         to determine the statement's vector type and the vector type that
23428         should be used for calculating nunits.  Deal with cases in which
23429         the type has to be deferred.
23430         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
23431         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
23432         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
23433         (vect_determine_vf_for_stmt): New functions, split out from...
23434         (vect_determine_vectorization_factor): ...here.
23435         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
23436         (vect_get_mask_type_for_stmt): New functions, split out from
23437         vect_determine_vectorization_factor.
23439 2018-05-16  Richard Biener  <rguenther@suse.de>
23441         * tree-cfg.c (verify_gimple_assign_ternary): Properly
23442         verify the [VEC_]COND_EXPR embedded comparison.
23444 2018-05-15  Martin Sebor  <msebor@redhat.com>
23446         PR tree-optimization/85753
23447         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
23448         RECORD_TYPE in addition to ARRAY_TYPE.
23450 2018-05-15  Martin Sebor  <msebor@redhat.com>
23452         PR middle-end/85643
23453         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
23455 2018-05-15  Richard Biener  <rguenther@suse.de>
23457         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
23458         add by_clobber_p one.  Change algorithm to collect all defs
23459         representing uses we need to walk and try reducing them to
23460         a single one before failing.
23461         (dse_dom_walker::dse_optimize_stmt): Adjust.
23463 2018-05-13  Mark Wielaard  <mark@klomp.org>
23465         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
23466         (size_of_loc_descr): Likewise.
23467         (output_loc_operands): Likewise.
23468         (output_loc_operands_raw): Likewise.
23469         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
23470         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
23471         (hash_loc_operands): Likewise.
23472         (compare_loc_operands): Likewise.
23474 2018-05-14  Mark Wielaard  <mark@klomp.org>
23476         * dwarf2out.c (count_index_addrs): New function.
23477         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
23479 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23481         PR tree-optimization/83648
23482         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
23483         return value as malloc candidate.
23485 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23487         PR ipa/85734
23488         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
23489         param as true in call to suggest_attribute.
23491 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
23493         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
23494         -mreadonly-in-sdata.
23496 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23498         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
23499         New pattern.
23500         (aarch64_crypto_aesd_fused): Likewise.
23502 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
23504         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
23505         (movsi_aarch64): Likewise.
23506         (load_pairsi): Likewise.
23507         (load_pairdi): Likewise.
23508         (store_pairsi): Likewise.
23509         (store_pairdi): Likewise.
23510         (load_pairsf): Likewise.
23511         (load_pairdf): Likewise.
23512         (store_pairsf): Likewise.
23513         (store_pairdf): Likewise.
23514         (zero_extend): Likewise.
23515         (trunc): Swap alternatives.
23516         (fcvt_target): Add '?' to prefer w over r.
23518 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
23520         PR target/85756
23521         * config/i386/i386.md: Disallow non-commutative arithmetics in
23522         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
23523         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
23524         in the peephole2 before it.
23526 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
23528         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
23529         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
23530         (ix86_handle_option): Handle -mcldemote.
23531         * config.gcc: New header.
23532         * config/i386/cldemoteintrin.h: New file.
23533         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
23534         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
23535         -mcldemote.
23536         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23537         OPTION_MASK_ISA_CLDEMOTE.
23538         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
23539         (ix86_valid_target_attribute_inner_p): Ditto.
23540         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
23541         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
23542         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
23543         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
23544         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
23545         (cldemote): New.
23546         * config/i386/i386.opt: Add -mcldemote.
23547         * config/i386/x86intrin.h: New header.
23548         * doc/invoke.texi: Add -mcldemote.
23550 2018-05-14  Richard Biener  <rguenther@suse.de>
23552         * doc/match-and-simplify.texi: Adjust :s documentation.
23554 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
23556         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
23557         intended memcpy size.
23558         (REORDER_45): Likewise.
23560 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
23562         * sort.cc: New file.
23563         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
23564         * vec.c (qsort_chk): Use gcc_qsort.
23565         * Makefile.in (OBJS-libcommon): Add sort.o.
23566         (build/sort.o): New target.  Use it...
23567         (BUILD_RTL): ... here, and...
23568         (build/gencfn-macros): ... here, and...
23569         (build/genmatch): ... here.
23571 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
23572             Chung-Ju Wu  <jasonwucj@gmail.com>
23574         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
23575         * config/nds32/nds32-graywolf.md: New file.
23576         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
23577         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
23578         pipeline.
23579         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
23580         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
23581         * config/nds32/nds32.md (pipeline_model): Add graywolf.
23582         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
23583         * config/nds32/pipelines.md: Include n15 settings.
23585 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
23586             Chung-Ju Wu  <jasonwucj@gmail.com>
23588         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
23589         * config/nds32/nds32-n13.md: New file.
23590         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
23591         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
23592         pipeline.
23593         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
23594         * config/nds32/nds32.md (pipeline_model): Add n13.
23595         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
23596         * config/nds32/pipelines.md: Include n13 settings.
23598 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
23599             Chung-Ju Wu  <jasonwucj@gmail.com>
23601         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
23602         * config/nds32/nds32-n10.md: New file.
23603         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
23604         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
23605         pipeline.
23606         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
23607         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
23608         * config/nds32/nds32.md (pipeline_model): Add n10.
23609         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
23610         * config/nds32/pipelines.md: Include n10 settings.
23612 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
23613             Kito Cheng  <kito.cheng@gmail.com>
23614             Chung-Ju Wu  <jasonwucj@gmail.com>
23616         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
23617         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23618         Add enum values for DSP extension instructions.
23619         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
23620         New constraints.
23621         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
23622         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
23623         New code iterators.
23624         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
23625         * config/nds32/nds32-dspext.md: New file for DSP implementation.
23626         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
23627         * config/nds32/nds32-intrinsic.md: Likewise.
23628         * config/nds32/nds32_intrinsic.h: Likewise.
23629         * config/nds32/nds32-md-auxiliary.c: Likewise.
23630         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
23631         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
23632         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
23633         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
23634         * config/nds32/nds32-protos.h: New declarations for DSP extension.
23635         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
23636         TYPE_DMAC in switch statement.
23637         * config/nds32/nds32.c: New checking and implementation for DSP
23638         extension instructions.
23639         * config/nds32/nds32.h: Likewise.
23640         * config/nds32/nds32.md: Likewise.
23641         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
23642         * config/nds32/predicates.md: Implement new predicates for DSP
23643         extension.
23645 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
23647         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
23648         Reformat alternatives and attributes so it is easier to identify
23649         which constraints/attributes go with which instruction.
23650         (mov<mode>_hardfloat32, FMOVE64): Likewise.
23651         (mov<mode>_softfloat32, FMOVE64): Likewise.
23652         (mov<mode>_hardfloat64, FMOVE64): Likewise.
23653         (mov<mode>_softfloat64, FMOVE64): Likewise.
23655 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23657         * doc/extend.texi (PowerPC Built-in Functions): Rename this
23658         subsection.
23659         (Basic PowerPC Built-in Functions): The new name of the
23660         subsection previously known as "PowerPC Built-in Functions".
23661         (Basic PowerPC Built-in Functions Available on all Configurations):
23662         New subsubsection.
23663         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
23664         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
23665         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
23666         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
23668 2018-05-11  Martin Jambor  <mjambor@suse.cz>
23670         PR ipa/85655
23671         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
23672         single const.
23674 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
23676         PR target/85733
23677         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
23679 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
23681         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
23682         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
23683         (ix86_handle_option): Handle -mwaitpkg.
23684         * config.gcc: New header.
23685         * config/i386/cpuid.h (bit_WAITPKG): New bit.
23686         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
23687         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
23688         function type.
23689         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23690         OPTION_MASK_ISA_WAITPKG.
23691         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
23692         (ix86_option_override_internal): Add PTA_WAITPKG.
23693         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
23694         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
23695         IX86_BUILTIN_TPAUSE.
23696         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
23697         __builtin_ia32_umwait and __builtin_ia32_tpause.
23698         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
23699         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
23700         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
23701         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
23702         UNSPECV_TPAUSE): New.
23703         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
23704         * config/i386/i386.opt: Add -mwaitpkg.
23705         * config/i386/waitpkgintrin.h: New file.
23706         * config/i386/x86intrin.h: New header.
23707         * doc/invoke.texi: Add -mwaitpkg.
23709 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
23711         PR target/85606
23712         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
23713         equivalent.
23714         (cortex-m0): Use armv6s-m isa.
23715         (cortex-m0plus): Likewise.
23716         (cortex-m1): Likewise.
23717         (cortex-m0.small-multiply): Likewise.
23718         (cortex-m0plus.small-multiply): Likewise.
23719         (cortex-m1.small-multiply): Likewise.
23721 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
23722             Jakub Jelinek  <jakub@redhat.com>
23724         PR tree-optimization/85692
23725         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
23726         source permute as well.
23728 2018-05-11  Martin Liska  <mliska@suse.cz>
23730         PR sanitizer/85556
23731         * doc/extend.texi: Document LLVM style format for no_sanitize
23732         attribute.
23734 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
23736         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
23737         mode_supports_vsx_dform_quad to mode_supports_dq_form.
23738         (mode_supports_vsx_dform_quad): Likewise.
23739         (mode_supports_vmx_dform): Move these functions to be next to the
23740         other mode_supports functions.
23741         (mode_supports_dq_form): Likewise.
23742         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
23743         mode_supports_dq_form.
23744         (reg_offset_addressing_ok_p): Likewise.
23745         (offsettable_ok_by_alignment): Likewise.
23746         (rs6000_legitimate_offset_address_p): Likewise.
23747         (legitimate_lo_sum_address_p): Likewise.
23748         (rs6000_legitimize_address): Likewise.
23749         (rs6000_legitimize_reload_address): Likewise.
23750         (rs6000_secondary_reload_inner): Likewise.
23751         (rs6000_preferred_reload_class): Likewise.
23752         (rs6000_output_move_128bit): Likewise.
23754 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
23756         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
23757         Generate SImode target register for null target.
23758         <case IX86_BUILTIN_XGETBV>: Ditto.
23759         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
23760         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
23762 2018-05-10  Carl Love  <cel@us.ibm.com>
23764         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23765         dcbtt and dcbtstt if operands[2] is 0.
23767 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
23769         PR target/85693
23770         * config/i386/sse.md (usadv64qi): New expander.
23772 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
23774         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
23775         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
23776         -maltivec=be support.
23777         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
23778         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
23779         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
23780         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
23781         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
23782         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
23783         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
23784         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
23785         altivec_vsumsws): Adjust.
23786         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
23787         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
23788         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
23789         support.
23790         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
23791         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
23792         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
23793         (altivec_lve<VI_char>x): Delete expand.
23794         (*altivec_lve<VI_char>x_internal): Rename to...
23795         (altivec_lve<VI_char>x): ... this.
23796         (altivec_lvxl_<mode>): Delete expand.
23797         (*altivec_lvxl_<mode>_internal): Rename to ...
23798         (altivec_lvxl_<mode>): ... this.
23799         (altivec_stvxl_<mode>): Delete expand.
23800         (*altivec_stvxl_<mode>_internal): Rename to ...
23801         (altivec_stvxl_<mode>): ... this.
23802         (altivec_stve<VI_char>x): Delete expand.
23803         (*altivec_stve<VI_char>x_internal): Rename to ...
23804         (altivec_stve<VI_char>x): ... this.
23805         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
23806         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
23807         reduc_plus_scal_<mode>): Adjust.
23808         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
23809         comment.
23810         (rs6000_cpu_cpp_builtins): Adjust.
23811         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
23812         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
23813         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
23814         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
23815         -maltivec=be support.
23816         (rs6000_split_vec_extract_var): Adjust.
23817         (rs6000_split_v4si_init): Adjust.
23818         (swap_selector_for_mode): Delete.
23819         (altivec_expand_lvx_be, altivec_expand_stvx_be,
23820         altivec_expand_stvex_be): Delete.
23821         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
23822         -maltivec=be support.
23823         (rs6000_gimple_fold_builtin): Ditto.
23824         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
23825         Adjust.
23826         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
23827         (TARGET_DIRECT_MOVE_64BIT): Adjust.
23828         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
23829         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
23830         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
23831         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
23832         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
23833         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
23834         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
23835         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
23836         anonymous split): Adjust.
23837         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
23838         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
23840 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
23842         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
23843         when --with-gxx-include-dir is also specified.
23844         * configure: Regenerate.
23846 2018-05-09  Jim Wilson  <jimw@sifive.com>
23848         PR target/84797
23849         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
23850         * config/riscv/t-withmultilib: New.
23851         * config/riscv/withmultilib.h: New.
23852         * doc/install.texi: Document RISC-V --with-multilib-list support.
23854 2018-05-09  Richard Biener  <rguenther@suse.de>
23856         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
23857         vector.
23858         (vect_bb_vectorization_profitable_p): Adjust.  Compute
23859         actual scalar cost using the cost vector and the add_stmt_cost
23860         machinery.
23862 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23864         PR rtl-optimization/85645
23865         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
23866         in the REG_CFA_REGISTER note for LR, don't leave it empty.
23868 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23870         PR rtl-optimization/85645
23871         * shrink-wrap.c (spread_components): Return a boolean saying if
23872         anything was changed.
23873         (try_shrink_wrapping_separate): Iterate spread_components until
23874         nothing changes anymore.
23876 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23878         PR rtl-optimization/85645
23879         * regrename.c (build_def_use): Also kill the chains that include the
23880         destination of a REG_CFA_REGISTER note.
23882 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23884         PR rtl-optimization/85645
23885         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
23886         insn that has a REG_CFA_REGISTER note.
23888 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
23890         * cfgexpand.c (expand_clobber): New function.
23891         (expand_gimple_stmt_1): Use it.
23892         * tree-vect-stmts.c (vect_clobber_variable): New function,
23893         split out from...
23894         (vectorizable_simd_clone_call): ...here.
23895         (vectorizable_store): Emit a clobber either side of an
23896         IFN_STORE_LANES sequence.
23897         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
23899 2018-05-09  Tom de Vries  <tom@codesourcery.com>
23901         PR target/85626
23902         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
23903         (define_insn "trap_if_false"): Add exit after trap.
23905 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
23907         PR rtl-optimization/85638
23908         * bb-reorder.c: Include common/common-target.h.
23909         (create_forwarder_block): New function extracted from...
23910         (fix_up_crossing_landing_pad): ...here.  Rename into...
23911         (dw2_fix_up_crossing_landing_pad): ...this.
23912         (sjlj_fix_up_crossing_landing_pad): New function.
23913         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
23914         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
23915         from both partitions and exit the loop after one iteration.
23917 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23919         Revert:
23920         * doc/extend.texi (PowerPC Built-in Functions): Rename this
23921         subsection.
23922         (Basic PowerPC Built-in Functions): The new name of the
23923         subsection previously known as "PowerPC Built-in Functions".
23924         (Basic PowerPC Built-in Functions Available on all Configurations):
23925         New subsubsection.
23926         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23927         subsubsection.
23928         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23929         subsubsection.
23930         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23931         subsubsection.
23932         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23933         subsubsection.
23935 2018-05-08  Jim Wilson  <jimw@sifive.com>
23937         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
23938         (LD_EMUL_SUFFIX): New.
23939         (LINK_SPEC): Use it.
23941 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23943         * doc/extend.texi (PowerPC Built-in Functions): Rename this
23944         subsection.
23945         (Basic PowerPC Built-in Functions): The new name of the
23946         subsection previously known as "PowerPC Built-in Functions".
23947         (Basic PowerPC Built-in Functions Available on all Configurations):
23948         New subsubsection.
23949         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23950         subsubsection.
23951         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23952         subsubsection.
23953         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23954         subsubsection.
23955         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23956         subsubsection.
23958 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
23960         PR target/85683
23961         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
23962         after cmpelim optimization.
23964 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
23966         * config.gcc: Support "goldmont".
23967         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
23968         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23969         PROCESSOR_GOLDMONT.
23970         * config/i386/i386.c (m_GOLDMONT): Define.
23971         (processor_target_table): Add "goldmont".
23972         (PTA_GOLDMONT): Define.
23973         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
23974         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
23975         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
23976         (fold_builtin_cpu): Add "goldmont".
23977         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
23978         (ix86_option_override_internal): Add "goldmont".
23979         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
23980         (processor_type): Add PROCESSOR_GOLDMONT.
23981         * config/i386/i386.md: Add CPU "glm".
23982         * config/i386/glm.md: New file.
23983         * config/i386/x86-tune.def: Add m_GOLDMONT.
23984         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
23986 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
23988         PR target/85572
23989         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
23990         E_V4DImode.
23991         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
23992         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
23993         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
23995         PR target/85317
23996         * config/i386/i386.c (ix86_fold_builtin): Handle
23997         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
23999         PR target/85480
24000         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
24001         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
24003 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
24005         PR target/85658
24006         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
24007         (check_arch): Likewise.
24008         (check_fpu): Return the result rather than printing it.
24009         (end arch): Fix operator precedence.
24010         (end cpu): Likewise.
24011         (END): Print the result from check_fpu.
24013 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
24014             Alan Hayward  <alan.hayward@arm.com>
24015             David Sherwood  <david.sherwood@arm.com>
24017         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
24018         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
24019         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
24020         (*fcmuo<mode>_and): New patterns.
24022 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
24024         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
24025         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
24026         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
24027         (cmp_op, sve_imm_con): New code attributes.
24028         (SVE_COND_INT_CMP, imm_con): Delete.
24029         (cmp_op): Remove above unspecs from int attribute.
24030         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
24031         to...
24032         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
24033         comparison-specific unspecs.
24034         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
24035         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
24036         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
24037         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
24038         (*vec_fcm<cmp_op><mode>): Rename to...
24039         (*fcm<cmp_op><mode>): ...this and adjust likewise.
24040         (*vec_fcmuo<mode>): Rename to...
24041         (*fcmuo<mode>): ...this and adjust likewise.
24042         (*pred_fcm<cmp_op><mode>): New pattern.
24043         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
24044         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
24045         functions.
24046         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
24047         and UNORDERED.
24048         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
24049         (aarch64_emit_sve_predicated_cond): New function.
24050         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
24051         (aarch64_emit_unspec_cond_or): Replace with...
24052         (aarch64_emit_sve_or_conds): ...this new function.  Use
24053         aarch64_emit_sve_ptrue_op for the individual comparisons and
24054         aarch64_emit_binop to OR them together.
24055         (aarch64_emit_inverted_unspec_cond): Replace with...
24056         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
24057         aarch64_emit_sve_ptrue_op for the comparison and
24058         aarch64_emit_unop to invert the result.
24059         (aarch64_expand_sve_vec_cmp_float): Update after the above
24060         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
24062 2018-05-07  Nathan Sidwell  <nathan@acm.org>
24064         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
24065         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
24066         (Backwards Compatibility): Likewise.
24068 2018-05-07  Luis Machado  <luis.machado@linaro.org>
24070         PR bootstrap/85681
24071         Revert:
24072         2018-05-07  Luis Machado  <luis.machado@linaro.org>
24074         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
24075         <prefetch_dynamic_strides>: New const bool field.
24076         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
24077         prefetch_dynamic_strides.
24078         (exynosm1_prefetch_tune): Likewise.
24079         (thunderxt88_prefetch_tune): Likewise.
24080         (thunderx_prefetch_tune): Likewise.
24081         (thunderx2t99_prefetch_tune): Likewise.
24082         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
24083         to false.
24084         (aarch64_override_options_internal): Update to set
24085         PARAM_PREFETCH_DYNAMIC_STRIDES.
24086         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
24087         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
24088         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
24089         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
24090         prefetch-dynamic-strides setting.
24092         2018-05-07  Luis Machado  <luis.machado@linaro.org>
24094         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
24095         <minimum_stride>: New const int field.
24096         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
24097         minimum_stride field.
24098         (exynosm1_prefetch_tune): Likewise.
24099         (thunderxt88_prefetch_tune): Likewise.
24100         (thunderx_prefetch_tune): Likewise.
24101         (thunderx2t99_prefetch_tune): Likewise.
24102         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
24103         (aarch64_override_options_internal): Update to set
24104         PARAM_PREFETCH_MINIMUM_STRIDE.
24105         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
24106         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
24107         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
24108         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
24109         stride is constant and is below the minimum stride threshold.
24111 2018-05-07  Luis Machado  <luis.machado@linaro.org>
24113         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
24114         to 512.
24116 2018-05-07  Luis Machado  <luis.machado@linaro.org>
24118         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
24119         <prefetch_dynamic_strides>: New const bool field.
24120         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
24121         prefetch_dynamic_strides.
24122         (exynosm1_prefetch_tune): Likewise.
24123         (thunderxt88_prefetch_tune): Likewise.
24124         (thunderx_prefetch_tune): Likewise.
24125         (thunderx2t99_prefetch_tune): Likewise.
24126         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
24127         to false.
24128         (aarch64_override_options_internal): Update to set
24129         PARAM_PREFETCH_DYNAMIC_STRIDES.
24130         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
24131         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
24132         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
24133         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
24134         prefetch-dynamic-strides setting.
24136 2018-05-07  Luis Machado  <luis.machado@linaro.org>
24138         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
24139         <minimum_stride>: New const int field.
24140         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
24141         minimum_stride field.
24142         (exynosm1_prefetch_tune): Likewise.
24143         (thunderxt88_prefetch_tune): Likewise.
24144         (thunderx_prefetch_tune): Likewise.
24145         (thunderx2t99_prefetch_tune): Likewise.
24146         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
24147         (aarch64_override_options_internal): Update to set
24148         PARAM_PREFETCH_MINIMUM_STRIDE.
24149         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
24150         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
24151         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
24152         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
24153         stride is constant and is below the minimum stride threshold.
24155 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
24157         PR c++/85659
24158         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
24159         the type is addressable.  Don't force op into register if it has
24160         BLKmode.
24162 2018-05-05  Roland McGrath  <mcgrathr@google.com>
24164         PR other/77609
24165         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
24166         any section for which we don't know a specific type it should have,
24167         regardless of name.  Previously this was done only for the exact
24168         names ".init_array", ".fini_array", and ".preinit_array".
24169         (default_elf_asm_named_section): Add comment about
24170         relationship with default_section_type_flags and SECTION_NOTYPE.
24171         (get_section): Don't consider it a type conflict if one side has
24172         SECTION_NOTYPE and the other doesn't, as long as neither has the
24173         SECTION_BSS et al used in the default_section_type_flags logic.
24175 2018-05-05  Tom de Vries  <tom@codesourcery.com>
24177         PR target/85653
24178         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
24179         (workaround_barsyncs): New function.
24180         (nvptx_reorg): Use workaround_barsyncs.
24181         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
24182         (define_expand "nvptx_membar_cta"): New define_expand.
24183         (define_insn "*nvptx_membar_cta"): New insn.
24185 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
24187         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
24188         To improve optimization opportunities.
24189         * builtin-types.def: The new needed builtin types for the above.
24191 2018-05-04  Richard Biener  <rguenther@suse.de>
24193         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
24194         * gimple-ssa-store-merging.c
24195         (imm_store_chain_info::output_merged_store): Remove redundant create,
24196         release split_store vector contents on failure.
24197         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
24198         scalar stmt vector on cache hit.
24200 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
24202         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
24203         Xilinx FP support.
24204         * config.gcc (powerpc-xilinx-eabi*): Remove.
24205         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
24206         support.
24207         (fusion_addis_mem_combo_load): Ditto.
24208         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
24209         FP support.
24210         (rs6000_cpu_cpp_builtins): Ditto.
24211         * config/rs6000/rs6000-linux.c
24212         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
24213         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
24214         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
24215         support.
24216         (rs6000_setup_reg_addr_masks): Ditto.
24217         (rs6000_init_hard_regno_mode_ok): Ditto.
24218         (rs6000_option_override_internal): Ditto.
24219         (legitimate_lo_sum_address_p): Ditto.
24220         (rs6000_legitimize_address): Ditto.
24221         (rs6000_legitimize_reload_address): Ditto.
24222         (rs6000_legitimate_address_p): Ditto.
24223         (abi_v4_pass_in_fpr): Ditto.
24224         (setup_incoming_varargs): Ditto.
24225         (rs6000_gimplify_va_arg): Ditto.
24226         (rs6000_split_multireg_move): Ditto.
24227         (rs6000_savres_strategy): Ditto.
24228         (rs6000_emit_prologue_components): Ditto.
24229         (rs6000_emit_epilogue_components): Ditto.
24230         (rs6000_emit_prologue): Ditto.
24231         (rs6000_emit_epilogue): Ditto.
24232         (rs6000_elf_file_end): Ditto.
24233         (rs6000_function_value): Ditto.
24234         (rs6000_libcall_value): Ditto.
24235         * config/rs6000/rs6000.h: Ditto.
24236         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
24237         (TARGET_MINMAX): ... this.  New.
24238         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
24239         * config/rs6000/rs6000.md: Remove Xilinx FP support.
24240         (*movsi_internal1_single): Delete.
24241         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
24242         mfpu=, mxilinx-fpu): Delete.
24243         * config/rs6000/singlefp.h: Delete.
24244         * config/rs6000/sysv4.h: Remove Xilinx FP support.
24245         * config/rs6000/t-rs6000: Ditto.
24246         * config/rs6000/t-xilinx: Delete.
24247         * config/rs6000/titan.md: Adjust for fp_type removal.
24248         * config/rs6000/vsx.md: Remove Xilinx FP support.
24249         (VStype_simple): Delete.
24250         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
24251         * config/rs6000/xfpu.h: Delete.
24252         * config/rs6000/xfpu.md: Delete.
24253         * config/rs6000/xilinx.h: Delete.
24254         * config/rs6000/xilinx.opt: Delete.
24255         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
24256         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
24258 2018-05-04  Tom de Vries  <tom@codesourcery.com>
24260         PR libgomp/85639
24261         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
24262         if ignore == 0.
24264 2018-05-04  Richard Biener  <rguenther@suse.de>
24266         PR middle-end/85627
24267         * tree-complex.c (update_complex_assignment): We are always in SSA form.
24268         (expand_complex_div_wide): Likewise.
24269         (expand_complex_operations_1): Likewise.
24270         (expand_complex_libcall): Preserve EH info of the original stmt.
24271         (tree_lower_complex): Handle removed blocks.
24272         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
24273         on complex multiplication and division libcall builtins.
24275 2018-05-04  Richard Biener  <rguenther@suse.de>
24277         PR middle-end/85574
24278         * fold-const.c (negate_expr_p): Restrict negation of operand
24279         zero of a division to when we know that can happen without
24280         overflow.
24281         (fold_negate_expr_1): Likewise.
24283 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
24285         PR libstdc++/85466
24286         * real.h (real_nextafter): Declare.
24287         * real.c (real_nextafter): New function.
24288         * fold-const-call.c (fold_const_nextafter): New function.
24289         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
24290         CASE_CFN_NEXTTOWARD.
24291         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
24292         even when arg1_mode is different from arg0_mode.
24294 2018-05-03  Nathan Sidwell  <nathan@acm.org>
24296         * doc/extend.texi (Deprecated Features): Remove
24297         -ffriend-injection.
24298         (Backwards Compatibility): Likewise.
24299         * doc/invoke.texi (C++ Language Options): Likewise.
24300         (C++ Dialect Options): Likewise.
24302 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
24304         PR target/85530
24305         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
24306         _mm512_mask_mullox_epi64): New intrinsics.
24308 2018-05-03  Tom de Vries  <tom@codesourcery.com>
24310         PR testsuite/85106
24311         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24312         dump files): Add offload-tree.
24314 2018-05-03  Richard Biener  <rguenther@suse.de>
24316         PR tree-optimization/85615
24317         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
24318         to loops not nested in BBs loop father to avoid creating multi-entry
24319         loops.
24321 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24323         PR tree-optimization/70291
24324         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
24325         arguments.  Change return type to tree.  Emit libcall as a new
24326         statement rather than replacing existing one when inplace_p is true.
24327         (expand_complex_multiplication_components): New function.
24328         (expand_complex_multiplication): Expand floating-point complex
24329         multiplication using the above.
24330         (expand_complex_division): Rename inner_type parameter to type.
24331         Update expand_complex_libcall call-site.
24332         (expand_complex_operations_1): Update expand_complex_multiplication
24333         and expand_complex_division call-sites.
24335 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
24337         PR target/85582
24338         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24339         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24340         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
24341         the highest significant bit of the shift count mask is clear.  In
24342         check whether and[sq]i3 is needed verify that all significant bits
24343         of the shift count other than the highest are set.
24345 2018-05-02  Tom de Vries  <tom@codesourcery.com>
24347         PR libgomp/82428
24348         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
24349         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
24350         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
24351         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
24352         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
24353         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
24354         __builtin_goacc_parlevel_size.
24356 2018-05-02  Richard Biener  <rguenther@suse.de>
24358         PR tree-optimization/85597
24359         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
24360         do not use split vect_get_vec_defs call but call vect_get_slp_defs
24361         directly.
24363 2018-05-02  Tom de Vries  <tom@codesourcery.com>
24365         PR testsuite/85106
24366         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24367         dump files): Add ltrans-tree.
24369 2018-05-02  Tom de Vries  <tom@codesourcery.com>
24371         PR testsuite/85106
24372         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24373         dump files): Add wpa-ipa.
24375 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
24377         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
24378         powerpc*-*-linux*paired* target.
24379         * config/rs6000/750cl.h: Delete.
24380         * config/rs6000/paired.h: Delete.
24381         * config/rs6000/paired.md: Delete.
24382         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
24383         float support.
24384         * config/rs6000/rs6000-builtin.def: Remove paired float support.
24385         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
24386         comment.  Remove paired float support.
24387         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
24388         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
24389         VECTOR_PAIRED.
24390         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
24391         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
24392         declarations.
24393         * config/rs6000/rs6000.c: Remove paired float support.
24394         (paired_expand_vector_init, paired_expand_vector_move,
24395         paired_emit_vector_compare, paired_emit_vector_cond_expr,
24396         (paired_expand_lv_builtin, paired_expand_stv_builtin,
24397         paired_expand_builtin, paired_expand_predicate_builtin,
24398         paired_init_builtins): Delete.
24399         * config/rs6000/rs6000.h: Remove paired float support.
24400         * config/rs6000/rs6000.md: Remove paired float support.
24401         (move_from_CR_ov_bit): Delete.
24402         * config/rs6000/rs6000.opt (mpaired): Delete.
24403         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
24404         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
24406 2018-05-02  Richard Biener  <rguenther@suse.de>
24408         PR middle-end/85567
24409         * gimplify.c (gimplify_save_expr): When in SSA form allow
24410         SAVE_EXPRs to compute to SSA vars.
24412 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
24414         PR target/85582
24415         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24416         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24417         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
24418         clobber operands[2], instead use a new pseudo.  Formatting fixes.
24420 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
24422         PR tree-optimization/85586
24423         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
24424         exit early for statements in the same group if the accesses are
24425         not strided.
24427 2018-05-02  Tom de Vries  <tom@codesourcery.com>
24429         PR lto/85451
24430         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
24431         error message.
24433 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
24435         PR tree-optimization/85143
24436         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
24438 2018-05-01  Tom de Vries  <tom@codesourcery.com>
24440         PR lto/85451
24441         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
24442         not found" error message.
24444 2018-05-01  Tom de Vries  <tom@codesourcery.com>
24446         PR other/83786
24447         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
24448         * vec.c (test_ordered_remove_if): New function.
24449         (vec_c_tests): Call test_ordered_remove_if.
24450         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
24451         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
24452         * tree-vect-patterns.c (vect_pattern_recog_1): Use
24453         VEC_ORDERED_REMOVE_IF.
24455 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
24457         PR tree-optimization/82665
24458         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
24459         pointer subtraction where arguments come from a memchr call.
24461 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
24463         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
24464         --push-state --as-needed and --pop-state instead of --as-needed and
24465         --no-as-needed if ld supports it.
24466         * configure: Regenerated.
24468         PR web/85578
24469         * doc/install.texi2html: Replace _002d with - and _002a with * in
24470         generated html files using sed.
24472 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
24474         PR c++/85523
24475         * gcc-rich-location.c (blank_line_before_p): New function.
24476         (use_new_line): New function.
24477         (gcc_rich_location::add_fixit_insert_formatted): New function.
24478         * gcc-rich-location.h
24479         (gcc_rich_location::add_fixit_insert_formatted): New function.
24481 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
24483         * selftest.c (assert_streq): Rename "expected" and "actual" to
24484         "val1" and "val2".  Extend NULL-handling to cover both inputs
24485         symmetrically, while still requiring both to be non-NULL for a pass.
24486         * selftest.h (assert_streq): Rename "expected" and "actual" to
24487         "val1" and "val2".
24488         (ASSERT_EQ): Likewise.
24489         (ASSERT_EQ_AT): Likewise.
24490         (ASSERT_KNOWN_EQ): Likewise.
24491         (ASSERT_KNOWN_EQ_AT): Likewise.
24492         (ASSERT_NE): Likewise.
24493         (ASSERT_MAYBE_NE): Likewise.
24494         (ASSERT_MAYBE_NE_AT): Likewise.
24495         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
24496         the assertion to pass.
24497         (ASSERT_STREQ_AT): Likewise.
24499 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
24501         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
24502         interaction with -pie.
24504 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
24506         * selftest.h: Fix alphabetization of per-source-file selftest
24507         declarations.
24509 2018-04-30  Jason Merrill  <jason@redhat.com>
24511         PR c++/61982 - dead stores to destroyed objects.
24512         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
24513         of clobber.
24515 2018-04-30  Jason Merrill  <jason@redhat.com>
24517         * tree.c (build_clobber): New.
24518         * tree.h: Declare it.
24519         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
24521 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
24523         * diagnostic-show-locus.c (layout::layout): Update for
24524         location_get_source_line returning a char_span.
24525         (struct char_span): Move to input.h.
24526         (struct correction): Update for fields in char_span becoming
24527         private.
24528         (struct source_line): Update for location_get_source_line
24529         returning a char_span.
24530         (layout::print_line): Likewise.
24531         * edit-context.c (edited_file::print_content): Likewise.
24532         (edited_file::print_diff_hunk): Likewise.
24533         (edited_file::print_run_of_changed_lines): Likewise.
24534         (edited_file::get_num_lines): Likewise.
24535         (edited_line::edited_line): Likewise.
24536         * final.c (asm_show_source): Likewise.
24537         * input.c (location_get_source_line): Convert return type
24538         from const char * to char_span, losing the final "line_len"
24539         param.
24540         (dump_location_info): Update for the above.
24541         (get_substring_ranges_for_loc): Likewise.  Use a char_span
24542         when handling the literal within the line.
24543         (test_reading_source_line): Update for location_get_source_line
24544         returning a char_span.
24545         * input.h (class char_span): Move here from
24546         diagnostic-show-locus.c, converting from a struct to a class.
24547         Make data members private.
24548         (char_span::operator bool): New.
24549         (char_span::length): New.
24550         (char_span::get_buffer): New.
24551         (char_span::operator[]): New.
24552         (char_span::subspan): Make const.
24553         (char_span::xstrdup): New.
24554         (location_get_source_line): Convert return type from const char *
24555         to char_span, losing the final "line_size" param.
24557 2018-04-30  Jan Hubicka  <jh@suse.cz>
24559         * lto-wrapper.c (ltrans_priorities): New static var.
24560         (cmp_priority): New.
24561         (run_gcc): Read priorities and if doing parallel build order
24562         the Makefile by them.
24564 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
24566         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
24568 2018-04-30  Richard Biener  <rguenther@suse.de>
24570         * tree-cfg.c (verify_address): Remove base argument, add
24571         flag whether to check TREE_ADDRESSABLE and do that.
24572         (verify_expr): Remove.
24573         (verify_types_in_gimple_reference): Add pieces from verify_expr.
24574         (verify_gimple_assign_single): Likewise.
24575         (verify_gimple_switch): Likewise.
24576         (verify_expr_location_1): Dereference tp once.  Add (disabled)
24577         piece from verify_expr.
24578         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
24580 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
24582         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
24584 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
24586         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
24587         (small_data_pattern): Likewise.
24588         (arc_rewrite_small_data): Likewise.
24589         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
24590         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
24591         (get_symbol_alignment): New function.
24592         (legitimate_small_data_address_p): Likewise.
24593         (legitimate_scaled_address): Update, call
24594         legitimate_small_data_address_p.
24595         (output_sdata): New static variable.
24596         (arc_print_operand): Update how we handle small data operands.
24597         (arc_print_operand_address): Likewise.
24598         (arc_legitimate_address_p): Update, use
24599         legitimate_small_data_address_p.
24600         (arc_rewrite_small_data_p): Remove.
24601         (arc_rewrite_small_data_1): Likewise.
24602         (arc_rewrite_small_data): Likewise.
24603         (small_data_pattern): Likewise.
24604         (compact_sda_memory_operand): Update to use
24605         legitimate_small_data_address_p and get_symbol_alignment.
24606         (prepare_move_operands): Don't rewite sdata pattern.
24607         (prepare_extend_operands): Remove.
24608         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
24609         pattern.
24610         (zero_extendqisi2): Likewise.
24611         (zero_extendhisi2): Likewise.
24612         (extendqihi2): Likewise.
24613         (extendqisi2): Likewise.
24614         (extendhisi2): Likewise.
24615         (addsi3): Likewise.
24616         (subsi3): Likewise.
24617         (andsi3): Likewise.
24618         * config/arc/constraints.md (Usd): Change it to memory constraint.
24620 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
24622         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
24623         as source of std instructions.
24624         * config/arc/arc.md (movsi_insn): Update pattern predicate to
24625         allow 6-bit constants as source for store instructions.
24626         (movdi_insn): Update instruction pattern to allow 6-bit constants
24627         as source for store instructions.
24629 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
24631         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
24633 2018-04-30  Nathan Sidwell  <nathan@acm.org>
24634             Sandra Loosemore  <sandra@codesourcery.com>
24636         * dumpfile.c (dump_open): Allow '-' for stdout.
24637         * doc/invoke.texi (Developer Options): Document dump filename
24638         determination early.  Document stdin/stdout selection.
24640 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
24642         Microblaze Target: PIC data text relative
24644         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
24645         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
24646         Add declaration.
24647         * config/microblaze/microblaze.h (microblaze_constant_address_p):
24648         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
24649         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
24650         New addressing mode for data-text relative position indepenedent code.
24651         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
24652         'ADDRESS_SYMBOLIC_TXT_REL'.
24653         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
24654         (microblaze_legitimate_pic_operand): Exclude function calls from
24655         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
24656         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
24657         addresses cases.
24658         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24659         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24660         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
24661         for 'address + offset'.
24662         (microblaze_expand_prologue): Add new function prologue call for
24663         'r20' assignation.
24664         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
24665         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
24666         table in case of TARGET_PIC_DATA_TEXT_REL.
24667         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
24668         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
24669         Add new macros 'UNSPEC_TEXT',
24670         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
24671         + exclude function calls from 'UNSPEC_PLT' in case of data text
24672         relative mode.
24673         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24674         new target hook for generating address diff vector tables in case of
24675         flag_pic.
24676         * doc/tm.texi : Regenerate.
24677         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
24678         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
24679         of addr diff vector generation.
24680         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24681         target hook definition.
24682         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
24683         Add default function for generate_pic_addr_diff_vec -> flag_pic.
24684         * doc/invoke.texi (Add new pic option): Add new microblaze pic
24685         option for data text relative.
24687 2018-04-30  Richard Biener  <rguenther@suse.de>
24689         * tree-chrec.h (evolution_function_is_constant_p): Remove
24690         redundant check.
24691         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
24693 2018-04-30  Richard Biener  <rguenther@suse.de>
24695         PR bootstrap/85571
24696         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
24698 2018-04-30  Richard Biener  <rguenther@suse.de>
24700         PR tree-optimization/28364
24701         PR tree-optimization/85275
24702         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
24703         copying first exit test.
24705 2018-04-28  Mark Wielaard  <mark@klomp.org>
24707         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
24708         dwarf_version >= 5.
24709         (dwarf_AT): Handle DW_AT_addr_base.
24710         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
24712 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
24714         PR target/84431
24715         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
24716         (*ashl<dwi>3_doubleword_mask_1): Ditto.
24717         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
24718         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
24720 2018-04-28  Richard Biener  <rguenther@suse.de>
24722         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
24723         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
24724         to reflect use.  Only add interesting stmts.
24726 2018-04-27  Martin Jambor  <mjambor@suse.cz>
24728         PR ipa/85549
24729         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
24730         the jump function allows for passing through aggregate values.
24732 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
24734         * input.h (in_system_header_at): Convert from macro to inline
24735         function.
24736         (from_macro_expansion_at): Likewise.
24737         (from_macro_definition_at): Likewise.
24739 2018-04-27  Jeff Law  <law@redhat.com>
24741         * config.gcc: Mark tile* targets as deprecated/obsolete.
24743 2018-04-27  Richard Biener  <rguenther@suse.de>
24745         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
24746         fix for ILP32.
24748 2018-04-27  Richard Biener  <rguenther@suse.de>
24750         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
24752 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
24754         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
24755         with Yd constraint. Set "preferred_for_speed" attribute from
24756         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
24757         with Yd constraint.
24758         (*movdi_internal): Ditto.
24759         (movti_interunit splitters): Remove
24760         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
24761         (movdi_interunit splitters): Ditto.
24762         * config/i386/constraints.md (Ye): Remove.
24763         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
24765 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24767         PR target/85512
24768         * config/aarch64/constraints.md (Usg): Limit to 31.
24769         (Usj): Limit to 63.
24771 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
24773         PR tree-optimization/85529
24774         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
24775         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
24776         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
24777         zero extension or masking of the MSB bit.
24778         (optimize_range_tests): Add FIRST_BB argument, pass it through
24779         to optimize_range_tests_var_bound.
24780         (maybe_optimize_range_tests, reassociate_bb): Adjust
24781         optimize_range_tests callers.
24783 2018-04-26  Richard Biener  <rguenther@suse.de>
24784             Jakub Jelinek  <jakub@redhat.com>
24786         * cgraph.h (symbol_table): Just declare debug method here.
24787         * symtab.c (symbol_table::debug): Define.
24789 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
24791         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
24793 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
24795         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
24796         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
24797         (*movdi_internal): Substitute Yi and Yj constraint with x
24798         and Ym and Yn constraint with y constraint.  Update "isa"
24799         attribute and set "preferred_for_speed" attribute from
24800         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
24801         (*movsi_internal): Ditto.
24802         (*movdf_internal): Ditto.
24803         (*movsf_internal): Ditto.
24804         (*zero_extendsidi2): Ditto.
24805         * config/i386/sse.md (vec_set<mode>_0): Ditto.
24806         (sse2_loadld): Ditto.
24807         (*vec_extract<ssevecmodelower>_0): Ditto.
24808         (*vec_extractv4si_0_zext_sse4): Ditto.
24809         (vec_concatv2di): Ditto.
24810         (*vec_dup<mode>): Ditto.
24811         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
24812         * config/i386/constraints.md (Yi): Remove.
24813         (Yj): Remove.
24814         (Ym): Remove.
24815         (Yn): Remove.
24817 2018-04-26  Nathan Sidwell  <nathan@acm.org>
24819         * dumpfile.c (dump_open): New.
24820         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
24821         (dump_finish): Detect stdio/stderr by value not name.
24823 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
24825         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
24827 2018-04-26  Tom de Vries  <tom@codesourcery.com>
24829         PR target/84952
24830         * config/nvptx/nvptx.c (verify_neutering_jumps)
24831         (verify_neutering_labels): New function
24832         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
24834 2018-04-26  Tom de Vries  <tom@codesourcery.com>
24836         PR target/84025
24837         * config/nvptx/nvptx.c (needs_neutering_p): New function.
24838         (nvptx_single): Use needs_neutering_p to skip over insns that do not
24839         need neutering.
24841 2018-04-26  Richard Biener  <rguenther@suse.de>
24842             Tom de Vries  <tom@codesourcery.com>
24844         PR lto/85422
24845         * lto-streamer-out.c (output_function): Fixup loops if required to match
24846         discovery done in the reader.
24848 2018-04-26  Richard Biener  <rguenther@suse.de>
24850         PR tree-optimization/85116
24851         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
24852         have a loop exit from the single latch predecessor.  Remove
24853         case of header with just condition.
24854         (ch_base::copy_headers): Exclude infinite loops from any
24855         processing.
24856         (pass_ch::execute): Record exits.
24858 2018-04-26  Richard Biener  <rguenther@suse.de>
24860         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
24861         prologue cost vector and pass it to vect_get_load_cost.
24862         (vect_get_peeling_costs_all_drs): Likewise.
24863         (vect_peeling_hash_get_lowest_cost): Likewise.
24864         (vect_enhance_data_refs_alignment): Likewise.
24866 2018-04-26  Richard Biener  <rguenther@suse.de>
24868         PR middle-end/85450
24869         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
24870         checking of integer<->pointer conversions.
24871         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
24872         sign-/zero-extending pointer types.
24873         (expand_omp_for_static_chunk): Likewise.
24875 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
24876             Jean Lee  <xiaoyur347@gmail.com>
24878         * config/mips/mips.c (mips_asan_shadow_offset): New function.
24879         (TARGET_ASAN_SHADOW_OFFSET): Define.
24880         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
24881         true for -fsanitize=address.
24883 2018-04-25  Mark Wielaard  <mark@klomp.org>
24885         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
24886         shorter ones.
24888 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
24890         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
24891         than "alu", remove explicit "memory" and "imm_disp" attributes.
24892         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
24894         PR middle-end/85414
24895         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
24896         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
24897         gen_lowpart_no_emit.
24899 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
24901         PR target/85473
24902         * config/i386/i386.c (ix86_expand_builtin): Change memory
24903         operand to XI, extend p0 to Pmode.
24904         * config/i386/i386.md: Change unspec volatile and operand
24905         1 mode to XI, change operand 0 mode to P.
24907 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24909         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
24910         GET_MODE_MASK before any checking.
24911         (nds32_can_use_bset_p): Likewise.
24912         (nds32_can_use_btgl_p): Likewise.
24914 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24916         * config/nds32/nds32-doubleword.md: New define_split pattern for
24917         illegal register number.
24919 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24921         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
24923 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24925         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
24927 2018-04-25  Richard Biener  <rguenther@suse.de>
24929         * lto-streamer.h (LTO_major_version): Bump to 8.
24931 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
24933         * BASE-VER: Set to 9.0.0.
24935 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
24937         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
24938         in __abskf2 and __powikf2.
24940 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24942         PR target/85512
24943         * config/aarch64/constraints.md (Usg, Usj): New constraints.
24944         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
24945         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
24946         Use the above on operand 2.  Reindent.
24947         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
24949 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
24951         PR target/85485
24952         * common/config/i386/i386-common.c (ix86_handle_option): Don't
24953         handle OPT_mcet.
24954         * config/i386/i386.opt (mcet): Removed.
24955         * doc/install.texi: Remove -mcet documentation.
24956         * doc/invoke.texi: Likewise.
24958 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
24960         PR target/85485
24961         * doc/install.texi: Remove -mcet from bootstrap-cet.
24963 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24965         PR target/85511
24966         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
24967         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
24968         if TARGET_64BIT.
24970         PR target/85503
24971         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
24972         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
24973         containing a CONST_VECTOR.
24975 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
24977         * doc/install.texi: Update newlib dependency for nvptx.
24979 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24981         PR target/85508
24982         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
24983         instead of INTVAL when shifting x left.
24985 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
24987         PR tree-optimization/85478
24988         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
24989         vect_grouped_store_supported for single element vectors.
24991 2018-04-24  Richard Biener  <rguenther@suse.de>
24993         PR target/85491
24994         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
24995         load cost increase to the case of non-constant step.
24997 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24999         PR target/84828
25000         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
25001         destination if any_malformed_asm.
25003 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
25005         PR middle-end/85496
25006         * expr.c (store_field): In the bitfield case, if the value comes from
25007         a function call and is returned in registers by means of a PARALLEL,
25008         do not change the mode of the temporary unless BLKmode and VOIDmode.
25010 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
25012         PR rtl-optimization/85423
25013         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
25014         dependencies to debug insns when the previous insn is non-debug.
25016 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
25018         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
25019         enums into a single definition.
25020         (fls): Fix predicates and printing.
25021         (seti): Likewise.
25023 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
25025         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
25026         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
25027         and short u6 immediate.
25028         (check_if_valid_sleep_operand): Remove.
25029         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
25031 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
25033         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
25034         flag_always_save_lp condition.
25035         * config/nds32/nds32.opt (malways-save-lp): New option.
25037 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
25039         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
25040         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
25041         * config/nds32/nds32.h
25042         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
25043         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
25045 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
25047         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
25048         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
25050 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
25051             Chung-Ju Wu  <jasonwucj@gmail.com>
25053         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
25054         Declare.
25055         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
25056         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
25058 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
25060         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
25062 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
25064         * config/nds32/nds32-protos.h (nds32_data_alignment,
25065         nds32_local_alignment): Declare.
25066         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
25067         nds32_local_alignment): New functions.
25068         (TARGET_CONSTANT_ALIGNMENT): Define.
25069         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
25071 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
25073         * config/nds32/nds32.c
25074         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
25075         (TARGET_MODES_TIEABLE_P): Likewise.
25077 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
25079         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
25080         level Ofast and Og.
25082 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
25083             Chung-Ju Wu  <jasonwucj@gmail.com>
25085         * config/nds32/constants.md (unspec_volatile_element): Add enum values
25086         for unaligned access.
25087         * config/nds32/nds32-intrinsic.c: Implementation of expanding
25088         unaligned access.
25089         * config/nds32/nds32-intrinsic.md: Likewise.
25090         * config/nds32/nds32_intrinsic.h: Likewise.
25091         * config/nds32/nds32.h (nds32_builtins): Likewise.
25092         * config/nds32/nds32.opt (munaligned-access): New option.
25093         * config/nds32/nds32.c (nds32_asm_file_start): Display
25094         flag_unaligned_access status.
25096 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
25098         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
25099         -mno-relax is present.
25100         * config/riscv/linux.h (LINK_SPEC): Ditto.
25102 2018-04-20  Martin Sebor  <msebor@redhat.com>
25104         PR c/85365
25105         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
25106         for null pointers.
25107         (gimple_fold_builtin_stxcpy_chk): Same.
25108         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
25110 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
25112         PR target/85456
25113         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
25114         __powikf2 when long double is IEEE 128-bit.
25116 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
25118         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
25119         step to make sure stack always aligned.
25121 2018-04-20  Carl Love  <cel@us.ibm.com>
25123         PR target/83402
25124         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
25125         size check for arg0.
25127 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
25128             Tom de Vries  <tom@codesourcery.com>
25130         PR target/85445
25131         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
25132         Emit insns for calls too.
25133         (nvptx_find_par): Always look for worker-level predecessor insn.
25134         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
25135         calls.
25136         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
25137         (nvptx_process_pars): Propagate frames for calls.
25139 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
25141         PR target/85469
25142         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
25143         Removed.
25144         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
25145         (ix86_handle_option): Don't handle OPT_mibt.
25146         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
25147         __SHSTK__.
25148         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
25149         has_ibt and ibt.
25150         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
25151         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
25152         (ix86_target_macros): Define __CET__ with flag_cf_protection
25153         for -fcf-protection.
25154         * config/i386/i386.c (isa2_opts): Remove -mibt.
25155         * config/i386/i386.h (TARGET_IBT): Removed.
25156         (TARGET_IBT_P): Likewise.
25157         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
25158         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
25159         * config/i386/i386.opt (mcet): Update help message.
25160         (mshstk): Likewise.
25161         (mibt): Removed.
25162         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
25163         -mcet as an alias for -mshstk.
25165 2018-04-20  Richard Biener  <rguenther@suse.de>
25167         PR middle-end/85475
25168         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
25169         complexity by forcing a single use of the multiply operand.
25171 2018-04-20  Martin Jambor  <mjambor@suse.cz>
25173         ipa/85449
25174         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
25175         recursion dependency to only apply to non-clones.
25177 2018-04-20  Martin Jambor  <mjambor@suse.cz>
25179         ipa/85447
25180         * ipa-cp.c (create_specialized_node): Check that clones of
25181         self-recursive edges exist during IPA-CP.
25183 2018-04-19  Toon Moene  <toon@moene.org>
25185         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
25186         by -O3.
25188 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
25190         PR tree-optimization/85467
25191         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
25192         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
25193         VECTOR_CST element to type.
25195 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
25197         PR target/85397
25198         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
25199         * config/i386/i386.md (builtin_setjmp_setup): Removed.
25200         (builtin_longjmp): Likewise.
25201         (save_stack_nonlocal): New pattern.
25202         (restore_stack_nonlocal): Likewise.
25204 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
25206         PR target/85404
25207         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
25208         Replace ASM_OUTPUT_LABEL with fprintf.
25210 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
25212         PR target/85417
25213         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
25214         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
25215         * config/i386/i386-c.c (ix86_target_macros_internal): Also
25216         define __IBT__ and __SHSTK__ for -fcf-protection.
25217         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
25218         TARGET_IBT.
25219         (ix86_trampoline_init): Likewise.
25220         (x86_output_mi_thunk): Likewise.
25221         (ix86_notrack_prefixed_insn_p): Likewise.
25222         (ix86_option_override_internal): Don't disallow -fcf-protection.
25223         * config/i386/i386.md (rdssp<mode>): Also enable for
25224         -fcf-protection.
25225         (incssp<mode>): Likewise.
25226         (nop_endbr): Likewise.
25227         * config/i386/i386.opt (mcet): Change help message to built-in
25228         functions only.
25229         (mibt): Likewise.
25230         (mshstk): Likewise.
25231         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
25232         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
25233         enable CET built-in functions.
25235 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
25237         * common/config/i386/i386-common.c
25238         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
25239         OPTION_MASK_ISA_MOVDIRI_UNSET,
25240         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
25241         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
25242         * config.gcc (movdirintrin.h): New header.
25243         * config/i386/cpuid.h (bit_MOVDIRI,
25244         bit_MOVDIR64B): New bits.
25245         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
25246         and -mmvodir64b.
25247         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
25248         (VOID, PVOID, PCVOID)): New function types.
25249         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
25250         __builtin_ia32_directstoreu_u64,
25251         __builtin_ia32_movdir64b): New builtins.
25252         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
25253         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
25254         and -mmovdiri.
25255         (ix86_valid_target_attribute_inner_p): Ditto.
25256         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
25257         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
25258         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
25259         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
25260         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
25261         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
25262         (movdiri<mode>, movdir64b_<mode>): New.
25263         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
25264         * config/i386/immintrin.h: Include movdirintrin.h.
25265         * config/i386/movdirintrin.h: New file.
25266         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
25268 2018-04-19  Richard Biener  <rguenther@suse.de>
25270         PR middle-end/85455
25271         * cfg.c (clear_bb_flags): When loop state says we have
25272         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
25274 2018-04-19  Richard Biener  <rguenther@suse.de>
25276         PR tree-optimization/84737
25277         * tree-vect-data-refs.c (vect_copy_ref_info): New function
25278         copying restrict info.
25279         (vect_setup_realignment): Use it.
25280         * tree-vectorizer.h (vect_copy_ref_info): Declare.
25281         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
25282         the first DR to all generated stores.
25283         (vectorizable_load): Likewise for loads.
25285 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
25287         PR tree-optimization/85446
25288         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
25289         the integral and pointer types to have the same precision.
25291         * doc/install.texi: Document --disable-cet being the default and
25292         --enable-cet=auto.
25294 2018-04-18  Martin Liska  <mliska@suse.cz>
25296         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
25297         style.
25299 2018-04-18  Martin Liska  <mliska@suse.cz>
25301         Revert
25302         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
25304         PR ipa/83983
25305         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
25306         arguments if they are comparable.
25308 2018-04-18  Martin Liska  <mliska@suse.cz>
25310         Revert
25311         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
25313         PR lto/84805
25314         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
25315         incomplete types.
25317 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
25319         PR target/85388
25320         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
25321         ENDBR after calling __morestack.
25323 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
25325         PR jit/85384
25326         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
25327         by using gcc_base_ver to generate a gcc_driver_version, and use
25328         it when generating GCC_DRIVER_NAME.
25329         * configure: Regenerate.
25331 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
25333         PR target/81084
25334         * config.gcc: Obsolete powerpc*-*-*spe*.
25336 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
25338         PR debug/84637
25339         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
25340         (stabstr_D): Change type of unum from unsigned int to
25341         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
25342         type.
25344 2018-04-17  Jim Wilson  <jimw@sifive.com>
25346         PR 84856
25347         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
25348         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
25349         Set arg_pointer_offset after using pretend_args_size.
25351 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
25353         PR rtl-optimization/85431
25354         * dse.c (record_store): Ignore zero width stores.
25356         PR sanitizer/85230
25357         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
25358         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
25359         __builtin_stack_restore rather than after it.
25360         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
25361         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
25362         argument instead of virtual_dynamic_stack_rtx.
25364 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25366         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
25367         New prototype.
25368         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25369         Add note to error message to explain internal mapping of overloaded
25370         built-in function name to non-overloaded built-in function name.
25371         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
25372         function.
25374 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25376         PR target/85424
25377         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
25378         where the inputs overlap with the output.
25380 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
25382         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
25383         (=v, v) alternative and explicit "memory" attribute.
25384         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
25385         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25386         attributes.
25387         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
25388         "sselog1" type instead of "sselog".
25389         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
25390         "sselog".  Remove explicit "memory" attribute.
25391         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
25392         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25393         attributes.
25394         (vec_extract_hi_v32hi): Merge all alternatives into one, use
25395         "sselog1" type instead of "sselog".  Remove explicit "memory"
25396         attribute.
25397         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
25398         use "sselog1" type instead of "sselog".  Remove explicit "memory"
25399         attribute.
25400         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
25401         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25402         attributes.
25403         (vec_extract_hi_v64qi): Merge all alternatives into one, use
25404         "sselog1" type instead of "sselog".  Remove explicit "memory"
25405         attribute.
25406         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
25407         use "sselog1" type instead of "sselog".  Remove explicit "memory"
25408         attribute.
25410         PR target/85430
25411         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
25413         PR middle-end/85414
25414         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
25415         on a SUBREG.
25417 2018-04-17  Martin Jambor  <mjambor@suse.cz>
25419         PR ipa/85421
25420         * ipa-cp.c (create_specialized_node): Call
25421         expand_all_artificial_thunks if necessary.
25423 2018-04-17  Martin Liska  <mliska@suse.cz>
25425         PR lto/85405
25426         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
25427         in message, remote space in between '_G' and '('.
25429 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
25431         PR target/85281
25432         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
25433         avx512f_vmcmp<mode>3<round_saeonly_name>,
25434         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
25435         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
25436         avx512f_rndscale<mode><round_saeonly_name>,
25437         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
25438         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
25439         Use %<iptr>2 instead of %2 for -masm=intel.
25440         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
25441         avx512f_vcvttss2usi<round_saeonly_name>,
25442         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
25443         -masm=intel.
25444         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
25445         avx512f_vcvttsd2usi<round_saeonly_name>,
25446         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
25447         Use %q1 instead of %1 for -masm=intel.
25448         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
25449         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
25450         of %3 for -masm=intel.
25451         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
25452         -masm=intel.
25453         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
25454         -masm=intel.
25455         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
25456         -masm=intel.
25457         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
25458         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
25459         %g1.
25460         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
25461         -masm=intel.
25462         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
25463         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
25464         %g1 and one with %0 and %1.
25465         (avx512er_vmrcp28<mode><round_saeonly_name>,
25466         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
25467         %1 for -masm=intel.
25468         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
25469         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
25470         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
25471         of %0 and %{%4%} for -masm=intel.
25472         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
25473         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
25474         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
25475         order of %0 and %{%5%}%{z%} for -masm=intel.
25477 2018-04-17  Jan Hubicka  <jh@suse.cz>
25479         PR lto/85405
25480         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
25482 2018-04-17  Martin Liska  <mliska@suse.cz>
25484         PR ipa/85329
25485         * multiple_target.c (create_dispatcher_calls): Set apostrophes
25486         for target_clone error message.  Make default implementation
25487         clone to be a local declaration.
25488         (separate_attrs): Add new argument and check for an empty
25489         string.
25490         (expand_target_clones): Handle it.
25491         (ipa_target_clone): Make redirection just for target_clones
25492         functions.
25494 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
25495             Tom de Vries  <tom@codesourcery.com>
25497         PR middle-end/84955
25498         * omp-expand.c (expand_oacc_for): Add dummy false branch for
25499         tiled basic blocks without omp continue statements.
25501 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
25503         PR target/83660
25504         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
25505         vec_extract expression as having side effects to make sure it gets
25506         a cleanup point.
25508 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
25510         PR target/85403
25511         * config/i386/i386.c (get_builtin_code_for_version): Check
25512         error_mark_node.
25514 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
25516         PR target/84331
25517         * config.gcc: Support "skylake".
25518         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
25519         PROCESSOR_SKYLAKE.
25520         * config/i386/i386.c (m_SKYLAKE): Define.
25521         (processor_target_table): Add "skylake".
25522         (ix86_option_override_internal): Add "skylake".
25523         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
25524         PROCESSOR_CANNONLAKE.
25525         (get_builtin_code_for_version): Fix priority for
25526         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
25527         PROCESSOR_SKYLAKE-AVX512.
25528         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
25529         (processor_type): Add PROCESSOR_SKYLAKE.
25531 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
25532             Jason Merrill  <jason@redhat.com>
25534         PR c++/85112
25535         * convert.c (convert_to_integer_1): Use direct recursion for
25536         enumeral types and types with a precision less than the number
25537         of bits in their mode.
25539 2018-04-16  Julia Koval  <julia.koval@intel.com>
25541         PR target/84413
25542         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
25543         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
25545 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
25547         PR target/85293
25548         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
25549         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
25550         and -mno-direct-move.
25552 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
25554         PR target/83402
25555         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
25556         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
25557         Ensure negative shifts result in {0}.
25559 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
25561         PR rtl-optimization/79916
25562         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
25563         regs (if any) to define how to gnerate SD moves when LRA is in
25564         progress.
25566 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
25568         PR rtl-optimization/85393
25569         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
25570         * except.c (expand_dw2_landing_pad_for_region): Make static.
25571         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
25572         a label and unconditional jump to old_bb, rather than
25573         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
25574         basic block.
25576         PR rtl-optimization/85376
25577         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
25578         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
25579         instead of a specific value.
25581 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
25582             Bin Cheng  <bin.cheng@arm.com>
25584         PR tree-optimization/82965
25585         PR tree-optimization/83991
25586         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
25587         by_profile_only parameter.
25588         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
25589         information if the loop was predicted to iterate too many times.
25590         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
25592 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
25594         PR lto/71991
25595         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
25596         always inline.
25598 2018-04-13  Martin Liska  <mliska@suse.cz>
25599             Jakub Jelinek  <jakub@redhat.com>
25601         PR middle-end/81657
25602         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
25603         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
25604         * builtins.c (expand_builtin_memory_copy_args): Use
25605         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
25606         handle dest_addr == pc_rtx.
25608 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
25610         PR target/85291
25611         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
25612         asked to not generate direct moves.
25613         (fix_trunc<mode>si2_stfiwx): Similar.
25614         (fix_trunc<mode>si2_internal): Similar.
25616 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
25618         PR debug/83157
25619         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
25620         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
25621         lookup if dest in some wider mode is known to be const0_rtx and
25622         if so, record permanent equivalence for it to be ZERO_EXTEND of
25623         the narrower mode destination.
25625 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
25627         * lto-streamer-out.c (output_function): Revert 259346.
25628         * omp-expand.c (expand_oacc_for): Likewise.
25630 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
25632         PR rtl-optimization/85354
25633         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
25634         * sel-sched.c (sel_global_init): ... here.
25636 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
25638         PR target/85238
25639         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
25640         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
25641         mode for PE-COFF targets.
25642         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
25643         (i386_pe_asm_lto_end): Likewise.
25644         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
25645         (TARGET_ASM_LTO_END): Likewise.
25646         * config/i386/winnt.c (saved_debug_info_level): New static variable.
25647         (i386_pe_asm_lto_start): New function.
25648         (i386_pe_asm_lto_end): Likewise.
25650 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
25651             Richard Biener  <rguenther@suse.de>
25653         PR middle-end/84955
25654         * lto-streamer-out.c (output_function): Fix CFG loop state before
25655         streaming out.
25656         * omp-expand.c (expand_oacc_for): Handle calls to internal
25657         functions like regular functions.
25659 2018-04-12  Richard Biener  <rguenther@suse.de>
25661         PR lto/85371
25662         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
25663         for the early LTO debug to properly generate references to it
25664         during DIE emission.  Do not re-use that for the skeleton for
25665         split-dwarf.
25666         (dwarf2out_early_finish): Likewise.
25668 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
25670         PR target/85328
25671         * config/i386/sse.md
25672         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
25673         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
25674         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
25675         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
25676         and output is a reg, avoid creating invalid lowpart subreg, but
25677         instead split into a 512-bit move.  Don't split if not AVX512VL,
25678         input is xmm16+ reg and output is a mem.
25679         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
25680         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
25681         xmm16+ reg and output is a mem.
25683 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25685         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
25686         also for flag_dwarf2_cfi_asm.
25688 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
25690         PR rtl-optimization/85342
25691         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
25692         a bool scalar var inside of the loop instead.  Don't try to update
25693         recog_data.operand after failed apply_change_group.
25695 2018-04-12  Tom de Vries  <tom@codesourcery.com>
25697         PR target/85296
25698         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
25699         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
25700         array with flexible array member as array without given dimension.
25701         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
25702         argument for undefined param to true.
25704 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
25706         PR target/85321
25707         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25708         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
25709         from PowerPC section.
25710         * config/rs6000/sysv4.opt (mcall-): Improve help text.
25711         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
25712         help text that is too long.
25713         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
25714         help text that is too long.
25715         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
25716         help text that is too long.
25718 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
25720         * config/alpha/alpha.md (stack_probe_internal): Rename
25721         from "probe_stack".  Update all callers.
25723 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
25725         PR rtl-optimization/84566
25726         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
25727         sched_macro_fuse_insns.
25729 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
25731         PR target/84301
25732         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
25733         (compute_block_dependences): ... from here.
25735 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
25737         PR tree-optimization/85331
25738         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
25739         from int to HOST_WIDE_INT.
25741 2018-04-11  Martin Jambor  <mjambor@suse.cz>
25743         PR ipa/84149
25744         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
25745         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
25746         not the same as the source val.
25747         (cgraph_edge_brings_value_p): New parameter.
25748         (gather_edges_for_value): Pass destination value to
25749         cgraph_edge_brings_value_p.
25750         (perhaps_add_new_callers): Likewise.
25751         (get_info_about_necessary_edges): Likewise and exclude values brought
25752         only by self-recursive edges.
25753         (create_specialized_node): Redirect only clones of self-calling edges.
25754         (+self_recursive_pass_through_p): New function.
25755         (find_more_scalar_values_for_callers_subset): Use it.
25756         (find_aggregate_values_for_callers_subset): Likewise.
25757         (known_aggs_to_agg_replacement_list): Removed.
25758         (decide_whether_version_node): Re-calculate known constants for all
25759         remaining context clones.
25761 2018-04-11  Richard Biener  <rguenther@suse.de>
25763         PR lto/85339
25764         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
25765         from early DWARF output.
25766         (dwarf2out_early_finish): Output line info unconditionally into
25767         early DWARF and add reference to it.
25769 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
25771         PR target/85281
25772         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
25773         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
25774         other than V2DFmode using iptr mode attribute.
25775         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
25777 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
25779         PR rtl-optimization/84659
25780         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
25782 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
25784         PR debug/85302
25785         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
25786         SIZEP is NULL.
25787         (output_loc_list): Pass address of a dummy size variable even in the
25788         locview handling loop.
25789         (index_location_lists): Add comment on why skip_loc_list_entry can't
25790         call size_of_locs.
25792 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25794         PR target/85261
25795         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
25796         into register.
25798 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
25800         PR target/85321
25801         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25802         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
25803         and -mstring-compare-inline-limit.
25805 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
25807         PR target/85287
25808         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
25809         for stack clash protection in a register whenever we need it to be in
25810         a register.
25812 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
25814         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
25815         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
25817 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
25819         PR target/85321
25820         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
25821         the help text.
25822         (mlong-double-): Ditto.
25823         * config/rs6000/sysv4.opt (msdata=): Ditto.
25824         (mtls-size=): Ditto.
25826 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25828         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
25829         erroneous entries for
25830         "vector int vec_ldl (int, long int *)", and
25831         "vector unsigned int vec_ldl (int, unsigned long int *)".
25832         Add comments and entries for
25833         "vector bool char vec_ldl (int, bool char *)",
25834         "vector bool short vec_ldl (int, bool short *)",
25835         "vector bool int vec_ldl (int, bool int *)",
25836         "vector bool long long vec_ldl (int, bool long long *)",
25837         "vector pixel vec_ldl (int, pixel *)",
25838         "vector long long vec_ldl (int, long long *)",
25839         "vector unsigned long long vec_ldl (int, unsigned long long *)".
25840         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
25841         type tree bool_long_long_type_node and correct definition of
25842         bool_V2DI_type_node to make reference to this new type tree.
25843         (rs6000_mangle_type): Replace erroneous reference to
25844         bool_long_type_node with bool_long_long_type_node.
25845         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
25846         comments to emphasize sign distinctions for char and int types and
25847         replace RS6000_BTI_bool_long constant with
25848         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
25849         use of RS6000_BTI_pixel.
25850         (bool_long_type_node): Remove this macro definition.
25851         (bool_long_long_type_node): New macro definition
25853 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
25855         PR rtl-optimization/85300
25856         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
25857         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
25858         simplify_unary_operation fails.
25860 2018-04-10  Martin Liska  <mliska@suse.cz>
25862         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
25863         cgraph_edge and ipa_ref.
25865 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
25867         PR target/85177
25868         PR target/85255
25869         * config/i386/sse.md
25870         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
25871         computation of the VEC_MERGE selector from mask.
25872         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
25873         Fix decoding of the VEC_MERGE selector into mask.
25875 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
25877         PR tree-optimization/85286
25878         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
25880 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
25882         * final.c (final_1): Set insn_last_address as well as
25883         insn_current_address.
25885 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25887         PR target/85173
25888         * explow.c (emit_stack_probe): Call validize_mem on memory location
25889         before passing it to gen_probe_stack.  Create address operand and
25890         legitimize it for the probe_stack_address case.
25892 2018-04-09  Jan Hubicka  <jh@suse.cz>
25894         PR lto/85078
25895         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
25896         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
25897         * tree.c (free_lang_data_in_type): Fix handling of binfos;
25898         walk basetypes.
25899         (free_lang_data): Rebuild type inheritance graph.
25901 2018-04-09  Martin Sebor  <msebor@redhat.com>
25903         * invoke.texi (-finline-small-functions): Mention other optimization
25904         options.
25905         (-findirect-inlining, -fpartial-inlining): Same.
25906         (-finline-functions-called-once): Same.
25907         (-freorder-blocks-and-partition): Same.
25909 2018-04-09  Jan Hubicka  <jh@suse.cz>
25911         PR rtl/84058
25912         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
25913         jumps; choose last target that matches the criteria (i.e.
25914         no partition changes for non-crossing jumps).
25915         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
25916         support for redirecting crossing jumps to non-crossing.
25918 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
25920         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
25921         also for naked functions.
25923 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
25925         * config/arc/arc.md (add_shift): New pattern.
25926         (add_shift2): Likewise.
25927         (sub_shift): Likewise.
25928         (sub_shift_cmp0_noout): Likewise.
25929         (compare_si_ashiftsi): Likewise.
25930         (xbfu_cmp0_noout): New combine pattern.
25931         (xbfu_cmp0"): Likewise.
25932         (movsi_set_cc_insn): Place the predicable variant first.
25933         (commutative_binary_cmp0_noout): Remove clobber.
25934         (commutative_binary_cmp0): New pattern.
25935         (noncommutative_binary_cmp0): Likewise.
25936         (noncommutative_binary_cmp0_noout): Likewise.
25937         (noncommutative_binary_comparison_result_used): Removed.
25938         (rsub_cmp0): New pattern.
25939         (rsub_cmp0_noout): Likewise.
25940         (extzvsi): Changed, keep only meaningful variants.
25941         (SQH, SEZ): New iterators.
25942         (SQH_postfix): New mode attribute.
25943         (SEZ_prefix): New code attribute.
25944         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
25945         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
25946         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
25947         of numerical value.
25948         (noncommutative_operator): Check the availability of barrel
25949         shifter option.
25951 2018-04-09  Richard Biener  <rguenther@suse.de>
25953         PR tree-optimization/85284
25954         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
25955         Only use the niter constraining form of simple_iv when the exit
25956         is always executed.
25958 2018-04-09  Tom de Vries  <tom@codesourcery.com>
25960         PR target/84041
25961         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
25962         (define_expand "*memory_barrier"): New define_expand.
25963         (define_insn "memory_barrier"): New insn.
25965 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25967         PR rtl-optimization/80463
25968         PR rtl-optimization/83972
25969         PR rtl-optimization/83480
25971         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
25972         correct producer for the insn.
25973         (tidy_control_flow): Fixup seqnos in case of debug insns.
25975 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25977         PR rtl-optimization/83913
25979         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
25980         different sched-times when merging exprs.
25982 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25984         PR rtl-optimization/83962
25986         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
25987         tidy_fallthru_edge and tidy_control_flow.
25989 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25991         PR rtl-optimization/83530
25993         * sel-sched.c (force_next_insn): New global variable.
25994         (remove_insn_for_debug): When force_next_insn is true, also leave only
25995         next insn in the ready list.
25996         (sel_sched_region): When the region wasn't scheduled, make another pass
25997         over it with force_next_insn set to 1.
25999 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
26001         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
26002         into tm_file.
26003         * config/nds32/constants.md (unspec_volatile_element): Add enum values
26004         for interrupt control.
26005         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
26006         functions for interrupt control.
26007         * config/nds32/nds32-intrinsic.md: Likewise.
26008         * config/nds32/nds32_intrinsic.h: Likewise.
26009         * config/nds32/nds32.h (nds32_builtins): Likewise.
26011 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
26013         * config/nds32/nds32.c (nds32_init_machine_status,
26014         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
26015         strict_aligned_p field.
26016         (nds32_expand_to_rtl_hook): New function.
26017         (TARGET_EXPAND_TO_RTL_HOOK): Define.
26018         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
26020 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
26021             Chung-Ju Wu  <jasonwucj@gmail.com>
26023         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
26024         * config/nds32/nds32-n7.md: New file.
26025         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
26026         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
26027         pipeline.
26028         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
26029         * config/nds32/nds32.md (pipeline_model): Add n7.
26030         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
26031         * config/nds32/pipelines.md: Include n7 settings.
26033 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
26034             Chung-Ju Wu  <jasonwucj@gmail.com>
26036         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
26037         * config/nds32/nds32-e8.md: New file.
26038         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
26039         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
26040         pipeline.
26041         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
26042         * config/nds32/nds32.md (pipeline_model): Add e8.
26043         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
26044         * config/nds32/pipelines.md: Include e8 settings.
26046 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
26047             Chung-Ju Wu  <jasonwucj@gmail.com>
26049         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
26050         * config/nds32/nds32-n8.md: New file.
26051         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
26052         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
26053         pipeline.
26054         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
26055         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
26056         * config/nds32/nds32.md (pipeline_model): Add n8.
26057         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
26058         * config/nds32/pipelines.md: Include n8 settings.
26060 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
26061             Chung-Ju Wu  <jasonwucj@gmail.com>
26063         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
26064         * config/nds32/nds32-n9-2r1w.md: New file.
26065         * config/nds32/nds32-n9-3r2w.md: New file.
26066         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
26067         nds32_register_ports): New or modify for cpu n9.
26068         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
26069         pipeline.
26070         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
26071         * config/nds32/nds32-utils.c: New file.
26072         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
26073         TARGET_MUL_SLOW): Define.
26074         * config/nds32/nds32.md (pipeline_model): New attribute.
26075         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
26076         New options that support cpu n9.
26077         * config/nds32/pipelines.md: Include n9 settings.
26078         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
26080 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
26082         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
26083         information if necessary.
26084         (output_cond_branch_compare_zero): Likewise.
26085         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
26086         (nds32_target_alignment): Refine for alignment.
26087         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
26088         (FUNCTION_BOUNDARY): Modify.
26089         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
26090         align case.
26091         * config/nds32/nds32.opt (malways-align, malign-functions): New.
26093 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
26095         * config/nds32/constants.md (unspec_volatile_element): Add values for
26096         TLB operation and data prefetch.
26097         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
26098         functions for TLB operation and data prefetch.
26099         * config/nds32/nds32-intrinsic.md: Likewise.
26100         * config/nds32/nds32_intrinsic.h: Likewise.
26101         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
26102         (nds32_print_operand): Likewise.
26103         * config/nds32/nds32.h (nds32_builtins): Likewise.
26105 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
26106         Andrew Pinski  <pinsika@gcc.gnu.org>
26108         PR middle-end/82976
26109         * match.pd: Use constant_boolean_node of correct type instead of
26110         boolean_true_node or boolean_false_node for simplifying
26111         pointer comparisons to zero.
26113 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
26115         PR tree-optimization/80021
26116         * tree.c (verify_type_variant): Make error call in verify_variant_match
26117         translatable and remove final full stop.
26119 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
26121         * config/nds32/constants.md (unspec_volatile_element): Add
26122         UNSPEC_VOLATILE_EH_RETURN.
26123         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
26124         nds32_output_stack_pop): Support dwarf exception handling process.
26125         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
26126         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
26127         exception handling process.
26128         (nds32_compute_stack_frame): Likewise.
26129         (nds32_return_addr_rtx): Likewise.
26130         (nds32_initial_elimination_offset): Likewise.
26131         (nds32_expand_prologue): Likewise.
26132         (nds32_expand_epilogue): Likewise.
26133         (nds32_dynamic_chain_address): New function.
26134         * config/nds32/nds32.h (machine_function): Add fields for dwarf
26135         exception handling.
26136         (DYNAMIC_CHAIN_ADDRESS): Define.
26137         (EH_RETURN_DATA_REGNO): Define.
26138         (EH_RETURN_STACKADJ_RTX): Define.
26139         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
26140         patterns for dwarf exception handling.
26142 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
26144         * config/nds32/nds32.h: Clean up obsolete macros.
26146 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
26148         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
26149         Add enum values for particular instructions.
26150         * config/nds32/nds32-intrinsic.c: Implementation of expanding
26151         particular intrinsic functions.
26152         * config/nds32/nds32-intrinsic.md: Likewise.
26153         * config/nds32/nds32_intrinsic.h: Likewise.
26154         * config/nds32/nds32.h (nds32_builtins): Likewise.
26155         * config/nds32/nds32.md (type): Add pbsad and pbsada.
26156         (btst, ave): New patterns for particular instructions.
26158 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
26160         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
26161         Add enum values for atomic load/store and memory sync.
26162         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
26163         and memory sync.
26164         * config/nds32/nds32-intrinsic.md: Likewise.
26165         * config/nds32/nds32_intrinsic.h: Likewise.
26166         * config/nds32/nds32.h (nds32_builtins): Likewise.
26168 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
26170         PR tree-optimization/85257
26171         * fold-const.c (native_encode_vector): If not all elts could fit
26172         and off is -1, return 0 rather than offset.
26173         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
26174         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
26175         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
26176         adjust buffer in native_interpret_expr call.
26178 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
26180         * config/nds32/constants.md (unspec_volatile_element): Add cache
26181         control enum values.
26182         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
26183         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
26184         * config/nds32/nds32.c (nds32_cctl_names): New.
26185         (nds32_print_operand): Handle cache control register names.
26186         * config/nds32/nds32.h (nds32_builtins): New enum values.
26187         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
26188         macros.
26189         * config/nds32/nds32.md (type): Add mmu.
26190         * config/nds32/pipelines.md (simple_insn): Add mmu.
26192 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
26194         * config/nds32/nds32.md (type): Remove call.
26195         * config/nds32/pipelines.md (simple_insn): Likewise.
26197 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
26199         * config/nds32/constants.md (unspec_volatile_element): Add
26200         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
26201         UNSPEC_VOLATILE_FMFCFG.
26202         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
26203         description for fmfcfg and fmfcsr.
26204         (bdesc_1arg): Add fmtcsr.
26205         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
26206         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
26207         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
26208         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
26209         unspec_fmfcfg): New patterns.
26210         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
26211         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
26212         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
26213         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
26214         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
26215         __nds32__fmfcfg): Define.
26217 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
26219         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
26220         intrinsic register names.
26221         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
26222         intrinsic register enum values and macros.
26224 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
26226         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
26227         for load/store addressing form.
26228         (nds32_print_operand_address): Likewise.
26230 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
26232         PR target/85196
26233         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
26234         based on LABEL_REF.  Remove useless assertion.
26235         (pic_address_needs_scratch): Fix formatting.
26236         (sparc_legitimize_pic_address): Minor tweaks.
26237         (sparc_delegitimize_address): Adjust assertion accordingly.
26238         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
26239         into symbolic_operand.
26240         (movsi_high_pic_label_ref): Likewise.
26241         (movsi_lo_sum_pic_label_ref): Likewise.
26242         (movdi_pic_label_ref): Likewise.
26243         (movdi_high_pic_label_ref): Likewise.
26244         (movdi_lo_sum_pic_label_ref): Likewise.
26246 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
26248         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
26249         custom LIB_SPEC setup.
26251 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
26252             Kito Cheng  <kito.cheng@gmail.com>
26254         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
26255         * config/riscv/freebsd.h: New.
26257 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
26259         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
26260         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
26261         file.
26263 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
26264             Kito Cheng  <kito.cheng@gmail.com>
26266         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
26267         nds32_output_call, nds32_symbol_binds_local_p): New functions.
26268         * config/nds32/nds32-protos.h (nds32_output_call,
26269         nds32_output_return): Declare.
26270         * config/nds32/nds32.md: Refine all the call and return patterns.
26272 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
26274         PR debug/85252
26275         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
26276         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
26278         PR rtl-optimization/84872
26279         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
26280         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
26281         EDGE_CROSSING edge.
26283 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
26285         * expr.c (copy_blkmode_to_reg): Revert 254862.
26286         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
26288 2018-04-06  Richard Biener  <rguenther@suse.de>
26290         PR middle-end/85244
26291         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
26292         after seeing a component reference with an adjacent field.  Treat
26293         refs to arrays at struct end of external decls similar to
26294         refs to unconstrained commons.
26296 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
26298         PR sanitizer/85213
26299         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
26300         look through SAVE_EXPRs with non-side-effects argument.  Adjust
26301         recursive calls.
26302         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
26303         save_p here.
26305 2018-04-06  Richard Biener  <rguenther@suse.de>
26307         PR middle-end/85180
26308         * alias.c (find_base_term): New wrapper around find_base_term
26309         unwinding CSELIB_VAL_PTR changes.
26310         (find_base_term): Do not restore CSELIB_VAL_PTR during the
26311         recursion.
26313 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26315         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
26316         instructions.
26317         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
26318         constant definitions.
26319         ("nop"): lr 0,0 -> nopr r0
26320         ("nop_lr0", "nop_lr1"): New insn definitions.
26322 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
26324         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
26325         NDS32_V3PUSH_AVAILABLE_P macro.
26327 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
26328             Chung-Ju Wu  <jasonwucj@gmail.com>
26330         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
26331         (nds32*-*-*): Add float and fpu_config into supported_defaults.
26332         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
26333         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
26334         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
26335         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
26336         * config/nds32/constraints.md: New constraints and checking for hard
26337         float configuration.
26338         * config/nds32/iterators.md: New mode iterator and attribute for hard
26339         float configuration.
26340         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
26341         patterns.
26342         * config/nds32/nds32-fpu.md: New file.
26343         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
26344         deal with hard float code generation.
26345         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
26346         ARCH_V3S.
26347         (abi_type, float_reg_number): New enum type.
26348         * config/nds32/nds32-predicates.c: New predicates for hard float.
26349         * config/nds32/nds32-protos.h: Declare functions for hard float.
26350         * config/nds32/nds32.c: Implementation for hard float configuration.
26351         * config/nds32/nds32.h: Definitions for hard float configuration.
26352         * config/nds32/nds32.md: Include hard float machine description and
26353         modify patterns for hard float configuration.
26354         * config/nds32/nds32.opt: New options for hard float configuration.
26355         * config/nds32/predicates.md: New predicates for hard float
26356         configuration.
26358 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
26360         * common/config/nds32/nds32-common.c
26361         (nds32_option_optimization_table): Enable -mreleax-hint by default.
26363 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
26365         PR middle-end/85195
26366         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
26367         CONSTRUCTOR_ELT (ctor, ...)->value.
26369 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
26371         PR target/85193
26372         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
26374 2018-04-05  Tom de Vries  <tom@codesourcery.com>
26376         PR target/85204
26377         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
26378         cond jump.
26380 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
26381             Kito Cheng  <kito.cheng@gmail.com>
26383         * config/nds32/constraints.md (U33): Fine-tune checking condition.
26384         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
26385         * config/nds32/nds32.h (nds32_16bit_address_type): Add
26386         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
26388 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
26389             Kito Cheng  <kito.cheng@gmail.com>
26391         * config/nds32/constraints.md (Ufe): New memory constraint.
26392         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
26393         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
26394         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
26395         operands.
26396         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
26397         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
26399 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26401         * config/nds32/nds32.md: Use optimize_size in the condition for
26402         alu-shift instructions.
26404 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26406         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
26408 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26410         * config/nds32/nds32.md (negsi2): Refine pattern.
26412 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
26413             Chung-Ju Wu  <jasonwucj@gmail.com>
26415         * config/nds32/iterators.md (shift_rotate): New code iterator.
26416         (shift): New code attribute.
26417         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
26418         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
26419         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
26420         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
26421         bit-wise operations.
26422         (andsi3, *andsi3): Ditto.
26423         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
26424         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
26425         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
26426         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
26427         nds32_ior_operand, nds32_xor_operand): New predicates.
26429 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26431         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
26432         (addsi3, subsi3): ... this.
26434 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26436         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
26438 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26440         * config/nds32/nds32.md: Adjust indention.
26442 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
26444         * config/nds32/nds32.md (feature): New attribute.
26446 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
26448         * config/nds32/nds32.md (subtype): New attribute.
26450 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
26452         PR target/85203
26453         * config/arm/arm-builtins.c (arm_expand_builtin): Change
26454         expansion to perform a bitwise AND of the argument followed by a
26455         boolean negation of the result.
26457 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
26459         PR rtl-optimization/84878
26460         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
26461         the basic block.  Assert the use reference is not artificial and that
26462         it has an associated insn.
26464 2018-04-04  Michael Matz  <matz@suse.de>
26466         * builtins.c (compute_objsize): Pass correct operand
26467         to array_at_struct_end_p.
26469 2018-04-04  Richard Biener  <rguenther@suse.de>
26471         PR lto/85176
26472         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
26473         from contexts for DINFO_LEVEL_TERSE and below.
26475 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
26477         * config/nds32/nds32-doubleword.md (move_<mode>): Require
26478         resiter_operand condition.
26479         * config/nds32/nds32.md (*move<mode>): Ditto.
26481 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
26482             Monk Chiang  <sh.chiang04@gmail.com>
26484         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
26486 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
26488         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
26490 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
26491             Kito Cheng  <kito.cheng@gmail.com>
26493         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
26494         nds32_cond_code_str, output_cond_branch,
26495         output_cond_branch_compare_zero, nds32_expand_cbranch,
26496         nds32_expand_cstore, nds32_expand_movcc,
26497         nds32_output_cbranchsi4_equality_zero,
26498         nds32_output_cbranchsi4_equality_reg,
26499         nds32_output_cbranchsi4_equality_reg_or_const_int,
26500         nds32_output_cbranchsi4_greater_less_zero: New functions.
26501         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
26502         nds32_expand_cstore, nds32_expand_movcc,
26503         nds32_output_cbranchsi4_equality_zero,
26504         nds32_output_cbranchsi4_equality_reg,
26505         nds32_output_cbranchsi4_equality_reg_or_const_int,
26506         nds32_output_cbranchsi4_greater_less_zero): Declare.
26507         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
26508         nds32_rimm11s_operand): New predicates.
26509         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
26510         * config/nds32/nds32.md: Rewrite all the branch and conditional move
26511         patterns.
26513 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
26515         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
26516         * config/nds32/nds32.md: Ditto.
26517         * config/nds32/pipelines.md: Ditto.
26519 2018-04-04  Richard Biener  <rguenther@suse.de>
26521         PR tree-optimization/85168
26522         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
26523         propagating abnormals.
26525 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
26527         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
26529 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
26530             Kito Cheng  <kito.cheng@gmail.com>
26532         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
26533         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
26534         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
26535         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
26536         * config/nds32/nds32.md (sibcall_internal): New.
26537         (sibcall_register): Remove.
26538         (sibcall_immediate): Remove.
26539         (sibcall_value_internal): New.
26540         (sibcall_value_register): Remove.
26541         (sibcall_value_immediate): Remove.
26542         * config/nds32/predicates.md (nds32_general_register_operand): New.
26543         (nds32_call_address_operand): New.
26545 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
26547         PR rtl-optimization/85167
26548         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
26549         bb_defs if *split_p, instead preinitialize it to NULL.
26551         PR tree-optimization/85156
26552         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
26553         evaluating the argument multiple times.
26555 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
26557         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
26558         than vector.
26559         (_mm_cvtpd_ps): Likewise.
26560         (_mm_cvttpd_epi32): Likewise.
26561         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
26562         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
26563         vector, pixel, and bool following altivec.h include.
26565 2018-04-03  Martin Sebor  <msebor@redhat.com>
26567         * doc/extend.texi (Common Function Attributes): Clarify.
26568         (const attribute): Likewise.
26569         (pure attribute): Likewise.
26571 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
26573         PR target/85169
26574         * config/i386/i386.c (ix86_expand_vector_set): Use
26575         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
26577 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
26579         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
26580         instructions when changing rounding bits to preserve precision bits
26581         in the x87 control word.
26583 2018-04-03  Martin Liska  <mliska@suse.cz>
26585         PR tree-optimization/82491
26586         * rtl.h (strip_offset_and_add): Replace += suboffset with
26587         poly_uint64 () + suboffset.
26589 2018-03-29  Martin Liska  <mliska@suse.cz>
26590             Martin Jambor  <mjambor@suse.cz>
26592         PR ipa/84947
26593         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
26594         param_type is not an integral or pointer type.
26596 2018-04-03  Richard Biener  <rguenther@suse.de>
26598         * sese.h (recompute_all_dominators): Remove.
26600 2018-04-02  Martin Sebor  <msebor@redhat.com>
26602         * doc/invoke.texi (-Wrestrict): Fix typos.
26604 2018-04-02  Jim Wilson  <jimw@sifive.com>
26606         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
26607         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
26608         (<optab>di3, <optab>si3_extend): Likewise.
26609         (<optab>si3_mask, <optab>si3_mask_1): New.
26610         (<optab>di3_mask, <optab>di3_mask_1): New.
26611         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
26612         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
26613         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
26615 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
26617         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
26618         example.
26620 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
26622         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
26623         (nds32_canonicalize_comparison): New function.
26625 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
26626             Kito Cheng  <kito.cheng@gmail.com>
26627             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
26629         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
26630         * config/nds32/constants.md (unspec_volatile_element): Add
26631         UNSPEC_VOLATILE_RELAX_GROUP.
26632         * config/nds32/nds32-relax-opt.c: New file.
26633         * config/nds32/nds32-predicates.c
26634         (nds32_symbol_load_store_p): New function.
26635         * config/nds32/nds32-protos.h
26636         (nds32_symbol_load_store_p): Declare function.
26637         (make_pass_nds32_relax_opt): Declare new rtl pass function.
26638         * config/nds32/nds32.c
26639         (nds32_register_pass): New function to register pass.
26640         (nds32_register_passes): New function to register passes.
26641         * config/nds32/nds32.md (relax_group): New pattern.
26642         * config/nds32/nds32.opt (mrelax-hint): New option.
26643         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
26645 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
26647         * config/nds32/t-nds32: Modify files dependency.
26649 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
26651         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
26652         (PROFILE_HOOK): Define its implementation.
26654 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
26656         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
26657         type and 32-bit size.
26659 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
26661         PR middle-end/85090
26662         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
26663         (V_128_256): New mode iterator.
26664         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
26665         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
26666         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
26667         of V.
26668         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
26669         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
26671 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
26673         PR target/83315
26674         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
26675         NaN inputs correctly.
26677 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
26679         PR target/80546
26680         * config/rs6000/vsx.md (??r): New mode attribute.
26681         (*vsx_mov<mode>_64bit): Use it.
26682         (*vsx_mov<mode>_32bit): Likewise.
26684 2018-03-30  Martin Sebor  <msebor@redhat.com>
26686         PR tree-optimization/84818
26687         * builtins.c (check_access): Use warning_n.
26689 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26691         PR target/83822
26692         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
26693         condition.
26694         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
26695         condition.
26697 2018-03-30  Julia Koval  <julia.koval@intel.com>
26699         PR target/84413
26700         * x86-tune.def (movx, partial_reg_dependency): Enable for
26701         m_SKYLAKE_AVX512.
26703 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
26705         PR inline-asm/84985
26706         * lra-constraints.c (process_alt_operands): Move setting
26707         this_alternative_matches below.
26709 2018-03-29  Martin Liska  <mliska@suse.cz>
26711         PR lto/84995.
26712         * doc/invoke.texi: Document how LTO works with debug info.
26713         Describe auto-load support of binutils.  Mention 'x86-64'
26714         as valid option value of -march option.
26716 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
26718         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
26720         PR c/85094
26721         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
26722         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
26723         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
26724         checking.
26726 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
26728         PR target/84912
26729         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
26730         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
26731         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
26732         for RS6000_BTM_POWERPC64.
26733         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
26734         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
26735         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
26736         definition.
26737         (DIVDE): Use it.
26738         (DIVDEU): Likewise.
26740 2018-03-28  Carl Love  <cel@us.ibm.com>
26742         Revert
26743         2017-09-27  Carl Love  <cel@us.ibm.com>
26745         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
26746         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
26747         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
26748         fctiw instruction.
26750 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26752         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
26753         instead of __vector bool.
26754         (_mm_max_pu8): Likewise.
26755         (_mm_min_pi16): Likewise.
26757 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
26759         PR target/84912
26760         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
26761         (DIVWEUO): Likewise.
26762         (DIVDEO): Likewise.
26763         (DIVDEUO): Likewise.
26764         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
26765         DIVWEUO and DIVDEUO.
26766         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
26767         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
26768         (div_extend): Likewise.
26769         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
26770         builtin function.
26771         (__builtin_divweuo): Likewise.
26772         (__builtin_divdeo): Likewise.
26773         (__builtin_divdeuo): Likewise.
26775 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
26777         PR target/85095
26778         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
26779         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
26781         PR tree-optimization/82004
26782         * gimple-match-head.c (optimize_pow_to_exp): New function.
26783         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
26784         Don't fold to exp if optimize_pow_to_exp is false.
26786 2018-03-28  Martin Liska  <mliska@suse.cz>
26788         PR other/84819
26789         * calls.c (initialize_argument_information): Fix trailing space.
26790         * common.opt: Fix typo and provide better explanation for
26791         -fsanitize-coverage option.
26792         * config/i386/i386.opt: Fix typo.
26794 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
26795             Martin Liska  <mliska@suse.cz>
26797         PR sanitizer/85081
26798         * gimplify.c (asan_poison_variable): Don't do the check for
26799         gimplify_omp_ctxp here.
26800         (gimplify_decl_expr): Do it here.
26801         (gimplify_target_expr): Likewise.
26803 2018-03-28  Martin Liska  <mliska@suse.cz>
26805         PR target/84988
26806         * config/i386/i386.c (ix86_function_arg_advance): Do not call
26807         chkp_type_bounds_count if MPX is not enabled.
26809 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
26811         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
26813 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
26815         PR target/84914
26816         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
26817         function to create the function decl for complex long double
26818         multiply and divide for -mabi=ieeelongdouble.
26819         (init_float128_ieee): Call it.
26821 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
26823         PR target/85044
26824         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
26825         -fcf-protection=branch -mibt.
26826         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
26828 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26830         PR target/81863
26831         * config/arm/arm.c (arm_valid_symbolic_address): Handle
26832         arm_word_relocations.
26834 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
26836         PR target/85056
26837         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
26838         extern array declarations.
26840 2018-03-27  Richard Biener  <rguenther@suse.de>
26842         PR middle-end/84067
26843         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
26844         explicit single_use checks.
26846 2018-03-27  Richard Biener  <rguenther@suse.de>
26848         PR tree-optimization/85082
26849         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
26850         Valueize the VUSE.
26852 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26854         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
26855         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
26856         Turn on fasynchronous-unwind-tables and funwind-tables.
26858 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
26860         PR target/85073
26861         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
26862         (*bmi_blsr_<mode>_ccz): Ditto.
26864 2018-03-26  Tom de Vries  <tom@codesourcery.com>
26866         PR tree-optimization/85063
26867         * omp-general.c (offloading_function_p): New function.  Factor out
26868         of ...
26869         * omp-offload.c (pass_omp_target_link::gate): ... here.
26870         * omp-general.h (offloading_function_p): Declare.
26871         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
26872         with attribute omp declare target for offloading functions.
26874 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
26876         PR tree-optimization/84005
26877         * tree-data-ref.h (get_base_for_alignment): Declare.
26878         * tree-data-ref.c (get_base_for_alignment_1): New function.
26879         (get_base_for_alignment): Likewise.
26880         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26881         get_base_for_alignment to find a suitable base object, instead
26882         of always using drb->base_address.
26884 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
26886         PR inline-asm/85022
26887         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
26888         known size by default.
26890 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
26892         PR inline-asm/85030
26893         * lra-constraints.c (process_alt_operands): Don't match BLKmode
26894         and non BLKmode operands.
26896 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26898         PR target/85026
26899         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
26900         Clean up attributes.
26902 2018-03-23  Richard Biener  <rguenther@suse.de>
26904         PR debug/85020
26905         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
26906         we are going to emit early debug for LTO.
26908 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
26910         PR inline-asm/85034
26911         * function.c (match_asm_constraints_1): Don't optimize if input
26912         doesn't satisfy general_operand predicate for output's mode.
26914         PR inline-asm/85022
26915         * alias.c (write_dependence_p): Don't require for x_canonicalized
26916         non-VOIDmode if x has VOIDmode.
26918         PR sanitizer/85029
26919         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
26920         just don't try to optimize it rather than assert it never happens.
26922 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26924         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
26925         macro expansions for definition of ST_INTERNAL_<mode> and
26926         LD_INTERNAL_<mode> builtins.
26927         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
26928         Remove prototype.
26929         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
26930         function.
26931         (altivec_expand_st_builtin): Likewise.
26932         (altivec_expand_builtin): Remove calls to deleted functions.
26933         (rs6000_address_for_altivec): Delete this function.
26934         * config/rs6000/vector.md: Remove expands for
26935         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
26937 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
26939         PR target/84826
26940         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
26941         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
26942         re-computing once computed.
26943         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
26944         (arm_init_machine_status): Initialize
26945         machine->static_chain_stack_bytes.
26947 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26949         PR target/84760
26950         * doc/extend.texi: Add four new prototypes for vec_ld.
26951         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
26952         definitions for more logical presentation.
26953         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
26954         entries for V1TI variants of __builtin_altivec_ld builtin.
26955         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
26956         handling of V1TI variant of LVX icode pattern.
26957         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
26958         (rs6000_gimple_fold_builtin): Likewise.
26959         (altivec_init_builtins): Add code to define
26960         __builtin_altivec_lvx_v1ti function.
26962 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
26964         PR inline-asm/84941
26965         * function.c (match_asm_constraints_1): Don't do the optimization
26966         if input isn't a REG, SUBREG, MEM or constant.
26968 2018-03-22  Tom de Vries  <tom@codesourcery.com>
26970         PR tree-optimization/84956
26971         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
26972         bb_has_abnormal_pred.
26974 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
26976         PR sanitizer/85018
26977         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
26978         DECL_INITIAL (decl) to decl at the end.
26979         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
26980         adjust the comment.
26982 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
26984         * doc/extend.texi (__builtin_tgmath): Document when complex
26985         integer types are treated as _Complex _Float64.
26987 2018-03-21  Tom de Vries  <tom@codesourcery.com>
26989         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
26991 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
26993         PR tree-optimization/84960
26994         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
26995         if it is ENTRY block, move them into single succ of ENTRY in that case.
26997 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
26999         PR tree-optimization/84811
27000         * poly-int.h (poly_span_traits): Remove the T3 parameter and
27001         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
27002         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
27003         (known_subrange_p): Update accordingly.  Cast each value involved
27004         in the size comparison, rather than casting the result of the
27005         subtraction.
27007 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
27009         PR tree-optimization/84982
27010         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
27011         by flipping the least significant bit rather than all bits from
27012         bitpos to bitpos + bitsize - 1.
27014 2018-03-21  Nathan Sidwell  <nathan@acm.org>
27016         * doc/extend.texi (Deprecated Features): Remove mention of
27017         long-deleted deprecations.
27019 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27021         PR jit/84288
27022         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
27023         * configure: Regenerate.
27025 2018-03-21  Tom de Vries  <tom@codesourcery.com>
27027         PR tree-optimization/83126
27028         * tree-parloops.c (num_phis): New function.
27029         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
27031 2018-03-21  Nathan Sidwell  <nathan@acm.org>
27033         * doc/extend.texi (Deprecated Features): Update deprecated flags,
27034         mention anon-struct/union members and trailing attributes.
27036 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
27038         PR tree-optimization/84969
27039         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
27040         builtin memset partitions if they set different rhs values.
27042 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
27044         PR rtl-optimization/84989
27045         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
27046         VEC_DUPLICATE with scalar result mode.
27048 2018-03-21  Martin Liska  <mliska@suse.cz>
27050         PR ipa/84963
27051         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
27052         not intended return statement.
27054 2018-03-21  Martin Liska  <mliska@suse.cz>
27056         PR target/84988
27057         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
27058         (chkp_find_bound_slots_1): Limit number of iterations.
27060 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
27062         PR target/84838
27063         * Minor grammar fixes for x86 options.
27065 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
27067         PR debug/84875
27068         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
27069         holding REG_CFA_RESTORE notes, instead turn them into a USE.
27071 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
27073         PR target/83789
27074         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
27075         (altivec_lvx_<mode>_1op): Likewise.
27076         (altivec_stvx_<mode>_2op): Likewise.
27077         (altivec_stvx_<mode>_1op): Likewise.
27078         (altivec_lvx_<VM2:mode>): New define_expand.
27079         (altivec_stvx_<VM2:mode>): Likewise.
27080         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
27081         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
27082         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
27083         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
27084         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
27085         (rs6000_gen_lvx): Likewise.
27086         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
27087         (altivec_expand_stv_builtin): Likewise.
27088         (altivec_expand_builtin): Likewise.
27089         * config/rs6000/vector.md: Likewise.
27091 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27093         PR target/82518
27094         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
27095         BYTES_BIG_ENDIAN.
27097 2018-03-20  Richard Biener  <rguenther@suse.de>
27099         PR target/84986
27100         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
27101         sign-conversions as zero, fall back to standard scalar_stmt
27102         cost for the rest.
27104 2018-03-20  Martin Liska  <mliska@suse.cz>
27106         PR ipa/84825
27107         * predict.c (rebuild_frequencies): Handle case when we have
27108         PROFILE_ABSENT, but flag_guess_branch_prob is false.
27110 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
27112         PR target/84990
27113         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
27114         flag_section_anchors.
27115         * varasm.c (use_blocks_for_decl_p): Remove hack for
27116         dw2_force_const_mem.
27118         PR target/84845
27119         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
27120         to ...
27121         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
27122         be created, use lowpart_subreg of operands[0] rather than operands[0]
27123         itself.
27124         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
27125         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
27126         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
27127         and n constraint instead of aarch64_shift_imm_di and Usd.
27128         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
27129         (*aarch64_<optab>_reg_minus<mode>3): ... this.
27131 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
27133         PR target/82989
27134         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
27135         to favor GPR over NEON registers.
27136         (<shift>di3_neon): Likewise.
27138 2018-03-20  Tom de Vries  <tom@codesourcery.com>
27140         PR target/84952
27141         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
27142         (nvptx_process_pars): Emit bar.sync asap and alap.
27144 2018-03-20  Tom de Vries  <tom@codesourcery.com>
27146         PR target/84954
27147         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
27148         seen_label if seen_label is already set.
27150 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
27152         PR target/84945
27153         * config/i386/i386.c (fold_builtin_cpu): For features above 31
27154         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
27155         Use 1U instead of 1.  Formatting fixes.
27157         PR c/84953
27158         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
27159         instead of TREE_TYPE (s1) for the return value.
27161 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
27163         PR tree-optimization/84946
27164         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
27165         bitsize + bitsize in poly_uint64 rather than poly_int64.
27167         PR sanitizer/78651
27168         * dwarf2asm.c: Include fold-const.c.
27169         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
27170         of decl rather than decl itself.
27172         PR rtl-optimization/84643
27173         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
27175 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
27177         PR sanitizer/78651
27178         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
27179         calling assemble_variable.
27181 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
27183         PR target/81647
27184         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
27185         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
27187 2018-03-19  Jim Wilson  <jimw@sifive.com>
27189         PR bootstrap/84856
27190         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
27191         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
27192         (riscv_first_stack_step): Likewise.
27193         (riscv_option_override): Use STACK_BOUNDARY instead of
27194         MIN_STACK_BOUNDARY.
27195         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
27196         MIN_STACK_BOUNDARY.
27197         (BIGGEST_ALIGNMENT): Set to 128.
27198         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
27199         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
27200         STACK_BOUNDARY.
27202 2018-03-19  Richard Biener  <rguenther@suse.de>
27204         PR tree-optimization/84933
27205         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
27206         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
27208 2018-03-19  Richard Biener  <rguenther@suse.de>
27210         PR tree-optimization/84859
27211         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
27212         (cond_if_else_store_replacement): Perform sinking operation on
27213         single-store BBs regardless of MAX_STORES_TO_SINK setting.
27214         Generalize what a BB with a single eligible store is.
27216 2018-03-19  Richard Biener  <rguenther@suse.de>
27218         PR tree-optimization/84929
27219         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
27220         chrec_is_positive against non-chrec arg.
27222 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
27224         PR target/84711
27225         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
27227 2018-03-18  Martin Liska  <mliska@suse.cz>
27229         PR rtl-optimization/84635
27230         * regrename.c (build_def_use): Use matches_mode only when
27231         matches >= 0.
27233 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
27235         PR tree-optimization/84913
27236         * tree-vect-loop.c (vectorizable_reduction): Don't try to
27237         vectorize chains of COND_EXPRs.
27239 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
27241         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
27243 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
27245         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
27247 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
27249         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
27251 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
27252             Kito Cheng  <kito.cheng@gmail.com>
27254         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
27255         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
27256         (nds32_adjust_reg_alloc_order): New function.
27257         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
27259 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
27261         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
27262         nds32_print_operand, nds32_print_operand_address): Use
27263         HOST_WIDE_INT_PRINT_DEC instead.
27265 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
27267         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
27269 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
27271         PR target/84902
27272         * config/i386/i386.c (initial_ix86_tune_features,
27273         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
27274         unsigned long long.
27275         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
27276         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
27277         rather than 1u << ix86_tune.  Formatting fix.
27278         (ix86_option_override_internal): Change ix86_arch_mask from
27279         unsigned int to unsigned HOST_WIDE_INT, initialize to
27280         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
27281         (ix86_function_specific_restore): Likewise.
27283 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
27285         PR target/84899
27286         * postreload.c (reload_combine_recognize_pattern): Perform
27287         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
27288         truncate_int_for_mode the result for the destination's mode.
27290         PR c/84909
27291         * hsa-gen.c (mem_type_for_type): Fix comment typo.
27292         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
27293         Likewise.
27294         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27295         Likewise.
27297 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
27299         PR target/84876
27300         * lra-assigns.c (lra_split_hard_reg_for): Don't use
27301         regno_allocno_class_array and sorted_pseudos.
27302         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
27303         insns where regno is used.
27305 2018-03-16  Martin Liska  <mliska@suse.cz>
27307         PR ipa/84833
27308         * multiple_target.c (create_dispatcher_calls): Redirect
27309         reference in the symbol table.
27311 2018-03-16  Martin Liska  <mliska@suse.cz>
27313         PR ipa/84722
27314         * multiple_target.c (create_dispatcher_calls): Redirect also
27315         an alias.
27317 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
27319         PR c++/79937
27320         PR c++/82410
27321         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
27322         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
27323         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
27325 2018-03-16  Julia Koval  <julia.koval@intel.com>
27327         * doc/invoke.texi (Skylake Server): Add CLWB.
27328         Cannonlake): Remove CLWB.
27330 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
27332         PR tree-optimization/84841
27333         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
27334         1 << 3.
27335         (FLOAT_ONE_CONST_TYPE): Define.
27336         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
27337         (sort_by_operand_rank): Put entries with higher constant_type last
27338         rather than first to match comments.
27340 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
27342         * config/nios2/nios2.md (movsi_internal): Fix thinko in
27343         split predicate.
27345 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
27347         PR c++/79085
27348         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
27349         check and use address of target always.
27351 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
27353         PR target/84574
27354         * config/i386/i386.c (indirect_thunk_needed): Update comments.
27355         (indirect_thunk_bnd_needed): Likewise.
27356         (indirect_thunks_used): Likewise.
27357         (indirect_thunks_bnd_used): Likewise.
27358         (indirect_return_needed): New.
27359         (indirect_return_bnd_needed): Likewise.
27360         (output_indirect_thunk_function): Add a bool argument for
27361         function return.
27362         (output_indirect_thunk_function): Don't generate alias for
27363         function return thunk.
27364         (ix86_code_end): Call output_indirect_thunk_function to generate
27365         function return thunks.
27366         (ix86_output_function_return): Set indirect_return_bnd_needed
27367         and indirect_return_needed instead of indirect_thunk_bnd_needed
27368         and indirect_thunk_needed.
27370 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
27372         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
27373         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
27374         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
27376 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
27377             Paul Hua  <paul.hua.gm@gmail.com>
27379         PR c/84852
27380         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
27382 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
27384         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
27385         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
27386         resp. SFmode cases.
27388 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
27390         PR target/84711
27391         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
27392         instead of GET_MODE_SIZE when comparing Units.
27394 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
27396         PR target/68256
27397         * varasm.c (hash_section): Return an unchangeble hash value
27398         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
27399         Return !aarch64_can_use_per_function_literal_pools_p ().
27401 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
27403         PR target/84860
27404         * optabs.c (emit_conditional_move): Pass address of cmode's copy
27405         rather than address of cmode as last argument to prepare_cmp_insn.
27407 2018-03-15  Julia Koval  <julia.koval@intel.com>
27409         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
27410         F_AVX512VNNI, F_AVX512BITALG): New.
27412 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
27414         PR target/83451
27415         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
27416         insn for floating-point loads and stores.
27418 2018-03-14  Carl Love  <cel@us.ibm.com>
27420         * config/rs6000/rs6000-c.c: Add macro definitions for
27421         ALTIVEC_BUILTIN_VEC_PERMXOR.
27422         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
27423         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
27424         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
27425         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
27426         UNSPEC_VPERMXOR.
27427         * config/doc/extend.texi: Add prototypes for vec_permxor.
27429 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
27431         PR c/84852
27432         * diagnostic-show-locus.c (class layout_point): Convert m_line
27433         from int to linenum_type.
27434         (line_span::comparator): Use linenum "compare" function when
27435         comparing line numbers.
27436         (test_line_span): New function.
27437         (layout_range::contains_point): Convert param "row" from int to
27438         linenum_type.
27439         (layout_range::intersects_line_p): Likewise.
27440         (layout::will_show_line_p): Likewise.
27441         (layout::print_source_line): Likewise.
27442         (layout::should_print_annotation_line_p): Likewise.
27443         (layout::print_annotation_line): Likewise.
27444         (layout::print_leading_fixits): Likewise.
27445         (layout::annotation_line_showed_range_p): Likewise.
27446         (struct line_corrections): Likewise for field m_row.
27447         (line_corrections::line_corrections): Likewise for param "row".
27448         (layout::print_trailing_fixits): Likewise.
27449         (layout::get_state_at_point): Likewise.
27450         (layout::get_x_bound_for_row): Likewise.
27451         (layout::print_line): Likewise.
27452         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
27453         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
27454         * input.c (selftest::test_linenum_comparisons): New function.
27455         (selftest::input_c_tests): Call it.
27456         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
27457         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
27458         * selftest.h (ASSERT_GT): New macro.
27459         (ASSERT_GT_AT): New macro.
27460         (ASSERT_LT): New macro.
27461         (ASSERT_LT_AT): New macro.
27463 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
27465         PR rtl-optimization/84780
27466         * combine.c (distribute_links): Don't make a link based on pc_rtx.
27468 2018-03-14  Martin Liska  <mliska@suse.cz>
27470         * tree.c (record_node_allocation_statistics): Use
27471         get_stats_node_kind.
27472         (get_stats_node_kind): New function extracted from
27473         record_node_allocation_statistics.
27474         (free_node): Use get_stats_node_kind.
27476 2018-03-14  Richard Biener  <rguenther@suse.de>
27478         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
27479         that the value-set of ANTIC_IN doesn't grow.
27481         Revert
27482         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27483         member.
27484         (BB_VISITED_WITH_VISITED_SUCCS): New define.
27485         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27487 2018-03-14  Julia Koval  <julia.koval@intel.com>
27489         * config.gcc (icelake-client, icelake-server): New.
27490         (icelake): Remove.
27491         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
27492         (initial_ix86_arch_features): Ditto.
27493         (PTA_SKYLAKE): Add SGX.
27494         (PTA_ICELAKE): Remove.
27495         (PTA_ICELAKE_CLIENT): New.
27496         (PTA_ICELAKE_SERVER): New.
27497         (ix86_option_override_internal): Split up icelake on icelake client and
27498         icelake server.
27499         (get_builtin_code_for_version): Ditto.
27500         (fold_builtin_cpu): Ditto.
27501         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
27502         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
27503         * config/i386/i386.h (processor_type): Ditto.
27504         * doc/invoke.texi: Ditto.
27506 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
27508         PR sanitizer/83392
27509         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
27510         INTEGER_CST offset, add it together with bitpos / 8 and
27511         sign extend based on POINTER_SIZE.
27513         PR target/84844
27514         Revert
27515         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
27517         PR target/78090
27518         * config/i386/constraints.md (Yc): New register constraint.
27519         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
27520         Use Yc constraint for alternative 2 of operand 0.  Remove
27521         preferred_for_speed attribute.
27523 2018-03-14  Richard Biener  <rguenther@suse.de>
27525         PR tree-optimization/84830
27526         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
27527         with the old one to avoid oscillations.
27529 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
27531         PR target/83712
27532         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
27533         pseudos.
27534         (assign_by_spills): Return a flag of reload assignment failure.
27535         Do not process the reload assignment failures.  Do not spill other
27536         reload pseudos if they has the same reg class.  Update n if
27537         necessary.
27538         (lra_assign): Add a return arg.  Set up from the result of
27539         assign_by_spills call.
27540         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27541         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
27542         usage_insns if it is not NULL.
27543         (spill_hard_reg_in_range): New function.
27544         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27545         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27546         function prototypes.
27547         (lra_assign): Change prototype.
27548         * lra.c (lra): Add code to deal with fails by splitting hard reg
27549         live ranges.
27551 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
27553         * config/riscv/riscv.opt (mrelax): New option.
27554         * config/riscv/riscv.c (riscv_file_start): Emit ".option
27555         "norelax" when riscv_mrelax is disabled.
27556         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
27558 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27560         PR target/84743
27561         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
27562         reassociation for int modes.
27564 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
27566         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
27567         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
27568         for big-endian.
27569         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
27570         * config/aarch64/aarch64-sve.md
27571         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
27572         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
27573         (*extend<mode><Vwide>2): Rename to...
27574         (aarch64_sve_extend<mode><Vwide>2): ...this.
27575         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
27576         renaming the old pattern to...
27577         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
27578         unsigned packs.
27579         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
27580         define_expand, renaming the old pattern to...
27581         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
27582         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
27583         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
27584         account when deciding which SVE instruction the optab should use.
27585         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
27587 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
27589         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
27590         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
27591         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
27592         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
27593         (tlsdesc_small_<mode>): Turn a define_expand and use
27594         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
27595         (tlsdesc_small_advsimd_<mode>): ...this.
27596         (tlsdesc_small_sve_<mode>): New pattern.
27598 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
27600         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
27601         (UNSPEC_UMUL_HIGHPART): New constants.
27602         (MUL_HIGHPART): New int iteraor.
27603         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
27604         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
27605         define_expand.
27606         (*<su>mul<mode>3_highpart): New define_insn.
27608 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
27610         PR lto/84805
27611         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
27612         incomplete types.
27614 2018-03-13  Martin Liska  <mliska@suse.cz>
27616         PR ipa/84658.
27617         * (sem_item_optimizer::sem_item_optimizer): Initialize new
27618         vector.
27619         (sem_item_optimizer::~sem_item_optimizer): Release it.
27620         (sem_item_optimizer::merge_classes): Register variable aliases.
27621         (sem_item_optimizer::fixup_pt_set): New function.
27622         (sem_item_optimizer::fixup_points_to_sets): Likewise.
27623         * ipa-icf.h: Declare new variables and functions.
27625 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
27627         PR middle-end/84834
27628         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
27629         integer_pow2p@2 and test integer_pow2p in condition.
27630         (A < 0 ? C : 0): Similarly for @1.
27632         PR middle-end/84831
27633         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
27634         characters starting at p contain '\0' character, don't look beyond
27635         that.
27637         PR target/84827
27638         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
27639         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
27641         PR target/84828
27642         * reg-stack.c (change_stack): Change update_end var from int to
27643         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
27644         also call set_block_for_insn on the newly added insns and rescan.
27646         PR target/84786
27647         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
27648         on the last operand.
27650         PR c++/84704
27651         * tree.c (stabilize_reference_1): Return save_expr (e) for
27652         STATEMENT_LIST even if it doesn't have side-effects.
27654 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
27656         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
27658 2018-03-12  Renlin Li  <renlin.li@arm.com>
27660         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
27661         aarch64_output_scalar_simd_mov_immediate.
27663 2018-03-12  Martin Sebor  <msebor@redhat.com>
27665         PR tree-optimization/83456
27666         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
27667         for perfectly overlapping calls to memcpy.
27668         (gimple_fold_builtin_memory_chk): Same.
27669         (gimple_fold_builtin_strcpy): Handle no-warning.
27670         (gimple_fold_builtin_stxcpy_chk): Same.
27671         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
27673 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
27675         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
27676         parameter.  Use it for SFmode.
27677         (rs6000_function_arg_advance_1): Adjust.
27678         (rs6000_function_arg): Adjust.
27679         (rs6000_gimplify_va_arg): Pass false for that new parameter.
27681 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
27683         PR rtl-optimization/84169
27684         PR rtl-optimization/84780
27685         * combine.c (can_combine_p): Check for a 2-insn combination whether
27686         the destination register is used between the two insns, too.
27688 2018-03-12  Richard Biener  <rguenther@suse.de>
27690         PR tree-optimization/84803
27691         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
27692         for refs DR analysis didn't process.
27694 2018-03-12  Richard Biener  <rguenther@suse.de>
27696         PR tree-optimization/84777
27697         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
27698         force-vectorize loops ignore whether we are optimizing for size.
27700 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
27702         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
27703         (TARGET_MD_ASM_ADJUST): Define.
27705 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
27706             Kito Cheng  <kito.cheng@gmail.com>
27707             Chung-Ju Wu  <jasonwucj@gmail.com>
27709         * config/nds32/nds32.c (nds32_compute_stack_frame,
27710         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
27711         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
27712         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
27713         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
27714         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
27715         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
27716         * config/nds32/nds32.md (prologue, epilogue): Use macro
27717         NDS32_V3PUSH_AVAILABLE_P to do checking.
27719 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
27721         PR debug/58150
27722         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
27723         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
27724         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
27725         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
27726         addition of most attributes on !orig_type_die or the attribute not
27727         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
27729 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
27730             Chung-Ju Wu  <jasonwucj@gmail.com>
27732         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
27733         __NDS32_VH__ macro.
27734         * config/nds32/nds32.opt (mvh): New option.
27736 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
27737             Chung-Ju Wu  <jasonwucj@gmail.com>
27739         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
27740         function.
27741         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
27742         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
27743         definition.
27745 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
27746             Chung-Ju Wu  <jasonwucj@gmail.com>
27748         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
27749         function.
27750         * config/nds32/nds32-multiple.md (strlensi): New pattern.
27751         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
27753 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
27754             Kito Cheng  <kito.cheng@gmail.com>
27755             Chung-Ju Wu  <jasonwucj@gmail.com>
27757         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
27758         UNSPEC_FFMISM and UNSPEC_FLMISM.
27759         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
27760         for ffb, ffmism and flmism.
27761         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
27762         (unspec_ffmism): Ditto.
27763         (unspec_flmism): Ditto.
27764         (nds32_expand_builtin_impl): Check if string extension is available.
27765         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
27766         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
27768 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
27770         Reverting patch:
27771         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
27773         PR target/83712
27774         * lra-assigns.c (assign_by_spills): Return a flag of reload
27775         assignment failure.  Do not process the reload assignment
27776         failures.  Do not spill other reload pseudos if they has the same
27777         reg class.
27778         (lra_assign): Add a return arg.  Set up from the result of
27779         assign_by_spills call.
27780         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27781         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
27782         usage_insns if it is not NULL.
27783         (spill_hard_reg_in_range): New function.
27784         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27785         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27786         function prototypes.
27787         (lra_assign): Change prototype.
27788         * lra.c (lra): Add code to deal with fails by splitting hard reg
27789         live ranges.
27791 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
27793         PR target/84807
27794         * config/i386/i386.opt: Replace Enforcment with Enforcement.
27796 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
27798         PR debug/84620
27799         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
27800         (dw_val_node): Add val_symbolic_view.
27801         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
27802         (symview_upper_bound): New.
27803         (new_line_info_table): Initialize symviews_since_reset.
27804         (dwarf2out_source_line): Count symviews_since_reset and set
27805         symview_upper_bound.
27806         (dw_val_equal_p): Handle symview.
27807         (add_AT_symview): New.
27808         (print_dw_val): Handle symview.
27809         (attr_checksum, attr_checksum_ordered): Likewise.
27810         (same_dw_val_p, size_of_die): Likewise.
27811         (value_format, output_die): Likewise.
27812         (add_high_low_attributes): Use add_AT_symview for entry_view.
27813         (dwarf2out_finish): Reset symview_upper_bound, clear
27814         zero_view_p.
27816 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
27818         PR target/83969
27819         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
27820         Add strict argument and use it.
27821         (rs6000_split_multireg_move): Update for new strict argument.
27822         (mem_operand_gpr): Disallow all non-offsettable addresses.
27823         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
27825 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
27827         PR target/84772
27828         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
27829         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
27830         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
27832         PR c++/84767
27833         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
27834         decl, use remap_type if we want to use the type.
27836 2018-03-09  Martin Sebor  <msebor@redhat.com>
27838         PR tree-optimization/84526
27839         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27840         Remove dead code.
27841         (builtin_access::generic_overlap): Be prepared to handle non-array
27842         base objects.
27844 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
27846         PR rtl-optimization/84682
27847         * lra-constraints.c (process_address_1): Check is_address flag
27848         for address constraints.
27849         (process_alt_operands): Likewise.
27850         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
27851         preprocess_constraints.
27852         * recog.h (preprocess_constraints): Add oploc parameter.
27853         Adjust callers.
27854         * recog.c (preprocess_constraints): Test address_operand for
27855         CT_ADDRESS constraints.
27857 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
27859         PR target/83712
27860         * lra-assigns.c (assign_by_spills): Return a flag of reload
27861         assignment failure.  Do not process the reload assignment
27862         failures.  Do not spill other reload pseudos if they has the same
27863         reg class.
27864         (lra_assign): Add a return arg.  Set up from the result of
27865         assign_by_spills call.
27866         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27867         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
27868         usage_insns if it is not NULL.
27869         (spill_hard_reg_in_range): New function.
27870         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27871         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27872         function prototypes.
27873         (lra_assign): Change prototype.
27874         * lra.c (lra): Add code to deal with fails by splitting hard reg
27875         live ranges.
27877 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27879         PR target/83193
27880         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
27881         Accept complain bool parameter.  Only emit errors if it is true.
27882         (arm_parse_cpu_option_name): Likewise.
27883         (arm_target_thumb_only): Adjust callers of the above.
27884         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
27885         prototype to take a default true bool parameter.
27886         (arm_parse_arch_option_name): Likewise.
27888 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
27889             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
27891         PR jit/64089
27892         PR jit/84288
27893         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
27894         * configure: Regenerate.
27895         * configure.ac ("linker --version-script option"): New.
27896         ("linker soname option"): New.
27898 2018-03-09  Richard Biener  <rguenther@suse.de>
27900         PR tree-optimization/84775
27901         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
27902         immediate uses of predicate stmts and mark them modified.
27904         Revert
27905         PR tree-optimization/84178
27906         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27907         to caller.
27908         (version_loop_for_if_conversion): Delay update_ssa call.
27909         (tree_if_conversion): Delay update_ssa until after predicate
27910         insertion.
27912 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
27914         PR target/84763
27915         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
27916         when the function accesses prior frames.
27918 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
27920         PR debug/84456
27921         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
27922         gen_llsym, otherwise call maybe_gen_llsym.
27924         PR inline-asm/84742
27925         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
27926         has ',' character inside of it.
27928 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27930         PR target/84748
27931         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
27932         as clobbering CC_REGNUM.
27934 2018-03-08  Richard Biener  <rguenther@suse.de>
27936         PR middle-end/84552
27937         * tree-scalar-evolution.c: Include tree-into-ssa.h.
27938         (follow_copies_to_constant): Do not follow SSA names registered
27939         for update.
27941 2018-03-08  Richard Biener  <rguenther@suse.de>
27943         PR tree-optimization/84178
27944         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27945         to caller.
27946         (version_loop_for_if_conversion): Delay update_ssa call.
27947         (tree_if_conversion): Delay update_ssa until after predicate
27948         insertion.
27950 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
27952         PR tree-optimization/84178
27953         * tree-if-conv.c (release_bb_predicate): Remove the
27954         the assertion that the stmts have NULL use_ops.
27955         Discard the statements, asserting that they haven't
27956         yet been added to a BB.
27958 2018-03-08  Richard Biener  <rguenther@suse.de>
27960         PR tree-optimization/84746
27961         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
27962         (phi_translate): Pass in destination ANTIC_OUT set.
27963         (phi_translate_1): Likewise.  For a simplified result lookup
27964         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
27965         (phi_translate_set): Adjust.
27966         (do_pre_regular_insertion): Likewise.
27967         (do_pre_partial_partial_insertion): Likewise.
27969 2018-03-08  Martin Liska  <mliska@suse.cz>
27971         PR gcov-profile/84735
27972         * doc/gcov.texi: Document usage of profile files.
27973         * gcov-io.h: Document changes in the format.
27975 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
27977         PR debug/84404
27978         PR debug/84408
27979         * dwarf2out.c (struct dw_line_info_table): Update comments for
27980         view == -1.
27981         (FORCE_RESET_NEXT_VIEW): New.
27982         (FORCE_RESETTING_VIEW_P): New.
27983         (RESETTING_VIEW_P): Check for -1 too.
27984         (ZERO_VIEW_P): Likewise.
27985         (new_line_info_table): Force-reset next view.
27986         (dwarf2out_begin_function): Likewise.
27987         (dwarf2out_source_line): Simplify zero_view_p initialization.
27988         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
27989         view directly.  Omit view when omitting .loc at line 0.
27991 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
27993         PR tree-optimization/84740
27994         * tree-switch-conversion.c (process_switch): Call build_constructors
27995         only if info.phi_count is non-zero.
27997         PR tree-optimization/84739
27998         * tree-tailcall.c (find_tail_calls): Check call arguments against
27999         DECL_ARGUMENTS (current_function_decl) rather than
28000         DECL_ARGUMENTS (func) when checking for tail recursion.
28002 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
28004         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
28005         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
28006         Volker Reichelt's entry and add entries for people that perform
28007         GCC fuzzy testing and report numerous bugs.
28009 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
28011         PR target/82411
28012         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
28013         readonly data in sdata, if that is disabled.
28014         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
28015         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
28016         -mreadonly-in-sdata option.
28018 2018-03-07  Martin Sebor  <msebor@redhat.com>
28020         PR tree-optimization/84468
28021         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
28022         basic block when looking for nul assignment.
28024 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
28026         PR target/84277
28027         * except.h (output_function_exception_table): Adjust prototype.
28028         * except.c (output_function_exception_table): Remove FNNAME parameter
28029         and add SECTION parameter.  Ouput one part of the table at a time.
28030         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
28031         the first part of the exception table and emit unwind directives.
28032         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
28033         (i386_pe_seh_cold_init): Likewise.
28034         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
28035         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
28036         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
28037         (ix86_output_call_insn): Emit a nop in one more case for SEH.
28038         * config/i386/winnt.c: Include except.h.
28039         (struct seh_frame_state): Add reg_offset, after_prologue and
28040         in_cold_section fields.
28041         (i386_pe_seh_end_prologue): Set seh->after_prologue.
28042         (i386_pe_seh_cold_init): New function.
28043         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
28044         to seh->in_cold_section.
28045         (seh_emit_push): Record the offset of the push.
28046         (seh_emit_save): Record the offet of the save.
28047         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
28048         Test seh->after_prologue to disregard the epilogue.
28049         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
28050         (i386_pe_end_cold_function): New function.
28052 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
28054         PR fortran/84565
28055         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
28056         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
28058         PR c++/84704
28059         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
28060         on tmp_var.
28061         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
28062         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
28064         PR middle-end/84723
28065         * multiple_target.c: Include tree-inline.h and intl.h.
28066         (expand_target_clones): Diagnose and fail if node->definition and
28067         !tree_versionable_function_p (node->decl).
28069 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
28071         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
28072         sprint_ul.
28073         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
28074         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
28075         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
28077 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
28079         PR target/84710
28080         * combine.c (try_combine): Use reg_or_subregno instead of handling
28081         just paradoxical SUBREGs and REGs.
28083 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
28085         * config/arc/arc.c (arc_finalize_pic): Remove function.
28086         (arc_must_save_register): We use single base PIC register, remove
28087         checks to save/restore the PIC register.
28088         (arc_expand_prologue): Likewise.
28089         * config/arc/arc-protos.h (arc_set_default_type_attributes):
28090         Remove.
28091         (arc_verify_short): Likewise.
28092         (arc_attr_type): Likewise.
28093         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
28094         (walk_stores): Likewise.
28095         (arc_address_cost): Make it static.
28096         (arc_verify_short): Likewise.
28097         (branch_dest): Likewise.
28098         (arc_attr_type): Likewise.
28099         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
28100         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
28101         (arc_final_prescan_insn): Remove inserting the nops due to
28102         hardware hazards.  It is done in reorg step.
28103         (insn_length_variant_t): Remove.
28104         (insn_length_parameters_t): Likewise.
28105         (arc_insn_length_parameters): Likewise.
28106         (arc_get_insn_variants): Likewise.
28107         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
28109 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
28111         PR inline-asm/84683
28112         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
28113         assertion failure.
28115         PR tree-optimization/84687
28116         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
28117         on new_node->decl.
28118         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
28120 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28122         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
28123         Rename to ppc_speculation_barrier.
28124         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
28125         __builtin_ppc_speculation_barrier.
28127 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
28129         PR target/84700
28130         * combine.c (combine_simplify_rtx): Don't try to simplify if
28131         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
28132         are equal to x.
28134 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
28136         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
28137         to 32 bytes when compiling for POWER9.
28139 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
28141         PR target/84564
28142         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
28143         regparm >= 3 with no arg reg available also for calls with
28144         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
28146         PR target/84524
28147         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
28148         orig,vex.
28149         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
28151 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
28153         PR target/84264
28154         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
28156 2018-03-05  Richard Biener  <rguenther@suse.de>
28158         PR tree-optimization/84486
28159         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
28160         When inserting a __builtin_assume_aligned call set the LHS
28161         SSA name alignment info accordingly.
28163 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
28165         PR tree-optimization/84114
28166         * config/aarch64/aarch64.c (aarch64_reassociation_width)
28167         Avoid reassociation of FLOAT_MODE addition.
28169 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
28171         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
28172         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
28173         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
28174         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
28175         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
28176         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
28177         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
28178         and -mwbnoinvd.
28179         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
28180         __builtin_ia32_wbinvd): New builtins.
28181         (SPECIAL_ARGS2): New.
28182         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
28183         (SPECIAL_ARGS2): New.
28184         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
28185         (ix86_valid_target_attribute_inner_p): Ditto.
28186         (ix86_init_mmx_sse_builtins): Add special_args2.
28187         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
28188         TARGET_WBNOINVD_P): New.
28189         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
28190         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
28191         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
28192         * config/i386/immintrin.h (_wbinvd): New intrinsic.
28193         * config/i386/pconfigintrin.h: New file.
28194         * config/i386/wbnoinvdintrin.h: Ditto.
28195         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
28196         wbnoinvdintrin.h.
28197         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
28199 2018-03-05  Richard Biener  <rguenther@suse.de>
28201         PR tree-optimization/84670
28202         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
28203         member.
28204         (BB_VISITED_WITH_VISITED_SUCCS): New define.
28205         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
28206         (compute_antic_aux): Only assert the number of values in ANTIC_IN
28207         doesn't grow if all successors (recursively) were visited at least
28208         once.
28210 2018-03-05  Richard Biener  <rguenther@suse.de>
28212         PR tree-optimization/84650
28213         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
28214         if executed in the loop pipeline.
28216 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
28218         * doc/configfiles.texi (Configuration Files): Move info about
28219         conditionalizing $target-protos.h to...
28220         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
28221         differs from $target-protos.h.
28223 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
28224             Chung-Ju Wu  <jasonwucj@gmail.com>
28226         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
28227         * config/nds32/nds32-multiple.md (setmemsi): Define.
28228         * config/nds32/nds32-memory-manipulation.c
28229         (nds32_gen_dup_4_byte_to_word_value): New.
28230         (emit_setmem_word_loop): New.
28231         (emit_setmem_byte_loop): New.
28232         (nds32_expand_setmem_loop): New.
28233         (nds32_expand_setmem_loop_v3m): New.
28234         (nds32_expand_setmem_unroll): New.
28235         (nds32_expand_setmem): New.
28237 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
28238             Chung-Ju Wu  <jasonwucj@gmail.com>
28240         * config/nds32/nds32-memory-manipulation.c
28241         (nds32_emit_load_store): New.
28242         (nds32_emit_post_inc_load_store): New.
28243         (nds32_emit_mem_move): New.
28244         (nds32_emit_mem_move_block): New.
28245         (nds32_expand_movmemsi_loop_unknown_size): New.
28246         (nds32_expand_movmemsi_loop_known_size): New.
28247         (nds32_expand_movmemsi_loop): New.
28248         (nds32_expand_movmemsi_unroll): New.
28249         (nds32_expand_movmemqi): Rename ...
28250         (nds32_expand_movmemsi): ... to this.
28251         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
28252         (movmemsi): ... to this.
28253         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
28254         (nds32_expand_movmemsi): ... to this.
28256 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
28257             Monk Chiang  <sh.chiang04@gmail.com>
28258             Chung-Ju Wu  <jasonwucj@gmail.com>
28260         * config/nds32/nds32-protos.h
28261         (nds32_expand_load_multiple): New arguments.
28262         (nds32_expand_store_multiple): Ditto.
28263         (nds32_valid_multiple_load_store): Rename ...
28264         (nds32_valid_multiple_load_store_p): ... to this.
28265         * config/nds32/nds32-memory-manipulation.c
28266         (nds32_expand_load_multiple): Refine implementation.
28267         (nds32_expand_store_multiple): Ditto.
28268         * config/nds32/nds32-multiple.md
28269         (load_multiple): Update nds32_expand_load_multiple interface.
28270         (store_multiple): Update nds32_expand_store_multiple interface.
28271         * config/nds32/nds32-predicates.c
28272         (nds32_valid_multiple_load_store): Rename ...
28273         (nds32_valid_multiple_load_store_p): ... to this and refine
28274         implementation.
28275         * config/nds32/predicates.md
28276         (nds32_load_multiple_and_update_address_operation): New predicate.
28277         (nds32_store_multiple_and_update_address_operation): New predicate.
28279 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
28280             Chung-Ju Wu  <jasonwucj@gmail.com>
28282         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
28283         (combo): New attribute.
28284         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
28286 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
28288         * config/nds32/nds32.opt: Change -mcmodel= default value.
28290 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
28291             Monk Chiang  <sh.chiang04@gmail.com>
28292             Chung-Ju Wu  <jasonwucj@gmail.com>
28294         * config/nds32/constants.md (unspec_element): New enum.
28295         * config/nds32/constraints.md (Umw): New constraint.
28296         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
28297         * config/nds32/nds32-intrinsic.md: Likewise.
28298         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
28299         (nds32_valid_smw_lwm_base_p): New.
28300         (nds32_output_smw_single_word): New.
28301         (nds32_output_lmw_single_word): New.
28302         (nds32_expand_unaligned_load): New.
28303         (nds32_expand_unaligned_store): New.
28304         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
28305         (nds32_output_smw_single_word): Declare.
28306         (nds32_output_lmw_single_word): Declare.
28307         (nds32_expand_unaligned_load): Declare.
28308         (nds32_expand_unaligned_store): Declare.
28309         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
28310         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
28311         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
28312         NDS32_BUILTIN_UASTORE_DW.
28313         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
28314         predicate.
28316 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
28317             Kito Cheng  <kito.cheng@gmail.com>
28318             Chung-Ju Wu  <jasonwucj@gmail.com>
28320         * config/nds32/nds32-intrinsic.c
28321         (nds32_expand_builtin_null_ftype_reg): Delete.
28322         (nds32_expand_builtin_reg_ftype_imm): Ditto.
28323         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
28324         (nds32_read_argument): New.
28325         (nds32_legitimize_target): Ditto.
28326         (nds32_legitimize_argument): Ditto.
28327         (nds32_check_constant_argument): Ditto.
28328         (nds32_expand_unop_builtin): Ditto.
28329         (nds32_expand_unopimm_builtin): Ditto.
28330         (nds32_expand_binop_builtin): Ditto.
28331         (nds32_builtin_decl_impl): Ditto.
28332         (builtin_description): Ditto.
28333         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
28334         (nds32_init_builtins_impl): Ditto.
28335         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
28336         (nds32_builtin_decl): New.
28337         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
28338         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
28340 2018-03-02  Jeff Law  <law@redhat.com>
28342         * reorg.c (stop_search_p): Handle DEBUG_INSN.
28343         (redundant_insn, fill_simple_delay_slots): Likewise.
28344         (fill_slots_from_thread): Likewise.
28345         * resource.c (mark_referenced_resources): Likewise.
28346         (mark_set_resources, find_dead_or_set_registers): Likewise.
28348 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
28350         * substring-locations.h (format_warning_va): Formatting fix for
28351         ATTRIBUTE_GCC_DIAG.
28352         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
28353         argument.
28354         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
28355         * substring-locations.c: Include intl.h.
28356         (format_warning_va): Turned into small wrapper around
28357         format_warning_n_va, renamed to ...
28358         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
28359         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
28360         use ngettext.
28361         (format_warning_at_substring_n): New function.
28362         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
28363         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
28364         format_warning_at_substring with just a shorter name instead of
28365         const function pointer.
28366         (fmtwarn_n): New function.
28367         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
28368         appropriate, get rid of all the fmtstr temporaries, move conditionals
28369         with G_() wrapped string literals directly into fmtwarn arguments,
28370         cast dir.len to (int), formatting fixes.
28372 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
28374         * doc/invoke.texi: Remove "Cilk Plus" references.
28376 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
28377             Richard Biener  <rguenther@suse.de>
28379         PR ipa/84628
28380         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
28381         for error or warning attributes if CALL_FROM_THUNK_P is set.
28382         Formatting fixes.
28384 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
28386         PR target/56540
28387         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
28388         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
28390         PR target/56540
28391         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
28392         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
28394         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
28395         instead of -1U in last predictors element's probability member.
28397 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
28399         PR ipa/83983
28400         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
28401         arguments if they are comparable.
28403 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
28405         PR tree-optimization/84634
28406         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
28407         masks and masked_loop_p with a single loop_masks, making sure it's
28408         null for bb vectorization.
28410 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
28412         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
28413         (vect_analyze_data_ref_access): Use loop->safe_len rather than
28414         loop->force_vectorize to check whether there is no alias.
28416 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
28418         PR target/84614
28419         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
28420         prototypes.
28421         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
28422         comments.
28423         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
28424         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
28425         instead of a loop around prev_real_insn.
28426         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
28427         prev_real_insn.
28429         PR inline-asm/84625
28430         * config/i386/i386.c (ix86_print_operand): Use conditional
28431         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
28432         zero vector.
28434 2018-03-02  Richard Biener  <rguenther@suse.de>
28436         PR tree-optimization/84427
28437         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
28438         (bitmap_set_subtract_values): Rewrite to handle multiple
28439         exprs per value.
28440         (clean): Likewise.
28441         (prune_clobbered_mems): Likewise.
28442         (phi_translate): Take edge instead of pred/phiblock.
28443         (phi_translate_1): Likewise.
28444         (phi_translate_set): Likewise.  Insert all translated
28445         exprs for a value into the set, keeping possibly multiple
28446         expressions per value.
28447         (compute_antic_aux): Adjust for phi_translate changes.
28448         When intersecting union the expressions and prune those
28449         not in the final value set, keeping possibly multiple
28450         expressions per value.  Do not use value-insertion
28451         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
28452         all expressions.  Add verification that the value-sets
28453         only shrink during iteration.
28454         (compute_partial_antic_aux): Adjust for the phi_translate changes.
28455         (do_pre_regular_insertion): Likewise.
28456         (do_pre_partial_partial_insertion): Likewise.
28458 2018-03-02  Richard Biener  <rguenther@suse.de>
28460         PR target/82005
28461         * config/darwin.c (saved_debug_info_level): New static global.
28462         (darwin_asm_lto_start): Disable debug info generation for LTO out.
28463         (darwin_asm_lto_end): Restore debug info generation settings.
28465 2018-03-01  Martin Liska  <mliska@suse.cz>
28467         PR sanitizer/82484
28468         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
28469         volatile arguments.
28471 2018-03-01  Richard Biener  <rguenther@suse.de>
28473         PR debug/84645
28474         * dwarf2out.c (gen_variable_die): Properly handle late VLA
28475         type annotation with LTO when debug was disabled at compile-time.
28477 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
28479         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
28480         XINT with INTVAL.
28481         (mips_final_postscan_insn): Likewise.
28483 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
28485         PR rtl-optimization/84528
28486         * alias.c (init_alias_target): Add commentary.
28487         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
28488         a unique base value if the frame pointer is not eliminated
28489         to the stack pointer.
28491 2018-03-01  Tom de Vries  <tom@codesourcery.com>
28493         PR rtl-optimization/83327
28494         * lra-int.h (hard_regs_spilled_into): Declare.
28495         * lra.c (hard_regs_spilled_into): Define.
28496         (init_reg_info): Init hard_regs_spilled_into.
28497         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
28498         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
28499         (process_bb_lives): Handle hard_regs_spilled_into.
28500         (lra_create_live_ranges_1): Before doing liveness propagation, clear
28501         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
28503 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
28505         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
28506         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
28507         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
28508         * config/rs6000/aix72.h: New file.
28510 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
28512         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
28513         instead of warning_at with conditional singular and plural messages
28514         where possible.
28516         PR target/52991
28517         * stor-layout.c (update_alignment_for_field): For
28518         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
28519         && !DECL_PACKED (field), do the alignment update, just use
28520         only desired_align instead of MAX (type_align, desired_align)
28521         as the alignment.
28522         (place_field): Don't do known_align < desired_align handling
28523         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
28524         is non-NULL, instead do it after rli->prev_field handling and
28525         only if not within a bitfield word.  For DECL_PACKED (field)
28526         use type_align of BITS_PER_UNIT.
28528 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
28530         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
28531         superfluous parentheses and trailing spaces.
28533 2018-02-28  Richard Biener  <rguenther@suse.de>
28535         PR tree-optimization/84584
28536         * graphite-scop-detection.c (scop_detection::add_scop): Discard
28537         SCoPs with fake exit edge.
28539 2018-02-28  Martin Liska  <mliska@suse.cz>
28541         PR testsuite/84597
28542         * timevar.c (timer::print): Fix format to properly print 100%
28543         values.
28545 2018-02-28  Richard Biener  <rguenther@suse.de>
28547         PR middle-end/84607
28548         * genmatch.c (capture_info::walk_match): Do not mark
28549         captured expressions without operands as expr_p given
28550         they act more like predicates and should be subject to
28551         "lost tail" side-effect preserving.
28553 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
28555         PR rtl-optimization/81611
28556         * auto-inc-dec.c (attempt_change): Move dead note from
28557         mem_insn if it's the next use of regno
28558         (find_address): Take address use of reg holding
28559         non-incremented value.  Add parm to limit search to the named
28560         reg only.
28561         (merge_in_block): Attempt to use a mem insn that is the next
28562         use of the original regno.
28564 2018-02-27  Martin Sebor  <msebor@redhat.com>
28566         PR c++/83871
28567         * doc/invoke.texi (-Wmissing-attributes): New option.
28568         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
28570 2018-02-27  Martin Sebor  <msebor@redhat.com>
28572         PR translation/84207
28573         * diagnostic-core.h (warning_n, error_n, inform_n): Change
28574         n argument to unsigned HOST_WIDE_INT.
28575         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
28576         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
28577         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
28578         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
28580 2018-02-27  Richard Biener  <rguenther@suse.de>
28582         PR tree-optimization/84512
28583         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
28584         Do not use the estimate returned from record_stmt_cost for
28585         the scalar iteration cost but sum properly using add_stmt_cost.
28587 2018-02-27  Richard Biener  <rguenther@suse.de>
28589         PR tree-optimization/84466
28590         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28591         Adjust last change to less strictly validate use operands.
28593 2018-02-27  Martin Liska  <mliska@suse.cz>
28595         PR gcov-profile/84548
28596         * gcov.c (process_file): Allow partial overlap and consider it
28597         also as group functions.
28598         (output_lines): Properly calculate range of lines for a group.
28600 2018-02-27  Martin Liska  <mliska@suse.cz>
28602         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
28603         'ggc' suffixes.  Change first column width.
28604         (timer::print): Fix formatting of the column.
28606 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
28608         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
28609         preserve inline entry blocks for the sake of debug inline
28610         entry point markers alone.
28611         (remove_unused_locals): Suggest in comments a better place to
28612         force the preservation of inline entry blocks that are
28613         otherwise unused, but do not preserve them.
28615 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
28617         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
28619 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
28621         PR target/84039
28622         * config/i386/constraints.md (Bs): Replace
28623         ix86_indirect_branch_register with
28624         TARGET_INDIRECT_BRANCH_REGISTER.
28625         (Bw): Likewise.
28626         * config/i386/i386.md (indirect_jump): Likewise.
28627         (tablejump): Likewise.
28628         (*sibcall_memory): Likewise.
28629         (*sibcall_value_memory): Likewise.
28630         Peepholes of indirect call and jump via memory: Likewise.
28631         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
28632         (*sibcall_value_GOT_32): Likewise.
28633         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28634         (GOT_memory_operand): Likewise.
28635         (call_insn_operand): Likewise.
28636         (sibcall_insn_operand): Likewise.
28637         (GOT32_symbol_operand): Likewise.
28638         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
28640 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
28642         PR rtl-optimization/83496
28643         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
28644         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
28645         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
28646         redundant insn, if any.
28647         (relax_delay_slots): Likewise.
28648         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
28650 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
28652         PR tree-optimization/83965
28653         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
28654         that grouped statements are part of a reduction chain.  Return
28655         true if the statement is not marked as a reduction itself but
28656         is part of a group.
28657         (vect_recog_dot_prod_pattern): Don't check whether the statement
28658         is part of a group here.
28659         (vect_recog_sad_pattern): Likewise.
28660         (vect_recog_widen_sum_pattern): Likewise.
28662 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
28664         PR debug/84545
28665         * final.c (rest_of_clean_state): Also look for calls inside sequences.
28667 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
28669         PR target/84530
28670         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
28671         the bool argument.
28672         (ix86_output_indirect_function_return): New prototype.
28673         (ix86_split_simple_return_pop_internal): Likewise.
28674         * config/i386/i386.c (indirect_return_via_cx): New.
28675         (indirect_return_via_cx_bnd): Likewise.
28676         (indirect_thunk_name): Handle return va CX_REG.
28677         (output_indirect_thunk_function): Create alias for
28678         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
28679         (ix86_output_indirect_jmp): Remove the bool argument.
28680         (ix86_output_indirect_function_return): New function.
28681         (ix86_split_simple_return_pop_internal): Likewise.
28682         * config/i386/i386.md (*indirect_jump): Don't pass false
28683         to ix86_output_indirect_jmp.
28684         (*tablejump_1): Likewise.
28685         (simple_return_pop_internal): Change it to define_insn_and_split.
28686         Call ix86_split_simple_return_pop_internal to split it for
28687         -mfunction-return=.
28688         (simple_return_indirect_internal): Call
28689         ix86_output_indirect_function_return instead of
28690         ix86_output_indirect_jmp.
28692 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
28694         PR bootstrap/84405
28695         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
28696         memset and value initialization afterwards.
28698 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
28700         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
28702 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28704         PR target/84521
28705         * common/config/aarch64/aarch64-common.c
28706         (aarch_option_optimization_table[]): Switch
28707         off fomit-frame-pointer
28709 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
28710             Chung-Ju Wu  <jasonwucj@gmail.com>
28712         * config/nds32/nds32-multiple.md (load_multiple): Disallow
28713         volatile memory.
28714         (store_multiple): Ditto.
28716 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
28718         * config.gcc: Add --with-cpu support for nds32 target.
28719         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
28720         * config/nds32/nds32.opt: Add -mcpu= option.
28722 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
28724         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
28725         isel=yes): Warn for these deprecated options.
28727 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
28729         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
28730         ISA_2_5_MASKS_EMBEDDED.
28732 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
28734         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
28735         p->max as pointers rather than using iterative_hash_expr.
28737 2018-02-23  Carl Love  <cel@us.ibm.com>
28739         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
28740         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
28741         BU_P8V_OVERLOAD_2.
28742         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
28743         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
28744         P8V_BUILTIN_VEC_VUNSIGNED2.
28746 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
28748         PR target/81572
28749         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
28750         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
28751         LRA_UNKNOWN_ALT.
28752         * lra-constraints.c (curr_insn_transform): Set up
28753         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
28754         LRA_UNKNOWN_ALT.
28755         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
28756         * lra-eliminations.c (spill_pseudos): Ditto.
28757         (process_insn_for_elimination): Ditto.
28758         * lra-lives.c (reg_early_clobber_p): Use the new macros.
28759         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
28760         LRA_NON_CLOBBERED_ALT.
28762 2018-02-22  Martin Sebor  <msebor@redhat.com>
28764         PR tree-optimization/84480
28765         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
28766         to maybe_diag_stxncpy_trunc.  Call it.
28767         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
28768         from gimple_fold_builtin_strcpy.  Print inlining stack.
28769         (handle_builtin_stxncpy): Print inlining stack.
28770         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
28772 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
28774         PR target/84176
28775         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
28776         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
28777         and -fcheck-pointer-bounds are used together.
28778         (indirect_thunk_prefix): New enum.
28779         (indirect_thunk_need_prefix): New function.
28780         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
28781         "_nt" instead of "_bnd" for NOTRACK prefix.
28782         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
28783         (output_indirect_thunk_function): Likewise.
28784         (): Likewise.
28785         (ix86_code_end): Update output_indirect_thunk_function calls.
28786         (ix86_output_indirect_branch_via_reg): Replace
28787         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
28788         (ix86_output_indirect_branch_via_push): Likewise.
28789         (ix86_output_function_return): Likewise.
28790         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
28791         incompatible with -fcf-protection=branch and
28792         -fcheck-pointer-bounds.
28794 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
28796         PR target/83335
28797         * config/aarch64/aarch64.c (aarch64_print_address_internal):
28798         Change gcc_assert call to output_operand_lossage.
28800 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
28802         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
28804 2018-02-22  DJ Delorie  <dj@redhat.com>
28805             Sebastian Perta  <sebastian.perta@renesas.com>
28806             Oleg Endo  <olegendo@gcc.gnu.org>
28808         * config/rx/rx.c (rx_rtx_costs): New function.
28809         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
28811 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
28813         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
28815 2018-02-22  Martin Liska  <mliska@suse.cz>
28817         PR driver/83193
28818         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
28819         Add "native" as a possible value.
28821 2018-02-22  Martin Liska  <mliska@suse.cz>
28823         PR driver/83193
28824         * config/i386/i386.c (ix86_option_override_internal):
28825         Add "native" as a possible value for -march and -mtune.
28827 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
28829         PR target/84502
28830         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
28831         to all type variants.
28833         PR tree-optimization/84503
28834         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
28835         width as info->bitpos + info->bitsize - start.
28836         (merged_store_group::merge_overlapping): Simplify width computation.
28837         (check_no_overlap): New function.
28838         (imm_store_chain_info::try_coalesce_bswap): Compute expected
28839         start + width and last_order of the group, fail if check_no_overlap
28840         fails.
28841         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
28842         to group if check_no_overlap fails.
28844 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
28846         * config/rs6000/altivec.md: Delete contraint arguments to
28847         define_expand, define_split, and define_peephole2, and in
28848         define_insn_and_split if always unused.
28849         * config/rs6000/darwin.md: Ditto.
28850         * config/rs6000/dfp.md: Ditto.
28851         * config/rs6000/rs6000.md: Ditto.
28852         * config/rs6000/sync.md: Ditto.
28853         * config/rs6000/vector.md: Ditto.
28854         * config/rs6000/vsx.md: Ditto.
28856 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
28858         * config/rs6000/altivec.md: Write output control strings as braced
28859         blocks instead of double-quoted strings.
28860         * config/rs6000/darwin.md: Ditto.
28861         * config/rs6000/rs6000.md: Ditto.
28862         * config/rs6000/vector.md: Ditto.
28863         * config/rs6000/vsx.md: Ditto.
28865 2018-02-21  Jason Merrill  <jason@redhat.com>
28867         PR c++/84314 - ICE with templates and fastcall attribute.
28868         * attribs.c (build_type_attribute_qual_variant): Remove assert.
28870 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
28872         * ipa-cp.c (determine_versionability): Fix comment typos.
28874 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
28876         PR c/84229
28877         * ipa-cp.c (determine_versionability): Do not version functions caling
28878         va_arg_pack.
28880 2018-02-21  Martin Liska  <mliska@suse.cz>
28882         PR driver/83193
28883         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
28884         Add "native" as a possible value.
28885         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
28886         the macro when native cpu detection is available.
28888 2018-02-21  Martin Liska  <mliska@suse.cz>
28890         PR driver/83193
28891         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
28892         Add "native" as a possible value.
28893         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
28894         when native cpu detection is available.
28896 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
28897             Martin Sebor  <msebor@redhat.com>
28899         PR tree-optimization/84478
28900         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
28901         false.
28902         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
28903         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
28904         support which is conservatively correct, for 2 only stay conservative
28905         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
28906         argument to the 2 argument get_range_strlen, adjust 6 arg
28907         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
28908         false.
28909         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
28910         (gimple_fold_builtin_strlen): Pass true as last argument to
28911         get_range_strlen.
28913 2018-02-20  Martin Sebor  <msebor@redhat.com>
28915         PR middle-end/84095
28916         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
28917         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
28918         (builtin_memref::builtin_memref): Factor out parts into
28919         set_base_and_offset and call it.
28921 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
28923         PR middle-end/84406
28924         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
28925         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
28926         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
28927         search at the associated MODE_INT.
28929 2018-02-20  Jeff Law  <law@redhat.com>
28931         PR middle-end/82123
28932         PR tree-optimization/81592
28933         PR middle-end/79257
28934         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
28935         for range data rather than using global data.
28936         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
28937         range data rather than using global data.
28938         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
28939         pass it to children as needed.
28940         (struct directive::fmtresult): Similarly.
28941         (struct directive::set_width): Similarly.
28942         (struct directive::set_precision): Similarly.
28943         (format_integer, format_directive, parse_directive): Similarly.
28944         (format_none): Accept unnamed vr_values parameter.
28945         (format_percent, format_floating, format_character): Similarly.
28946         (format_string, format_plain): Similarly.
28947         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
28948         the EVRP range analyzer for range data rather than using global data.
28949         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
28950         gimple-ssa-evrp-analyze.h
28951         (class sprintf_dom_walker): Add after_dom_children member function.
28952         Add evrp_range_analyzer member.
28953         (sprintf_dom_walker::before_dom_children): Call into the EVRP
28954         range analyzer as needed.
28955         (sprintf_dom_walker::after_dom_children): New member function.
28956         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
28957         if not optimizing.
28958         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
28959         (evrp_range_analyzer::pop_to_marker): Likewise.
28961 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
28963         PR tree-optimization/84419
28964         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
28965         with the required type if its current type is compatible but
28966         different.
28968 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
28970         PR middle-end/82004
28971         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
28972         after vectorization.
28974 2018-02-20  Martin Liska  <mliska@suse.cz>
28976         PR driver/83193
28977         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
28978         possible values if we don't have a hint.
28980 2018-02-20  Martin Liska  <mliska@suse.cz>
28982         PR c/84310
28983         PR target/79747
28984         * final.c (shorten_branches): Build align_tab array with one
28985         more element.
28986         * opts.c (finish_options): Add alignment option limit check.
28987         (MAX_CODE_ALIGN): Likewise.
28988         (MAX_CODE_ALIGN_VALUE): Likewise.
28989         * doc/invoke.texi: Document maximum allowed option value for
28990         all -falign-* options.
28992 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
28994         PR target/84146
28995         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
28996         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
28997         * var-tracking.c (emit_note_insn_var_location): Remove all references
28998         to NOTE_INSN_CALL_ARG_LOCATION.
28999         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
29000         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
29001         Use copy_rtx_if_shared.
29002         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
29003         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
29004         (dwarf2out_var_location): Remove handling of
29005         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
29006         on call_insn.
29007         * final.c (final_scan_insn): Remove all references to
29008         NOTE_INSN_CALL_ARG_LOCATION.
29009         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
29010         before dumping final insns.
29011         * except.c (emit_note_eh_region_end): Remove all references to
29012         NOTE_INSN_CALL_ARG_LOCATION.
29013         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
29014         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
29015         * config/arc/arc.c (hwloop_optimize): Likewise.
29016         * config/arm/arm.c (create_fix_barrier): Likewise.
29017         * config/s390/s390.c (s390_chunkify_start): Likewise.
29018         * config/sh/sh.c (find_barrier): Likewise.
29019         * config/i386/i386.c (rest_of_insert_endbranch,
29020         ix86_seh_fixup_eh_fallthru): Likewise.
29021         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
29022         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
29023         * config/frv/frv.c (frv_function_prologue): Likewise.
29024         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
29025         reg note.
29026         (note_outside_basic_block_p): Remove all references to
29027         NOTE_INSN_CALL_ARG_LOCATION.
29028         * gengtype.c (adjust_field_rtx_def): Likewise.
29029         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
29030         Likewise.
29031         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
29032         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
29034         PR c++/84444
29035         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
29036         is ADDR_EXPR.
29038         PR tree-optimization/84452
29039         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
29040         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
29041         is NULL.
29043 2018-02-19  Martin Liska  <mliska@suse.cz>
29045         PR sanitizer/82183
29046         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
29048 2018-02-19  Martin Liska  <mliska@suse.cz>
29049             Richard Sandiford  <richard.sandiford@linaro.org>
29051         PR tree-optimization/82491
29052         * gimple-fold.c (get_base_constructor): Make earlier bail out
29053         to prevent ubsan.
29055 2018-02-19  Carl Love  <cel@us.ibm.com>
29057         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
29058         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
29059         BU_P8V_OVERLOAD_1.
29060         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
29061         P8V_BUILTIN_VEC_NEG.
29063 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
29065         * config/rl78/rl78.md (movdf): New define expand.
29067 2018-02-19  Martin Liska  <mliska@suse.cz>
29069         PR other/80589
29070         * doc/invoke.texi: Fix typo.
29071         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
29073 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
29075         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
29076         handle rs6000_single_float and rs6000_double_float specially for
29077         e500 family CPUs.
29079 2018-02-16  Jeff Law  <law@redhat.com>
29081         * config/rx/rx.c (add_pop_cfi_notes): New function.;
29082         (pop_regs): Use it.
29084 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
29086         PR ipa/84425
29087         * ipa-inline.c (inline_small_functions): Fix a typo.
29089 2018-02-16  Nathan Sidwell  <nathan@acm.org>
29091         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
29093 2018-02-16  Carl Love  <cel@us.ibm.com>
29095         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
29096         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
29097         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
29098         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
29099         expansion to P8V_BUILTIN_VEC_FLOAT2.
29101 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
29103         PR rtl-optimization/70023
29104         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
29105         src_regno into account.
29107 2018-02-16  Carl Love  <cel@us.ibm.com>
29109         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
29110         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
29111         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
29112         * config/rs6000/rs6000.c: Remove case statements for
29113         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
29114         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
29115         and P9V_BUILTIN_VEC_VINSERT4B.
29116         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
29117         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
29118         * config/rs6000/vsx.md:
29119         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
29120         vec_insert4b.
29122 2018-02-16  Carl Love  <cel@us.ibm.com>
29124         * config/rs6000/altivec.h: Add builtin names vec_extract4b
29125         vec_insert4b.
29126         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
29127         definitions.
29128         * config/rs6000/rs6000-c.c: Add the definitions for
29129         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
29130         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
29131         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
29132         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
29133         definition for insert4b and define insn *insert3b_internal.
29134         * doc/extend.texi: Add documentation for vec_extract4b.
29136 2018-02-16  Nathan Sidwell  <nathan@acm.org>
29138         * doc/extend.texi (Backwards Compatibility): Mention friend
29139         injection.  Note for-scope is deprecated.
29140         * doc/invoke.texi (-ffriend-injection): Deprecate.
29142 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
29144         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
29145         that moved to I2, also allow destinations that are a paradoxical
29146         subreg (instead of a normal reg).
29148 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
29150         PR target/83831
29151         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
29152         to QImode.
29154 2018-02-16  Richard Biener  <rguenther@suse.de>
29156         PR tree-optimization/84037
29157         PR tree-optimization/84016
29158         PR target/82862
29159         * config/i386/i386.c (ix86_builtin_vectorization_cost):
29160         Adjust vec_construct for the fact we need additional higher latency
29161         128bit inserts for AVX256 and AVX512 vector builds.
29162         (ix86_add_stmt_cost): Scale vector construction cost for
29163         elementwise loads.
29165 2018-02-16  Richard Biener  <rguenther@suse.de>
29167         PR tree-optimization/84417
29168         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
29169         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
29170         (non_rewritable_lvalue_p): Likewise, use poly-ints.
29172 2018-02-16  Martin Liska  <mliska@suse.cz>
29174         PR sanitizer/84307
29175         * internal-fn.def (ASAN_CHECK): Set proper flags.
29176         (ASAN_MARK): Likewise.
29178 2018-02-16  Julia Koval  <julia.koval@intel.com>
29180         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
29181         from PTA_CANNONLAKE.
29183 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
29185         PR target/84272
29186         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
29187         Use ++iter rather than iter++ for std::list iterators.
29188         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
29189         defer deleting them until all nodes in the forest are processed.  Do
29190         free even leaf nodes.  Change to_process into auto_vec.
29192         PR bootstrap/84405
29193         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
29194         * vec.h (vec_default_construct): Use memset instead of placement new
29195         if BROKEN_VALUE_INITIALIZATION is defined.
29196         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
29197         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
29198         is defined.
29200         PR rtl-optimization/83723
29201         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
29202         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
29203         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
29204         recursive calls.
29205         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
29206         callers.
29207         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
29209 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
29211         PR rtl-optimization/81443
29212         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
29213         from inner REGs to paradoxical SUBREGs.
29215 2018-02-16  Richard Biener  <rguenther@suse.de>
29217         PR tree-optimization/84399
29218         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
29219         For operands we can analyze at their definition make sure we can
29220         analyze them at each use as well.
29222 2018-02-16  Richard Biener  <rguenther@suse.de>
29224         PR tree-optimization/84190
29225         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
29226         volatile accesses if the decl isn't volatile.
29228 2018-02-15  Jason Merrill  <jason@redhat.com>
29230         PR c++/84314 - ICE with templates and fastcall attribute.
29231         * attribs.c (build_type_attribute_qual_variant): Don't clobber
29232         TYPE_CANONICAL on an existing type.
29234 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
29236         PR tree-optimization/84383
29237         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
29238         dstoff nor call operand_equal_p if dstbase is NULL.
29240         PR tree-optimization/84334
29241         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
29242         also a CONSTANT_CLASS_P, punt.
29244 2018-02-14  Jim Wilson  <jimw@sifive.com>
29246         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
29247         first SMALL_OPERAND check.  New local min_second_step.  Move assert
29248         to where locals are set.  Add TARGET_RVC support.
29249         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
29251 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
29253         * doc/invoke.texi: Correct -Wformat-overflow code sample.
29255 2018-02-14  Martin Sebor  <msebor@redhat.com>
29257         PR tree-optimization/83698
29258         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
29259         arrays constrain the offset range to their bounds.
29260         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
29261         (builtin_access::overlap): Avoid setting the size of overlap if it's
29262         already been set.
29263         (maybe_diag_overlap): Also consider arrays when deciding what values
29264         of offsets to include in diagnostics.
29266 2018-02-14  Martin Sebor  <msebor@redhat.com>
29268         PR c/84108
29269         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
29270         that correspond to the kind of a declaration.
29272 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
29274         PR target/83984
29275         * config/pa/pa.md: Load address of PIC label using the linkage table
29276         if the label is nonlocal.
29278 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
29280         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
29281         warning message if user requests -maltivec=be.
29282         * doc/invoke.texi: Document deprecation of -maltivec=be.
29284 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
29286         PR target/84220
29287         * config/rs6000/rs6000-c.c: Update definitions for
29288         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
29289         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
29291 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
29293         PR target/84239
29294         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
29295         add _get_ssp intrinsics. Remove argument from
29296         __builtin_ia32_rdssp[d|q].
29297         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
29298         * config/i386/i386-builtin.def: Remove argument from
29299         __builtin_ia32_rdssp[d|q].
29300         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
29301         ix86_expand_special_args_builtin for _rdssp[d|q].
29302         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
29303         Clear register before usage.
29304         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
29305         Add documentation for new _get_ssp and _inc_ssp intrinsics.
29307 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
29309         PR tree-optimization/84357
29310         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
29311         operand 1 of an ARRAY_REF too.
29313 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
29315         PR target/83831
29316         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
29317         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
29318         declarations.
29319         (set_of_reg): New struct.
29320         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
29321         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
29322         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
29323         functions.
29324         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
29325         Split into bitclr, bitset, bitinvert patterns if appropriate.
29326         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
29327         use rx_fuse_in_memory_bitop.
29328         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
29329         to named insn, correct maximum insn length.
29331 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
29333         PR target/79242
29334         * machmode.def: Define a complex mode for PARTIAL_INT.
29335         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
29336         MODE_PARTIAL_INT.
29337         * doc/rtl.texi: Document CSPImode.
29338         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
29339         handling.
29340         (msp430_hard_regno_nregs_with_padding): Likewise.
29342 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
29344         PR target/84279
29345         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
29347 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
29349         PR rtl-optimization/84169
29350         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
29351         we generated a parallel as new i3 and we split that to new i2 and i3
29352         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
29353         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
29354         those to i2, not i1.  Partially rewrite this scan code.
29356 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
29358         PR c/82210
29359         * stor-layout.c (place_field): For variable length fields, adjust
29360         offset_align afterwards not just based on the field's alignment,
29361         but also on the size.
29363         PR middle-end/84309
29364         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
29365         of exps and logs in the use_exp2 case.
29367 2018-02-13  Jeff Law  <law@redhat.com>
29369         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
29370         entry for "vector".
29372         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
29373         ARGS as unused.
29375 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
29377         PR debug/84342
29378         PR debug/84319
29379         * common.opt (gas-loc-support, gas-locview-support): New.
29380         (ginline-points, ginternal-reset-location-views): New.
29381         * doc/invoke.texi: Document them.  Use @itemx where intended.
29382         (gvariable-location-views): Adjust.
29383         * target.def (reset_location_view): New.
29384         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
29385         (TARGET_RESET_LOCATION_VIEW): New.
29386         * doc/tm.texi: Rebuilt.
29387         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
29388         (dwarf2out_default_as_locview_support): New.
29389         (output_asm_line_debug_info): Use option variables.
29390         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
29391         (output_loc_list): Likewise.
29392         (add_high_low_attributes): Check option variables.
29393         Don't output entry view attribute in strict mode.
29394         (gen_inlined_subroutine_die): Check option variables.
29395         (dwarf2out_inline_entry): Likewise.
29396         (init_sections_and_labels): Likewise.
29397         (dwarf2out_early_finish): Likewise.
29398         (maybe_reset_location_view): New, from...
29399         (dwarf2out_var_location): ... here.  Call it.
29400         * debug.h (dwarf2out_default_as_loc_support): Declare.
29401         (dwarf2out_default_as_locview_support): Declare.
29402         * hooks.c (hook_int_rtx_insn_0): New.
29403         * hooks.h (hook_int_rtx_insn_0): Declare.
29404         * toplev.c (process_options): Take -gas-loc-support and
29405         -gas-locview-support from dwarf2out.  Enable
29406         -gvariable-location-views by default only with locview
29407         assembler support.  Enable -ginternal-reset-location-views by
29408         default only if the target defines the corresponding hook.
29409         Enable -ginline-points by default if location views are
29410         enabled; force it disabled if statement frontiers are
29411         disabled.
29412         * tree-inline.c (expand_call_inline): Check option variables.
29413         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
29415 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
29417         PR tree-optimization/84321
29418         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
29419         handling.  Also check whether the anti-range contains any values
29420         that satisfy the mask; switch to a VR_RANGE if not.
29422 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
29424         PR sanitizer/84340
29425         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
29427 2018-02-13  Martin Jambor  <mjambor@suse.cz>
29429         PR c++/83990
29430         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
29431         of call statements, also set location of a load to a temporary.
29433 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
29435         * config/rl78/rl78.c (add_vector_labels): New function.
29436         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
29437         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
29438         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
29439         which checks that no arguments are passed.
29440         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
29441         * doc/extend.texi: Documentation for the new attribute.
29443 2018-02-13  Andreas Schwab  <schwab@suse.de>
29445         * config/riscv/linux.h (CPP_SPEC): Define.
29447 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
29449         PR target/84335
29450         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
29451         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
29452         OPTION_MASK_ISA_AES as first argument to def_builtin_const
29453         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
29454         instead of OPTION_MASK_ISA_PCLMUL as first argument to
29455         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
29456         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
29457         temporarily for AES and PCLMUL builtins.
29459         PR tree-optimization/84339
29460         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
29461         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
29462         Formatting fixes.
29464         PR middle-end/84309
29465         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
29466         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
29467         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
29468         inline function.
29469         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
29470         inline function.
29471         * omp-simd-clone.h: New file.
29472         * omp-simd-clone.c: Include omp-simd-clone.h.
29473         (expand_simd_clones): No longer static.
29474         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
29475         cgraph.h and omp-simd-clone.h.
29476         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
29477         (vect_recog_widen_shift_pattern): Formatting fix.
29478         (vect_pattern_recog_1): Don't check optab for calls.
29480         PR target/84336
29481         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
29482         operands[2] into a REG before using gen_lowpart on it.
29484 2018-02-12  Jeff Law  <law@redhat.com>
29486         PR target/83760
29487         * config/sh/sh.c (find_barrier): Consider a sibling call
29488         a barrier as well.
29490         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
29491         successfully back substituting a reg.
29493 2018-02-12  Richard Biener  <rguenther@suse.de>
29495         PR tree-optimization/84037
29496         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
29497         parameter, move visited init to caller.
29498         (vect_slp_analyze_operations): Separate cost from validity
29499         check, initialize visited once for all instances.
29500         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
29501         for all instances.
29502         * tree-vect-stmts.c (vect_model_simple_cost): Make early
29503         out an assert.
29504         (vect_model_promotion_demotion_cost): Likewise.
29505         (vectorizable_bswap): Guard cost modeling with !slp_node
29506         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
29507         SLP stmts.
29508         (vectorizable_call): Likewise.
29509         (vectorizable_conversion): Likewise.
29510         (vectorizable_assignment): Likewise.
29511         (vectorizable_shift): Likewise.
29512         (vectorizable_operation): Likewise.
29513         (vectorizable_store): Likewise.
29514         (vectorizable_load): Likewise.
29515         (vectorizable_condition): Likewise.
29516         (vectorizable_comparison): Likewise.
29518 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
29520         PR sanitizer/84307
29521         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
29522         (ASAN_MARK): Fix fnspec to account for return value, change pointer
29523         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
29525 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
29527         PR middle-end/83665
29528         * params.def (inline-min-speedup): Increase from 8 to 15.
29529         (max-inline-insns-auto): Decrease from 40 to 30.
29530         * ipa-split.c (consider_split): Add some buffer for function to
29531         be considered inlining candidate.
29532         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
29533         default values.
29535 2018-02-12  Richard Biener  <rguenther@suse.de>
29537         PR tree-optimization/84037
29538         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
29539         matched stmts if we cannot swap the non-matched ones.
29541 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
29543         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
29544         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
29545         _mm_maskz_scalef_round_ss): New intrinsics.
29546         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
29547         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
29548         __builtin_ia32_scalefss_round): Remove.
29549         (__builtin_ia32_scalefsd_mask_round,
29550         __builtin_ia32_scalefss_mask_round): New intrinsics.
29551         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
29552         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
29553         ((match_operand:VF_128 2 "<round_nimm_predicate>"
29554         "<round_constraint>")): Changed to ...
29555         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
29556         "<round_scalar_constraint>")): ... this.
29557         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
29558         %0, %1, %2<round_op3>}"): Changed to ...
29559         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
29560         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
29561         %2<round_scalar_mask_op3>}"): ... this.
29562         * config/i386/subst.md (round_scalar_nimm_predicate): New.
29564 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
29566         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
29567         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
29568         (_mm_maskz_sqrt_round_ss): New intrinsics.
29569         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
29570         (__builtin_ia32_sqrtsd_mask_round)
29571         (__builtin_ia32_sqrtss_mask_round): New builtins.
29572         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
29573         (__builtin_ia32_sqrtss_round): Remove.
29574         (__builtin_ia32_sqrtsd_mask_round)
29575         (__builtin_ia32_sqrtss_mask_round): New builtins.
29576         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
29577         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
29578         ((match_operand:VF_128 1 "vector_operand"
29579         "xBm,<round_constraint>")): Changed to ...
29580         ((match_operand:VF_128 1 "vector_operand"
29581         "xBm,<round_scalar_constraint>")): ... this.
29582         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
29583         %0, %2, %<iptr>1<round_op3>}): Changed to ...
29584         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
29585         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
29586         %<iptr>1<round_scalar_mask_op3>}): ... this.
29587         ((set_attr "prefix" "<round_prefix>")): Changed to ...
29588         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
29590 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
29592         PR target/84266
29593         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
29594         Cast vec_cmpeq result to correct type.
29595         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
29596         Cast vec_cmpgt result to correct type.
29598 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
29600         * final.c (final_scan_insn_1): Renamed from...
29601         (final_scan_insn): ... this.  New wrapper, to recover
29602         seen from the outermost call in recursive ones.
29603         * config/sparc/sparc.c (output_return): Drop seen from call.
29604         (output_sibcall): Likewise.
29605         * config/visium/visium.c (output_branch): Likewise.
29607 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
29609         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
29610         function label.
29612 2018-02-10  Alan Modra  <amodra@gmail.com>
29614         PR target/84300
29615         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
29616         Specify LR as an input.
29618 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
29620         PR sanitizer/83987
29621         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
29622         remove_member_access_dummy_vars): New functions.
29623         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
29624         lower_omp_1, execute_lower_omp): Use them.
29626         PR rtl-optimization/84308
29627         * shrink-wrap.c (spread_components): Release todo vector.
29629 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
29631         PR rtl-optimization/57193
29632         * ira-color.c (struct allocno_color_data): Add member
29633         conflict_allocno_hard_prefs.
29634         (update_conflict_allocno_hard_prefs): New.
29635         (bucket_allocno_compare_func): Add a preference based on
29636         conflict_allocno_hard_prefs.
29637         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
29638         (color_allocnos): Remove a dead code.  Initiate
29639         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
29641 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
29643         PR target/84226
29644         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
29645         constraint from =wa to wa.  Avoid a subreg on the output operand,
29646         instead use a pseudo and subreg it in a move.
29647         (p9_xxbrd_<mode>): Changed to ...
29648         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
29649         (p9_xxbrd_v2df): New expander.
29650         (p9_xxbrw_<mode>): Changed to ...
29651         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
29652         (p9_xxbrw_v4sf): New expander.
29654 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
29656         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
29658 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
29660         PR target/83926
29661         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
29662         multiply in 32-bit mode.
29663         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
29664         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
29665         mode.
29667 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
29669         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
29670         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
29671         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
29672         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
29674 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
29676         PR lto/84213
29677         * dwarf2out.c (is_trivial_indirect_ref): New function.
29678         (dwarf2out_late_global_decl): Do not generate a location
29679         attribute for variables that have a non-trivial DECL_VALUE_EXPR
29680         and that are not defined in the current unit.
29682 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
29684         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
29685         instead of a libcall for UNORDERED.
29687 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
29689         PR target/82641
29690         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
29691         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
29693 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29695         PR target/PR84295
29696         * config/s390/s390.c (s390_set_current_function): Invoke
29697         s390_indirect_branch_settings also if fndecl didn't change.
29699 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
29701         * config/rs6000/rs6000.md (blockage): Set length to zero.
29703 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
29705         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
29707 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
29709         PR sanitizer/84285
29710         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
29711         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
29712         -static-lib*san.
29714         PR debug/84252
29715         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
29716         PARALLEL incoming that failed vt_get_decl_and_offset check.
29718         PR middle-end/84237
29719         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
29720         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
29721         TREE_READONLY bit.
29722         (get_variable_section): For decls in named .bss* sections pass true as
29723         second argument to bss_initializer_p.
29725 2018-02-09  Marek Polacek  <polacek@redhat.com>
29726             Jakub Jelinek  <jakub@redhat.com>
29728         PR c++/83659
29729         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
29730         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
29731         Sync some changes from cxx_fold_indirect_ref.
29733 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
29735         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
29736         markers.
29737         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
29738         (BLOCK_INLINE_ENTRY_LABEL): New.
29739         (dwarf2out_var_location): Disregard inline entry markers.
29740         (inline_entry_data): New struct.
29741         (inline_entry_data_hasher): New hashtable type.
29742         (inline_entry_data_hasher::hash): New.
29743         (inline_entry_data_hasher::equal): New.
29744         (inline_entry_data_table): New variable.
29745         (add_high_low_attributes): Add DW_AT_entry_pc and
29746         DW_AT_GNU_entry_view attributes if a pending entry is found
29747         in inline_entry_data_table.  Add old entry_pc attribute only
29748         if debug nonbinding markers are disabled.
29749         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
29750         markers are enabled.
29751         (block_within_block_p, dwarf2out_inline_entry): New.
29752         (dwarf2out_finish): Check that no entries remained in
29753         inline_entry_data_table.
29754         * final.c (reemit_insn_block_notes): Handle inline entry notes.
29755         (final_scan_insn, notice_source_line): Likewise.
29756         (rest_of_clean_state): Skip inline entry markers.
29757         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
29758         markers.
29759         * gimple.c (gimple_build_debug_inline_entry): New.
29760         * gimple.h (enum gimple_debug_subcode): Add
29761         GIMPLE_DEBUG_INLINE_ENTRY.
29762         (gimple_build_debug_inline_entry): Declare.
29763         (gimple_debug_inline_entry_p): New.
29764         (gimple_debug_nonbind_marker_p): Adjust.
29765         * insn-notes.def (INLINE_ENTRY): New.
29766         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
29767         inline entry marker notes.
29768         (print_insn): Likewise.
29769         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
29770         (INSN_DEBUG_MARKER_KIND): Likewise.
29771         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
29772         * tree-inline.c (expand_call_inline): Build and insert
29773         debug_inline_entry stmt.
29774         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
29775         inline entry blocks early, if nonbind markers are enabled.
29776         (dump_scope_block): Dump fragment info.
29777         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
29778         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
29779         (gimple_build_debug_inline_entry): New.
29780         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
29781         Enable/disable inline entry points too.
29782         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
29783         (DEBUG_INSN): Describe inline entry markers.
29785         * common.opt (gvariable-location-views): New.
29786         (gvariable-location-views=incompat5): New.
29787         * config.in: Rebuilt.
29788         * configure: Rebuilt.
29789         * configure.ac: Test assembler for view support.
29790         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
29791         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
29792         * dwarf2out.c (var_loc_view): New typedef.
29793         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
29794         (dwarf2out_locviews_in_attribute): New.
29795         (dwarf2out_locviews_in_loclist): New.
29796         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
29797         (enum dw_line_info_opcode): Add LI_adv_address.
29798         (struct dw_line_info_table): Add view.
29799         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
29800         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
29801         (zero_view_p): New variable.
29802         (ZERO_VIEW_P): New macro.
29803         (output_asm_line_debug_info): New.
29804         (struct var_loc_node): Add view.
29805         (add_AT_view_list, AT_loc_list): New.
29806         (add_var_loc_to_decl): Add view param.  Test it against last.
29807         (new_loc_list): Add view params.  Record them.
29808         (AT_loc_list_ptr): Handle loc and view lists.
29809         (view_list_to_loc_list_val_node): New.
29810         (print_dw_val): Handle dw_val_class_view_list.
29811         (size_of_die): Likewise.
29812         (value_format): Likewise.
29813         (loc_list_has_views): New.
29814         (gen_llsym): Set vl_symbol too.
29815         (maybe_gen_llsym, skip_loc_list_entry): New.
29816         (dwarf2out_maybe_output_loclist_view_pair): New.
29817         (output_loc_list): Output view list or entries too.
29818         (output_view_list_offset): New.
29819         (output_die): Handle dw_val_class_view_list.
29820         (output_dwarf_version): New.
29821         (output_compilation_unit_header): Use it.
29822         (output_skeleton_debug_sections): Likewise.
29823         (output_rnglists, output_line_info): Likewise.
29824         (output_pubnames, output_aranges): Update version comments.
29825         (output_one_line_info_table): Output view numbers in asm comments.
29826         (dw_loc_list): Determine current endview, pass it to new_loc_list.
29827         Call maybe_gen_llsym.
29828         (loc_list_from_tree_1): Adjust.
29829         (add_AT_location_description): Create view list attribute if
29830         needed, check it's absent otherwise.
29831         (convert_cfa_to_fb_loc_list): Adjust.
29832         (maybe_emit_file): Call output_asm_line_debug_info for test.
29833         (dwarf2out_var_location): Reset views as needed.  Precompute
29834         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
29835         attribute.  Set view.
29836         (new_line_info_table): Reset next view.
29837         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
29838         (dwarf2out_source_line): Likewise.  Output view resets and labels to
29839         the assembler, or select appropriate line info opcodes.
29840         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
29841         (optimize_string_length): Catch it.  Adjust.
29842         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
29843         dw_val_class_view_list, and remove it if no longer needed.
29844         (hash_loc_list): Hash view numbers.
29845         (loc_list_hasher::equal): Compare them.
29846         (optimize_location_lists): Check whether a view list symbol is
29847         needed, and whether the locview attribute is present, and
29848         whether they match.  Remove the locview attribute if no longer
29849         needed.
29850         (index_location_lists): Call skip_loc_list_entry for test.
29851         (dwarf2out_finish): Call output_asm_line_debug_info for test.
29852         Use output_dwarf_version.
29853         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
29854         (struct dw_val_node): Add val_view_list.
29855         * final.c (SEEN_NEXT_VIEW): New.
29856         (set_next_view_needed): New.
29857         (clear_next_view_needed): New.
29858         (maybe_output_next_view): New.
29859         (final_start_function): Rename to...
29860         (final_start_function_1): ... this.  Take pointer to FIRST,
29861         add SEEN parameter.  Emit param bindings in the initial view.
29862         (final_start_function): Reintroduce SEEN-less interface.
29863         (final): Rename to...
29864         (final_1): ... this.  Take SEEN parameter.  Output final pending
29865         next view at the end.
29866         (final): Reintroduce seen-less interface.
29867         (final_scan_insn): Output pending next view before switching
29868         sections or ending a block.  Mark the next view as needed when
29869         outputting variable locations.  Notify debug backend of section
29870         changes, and of location view changes.
29871         (rest_of_handle_final): Adjust.
29872         * toplev.c (process_options): Autodetect value for debug variable
29873         location views option.  Warn on incompat5 without -gdwarf-5.
29874         * doc/invoke.texi (gvariable-location-views): New.
29875         (gvariable-location-views=incompat5): New.
29876         (gno-variable-location-views): New.
29878 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
29880         PR tree-optimization/84136
29881         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
29882         that the result of find_edge is non-NULL.
29884 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
29886         PR target/83008
29887         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
29888         storing integer register in SImode.  Fix cost of 256 and 512
29889         byte aligned SSE register store.
29891 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
29893         * config/i386/i386.c (ix86_multiplication_cost): Fix
29894         multiplication cost for TARGET_AVX512DQ.
29896 2018-02-08  Marek Polacek  <polacek@redhat.com>
29898         PR tree-optimization/84238
29899         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
29900         get_range_strlen.
29902 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
29904         PR tree-optimization/84265
29905         * tree-vect-stmts.c (vectorizable_store): Don't treat
29906         VMAT_CONTIGUOUS accesses as grouped.
29907         (vectorizable_load): Likewise.
29909 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
29911         PR tree-optimization/81635
29912         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
29913         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
29914         (test_round_for_mask): New functions.
29915         (wide_int_cc_tests): Call test_round_for_mask.
29916         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
29917         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
29918         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
29919         range returned by get_range_info.
29921 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
29923         PR ipa/81360
29924         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
29925         * symtab.c: Include builtins.h
29926         (symtab_node::output_to_lto_symbol_table_p): Move here
29927         from lto-streamer-out.c:output_symbol_p.
29928         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
29929         (output_symbol_p): Move all logic to symtab.c
29930         (produce_symtab): Update.
29932 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29934         * config/s390/s390-opts.h (enum indirect_branch): Define.
29935         * config/s390/s390-protos.h (s390_return_addr_from_memory)
29936         (s390_indirect_branch_via_thunk)
29937         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
29938         (enum s390_indirect_branch_type): Define.
29939         * config/s390/s390.c (struct s390_frame_layout, struct
29940         machine_function): Remove.
29941         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
29942         (indirect_branch_table_label_no, indirect_branch_table_name):
29943         Define variables.
29944         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
29945         (enum s390_indirect_branch_option): Define.
29946         (s390_return_addr_from_memory): New function.
29947         (s390_handle_string_attribute): New function.
29948         (s390_attribute_table): Add new attribute handler.
29949         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
29950         (s390_indirect_branch_via_thunk): New function.
29951         (s390_indirect_branch_via_inline_thunk): New function.
29952         (s390_function_ok_for_sibcall): When jumping via thunk disallow
29953         sibling call optimization for non z10 compiles.
29954         (s390_emit_call): Force indirect branch target to be a single
29955         register.  Add r1 clobber for non-z10 compiles.
29956         (s390_emit_epilogue): Emit return jump via return_use expander.
29957         (s390_reorg): Handle JUMP_INSNs as execute targets.
29958         (s390_option_override_internal): Perform validity checks for the
29959         new command line options.
29960         (s390_indirect_branch_attrvalue): New function.
29961         (s390_indirect_branch_settings): New function.
29962         (s390_set_current_function): Invoke s390_indirect_branch_settings.
29963         (s390_output_indirect_thunk_function):  New function.
29964         (s390_code_end): Implement target hook.
29965         (s390_case_values_threshold): Implement target hook.
29966         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
29967         macros.
29968         * config/s390/s390.h (struct s390_frame_layout)
29969         (struct machine_function): Move here from s390.c.
29970         (TARGET_INDIRECT_BRANCH_NOBP_RET)
29971         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
29972         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
29973         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
29974         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
29975         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
29976         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
29977         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
29978         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
29979         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
29980         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
29981         (mnemonic attribute): Add values which aren't recognized
29982         automatically.
29983         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
29984         pattern for branch conversion.  Fix mnemonic attribute.
29985         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
29986         indirect branch via thunk if requested.
29987         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
29988         ("*indirect_jump"): Disable for branch conversion using out of
29989         line thunks.
29990         ("indirect_jump_via_thunk<mode>_z10")
29991         ("indirect_jump_via_thunk<mode>")
29992         ("indirect_jump_via_inlinethunk<mode>_z10")
29993         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
29994         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
29995         ("casesi_jump_via_inlinethunk<mode>_z10")
29996         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
29997         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
29998         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
29999         ("*indirect2_jump"): Disable for branch conversion.
30000         ("casesi_jump"): Turn into expander and expand patterns for branch
30001         conversion.
30002         ("return_use"): New expander.
30003         ("*return"): Emit return via thunk and rename it to ...
30004         ("*return<mode>"): ... this one.
30005         * config/s390/s390.opt: Add new options and and enum for the
30006         option values.
30008 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
30010         * lra-constraints.c (match_reload): Unconditionally use
30011         gen_lowpart_SUBREG, rather than selecting between that
30012         and equivalent gen_rtx_SUBREG code.
30014 2018-02-08  Richard Biener  <rguenther@suse.de>
30016         PR tree-optimization/84233
30017         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
30018         changed flag instead of boguously re-using phi_inserted.
30020 2018-02-08  Martin Jambor  <mjambor@suse.cz>
30022         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
30023         static local variables.
30025 2018-02-08  Richard Biener  <rguenther@suse.de>
30027         PR tree-optimization/84278
30028         * tree-vect-stmts.c (vectorizable_store): When looking for
30029         smaller vector types to perform grouped strided loads/stores
30030         make sure the mode is supported by the target.
30031         (vectorizable_load): Likewise.
30033 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
30035         * config/aarch64/aarch64.c (aarch64_components_for_bb):
30036         Increase LDP/STP opportunities by adding adjacent callee-saves.
30038 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
30040         PR rtl-optimization/84068
30041         PR rtl-optimization/83459
30042         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
30044 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
30046         PR tree-optimization/84224
30047         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
30048         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
30049         non-zero arguments.
30051 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
30053         PR target/84113
30054         * config/rs6000/altivec.md (*restore_world): Remove LR use.
30055         * config/rs6000/predicates.md (restore_world_operation): Adjust op
30056         count, remove one USE.
30058 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
30060         * doc/install.texi (Configuration): Document the
30061         --with-long-double-format={ibm,ieee} PowerPC configuration
30062         options.
30064         PR target/84154
30065         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
30066         Convert from define_expand to be define_insn_and_split.  Rework
30067         float/double/_Float128 conversions to QI/HI/SImode to work with
30068         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
30069         conversions to QI/HImode types did a store and then a load to
30070         truncate the value.  For conversions to VSX registers, don't split
30071         the insn, instead emit the code directly.  Use the code iterator
30072         any_fix to combine signed and unsigned conversions.
30073         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
30074         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
30075         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
30076         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
30077         (fix_<mode>di2_hw): Likewise.
30078         (fixuns_<mode>di2_hw): Likewise.
30079         (fix_<mode>si2_hw): Likewise.
30080         (fixuns_<mode>si2_hw): Likewise.
30081         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
30082         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
30083         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
30084         fix<uns>_trunc<SFDF:mode>si2_p8.
30085         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
30086         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
30087         (fix<uns>_<mode>_mem): Likewise.
30088         (fctiw<u>z_<mode>_mem): Likewise.
30089         (fix<uns>_<mode>_mem): Likewise.
30090         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
30091         the register allocator from doing a direct move to the GPRs to do
30092         a store, and instead use the ISA 3.0 store byte/half-word from
30093         vector register instruction.  For IEEE 128-bit floating point,
30094         also optimize stores of 32-bit ints.
30095         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
30097 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
30099         * genextract.c (push_pathstr_operand): New function to support
30100         [a-zA-Z].
30101         (walk_rtx): Call push_pathstr_operand.
30102         (print_path): Support [a-zA-Z].
30104 2018-02-07  Richard Biener  <rguenther@suse.de>
30106         PR tree-optimization/84037
30107         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
30108         (cse_and_gimplify_to_preheader): Declare.
30109         (vect_get_place_in_interleaving_chain): Likewise.
30110         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
30111         ivexpr_map.
30112         (_loop_vec_info::~_loop_vec_info): Delete it.
30113         (cse_and_gimplify_to_preheader): New function.
30114         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
30115         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
30116         (vectorizable_load): Likewise.  For grouped stores always base
30117         the IV on the first element.
30118         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
30119         condition before gimplifying.
30121 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
30123         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
30124         *DIV_EXPR and *MOD_EXPR.
30126 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
30128         PR target/84248
30129         * config/i386/i386.c (ix86_option_override_internal): Mask out
30130         the CF_SET bit when checking -fcf-protection.
30132 2018-02-07  Tom de Vries  <tom@codesourcery.com>
30134         PR libgomp/84217
30135         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
30136         enough.
30138 2018-02-07  Richard Biener  <rguenther@suse.de>
30140         PR tree-optimization/84204
30141         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
30142         this place.
30144         PR tree-optimization/84205
30145         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
30146         special-case isl_ast_op_zdiv_r.
30148         PR tree-optimization/84223
30149         * graphite-scop-detection.c (gather_bbs::before_dom_children):
30150         Only add conditions from within the region.
30151         (gather_bbs::after_dom_children): Adjust.
30153 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
30155         PR target/84209
30156         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
30157         * config/avr/avr.md: Only post-reload split REG-REG moves if
30158         either register is GENERAL_REG_P.
30160 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
30162         PR tree-optimization/84235
30163         * tree-ssa-scopedtables.c
30164         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
30165         if the subtraction is performed in floating point type where NaNs are
30166         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
30167         build 1.  Formatting fix.
30169 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
30171         PR target/84146
30172         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
30173         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
30174         and skip it regardless of bb boundaries.  Use CALL_P macro,
30175         don't test INSN_P (insn) together with CALL_P or JUMP_P check
30176         unnecessarily, formatting fix.
30178 2018-02-06  Michael Collison  <michael.collison@arm.com>
30180         * config/arm/thumb2.md:
30181         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
30182         (*thumb_mov_notscc): Ditto.
30184 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
30186         PR target/84154
30187         * config/rs6000/rs6000.md (su code attribute): Use "u" for
30188         unsigned_fix, not "s".
30190 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30192         * configure.ac (gcc_fn_eh_frame_ro): New function.
30193         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
30194         correct .eh_frame permissions.
30195         * configure: Regenerate.
30197 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
30199         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
30200         irrelevant options.
30202 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30204         * config/rs6000/rs6000.c (rs6000_option_override_internal):
30205         Display warning message for -mno-speculate-indirect-jumps.
30207 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
30209         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
30210         Undocumented.
30211         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
30213 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
30215         PR tree-optimization/84225
30216         * tree-eh.c (find_trapping_overflow): Only call
30217         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
30219 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
30221         PR target/84145
30222         * config/i386/i386.c: Reimplement the check of possible options
30223         -mibt/-mshstk conbination. Change error messages.
30224         * doc/invoke.texi: Fix a typo: remove extra '='.
30226 2018-02-06  Marek Polacek  <polacek@redhat.com>
30228         PR tree-optimization/84228
30229         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
30231 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
30233         PR target/82641
30234         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
30235         emitted arch directives.
30236         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
30237         __ARM_FEATURE_COPROC before changing architectures.
30239 2018-02-06  Richard Biener  <rguenther@suse.de>
30241         * config/i386/i386.c (print_reg): Fix typo.
30242         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
30244 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
30246         * configure: Regenerate.
30248 2018-02-05  Martin Sebor  <msebor@redhat.com>
30250         PR tree-optimization/83369
30251         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
30252         inlining context.
30254 2018-02-05  Martin Liska  <mliska@suse.cz>
30256         * doc/invoke.texi: Cherry-pick upstream r323995.
30258 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
30260         * ira.c (ira_init_register_move_cost): Adjust comment.
30262 2018-02-05  Martin Liska  <mliska@suse.cz>
30264         PR gcov-profile/84137
30265         * doc/gcov.texi: Fix typo in documentation.
30267 2018-02-05  Martin Liska  <mliska@suse.cz>
30269         PR gcov-profile/83879
30270         * doc/gcov.texi: Document necessity of --dynamic-list-data when
30271         using dlopen functionality.
30273 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
30275         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
30276         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
30277         _mm_maskz_range_ss, _mm_mask_range_round_ss,
30278         _mm_maskz_range_round_ss): New intrinsics.
30279         (__builtin_ia32_rangesd128_round)
30280         (__builtin_ia32_rangess128_round): Remove.
30281         (__builtin_ia32_rangesd128_mask_round,
30282         __builtin_ia32_rangess128_mask_round): New builtins.
30283         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
30284         __builtin_ia32_rangess128_round): Remove.
30285         (__builtin_ia32_rangesd128_mask_round,
30286         __builtin_ia32_rangess128_mask_round): New builtins.
30287         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
30288         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
30289         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
30290         "<round_saeonly_constraint>")): Changed to ...
30291         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
30292         "<round_saeonly_scalar_constraint>")): ... this.
30293         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
30294         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
30295         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
30296         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
30297         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
30299 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
30301         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
30302         options.
30303         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
30304         Remove all values except native, 8540 and 8548.
30306 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
30308         * config/i386/i386.c (ix86_output_function_return): Pass
30309         INVALID_REGNUM, instead of -1, as invalid register number to
30310         indirect_thunk_name and output_indirect_thunk.
30312 2018-02-02  Julia Koval  <julia.koval@intel.com>
30314         * config.gcc: Add -march=icelake.
30315         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
30316         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
30317         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
30318         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
30319         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
30320         (processor_target_table): Add icelake.
30321         (ix86_option_override_internal): Handle new PTAs.
30322         (get_builtin_code_for_version): Handle icelake.
30323         (M_INTEL_COREI7_ICELAKE): New.
30324         (fold_builtin_cpu): Handle icelake.
30325         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
30326         * doc/invoke.texi: Add -march=icelake.
30328 2018-02-02  Julia Koval  <julia.koval@intel.com>
30330         * config/i386/i386.c (ix86_option_override_internal): Change flags type
30331         to wide_int_bitmask.
30332         * wide-int-bitmask.h: New.
30334 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
30336         PR target/84066
30337         * config/i386/i386.md: Replace Pmode with word_mode in
30338         builtin_setjmp_setup and builtin_longjmp to support x32.
30340 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
30342         PR target/56010
30343         PR target/83743
30344         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
30345         #include "opts.h".
30346         (rs6000_supported_cpu_names): New static variable.
30347         (linux_cpu_translation_table): Likewise.
30348         (elf_platform) <cpu>: Define new static variable and use it.
30349         Translate kernel AT_PLATFORM name to canonical name if needed.
30350         Error if platform name is unknown.
30352 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
30354         PR target/84089
30355         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
30357 2018-02-01  Jeff Law  <law@redhat.com>
30359         PR target/84128
30360         * config/i386/i386.c (release_scratch_register_on_entry): Add new
30361         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
30362         the scratch if RELEASE_VIA_POP is false.
30363         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
30364         If we have to save a temporary register, decrement SIZE appropriately.
30365         Pass new arguments to release_scratch_register_on_entry.
30366         (ix86_adjust_stack_and_probe): Likewise.
30367         (ix86_emit_probe_stack_range): Pass new arguments to
30368         release_scratch_register_on_entry.
30370 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
30372         PR rtl-optimization/84157
30373         * combine.c (change_zero_ext): Use REG_P predicate in
30374         front of HARD_REGISTER_P predicate.
30376 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
30378         * config/avr/avr.c (avr_option_override): Move disabling of
30379         -fdelete-null-pointer-checks to...
30380         * common/config/avr/avr-common.c (avr_option_optimization_table):
30381         ...here.
30383 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30385         PR tree-optimization/81635
30386         * tree-data-ref.c (split_constant_offset_1): For types that
30387         wrap on overflow, try to use range info to prove that wrapping
30388         cannot occur.
30390 2018-02-01  Renlin Li  <renlin.li@arm.com>
30392         PR target/83370
30393         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
30394         TAILCALL_ADDR_REGS.
30395         (aarch64_register_move_cost): Likewise.
30396         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
30397         TAILCALL_ADDR_REGS.
30398         (REG_CLASS_NAMES): Likewise.
30399         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
30400         TAILCALL_ADDR_REGS. Remove IP registers.
30401         * config/aarch64/aarch64.md (Ucs): Update register constraint.
30403 2018-02-01  Richard Biener  <rguenther@suse.de>
30405         * domwalk.h (dom_walker::dom_walker): Add additional constructor
30406         for specifying RPO order and allow NULL for that.
30407         * domwalk.c (dom_walker::dom_walker): Likewise.
30408         (dom_walker::walk): Handle NULL RPO order.
30409         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
30410         in RPO order.
30411         (rewrite_update_dom_walker): Likewise.
30412         (mark_def_dom_walker): Likewise.
30414 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30416         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
30417         (aarch64_maybe_expand_sve_subreg_move): Declare.
30418         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
30419         * config/aarch64/predicates.md (aarch64_any_register_operand): New
30420         predicate.
30421         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
30422         that are semantically a reverse operation.
30423         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
30424         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
30425         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
30426         functions.
30427         (aarch64_can_change_mode_class): For big-endian, forbid changes
30428         between two SVE modes if they have different element sizes.
30430 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30432         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
30433         the TImode handling for big-endian targets.
30435 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30437         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
30438         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
30439         not just bytes.
30440         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
30441         Remove BSWAP handing for big-endian targets and use the form of
30442         LD1RQ appropariate for the mode.
30444 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30446         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
30447         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
30448         duplicated element.
30450 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
30452         PR tearget/83845
30453         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
30454         check for operands that need to go through aarch64_sve_reload_be.
30456 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
30458         PR tree-optimization/81661
30459         PR tree-optimization/84117
30460         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
30461         * tree-eh.c: Include gimplify.h.
30462         (find_trapping_overflow, replace_trapping_overflow,
30463         rewrite_to_non_trapping_overflow): New functions.
30464         * tree-vect-loop.c: Include tree-eh.h.
30465         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
30466         * tree-data-ref.c: Include tree-eh.h.
30467         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
30469 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
30471         PR rtl-optimization/84123
30472         * combine.c (change_zero_ext): Check if hard register satisfies
30473         can_change_dest_mode before calling gen_lowpart_SUBREG.
30475 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
30477         PR target/82444
30478         * ira.c (ira_init_register_move_cost): Remove assert.
30480 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
30482         PR rtl-optimization/84071
30483         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
30484         * doc/tm.texi: Regenerate.
30486 2018-01-31  Richard Biener  <rguenther@suse.de>
30488         PR tree-optimization/84132
30489         * tree-data-ref.c (analyze_miv_subscript): Properly
30490         check whether evolution_function_is_affine_multivariate_p
30491         before calling gcd_of_steps_may_divide_p.
30493 2018-01-31  Julia Koval  <julia.koval@intel.com>
30495         PR target/83618
30496         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
30497         * config/i386/i386.md (rdpid_rex64) New.
30498         (rdpid): Make 32bit only.
30500 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
30502         PR lto/84105
30503         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
30504         an IDENTIFIER_NODE for FUNCTION_TYPE's.
30506 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
30508         Revert
30509         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
30511         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
30513 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
30515         PR rtl-optimization/84071
30516         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
30517         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
30519 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
30521         * config/arc/arc.c (arc_handle_aux_attribute): New function.
30522         (arc_attribute_table): Add 'aux' attribute.
30523         (arc_in_small_data_p): Consider aux like variables.
30524         (arc_is_aux_reg_p): New function.
30525         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
30526         (arc_get_aux_arg): New function.
30527         (prepare_move_operands): Handle aux-register access.
30528         (arc_handle_aux_attribute): New function.
30529         * doc/extend.texi (ARC Variable attributes): Add subsection.
30531 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
30533         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
30534         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
30535         (arc_attribute_table): Add 'uncached' attribute.
30536         (arc_print_operand): Print '.di' flag for uncached memory
30537         accesses.
30538         (arc_in_small_data_p): Do not consider for small data the uncached
30539         types.
30540         (arc_is_uncached_mem_p): New function.
30541         * config/arc/predicates.md (compact_store_memory_operand): Check
30542         for uncached memory accesses.
30543         (nonvol_nonimm_operand): Likewise.
30544         * doc/extend.texi (ARC Type Attribute): New subsection.
30546 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
30548         PR c/84100
30549         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
30550         falign-loops=): Add Optimization flag.
30552 2018-01-30  Jeff Law  <law@redhat.com>
30554         PR target/84064
30555         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
30556         INT_REGISTERS_SAVED.  Check it prior to calling
30557         get_scratch_register_on_entry.
30558         (ix86_adjust_stack_and_probe): Similarly.
30559         (ix86_emit_probe_stack_range): Similarly.
30560         (ix86_expand_prologue): Corresponding changes.
30562 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30564         PR target/40411
30565         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
30566         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
30568 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
30570         PR target/84112
30571         * lra-constraints.c (curr_insn_transform): Process AND in the
30572         address.
30574 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
30576         PR rtl-optimization/83986
30577         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
30578         dependence against last_pending_memory_flush in addition to
30579         pending_jump_insns.
30581 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
30583         PR tree-optimization/81611
30584         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
30585         copies.
30587 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30589         PR target/83758
30590         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
30591         a reg rtx.
30593 2018-01-30  Richard Biener  <rguenther@suse.de>
30594             Jakub Jelinek  <jakub@redhat.com>
30596         PR tree-optimization/84111
30597         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
30598         inner loops added during recursion, as they don't have up-to-date
30599         SSA form.
30601 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
30603         PR ipa/81360
30604         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
30605         (can_inline_edge_by_limits_p): ... here.
30606         (can_early_inline_edge_p, check_callers,
30607         update_caller_keys, update_callee_keys, recursive_inlining,
30608         add_new_edges_to_heap, speculation_useful_p,
30609         inline_small_functions,
30610         inline_small_functions, flatten_function,
30611         inline_to_all_callers_1): Update.
30613 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
30615         * profile-count.c (profile_count::combine_with_ipa_count): Handle
30616         zeros correctly.
30618 2018-01-30  Richard Biener  <rguenther@suse.de>
30620         PR tree-optimization/83008
30621         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
30622         invariant and constant vector uses in stmts when they need
30623         more than one stmt.
30625 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30627         PR bootstrap/84017
30628         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
30629         * configure: Regenerate.
30631 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
30633         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
30634         pattern.
30635         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
30636         Use gen_rtx_REG rather than gen_lowpart.
30638 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
30640         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
30641         rather than 0 when creating partial subregs.
30643 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
30645         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
30646         of usage.
30648 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
30650         PR target/81550
30651         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
30652         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
30653         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
30654         flags.  This restores the settings used before the 2017-07-24.
30655         Turning off pre increment/decrement/modify allows IVOPTS to
30656         optimize DF/SF loops where the index is an int.
30658 2018-01-29  Richard Biener  <rguenther@suse.de>
30659             Kelvin Nilsen  <kelvin@gcc.gnu.org>
30661         PR bootstrap/80867
30662         * tree-vect-stmts.c (vectorizable_call): Don't call
30663         targetm.vectorize_builtin_md_vectorized_function if callee is
30664         NULL.
30666 2018-01-22  Carl Love  <cel@us.ibm.com>
30668         * doc/extend.tex: Fix typo in second arg in
30669         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
30671 2018-01-29  Richard Biener  <rguenther@suse.de>
30673         PR tree-optimization/84086
30674         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
30675         (flush_ssaname_freelist): When SSA names were released reset
30676         the SCEV hash table.
30678 2018-01-29  Richard Biener  <rguenther@suse.de>
30680         PR tree-optimization/84057
30681         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
30682         removed paths when removing edges.
30684 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
30686         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
30687         -mfunction-return=@var{choice}.
30689 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
30691         PR diagnostic/84034
30692         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
30693         Handle CR like TAB.
30694         (layout::print_source_line): Likewise.
30695         (test_get_line_width_without_trailing_whitespace): Add test cases.
30697 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
30699         PR middle-end/84040
30700         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
30701         debug insns.
30703 2018-01-26  Jim Wilson  <jimw@sifive.com>
30705         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
30707         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
30708         specified.
30710 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30712         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
30713         and CMP + SUB-immediate -> SUBS.
30715 2018-01-26  Martin Sebor  <msebor@redhat.com>
30717         PR tree-optimization/83896
30718         * tree-ssa-strlen.c (get_string_len): Rename...
30719         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
30720         Avoid assuming length is constant.
30721         (handle_char_store): Use HOST_WIDE_INT for string length.
30723 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
30725         PR target/81763
30726         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
30727         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
30729 2018-01-26  Richard Biener  <rguenther@suse.de>
30731         PR rtl-optimization/84003
30732         * dse.c (record_store): Only record redundant stores when
30733         the earlier store aliases at least all accesses the later one does.
30735 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
30737         PR rtl-optimization/83985
30738         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
30739         REG_CFA_RESTORE insns.
30740         (delete_unmarked_insns): Don't ignore separate shrink wrapping
30741         REG_CFA_RESTORE insns here.
30743         PR c/83989
30744         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
30745         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
30747 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
30749         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
30750         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
30751         (arc_init): Likewise.
30752         (arc_override_options): Likewise.
30753         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
30754         value.
30755         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
30756         support.
30757         * config/arc/arc.h (TARGET_DBNZ): Define.
30758         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
30759         properly set the tune attribute.
30760         (dbnz): Use TARGET_DBNZ guard.
30761         * config/arc/arc.opt (mtune): Add core3 option.
30763 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
30765         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
30766         recognize new pic like addresses.
30767         (arc_delegitimize_address): Clean up.
30769 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
30771         * config/arc/arc-arches.def: Option mrf16 valid for all
30772         architectures.
30773         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
30774         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
30775         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
30776         * config/arc/arc-tables.opt: Regenerate.
30777         * config/arc/arc.c (arc_conditional_register_usage): Handle
30778         reduced register file case.
30779         (arc_file_start): Set must have build attributes.
30780         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
30781         mrf16 option value.
30782         * config/arc/arc.opt (mrf16): Add new option.
30783         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
30784         * config/arc/genmultilib.awk: Handle new mrf16 option.
30785         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
30786         * config/arc/t-multilib: Regenerate.
30787         * doc/invoke.texi (ARC Options): Document mrf16 option.
30789 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
30791         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
30792         * config/arc/arc.c (arc_handle_secure_attribute): New function.
30793         (arc_attribute_table): Add 'secure_call' attribute.
30794         (arc_print_operand): Print secure call operand.
30795         (arc_function_ok_for_sibcall): Don't optimize tail calls when
30796         secure.
30797         (arc_is_secure_call_p): New function.  * config/arc/arc.md
30798         (call_i): Add support for sjli instruction.
30799         (call_value_i): Likewise.
30800         * config/arc/constraints.md (Csc): New constraint.
30802 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
30803             John Eric Martin  <John.Martin@emmicro-us.com>
30805         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
30806         * config/arc/arc.c (_arc_jli_section): New struct.
30807         (arc_jli_section): New type.
30808         (rc_jli_sections): New static variable.
30809         (arc_handle_jli_attribute): New function.
30810         (arc_attribute_table): Add jli_always and jli_fixed attribute.
30811         (arc_file_end): New function.
30812         (TARGET_ASM_FILE_END): Define.
30813         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
30814         (arc_add_jli_section): New function.
30815         (jli_call_scan): Likewise.
30816         (arc_reorg): Call jli_call_scan.
30817         (arc_output_addsi): Remove 'S' from printing asm operand.
30818         (arc_is_jli_call_p): New function.
30819         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
30820         operand.
30821         (movhi_insn): Likewise.
30822         (movsi_insn): Likewise.
30823         (movsi_set_cc_insn): Likewise.
30824         (loadqi_update): Likewise.
30825         (load_zeroextendqisi_update): Likewise.
30826         (load_signextendqisi_update): Likewise.
30827         (loadhi_update): Likewise.
30828         (load_zeroextendhisi_update): Likewise.
30829         (load_signextendhisi_update): Likewise.
30830         (loadsi_update): Likewise.
30831         (loadsf_update): Likewise.
30832         (movsicc_insn): Likewise.
30833         (bset_insn): Likewise.
30834         (bxor_insn): Likewise.
30835         (bclr_insn): Likewise.
30836         (bmsk_insn): Likewise.
30837         (bicsi3_insn): Likewise.
30838         (cmpsi_cc_c_insn): Likewise.
30839         (movsi_ne): Likewise.
30840         (movsi_cond_exec): Likewise.
30841         (clrsbsi2): Likewise.
30842         (norm_f): Likewise.
30843         (normw): Likewise.
30844         (swap): Likewise.
30845         (divaw): Likewise.
30846         (flag): Likewise.
30847         (sr): Likewise.
30848         (kflag): Likewise.
30849         (ffs): Likewise.
30850         (ffs_f): Likewise.
30851         (fls): Likewise.
30852         (call_i): Remove 'S' asm letter, add jli instruction.
30853         (call_value_i): Likewise.
30854         * config/arc/arc.op (mjli-always): New option.
30855         * config/arc/constraints.md (Cji): New constraint.
30856         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
30857         operand.
30858         (subsf3_fpx): Likewise.
30859         (mulsf3_fpx): Likewise.
30860         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
30861         asm operand.
30862         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
30863         function attrbutes.
30864         * doc/invoke.texi (ARC): Document mjli-always option.
30866 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
30868         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
30869         avoid addition with 0 and use incw and decw where possible.
30871 2018-01-26  Richard Biener  <rguenther@suse.de>
30873         PR tree-optimization/81082
30874         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
30875         association if it requires casting to unsigned.
30876         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
30877         from fold_plusminus_mult_expr to catch important cases late when
30878         range info is available.
30880 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30882         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
30883         * configure.ac (hidden_linkonce): New test.
30884         * configure: Regenerate.
30885         * config.in: Regenerate.
30887 2018-01-26  Julia Koval  <julia.koval@intel.com>
30889         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
30890         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
30891         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
30892         _mm_mask_bitshuffle_epi64_mask): Fix type.
30893         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
30894         USI_FTYPE_V4DI_V4DI_USI): Remove.
30895         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
30896         __builtin_ia32_vpshufbitqmb256_mask,
30897         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
30898         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
30899         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
30901 2018-01-26  Alan Modra  <amodra@gmail.com>
30903         PR target/84033
30904         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
30905         UNSPEC_VBPERMQ.  Sort other unspecs.
30907 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
30909         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
30911 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
30913         PR middle-end/83055
30914         * predict.c (drop_profile): Do not push/pop cfun; update also
30915         node->count.
30916         (handle_missing_profiles): Fix logic looking for zero profiles.
30918 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
30920         PR middle-end/83977
30921         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
30922         on functions with #pragma omp declare simd or functions with simd
30923         attribute.
30924         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
30925         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
30926         Remove trailing \n from warning_at calls.
30928 2018-01-25  Tom de Vries  <tom@codesourcery.com>
30930         PR target/84028
30931         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30932         for neutered workers.
30934 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
30936         PR target/68467
30937         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
30938         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
30940 2018-01-24  Jeff Law  <law@redhat.com>
30942         PR target/83994
30943         * i386.c (get_probe_interval): Move to earlier point.
30944         (ix86_compute_frame_layout): If -fstack-clash-protection and
30945         the frame is larger than the probe interval, then use pushes
30946         to save registers rather than reg->mem moves.
30947         (ix86_expand_prologue): Remove conditional for int_registers_saved
30948         assertion.
30950 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
30952         PR target/84014
30953         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
30954         min/max for never referenced object.
30956 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
30958         PR middle-end/83977
30959         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
30960         here.
30961         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
30962         attributes from DECL_ATTRIBUTES (decl) without affecting
30963         DECL_ATTRIBUTES (current_function_decl).
30964         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
30965         functions with non-NULL DECL_ABSTRACT_ORIGIN.
30967 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
30969         PR tree-optimization/83979
30970         * fold-const.c (fold_comparison): Use constant_boolean_node
30971         instead of boolean_{true,false}_node.
30973 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
30975         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
30976         with zero counts.
30978 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30980         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30981         Simplify the clause that sets the length attribute.
30982         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30983         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
30984         clause that sets the length attribute.
30985         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30987 2018-01-24  Tom de Vries  <tom@codesourcery.com>
30989         PR target/83589
30990         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
30991         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
30992         Add strict parameter.
30993         (prevent_branch_around_nothing): Insert dummy insn between branch to
30994         label and label with no ptx insn inbetween.
30995         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
30997 2018-01-24  Tom de Vries  <tom@codesourcery.com>
30999         PR target/81352
31000         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
31001         for neutered threads in warp.
31002         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
31004 2018-01-24  Richard Biener  <rguenther@suse.de>
31006         PR tree-optimization/83176
31007         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
31008         operands.
31010 2018-01-24  Richard Biener  <rguenther@suse.de>
31012         PR tree-optimization/82819
31013         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
31014         code generating pluses that are no-ops in the target precision.
31016 2018-01-24  Richard Biener  <rguenther@suse.de>
31018         PR middle-end/84000
31019         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
31021 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
31023         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
31024         to merge probabilities.
31025         * predict.c (probably_never_executed): Also mark as cold functions
31026         with global 0 profile and guessed local profile.
31027         * profile-count.c (profile_probability::combine_with_count): New
31028         member function.
31029         * profile-count.h (profile_probability::operator*,
31030         profile_probability::operator*=, profile_probability::operator/,
31031         profile_probability::operator/=): Reduce precision to adjusted
31032         and set value to guessed on contradictory divisions.
31033         (profile_probability::combine_with_freq): Remove.
31034         (profile_probability::combine_wiht_count): Declare.
31035         (profile_count::force_nonzero):: Set to adjusted.
31036         (profile_count::probability_in):: Set quality to adjusted.
31037         * tree-ssa-tail-merge.c (replace_block_by): Use
31038         combine_with_count.
31040 2018-01-23  Andrew Waterman  <andrew@sifive.com>
31041             Jim Wilson  <jimw@sifive.com>
31043         * config/riscv/riscv.c (riscv_stack_boundary): New.
31044         (riscv_option_override): Set riscv_stack_boundary.  Handle
31045         riscv_preferred_stack_boundary_arg.
31046         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
31047         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
31048         (STACK_BOUNDARY): Set to riscv_stack_boundary.
31049         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
31050         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
31051         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
31053 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
31055         PR target/83905
31056         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
31057         of struct ix86_frame.
31058         (ix86_expand_epilogue): Likewise.  Add a local variable for
31059         the reg_save_offset field in struct ix86_frame.
31061 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
31063         PR tree-optimization/82604
31064         * tree-loop-distribution.c (enum partition_kind): New enum item
31065         PKIND_PARTIAL_MEMSET.
31066         (partition_builtin_p): Support above new enum item.
31067         (generate_code_for_partition): Ditto.
31068         (compute_access_range): Differentiate cases that equality can be
31069         proven at all loops, the innermost loops or no loops.
31070         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
31071         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
31072         (finalize_partitions, distribute_loop): Don't fuse partition of
31073         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
31074         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
31075         parloop is enabled.
31077 2018-01-23  Martin Liska  <mliska@suse.cz>
31079         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
31080         order to ignore the predictor.
31081         (PRED_POLYMORPHIC_CALL): Likewise.
31082         (PRED_RECURSIVE_CALL): Likewise.
31084 2018-01-23  Martin Liska  <mliska@suse.cz>
31086         * tree-profile.c (tree_profiling): Print function header to
31087         aware reader which function we are working on.
31088         * value-prof.c (gimple_find_values_to_profile): Do not print
31089         not interesting value histograms.
31091 2018-01-23  Martin Liska  <mliska@suse.cz>
31093         * profile-count.h (enum profile_quality): Add
31094         profile_uninitialized as the first value. Do not number values
31095         as they are zero based.
31096         (profile_count::verify): Update sanity check.
31097         (profile_probability::verify): Likewise.
31099 2018-01-23  Nathan Sidwell  <nathan@acm.org>
31101         * doc/invoke.texi (ffor-scope): Deprecate.
31103 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
31105         PR tree-optimization/83510
31106         * domwalk.c (set_all_edges_as_executable): New function.
31107         (dom_walker::dom_walker): Convert bool param
31108         "skip_unreachable_blocks" to enum reachability.  Move setup of
31109         edge flags to set_all_edges_as_executable and only do it when
31110         reachability is REACHABLE_BLOCKS.
31111         * domwalk.h (enum dom_walker::reachability): New enum.
31112         (dom_walker::dom_walker): Convert bool param
31113         "skip_unreachable_blocks" to enum reachability.
31114         (set_all_edges_as_executable): New decl.
31115         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
31116         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
31117         "reachability".
31118         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
31119         but converting true to REACHABLE_BLOCKS.
31120         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
31121         * tree-vrp.c
31122         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
31123         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
31124         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
31125         REACHABLE_BLOCKS.
31126         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
31127         if check_all_array_refs will be called.
31129 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
31131         * tree.c (selftest::test_location_wrappers): Add more test
31132         coverage.
31134 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
31136         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
31137         (selftest::test_bit_in_range): Likewise.
31139 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
31141         PR testsuite/83888
31142         * doc/sourcebuild.texi (vect_float): Say that the selector
31143         only describes the situation when -funsafe-math-optimizations is on.
31144         (vect_float_strict): Document.
31146 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
31148         PR tree-optimization/83965
31149         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
31150         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
31151         instead of checking only for a reduction.
31152         (vect_recog_widen_sum_pattern): Likewise.
31154 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
31156         * predict.c (probably_never_executed): Only use precise profile info.
31157         (compute_function_frequency): Skip after inlining hack since we now
31158         have quality checking.
31160 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
31162         * profile-count.h (profile_probability::very_unlikely,
31163         profile_probability::unlikely, profile_probability::even): Set
31164         precision to guessed.
31166 2018-01-23  Richard Biener  <rguenther@suse.de>
31168         PR tree-optimization/83963
31169         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
31170         Properly terminate dominator walk when crossing the exit edge not
31171         when visiting its source block.
31173 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
31175         PR c++/83918
31176         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
31177         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
31179 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
31181         PR tree-optimization/83957
31182         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
31183         semicolon after for body surrounded by braces.
31185         PR tree-optimization/83081
31186         * profile-count.h (profile_probability::split): New method.
31187         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
31188         Use profile_probability::split.
31189         (do_compare_rtx_and_jump): Fix adjustment of probabilities
31190         when splitting a single conditional jump into 2.
31192 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
31194         PR tree-optimization/69452
31195         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
31196         decl.
31198 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31200         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
31201         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
31202         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
31204 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31206         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
31207         declaration.
31208         * config/rl78/rl78.md (movdi): New define_expand.
31209         * config/rl78/rl78.c (rl78_split_movdi): New function.
31211 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
31213         PR target/83862
31214         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
31215         no longer used.
31216         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
31217         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
31218         128-bit to produce an UNSPEC move to get the double word with the
31219         signbit and then a shift directly to do signbit.
31220         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
31221         implementation with a new version that just does either a direct
31222         move or a regular move.  Move memory interface to separate insns.
31223         Move insns so they are next to the expander.
31224         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
31225         with signbit move.  Split big and little endian case.
31226         (signbit<mode>2_dm_mem_le): Likewise.
31227         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
31228         (signbit<mode>2_dm2): Likewise.
31230 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31232         * config/rl78/rl78.md (anddi3): New define_expand.
31234 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31236         * config/rl78/rl78.md (umindi3): New define_expand.
31238 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31240         * config/rl78/rl78.md (smindi3): New define_expand.
31242 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31244         * config/rl78/rl78.md (smaxdi3): New define_expand.
31246 2018-01-22  Carl Love  <cel@us.ibm.com>
31248         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
31249         LVX_V1TI): Add macro expansion.
31250         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
31251         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
31252         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
31253         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
31254         Change check to determine if the instruction is a byte reversing
31255         entry.  Fix typo in comment.
31256         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
31257         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
31258         Add def_builtin calls for new builtins.
31259         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
31260         Add define_insn expansion.
31262 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31264         * config/rl78/rl78.md (umaxdi3): New define_expand.
31266 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
31268         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
31269         for non-QImode registers.
31271 2018-01-22  Richard Biener  <rguenther@suse.de>
31273         PR tree-optimization/83963
31274         * graphite-scop-detection.c (scop_detection::get_sese): Delay
31275         including the loop exit block.
31276         (scop_detection::merge_sese): Likewise.
31277         (scop_detection::add_scop): Do it here instead.
31279 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31281         * doc/sourcebuild.texi (arm_softfloat): Document.
31283 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
31285         PR gcc/77734
31286         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
31287         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
31288         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
31290 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31291             David Edelsohn  <dje.gcc@gmail.com>
31293         PR target/83946
31294         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
31295         Change "crset eq" to "crset 2".
31296         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
31297         (*call_indirect_aix<mode>_nospec): Likewise.
31298         (*call_value_indirect_aix<mode>_nospec): Likewise.
31299         (*call_indirect_elfv2<mode>_nospec): Likewise.
31300         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
31301         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
31302         change assembly output from . to $.
31303         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
31304         (indirect_jump<mode>_nospec): Change assembly output from . to $.
31305         (*tablejump<mode>_internal1_nospec): Likewise.
31307 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
31309         PR target/80870
31310         * config/sh/sh_optimize_sett_clrt.cc:
31311         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
31313 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
31315         PR tree-optimization/83940
31316         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
31317         offset_dt to vect_constant_def rather than vect_unknown_def_type.
31318         (vect_check_load_store_mask): Add a mask_dt_out parameter and
31319         use it to pass back the definition type.
31320         (vect_check_store_rhs): Likewise rhs_dt_out.
31321         (vect_build_gather_load_calls): Add a mask_dt argument and use
31322         it instead of a call to vect_is_simple_use.
31323         (vectorizable_store): Update calls to vect_check_load_store_mask
31324         and vect_check_store_rhs.  Use the dt returned by the latter instead
31325         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
31326         instead of calls to vect_is_simple_use.  Pass the scalar rather
31327         than the vector operand to vect_is_simple_use when handling
31328         second and subsequent copies of an rhs value.
31329         (vectorizable_load): Update calls to vect_check_load_store_mask
31330         and vect_build_gather_load_calls.  Use the cached mask_dt and
31331         gs_info.offset_dt instead of calls to vect_is_simple_use.
31333 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
31335         PR middle-end/83945
31336         * tree-emutls.c: Include gimplify.h.
31337         (lower_emutls_2): New function.
31338         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
31339         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
31340         it before further processing.
31342         PR target/83930
31343         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
31344         UINTVAL (trueop1) instead of INTVAL (op1).
31346 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
31348         PR debug/81570
31349         PR debug/83728
31350         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
31351         INCOMING_FRAME_SP_OFFSET if not defined.
31352         (scan_trace): Add ENTRY argument.  If true and
31353         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
31354         emit a note to adjust the CFA offset.
31355         (create_cfi_notes): Adjust scan_trace callers.
31356         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
31357         INCOMING_FRAME_SP_OFFSET in the CIE.
31358         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
31359         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
31360         Likewise.
31361         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
31362         * doc/tm.texi: Regenerated.
31364 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
31366         PR rtl-optimization/83147
31367         * lra-constraints.c (remove_inheritance_pseudos): Use
31368         lra_substitute_pseudo_within_insn.
31370 2018-01-19  Tom de Vries  <tom@codesourcery.com>
31371             Cesar Philippidis  <cesar@codesourcery.com>
31373         PR target/83920
31374         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
31376 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
31378         PR target/83790
31379         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
31380         spaces for function labels.
31382 2018-01-19  Martin Liska  <mliska@suse.cz>
31384         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
31385         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
31386         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
31387         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
31388         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
31389         (PRED_CONST_RETURN): Change from 69 to 65.
31390         (PRED_NULL_RETURN): Change from 91 to 71.
31391         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
31392         (PRED_LOOP_GUARD): Change from 66 to 73.
31394 2018-01-19  Martin Liska  <mliska@suse.cz>
31396         * predict.c (predict_insn_def): Add new assert.
31397         (struct branch_predictor): Change type to signed integer.
31398         (test_prediction_value_range): Amend test to cover
31399         PROB_UNINITIALIZED.
31400         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
31401         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
31402         (PRED_LOOP_ITERATIONS_MAX): Likewise.
31403         (PRED_LOOP_IV_COMPARE): Likewise.
31404         * predict.h (PROB_UNINITIALIZED): Define new constant.
31406 2018-01-19  Martin Liska  <mliska@suse.cz>
31408         * predict.c (dump_prediction): Add new format for
31409         analyze_brprob.py script which is enabled with -details
31410         suboption.
31411         * profile-count.h (precise_p): New function.
31413 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
31415         PR tree-optimization/83922
31416         * tree-vect-loop.c (vect_verify_full_masking): Return false if
31417         there are no statements that need masking.
31418         (vect_active_double_reduction_p): New function.
31419         (vect_analyze_loop_operations): Use it when handling phis that
31420         are not in the loop header.
31422 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
31424         PR tree-optimization/83914
31425         * tree-vect-loop.c (vectorizable_induction): Don't convert
31426         init_expr or apply the peeling adjustment for inductions
31427         that are nested within the vectorized loop.
31429 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31431         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
31432         instead of NEG.
31434 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
31436         PR sanitizer/81715
31437         PR testsuite/83882
31438         * function.h (gimplify_parameters): Add gimple_seq * argument.
31439         * function.c: Include gimple.h and options.h.
31440         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
31441         for the added local temporaries if needed.
31442         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
31443         if there are any parameter cleanups, wrap whole body into a
31444         try/finally with the cleanups.
31446 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
31448         PR target/82964
31449         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
31450         Use GET_MODE_CLASS for scalar floating point.
31452 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
31454         PR ipa/82256
31455         patch by PaX Team
31456         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
31457         Fix call of call_cgraph_insertion_hooks.
31459 2018-01-18  Martin Sebor  <msebor@redhat.com>
31461         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
31463 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
31465         PR ipa/83619
31466         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
31467         frequencies.
31469 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
31471         PR other/70268
31472         * common.opt: (-ffile-prefix-map): New option.
31473         * opts.c (common_handle_option): Defer it.
31474         * opts-global.c (handle_common_deferred_options): Handle it.
31475         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
31476         * file-prefix-map.h: New file.
31477         (remap_debug_filename, add_debug_prefix_map): ...here.
31478         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
31479         * final.c (debug_prefix_map, add_debug_prefix_map
31480         remap_debug_filename): Move to...
31481         * file-prefix-map.c: New file.
31482         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
31483         generalize, get rid of alloca(), use strrchr() instead of strchr().
31484         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
31485         Implement in terms of add_prefix_map().
31486         (remap_macro_filename, remap_debug_filename): Implement in term of
31487         remap_filename().
31488         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
31489         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
31490         * dbxout.c: Include file-prefix-map.h.
31491         * varasm.c: Likewise.
31492         * vmsdbgout.c: Likewise.
31493         * xcoffout.c: Likewise.
31494         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
31495         * doc/cppopts.texi (-fmacro-prefix-map): Document.
31496         * doc/invoke.texi (-ffile-prefix-map): Document.
31497         (-fdebug-prefix-map): Update description.
31499 2018-01-18  Martin Liska  <mliska@suse.cz>
31501         * config/i386/i386.c (indirect_thunk_name): Document that also
31502         lfence is emitted.
31503         (output_indirect_thunk): Document why both instructions
31504         (pause and lfence) are generated.
31506 2018-01-18  Richard Biener  <rguenther@suse.de>
31508         PR tree-optimization/83887
31509         * graphite-scop-detection.c
31510         (scop_detection::get_nearest_dom_with_single_entry): Remove.
31511         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
31512         (scop_detection::merge_sese): Re-implement with a flood-fill
31513         algorithm that properly finds a SESE region if it exists.
31515 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
31517         PR c/61240
31518         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
31519         pointer_diff optimizations use view_convert instead of convert.
31521 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31523         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
31524         Generate different code for -mno-speculate-indirect-jumps.
31525         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
31526         (*call_indirect_aix<mode>): Disable for
31527         -mno-speculate-indirect-jumps.
31528         (*call_indirect_aix<mode>_nospec): New define_insn.
31529         (*call_value_indirect_aix<mode>): Disable for
31530         -mno-speculate-indirect-jumps.
31531         (*call_value_indirect_aix<mode>_nospec): New define_insn.
31532         (*sibcall_nonlocal_sysv<mode>): Generate different code for
31533         -mno-speculate-indirect-jumps.
31534         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
31536 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
31538         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
31539         long double type, set the flags for noting the default long double
31540         type, even if we don't pass or return a long double type.
31542 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
31544         PR ipa/83051
31545         * ipa-inline.c (flatten_function): Do not overwrite final inlining
31546         failure.
31548 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
31550         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
31551         support for merge[hl].
31552         (fold_mergehl_helper): New helper function.
31553         (tree-vector-builder.h): New #include for tree_vector_builder usage.
31554         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
31555         (altivec_vmrglw_direct): Add xxmrglw insn.
31557 2018-01-17  Andrew Waterman  <andrew@sifive.com>
31559         * config/riscv/riscv.c (riscv_conditional_register_usage): If
31560         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
31562 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
31564         PR lto/83121
31565         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
31566         call the lto_location_cache before reading the
31567         DECL_SOURCE_LOCATION of the types.
31569 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
31570             Richard Sandiford  <richard.sandiford@linaro.org>
31572         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
31573         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
31574         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
31575         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
31576         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
31577         Add declaration.
31578         * config/aarch64/constraints.md (aarch64_movti_operand):
31579         Limit immediates.
31580         * config/aarch64/predicates.md (Uti): Add new constraint.
31582 2018-01-17  Carl Love  <cel@us.ibm.com>
31584         * config/rs6000/vsx.md (define_expand xl_len_r,
31585         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
31586         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
31587         lxvll.
31588         (define_expand, define_insn): Move the shift left from  the
31589         define_insn to the define_expand for lxvl and stxvl instructions.
31590         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
31591         and XL_LEN_R definitions to PURE.
31593 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
31595         * config/i386/i386.c (indirect_thunk_name): Declare regno
31596         as unsigned int.  Compare regno with INVALID_REGNUM.
31597         (output_indirect_thunk): Ditto.
31598         (output_indirect_thunk_function): Ditto.
31599         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
31600         in the call to output_indirect_thunk_function.
31602 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
31604         PR middle-end/83884
31605         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
31606         rather than the size of inner_type to determine the stack slot size
31607         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
31609 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
31611         PR target/83546
31612         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
31613         to PTA_SILVERMONT.
31615 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
31617         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
31618         endian Linux systems to optionally enable multilibs for selecting
31619         the long double type if the user configured an explicit type.
31620         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
31621         have no long double multilibs if not defined.
31622         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
31623         warn if the user used -mabi={ieee,ibm}longdouble and we built
31624         multilibs for long double.
31625         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
31626         appropriate multilib option.
31627         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
31628         multilib options.
31629         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
31630         for building long double multilibs.
31631         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
31633 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
31635         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
31636         copies.
31638         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
31639         64 bits.
31640         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
31641         128 bits.
31643         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
31644         variables.
31646         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
31647         return value.
31649 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
31651         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
31652         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
31654 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
31656         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
31657         different rtl trees depending on TARGET_64BIT.
31658         (rs6000_gen_lvx): Likewise.
31660 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
31662         * config/visium/visium.md (nop): Tweak comment.
31663         (hazard_nop): Likewise.
31665 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31667         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
31668         -mspeculate-indirect-jumps.
31669         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
31670         for -mno-speculate-indirect-jumps.
31671         (*call_indirect_elfv2<mode>_nospec): New define_insn.
31672         (*call_value_indirect_elfv2<mode>): Disable for
31673         -mno-speculate-indirect-jumps.
31674         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
31675         (indirect_jump): Emit different RTL for
31676         -mno-speculate-indirect-jumps.
31677         (*indirect_jump<mode>): Disable for
31678         -mno-speculate-indirect-jumps.
31679         (*indirect_jump<mode>_nospec): New define_insn.
31680         (tablejump): Emit different RTL for
31681         -mno-speculate-indirect-jumps.
31682         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
31683         (tablejumpsi_nospec): New define_expand.
31684         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
31685         (tablejumpdi_nospec): New define_expand.
31686         (*tablejump<mode>_internal1): Disable for
31687         -mno-speculate-indirect-jumps.
31688         (*tablejump<mode>_internal1_nospec): New define_insn.
31689         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
31690         option.
31692 2018-01-16  Artyom Skrobov tyomitch@gmail.com
31694         * caller-save.c (insert_save): Drop unnecessary parameter.  All
31695         callers updated.
31697 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
31698             Richard Biener  <rguenth@suse.de>
31700         PR libgomp/83590
31701         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
31702         return early, inline manually is_gimple_sizepos.  Make sure if we
31703         call gimplify_expr we don't end up with a gimple constant.
31704         * tree.c (variably_modified_type_p): Don't return true for
31705         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
31706         * gimplify.h (is_gimple_sizepos): Remove.
31708 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
31710         PR tree-optimization/83857
31711         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
31712         vectorizable_live_operation for pure SLP statements.
31713         (vectorizable_live_operation): Handle PHIs.
31715 2018-01-16  Richard Biener  <rguenther@suse.de>
31717         PR tree-optimization/83867
31718         * tree-vect-stmts.c (vect_transform_stmt): Precompute
31719         nested_in_vect_loop_p since the scalar stmt may get invalidated.
31721 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
31723         PR c/83844
31724         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
31725         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
31726         If off is not INTEGER_CST, issue a may not be aligned warning
31727         rather than isn't aligned.  Use isn%'t rather than isn't.
31728         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
31729         into MULT_EXPR.
31730         <case MULT_EXPR>: Improve the case when bottom and one of the
31731         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
31732         operand, in that case check if the other operand is multiple of
31733         bottom divided by the INTEGER_CST operand.
31735 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
31737         PR target/83858
31738         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
31739         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
31740         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
31741         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
31742         * config/pa/pa.c (pa_function_arg_advance): Likewise.
31743         (pa_function_arg, pa_arg_partial_bytes): Likewise.
31744         (pa_function_arg_size): New function.
31746 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
31748         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
31749         in a separate statement.
31751 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
31753         PR tree-optimization/83847
31754         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
31755         group gathers and scatters.
31757 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
31759         PR rtl-optimization/86620
31760         * params.def (max-sched-ready-insns): Bump minimum value to 1.
31762         PR rtl-optimization/83213
31763         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
31764         to last if both are JUMP_INSNs.
31766         PR tree-optimization/83843
31767         * gimple-ssa-store-merging.c
31768         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
31769         store_immediate_info for bswap/nop orig_stores.
31771 2018-01-15  Andrew Waterman  <andrew@sifive.com>
31773         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
31774         !TARGET_MUL.
31775         <UDIV>: Increase cost if !TARGET_DIV.
31777 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
31779         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
31780         (define_attr "cr_logical_3op"): New.
31781         (cceq_ior_compare): Adjust.
31782         (cceq_ior_compare_complement): Adjust.
31783         (*cceq_rev_compare): Adjust.
31784         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
31785         (is_cracked_insn): Adjust.
31786         (insn_must_be_first_in_group): Adjust.
31787         * config/rs6000/40x.md: Adjust.
31788         * config/rs6000/440.md: Adjust.
31789         * config/rs6000/476.md: Adjust.
31790         * config/rs6000/601.md: Adjust.
31791         * config/rs6000/603.md: Adjust.
31792         * config/rs6000/6xx.md: Adjust.
31793         * config/rs6000/7450.md: Adjust.
31794         * config/rs6000/7xx.md: Adjust.
31795         * config/rs6000/8540.md: Adjust.
31796         * config/rs6000/cell.md: Adjust.
31797         * config/rs6000/e300c2c3.md: Adjust.
31798         * config/rs6000/e500mc.md: Adjust.
31799         * config/rs6000/e500mc64.md: Adjust.
31800         * config/rs6000/e5500.md: Adjust.
31801         * config/rs6000/e6500.md: Adjust.
31802         * config/rs6000/mpc.md: Adjust.
31803         * config/rs6000/power4.md: Adjust.
31804         * config/rs6000/power5.md: Adjust.
31805         * config/rs6000/power6.md: Adjust.
31806         * config/rs6000/power7.md: Adjust.
31807         * config/rs6000/power8.md: Adjust.
31808         * config/rs6000/power9.md: Adjust.
31809         * config/rs6000/rs64.md: Adjust.
31810         * config/rs6000/titan.md: Adjust.
31812 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31814         * config/i386/predicates.md (indirect_branch_operand): Rewrite
31815         ix86_indirect_branch_register logic.
31817 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31819         * config/i386/constraints.md (Bs): Update
31820         ix86_indirect_branch_register check.  Don't check
31821         ix86_indirect_branch_register with GOT_memory_operand.
31822         (Bw): Likewise.
31823         * config/i386/predicates.md (GOT_memory_operand): Don't check
31824         ix86_indirect_branch_register here.
31825         (GOT32_symbol_operand): Likewise.
31827 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31829         * config/i386/predicates.md (constant_call_address_operand):
31830         Rewrite ix86_indirect_branch_register logic.
31831         (sibcall_insn_operand): Likewise.
31833 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31835         * config/i386/constraints.md (Bs): Replace
31836         ix86_indirect_branch_thunk_register with
31837         ix86_indirect_branch_register.
31838         (Bw): Likewise.
31839         * config/i386/i386.md (indirect_jump): Likewise.
31840         (tablejump): Likewise.
31841         (*sibcall_memory): Likewise.
31842         (*sibcall_value_memory): Likewise.
31843         Peepholes of indirect call and jump via memory: Likewise.
31844         * config/i386/i386.opt: Likewise.
31845         * config/i386/predicates.md (indirect_branch_operand): Likewise.
31846         (GOT_memory_operand): Likewise.
31847         (call_insn_operand): Likewise.
31848         (sibcall_insn_operand): Likewise.
31849         (GOT32_symbol_operand): Likewise.
31851 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
31853         PR middle-end/83837
31854         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
31855         type rather than type addr's type points to.
31856         (expand_omp_atomic_mutex): Likewise.
31857         (expand_omp_atomic): Likewise.
31859 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31861         PR target/83839
31862         * config/i386/i386.c (output_indirect_thunk_function): Use
31863         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
31864         for  __x86_return_thunk.
31866 2018-01-15  Richard Biener  <rguenther@suse.de>
31868         PR middle-end/83850
31869         * expmed.c (extract_bit_field_1): Fix typo.
31871 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31873         PR target/83687
31874         * config/arm/iterators.md (VF): New mode iterator.
31875         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
31876         Remove integer-related logic from pattern.
31877         (neon_vabd<mode>_3): Likewise.
31879 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
31881         PR middle-end/82694
31882         * common.opt (fstrict-overflow): No longer an alias.
31883         (fwrapv-pointer): New option.
31884         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
31885         also for pointer types based on flag_wrapv_pointer.
31886         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
31887         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
31888         opts->x_flag_wrapv got set.
31889         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
31890         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
31891         POINTER_TYPE_OVERFLOW_UNDEFINED.
31892         * match.pd: Likewise in address comparison pattern.
31893         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
31895 2018-01-15  Richard Biener  <rguenther@suse.de>
31897         PR lto/83804
31898         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
31899         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
31900         Reset type names to their identifier if their TYPE_DECL doesn't
31901         have linkage (and thus is used for ODR and devirt).
31902         (save_debug_info_for_decl): Remove.
31903         (save_debug_info_for_type): Likewise.
31904         (add_tree_to_fld_list): Adjust.
31905         * tree-pretty-print.c (dump_generic_node): Make dumping of
31906         type names more robust.
31908 2018-01-15  Richard Biener  <rguenther@suse.de>
31910         * BASE-VER: Bump to 8.0.1.
31912 2018-01-14  Martin Sebor  <msebor@redhat.com>
31914         PR other/83508
31915         * builtins.c (check_access): Avoid warning when the no-warning bit
31916         is set.
31918 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
31920         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
31921         * ira-color (allocno_hard_regs_compare): Likewise.
31923 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
31925         PR target/83013
31926         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
31927         Use .pushsection/.popsection.
31929 2018-01-14  Martin Sebor  <msebor@redhat.com>
31931         PR c++/81327
31932         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
31934 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
31936         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
31937         entry from extra_headers.
31938         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
31939         extra_headers, make the list bitwise identical to the i?86-*-* one.
31941 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31943         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
31944         -mcmodel=large with -mindirect-branch=thunk,
31945         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
31946         -mfunction-return=thunk-extern.
31947         * doc/invoke.texi: Document -mcmodel=large is incompatible with
31948         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
31949         -mfunction-return=thunk and -mfunction-return=thunk-extern.
31951 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31953         * config/i386/i386.c (print_reg): Print the name of the full
31954         integer register without '%'.
31955         (ix86_print_operand): Handle 'V'.
31956         * doc/extend.texi: Document 'V' modifier.
31958 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31960         * config/i386/constraints.md (Bs): Disallow memory operand for
31961         -mindirect-branch-register.
31962         (Bw): Likewise.
31963         * config/i386/predicates.md (indirect_branch_operand): Likewise.
31964         (GOT_memory_operand): Likewise.
31965         (call_insn_operand): Likewise.
31966         (sibcall_insn_operand): Likewise.
31967         (GOT32_symbol_operand): Likewise.
31968         * config/i386/i386.md (indirect_jump): Call convert_memory_address
31969         for -mindirect-branch-register.
31970         (tablejump): Likewise.
31971         (*sibcall_memory): Likewise.
31972         (*sibcall_value_memory): Likewise.
31973         Disallow peepholes of indirect call and jump via memory for
31974         -mindirect-branch-register.
31975         (*call_pop): Replace m with Bw.
31976         (*call_value_pop): Likewise.
31977         (*sibcall_pop_memory): Replace m with Bs.
31978         * config/i386/i386.opt (mindirect-branch-register): New option.
31979         * doc/invoke.texi: Document -mindirect-branch-register option.
31981 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31983         * config/i386/i386-protos.h (ix86_output_function_return): New.
31984         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
31985         set function_return_type.
31986         (indirect_thunk_name): Add ret_p to indicate thunk for function
31987         return.
31988         (output_indirect_thunk_function): Pass false to
31989         indirect_thunk_name.
31990         (ix86_output_indirect_branch_via_reg): Likewise.
31991         (ix86_output_indirect_branch_via_push): Likewise.
31992         (output_indirect_thunk_function): Create alias for function
31993         return thunk if regno < 0.
31994         (ix86_output_function_return): New function.
31995         (ix86_handle_fndecl_attribute): Handle function_return.
31996         (ix86_attribute_table): Add function_return.
31997         * config/i386/i386.h (machine_function): Add
31998         function_return_type.
31999         * config/i386/i386.md (simple_return_internal): Use
32000         ix86_output_function_return.
32001         (simple_return_internal_long): Likewise.
32002         * config/i386/i386.opt (mfunction-return=): New option.
32003         (indirect_branch): Mention -mfunction-return=.
32004         * doc/extend.texi: Document function_return function attribute.
32005         * doc/invoke.texi: Document -mfunction-return= option.
32007 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
32009         * config/i386/i386-opts.h (indirect_branch): New.
32010         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
32011         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
32012         with local indirect jump when converting indirect call and jump.
32013         (ix86_set_indirect_branch_type): New.
32014         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
32015         (indirectlabelno): New.
32016         (indirect_thunk_needed): Likewise.
32017         (indirect_thunk_bnd_needed): Likewise.
32018         (indirect_thunks_used): Likewise.
32019         (indirect_thunks_bnd_used): Likewise.
32020         (INDIRECT_LABEL): Likewise.
32021         (indirect_thunk_name): Likewise.
32022         (output_indirect_thunk): Likewise.
32023         (output_indirect_thunk_function): Likewise.
32024         (ix86_output_indirect_branch_via_reg): Likewise.
32025         (ix86_output_indirect_branch_via_push): Likewise.
32026         (ix86_output_indirect_branch): Likewise.
32027         (ix86_output_indirect_jmp): Likewise.
32028         (ix86_code_end): Call output_indirect_thunk_function if needed.
32029         (ix86_output_call_insn): Call ix86_output_indirect_branch if
32030         needed.
32031         (ix86_handle_fndecl_attribute): Handle indirect_branch.
32032         (ix86_attribute_table): Add indirect_branch.
32033         * config/i386/i386.h (machine_function): Add indirect_branch_type
32034         and has_local_indirect_jump.
32035         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
32036         to true.
32037         (tablejump): Likewise.
32038         (*indirect_jump): Use ix86_output_indirect_jmp.
32039         (*tablejump_1): Likewise.
32040         (simple_return_indirect_internal): Likewise.
32041         * config/i386/i386.opt (mindirect-branch=): New option.
32042         (indirect_branch): New.
32043         (keep): Likewise.
32044         (thunk): Likewise.
32045         (thunk-inline): Likewise.
32046         (thunk-extern): Likewise.
32047         * doc/extend.texi: Document indirect_branch function attribute.
32048         * doc/invoke.texi: Document -mindirect-branch= option.
32050 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
32052         PR ipa/83051
32053         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
32055 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
32057         * ipa-inline.c (want_inline_small_function_p): Return false if
32058         inlining has already failed with CIF_FINAL_ERROR.
32059         (update_caller_keys): Call want_inline_small_function_p before
32060         can_inline_edge_p.
32061         (update_callee_keys): Likewise.
32063 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32065         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
32066         New function.
32067         (rs6000_quadword_masked_address_p): Likewise.
32068         (quad_aligned_load_p): Likewise.
32069         (quad_aligned_store_p): Likewise.
32070         (const_load_sequence_p): Add comment to describe the outer-most loop.
32071         (mimic_memory_attributes_and_flags): New function.
32072         (rs6000_gen_stvx): Likewise.
32073         (replace_swapped_aligned_store): Likewise.
32074         (rs6000_gen_lvx): Likewise.
32075         (replace_swapped_aligned_load): Likewise.
32076         (replace_swapped_load_constant): Capitalize argument name in
32077         comment describing this function.
32078         (rs6000_analyze_swaps): Add a third pass to search for vector loads
32079         and stores that access quad-word aligned addresses and replace
32080         with stvx or lvx instructions when appropriate.
32081         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
32082         New function prototype.
32083         (rs6000_quadword_masked_address_p): Likewise.
32084         (rs6000_gen_lvx): Likewise.
32085         (rs6000_gen_stvx): Likewise.
32086         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
32087         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
32088         when memory address is aligned.
32089         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
32090         this split to select lvx instruction when memory address is aligned.
32091         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
32092         instruction when memory address is aligned.
32093         (*vsx_le_perm_load_v16qi): Likewise.
32094         (four unnamed splitters): Modify to select the stvx instruction
32095         when memory is aligned.
32097 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
32099         * predict.c (determine_unlikely_bbs): Handle correctly BBs
32100         which appears in the queue multiple times.
32102 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32103             Alan Hayward  <alan.hayward@arm.com>
32104             David Sherwood  <david.sherwood@arm.com>
32106         * tree-vectorizer.h (vec_lower_bound): New structure.
32107         (_loop_vec_info): Add check_nonzero and lower_bounds.
32108         (LOOP_VINFO_CHECK_NONZERO): New macro.
32109         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
32110         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
32111         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
32112         fields.  Make seg_len the distance travelled, not including the
32113         access size.
32114         (dr_direction_indicator): Declare.
32115         (dr_zero_step_indicator): Likewise.
32116         (dr_known_forward_stride_p): Likewise.
32117         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
32118         tree-ssanames.h.
32119         (runtime_alias_check_p): Allow runtime alias checks with
32120         variable strides.
32121         (operator ==): Compare access_size and align.
32122         (prune_runtime_alias_test_list): Rework for new distinction between
32123         the access_size and seg_len.
32124         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
32125         segment lengths.
32126         (get_segment_min_max): New function.
32127         (create_intersect_range_checks): Use it.
32128         (dr_step_indicator): New function.
32129         (dr_direction_indicator): Likewise.
32130         (dr_zero_step_indicator): Likewise.
32131         (dr_known_forward_stride_p): Likewise.
32132         * tree-loop-distribution.c (data_ref_segment_size): Return
32133         DR_STEP * (niters - 1).
32134         (compute_alias_check_pairs): Update call to the dr_with_seg_len
32135         constructor.
32136         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
32137         (vect_preserves_scalar_order_p): New function, split out from...
32138         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
32139         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
32140         (vect_vfa_access_size): New function.
32141         (vect_vfa_align): Likewise.
32142         (vect_compile_time_alias): Take access_size_a and access_b arguments.
32143         (dump_lower_bound): New function.
32144         (vect_check_lower_bound): Likewise.
32145         (vect_small_gap_p): Likewise.
32146         (vectorizable_with_step_bound_p): Likewise.
32147         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
32148         depencies if the vectorization factor is 1.  Convert the checks
32149         for nonzero steps into checks on the bounds of DR_STEP.  Try using
32150         a bunds check for variable steps if the minimum required step is
32151         relatively small. Update calls to the dr_with_seg_len
32152         constructor and to vect_compile_time_alias.
32153         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
32154         function.
32155         (vect_loop_versioning): Call it.
32156         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
32157         when retrying.
32158         (vect_estimate_min_profitable_iters): Account for any bounds checks.
32160 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32161             Alan Hayward  <alan.hayward@arm.com>
32162             David Sherwood  <david.sherwood@arm.com>
32164         * doc/sourcebuild.texi (vect_scatter_store): Document.
32165         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
32166         optabs.
32167         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
32168         Document.
32169         * genopinit.c (main): Add supports_vec_scatter_store and
32170         supports_vec_scatter_store_cached to target_optabs.
32171         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
32172         IFN_MASK_SCATTER_STORE.
32173         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
32174         functions.
32175         * internal-fn.h (internal_store_fn_p): Declare.
32176         (internal_fn_stored_value_index): Likewise.
32177         * internal-fn.c (scatter_store_direct): New macro.
32178         (expand_scatter_store_optab_fn): New function.
32179         (direct_scatter_store_optab_supported_p): New macro.
32180         (internal_store_fn_p): New function.
32181         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
32182         IFN_MASK_SCATTER_STORE.
32183         (internal_fn_mask_index): Likewise.
32184         (internal_fn_stored_value_index): New function.
32185         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
32186         for scatter stores.
32187         * optabs-query.h (supports_vec_scatter_store_p): Declare.
32188         * optabs-query.c (supports_vec_scatter_store_p): New function.
32189         * tree-vectorizer.h (vect_get_store_rhs): Declare.
32190         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
32191         true for scatter stores.
32192         (vect_gather_scatter_fn_p): Handle scatter stores too.
32193         (vect_check_gather_scatter): Consider using scatter stores if
32194         supports_vec_scatter_store_p.
32195         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
32196         scatter stores too.
32197         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
32198         internal_fn_stored_value_index.
32199         (check_load_store_masking): Handle scatter stores too.
32200         (vect_get_store_rhs): Make public.
32201         (vectorizable_call): Use internal_store_fn_p.
32202         (vectorizable_store): Handle scatter store internal functions.
32203         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
32204         when deciding whether the end of the group has been reached.
32205         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
32206         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
32207         (mask_scatter_store<mode>): New insns.
32209 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32210             Alan Hayward  <alan.hayward@arm.com>
32211             David Sherwood  <david.sherwood@arm.com>
32213         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
32214         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
32215         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
32216         function.
32217         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
32218         Use vect_truncate_gather_scatter_offset if we can't treat the
32219         operation as a normal gather load or scatter store.
32220         (get_group_load_store_type): Take the gather_scatter_info
32221         as argument.  Try using a gather load or scatter store for
32222         single-element groups.
32223         (get_load_store_type): Update calls to get_group_load_store_type
32224         and vect_use_strided_gather_scatters_p.
32226 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32227             Alan Hayward  <alan.hayward@arm.com>
32228             David Sherwood  <david.sherwood@arm.com>
32230         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
32231         optional tree argument.
32232         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
32233         null target hooks.
32234         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
32235         but continue to use the current value as a fallback.
32236         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
32237         to compare the updates.
32238         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
32239         (get_load_store_type): Use it when handling a strided access.
32240         (vect_get_strided_load_store_ops): New function.
32241         (vect_get_data_ptr_increment): Likewise.
32242         (vectorizable_load): Handle strided gather loads.  Always pass
32243         a step to vect_create_data_ref_ptr and bump_vector_ptr.
32245 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32246             Alan Hayward  <alan.hayward@arm.com>
32247             David Sherwood  <david.sherwood@arm.com>
32249         * doc/md.texi (gather_load@var{m}): Document.
32250         (mask_gather_load@var{m}): Likewise.
32251         * genopinit.c (main): Add supports_vec_gather_load and
32252         supports_vec_gather_load_cached to target_optabs.
32253         * optabs-tree.c (init_tree_optimization_optabs): Use
32254         ggc_cleared_alloc to allocate target_optabs.
32255         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
32256         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
32257         functions.
32258         * internal-fn.h (internal_load_fn_p): Declare.
32259         (internal_gather_scatter_fn_p): Likewise.
32260         (internal_fn_mask_index): Likewise.
32261         (internal_gather_scatter_fn_supported_p): Likewise.
32262         * internal-fn.c (gather_load_direct): New macro.
32263         (expand_gather_load_optab_fn): New function.
32264         (direct_gather_load_optab_supported_p): New macro.
32265         (direct_internal_fn_optab): New function.
32266         (internal_load_fn_p): Likewise.
32267         (internal_gather_scatter_fn_p): Likewise.
32268         (internal_fn_mask_index): Likewise.
32269         (internal_gather_scatter_fn_supported_p): Likewise.
32270         * optabs-query.c (supports_at_least_one_mode_p): New function.
32271         (supports_vec_gather_load_p): Likewise.
32272         * optabs-query.h (supports_vec_gather_load_p): Declare.
32273         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
32274         and memory_type field.
32275         (NUM_PATTERNS): Bump to 15.
32276         * tree-vect-data-refs.c: Include internal-fn.h.
32277         (vect_gather_scatter_fn_p): New function.
32278         (vect_describe_gather_scatter_call): Likewise.
32279         (vect_check_gather_scatter): Try using internal functions for
32280         gather loads.  Recognize existing calls to a gather load function.
32281         (vect_analyze_data_refs): Consider using gather loads if
32282         supports_vec_gather_load_p.
32283         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
32284         (vect_get_gather_scatter_offset_type): Likewise.
32285         (vect_convert_mask_for_vectype): Likewise.
32286         (vect_add_conversion_to_patterm): Likewise.
32287         (vect_try_gather_scatter_pattern): Likewise.
32288         (vect_recog_gather_scatter_pattern): New pattern recognizer.
32289         (vect_vect_recog_func_ptrs): Add it.
32290         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
32291         internal_fn_mask_index and internal_gather_scatter_fn_p.
32292         (check_load_store_masking): Take the gather_scatter_info as an
32293         argument and handle gather loads.
32294         (vect_get_gather_scatter_ops): New function.
32295         (vectorizable_call): Check internal_load_fn_p.
32296         (vectorizable_load): Likewise.  Handle gather load internal
32297         functions.
32298         (vectorizable_store): Update call to check_load_store_masking.
32299         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
32300         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
32301         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
32302         (aarch64_gather_scale_operand_d): New predicates.
32303         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
32304         (mask_gather_load<mode>): New insns.
32306 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32307             Alan Hayward  <alan.hayward@arm.com>
32308             David Sherwood  <david.sherwood@arm.com>
32310         * optabs.def (fold_left_plus_optab): New optab.
32311         * doc/md.texi (fold_left_plus_@var{m}): Document.
32312         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
32313         * internal-fn.c (fold_left_direct): Define.
32314         (expand_fold_left_optab_fn): Likewise.
32315         (direct_fold_left_optab_supported_p): Likewise.
32316         * fold-const-call.c (fold_const_fold_left): New function.
32317         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
32318         * tree-parloops.c (valid_reduction_p): New function.
32319         (gather_scalar_reductions): Use it.
32320         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
32321         (vect_finish_replace_stmt): Declare.
32322         * tree-vect-loop.c (fold_left_reduction_fn): New function.
32323         (needs_fold_left_reduction_p): New function, split out from...
32324         (vect_is_simple_reduction): ...here.  Accept reductions that
32325         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
32326         (vect_force_simple_reduction): Also store the reduction type in
32327         the assignment's STMT_VINFO_REDUC_TYPE.
32328         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
32329         (merge_with_identity): New function.
32330         (vect_expand_fold_left): Likewise.
32331         (vectorize_fold_left_reduction): Likewise.
32332         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
32333         scalar phi in place for it.  Check for target support and reject
32334         cases that would reassociate the operation.  Defer the transform
32335         phase to vectorize_fold_left_reduction.
32336         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
32337         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
32338         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
32340 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32342         * tree-if-conv.c (predicate_mem_writes): Remove redundant
32343         call to ifc_temp_var.
32345 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32346             Alan Hayward  <alan.hayward@arm.com>
32347             David Sherwood  <david.sherwood@arm.com>
32349         * target.def (legitimize_address_displacement): Take the original
32350         offset as a poly_int.
32351         * targhooks.h (default_legitimize_address_displacement): Update
32352         accordingly.
32353         * targhooks.c (default_legitimize_address_displacement): Likewise.
32354         * doc/tm.texi: Regenerate.
32355         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
32356         as an argument, moving assert of ad->disp == ad->disp_term to...
32357         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
32358         Try calling targetm.legitimize_address_displacement before expanding
32359         the address rather than afterwards, and adjust for the new interface.
32360         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
32361         Match the new hook interface.  Handle SVE addresses.
32362         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
32363         new hook interface.
32365 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32367         * Makefile.in (OBJS): Add early-remat.o.
32368         * target.def (select_early_remat_modes): New hook.
32369         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
32370         * doc/tm.texi: Regenerate.
32371         * targhooks.h (default_select_early_remat_modes): Declare.
32372         * targhooks.c (default_select_early_remat_modes): New function.
32373         * timevar.def (TV_EARLY_REMAT): New timevar.
32374         * passes.def (pass_early_remat): New pass.
32375         * tree-pass.h (make_pass_early_remat): Declare.
32376         * early-remat.c: New file.
32377         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
32378         function.
32379         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
32381 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32382             Alan Hayward  <alan.hayward@arm.com>
32383             David Sherwood  <david.sherwood@arm.com>
32385         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
32386         vfm1 with a bound_epilog parameter.
32387         (vect_do_peeling): Update calls accordingly, and move the prologue
32388         call earlier in the function.  Treat the base bound_epilog as 0 for
32389         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
32390         this base when peeling for gaps.
32391         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
32392         with fully-masked loops.
32393         (vect_estimate_min_profitable_iters): Handle the single peeled
32394         iteration in that case.
32396 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32397             Alan Hayward  <alan.hayward@arm.com>
32398             David Sherwood  <david.sherwood@arm.com>
32400         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
32401         single-element interleaving even if the size is not a power of 2.
32402         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
32403         accesses for single-element interleaving if the group size is
32404         not a power of 2.
32406 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32407             Alan Hayward  <alan.hayward@arm.com>
32408             David Sherwood  <david.sherwood@arm.com>
32410         * doc/md.texi (fold_extract_last_@var{m}): Document.
32411         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
32412         * optabs.def (fold_extract_last_optab): New optab.
32413         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
32414         * internal-fn.c (fold_extract_direct): New macro.
32415         (expand_fold_extract_optab_fn): Likewise.
32416         (direct_fold_extract_optab_supported_p): Likewise.
32417         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
32418         * tree-vect-loop.c (vect_model_reduction_cost): Handle
32419         EXTRACT_LAST_REDUCTION.
32420         (get_initial_def_for_reduction): Do not create an initial vector
32421         for EXTRACT_LAST_REDUCTION reductions.
32422         (vectorizable_reduction): Leave the scalar phi in place for
32423         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
32424         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
32425         epilogue code for EXTRACT_LAST_REDUCTION and defer the
32426         transform phase to vectorizable_condition.
32427         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
32428         split out from...
32429         (vect_finish_stmt_generation): ...here.
32430         (vect_finish_replace_stmt): New function.
32431         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
32432         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
32433         pattern.
32434         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
32436 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32437             Alan Hayward  <alan.hayward@arm.com>
32438             David Sherwood  <david.sherwood@arm.com>
32440         * doc/md.texi (extract_last_@var{m}): Document.
32441         * optabs.def (extract_last_optab): New optab.
32442         * internal-fn.def (EXTRACT_LAST): New internal function.
32443         * internal-fn.c (cond_unary_direct): New macro.
32444         (expand_cond_unary_optab_fn): Likewise.
32445         (direct_cond_unary_optab_supported_p): Likewise.
32446         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
32447         loops using EXTRACT_LAST.
32448         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
32449         (extract_last_<mode>): ...this optab.
32450         (vec_extract<mode><Vel>): Update accordingly.
32452 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32453             Alan Hayward  <alan.hayward@arm.com>
32454             David Sherwood  <david.sherwood@arm.com>
32456         * target.def (empty_mask_is_expensive): New hook.
32457         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
32458         * doc/tm.texi: Regenerate.
32459         * targhooks.h (default_empty_mask_is_expensive): Declare.
32460         * targhooks.c (default_empty_mask_is_expensive): New function.
32461         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
32462         if the target says that empty masks are expensive.
32463         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
32464         New function.
32465         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
32467 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32468             Alan Hayward  <alan.hayward@arm.com>
32469             David Sherwood  <david.sherwood@arm.com>
32471         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
32472         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
32473         (vect_use_loop_mask_for_alignment_p): New function.
32474         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
32475         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
32476         niters_skip argument.  Make sure that the first niters_skip elements
32477         of the first iteration are inactive.
32478         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
32479         Update call to vect_set_loop_masks_directly.
32480         (get_misalign_in_elems): New function, split out from...
32481         (vect_gen_prolog_loop_niters): ...here.
32482         (vect_update_init_of_dr): Take a code argument that specifies whether
32483         the adjustment should be added or subtracted.
32484         (vect_update_init_of_drs): Likewise.
32485         (vect_prepare_for_masked_peels): New function.
32486         (vect_do_peeling): Skip prologue peeling if we're using a mask
32487         instead.  Update call to vect_update_inits_of_drs.
32488         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32489         mask_skip_niters.
32490         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
32491         alignment.  Do not include the number of peeled iterations in
32492         the minimum threshold in that case.
32493         (vectorizable_induction): Adjust the start value down by
32494         LOOP_VINFO_MASK_SKIP_NITERS iterations.
32495         (vect_transform_loop): Call vect_prepare_for_masked_peels.
32496         Take the number of skipped iterations into account when calculating
32497         the loop bounds.
32498         * tree-vect-stmts.c (vect_gen_while_not): New function.
32500 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32501             Alan Hayward  <alan.hayward@arm.com>
32502             David Sherwood  <david.sherwood@arm.com>
32504         * doc/sourcebuild.texi (vect_fully_masked): Document.
32505         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
32506         default value to 0.
32507         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
32508         split out from...
32509         (vect_analyze_loop_2): ...here. Don't check the vectorization
32510         factor against the number of loop iterations if the loop is
32511         fully-masked.
32513 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32514             Alan Hayward  <alan.hayward@arm.com>
32515             David Sherwood  <david.sherwood@arm.com>
32517         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
32518         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
32519         (dump_groups): Update accordingly.
32520         (iv_use::mem_type): New member variable.
32521         (address_p): New function.
32522         (record_use): Add a mem_type argument and initialize the new
32523         mem_type field.
32524         (record_group_use): Add a mem_type argument.  Use address_p.
32525         Remove obsolete null checks of base_object.  Update call to record_use.
32526         (find_interesting_uses_op): Update call to record_group_use.
32527         (find_interesting_uses_cond): Likewise.
32528         (find_interesting_uses_address): Likewise.
32529         (get_mem_type_for_internal_fn): New function.
32530         (find_address_like_use): Likewise.
32531         (find_interesting_uses_stmt): Try find_address_like_use before
32532         calling find_interesting_uses_op.
32533         (addr_offset_valid_p): Use the iv mem_type field as the type
32534         of the addressed memory.
32535         (add_autoinc_candidates): Likewise.
32536         (get_address_cost): Likewise.
32537         (split_small_address_groups_p): Use address_p.
32538         (split_address_groups): Likewise.
32539         (add_iv_candidate_for_use): Likewise.
32540         (autoinc_possible_for_pair): Likewise.
32541         (rewrite_groups): Likewise.
32542         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
32543         (determine_group_iv_cost): Update after split of USE_ADDRESS.
32544         (get_alias_ptr_type_for_ptr_address): New function.
32545         (rewrite_use_address): Rewrite address uses in calls that were
32546         identified by find_address_like_use.
32548 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32549             Alan Hayward  <alan.hayward@arm.com>
32550             David Sherwood  <david.sherwood@arm.com>
32552         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
32553         TARGET_MEM_REFs.
32554         * gimple-expr.h (is_gimple_addressable: Likewise.
32555         * gimple-expr.c (is_gimple_address): Likewise.
32556         * internal-fn.c (expand_call_mem_ref): New function.
32557         (expand_mask_load_optab_fn): Use it.
32558         (expand_mask_store_optab_fn): Likewise.
32560 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32561             Alan Hayward  <alan.hayward@arm.com>
32562             David Sherwood  <david.sherwood@arm.com>
32564         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
32565         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
32566         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
32567         (cond_umax@var{mode}): Document.
32568         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
32569         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
32570         (cond_umin_optab, cond_umax_optab): New optabs.
32571         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
32572         (COND_IOR, COND_XOR): New internal functions.
32573         * internal-fn.h (get_conditional_internal_fn): Declare.
32574         * internal-fn.c (cond_binary_direct): New macro.
32575         (expand_cond_binary_optab_fn): Likewise.
32576         (direct_cond_binary_optab_supported_p): Likewise.
32577         (get_conditional_internal_fn): New function.
32578         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
32579         Cope with reduction statements that are vectorized as calls rather
32580         than assignments.
32581         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
32582         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
32583         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
32584         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
32585         (UNSPEC_COND_EOR): New unspecs.
32586         (optab): Add mappings for them.
32587         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
32588         (sve_int_op, sve_fp_op): New int attributes.
32590 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32591             Alan Hayward  <alan.hayward@arm.com>
32592             David Sherwood  <david.sherwood@arm.com>
32594         * optabs.def (while_ult_optab): New optab.
32595         * doc/md.texi (while_ult@var{m}@var{n}): Document.
32596         * internal-fn.def (WHILE_ULT): New internal function.
32597         * internal-fn.h (direct_internal_fn_supported_p): New override
32598         that takes two types as argument.
32599         * internal-fn.c (while_direct): New macro.
32600         (expand_while_optab_fn): New function.
32601         (convert_optab_supported_p): Likewise.
32602         (direct_while_optab_supported_p): New macro.
32603         * wide-int.h (wi::udiv_ceil): New function.
32604         * tree-vectorizer.h (rgroup_masks): New structure.
32605         (vec_loop_masks): New typedef.
32606         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
32607         and fully_masked_p.
32608         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
32609         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
32610         (vect_max_vf): New function.
32611         (slpeel_make_loop_iterate_ntimes): Delete.
32612         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
32613         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
32614         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
32615         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
32616         internal-fn.h, stor-layout.h and optabs-query.h.
32617         (vect_set_loop_mask): New function.
32618         (add_preheader_seq): Likewise.
32619         (add_header_seq): Likewise.
32620         (interleave_supported_p): Likewise.
32621         (vect_maybe_permute_loop_masks): Likewise.
32622         (vect_set_loop_masks_directly): Likewise.
32623         (vect_set_loop_condition_masked): Likewise.
32624         (vect_set_loop_condition_unmasked): New function, split out from
32625         slpeel_make_loop_iterate_ntimes.
32626         (slpeel_make_loop_iterate_ntimes): Rename to..
32627         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
32628         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
32629         (vect_do_peeling): Update call accordingly.
32630         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
32631         loops.
32632         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32633         mask_compare_type, can_fully_mask_p and fully_masked_p.
32634         (release_vec_loop_masks): New function.
32635         (_loop_vec_info): Use it to free the loop masks.
32636         (can_produce_all_loop_masks_p): New function.
32637         (vect_get_max_nscalars_per_iter): Likewise.
32638         (vect_verify_full_masking): Likewise.
32639         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
32640         retries, and free the mask rgroups before retrying.  Check loop-wide
32641         reasons for disallowing fully-masked loops.  Make the final decision
32642         about whether use a fully-masked loop or not.
32643         (vect_estimate_min_profitable_iters): Do not assume that peeling
32644         for the number of iterations will be needed for fully-masked loops.
32645         (vectorizable_reduction): Disable fully-masked loops.
32646         (vectorizable_live_operation): Likewise.
32647         (vect_halve_mask_nunits): New function.
32648         (vect_double_mask_nunits): Likewise.
32649         (vect_record_loop_mask): Likewise.
32650         (vect_get_loop_mask): Likewise.
32651         (vect_transform_loop): Handle the case in which the final loop
32652         iteration might handle a partial vector.  Call vect_set_loop_condition
32653         instead of slpeel_make_loop_iterate_ntimes.
32654         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
32655         (check_load_store_masking): New function.
32656         (prepare_load_store_mask): Likewise.
32657         (vectorizable_store): Handle fully-masked loops.
32658         (vectorizable_load): Likewise.
32659         (supportable_widening_operation): Use vect_halve_mask_nunits for
32660         booleans.
32661         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
32662         (vect_gen_while): New function.
32663         * config/aarch64/aarch64.md (umax<mode>3): New expander.
32664         (aarch64_uqdec<mode>): New insn.
32666 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32667             Alan Hayward  <alan.hayward@arm.com>
32668             David Sherwood  <david.sherwood@arm.com>
32670         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
32671         (reduc_xor_scal_optab): New optabs.
32672         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
32673         (reduc_xor_scal_@var{m}): Document.
32674         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
32675         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
32676         internal functions.
32677         * fold-const-call.c (fold_const_call): Handle them.
32678         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
32679         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
32680         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
32681         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
32682         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
32683         (UNSPEC_XORV): New unspecs.
32684         (optab): Add entries for them.
32685         (BITWISEV): New int iterator.
32686         (bit_reduc_op): New int attributes.
32688 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32689             Alan Hayward  <alan.hayward@arm.com>
32690             David Sherwood  <david.sherwood@arm.com>
32692         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
32693         * internal-fn.def (VEC_SHL_INSERT): New internal function.
32694         * optabs.def (vec_shl_insert_optab): New optab.
32695         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
32696         (duplicate_and_interleave): Likewise.
32697         * tree-vect-loop.c: Include internal-fn.h.
32698         (neutral_op_for_slp_reduction): New function, split out from
32699         get_initial_defs_for_reduction.
32700         (get_initial_def_for_reduction): Handle option 2 for variable-length
32701         vectors by loading the neutral value into a vector and then shifting
32702         the initial value into element 0.
32703         (get_initial_defs_for_reduction): Replace the code argument with
32704         the neutral value calculated by neutral_op_for_slp_reduction.
32705         Use gimple_build_vector for constant-length vectors.
32706         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
32707         but the first group_size elements have a neutral value.
32708         Use duplicate_and_interleave otherwise.
32709         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
32710         Update call to get_initial_defs_for_reduction.  Handle SLP
32711         reductions for variable-length vectors by creating one vector
32712         result for each scalar result, with the elements associated
32713         with other scalar results stubbed out with the neutral value.
32714         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
32715         Require IFN_VEC_SHL_INSERT for double reductions on
32716         variable-length vectors, or SLP reductions that have
32717         a neutral value.  Require can_duplicate_and_interleave_p
32718         support for variable-length unchained SLP reductions if there
32719         is no neutral value, such as for MIN/MAX reductions.  Also require
32720         the number of vector elements to be a multiple of the number of
32721         SLP statements when doing variable-length unchained SLP reductions.
32722         Update call to vect_create_epilog_for_reduction.
32723         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
32724         and remove initial values.
32725         (duplicate_and_interleave): Make public.
32726         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
32727         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
32729 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32730             Alan Hayward  <alan.hayward@arm.com>
32731             David Sherwood  <david.sherwood@arm.com>
32733         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
32734         (can_duplicate_and_interleave_p): New function.
32735         (vect_get_and_check_slp_defs): Take the vector of statements
32736         rather than just the current one.  Remove excess parentheses.
32737         Restriction rejectinon of vect_constant_def and vect_external_def
32738         for variable-length vectors to boolean types, or types for which
32739         can_duplicate_and_interleave_p is false.
32740         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
32741         (duplicate_and_interleave): New function.
32742         (vect_get_constant_vectors): Use gimple_build_vector for
32743         constant-length vectors and suitable variable-length constant
32744         vectors.  Use duplicate_and_interleave for other variable-length
32745         vectors.  Don't defer the update when inserting new statements.
32747 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32748             Alan Hayward  <alan.hayward@arm.com>
32749             David Sherwood  <david.sherwood@arm.com>
32751         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
32752         min_profitable_iters doesn't go negative.
32754 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32755             Alan Hayward  <alan.hayward@arm.com>
32756             David Sherwood  <david.sherwood@arm.com>
32758         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
32759         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
32760         * optabs.def (vec_mask_load_lanes_optab): New optab.
32761         (vec_mask_store_lanes_optab): Likewise.
32762         * internal-fn.def (MASK_LOAD_LANES): New internal function.
32763         (MASK_STORE_LANES): Likewise.
32764         * internal-fn.c (mask_load_lanes_direct): New macro.
32765         (mask_store_lanes_direct): Likewise.
32766         (expand_mask_load_optab_fn): Handle masked operations.
32767         (expand_mask_load_lanes_optab_fn): New macro.
32768         (expand_mask_store_optab_fn): Handle masked operations.
32769         (expand_mask_store_lanes_optab_fn): New macro.
32770         (direct_mask_load_lanes_optab_supported_p): Likewise.
32771         (direct_mask_store_lanes_optab_supported_p): Likewise.
32772         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
32773         parameter.
32774         (vect_load_lanes_supported): Likewise.
32775         * tree-vect-data-refs.c (strip_conversion): New function.
32776         (can_group_stmts_p): Likewise.
32777         (vect_analyze_data_ref_accesses): Use it instead of checking
32778         for a pair of assignments.
32779         (vect_store_lanes_supported): Take a masked_p parameter.
32780         (vect_load_lanes_supported): Likewise.
32781         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
32782         vect_store_lanes_supported and vect_load_lanes_supported.
32783         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
32784         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
32785         parameter.  Don't allow gaps for masked accesses.
32786         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
32787         and vect_load_lanes_supported.
32788         (get_load_store_type): Take a masked_p parameter and update
32789         call to get_group_load_store_type.
32790         (vectorizable_store): Update call to get_load_store_type.
32791         Handle IFN_MASK_STORE_LANES.
32792         (vectorizable_load): Update call to get_load_store_type.
32793         Handle IFN_MASK_LOAD_LANES.
32795 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32796             Alan Hayward  <alan.hayward@arm.com>
32797             David Sherwood  <david.sherwood@arm.com>
32799         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
32800         modes for SVE.
32801         * config/aarch64/aarch64-protos.h
32802         (aarch64_sve_struct_memory_operand_p): Declare.
32803         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
32804         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
32805         (VPRED, vpred): Handle SVE structure modes.
32806         * config/aarch64/constraints.md (Utx): New constraint.
32807         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
32808         (aarch64_sve_struct_nonimmediate_operand): New predicates.
32809         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
32810         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
32811         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
32812         structure modes.  Split into pieces after RA.
32813         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
32814         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
32815         New patterns.
32816         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
32817         SVE structure modes.
32818         (aarch64_classify_address): Likewise.
32819         (sizetochar): Move earlier in file.
32820         (aarch64_print_operand): Handle SVE register lists.
32821         (aarch64_array_mode): New function.
32822         (aarch64_sve_struct_memory_operand_p): Likewise.
32823         (TARGET_ARRAY_MODE): Redefine.
32825 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32826             Alan Hayward  <alan.hayward@arm.com>
32827             David Sherwood  <david.sherwood@arm.com>
32829         * target.def (array_mode): New target hook.
32830         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
32831         * doc/tm.texi: Regenerate.
32832         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
32833         * hooks.c (hook_optmode_mode_uhwi_none): New function.
32834         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
32835         targetm.array_mode.
32836         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
32837         type sizes.
32839 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32840             Alan Hayward  <alan.hayward@arm.com>
32841             David Sherwood  <david.sherwood@arm.com>
32843         * fold-const.c (fold_binary_loc): Check the argument types
32844         rather than the result type when testing for a vector operation.
32846 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32848         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
32849         * doc/tm.texi: Regenerate.
32851 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32852             Alan Hayward  <alan.hayward@arm.com>
32853             David Sherwood  <david.sherwood@arm.com>
32855         * doc/invoke.texi (-msve-vector-bits=): Document new option.
32856         (sve): Document new AArch64 extension.
32857         * doc/md.texi (w): Extend the description of the AArch64
32858         constraint to include SVE vectors.
32859         (Upl, Upa): Document new AArch64 predicate constraints.
32860         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
32861         enum.
32862         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
32863         (msve-vector-bits=): New option.
32864         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
32865         SVE when these are disabled.
32866         (sve): New extension.
32867         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
32868         modes.  Adjust their number of units based on aarch64_sve_vg.
32869         (MAX_BITSIZE_MODE_ANY_MODE): Define.
32870         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
32871         aarch64_addr_query_type.
32872         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
32873         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
32874         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
32875         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
32876         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
32877         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
32878         (aarch64_simd_imm_zero_p): Delete.
32879         (aarch64_check_zero_based_sve_index_immediate): Declare.
32880         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32881         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32882         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32883         (aarch64_sve_float_mul_immediate_p): Likewise.
32884         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32885         rather than an rtx.
32886         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
32887         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
32888         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
32889         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
32890         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
32891         (aarch64_regmode_natural_size): Likewise.
32892         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
32893         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
32894         left one place.
32895         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
32896         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
32897         for VG and the SVE predicate registers.
32898         (V_ALIASES): Add a "z"-prefixed alias.
32899         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
32900         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
32901         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
32902         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
32903         (REG_CLASS_NAMES): Add entries for them.
32904         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
32905         and the predicate registers.
32906         (aarch64_sve_vg): Declare.
32907         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
32908         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
32909         (REGMODE_NATURAL_SIZE): Define.
32910         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
32911         SVE macros.
32912         * config/aarch64/aarch64.c: Include cfgrtl.h.
32913         (simd_immediate_info): Add a constructor for series vectors,
32914         and an associated step field.
32915         (aarch64_sve_vg): New variable.
32916         (aarch64_dbx_register_number): Handle VG and the predicate registers.
32917         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
32918         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
32919         (VEC_ANY_DATA, VEC_STRUCT): New constants.
32920         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
32921         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
32922         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
32923         (aarch64_get_mask_mode): New functions.
32924         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
32925         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
32926         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
32927         predicate modes and predicate registers.  Explicitly restrict
32928         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
32929         to store a vector mode if it is recognized by
32930         aarch64_classify_vector_mode.
32931         (aarch64_regmode_natural_size): New function.
32932         (aarch64_hard_regno_caller_save_mode): Return the original mode
32933         for predicates.
32934         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
32935         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
32936         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
32937         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
32938         functions.
32939         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
32940         does not overlap dest if the function is frame-related.  Handle
32941         SVE constants.
32942         (aarch64_split_add_offset): New function.
32943         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
32944         them aarch64_add_offset.
32945         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
32946         and update call to aarch64_sub_sp.
32947         (aarch64_add_cfa_expression): New function.
32948         (aarch64_expand_prologue): Pass extra temporary registers to the
32949         functions above.  Handle the case in which we need to emit new
32950         DW_CFA_expressions for registers that were originally saved
32951         relative to the stack pointer, but now have to be expressed
32952         relative to the frame pointer.
32953         (aarch64_output_mi_thunk): Pass extra temporary registers to the
32954         functions above.
32955         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
32956         IP0 and IP1 values for SVE frames.
32957         (aarch64_expand_vec_series): New function.
32958         (aarch64_expand_sve_widened_duplicate): Likewise.
32959         (aarch64_expand_sve_const_vector): Likewise.
32960         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
32961         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
32962         into the register, rather than emitting a SET directly.
32963         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
32964         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
32965         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
32966         (offset_9bit_signed_scaled_p): New functions.
32967         (aarch64_replicate_bitmask_imm): New function.
32968         (aarch64_bitmask_imm): Use it.
32969         (aarch64_cannot_force_const_mem): Reject expressions involving
32970         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
32971         (aarch64_classify_index): Handle SVE indices, by requiring
32972         a plain register index with a scale that matches the element size.
32973         (aarch64_classify_address): Handle SVE addresses.  Assert that
32974         the mode of the address is VOIDmode or an integer mode.
32975         Update call to aarch64_classify_symbol.
32976         (aarch64_classify_symbolic_expression): Update call to
32977         aarch64_classify_symbol.
32978         (aarch64_const_vec_all_in_range_p): New function.
32979         (aarch64_print_vector_float_operand): Likewise.
32980         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
32981         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
32982         and the FP immediates 1.0 and 0.5.
32983         (aarch64_print_address_internal): Handle SVE addresses.
32984         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
32985         (aarch64_regno_regclass): Handle predicate registers.
32986         (aarch64_secondary_reload): Handle big-endian reloads of SVE
32987         data modes.
32988         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
32989         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
32990         (aarch64_convert_sve_vector_bits): New function.
32991         (aarch64_override_options): Use it to handle -msve-vector-bits=.
32992         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32993         rather than an rtx.
32994         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
32995         Handle SVE vector and predicate modes.  Accept VL-based constants
32996         that need only one temporary register, and VL offsets that require
32997         no temporary registers.
32998         (aarch64_conditional_register_usage): Mark the predicate registers
32999         as fixed if SVE isn't available.
33000         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
33001         Return true for SVE vector and predicate modes.
33002         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
33003         rather than an unsigned int.  Handle SVE modes.
33004         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
33005         SVE modes.
33006         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
33007         if SVE is enabled.
33008         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
33009         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
33010         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
33011         (aarch64_sve_float_mul_immediate_p): New functions.
33012         (aarch64_sve_valid_immediate): New function.
33013         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
33014         Explicitly reject structure modes.  Check for INDEX constants.
33015         Handle PTRUE and PFALSE constants.
33016         (aarch64_check_zero_based_sve_index_immediate): New function.
33017         (aarch64_simd_imm_zero_p): Delete.
33018         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
33019         vector modes.  Accept constants in the range of CNT[BHWD].
33020         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
33021         ask for an Advanced SIMD mode.
33022         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
33023         (aarch64_simd_vector_alignment): Handle SVE predicates.
33024         (aarch64_vectorize_preferred_vector_alignment): New function.
33025         (aarch64_simd_vector_alignment_reachable): Use it instead of
33026         the vector size.
33027         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
33028         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
33029         functions.
33030         (MAX_VECT_LEN): Delete.
33031         (expand_vec_perm_d): Add a vec_flags field.
33032         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
33033         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
33034         (aarch64_evpc_ext): Don't apply a big-endian lane correction
33035         for SVE modes.
33036         (aarch64_evpc_rev): Rename to...
33037         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
33038         (aarch64_evpc_rev_global): New function.
33039         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
33040         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
33041         MAX_VECT_LEN.
33042         (aarch64_evpc_sve_tbl): New function.
33043         (aarch64_expand_vec_perm_const_1): Update after rename of
33044         aarch64_evpc_rev.  Handle SVE permutes too, trying
33045         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
33046         than aarch64_evpc_tbl.
33047         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
33048         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
33049         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
33050         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
33051         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
33052         (aarch64_expand_sve_vcond): New functions.
33053         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
33054         of aarch64_vector_mode_p.
33055         (aarch64_dwarf_poly_indeterminate_value): New function.
33056         (aarch64_compute_pressure_classes): Likewise.
33057         (aarch64_can_change_mode_class): Likewise.
33058         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
33059         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
33060         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
33061         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
33062         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
33063         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
33064         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
33065         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
33066         constraints.
33067         (Dn, Dl, Dr): Accept const as well as const_vector.
33068         (Dz): Likewise.  Compare against CONST0_RTX.
33069         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
33070         of "vector" where appropriate.
33071         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
33072         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
33073         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
33074         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
33075         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
33076         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
33077         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
33078         (v_int_equiv): Extend to SVE modes.
33079         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
33080         mode attributes.
33081         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
33082         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
33083         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
33084         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
33085         (SVE_COND_FP_CMP): New int iterators.
33086         (perm_hilo): Handle the new unpack unspecs.
33087         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
33088         attributes.
33089         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
33090         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
33091         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
33092         (aarch64_equality_operator, aarch64_constant_vector_operand)
33093         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
33094         (aarch64_sve_nonimmediate_operand): Likewise.
33095         (aarch64_sve_general_operand): Likewise.
33096         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
33097         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
33098         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
33099         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
33100         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
33101         (aarch64_sve_float_arith_immediate): Likewise.
33102         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
33103         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
33104         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
33105         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
33106         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
33107         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
33108         (aarch64_sve_float_arith_operand): Likewise.
33109         (aarch64_sve_float_arith_with_sub_operand): Likewise.
33110         (aarch64_sve_float_mul_operand): Likewise.
33111         (aarch64_sve_vec_perm_operand): Likewise.
33112         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
33113         (aarch64_mov_operand): Accept const_poly_int and const_vector.
33114         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
33115         as well as const_vector.
33116         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
33117         in file.  Use CONST0_RTX and CONSTM1_RTX.
33118         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
33119         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
33120         Use aarch64_simd_imm_zero.
33121         * config/aarch64/aarch64-sve.md: New file.
33122         * config/aarch64/aarch64.md: Include it.
33123         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
33124         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
33125         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
33126         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
33127         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
33128         (sve): New attribute.
33129         (enabled): Disable instructions with the sve attribute unless
33130         TARGET_SVE.
33131         (movqi, movhi): Pass CONST_POLY_INT operaneds through
33132         aarch64_expand_mov_immediate.
33133         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
33134         CNT[BHSD] immediates.
33135         (movti): Split CONST_POLY_INT moves into two halves.
33136         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
33137         Split additions that need a temporary here if the destination
33138         is the stack pointer.
33139         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
33140         (*add<mode>3_poly_1): New instruction.
33141         (set_clobber_cc): New expander.
33143 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
33145         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
33146         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
33147         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
33148         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
33149         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
33150         Change innermode from fixed_mode_size to machine_mode.
33151         (simplify_subreg): Update call accordingly.  Handle a constant-sized
33152         subreg of a variable-length CONST_VECTOR.
33154 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
33155             Alan Hayward  <alan.hayward@arm.com>
33156             David Sherwood  <david.sherwood@arm.com>
33158         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
33159         (add_offset_to_base): New function, split out from...
33160         (create_mem_ref): ...here.  When handling a scale other than 1,
33161         check first whether the address is valid without the offset.
33162         Add it into the base if so, leaving the index and scale as-is.
33164 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
33166         PR c++/83778
33167         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
33168         fold_for_warn before checking if arg2 is INTEGER_CST.
33170 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
33172         * config/rs6000/predicates.md (load_multiple_operation): Delete.
33173         (store_multiple_operation): Delete.
33174         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
33175         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
33176         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
33177         guarded by TARGET_STRING.
33178         (rs6000_output_load_multiple): Delete.
33179         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
33180         OPTION_MASK_STRING / TARGET_STRING handling.
33181         (print_operand) <'N', 'O'>: Add comment that these are unused now.
33182         (const rs6000_opt_masks) <"string">: Change mask to 0.
33183         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
33184         (MASK_STRING): Delete.
33185         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
33186         parts.  Simplify.
33187         (load_multiple): Delete.
33188         (*ldmsi8): Delete.
33189         (*ldmsi7): Delete.
33190         (*ldmsi6): Delete.
33191         (*ldmsi5): Delete.
33192         (*ldmsi4): Delete.
33193         (*ldmsi3): Delete.
33194         (store_multiple): Delete.
33195         (*stmsi8): Delete.
33196         (*stmsi7): Delete.
33197         (*stmsi6): Delete.
33198         (*stmsi5): Delete.
33199         (*stmsi4): Delete.
33200         (*stmsi3): Delete.
33201         (movmemsi_8reg): Delete.
33202         (corresponding unnamed define_insn): Delete.
33203         (movmemsi_6reg): Delete.
33204         (corresponding unnamed define_insn): Delete.
33205         (movmemsi_4reg): Delete.
33206         (corresponding unnamed define_insn): Delete.
33207         (movmemsi_2reg): Delete.
33208         (corresponding unnamed define_insn): Delete.
33209         (movmemsi_1reg): Delete.
33210         (corresponding unnamed define_insn): Delete.
33211         * config/rs6000/rs6000.opt (mno-string): New.
33212         (mstring): Replace by deprecation warning stub.
33213         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
33215 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
33217         * regrename.c (regrename_do_replace): If replacing the same
33218         reg multiple times, try to reuse last created gen_raw_REG.
33220         PR debug/81155
33221         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
33222         main to workaround a bug in GDB.
33224 2018-01-12  Tom de Vries  <tom@codesourcery.com>
33226         PR target/83737
33227         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
33229 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
33231         PR rtl-optimization/80481
33232         * ira-color.c (get_cap_member): New function.
33233         (allocnos_conflict_by_live_ranges_p): Use it.
33234         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
33235         (setup_slot_coalesced_allocno_live_ranges): Ditto.
33237 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
33239         PR target/83628
33240         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
33241         (*saddl_se_1): Ditto.
33242         (*ssubsi_1): Ditto.
33243         (*ssubl_se_1): Ditto.
33245 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
33247         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
33248         rather than wi::to_widest for DR_INITs.
33249         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
33250         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
33251         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
33252         INTEGER_CSTs.
33253         (vect_analyze_group_access_1): Note that here.
33255 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
33257         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
33258         polynomial type sizes.
33260 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
33262         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
33263         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
33264         (gimple_add_tmp_var): Likewise.
33266 2018-01-12  Martin Liska  <mliska@suse.cz>
33268         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
33269         (gimple_alloc_sizes): Likewise.
33270         (dump_gimple_statistics): Use PRIu64 in printf format.
33271         * gimple.h: Change uint64_t to int.
33273 2018-01-12  Martin Liska  <mliska@suse.cz>
33275         * tree-core.h: Use uint64_t instead of int.
33276         * tree.c (tree_node_counts): Likewise.
33277         (tree_node_sizes): Likewise.
33278         (dump_tree_statistics): Use PRIu64 in printf format.
33280 2018-01-12  Martin Liska  <mliska@suse.cz>
33282         * Makefile.in: As qsort_chk is implemented in vec.c, add
33283         vec.o to linkage of gencfn-macros.
33284         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
33285         passing the info to record_node_allocation_statistics.
33286         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
33287         and pass the info.
33288         * ggc-common.c (struct ggc_usage): Add operator== and use
33289         it in operator< and compare function.
33290         * mem-stats.h (struct mem_usage): Likewise.
33291         * vec.c (struct vec_usage): Remove operator< and compare
33292         function. Can be simply inherited.
33294 2018-01-12  Martin Jambor  <mjambor@suse.cz>
33296         PR target/81616
33297         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
33298         * tree-ssa-math-opts.c: Include domwalk.h.
33299         (convert_mult_to_fma_1): New function.
33300         (fma_transformation_info): New type.
33301         (fma_deferring_state): Likewise.
33302         (cancel_fma_deferring): New function.
33303         (result_of_phi): Likewise.
33304         (last_fma_candidate_feeds_initial_phi): Likewise.
33305         (convert_mult_to_fma): Added deferring logic, split actual
33306         transformation to convert_mult_to_fma_1.
33307         (math_opts_dom_walker): New type.
33308         (math_opts_dom_walker::after_dom_children): New method, body moved
33309         here from pass_optimize_widening_mul::execute, added deferring logic
33310         bits.
33311         (pass_optimize_widening_mul::execute): Moved most of code to
33312         math_opts_dom_walker::after_dom_children.
33313         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
33314         * config/i386/i386.c (ix86_option_override_internal): Added
33315         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
33317 2018-01-12  Richard Biener  <rguenther@suse.de>
33319         PR debug/83157
33320         * dwarf2out.c (gen_variable_die): Do not reset old_die for
33321         inline instance vars.
33323 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
33325         PR target/81819
33326         * config/rx/rx.c (rx_is_restricted_memory_address):
33327         Handle SUBREG case.
33329 2018-01-12  Richard Biener  <rguenther@suse.de>
33331         PR tree-optimization/80846
33332         * target.def (split_reduction): New target hook.
33333         * targhooks.c (default_split_reduction): New function.
33334         * targhooks.h (default_split_reduction): Declare.
33335         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
33336         target requests first reduce vectors by combining low and high
33337         parts.
33338         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
33339         (get_vectype_for_scalar_type_and_size): Export.
33340         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
33341         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
33342         * doc/tm.texi: Regenerate.
33343         * config/i386/i386.c (ix86_split_reduction): Implement
33344         TARGET_VECTORIZE_SPLIT_REDUCTION.
33346 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
33348         PR target/83368
33349         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
33350         in PIC mode except for TARGET_VXWORKS_RTP.
33351         * config/sparc/sparc.c: Include cfgrtl.h.
33352         (TARGET_INIT_PIC_REG): Define.
33353         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
33354         (sparc_pic_register_p): New predicate.
33355         (sparc_legitimate_address_p): Use it.
33356         (sparc_legitimize_pic_address): Likewise.
33357         (sparc_delegitimize_address): Likewise.
33358         (sparc_mode_dependent_address_p): Likewise.
33359         (gen_load_pcrel_sym): Remove 4th parameter.
33360         (load_got_register): Adjust call to above.  Remove obsolete stuff.
33361         (sparc_expand_prologue): Do not call load_got_register here.
33362         (sparc_flat_expand_prologue): Likewise.
33363         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
33364         (sparc_use_pseudo_pic_reg): New function.
33365         (sparc_init_pic_reg): Likewise.
33366         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
33367         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
33369 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
33371         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
33372         Add item for branch_cost.
33374 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
33376         PR rtl-optimization/83565
33377         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
33378         not extend the result to a larger mode for rotate operations.
33379         (num_sign_bit_copies1): Likewise.
33381 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33383         PR target/40411
33384         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
33385         -symbolic.
33386         Use values-Xc.o for -pedantic.
33387         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
33389 2018-01-12  Martin Liska  <mliska@suse.cz>
33391         PR ipa/83054
33392         * ipa-devirt.c (final_warning_record::grow_type_warnings):
33393         New function.
33394         (possible_polymorphic_call_targets): Use it.
33395         (ipa_devirt): Likewise.
33397 2018-01-12  Martin Liska  <mliska@suse.cz>
33399         * profile-count.h (enum profile_quality): Use 0 as invalid
33400         enum value of profile_quality.
33402 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
33404         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
33405         -mext-string options.
33407 2018-01-12  Richard Biener  <rguenther@suse.de>
33409         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
33410         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
33411         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
33412         Likewise.
33413         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
33415 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
33417         * configure.ac (--with-long-double-format): Add support for the
33418         configuration option to change the default long double format on
33419         PowerPC systems.
33420         * config.gcc (powerpc*-linux*-*): Likewise.
33421         * configure: Regenerate.
33422         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
33423         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
33424         used without modification.
33426 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
33428         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
33429         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
33430         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
33431         MISC_BUILTIN_SPEC_BARRIER.
33432         (rs6000_init_builtins): Likewise.
33433         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
33434         enum value.
33435         (speculation_barrier): New define_insn.
33436         * doc/extend.texi: Document __builtin_speculation_barrier.
33438 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
33440         PR target/83203
33441         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
33442         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
33443         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
33444         iterators.
33445         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
33446         integral modes instead of "ss" and "sd".
33447         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
33448         vectors with 32-bit and 64-bit elements.
33449         (vecdupssescalarmodesuffix): New mode attribute.
33450         (vec_dup<mode>): Use it.
33452 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
33454         PR target/83330
33455         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
33456         frame if argument is passed on stack.
33458 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
33460         PR target/82682
33461         * ree.c (combine_reaching_defs): Optimize also
33462         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
33463         reg2=any_extend(exp); reg1=reg2;, formatting fix.
33465 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
33467         PR middle-end/83189
33468         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
33470 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
33472         PR middle-end/83718
33473         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
33474         after they are computed.
33476 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
33478         PR tree-optimization/83695
33479         * gimple-loop-linterchange.cc
33480         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
33481         reset cached scev information after interchange.
33482         (pass_linterchange::execute): Remove call to scev_reset_htab.
33484 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33486         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
33487         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
33488         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
33489         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
33490         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
33491         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
33492         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
33493         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
33494         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
33495         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
33496         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
33497         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
33498         (V_lane_reg): Likewise.
33499         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
33500         New define_expand.
33501         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
33502         (vfmal_lane_low<mode>_intrinsic,
33503         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
33504         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
33505         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
33506         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
33507         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
33508         vfmsl_lane_high<mode>_intrinsic): New define_insns.
33510 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33512         * config/arm/arm-cpus.in (fp16fml): New feature.
33513         (ALL_SIMD): Add fp16fml.
33514         (armv8.2-a): Add fp16fml as an option.
33515         (armv8.3-a): Likewise.
33516         (armv8.4-a): Add fp16fml as part of fp16.
33517         * config/arm/arm.h (TARGET_FP16FML): Define.
33518         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
33519         when appropriate.
33520         * config/arm/arm-modes.def (V2HF): Define.
33521         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
33522         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
33523         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
33524         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
33525         vfmsl_low, vfmsl_high): New set of builtins.
33526         * config/arm/iterators.md (PLUSMINUS): New code iterator.
33527         (vfml_op): New code attribute.
33528         (VFMLHALVES): New int iterator.
33529         (VFML, VFMLSEL): New mode attributes.
33530         (V_reg): Define mapping for V2HF.
33531         (V_hi, V_lo): New mode attributes.
33532         (VF_constraint): Likewise.
33533         (vfml_half, vfml_half_selector): New int attributes.
33534         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
33535         define_expand.
33536         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
33537         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
33538         New define_insn.
33539         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
33540         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
33541         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
33542         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
33543         documentation.
33544         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
33545         Document new effective target and option set.
33547 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33549         * config/arm/arm-cpus.in (armv8_4): New feature.
33550         (ARMv8_4a): New fgroup.
33551         (armv8.4-a): New arch.
33552         * config/arm/arm-tables.opt: Regenerate.
33553         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
33554         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
33555         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
33556         Add matching rules for -march=armv8.4-a and extensions.
33557         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
33559 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
33561         PR target/81821
33562         * config/rx/rx.md (BW): New mode attribute.
33563         (sync_lock_test_and_setsi): Add mode suffix to insn output.
33565 2018-01-11  Richard Biener  <rguenther@suse.de>
33567         PR tree-optimization/83435
33568         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
33569         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
33570         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
33572 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33573             Alan Hayward  <alan.hayward@arm.com>
33574             David Sherwood  <david.sherwood@arm.com>
33576         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
33577         field.
33578         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
33579         (aarch64_print_address_internal): Use it to check for a zero offset.
33581 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33582             Alan Hayward  <alan.hayward@arm.com>
33583             David Sherwood  <david.sherwood@arm.com>
33585         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
33586         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
33587         Return a poly_int64 rather than a HOST_WIDE_INT.
33588         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
33589         rather than a HOST_WIDE_INT.
33590         * config/aarch64/aarch64.h (aarch64_frame): Protect with
33591         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
33592         hard_fp_offset, frame_size, initial_adjust, callee_offset and
33593         final_offset from HOST_WIDE_INT to poly_int64.
33594         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
33595         to_constant when getting the number of units in an Advanced SIMD
33596         mode.
33597         (aarch64_builtin_vectorized_function): Check for a constant number
33598         of units.
33599         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
33600         GET_MODE_SIZE.
33601         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
33602         attribute instead of GET_MODE_NUNITS.
33603         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
33604         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
33605         GET_MODE_SIZE for fixed-size registers.
33606         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
33607         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
33608         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
33609         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
33610         (aarch64_print_operand, aarch64_print_address_internal)
33611         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
33612         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
33613         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
33614         Handle polynomial GET_MODE_SIZE.
33615         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
33616         wider than SImode without modification.
33617         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
33618         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
33619         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
33620         passing and returning SVE modes.
33621         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
33622         rather than GEN_INT.
33623         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
33624         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
33625         (aarch64_allocate_and_probe_stack_space): Likewise.
33626         (aarch64_layout_frame): Cope with polynomial offsets.
33627         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
33628         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
33629         polynomial offsets.
33630         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
33631         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
33632         poly_int64 rather than a HOST_WIDE_INT.
33633         (aarch64_get_separate_components, aarch64_process_components)
33634         (aarch64_expand_prologue, aarch64_expand_epilogue)
33635         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
33636         (aarch64_anchor_offset): New function, split out from...
33637         (aarch64_legitimize_address): ...here.
33638         (aarch64_builtin_vectorization_cost): Handle polynomial
33639         TYPE_VECTOR_SUBPARTS.
33640         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
33641         GET_MODE_NUNITS.
33642         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
33643         number of elements from the PARALLEL rather than the mode.
33644         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
33645         rather than GET_MODE_BITSIZE.
33646         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
33647         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
33648         (aarch64_expand_vec_perm_const_1): Handle polynomial
33649         d->perm.length () and d->perm elements.
33650         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
33651         Apply to_constant to d->perm elements.
33652         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
33653         polynomial CONST_VECTOR_NUNITS.
33654         (aarch64_move_pointer): Take amount as a poly_int64 rather
33655         than an int.
33656         (aarch64_progress_pointer): Avoid temporary variable.
33657         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
33658         the mode attribute instead of GET_MODE.
33660 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33661             Alan Hayward  <alan.hayward@arm.com>
33662             David Sherwood  <david.sherwood@arm.com>
33664         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
33665         x exists before using it.
33666         (aarch64_add_constant_internal): Rename to...
33667         (aarch64_add_offset_1): ...this.  Replace regnum with separate
33668         src and dest rtxes.  Handle the case in which they're different,
33669         including when the offset is zero.  Replace scratchreg with an rtx.
33670         Use 2 additions if there is no spare register into which we can
33671         move a 16-bit constant.
33672         (aarch64_add_constant): Delete.
33673         (aarch64_add_offset): Replace reg with separate src and dest
33674         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
33675         Use aarch64_add_offset_1.
33676         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
33677         an rtx rather than an int.  Take the delta as a poly_int64
33678         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
33679         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
33680         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
33681         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
33682         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
33683         and aarch64_add_sp.
33684         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
33685         aarch64_add_constant.
33687 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33689         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
33690         Use scalar_float_mode.
33692 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33694         * config/aarch64/aarch64-simd.md
33695         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
33696         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
33697         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
33698         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
33699         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
33700         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
33701         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
33702         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
33703         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
33704         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
33706 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33708         PR target/83514
33709         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
33710         targ_options->x_arm_arch_string is non NULL.
33712 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
33714         * config/aarch64/aarch64.h
33715         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
33717 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
33719         PR target/82096
33720         * expmed.c (emit_store_flag_force): Swap if const op0
33721         and change VOIDmode to mode of op0.
33723 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
33725         PR rtl-optimization/83761
33726         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
33727         than bytes to mode_for_size.
33729 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
33731         PR middle-end/83189
33732         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
33733         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
33734         profile.
33736 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
33738         PR middle-end/83575
33739         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
33740         when in layout mode.
33741         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
33742         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
33743         partition fixup.
33745 2018-01-10  Michael Collison  <michael.collison@arm.com>
33747         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
33748         * config/aarch64/aarch64-option-extension.def: Add
33749         AARCH64_OPT_EXTENSION of 'fp16fml'.
33750         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33751         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
33752         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
33753         * config/aarch64/constraints.md (Ui7): New constraint.
33754         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
33755         (VFMLA_SEL_W): Ditto.
33756         (f16quad): Ditto.
33757         (f16mac1): Ditto.
33758         (VFMLA16_LOW): New int iterator.
33759         (VFMLA16_HIGH): Ditto.
33760         (UNSPEC_FMLAL): New unspec.
33761         (UNSPEC_FMLSL): Ditto.
33762         (UNSPEC_FMLAL2): Ditto.
33763         (UNSPEC_FMLSL2): Ditto.
33764         (f16mac): New code attribute.
33765         * config/aarch64/aarch64-simd-builtins.def
33766         (aarch64_fmlal_lowv2sf): Ditto.
33767         (aarch64_fmlsl_lowv2sf): Ditto.
33768         (aarch64_fmlalq_lowv4sf): Ditto.
33769         (aarch64_fmlslq_lowv4sf): Ditto.
33770         (aarch64_fmlal_highv2sf): Ditto.
33771         (aarch64_fmlsl_highv2sf): Ditto.
33772         (aarch64_fmlalq_highv4sf): Ditto.
33773         (aarch64_fmlslq_highv4sf): Ditto.
33774         (aarch64_fmlal_lane_lowv2sf): Ditto.
33775         (aarch64_fmlsl_lane_lowv2sf): Ditto.
33776         (aarch64_fmlal_laneq_lowv2sf): Ditto.
33777         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
33778         (aarch64_fmlalq_lane_lowv4sf): Ditto.
33779         (aarch64_fmlsl_lane_lowv4sf): Ditto.
33780         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
33781         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
33782         (aarch64_fmlal_lane_highv2sf): Ditto.
33783         (aarch64_fmlsl_lane_highv2sf): Ditto.
33784         (aarch64_fmlal_laneq_highv2sf): Ditto.
33785         (aarch64_fmlsl_laneq_highv2sf): Ditto.
33786         (aarch64_fmlalq_lane_highv4sf): Ditto.
33787         (aarch64_fmlsl_lane_highv4sf): Ditto.
33788         (aarch64_fmlalq_laneq_highv4sf): Ditto.
33789         (aarch64_fmlsl_laneq_highv4sf): Ditto.
33790         * config/aarch64/aarch64-simd.md:
33791         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
33792         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33793         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
33794         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33795         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
33796         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
33797         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
33798         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
33799         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
33800         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
33801         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
33802         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
33803         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
33804         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
33805         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
33806         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
33807         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
33808         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
33809         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
33810         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
33811         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
33812         (vfmlsl_low_u32): Ditto.
33813         (vfmlalq_low_u32): Ditto.
33814         (vfmlslq_low_u32): Ditto.
33815         (vfmlal_high_u32): Ditto.
33816         (vfmlsl_high_u32): Ditto.
33817         (vfmlalq_high_u32): Ditto.
33818         (vfmlslq_high_u32): Ditto.
33819         (vfmlal_lane_low_u32): Ditto.
33820         (vfmlsl_lane_low_u32): Ditto.
33821         (vfmlal_laneq_low_u32): Ditto.
33822         (vfmlsl_laneq_low_u32): Ditto.
33823         (vfmlalq_lane_low_u32): Ditto.
33824         (vfmlslq_lane_low_u32): Ditto.
33825         (vfmlalq_laneq_low_u32): Ditto.
33826         (vfmlslq_laneq_low_u32): Ditto.
33827         (vfmlal_lane_high_u32): Ditto.
33828         (vfmlsl_lane_high_u32): Ditto.
33829         (vfmlal_laneq_high_u32): Ditto.
33830         (vfmlsl_laneq_high_u32): Ditto.
33831         (vfmlalq_lane_high_u32): Ditto.
33832         (vfmlslq_lane_high_u32): Ditto.
33833         (vfmlalq_laneq_high_u32): Ditto.
33834         (vfmlslq_laneq_high_u32): Ditto.
33835         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
33836         (AARCH64_FL_FOR_ARCH8_4): New.
33837         (AARCH64_ISA_F16FML): New ISA flag.
33838         (TARGET_F16FML): New feature flag for fp16fml.
33839         (doc/invoke.texi): Document new fp16fml option.
33841 2018-01-10  Michael Collison  <michael.collison@arm.com>
33843         * config/aarch64/aarch64-builtins.c:
33844         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
33845         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33846         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
33847         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
33848         (AARCH64_ISA_SHA3): New ISA flag.
33849         (TARGET_SHA3): New feature flag for sha3.
33850         * config/aarch64/iterators.md (sha512_op): New int attribute.
33851         (CRYPTO_SHA512): New int iterator.
33852         (UNSPEC_SHA512H): New unspec.
33853         (UNSPEC_SHA512H2): Ditto.
33854         (UNSPEC_SHA512SU0): Ditto.
33855         (UNSPEC_SHA512SU1): Ditto.
33856         * config/aarch64/aarch64-simd-builtins.def
33857         (aarch64_crypto_sha512hqv2di): New builtin.
33858         (aarch64_crypto_sha512h2qv2di): Ditto.
33859         (aarch64_crypto_sha512su0qv2di): Ditto.
33860         (aarch64_crypto_sha512su1qv2di): Ditto.
33861         (aarch64_eor3qv8hi): Ditto.
33862         (aarch64_rax1qv2di): Ditto.
33863         (aarch64_xarqv2di): Ditto.
33864         (aarch64_bcaxqv8hi): Ditto.
33865         * config/aarch64/aarch64-simd.md:
33866         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
33867         (aarch64_crypto_sha512su0qv2di): Ditto.
33868         (aarch64_crypto_sha512su1qv2di): Ditto.
33869         (aarch64_eor3qv8hi): Ditto.
33870         (aarch64_rax1qv2di): Ditto.
33871         (aarch64_xarqv2di): Ditto.
33872         (aarch64_bcaxqv8hi): Ditto.
33873         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
33874         (vsha512h2q_u64): Ditto.
33875         (vsha512su0q_u64): Ditto.
33876         (vsha512su1q_u64): Ditto.
33877         (veor3q_u16): Ditto.
33878         (vrax1q_u64): Ditto.
33879         (vxarq_u64): Ditto.
33880         (vbcaxq_u16): Ditto.
33881         * config/arm/types.md (crypto_sha512): New type attribute.
33882         (crypto_sha3): Ditto.
33883         (doc/invoke.texi): Document new sha3 option.
33885 2018-01-10  Michael Collison  <michael.collison@arm.com>
33887         * config/aarch64/aarch64-builtins.c:
33888         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
33889         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33890         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
33891         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
33892         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
33893         (AARCH64_ISA_SM4): New ISA flag.
33894         (TARGET_SM4): New feature flag for sm4.
33895         * config/aarch64/aarch64-simd-builtins.def
33896         (aarch64_sm3ss1qv4si): Ditto.
33897         (aarch64_sm3tt1aq4si): Ditto.
33898         (aarch64_sm3tt1bq4si): Ditto.
33899         (aarch64_sm3tt2aq4si): Ditto.
33900         (aarch64_sm3tt2bq4si): Ditto.
33901         (aarch64_sm3partw1qv4si): Ditto.
33902         (aarch64_sm3partw2qv4si): Ditto.
33903         (aarch64_sm4eqv4si): Ditto.
33904         (aarch64_sm4ekeyqv4si): Ditto.
33905         * config/aarch64/aarch64-simd.md:
33906         (aarch64_sm3ss1qv4si): Ditto.
33907         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
33908         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
33909         (aarch64_sm4eqv4si): Ditto.
33910         (aarch64_sm4ekeyqv4si): Ditto.
33911         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
33912         (sm3part_op): Ditto.
33913         (CRYPTO_SM3TT): Ditto.
33914         (CRYPTO_SM3PART): Ditto.
33915         (UNSPEC_SM3SS1): New unspec.
33916         (UNSPEC_SM3TT1A): Ditto.
33917         (UNSPEC_SM3TT1B): Ditto.
33918         (UNSPEC_SM3TT2A): Ditto.
33919         (UNSPEC_SM3TT2B): Ditto.
33920         (UNSPEC_SM3PARTW1): Ditto.
33921         (UNSPEC_SM3PARTW2): Ditto.
33922         (UNSPEC_SM4E): Ditto.
33923         (UNSPEC_SM4EKEY): Ditto.
33924         * config/aarch64/constraints.md (Ui2): New constraint.
33925         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
33926         * config/arm/types.md (crypto_sm3): New type attribute.
33927         (crypto_sm4): Ditto.
33928         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
33929         (vsm3tt1aq_u32): Ditto.
33930         (vsm3tt1bq_u32): Ditto.
33931         (vsm3tt2aq_u32): Ditto.
33932         (vsm3tt2bq_u32): Ditto.
33933         (vsm3partw1q_u32): Ditto.
33934         (vsm3partw2q_u32): Ditto.
33935         (vsm4eq_u32): Ditto.
33936         (vsm4ekeyq_u32): Ditto.
33937         (doc/invoke.texi): Document new sm4 option.
33939 2018-01-10  Michael Collison  <michael.collison@arm.com>
33941         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
33942         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
33943         (AARCH64_FL_FOR_ARCH8_4): New.
33944         (AARCH64_FL_V8_4): New flag.
33945         (doc/invoke.texi): Document new armv8.4-a option.
33947 2018-01-10  Michael Collison  <michael.collison@arm.com>
33949         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33950         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
33951         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
33952         * config/aarch64/aarch64-option-extension.def: Add
33953         AARCH64_OPT_EXTENSION of 'sha2'.
33954         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
33955         (crypto): Disable sha2 and aes if crypto disabled.
33956         (crypto): Enable aes and sha2 if enabled.
33957         (simd): Disable sha2 and aes if simd disabled.
33958         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
33959         New flags.
33960         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
33961         (TARGET_SHA2): New feature flag for sha2.
33962         (TARGET_AES): New feature flag for aes.
33963         * config/aarch64/aarch64-simd.md:
33964         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
33965         conditional on TARGET_AES.
33966         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
33967         (aarch64_crypto_sha1hsi): Make pattern conditional
33968         on TARGET_SHA2.
33969         (aarch64_crypto_sha1hv4si): Ditto.
33970         (aarch64_be_crypto_sha1hv4si): Ditto.
33971         (aarch64_crypto_sha1su1v4si): Ditto.
33972         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
33973         (aarch64_crypto_sha1su0v4si): Ditto.
33974         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
33975         (aarch64_crypto_sha256su0v4si): Ditto.
33976         (aarch64_crypto_sha256su1v4si): Ditto.
33977         (doc/invoke.texi): Document new aes and sha2 options.
33979 2018-01-10  Martin Sebor  <msebor@redhat.com>
33981         PR tree-optimization/83781
33982         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
33983         as string arrays.
33985 2018-01-11  Martin Sebor  <msebor@gmail.com>
33986             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33988         PR tree-optimization/83501
33989         PR tree-optimization/81703
33991         * tree-ssa-strlen.c (get_string_cst): Rename...
33992         (get_string_len): ...to this.  Handle global constants.
33993         (handle_char_store): Adjust.
33995 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
33996             Jim Wilson  <jimw@sifive.com>
33998         * config/riscv/riscv-protos.h (riscv_output_return): New.
33999         * config/riscv/riscv.c (struct machine_function): New naked_p field.
34000         (riscv_attribute_table, riscv_output_return),
34001         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
34002         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
34003         (riscv_compute_frame_info): Only compute frame->mask if not a naked
34004         function.
34005         (riscv_expand_prologue): Add early return for naked function.
34006         (riscv_expand_epilogue): Likewise.
34007         (riscv_function_ok_for_sibcall): Return false for naked function.
34008         (riscv_set_current_function): New.
34009         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
34010         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
34011         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
34012         * doc/extend.texi (RISC-V Function Attributes): New.
34014 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
34016         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
34017         check for 128-bit long double before checking TCmode.
34018         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
34019         128-bit long doubles before checking TFmode or TCmode.
34020         (FLOAT128_IBM_P): Likewise.
34022 2018-01-10  Martin Sebor  <msebor@redhat.com>
34024         PR tree-optimization/83671
34025         * builtins.c (c_strlen): Unconditionally return zero for the empty
34026         string.
34027         Use -Warray-bounds for warnings.
34028         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
34029         for non-constant array indices with COMPONENT_REF, arrays of
34030         arrays, and pointers to arrays.
34031         (gimple_fold_builtin_strlen): Determine and set length range for
34032         non-constant character arrays.
34034 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
34036         PR middle-end/81897
34037         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
34038         empty blocks.
34040 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
34042         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
34044 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
34046         PR target/83399
34047         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
34048         VECTOR_MEM_ALTIVEC_OR_VSX_P.
34049         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
34050         indexed_or_indirect_operand predicate.
34051         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
34052         (*vsx_le_perm_load_v8hi): Likewise.
34053         (*vsx_le_perm_load_v16qi): Likewise.
34054         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
34055         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
34056         (*vsx_le_perm_store_v8hi): Likewise.
34057         (*vsx_le_perm_store_v16qi): Likewise.
34058         (eight unnamed splitters): Likewise.
34060 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
34062         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
34063         * config/rs6000/emmintrin.h: Likewise.
34064         * config/rs6000/mmintrin.h: Likewise.
34065         * config/rs6000/xmmintrin.h: Likewise.
34067 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
34069         PR c++/43486
34070         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
34071         "public_flag".
34072         * tree.c (tree_nop_conversion): Return true for location wrapper
34073         nodes.
34074         (maybe_wrap_with_location): New function.
34075         (selftest::check_strip_nops): New function.
34076         (selftest::test_location_wrappers): New function.
34077         (selftest::tree_c_tests): Call it.
34078         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
34079         (maybe_wrap_with_location): New decl.
34080         (EXPR_LOCATION_WRAPPER_P): New macro.
34081         (location_wrapper_p): New inline function.
34082         (tree_strip_any_location_wrapper): New inline function.
34084 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
34086         PR target/83735
34087         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
34088         stack_realign_offset for the largest alignment of stack slot
34089         actually used.
34090         (ix86_find_max_used_stack_alignment): New function.
34091         (ix86_finalize_stack_frame_flags): Use it.  Set
34092         max_used_stack_alignment if we don't realign stack.
34093         * config/i386/i386.h (machine_function): Add
34094         max_used_stack_alignment.
34096 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
34098         * config/arm/arm.opt (-mbranch-cost): New option.
34099         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
34100         account.
34102 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
34104         PR target/83629
34105         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
34106         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
34108 2018-01-10  Richard Biener  <rguenther@suse.de>
34110         PR debug/83765
34111         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
34112         early out so it also covers the case where we have a non-NULL
34113         origin.
34115 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
34117         PR tree-optimization/83753
34118         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
34119         for non-strided grouped accesses if the number of elements is 1.
34121 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
34123         PR target/81616
34124         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
34125         * i386.h (TARGET_USE_GATHER): Define.
34126         * x86-tune.def (X86_TUNE_USE_GATHER): New.
34128 2018-01-10  Martin Liska  <mliska@suse.cz>
34130         PR bootstrap/82831
34131         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
34132         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
34133         partitioning.
34134         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
34135         CLEANUP_NO_PARTITIONING is not set.
34137 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
34139         * doc/rtl.texi: Remove documentation of (const ...) wrappers
34140         for vectors, as a partial revert of r254296.
34141         * rtl.h (const_vec_p): Delete.
34142         (const_vec_duplicate_p): Don't test for vector CONSTs.
34143         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
34144         * expmed.c (make_tree): Likewise.
34146         Revert:
34147         * common.md (E, F): Use CONSTANT_P instead of checking for
34148         CONST_VECTOR.
34149         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
34150         checking for CONST_VECTOR.
34152 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
34154         PR middle-end/83575
34155         * predict.c (force_edge_cold): Handle in more sane way edges
34156         with no prediction.
34158 2018-01-09  Carl Love  <cel@us.ibm.com>
34160         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
34161         V4SI, V4SF types.
34162         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
34163         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
34164         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
34165         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
34166         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
34167         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
34168         * config/rs6000/rs6000-protos.h: Add extern defition for
34169         rs6000_generate_float2_double_code.
34170         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
34171         function.
34172         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
34173         (float2_v2df): Add define_expand.
34175 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
34177         PR target/83628
34178         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
34179         op_mode in the force_to_mode call.
34181 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
34183         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
34184         instead of checking each element individually.
34185         (aarch64_evpc_uzp): Likewise.
34186         (aarch64_evpc_zip): Likewise.
34187         (aarch64_evpc_ext): Likewise.
34188         (aarch64_evpc_rev): Likewise.
34189         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
34190         instead of checking each element individually.  Return true without
34191         generating rtl if
34192         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
34193         whether all selected elements come from the same input, instead of
34194         checking each element individually.  Remove calls to gen_rtx_REG,
34195         start_sequence and end_sequence and instead assert that no rtl is
34196         generated.
34198 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
34200         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
34201         order of HIGH and CONST checks.
34203 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
34205         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
34206         if the destination isn't an SSA_NAME.
34208 2018-01-09  Richard Biener  <rguenther@suse.de>
34210         PR tree-optimization/83668
34211         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
34212         move prologue...
34213         (canonicalize_loop_form): ... here, renamed from ...
34214         (canonicalize_loop_closed_ssa_form): ... this and amended to
34215         swap successor edges for loop exit blocks to make us use
34216         the RPO order we need for initial schedule generation.
34218 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
34220         PR tree-optimization/64811
34221         * match.pd: When optimizing comparisons with Inf, avoid
34222         introducing or losing exceptions from comparisons with NaN.
34224 2018-01-09  Martin Liska  <mliska@suse.cz>
34226         PR sanitizer/82517
34227         * asan.c (shadow_mem_size): Add gcc_assert.
34229 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
34231         Don't save registers in main().
34233         PR target/83738
34234         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
34235         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
34236         * config/avr/avr.c (avr_set_current_function): Don't error if
34237         naked, OS_task or OS_main are specified at the same time.
34238         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
34239         OS_main.
34240         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
34241         attribute.
34242         * common/config/avr/avr-common.c (avr_option_optimization_table):
34243         Switch on -mmain-is-OS_task for optimizing compilations.
34245 2018-01-09  Richard Biener  <rguenther@suse.de>
34247         PR tree-optimization/83572
34248         * graphite.c: Include cfganal.h.
34249         (graphite_transform_loops): Connect infinite loops to exit
34250         and remove fake edges at the end.
34252 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
34254         * ipa-inline.c (edge_badness): Revert accidental checkin.
34256 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
34258         PR ipa/80763
34259         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
34260         symbols; not inline clones.
34262 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
34264         PR target/83507
34265         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
34266         hard registers.  Formatting fixes.
34268         PR preprocessor/83722
34269         * gcc.c (try_generate_repro): Pass
34270         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
34271         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
34272         do_report_bug.
34274 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
34275             Kito Cheng  <kito.cheng@gmail.com>
34277         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
34278         (riscv_leaf_function_p): Delete.
34279         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
34281 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34283         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
34284         function.
34285         (do_ifelse): New function.
34286         (do_isel): New function.
34287         (do_sub3): New function.
34288         (do_add3): New function.
34289         (do_load_mask_compare): New function.
34290         (do_overlap_load_compare): New function.
34291         (expand_compare_loop): New function.
34292         (expand_block_compare): Call expand_compare_loop() when appropriate.
34293         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
34294         option description.
34295         (-mblock-compare-inline-loop-limit): New option.
34297 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34299         PR target/83677
34300         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
34301         Reverse order of second and third operands in first alternative.
34302         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
34303         of first and second elements in UNSPEC_VPERMR vector.
34304         (altivec_expand_vec_perm_le): Likewise.
34306 2018-01-08  Jeff Law  <law@redhat.com>
34308         PR rtl-optimizatin/81308
34309         * tree-switch-conversion.c (cfg_altered): New file scoped static.
34310         (process_switch): If group_case_labels makes a change, then set
34311         cfg_altered.
34312         (pass_convert_switch::execute): If a switch is converted, then
34313         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
34315         PR rtl-optimization/81308
34316         * recog.c (split_all_insns): Conditionally cleanup the CFG after
34317         splitting insns.
34319 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
34321         PR target/83663 - Revert r255946
34322         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
34323         generation for cases where splatting a value is not useful.
34324         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
34325         across a vec_duplicate and a paradoxical subreg forming a vector
34326         mode to a vec_concat.
34328 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34330         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
34331         -march=armv8.3-a variants.
34332         * config/arm/t-multilib: Likewise.
34333         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
34335 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34337         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
34338         to generate rtl.
34339         (cceq_ior_compare_complement): Give it a name so I can use it, and
34340         change boolean_or_operator predicate to boolean_operator so it can
34341         be used to generate a crand.
34342         (eqne): New code iterator.
34343         (bd/bd_neg): New code_attrs.
34344         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
34345         a single define_insn.
34346         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
34347         decrement (bdnzt/bdnzf/bdzt/bdzf).
34348         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
34349         with the new names of the branch decrement patterns, and added the
34350         names of the branch decrement conditional patterns.
34352 2018-01-08  Richard Biener  <rguenther@suse.de>
34354         PR tree-optimization/83563
34355         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
34356         cache.
34358 2018-01-08  Richard Biener  <rguenther@suse.de>
34360         PR middle-end/83713
34361         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
34363 2018-01-08  Richard Biener  <rguenther@suse.de>
34365         PR tree-optimization/83685
34366         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
34367         references to abnormals.
34369 2018-01-08  Richard Biener  <rguenther@suse.de>
34371         PR lto/83719
34372         * dwarf2out.c (output_indirect_strings): Handle empty
34373         skeleton_debug_str_hash.
34374         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
34376 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
34378         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
34379         (emit_store_direct): Likewise.
34380         (arc_trampoline_adjust_address): Likewise.
34381         (arc_asm_trampoline_template): New function.
34382         (arc_initialize_trampoline): Use asm_trampoline_template.
34383         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
34384         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
34385         * config/arc/arc.md (flush_icache): Delete pattern.
34387 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
34389         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
34390         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
34391         munaligned-access.
34393 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34395         PR target/83681
34396         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
34397         by not USED_FOR_TARGET.
34398         (make_pass_resolve_sw_modes): Likewise.
34400 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34402         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
34403         USED_FOR_TARGET.
34405 2018-01-08  Richard Biener  <rguenther@suse.de>
34407         PR middle-end/83580
34408         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
34410 2018-01-08  Richard Biener  <rguenther@suse.de>
34412         PR middle-end/83517
34413         * match.pd ((t * 2) / 2) -> t): Add missing :c.
34415 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
34417         PR middle-end/81897
34418         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
34419         basic blocks with a small number of successors.
34420         (convert_control_dep_chain_into_preds): Improve handling of
34421         forwarder blocks.
34422         (dump_predicates): Split apart into...
34423         (dump_pred_chain): ...here...
34424         (dump_pred_info): ...and here.
34425         (can_one_predicate_be_invalidated_p): Add debugging printfs.
34426         (can_chain_union_be_invalidated_p): Improve check for invalidation
34427         of paths.
34428         (uninit_uses_cannot_happen): Avoid unnecessary if
34429         convert_control_dep_chain_into_preds yielded nothing.
34431 2018-01-06  Martin Sebor  <msebor@redhat.com>
34433         PR tree-optimization/83640
34434         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
34435         subtracting negative offset from size.
34436         (builtin_access::overlap): Adjust offset bounds of the access to fall
34437         within the size of the object if possible.
34439 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
34441         PR rtl-optimization/83699
34442         * expmed.c (extract_bit_field_1): Restrict the vector usage of
34443         extract_bit_field_as_subreg to cases in which the extracted
34444         value is also a vector.
34446         * lra-constraints.c (process_alt_operands): Test for the equivalence
34447         substitutions when detecting a possible reload cycle.
34449 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
34451         PR debug/83480
34452         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
34453         by default if flag_selective_schedling{,2}.  Formatting fixes.
34455         PR rtl-optimization/83682
34456         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
34457         if it has non-VECTOR_MODE element mode.
34458         (vec_duplicate_p): Likewise.
34460         PR middle-end/83694
34461         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
34462         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
34464 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
34466         PR target/83604
34467         * config/i386/i386-builtin.def
34468         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
34469         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
34470         Require also OPTION_MASK_ISA_AVX512F in addition to
34471         OPTION_MASK_ISA_GFNI.
34472         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
34473         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
34474         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
34475         to OPTION_MASK_ISA_GFNI.
34476         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
34477         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
34478         OPTION_MASK_ISA_AVX512BW.
34479         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
34480         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
34481         addition to OPTION_MASK_ISA_GFNI.
34482         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
34483         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
34484         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
34485         to OPTION_MASK_ISA_GFNI.
34486         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
34487         a requirement for all ISAs rather than any of them with a few
34488         exceptions.
34489         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
34490         processing.
34491         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
34492         bitmasks to be enabled with 3 exceptions, instead of requiring any
34493         enabled ISA with lots of exceptions.
34494         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
34495         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
34496         Change avx512bw in isa attribute to avx512f.
34497         * config/i386/sgxintrin.h: Add license boilerplate.
34498         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
34499         to __AVX512F__ and __AVX512VL to __AVX512VL__.
34500         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
34501         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
34502         defined.
34503         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
34504         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
34505         temporarily sse2 rather than sse if not enabled already.
34507         PR target/83604
34508         * config/i386/sse.md (VI248_VLBW): Rename to ...
34509         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
34510         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
34511         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
34512         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
34513         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
34514         mode iterator instead of VI248_VLBW.
34516 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
34518         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
34519         (record_modified): Skip clobbers; add debug output.
34520         (param_change_prob): Use sreal frequencies.
34522 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
34524         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
34525         punt for user-aligned variables.
34527 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
34529         * tree-chrec.c (chrec_contains_symbols): Return true for
34530         POLY_INT_CST.
34532 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
34534         PR target/82439
34535         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
34536         of (x|y) == x for BICS pattern.
34538 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
34540         PR tree-optimization/83605
34541         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
34542         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
34543         can throw.
34545 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34547         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
34548         * config/epiphany/rtems.h: New file.
34550 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
34551             Uros Bizjak  <ubizjak@gmail.com>
34553         PR target/83554
34554         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
34555         QIreg_operand instead of register_operand predicate.
34556         * config/i386/i386.c (ix86_rop_should_change_byte_p,
34557         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
34558         comments instead of -fmitigate[-_]rop.
34560 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
34562         PR bootstrap/81926
34563         * cgraphunit.c (symbol_table::compile): Switch to text_section
34564         before calling assembly_start debug hook.
34565         * run-rtl-passes.c (run_rtl_passes): Likewise.
34566         Include output.h.
34568 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34570         * tree-vrp.c (extract_range_from_binary_expr_1): Check
34571         range_int_cst_p rather than !symbolic_range_p before calling
34572         extract_range_from_multiplicative_op_1.
34574 2018-01-04  Jeff Law  <law@redhat.com>
34576         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
34577         redundant test in assertion.
34579 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34581         * doc/rtl.texi: Document machine_mode wrapper classes.
34583 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34585         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
34586         using tree_to_uhwi.
34588 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34590         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
34591         the VEC_PERM_EXPR fold to fail.
34593 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
34595         PR debug/83585
34596         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
34597         to switched_sections.
34599 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34601         PR target/83680
34602         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
34603         test for d.testing.
34605 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
34607         PR target/83387
34608         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
34609         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
34611 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
34613         PR debug/83666
34614         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
34615         is BLKmode and bitpos not zero or mode change is needed.
34617 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
34619         PR target/83675
34620         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
34621         TARGET_VIS2.
34623 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
34625         PR target/83628
34626         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
34627         instead of MULT rtx.  Update all corresponding splitters.
34628         (*saddl_se): Ditto.
34629         (*ssub<modesuffix>): Ditto.
34630         (*ssubl_se): Ditto.
34631         (*cmp_sadd_di): Update split patterns.
34632         (*cmp_sadd_si): Ditto.
34633         (*cmp_sadd_sidi): Ditto.
34634         (*cmp_ssub_di): Ditto.
34635         (*cmp_ssub_si): Ditto.
34636         (*cmp_ssub_sidi): Ditto.
34637         * config/alpha/predicates.md (const23_operand): New predicate.
34638         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
34639         Look for ASHIFT, not MULT inner operand.
34640         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
34642 2018-01-04  Martin Liska  <mliska@suse.cz>
34644         PR gcov-profile/83669
34645         * gcov.c (output_intermediate_file): Add version to intermediate
34646         gcov file.
34647         * doc/gcov.texi: Document new field 'version' in intermediate
34648         file format. Fix location of '-k' option of gcov command.
34650 2018-01-04  Martin Liska  <mliska@suse.cz>
34652         PR ipa/82352
34653         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
34655 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
34657         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
34659 2018-01-03  Martin Sebor  <msebor@redhat.com>
34661         PR tree-optimization/83655
34662         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
34663         checking calls with invalid arguments.
34665 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34667         * tree-vect-stmts.c (vect_get_store_rhs): New function.
34668         (vectorizable_mask_load_store): Delete.
34669         (vectorizable_call): Return false for masked loads and stores.
34670         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
34671         instead of gimple_assign_rhs1.
34672         (vectorizable_load): Handle IFN_MASK_LOAD.
34673         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
34675 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34677         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
34678         split out from..,
34679         (vectorizable_mask_load_store): ...here.
34680         (vectorizable_load): ...and here.
34682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34684         * tree-vect-stmts.c (vect_build_all_ones_mask)
34685         (vect_build_zero_merge_argument): New functions, split out from...
34686         (vectorizable_load): ...here.
34688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34690         * tree-vect-stmts.c (vect_check_store_rhs): New function,
34691         split out from...
34692         (vectorizable_mask_load_store): ...here.
34693         (vectorizable_store): ...and here.
34695 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34697         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
34698         split out from...
34699         (vectorizable_mask_load_store): ...here.
34701 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34703         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
34704         (vect_model_store_cost): Take a vec_load_store_type instead of a
34705         vect_def_type.
34706         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
34707         (vect_model_store_cost): Take a vec_load_store_type instead of a
34708         vect_def_type.
34709         (vectorizable_mask_load_store): Update accordingly.
34710         (vectorizable_store): Likewise.
34711         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
34713 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34715         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
34716         IFN_MASK_LOAD calls here rather than...
34717         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
34719 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34720             Alan Hayward  <alan.hayward@arm.com>
34721             David Sherwood  <david.sherwood@arm.com>
34723         * expmed.c (extract_bit_field_1): For vector extracts,
34724         fall back to extract_bit_field_as_subreg if vec_extract
34725         isn't available.
34727 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34728             Alan Hayward  <alan.hayward@arm.com>
34729             David Sherwood  <david.sherwood@arm.com>
34731         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
34732         they are variable or constant sized.
34733         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
34734         slots for constant-sized data.
34736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34737             Alan Hayward  <alan.hayward@arm.com>
34738             David Sherwood  <david.sherwood@arm.com>
34740         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
34741         handling COND_EXPRs with boolean comparisons, try to find a better
34742         basis for the mask type than the boolean itself.
34744 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34746         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
34747         is calculated and how it can be overridden.
34748         * genmodes.c (max_bitsize_mode_any_mode): New variable.
34749         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
34750         if defined.
34751         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
34752         if nonzero.
34754 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34755             Alan Hayward  <alan.hayward@arm.com>
34756             David Sherwood  <david.sherwood@arm.com>
34758         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
34759         Remove the mode argument.
34760         (aarch64_simd_valid_immediate): Remove the mode and inverse
34761         arguments.
34762         * config/aarch64/iterators.md (bitsize): New iterator.
34763         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
34764         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
34765         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
34766         aarch64_simd_valid_immediate.
34767         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
34768         (aarch64_reg_or_bic_imm): Likewise.
34769         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
34770         with an insn_type enum and msl with a modifier_type enum.
34771         Replace element_width with a scalar_mode.  Change the shift
34772         to unsigned int.  Add constructors for scalar_float_mode and
34773         scalar_int_mode elements.
34774         (aarch64_vect_float_const_representable_p): Delete.
34775         (aarch64_can_const_movi_rtx_p)
34776         (aarch64_simd_scalar_immediate_valid_for_move)
34777         (aarch64_simd_make_constant): Update call to
34778         aarch64_simd_valid_immediate.
34779         (aarch64_advsimd_valid_immediate_hs): New function.
34780         (aarch64_advsimd_valid_immediate): Likewise.
34781         (aarch64_simd_valid_immediate): Remove mode and inverse
34782         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
34783         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
34784         and aarch64_float_const_representable_p on the result.
34785         (aarch64_output_simd_mov_immediate): Remove mode argument.
34786         Update call to aarch64_simd_valid_immediate and use of
34787         simd_immediate_info.
34788         (aarch64_output_scalar_simd_mov_immediate): Update call
34789         accordingly.
34791 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34792             Alan Hayward  <alan.hayward@arm.com>
34793             David Sherwood  <david.sherwood@arm.com>
34795         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
34796         (mode_nunits): Likewise CONST_MODE_NUNITS.
34797         * machmode.def (ADJUST_NUNITS): Document.
34798         * genmodes.c (mode_data::need_nunits_adj): New field.
34799         (blank_mode): Update accordingly.
34800         (adj_nunits): New variable.
34801         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
34802         parameter.
34803         (emit_mode_size_inline): Set need_bytesize_adj for all modes
34804         listed in adj_nunits.
34805         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
34806         listed in adj_nunits.  Don't emit case statements for such modes.
34807         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
34808         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
34809         nothing if adj_nunits is nonnull.
34810         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
34811         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
34812         (emit_mode_fbit): Update use of print_maybe_const_decl.
34813         (emit_move_size): Likewise.  Treat the array as non-const
34814         if adj_nunits.
34815         (emit_mode_adjustments): Handle adj_nunits.
34817 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34819         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
34820         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
34821         (VECTOR_MODES): Use it.
34822         (make_vector_modes): Take the prefix as an argument.
34824 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34825             Alan Hayward  <alan.hayward@arm.com>
34826             David Sherwood  <david.sherwood@arm.com>
34828         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
34829         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
34830         for MODE_VECTOR_BOOL.
34831         * machmode.def (VECTOR_BOOL_MODE): Document.
34832         * genmodes.c (VECTOR_BOOL_MODE): New macro.
34833         (make_vector_bool_mode): New function.
34834         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
34835         MODE_VECTOR_BOOL.
34836         * lto-streamer-in.c (lto_input_mode_table): Likewise.
34837         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
34838         Likewise.
34839         * stor-layout.c (int_mode_for_mode): Likewise.
34840         * tree.c (build_vector_type_for_mode): Likewise.
34841         * varasm.c (output_constant_pool_2): Likewise.
34842         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
34843         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
34844         for MODE_VECTOR_BOOL.
34845         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
34846         of mode class checks.
34847         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
34848         instead of a list of mode class checks.
34849         (expand_vector_scalar_condition): Likewise.
34850         (type_for_widest_vector_mode): Handle BImode as an inner mode.
34852 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34853             Alan Hayward  <alan.hayward@arm.com>
34854             David Sherwood  <david.sherwood@arm.com>
34856         * machmode.h (mode_size): Change from unsigned short to
34857         poly_uint16_pod.
34858         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
34859         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34860         or if measurement_type is not polynomial.
34861         (fixed_size_mode::includes_p): Check for constant-sized modes.
34862         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
34863         return a poly_uint16 rather than an unsigned short.
34864         (emit_mode_size): Change the type of mode_size from unsigned short
34865         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
34866         (emit_mode_adjustments): Cope with polynomial vector sizes.
34867         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34868         for GET_MODE_SIZE.
34869         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34870         for GET_MODE_SIZE.
34871         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
34872         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
34873         * caller-save.c (setup_save_areas): Likewise.
34874         (replace_reg_with_saved_mem): Likewise.
34875         * calls.c (emit_library_call_value_1): Likewise.
34876         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
34877         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
34878         (gen_lowpart_for_combine): Likewise.
34879         * convert.c (convert_to_integer_1): Likewise.
34880         * cse.c (equiv_constant, cse_insn): Likewise.
34881         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
34882         (cselib_subst_to_values): Likewise.
34883         * dce.c (word_dce_process_block): Likewise.
34884         * df-problems.c (df_word_lr_mark_ref): Likewise.
34885         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
34886         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
34887         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
34888         (rtl_for_decl_location): Likewise.
34889         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
34890         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
34891         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
34892         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
34893         (expand_expr_real_1): Likewise.
34894         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
34895         (pad_below): Likewise.
34896         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34897         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
34898         * ira.c (get_subreg_tracking_sizes): Likewise.
34899         * ira-build.c (ira_create_allocno_objects): Likewise.
34900         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
34901         (ira_sort_regnos_for_alter_reg): Likewise.
34902         * ira-costs.c (record_operand_costs): Likewise.
34903         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
34904         (resolve_simple_move): Likewise.
34905         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
34906         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
34907         (lra_constraints): Likewise.
34908         (CONST_POOL_OK_P): Reject variable-sized modes.
34909         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
34910         (add_pseudo_to_slot, lra_spill): Likewise.
34911         * omp-low.c (omp_clause_aligned_alignment): Likewise.
34912         * optabs-query.c (get_best_extraction_insn): Likewise.
34913         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34914         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
34915         (expand_mult_highpart, valid_multiword_target_p): Likewise.
34916         * recog.c (offsettable_address_addr_space_p): Likewise.
34917         * regcprop.c (maybe_mode_change): Likewise.
34918         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
34919         * regrename.c (build_def_use): Likewise.
34920         * regstat.c (dump_reg_info): Likewise.
34921         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
34922         (find_reloads, find_reloads_subreg_address): Likewise.
34923         * reload1.c (eliminate_regs_1): Likewise.
34924         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
34925         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
34926         (simplify_binary_operation_1, simplify_subreg): Likewise.
34927         * targhooks.c (default_function_arg_padding): Likewise.
34928         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
34929         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
34930         (verify_gimple_assign_ternary): Likewise.
34931         * tree-inline.c (estimate_move_cost): Likewise.
34932         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34933         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
34934         (get_address_cost_ainc): Likewise.
34935         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
34936         (vect_supportable_dr_alignment): Likewise.
34937         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34938         (vectorizable_reduction): Likewise.
34939         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
34940         (vectorizable_operation, vectorizable_load): Likewise.
34941         * tree.c (build_same_sized_truth_vector_type): Likewise.
34942         * valtrack.c (cleanup_auto_inc_dec): Likewise.
34943         * var-tracking.c (emit_note_insn_var_location): Likewise.
34944         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
34945         (ADDR_VEC_ALIGN): Likewise.
34947 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34948             Alan Hayward  <alan.hayward@arm.com>
34949             David Sherwood  <david.sherwood@arm.com>
34951         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
34952         unsigned short.
34953         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34954         or if measurement_type is polynomial.
34955         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
34956         * combine.c (make_extraction): Likewise.
34957         * dse.c (find_shift_sequence): Likewise.
34958         * dwarf2out.c (mem_loc_descriptor): Likewise.
34959         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
34960         (extract_bit_field, extract_low_bits): Likewise.
34961         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
34962         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
34963         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
34964         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
34965         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34966         * reload.c (find_reloads): Likewise.
34967         * reload1.c (alter_reg): Likewise.
34968         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
34969         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
34970         * tree-if-conv.c (predicate_mem_writes): Likewise.
34971         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
34972         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
34973         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
34974         * valtrack.c (dead_debug_insert_temp): Likewise.
34975         * varasm.c (mergeable_constant_section): Likewise.
34976         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
34978 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34979             Alan Hayward  <alan.hayward@arm.com>
34980             David Sherwood  <david.sherwood@arm.com>
34982         * expr.c (expand_assignment): Cope with polynomial mode sizes
34983         when assigning to a CONCAT.
34985 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34986             Alan Hayward  <alan.hayward@arm.com>
34987             David Sherwood  <david.sherwood@arm.com>
34989         * machmode.h (mode_precision): Change from unsigned short to
34990         poly_uint16_pod.
34991         (mode_to_precision): Return a poly_uint16 rather than an unsigned
34992         short.
34993         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
34994         or if measurement_type is not polynomial.
34995         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
34996         in which the mode is already known to be a scalar_int_mode.
34997         * genmodes.c (emit_mode_precision): Change the type of mode_precision
34998         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
34999         initializer.
35000         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
35001         for GET_MODE_PRECISION.
35002         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
35003         for GET_MODE_PRECISION.
35004         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
35005         as polynomial.
35006         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
35007         (expand_field_assignment, make_extraction): Likewise.
35008         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
35009         (get_last_value): Likewise.
35010         * convert.c (convert_to_integer_1): Likewise.
35011         * cse.c (cse_insn): Likewise.
35012         * expr.c (expand_expr_real_1): Likewise.
35013         * lra-constraints.c (simplify_operand_subreg): Likewise.
35014         * optabs-query.c (can_atomic_load_p): Likewise.
35015         * optabs.c (expand_atomic_load): Likewise.
35016         (expand_atomic_store): Likewise.
35017         * ree.c (combine_reaching_defs): Likewise.
35018         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
35019         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
35020         * tree.h (type_has_mode_precision_p): Likewise.
35021         * ubsan.c (instrument_si_overflow): Likewise.
35023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35024             Alan Hayward  <alan.hayward@arm.com>
35025             David Sherwood  <david.sherwood@arm.com>
35027         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
35028         polynomial numbers of units.
35029         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
35030         (valid_vector_subparts_p): New function.
35031         (build_vector_type): Remove temporary shim and take the number
35032         of units as a poly_uint64 rather than an int.
35033         (build_opaque_vector_type): Take the number of units as a
35034         poly_uint64 rather than an int.
35035         * tree.c (build_vector_from_ctor): Handle polynomial
35036         TYPE_VECTOR_SUBPARTS.
35037         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
35038         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
35039         (build_vector_from_val): If the number of units is variable,
35040         use build_vec_duplicate_cst for constant operands and
35041         VEC_DUPLICATE_EXPR otherwise.
35042         (make_vector_type): Remove temporary is_constant ().
35043         (build_vector_type, build_opaque_vector_type): Take the number of
35044         units as a poly_uint64 rather than an int.
35045         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
35046         VECTOR_CST_NELTS.
35047         * cfgexpand.c (expand_debug_expr): Likewise.
35048         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
35049         (store_constructor, expand_expr_real_1): Likewise.
35050         (const_scalar_mask_from_tree): Likewise.
35051         * fold-const-call.c (fold_const_reduction): Likewise.
35052         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
35053         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
35054         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
35055         (fold_relational_const): Likewise.
35056         (native_interpret_vector): Likewise.  Change the size from an
35057         int to an unsigned int.
35058         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
35059         TYPE_VECTOR_SUBPARTS.
35060         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
35061         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
35062         duplicating a non-constant operand into a variable-length vector.
35063         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
35064         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
35065         * ipa-icf.c (sem_variable::equals): Likewise.
35066         * match.pd: Likewise.
35067         * omp-simd-clone.c (simd_clone_subparts): Likewise.
35068         * print-tree.c (print_node): Likewise.
35069         * stor-layout.c (layout_type): Likewise.
35070         * targhooks.c (default_builtin_vectorization_cost): Likewise.
35071         * tree-cfg.c (verify_gimple_comparison): Likewise.
35072         (verify_gimple_assign_binary): Likewise.
35073         (verify_gimple_assign_ternary): Likewise.
35074         (verify_gimple_assign_single): Likewise.
35075         * tree-pretty-print.c (dump_generic_node): Likewise.
35076         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35077         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
35078         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
35079         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
35080         (vect_shift_permute_load_chain): Likewise.
35081         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
35082         (expand_vector_condition, optimize_vector_constructor): Likewise.
35083         (lower_vec_perm, get_compute_type): Likewise.
35084         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
35085         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
35086         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
35087         (vect_recog_mask_conversion_pattern): Likewise.
35088         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
35089         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
35090         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35091         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
35092         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
35093         (vectorizable_shift, vectorizable_operation, vectorizable_store)
35094         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
35095         (supportable_widening_operation): Likewise.
35096         (supportable_narrowing_operation): Likewise.
35097         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
35098         Likewise.
35099         * varasm.c (output_constant): Likewise.
35101 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35102             Alan Hayward  <alan.hayward@arm.com>
35103             David Sherwood  <david.sherwood@arm.com>
35105         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
35106         so that both the length == 3 and length != 3 cases set up their
35107         own permute vectors.  Add comments explaining why we know the
35108         number of elements is constant.
35109         (vect_permute_load_chain): Likewise.
35111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35112             Alan Hayward  <alan.hayward@arm.com>
35113             David Sherwood  <david.sherwood@arm.com>
35115         * machmode.h (mode_nunits): Change from unsigned char to
35116         poly_uint16_pod.
35117         (ONLY_FIXED_SIZE_MODES): New macro.
35118         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
35119         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
35120         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
35121         New typedefs.
35122         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
35123         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
35124         or if measurement_type is not polynomial.
35125         * genmodes.c (ZERO_COEFFS): New macro.
35126         (emit_mode_nunits_inline): Make mode_nunits_inline return a
35127         poly_uint16.
35128         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
35129         Use ZERO_COEFFS when emitting initializers.
35130         * data-streamer.h (bp_pack_poly_value): New function.
35131         (bp_unpack_poly_value): Likewise.
35132         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
35133         for GET_MODE_NUNITS.
35134         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
35135         for GET_MODE_NUNITS.
35136         * tree.c (make_vector_type): Remove temporary shim and make
35137         the real function take the number of units as a poly_uint64
35138         rather than an int.
35139         (build_vector_type_for_mode): Handle polynomial nunits.
35140         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
35141         * emit-rtl.c (const_vec_series_p_1): Likewise.
35142         (gen_rtx_CONST_VECTOR): Likewise.
35143         * fold-const.c (test_vec_duplicate_folding): Likewise.
35144         * genrecog.c (validate_pattern): Likewise.
35145         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
35146         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
35147         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
35148         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
35149         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
35150         * rtlanal.c (subreg_get_info): Likewise.
35151         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35152         (vect_grouped_load_supported): Likewise.
35153         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
35154         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
35155         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
35156         (simplify_const_unary_operation, simplify_binary_operation_1)
35157         (simplify_const_binary_operation, simplify_ternary_operation)
35158         (test_vector_ops_duplicate, test_vector_ops): Likewise.
35159         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
35160         instead of CONST_VECTOR_NUNITS.
35161         * varasm.c (output_constant_pool_2): Likewise.
35162         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
35163         explicit-encoded elements in the XVEC for variable-length vectors.
35165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35167         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
35169 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35170             Alan Hayward  <alan.hayward@arm.com>
35171             David Sherwood  <david.sherwood@arm.com>
35173         * coretypes.h (fixed_size_mode): Declare.
35174         (fixed_size_mode_pod): New typedef.
35175         * builtins.h (target_builtins::x_apply_args_mode)
35176         (target_builtins::x_apply_result_mode): Change type to
35177         fixed_size_mode_pod.
35178         * builtins.c (apply_args_size, apply_result_size, result_vector)
35179         (expand_builtin_apply_args_1, expand_builtin_apply)
35180         (expand_builtin_return): Update accordingly.
35182 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35184         * cse.c (hash_rtx_cb): Hash only the encoded elements.
35185         * cselib.c (cselib_hash_rtx): Likewise.
35186         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
35187         CONST_VECTOR encoding.
35189 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
35190             Jeff Law  <law@redhat.com>
35192         PR target/83641
35193         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
35194         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
35195         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
35196         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
35198         PR target/83641
35199         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
35200         explicitly probe *sp in a noreturn function if there were any callee
35201         register saves or frame pointer is needed.
35203 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
35205         PR debug/83621
35206         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
35207         BLKmode for ternary, binary or unary expressions.
35209         PR debug/83645
35210         * var-tracking.c (delete_vta_debug_insn): New inline function.
35211         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
35212         insns from get_insns () to NULL instead of each bb separately.
35213         Use delete_vta_debug_insn.  No longer static.
35214         (vt_debug_insns_local, variable_tracking_main_1): Adjust
35215         delete_vta_debug_insns callers.
35216         * rtl.h (delete_vta_debug_insns): Declare.
35217         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
35218         instead of variable_tracking_main.
35220 2018-01-03  Martin Sebor  <msebor@redhat.com>
35222         PR tree-optimization/83603
35223         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
35224         arguments past the endof the argument list in functions declared
35225         without a prototype.
35226         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
35227         Avoid checking when arguments are null.
35229 2018-01-03  Martin Sebor  <msebor@redhat.com>
35231         PR c/83559
35232         * doc/extend.texi (attribute const): Fix a typo.
35233         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
35234         issuing -Wsuggest-attribute for void functions.
35236 2018-01-03  Martin Sebor  <msebor@redhat.com>
35238         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
35239         offset_int::from instead of wide_int::to_shwi.
35240         (maybe_diag_overlap): Remove assertion.
35241         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
35242         * gimple-ssa-sprintf.c (format_directive): Same.
35243         (parse_directive): Same.
35244         (sprintf_dom_walker::compute_format_length): Same.
35245         (try_substitute_return_value): Same.
35247 2018-01-03  Jeff Law  <law@redhat.com>
35249         PR middle-end/83654
35250         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
35251         non-constant residual for zero at runtime and avoid probing in
35252         that case.  Reorganize code for trailing problem to mirror handling
35253         of the residual.
35255 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
35257         PR tree-optimization/83501
35258         * tree-ssa-strlen.c (get_string_cst): New.
35259         (handle_char_store): Call get_string_cst.
35261 2018-01-03  Martin Liska  <mliska@suse.cz>
35263         PR tree-optimization/83593
35264         * tree-ssa-strlen.c: Include tree-cfg.h.
35265         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
35266         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
35267         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
35268         to false.
35269         (strlen_dom_walker::before_dom_children): Call
35270         gimple_purge_dead_eh_edges. Dump tranformation with details
35271         dump flags.
35272         (strlen_dom_walker::before_dom_children): Update call by adding
35273         new argument cleanup_eh.
35274         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
35276 2018-01-03  Martin Liska  <mliska@suse.cz>
35278         PR ipa/83549
35279         * cif-code.def (VARIADIC_THUNK): New enum value.
35280         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
35281         thunks.
35283 2018-01-03  Jan Beulich  <jbeulich@suse.com>
35285         * sse.md (mov<mode>_internal): Tighten condition for when to use
35286         vmovdqu<ssescalarsize> for TI and OI modes.
35288 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
35290         Update copyright years.
35292 2018-01-03  Martin Liska  <mliska@suse.cz>
35294         PR ipa/83594
35295         * ipa-visibility.c (function_and_variable_visibility): Skip
35296         functions with noipa attribure.
35298 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
35300         * gcc.c (process_command): Update copyright notice dates.
35301         * gcov-dump.c (print_version): Ditto.
35302         * gcov.c (print_version): Ditto.
35303         * gcov-tool.c (print_version): Ditto.
35304         * gengtype.c (create_file): Ditto.
35305         * doc/cpp.texi: Bump @copying's copyright year.
35306         * doc/cppinternals.texi: Ditto.
35307         * doc/gcc.texi: Ditto.
35308         * doc/gccint.texi: Ditto.
35309         * doc/gcov.texi: Ditto.
35310         * doc/install.texi: Ditto.
35311         * doc/invoke.texi: Ditto.
35313 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35315         * vector-builder.h (vector_builder::m_full_nelts): Change from
35316         unsigned int to poly_uint64.
35317         (vector_builder::full_nelts): Update prototype accordingly.
35318         (vector_builder::new_vector): Likewise.
35319         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
35320         (vector_builder::operator ==): Likewise.
35321         (vector_builder::finalize): Likewise.
35322         * int-vector-builder.h (int_vector_builder::int_vector_builder):
35323         Take the number of elements as a poly_uint64 rather than an
35324         unsigned int.
35325         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
35326         from unsigned int to poly_uint64.
35327         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
35328         (vec_perm_indices::new_vector): Likewise.
35329         (vec_perm_indices::length): Likewise.
35330         (vec_perm_indices::nelts_per_input): Likewise.
35331         (vec_perm_indices::input_nelts): Likewise.
35332         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
35333         number of elements per input as a poly_uint64 rather than an
35334         unsigned int.  Use the original encoding for variable-length
35335         vectors, rather than clamping each individual element.
35336         For the second and subsequent elements in each pattern,
35337         clamp the step and base before clamping their sum.
35338         (vec_perm_indices::series_p): Handle polynomial element counts.
35339         (vec_perm_indices::all_in_range_p): Likewise.
35340         (vec_perm_indices_to_tree): Likewise.
35341         (vec_perm_indices_to_rtx): Likewise.
35342         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
35343         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
35344         (tree_vector_builder::new_binary_operation): Handle polynomial
35345         element counts.  Return false if we need to know the number
35346         of elements at compile time.
35347         * fold-const.c (fold_vec_perm): Punt if the number of elements
35348         isn't known at compile time.
35350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35352         * vec-perm-indices.h (vec_perm_builder): Change element type
35353         from HOST_WIDE_INT to poly_int64.
35354         (vec_perm_indices::element_type): Update accordingly.
35355         (vec_perm_indices::clamp): Handle polynomial element_types.
35356         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35357         (vec_perm_indices::all_in_range_p): Likewise.
35358         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
35359         than shwi trees.
35360         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
35361         polynomial vec_perm_indices element types.
35362         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
35363         * fold-const.c (fold_vec_perm): Likewise.
35364         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
35365         * tree-vect-generic.c (lower_vec_perm): Likewise.
35366         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35367         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
35368         element type to HOST_WIDE_INT.
35370 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35371             Alan Hayward  <alan.hayward@arm.com>
35372             David Sherwood  <david.sherwood@arm.com>
35374         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
35375         rather than an int.  Use plus_constant.
35376         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
35377         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
35379 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35380             Alan Hayward  <alan.hayward@arm.com>
35381             David Sherwood  <david.sherwood@arm.com>
35383         * calls.c (emit_call_1, expand_call): Change struct_value_size from
35384         a HOST_WIDE_INT to a poly_int64.
35386 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35387             Alan Hayward  <alan.hayward@arm.com>
35388             David Sherwood  <david.sherwood@arm.com>
35390         * calls.c (load_register_parameters): Cope with polynomial
35391         mode sizes.  Require a constant size for BLKmode parameters
35392         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
35393         forces a parameter to be padded at the lsb end in order to
35394         fill a complete number of words, require the parameter size
35395         to be ordered wrt UNITS_PER_WORD.
35397 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35398             Alan Hayward  <alan.hayward@arm.com>
35399             David Sherwood  <david.sherwood@arm.com>
35401         * reload1.c (spill_stack_slot_width): Change element type
35402         from unsigned int to poly_uint64_pod.
35403         (alter_reg): Treat mode sizes as polynomial.
35405 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35406             Alan Hayward  <alan.hayward@arm.com>
35407             David Sherwood  <david.sherwood@arm.com>
35409         * reload.c (complex_word_subreg_p): New function.
35410         (reload_inner_reg_of_subreg, push_reload): Use it.
35412 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35413             Alan Hayward  <alan.hayward@arm.com>
35414             David Sherwood  <david.sherwood@arm.com>
35416         * lra-constraints.c (process_alt_operands): Reject matched
35417         operands whose sizes aren't ordered.
35418         (match_reload): Refer to this check here.
35420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35421             Alan Hayward  <alan.hayward@arm.com>
35422             David Sherwood  <david.sherwood@arm.com>
35424         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
35425         that the mode size is in the set {1, 2, 4, 8, 16}.
35427 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35428             Alan Hayward  <alan.hayward@arm.com>
35429             David Sherwood  <david.sherwood@arm.com>
35431         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
35432         Use plus_constant instead of gen_rtx_PLUS.
35434 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35435             Alan Hayward  <alan.hayward@arm.com>
35436             David Sherwood  <david.sherwood@arm.com>
35438         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
35439         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
35440         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
35441         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
35442         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
35443         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
35444         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
35445         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
35446         * config/i386/i386.c (ix86_push_rounding): ...this new function.
35447         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
35448         a poly_int64.
35449         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
35450         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
35451         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
35452         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
35453         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
35454         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
35455         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
35456         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
35457         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
35458         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
35459         function.
35460         * expr.c (emit_move_resolve_push): Treat the input and result
35461         of PUSH_ROUNDING as a poly_int64.
35462         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
35463         (emit_push_insn): Likewise.
35464         * lra-eliminations.c (mark_not_eliminable): Likewise.
35465         * recog.c (push_operand): Likewise.
35466         * reload1.c (elimination_effects): Likewise.
35467         * rtlanal.c (nonzero_bits1): Likewise.
35468         * calls.c (store_one_arg): Likewise.  Require the padding to be
35469         known at compile time.
35471 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35472             Alan Hayward  <alan.hayward@arm.com>
35473             David Sherwood  <david.sherwood@arm.com>
35475         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
35476         Use plus_constant instead of gen_rtx_PLUS.
35478 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35479             Alan Hayward  <alan.hayward@arm.com>
35480             David Sherwood  <david.sherwood@arm.com>
35482         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
35483         rather than an int.
35485 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35486             Alan Hayward  <alan.hayward@arm.com>
35487             David Sherwood  <david.sherwood@arm.com>
35489         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
35490         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
35491         via stack temporaries.  Treat the mode size as polynomial too.
35493 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35494             Alan Hayward  <alan.hayward@arm.com>
35495             David Sherwood  <david.sherwood@arm.com>
35497         * expr.c (expand_expr_real_2): When handling conversions involving
35498         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
35499         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
35500         as a poly_uint64 too.
35502 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35503             Alan Hayward  <alan.hayward@arm.com>
35504             David Sherwood  <david.sherwood@arm.com>
35506         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
35508 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35509             Alan Hayward  <alan.hayward@arm.com>
35510             David Sherwood  <david.sherwood@arm.com>
35512         * combine.c (can_change_dest_mode): Handle polynomial
35513         REGMODE_NATURAL_SIZE.
35514         * expmed.c (store_bit_field_1): Likewise.
35515         * expr.c (store_constructor): Likewise.
35516         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
35517         and polynomial REGMODE_NATURAL_SIZE.
35518         (gen_lowpart_common): Likewise.
35519         * reginfo.c (record_subregs_of_mode): Likewise.
35520         * rtlanal.c (read_modify_subreg_p): Likewise.
35522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35523             Alan Hayward  <alan.hayward@arm.com>
35524             David Sherwood  <david.sherwood@arm.com>
35526         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
35527         numbers of elements.
35529 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35530             Alan Hayward  <alan.hayward@arm.com>
35531             David Sherwood  <david.sherwood@arm.com>
35533         * match.pd: Cope with polynomial numbers of vector elements.
35535 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35536             Alan Hayward  <alan.hayward@arm.com>
35537             David Sherwood  <david.sherwood@arm.com>
35539         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
35540         in a POINTER_PLUS_EXPR.
35542 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35543             Alan Hayward  <alan.hayward@arm.com>
35544             David Sherwood  <david.sherwood@arm.com>
35546         * omp-simd-clone.c (simd_clone_subparts): New function.
35547         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
35548         (ipa_simd_modify_function_body): Likewise.
35550 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35551             Alan Hayward  <alan.hayward@arm.com>
35552             David Sherwood  <david.sherwood@arm.com>
35554         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
35555         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
35556         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
35557         (expand_vector_condition, vector_element): Likewise.
35558         (subparts_gt): New function.
35559         (get_compute_type): Use subparts_gt.
35560         (count_type_subparts): Delete.
35561         (expand_vector_operations_1): Use subparts_gt instead of
35562         count_type_subparts.
35564 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35565             Alan Hayward  <alan.hayward@arm.com>
35566             David Sherwood  <david.sherwood@arm.com>
35568         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
35569         (vect_compile_time_alias): ...this new function.  Do the calculation
35570         on poly_ints rather than trees.
35571         (vect_prune_runtime_alias_test_list): Update call accordingly.
35573 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35574             Alan Hayward  <alan.hayward@arm.com>
35575             David Sherwood  <david.sherwood@arm.com>
35577         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
35578         numbers of units.
35579         (vect_schedule_slp_instance): Likewise.
35581 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35582             Alan Hayward  <alan.hayward@arm.com>
35583             David Sherwood  <david.sherwood@arm.com>
35585         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
35586         constant and extern definitions for variable-length vectors.
35587         (vect_get_constant_vectors): Note that the number of units
35588         is known to be constant.
35590 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35591             Alan Hayward  <alan.hayward@arm.com>
35592             David Sherwood  <david.sherwood@arm.com>
35594         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
35595         of units as polynomial.  Choose between WIDE and NARROW based
35596         on multiple_p.
35598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35599             Alan Hayward  <alan.hayward@arm.com>
35600             David Sherwood  <david.sherwood@arm.com>
35602         * tree-vect-stmts.c (simd_clone_subparts): New function.
35603         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
35605 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35606             Alan Hayward  <alan.hayward@arm.com>
35607             David Sherwood  <david.sherwood@arm.com>
35609         * tree-vect-stmts.c (vectorizable_call): Treat the number of
35610         vectors as polynomial.  Use build_index_vector for
35611         IFN_GOMP_SIMD_LANE.
35613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35614             Alan Hayward  <alan.hayward@arm.com>
35615             David Sherwood  <david.sherwood@arm.com>
35617         * tree-vect-stmts.c (get_load_store_type): Treat the number of
35618         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
35619         for variable-length vectors.
35620         (vectorizable_mask_load_store): Treat the number of units as
35621         polynomial, asserting that it is constant if the condition has
35622         already been enforced.
35623         (vectorizable_store, vectorizable_load): Likewise.
35625 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35626             Alan Hayward  <alan.hayward@arm.com>
35627             David Sherwood  <david.sherwood@arm.com>
35629         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
35630         of units as polynomial.  Punt if we can't tell at compile time
35631         which vector contains the final result.
35633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35634             Alan Hayward  <alan.hayward@arm.com>
35635             David Sherwood  <david.sherwood@arm.com>
35637         * tree-vect-loop.c (vectorizable_induction): Treat the number
35638         of units as polynomial.  Punt on SLP inductions.  Use an integer
35639         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
35640         cast of such a series for variable-length floating-point
35641         reductions.
35643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35644             Alan Hayward  <alan.hayward@arm.com>
35645             David Sherwood  <david.sherwood@arm.com>
35647         * tree.h (build_index_vector): Declare.
35648         * tree.c (build_index_vector): New function.
35649         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
35650         of units as polynomial, forcibly converting it to a constant if
35651         vectorizable_reduction has already enforced the condition.
35652         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
35653         to create a {1,2,3,...} vector.
35654         (vectorizable_reduction): Treat the number of units as polynomial.
35655         Choose vectype_in based on the largest scalar element size rather
35656         than the smallest number of units.  Enforce the restrictions
35657         relied on above.
35659 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35660             Alan Hayward  <alan.hayward@arm.com>
35661             David Sherwood  <david.sherwood@arm.com>
35663         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
35664         number of units as polynomial.
35666 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35667             Alan Hayward  <alan.hayward@arm.com>
35668             David Sherwood  <david.sherwood@arm.com>
35670         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
35671         * target.def (autovectorize_vector_sizes): Return the vector sizes
35672         by pointer, using vector_sizes rather than a bitmask.
35673         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
35674         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
35675         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
35676         Likewise.
35677         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
35678         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
35679         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
35680         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
35681         * omp-general.c (omp_max_vf): Likewise.
35682         * omp-low.c (omp_clause_aligned_alignment): Likewise.
35683         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
35684         * tree-vect-loop.c (vect_analyze_loop): Likewise.
35685         * tree-vect-slp.c (vect_slp_bb): Likewise.
35686         * doc/tm.texi: Regenerate.
35687         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
35688         to a poly_uint64.
35689         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
35690         the vector size as a poly_uint64 rather than an unsigned int.
35691         (current_vector_size): Change from an unsigned int to a poly_uint64.
35692         (get_vectype_for_scalar_type): Update accordingly.
35693         * tree.h (build_truth_vector_type): Take the size and number of
35694         units as a poly_uint64 rather than an unsigned int.
35695         (build_vector_type): Add a temporary overload that takes
35696         the number of units as a poly_uint64 rather than an unsigned int.
35697         * tree.c (make_vector_type): Likewise.
35698         (build_truth_vector_type): Take the number of units as a poly_uint64
35699         rather than an unsigned int.
35701 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35702             Alan Hayward  <alan.hayward@arm.com>
35703             David Sherwood  <david.sherwood@arm.com>
35705         * target.def (get_mask_mode): Take the number of units and length
35706         as poly_uint64s rather than unsigned ints.
35707         * targhooks.h (default_get_mask_mode): Update accordingly.
35708         * targhooks.c (default_get_mask_mode): Likewise.
35709         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
35710         * doc/tm.texi: Regenerate.
35712 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35713             Alan Hayward  <alan.hayward@arm.com>
35714             David Sherwood  <david.sherwood@arm.com>
35716         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
35717         * omp-general.c (omp_max_vf): Likewise.
35718         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
35719         (expand_omp_simd): Handle polynomial safelen.
35720         * omp-low.c (omplow_simd_context): Add a default constructor.
35721         (omplow_simd_context::max_vf): Change from int to poly_uint64.
35722         (lower_rec_simd_input_clauses): Update accordingly.
35723         (lower_rec_input_clauses): Likewise.
35725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35726             Alan Hayward  <alan.hayward@arm.com>
35727             David Sherwood  <david.sherwood@arm.com>
35729         * tree-vectorizer.h (vect_nunits_for_cost): New function.
35730         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
35731         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
35732         (vect_analyze_slp_cost): Likewise.
35733         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
35734         (vect_model_load_cost): Likewise.
35736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35737             Alan Hayward  <alan.hayward@arm.com>
35738             David Sherwood  <david.sherwood@arm.com>
35740         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
35741         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
35742         from an unsigned int * to a poly_uint64_pod *.
35743         (calculate_unrolling_factor): New function.
35744         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
35746 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35747             Alan Hayward  <alan.hayward@arm.com>
35748             David Sherwood  <david.sherwood@arm.com>
35750         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
35751         from an unsigned int to a poly_uint64.
35752         (_loop_vec_info::slp_unrolling_factor): Likewise.
35753         (_loop_vec_info::vectorization_factor): Change from an int
35754         to a poly_uint64.
35755         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
35756         (vect_get_num_vectors): New function.
35757         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
35758         (vect_get_num_copies): Use vect_get_num_vectors.
35759         (vect_analyze_data_ref_dependences): Change max_vf from an int *
35760         to an unsigned int *.
35761         (vect_analyze_data_refs): Change min_vf from an int * to a
35762         poly_uint64 *.
35763         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35764         than an unsigned HOST_WIDE_INT.
35765         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
35766         (vect_analyze_data_ref_dependence): Change max_vf from an int *
35767         to an unsigned int *.
35768         (vect_analyze_data_ref_dependences): Likewise.
35769         (vect_compute_data_ref_alignment): Handle polynomial vf.
35770         (vect_enhance_data_refs_alignment): Likewise.
35771         (vect_prune_runtime_alias_test_list): Likewise.
35772         (vect_shift_permute_load_chain): Likewise.
35773         (vect_supportable_dr_alignment): Likewise.
35774         (dependence_distance_ge_vf): Take the vectorization factor as a
35775         poly_uint64 rather than an unsigned HOST_WIDE_INT.
35776         (vect_analyze_data_refs): Change min_vf from an int * to a
35777         poly_uint64 *.
35778         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
35779         vfm1 as a poly_uint64 rather than an int.  Make the same change
35780         for the returned bound_scalar.
35781         (vect_gen_vector_loop_niters): Handle polynomial vf.
35782         (vect_do_peeling): Likewise.  Update call to
35783         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
35784         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
35785         be constant.
35786         * tree-vect-loop.c (vect_determine_vectorization_factor)
35787         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
35788         (vect_get_known_peeling_cost): Likewise.
35789         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
35790         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
35791         (vect_transform_loop): Likewise.  Use the lowest possible VF when
35792         updating the upper bounds of the loop.
35793         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
35794         rather than an int.
35795         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
35796         polynomial unroll factors.
35797         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
35798         (vect_make_slp_decision): Likewise.
35799         (vect_supported_load_permutation_p): Likewise, and polynomial
35800         vf too.
35801         (vect_analyze_slp_cost): Handle polynomial vf.
35802         (vect_slp_analyze_node_operations): Likewise.
35803         (vect_slp_analyze_bb_1): Likewise.
35804         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35805         than an unsigned HOST_WIDE_INT.
35806         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
35807         (vectorizable_load): Handle polynomial vf.
35808         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
35809         a poly_uint64.
35810         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
35812 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35813             Alan Hayward  <alan.hayward@arm.com>
35814             David Sherwood  <david.sherwood@arm.com>
35816         * match.pd: Handle bit operations involving three constants
35817         and try to fold one pair.
35819 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35821         * tree-vect-loop-manip.c: Include gimple-fold.h.
35822         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
35823         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
35824         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
35825         Add a path that uses a step of VF instead of 1, but disable it
35826         for now.
35827         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
35828         and niters_no_overflow parameters.  Update calls to
35829         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
35830         Create a new SSA name if the latter choses to use a ste other
35831         than zero, and return it via niters_vector_mult_vf_var.
35832         * tree-vect-loop.c (vect_transform_loop): Update calls to
35833         vect_do_peeling, vect_gen_vector_loop_niters and
35834         slpeel_make_loop_iterate_ntimes.
35835         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
35836         (vect_gen_vector_loop_niters): Update declarations after above changes.
35838 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
35840         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
35841         128-bit round to integer instructions.
35842         (ceil<mode>2): Likewise.
35843         (btrunc<mode>2): Likewise.
35844         (round<mode>2): Likewise.
35846 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35848         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
35849         unaligned VSX load/store on P8/P9.
35850         (expand_block_clear): Allow the use of unaligned VSX
35851         load/store on P8/P9.
35853 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35855         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
35856         New function.
35857         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
35858         swap associated with both a load and a store.
35860 2018-01-02  Andrew Waterman  <andrew@sifive.com>
35862         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
35863         * config/riscv/riscv.md (clear_cache): Use it.
35865 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
35867         * web.c: Remove out-of-date comment.
35869 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35871         * expr.c (fixup_args_size_notes): Check that any existing
35872         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
35873         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
35874         (emit_single_push_insn): ...here.
35876 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35878         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
35879         (const_vector_encoded_nelts): New function.
35880         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
35881         (const_vector_int_elt, const_vector_elt): Declare.
35882         * emit-rtl.c (const_vector_int_elt_1): New function.
35883         (const_vector_elt): Likewise.
35884         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
35885         of CONST_VECTOR_ELT.
35887 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35889         * expr.c: Include rtx-vector-builder.h.
35890         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
35891         directly on the tree encoding.
35892         (const_vector_from_tree): Likewise.
35893         * optabs.c: Include rtx-vector-builder.h.
35894         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
35895         sequence of "u" values.
35896         * vec-perm-indices.c: Include rtx-vector-builder.h.
35897         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
35898         directly on the vec_perm_indices encoding.
35900 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35902         * doc/rtl.texi (const_vector): Describe new encoding scheme.
35903         * Makefile.in (OBJS): Add rtx-vector-builder.o.
35904         * rtx-vector-builder.h: New file.
35905         * rtx-vector-builder.c: Likewise.
35906         * rtl.h (rtx_def::u2): Add a const_vector field.
35907         (CONST_VECTOR_NPATTERNS): New macro.
35908         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
35909         (CONST_VECTOR_DUPLICATE_P): Likewise.
35910         (CONST_VECTOR_STEPPED_P): Likewise.
35911         (CONST_VECTOR_ENCODED_ELT): Likewise.
35912         (const_vec_duplicate_p): Check for a duplicated vector encoding.
35913         (unwrap_const_vec_duplicate): Likewise.
35914         (const_vec_series_p): Check for a non-duplicated vector encoding.
35915         Say that the function only returns true for integer vectors.
35916         * emit-rtl.c: Include rtx-vector-builder.h.
35917         (gen_const_vec_duplicate_1): Delete.
35918         (gen_const_vector): Call gen_const_vec_duplicate instead of
35919         gen_const_vec_duplicate_1.
35920         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
35921         (gen_const_vec_duplicate): Use rtx_vector_builder.
35922         (gen_const_vec_series): Likewise.
35923         (gen_rtx_CONST_VECTOR): Likewise.
35924         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
35925         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35926         Build a new vector rather than modifying a CONST_VECTOR in-place.
35927         (handle_special_swappables): Update call accordingly.
35928         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
35929         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35930         Build a new vector rather than modifying a CONST_VECTOR in-place.
35931         (handle_special_swappables): Update call accordingly.
35933 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35935         * simplify-rtx.c (simplify_const_binary_operation): Use
35936         CONST_VECTOR_ELT instead of XVECEXP.
35938 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35940         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
35941         the selector elements to be different from the data elements
35942         if the selector is a VECTOR_CST.
35943         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
35944         ssizetype for the selector.
35946 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35948         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
35949         before testing each element individually.
35950         * tree-vect-generic.c (lower_vec_perm): Likewise.
35952 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35954         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
35955         * selftest-run-tests.c (selftest::run_tests): Call it.
35956         * vector-builder.h (vector_builder::operator ==): New function.
35957         (vector_builder::operator !=): Likewise.
35958         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
35959         (vec_perm_indices::all_from_input_p): New function.
35960         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35961         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
35962         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
35963         instead of reading the VECTOR_CST directly.  Detect whether both
35964         vector inputs are the same before constructing the vec_perm_indices,
35965         and update the number of inputs argument accordingly.  Use the
35966         utility functions added above.  Only construct sel2 if we need to.
35968 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35970         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
35971         the broadcast of the low byte.
35972         (expand_mult_highpart): Use an explicit encoding for the permutes.
35973         * optabs-query.c (can_mult_highpart_p): Likewise.
35974         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
35975         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35976         (vectorizable_bswap): Likewise.
35977         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
35978         explicit encoding for the power-of-2 permutes.
35979         (vect_permute_store_chain): Likewise.
35980         (vect_grouped_load_supported): Likewise.
35981         (vect_permute_load_chain): Likewise.
35983 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35985         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
35986         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
35987         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
35988         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35989         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
35990         (vect_gen_perm_mask_any): Likewise.
35992 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35994         * int-vector-builder.h: New file.
35995         * vec-perm-indices.h: Include int-vector-builder.h.
35996         (vec_perm_indices): Redefine as an int_vector_builder.
35997         (auto_vec_perm_indices): Delete.
35998         (vec_perm_builder): Redefine as a stand-alone class.
35999         (vec_perm_indices::vec_perm_indices): New function.
36000         (vec_perm_indices::clamp): Likewise.
36001         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
36002         (vec_perm_indices::new_vector): New function.
36003         (vec_perm_indices::new_expanded_vector): Update for new
36004         vec_perm_indices class.
36005         (vec_perm_indices::rotate_inputs): New function.
36006         (vec_perm_indices::all_in_range_p): Operate directly on the
36007         encoded form, without computing elided elements.
36008         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
36009         encoding.  Update for new vec_perm_indices class.
36010         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
36011         the given vec_perm_builder.
36012         (expand_vec_perm_var): Update vec_perm_builder constructor.
36013         (expand_mult_highpart): Use vec_perm_builder instead of
36014         auto_vec_perm_indices.
36015         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
36016         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
36017         or double series encoding as appropriate.
36018         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
36019         vec_perm_indices instead of auto_vec_perm_indices.
36020         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
36021         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
36022         (vect_permute_store_chain): Likewise.
36023         (vect_grouped_load_supported): Likewise.
36024         (vect_permute_load_chain): Likewise.
36025         (vect_shift_permute_load_chain): Likewise.
36026         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
36027         (vect_transform_slp_perm_load): Likewise.
36028         (vect_schedule_slp_instance): Likewise.
36029         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
36030         (vectorizable_mask_load_store): Likewise.
36031         (vectorizable_bswap): Likewise.
36032         (vectorizable_store): Likewise.
36033         (vectorizable_load): Likewise.
36034         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
36035         vec_perm_indices instead of auto_vec_perm_indices.  Use
36036         tree_to_vec_perm_builder to read the vector from a tree.
36037         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
36038         vec_perm_builder instead of a vec_perm_indices.
36039         (have_whole_vector_shift): Use vec_perm_builder and
36040         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
36041         truncation to calc_vec_perm_mask_for_shift.
36042         (vect_create_epilog_for_reduction): Likewise.
36043         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
36044         from auto_vec_perm_indices to vec_perm_indices.
36045         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
36046         instead of changing individual elements.
36047         (aarch64_vectorize_vec_perm_const): Use new_vector to install
36048         the vector in d.perm.
36049         * config/arm/arm.c (expand_vec_perm_d::perm): Change
36050         from auto_vec_perm_indices to vec_perm_indices.
36051         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
36052         instead of changing individual elements.
36053         (arm_vectorize_vec_perm_const): Use new_vector to install
36054         the vector in d.perm.
36055         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
36056         Update vec_perm_builder constructor.
36057         (rs6000_expand_interleave): Likewise.
36058         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
36059         (rs6000_expand_interleave): Likewise.
36061 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36063         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
36064         to qimode could truncate the indices.
36065         * optabs.c (expand_vec_perm_var): Likewise.
36067 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36069         * Makefile.in (OBJS): Add vec-perm-indices.o.
36070         * vec-perm-indices.h: New file.
36071         * vec-perm-indices.c: Likewise.
36072         * target.h (vec_perm_indices): Replace with a forward class
36073         declaration.
36074         (auto_vec_perm_indices): Move to vec-perm-indices.h.
36075         * optabs.h: Include vec-perm-indices.h.
36076         (expand_vec_perm): Delete.
36077         (selector_fits_mode_p, expand_vec_perm_var): Declare.
36078         (expand_vec_perm_const): Declare.
36079         * target.def (vec_perm_const_ok): Replace with...
36080         (vec_perm_const): ...this new hook.
36081         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
36082         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
36083         * doc/tm.texi: Regenerate.
36084         * optabs.def (vec_perm_const): Delete.
36085         * doc/md.texi (vec_perm_const): Likewise.
36086         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
36087         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
36088         expand_vec_perm for constant permutation vectors.  Assert that
36089         the mode of variable permutation vectors is the integer equivalent
36090         of the mode that is being permuted.
36091         * optabs-query.h (selector_fits_mode_p): Declare.
36092         * optabs-query.c: Include vec-perm-indices.h.
36093         (selector_fits_mode_p): New function.
36094         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
36095         is defined, instead of checking whether the vec_perm_const_optab
36096         exists.  Use targetm.vectorize.vec_perm_const instead of
36097         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
36098         fit in the vector mode before using a variable permute.
36099         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
36100         vec_perm_indices instead of an rtx.
36101         (expand_vec_perm): Replace with...
36102         (expand_vec_perm_const): ...this new function.  Take the selector
36103         as a vec_perm_indices rather than an rtx.  Also take the mode of
36104         the selector.  Update call to shift_amt_for_vec_perm_mask.
36105         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
36106         Use vec_perm_indices::new_expanded_vector to expand the original
36107         selector into bytes.  Check whether the indices fit in the vector
36108         mode before using a variable permute.
36109         (expand_vec_perm_var): Make global.
36110         (expand_mult_highpart): Use expand_vec_perm_const.
36111         * fold-const.c: Includes vec-perm-indices.h.
36112         * tree-ssa-forwprop.c: Likewise.
36113         * tree-vect-data-refs.c: Likewise.
36114         * tree-vect-generic.c: Likewise.
36115         * tree-vect-loop.c: Likewise.
36116         * tree-vect-slp.c: Likewise.
36117         * tree-vect-stmts.c: Likewise.
36118         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
36119         Delete.
36120         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
36121         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
36122         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
36123         (aarch64_vectorize_vec_perm_const): ...this new function.
36124         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
36125         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
36126         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
36127         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
36128         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
36129         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
36130         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
36131         into...
36132         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
36133         check for NEON modes.
36134         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
36135         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
36136         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
36137         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
36138         into...
36139         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
36140         the old VEC_PERM_CONST conditions.
36141         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
36142         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
36143         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
36144         (ia64_vectorize_vec_perm_const_ok): Merge into...
36145         (ia64_vectorize_vec_perm_const): ...this new function.
36146         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
36147         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
36148         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
36149         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
36150         * config/mips/mips.c (mips_expand_vec_perm_const)
36151         (mips_vectorize_vec_perm_const_ok): Merge into...
36152         (mips_vectorize_vec_perm_const): ...this new function.
36153         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
36154         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
36155         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
36156         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
36157         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
36158         (rs6000_expand_vec_perm_const): Delete.
36159         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
36160         Delete.
36161         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
36162         (altivec_expand_vec_perm_const_le): Take each operand individually.
36163         Operate on constant selectors rather than rtxes.
36164         (altivec_expand_vec_perm_const): Likewise.  Update call to
36165         altivec_expand_vec_perm_const_le.
36166         (rs6000_expand_vec_perm_const): Delete.
36167         (rs6000_vectorize_vec_perm_const_ok): Delete.
36168         (rs6000_vectorize_vec_perm_const): New function.
36169         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
36170         an element count and rtx array.
36171         (rs6000_expand_extract_even): Update call accordingly.
36172         (rs6000_expand_interleave): Likewise.
36173         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
36174         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
36175         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
36176         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
36177         (rs6000_expand_vec_perm_const): Delete.
36178         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
36179         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
36180         (altivec_expand_vec_perm_const_le): Take each operand individually.
36181         Operate on constant selectors rather than rtxes.
36182         (altivec_expand_vec_perm_const): Likewise.  Update call to
36183         altivec_expand_vec_perm_const_le.
36184         (rs6000_expand_vec_perm_const): Delete.
36185         (rs6000_vectorize_vec_perm_const_ok): Delete.
36186         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
36187         reference to the SPE evmerge intructions.
36188         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
36189         an element count and rtx array.
36190         (rs6000_expand_extract_even): Update call accordingly.
36191         (rs6000_expand_interleave): Likewise.
36192         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
36193         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
36194         new function.
36195         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
36197 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36199         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
36200         vector mode and that that mode matches the mode of the data
36201         being permuted.
36202         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
36203         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
36204         directly using expand_vec_perm_1 when forcing selectors into
36205         registers.
36206         (expand_vec_perm_var): New function, split out from expand_vec_perm.
36208 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36210         * optabs-query.h (can_vec_perm_p): Delete.
36211         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
36212         * optabs-query.c (can_vec_perm_p): Split into...
36213         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
36214         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
36215         particular selector is valid.
36216         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
36217         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
36218         (vect_grouped_load_supported): Likewise.
36219         (vect_shift_permute_load_chain): Likewise.
36220         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
36221         (vect_transform_slp_perm_load): Likewise.
36222         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
36223         (vectorizable_bswap): Likewise.
36224         (vect_gen_perm_mask_checked): Likewise.
36225         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
36226         implementations of variable permutation vectors into account
36227         when deciding which selector to use.
36228         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
36229         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
36230         with a false third argument.
36231         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
36232         to test whether the constant selector is valid and can_vec_perm_var_p
36233         to test whether a variable selector is valid.
36235 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36237         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
36238         * optabs-query.c (can_vec_perm_p): Likewise.
36239         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
36240         instead of vec_perm_indices.
36241         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
36242         (vect_gen_perm_mask_checked): Likewise,
36243         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
36244         (vect_gen_perm_mask_checked): Likewise,
36246 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
36248         * optabs-query.h (qimode_for_vec_perm): Declare.
36249         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
36250         (qimode_for_vec_perm): ...this new function.
36251         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
36253 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36255         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
36256         does not have a conditional at the top.
36258 2018-01-02  Richard Biener  <rguenther@suse.de>
36260         * ipa-inline.c (big_speedup_p): Fix expression.
36262 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
36264         PR target/81616
36265         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
36266         for generic 4->6.
36268 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
36270         PR target/81616
36271         Generic tuning.
36272         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
36273         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
36274         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
36275         cond_taken_branch_cost 3->4.
36277 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
36279         PR tree-optimization/83581
36280         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
36281         TODO_cleanup_cfg if any changes have been made.
36283         PR middle-end/83608
36284         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
36285         convert_modes if target mode has the right side, but different mode
36286         class.
36288         PR middle-end/83609
36289         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
36290         last argument when extracting from CONCAT.  If either from_real or
36291         from_imag is NULL, use expansion through memory.  If result is not
36292         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
36293         the parts directly to inner mode, if even that fails, use expansion
36294         through memory.
36296         PR middle-end/83623
36297         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
36298         check for bswap in mode rather than HImode and use that in expand_unop
36299         too.
36301 Copyright (C) 2018 Free Software Foundation, Inc.
36303 Copying and distribution of this file, with or without modification,
36304 are permitted in any medium without royalty provided the copyright
36305 notice and this notice are preserved.