1 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/69615
4 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5 of a cast from a same precision integral SSA_NAME in a bb dominated
6 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7 cast to utype if rhs2 has already a compatible type.
9 2018-06-07 Richard Biener <rguenther@suse.de>
11 PR tree-optimization/85935
12 * graphite-scop-detection.c (find_params_in_bb): Analyze
13 condition operands with respect to the correct loop. Assert
14 the analysis doesn't fail.
16 2018-06-04 Carl Love <cel@us.ibm.com>
18 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
19 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
20 as it is slightly cheaper.
21 (first_match_or_eos_index_<mode>):
22 Calculate index using natural element order.
23 (first_match_index_<mode>):
24 Calculate index using natural element order.
25 (first_match_or_eos_index_<mode>):
26 Calculate index using natural order.
27 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
29 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
30 P9V_BUILTIN_VCLZLSBB_V16QI.
31 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
34 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
36 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
37 indentation and line wrap for many prototypes. Add missing
38 @smallexample directives around block of prototypes for vec_xl and
41 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
43 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
44 track if we pass or return IEEE 128-bit floating point.
45 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
46 C++ mangling that is compatible with GCC 8.1.
47 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
48 (init_cumulative_args): Note if we pass or return IEEE 128-bit
50 (rs6000_function_arg_advance_1): Likewise.
51 (rs6000_mangle_type): Optionally generate mangled names that match
52 what GCC 8.1 generated for IEEE 128-bit floating point types.
53 (rs6000_globalize_decl_name): If we have an external function that
54 passes or returns IEEE 128-bit types, generate a weak reference
55 from the mangled name used in GCC 8.1 to the current mangled
57 (rs6000_init_builtins): Make __ibm128 use the long double type if
58 long double is IBM extended double. Make __float128 use the long
59 double type if long double is IEEE 128-bit.
62 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
63 macro for __ibm128 built-in functions.
64 (PACK_IF): Add __ibm128 pack/unpack functions.
65 (UNPACK_IF): Likewise.
66 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
67 enable long double built-in functions if long double is IEEE
68 128-bit floating point.
69 (rs6000_invalid_builtin): Update long double built-in function
71 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
72 functions, adjust the built-in function to use the long double
73 built-in function if __ibm128 and long double are the same type.
74 * doc/extend.texi (PowerPC builtins): Update documention for
75 __builtin_{,un}pack_longdouble. Add documentation for
76 __builtin_{,un}pack_ibm128.
78 2018-06-06 Jim Wilson <jimw@sifive.com>
80 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
81 (struct machine_function): New field interrupt_mode.
82 (riscv_handle_type_attribute): New function. Add forward declaration.
83 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
84 (riscv_expand_epilogue): Check interrupt_mode field.
85 (riscv_set_current_function): Check interrupt attribute args and
86 set interrupt_mode field.
87 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
88 (riscv_sret, riscv_uret): New.
89 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
90 new arguments to interrupt attribute.
92 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
95 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
96 Don't handle -mcpu=power8 if -mpower9-vector is also used.
98 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
100 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
101 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
102 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
103 several redundant entries.
105 2018-06-06 David Malcolm <dmalcolm@redhat.com>
107 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
108 type from "rtx" to "rtx_insn *".
109 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
110 for local "call_insn", removing cast.
111 (ix86_expand_call): Likewise, introducing a "call_insn" local.
113 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
115 PR tree-optimization/86066
116 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
117 for BIT_INSERT_EXPR stores.
119 2018-06-06 Richard Biener <rguenther@suse.de>
121 PR tree-optimization/86062
122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
124 of to be offsetted base.
126 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
128 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
129 to be static and remove check on interrupt attribute name.
131 gcc/testsuite/gcc.target/msp430/
132 * function-attributes-4.c: New test.
133 * static-interrupts.c: New test.
135 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
137 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
138 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
140 2018-06-05 Steve Ellcey <sellcey@cavium.com>
143 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
145 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
146 Remove second argument, change how error is called.
147 (aarch64_layout_arg): Remove second argument from
148 aarch64_err_no_fpadvsimd call.
149 (aarch64_init_cumulative_args): Ditto.
150 (aarch64_gimplify_va_arg_expr): Ditto.
151 * config/aarch64/aarch64.md (mov<mode>): Ditto.
153 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
155 * config/i386/i386.md (simple_return_indirect_internal): New expander.
156 (*simple_return_indirect_internal<mode>): Rename from
157 simple_return_indirect_internal. Use W mode iterator.
158 (rstorssp): New expander.
159 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
160 (clrssbsy): New expander.
161 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
163 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
165 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
167 (cmse_check_pointed_object): Likewise.
169 2018-06-05 Martin Liska <mliska@suse.cz>
171 PR gcov-profile/47618
172 * doc/invoke.texi: Document how -fprofile-dir format
175 2018-06-05 Richard Biener <rguenther@suse.de>
177 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
178 removal pretend DOM info isn't available so we do not update
179 it and only remove edges, not dominated blocks. Actually free
180 DOM info in case we removed something. Remove unreachable blocks.
181 (mfb_keep_latches): Work with either DOM info or marked backedges.
182 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
183 first. Mark backedges if DOM info isn't available.
184 (Re-)compute DOM info after cleanup_control_flow_pre.
186 2018-06-05 Richard Biener <rguenther@suse.de>
188 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
189 (locus_discrim_hasher::hash): Adjust.
190 (locus_discrim_hasher::equal): Likewise.
191 (next_discriminator_for_locus): Work on line directly.
192 (same_line_p): Pass in expanded locus1 as well.
193 (assign_discriminators): Avoid redundant location expansions.
195 2018-06-05 Richard Biener <rguenther@suse.de>
197 PR tree-optimization/86046
198 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
199 if required after clearing TREE_ADDRESSABLE.
201 2018-06-05 Richard Biener <rguenther@suse.de>
203 PR tree-optimization/86047
204 * tree-ssa-loop.c (for_each_index): Glob handling of all
205 decls and constants and really handle all of them.
207 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
210 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
211 qualifier_void_pointer and qualifier_const_void_pointer.
212 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
213 (arm_init_builtins): Handle the above.
214 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
215 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
216 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
219 2018-06-05 Martin Liska <mliska@suse.cz>
221 * auto-profile.c (read_autofdo_file): Do not use
222 gcov_ctr_summary struct.
223 (afdo_callsite_hot_enough_for_early_inline): Likewise.
224 * coverage.c (struct counts_entry): Likewise.
225 (read_counts_file): Read just single summary entry.
226 (get_coverage_counts): Use gcov_summary struct.
227 * coverage.h (get_coverage_counts): Likewise.
228 * gcov-dump.c (dump_working_sets): Likewise.
229 (tag_summary): Dump just single summary.
230 * gcov-io.c (gcov_write_summary): Write just histogram
232 (gcov_read_summary): Read just single summary.
233 (compute_working_sets): Use gcov_summary struct.
234 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
235 of GCOV_COUNTERS_SUMMABLE.
236 (GCOV_COUNTERS_SUMMABLE): Remove.
237 (GCOV_FIRST_VALUE_COUNTER): Replace with
238 GCOV_COUNTER_V_INTERVAL.
239 (struct gcov_ctr_summary): Remove.
240 (struct gcov_summary): Directly use fields of former
242 (compute_working_sets): Use gcov_summary struct.
243 * gcov.c (read_count_file): Do not use ctrs fields.
244 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
246 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
248 * profile.c: Likewise.
249 * profile.h: Likewise.
251 2018-06-05 Martin Liska <mliska@suse.cz>
253 PR gcov-profile/84846
254 * gcov.c (output_lines): Print working directory only
255 in intermediate format.
257 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
259 * config/s390/s390-builtin-types.def: Add void function type.
260 * config/s390/s390-builtins.def: Use the function type for the
263 2018-06-04 Jim Wilson <jimw@sifive.com>
265 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
267 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
268 and maybe_eh_return. Change regno to unsigned int. Use new args to
269 handle EH_RETURN_DATA_REGNO registers properly.
270 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
271 (riscv_expand_epilogue): Update comment. Change argument name and
272 type. Update code to use new name and type. Pass new args to
273 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
275 * config/riscv/riscv.md (NORMAL_RETURN): New.
276 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
277 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
278 (eh_return): Call gen_eh_return_internal and emit barrier.
279 (eh_return_internal): Call riscv_expand_epilogue.
281 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
283 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
284 bit_insertion field and declare can_be_merged_into method.
285 (merged_store_group::can_be_merged_into): New method.
286 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
287 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
288 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
290 2018-06-04 Richard Biener <rguenther@suse.de>
292 PR tree-optimization/85955
293 * builtins.c (fold_builtin_sincos): Convert pointers to
294 destination to appropriate type before dereferencing.
296 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
298 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
300 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
302 * expr.c (expand_expr_real_1): Force the operand into memory if
303 its TYPE_MODE is BLKmode and if there is no integer mode for
304 the number of bits being extracted.
306 2018-06-04 Jakub Jelinek <jakub@redhat.com>
310 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
311 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
313 2018-06-04 Richard Biener <rguenther@suse.de>
315 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
316 (cleanup_tree_cfg_noloop): ... single caller. Do
317 start_recording_case_labels later.
319 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
321 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
322 to _IMMINTRIN_H_INCLUDED.
323 * config/i386/pconfigintrin.h: Ditto.
324 * config/i386/waitpkgintrin.h: Ditto.
325 * config/i386/immintrin.h: Add includes for sgxintrin.h,
326 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
327 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
328 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
329 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
330 waitpkgintrin.h and cldemoteintrin.h.
332 2018-06-04 Richard Biener <rguenther@suse.de>
334 PR tree-optimization/86038
335 * tracer.c (find_best_successor): Check probability for
336 being initialized, bail out if not.
338 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
341 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
342 of bits to ignore when comparing architectures.
344 2018-06-04 Jakub Jelinek <jakub@redhat.com>
346 PR tree-optimization/69615
347 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
348 maximum or minimum of the type, try to merge it also as if
349 range1 is + [-, x - 1] or + [x + 1, -].
352 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
354 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
356 PR tree-optimization/86034
357 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
358 the unsigned bitfield type in a bit insertion sequence if it does not
359 have a larger precision than the bitfield size.
360 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
362 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
364 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
366 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
368 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
369 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
370 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
371 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
373 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
375 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
376 Disable -fdelete-null-pointer-checks for ELF toolchain.
378 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
379 Kito Cheng <kito.cheng@gmail.com>
381 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
382 (nds32le-*-*, nds32be-*-*): Integrate checking process.
383 (nds32*-*-*): Add glibc and uclibc conditions.
384 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
385 (TARGET_EXCEPT_UNWIND_INFO): Define.
386 * config/nds32/elf.h: New file.
387 * config/nds32/linux.h: New file.
388 * config/nds32/nds32-elf.opt: New file.
389 * config/nds32/nds32-linux.opt: New file.
390 * config/nds32/nds32-fp-as-gp.c
391 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
392 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
394 (nds32_asm_file_end): Ditto.
395 (nds32_print_operand): Ditto.
396 (nds32_insert_attributes): Ditto.
397 (nds32_init_libfuncs): New function.
398 (TARGET_HAVE_TLS): Define.
399 (TARGET_INIT_LIBFUNCS): Define.
400 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
402 (TARGET_ELF): Apply different mcmodel setting.
403 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
404 been migrated into elf.h and linux.h files.
405 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
406 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
407 (mcmodel): The content has been migrated into nds32-elf.opt and
408 nds32-linux.opt files.
409 * config/nds32/t-elf: New file.
410 * config/nds32/t-linux: New file.
412 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
413 Shiva Chen <shiva0217@gmail.com>
415 * config/nds32/constants.md (unspec_volatile_element): Add
416 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
417 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
419 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
420 (make_pass_nds32_fp_as_gp): Declare.
421 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
423 (nds32_asm_function_end_prologue): Remove unused asm output.
424 (nds32_asm_function_begin_epilogue): Remove unused asm output.
425 (nds32_asm_file_start): Output necessary fp_as_gp information.
426 (nds32_option_override): Adjust register usage.
427 (nds32_expand_prologue): Consider fp_as_gp situation.
428 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
429 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
432 (simple_return): Ditto.
433 (omit_fp_begin): Output special directive for fp_as_gp.
434 (omit_fp_end): Output special directive for fp_as_gp.
435 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
436 mforbid-fp-as-gp): New options.
438 2018-06-01 Mark Wielaard <mark@klomp.org>
440 * dwarf2out.c (dwarf2out_finish): Remove generation of
443 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
445 * gimple-ssa-store-merging.c: Include gimple-fold.h.
446 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
447 (struct merged_store_group): Add bit_insertion field.
448 (dump_char_array): Use standard hexadecimal format.
449 (merged_store_group::merged_store_group): Set bit_insertion to false.
450 (merged_store_group::apply_stores): Use optimal buffer size. Deal
451 with BIT_INSERT_EXPR stores. Move up code updating the mask and
452 also print the mask in the dump file.
453 (pass_store_merging::gate): Minor tweak.
454 (imm_store_chain_info::coalesce_immediate): Fix wrong association
455 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
456 stores with INTEGER_CST stores.
457 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
458 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
459 and use it throughout. Generate bit insertion sequences if need be.
460 (pass_store_merging::process_store): Remove redundant condition.
461 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
463 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
465 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
466 the 128-bit floating point types. Fix function comment.
468 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
470 * config/aarch64/aarch64-simd.md
471 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
473 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
476 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
478 PR tree-optimization/85989
479 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
481 (backprop::intersect_uses): Check it when deciding whether this
482 is a backedge reference.
483 (backprop::process_block): Add each phi to m_visited_phis
484 after visiting it, then clear it at the end.
486 2018-06-01 Richard Biener <rguenther@suse.de>
488 * tree-vectorizer.h (vect_dr_stmt): New function.
489 (vect_get_load_cost): Adjust.
490 (vect_get_store_cost): Likewise.
491 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
492 Use vect_dr_stmt instead of DR_SMTT.
493 (vect_record_base_alignments): Likewise.
494 (vect_calculate_target_alignment): Likewise.
495 (vect_compute_data_ref_alignment): Likewise and make static.
496 (vect_update_misalignment_for_peel): Likewise.
497 (vect_verify_datarefs_alignment): Likewise.
498 (vector_alignment_reachable_p): Likewise.
499 (vect_get_data_access_cost): Likewise. Pass down
500 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
501 (vect_get_peeling_costs_all_drs): Likewise.
502 (vect_peeling_hash_get_lowest_cost): Likewise.
503 (vect_enhance_data_refs_alignment): Likewise.
504 (vect_find_same_alignment_drs): Likewise.
505 (vect_analyze_data_refs_alignment): Likewise.
506 (vect_analyze_group_access_1): Likewise.
507 (vect_analyze_group_access): Likewise.
508 (vect_analyze_data_ref_access): Likewise.
509 (vect_analyze_data_ref_accesses): Likewise.
510 (vect_vfa_segment_size): Likewise.
511 (vect_small_gap_p): Likewise.
512 (vectorizable_with_step_bound_p): Likewise.
513 (vect_prune_runtime_alias_test_list): Likewise.
514 (vect_analyze_data_refs): Likewise.
515 (vect_supportable_dr_alignment): Likewise.
516 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
517 (vect_gen_prolog_loop_niters): Likewise.
518 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
519 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
521 (vect_recog_mask_conversion_pattern): Likewise.
522 (vect_try_gather_scatter_pattern): Likewise.
523 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
524 to vect_get_store_cost.
525 (vect_get_store_cost): Get stmt_info instead of DR.
526 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
527 (vect_get_load_cost): Get stmt_info instead of DR.
529 2018-06-01 Richard Biener <rguenther@suse.de>
532 * gimple-fold.c (var_decl_component_p): Also allow offsetted
533 vars wrapped in MEM_REFs.
535 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
537 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
538 Fix subreg tests so that we only return a choice between
539 GENERAL_REGS and FP_REGS if the original classes included both.
541 2018-06-01 Richard Biener <rguenther@suse.de>
544 * tree-ssa-structalias.c (get_function_part_constraint):
545 Handle NULL fi->decl.
546 (find_func_aliases_for_call): Properly handle indirect
548 (find_func_clobbers): Likewise.
549 (ipa_pta_execute): Likewise.
550 (create_variable_info_for): For functions that are ifunc_resolver
551 resolve to a varinfo that contains the result of the resolver call.
552 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
555 2018-05-31 Michael Collison <michael.collison@arm.com>
557 * config/aarch64/aarch64.md:
558 (*fix_to_zero_extenddfdi2): New pattern.
559 * gcc.target/aarch64/fix_extend1.c: New testcase.
561 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
565 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
566 and BUILT_IN_STRNCMP_EQ.
567 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
569 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
570 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
571 (gimple_fold_builtin): Likewise.
572 * tree-ssa-strlen.c (compute_string_length): New function.
573 (determine_min_obsize): New function.
574 (handle_builtin_string_cmp): New function to handle calls to
575 string compare functions.
576 (strlen_optimize_stmt): Add handling to builtin string compare
578 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
579 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
580 * tree.c (build_common_builtin_nodes): Add new defines of
581 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
583 2018-05-31 Jakub Jelinek <jakub@redhat.com>
586 * bb-reorder.c (pass_partition_blocks::gate): Return false for
587 functions with naked attribute.
589 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
591 * config/i386/sse.md (avx_vec_concat<mode>):
592 Substitute concat_tg_mode mode attribute with xtg_mode.
593 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
594 (concat_tg_mode): Remove mode attribute.
596 2018-05-31 Martin Sebor <msebor@redhat.com>
599 * calls.c (alloc_max_size): Correct a logic error/typo.
600 Treat excessive arguments as infinite. Warn for invalid arguments.
601 * doc/invoke.texi (-Walloc-size-larger-than): Update.
603 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
606 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
607 and movx for Haswell.
609 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
610 Cesar Philippidis <cesar@codesourcery.com>
613 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
614 when emitting error on private/firstprivate reductions.
615 * omp-low.c (lower_omp_target): Avoid reference-type processing
616 on pointers for firstprivate clause.
618 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
620 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
623 * config/aarch64/aarch64-simd.md
624 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
625 (aarch64_ld1_x3_<mode>): Likewise
626 (aarch64_st1x2<VALLDIF:mode>): Likewise
627 (aarch64_st1_x2_<mode>): Likewise
628 (aarch64_st1x3<VALLDIF:mode>): Likewise
629 (aarch64_st1_x3_<mode>): Likewise
630 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
631 (vld1_s8_x3): Likewise.
632 (vld1_u16_x3): Likewise.
633 (vld1_s16_x3): Likewise.
634 (vld1_u32_x3): Likewise.
635 (vld1_s32_x3): Likewise.
636 (vld1_u64_x3): Likewise.
637 (vld1_s64_x3): Likewise.
638 (vld1_f16_x3): Likewise.
639 (vld1_f32_x3): Likewise.
640 (vld1_f64_x3): Likewise.
641 (vld1_p8_x3): Likewise.
642 (vld1_p16_x3): Likewise.
643 (vld1_p64_x3): Likewise.
644 (vld1q_u8_x3): Likewise.
645 (vld1q_s8_x3): Likewise.
646 (vld1q_u16_x3): Likewise.
647 (vld1q_s16_x3): Likewise.
648 (vld1q_u32_x3): Likewise.
649 (vld1q_s32_x3): Likewise.
650 (vld1q_u64_x3): Likewise.
651 (vld1q_s64_x3): Likewise.
652 (vld1q_f16_x3): Likewise.
653 (vld1q_f32_x3): Likewise.
654 (vld1q_f64_x3): Likewise.
655 (vld1q_p8_x3): Likewise.
656 (vld1q_p16_x3): Likewise.
657 (vld1q_p64_x3): Likewise.
658 (vst1_s64_x2): Likewise.
659 (vst1_u64_x2): Likewise.
660 (vst1_f64_x2): Likewise.
661 (vst1_s8_x2): Likewise.
662 (vst1_p8_x2): Likewise.
663 (vst1_s16_x2): Likewise.
664 (vst1_p16_x2): Likewise.
665 (vst1_s32_x2): Likewise.
666 (vst1_u8_x2): Likewise.
667 (vst1_u16_x2): Likewise.
668 (vst1_u32_x2): Likewise.
669 (vst1_f16_x2): Likewise.
670 (vst1_f32_x2): Likewise.
671 (vst1_p64_x2): Likewise.
672 (vst1q_s8_x2): Likewise.
673 (vst1q_p8_x2): Likewise.
674 (vst1q_s16_x2): Likewise.
675 (vst1q_p16_x2): Likewise.
676 (vst1q_s32_x2): Likewise.
677 (vst1q_s64_x2): Likewise.
678 (vst1q_u8_x2): Likewise.
679 (vst1q_u16_x2): Likewise.
680 (vst1q_u32_x2): Likewise.
681 (vst1q_u64_x2): Likewise.
682 (vst1q_f16_x2): Likewise.
683 (vst1q_f32_x2): Likewise.
684 (vst1q_f64_x2): Likewise.
685 (vst1q_p64_x2): Likewise.
686 (vst1_s64_x3): Likewise.
687 (vst1_u64_x3): Likewise.
688 (vst1_f64_x3): Likewise.
689 (vst1_s8_x3): Likewise.
690 (vst1_p8_x3): Likewise.
691 (vst1_s16_x3): Likewise.
692 (vst1_p16_x3): Likewise.
693 (vst1_s32_x3): Likewise.
694 (vst1_u8_x3): Likewise.
695 (vst1_u16_x3): Likewise.
696 (vst1_u32_x3): Likewise.
697 (vst1_f16_x3): Likewise.
698 (vst1_f32_x3): Likewise.
699 (vst1_p64_x3): Likewise.
700 (vst1q_s8_x3): Likewise.
701 (vst1q_p8_x3): Likewise.
702 (vst1q_s16_x3): Likewise.
703 (vst1q_p16_x3): Likewise.
704 (vst1q_s32_x3): Likewise.
705 (vst1q_s64_x3): Likewise.
706 (vst1q_u8_x3): Likewise.
707 (vst1q_u16_x3): Likewise.
708 (vst1q_u32_x3): Likewise.
709 (vst1q_u64_x3): Likewise.
710 (vst1q_f16_x3): Likewise.
711 (vst1q_f32_x3): Likewise.
712 (vst1q_f64_x3): Likewise.
713 (vst1q_p64_x3): Likewise.
715 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
717 * config/msp430/msp430.c (msp430_output_labelref): Prepend
718 user_label_prefix to name.
720 * tree-core.h: Update comment about the format of NAME string
721 passed to handler in attribute_spec.
723 * config/msp430/msp430.md: Remove erroneous subreg expression from
724 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
725 zero_extend{q,h}isi2.
727 2018-05-30 Borislav Petkov <bp@suse.de>
729 * doc/extend.texi: Document some architecture specific
730 constraints and sort entries.
732 2018-05-30 Martin Sebor <msebor@redhat.com>
735 * builtins.c (expand_builtin_stpcpy_1): New function.
736 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
737 only if the former succeeds.
739 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
741 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
744 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
746 * doc/invoke.texi (-flinker-output): Document
748 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
750 * passes.c (ipa_write_summaries): Only modify statements if body
752 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
753 incrementally linking.
754 (ipa_passes): Likewise.
755 * lto-cgraph.c (lto_output_node): When incrementally linking do not
756 pass down resolution info.
757 * common.opt (flag_incremental_link): Update info.
758 * gcc.c (plugin specs): Turn flinker-output=* to
759 -plugin-opt=-linker-output-known
760 * toplev.c (compile_file): Also cut compilation when doing incremental
762 * flag-types. (enum lto_partition_model): Add
763 LTO_LINKER_OUTPUT_NOLTOREL.
764 (invoke.texi): Add -flinker-output docs.
765 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
766 link same way as WPA; do not stream in dead initializers.
768 * dwarf2out.c (dwarf2out_die_ref_for_decl,
769 darf2out_register_external_decl): Support incremental link.
771 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
773 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
775 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
777 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
778 it down to simple_object_copy_lto_debug_sections.
779 (run_gcc): Determine incremental LTO link time and configure
780 lto1 into non-wpa mode, disable renaming of debug sections.
782 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
784 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
785 descriptions of various incorrectly documented functions.
787 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
790 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
791 address check not strict.
793 2018-05-30 Richard Biener <rguenther@suse.de>
795 PR tree-optimization/85964
796 * tracer.c (better_p): Drop initialized count check, we only
797 call the function with initialized counts now.
798 (find_best_successor): Do find a best edge if one
799 has uninitialized count.
800 (find_best_predecessor): Likewise. Do BB frequency check only
801 if count is initialized.
803 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
805 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
806 (aarch64_ldrstr_offset_compare): New.
807 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
808 load/store orderings.
809 (aarch64_gen_adjusted_ldpstp): Likewise.
811 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
813 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
814 Check for subset of GENERAL_REGS and FP_REGS.
815 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
818 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
820 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
821 and wi::to_poly_offset. Add the current offset and then check
822 whether the sum fits, rather than using an unchecked addition of
823 a checked term. Check for a shwi rather than a uhwi.
824 * expr.c (get_bit_range): Use tree_to_poly_uint64.
825 (store_constructor): Use poly_int_tree_p.
826 (expand_expr_real_1): Likewise.
827 * function.c (assign_temp): Likewise.
828 * fold-const.c (const_binop): Use poly_int_tree_p and
830 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
832 * ipa-icf-gimple.c (func_checker::compare_operand): Use
833 to_poly_offset for MEM offsets.
834 * ipa-icf.c (sem_variable::equals): Likewise.
835 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
836 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
837 wi::to_poly_offset for BIT_FIELD_REF offsets.
838 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
840 * var-tracking.c (emit_note_insn_var_location): Use
843 2018-05-29 Jim Wilson <jimw@sifive.com>
845 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
847 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
850 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
851 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
853 (sse4_1_round<mode>2): Use nonimmediate_operand
854 for operand 1 predicate.
856 2018-05-29 Martin Sebor <msebor@redhat.com>
857 Richard Biener <rguenther@suse.de>
860 * calls.c (get_size_range): Call determine_value_range instead
862 * tree-vrp.h (determine_value_range): Declared new function.
863 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
865 2018-05-29 Richard Biener <rguenther@suse.de>
867 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
868 sure to use non-pattern stmts for get_earlier_stmt arguments.
869 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
870 called on pattern stmts.
871 (get_later_stmt): Likewise.
873 2018-05-29 Martin Liska <mliska@suse.cz>
875 PR gcov-profile/85759
876 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
879 2018-05-29 Jakub Jelinek <jakub@redhat.com>
881 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
883 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
887 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
888 VEC_PACK_FLOAT_EXPR): New tree codes.
889 * tree-pretty-print.c (op_code_prio): Handle
890 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
891 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
892 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
893 * tree-inline.c (estimate_operator_cost): Likewise.
894 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
895 * fold-const.c (const_binop): Likewise.
896 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
897 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
898 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
899 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
900 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
901 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
902 * expr.c (expand_expr_real_2): Likewise.
903 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
904 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
905 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
907 * optabs.c (expand_widen_pattern_expr): For
908 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
909 sign from result type rather than operand's type.
910 (expand_binop_directly): For vec_packu_float_optab and
911 vec_packs_float_optab allow result type to be different from operand's
913 * optabs-tree.c (optab_for_tree_code): Handle
914 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
915 VEC_PACK_FLOAT_EXPR. Formatting fixes.
916 * tree-vect-generic.c (expand_vector_operations_1): Handle
917 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
919 * tree-vect-stmts.c (supportable_widening_operation): Handle
921 (supportable_narrowing_operation): Handle FLOAT_EXPR.
922 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
923 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
924 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
925 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
927 (vec_pack<floatprefix>_float_<mode>): New expander.
928 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
930 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
931 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
932 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
933 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
934 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
936 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
937 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
938 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
939 VEC_PACK_FLOAT_EXPR): Document.
941 2018-05-29 Richard Biener <rguenther@suse.de>
943 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
945 (stmt_vec_info_vec): Make pointer.
946 (init_stmt_vec_info_vec): Remove.
947 (free_stmt_vec_info_vec): Likewise.
948 (set_stmt_vec_info_vec): New function.
949 (free_stmt_vec_infos): Likewise.
950 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
951 (set_vinfo_for_stmt): Likewise.
952 (get_earlier_stmt): Likewise.
953 (get_later_stmt): Likewise.
954 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
955 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
956 (vec_info::~vec_info): Free stmt_vec_infos.
957 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
958 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
959 (pass_slp_vectorize::execute): Likewise.
960 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
961 (free_stmt_vec_info_vec): Likewise.
962 (set_stmt_vec_info_vec): New function.
963 (free_stmt_vec_infos): Likewise.
964 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
965 the global stmt_vec_info_vec.
966 * tree-parloops.c (gather_scalar_reductions): Use
967 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
970 2018-05-29 Richard Biener <rguenther@suse.de>
972 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
974 2018-05-29 Martin Liska <mliska@suse.cz>
975 David Malcolm <dmalcolm@redhat.com>
977 * vec.c (test_reverse): New.
978 (vec_c_tests): Add new test.
979 * vec.h (vl_ptr>::reverse): New function.
981 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
983 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
985 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
988 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
990 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
992 2018-05-28 Richard Biener <rguenther@suse.de>
994 PR tree-optimization/85933
995 * tree-vect-data-refs.c (vect_record_base_alignments): Only
996 look at stmts marked as vectorizable.
998 2018-05-28 Richard Biener <rguenther@suse.de>
1000 PR tree-optimization/85934
1001 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1002 vector boolean check before scalar optimization.
1004 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1006 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1009 2018-05-28 Mark Wielaard <mark@klomp.org>
1011 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1012 if it is an expression containing a minus sign.
1014 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1016 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1018 2018-05-27 Paul Koning <ni1d@arrl.net>
1020 * config/pdp11/pdp11.md (truncsihi2): Remove.
1022 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1023 Chung-Ju Wu <jasonwucj@gmail.com>
1025 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1027 (unaligned_store_dw): Ditto.
1028 * config/nds32/nds32-memory-manipulation.c
1029 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1030 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1031 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1032 (emit_setmem_word_loop): Rename to ...
1033 (emit_setmem_doubleword_loop): ... this.
1034 (nds32_gen_dup_4_byte_to_word_value): New function.
1035 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1036 (nds32_expand_setmem_loop): Refine implementation.
1037 (nds32_expand_setmem_loop_v3m): Ditto.
1038 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1041 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1043 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1045 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1047 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1048 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1049 machine->attr_no_prologue_p.
1050 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1051 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1052 (nds32_expand_epilogue): Consider attr_naked_p.
1053 (nds32_expand_epilogue_v3pop): Likewise.
1054 (nds32_can_use_return_insn): Likewise.
1055 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1056 attr_no_prologue_p fields.
1057 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1059 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1062 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1064 * config/i386/sse.md
1065 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1067 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1069 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1071 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1073 (*<floatsuffix>floatv2div2sf2): Rename to ...
1074 (*float<floatunssuffix>v2div2sf2): ... this.
1075 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1076 (float<floatunssuffix>v2div2sf2_mask): ... this.
1077 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1078 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1079 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1081 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1083 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1085 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1087 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1089 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1091 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1092 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1093 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1094 gen_ufix_truncv8dfv8si2.
1095 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1096 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1097 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1098 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1099 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1100 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1101 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1102 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1104 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1108 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1110 2018-05-25 Jim Wilson <jimw@sifive.com>
1112 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1113 * config/riscv/riscv.c (struct machine_function): Add
1114 interrupt_handler_p and attribute_checked_p fields.
1115 (riscv_attribute_table): Add interrupt.
1116 (riscv_interrupt_type_p): New.
1117 (riscv_save_reg_p): Save extra regs for interrupt handler.
1118 (riscv_use_save_libcall): Return false for interrupt handler.
1119 (riscv_first_stack_step): Add forward declaration.
1120 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1121 for interrupt handler with large frame. Use it for saved reg list.
1122 (riscv_expand_prologue): Move flag_stack_usage_info support to
1123 eliminate duplication.
1124 (riscv_expand_epilogue): Generate mret for interrupt handler.
1125 (riscv_epilogue_uses): New.
1126 (riscv_can_use_return_insn): Return false for interrupt handler.
1127 (riscv_function_ok_for_sibcall): Likewise.
1128 (riscv_set_current_function): Add interrupt handler support.
1129 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1130 * config/riscv/riscv.md (UNSPECV_MRET): New.
1132 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1134 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1136 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1138 PR tree-optimization/85712
1139 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1140 this candidate has already been replaced in-situ by a copy.
1142 2018-05-25 Jason Merrill <jason@redhat.com>
1144 PR c++/80485 - inline function non-zero address.
1145 * symtab.c (nonzero_address): Check DECL_COMDAT.
1147 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1150 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1151 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1152 extension of SImode operation. Use const123_operand predicate.
1153 (*saddsi_1): Remove.
1154 (*saddl_se_1): Ditto.
1156 (*ssubl_se_1): Ditto.
1157 * config/alpha/predicates.md (const123_operand): New predicate.
1158 * config/alpha/constraints.md (P): Use IN_RANGE.
1160 2018-05-25 Richard Biener <rguenther@suse.de>
1162 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1164 (ref_maybe_used_by_stmt_p): Likewise.
1165 (stmt_may_clobber_ref_p): Likewise.
1166 (stmt_may_clobber_ref_p_1): Likewise.
1167 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1169 (ref_maybe_used_by_stmt_p): Likewise.
1170 (stmt_may_clobber_ref_p): Likewise.
1171 (stmt_may_clobber_ref_p_1): Likewise.
1172 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1173 the alias oracle to disambiguate DRs with stmts DR analysis
1175 (vect_analyze_data_refs): Do not give up on not analyzable
1176 DRs for BB vectorization. Remove code truncating the dataref
1179 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1182 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1183 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1184 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1186 2018-05-25 Richard Biener <rguenther@suse.de>
1188 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1189 function, combining stmt data ref gathering and fatal analysis
1191 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1192 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1193 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1194 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1195 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1197 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1199 PR tree-optimization/85720
1200 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1201 SCC if all partitions are builtins.
1202 (version_loop_by_alias_check): New parameter. Generate cancelable
1203 runtime alias check if all partitions are builtins.
1204 (distribute_loop): Update call to above function.
1206 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1208 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1209 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1210 (parm_default_def_partition_arg): Ditto.
1211 (set_parm_default_def_partition): Ditto.
1212 (get_parm_default_def_partitions): Ditto and make it static.
1213 (get_undefined_value_partitions): Ditto and make it static.
1214 (remove_ssa_form): Refactor call to init_var_map here.
1215 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1216 computation for loop region.
1217 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1218 (register_default_def): Delete.
1219 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1220 (parm_default_def_partition_arg): Ditto.
1221 (set_parm_default_def_partition): Ditto.
1222 (get_parm_default_def_partitions): Ditto and make it static.
1223 (get_undefined_value_partitions): Ditto and make it static.
1224 (coalesce_with_default, coalesce_with_default): Update comment.
1225 (create_coalesce_list_for_region): New func factored out from
1226 create_outofssa_var_map.
1227 (populate_coalesce_list_for_outofssa): New func factored out from
1228 create_outofssa_var_map and coalesce_ssa_name.
1229 (create_outofssa_var_map): Delete.
1230 (coalesce_ssa_name): Refactor to support live range computation.
1231 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1232 (get_parm_default_def_partitions): Delete.
1233 (get_undefined_value_partitions): Ditto.
1234 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1235 computation for loop region.
1236 (new_tree_live_info, loe_visit_block): Ditto.
1237 (live_worklist, set_var_live_on_entry): Ditto.
1238 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1239 * tree-ssa-live.h (struct _var_map): New fields.
1240 (init_var_map): Change decl.
1241 (region_contains_p): New.
1243 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1245 * tree-ssa-live.h (live_merge_and_clear): Delete.
1247 2018-05-25 Richard Biener <rguenther@suse.de>
1250 * tree-dump.c (dequeue_and_dump): Remove access to removed
1251 operand 2 of a SWITCH_EXPR.
1253 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1255 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1256 multiplication and division.
1257 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1258 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1259 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1260 (cond_udiv_optab, cond_umod_optab): New optabs.
1261 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1262 (IFN_COND_RDIV): New internal functions.
1263 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1264 TRUNC_MOD_EXPR and RDIV_EXPR.
1265 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1266 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1268 (SVE_INT_BINARY): Include mult.
1269 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1270 (optab, sve_int_op): Handle mult.
1271 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1273 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1274 for SVE_INT_BINARY_SD.
1276 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1278 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1279 (optab, sve_int_op): Handle div and udiv.
1280 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1281 for SVE_INT_BINARY_SD.
1282 (*<optab><mode>3): New insn for the same.
1284 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1286 * tree-vect-patterns.c: Include predict.h.
1287 (vect_recog_divmod_pattern): Restrict check for division support
1288 to when optimizing for size.
1290 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1292 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1293 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1294 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1295 (gimple_match_op::set_op): Likewise.
1296 (gimple_resimplify4): Declare.
1297 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1298 (expr::gen_transform): Likewise.
1299 (decision_tree::gen): Generate a simplification routine for 4 operands.
1300 * gimple-match-head.c (gimple_simplify): Add an overload for
1301 4 operands. In the top-level function, handle up to 4 call
1302 arguments and call gimple_resimplify4.
1303 (gimple_resimplify4): New function.
1304 (build_call_internal): Pass a fourth operand.
1305 (maybe_push_to_seq): Likewise.
1306 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1307 Fold VEC_COND_EXPRs of an operation and a default value into
1308 an IFN_COND_* function if possible.
1309 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1311 (SVE_COND_FP_BINARY): Include them.
1312 (optab, sve_fp_op): Handle them.
1313 (SVE_INT_BINARY_REV): New code iterator.
1314 (SVE_COND_FP_BINARY_REV): New int iterator.
1315 (commutative): New int attribute.
1316 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1318 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1320 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1321 (*cond_<optab><mode>): New patterns for reversed operands.
1323 2018-05-25 Richard Biener <rguenther@suse.de>
1325 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1326 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1327 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1328 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1329 * tree-vect-data-refs.c (everywhere): Adjust users.
1330 * tree-vect-loop.c (everywhere): Likewise.
1331 * tree-vect-slp.c (everywhere): Likewise.
1332 * tree-vect-stmts.c (everywhere): Likewise.
1333 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1335 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1337 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1339 (gcc_cv_as_section_exclude): ... this.
1340 Try Solaris as #exclude syntax.
1341 * configure: Regenerate.
1342 * config.in: Regenerate.
1343 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1345 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1346 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1348 * varasm.c (default_elf_asm_named_section): Don't check if
1349 HAVE_GAS_SECTION_EXCLUDE is defined.
1351 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1353 * doc/md.texi: Update the documentation of the cond_* optabs
1354 to mention the new final operand. Fix GET_MODE_NUNITS call.
1355 Describe the scalar case too.
1356 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1357 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1359 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1360 (get_conditional_internal_fn): Update comment.
1361 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1362 accumulator value as a final argument to conditional functions.
1363 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1364 a define_expand and add an "else" operand. Assert for now that
1365 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1366 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1367 (*cond_<optab><mode>): New patterns.
1368 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1369 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1370 (UNSPEC_COND_EOR): Delete.
1371 (optab): Remove associated mappings.
1372 (SVE_INT_BINARY): New code iterator.
1373 (sve_int_op): Remove int attribute and add "minus" to the code
1375 (SVE_COND_INT_OP): Delete.
1376 (SVE_COND_FP_OP): Rename to...
1377 (SVE_COND_FP_BINARY): ...this.
1379 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1381 * optabs.c (can_reuse_operands_p): New function.
1382 (maybe_legitimize_operands): Try to reuse the results for
1385 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1387 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1388 Add {q} suffix to insn mnemonic.
1390 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1392 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1393 (msp430_warn_func_return): New.
1395 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1397 * fold-const.c (tree_nonzero_bits): New function.
1398 * fold-const.h (tree_nonzero_bits): Likewise.
1399 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1400 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1402 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1406 * varasm.c (assemble_alias): Check ifunc_resolver only on
1409 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1412 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1413 when memory input operand is handled.
1415 2018-05-24 Luis Machado <luis.machado@linaro.org>
1417 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1419 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1421 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1423 * match.pd: Delay FMA folds until after vectorization.
1425 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1428 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1429 address check not strict.
1431 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1433 * gimple-match.h (gimple_match_op): New class.
1434 (mprts_hook): Replace parameters with a gimple_match_op *.
1435 (maybe_build_generic_op): Likewise.
1436 (gimple_simplified_result_is_gimple_val): Replace parameters with
1437 a const gimple_match_op *.
1438 (gimple_simplify): Replace code_helper * and tree * parameters with
1439 a gimple_match_op * parameter.
1440 (gimple_resimplify1): Replace code_helper *, tree and tree *
1441 parameters with a gimple_match_op * parameter.
1442 (gimple_resimplify2): Likewise.
1443 (gimple_resimplify3): Likewise.
1444 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1445 parameters with a gimple_match_op * parameter.
1446 * gimple-match-head.c (gimple_simplify): Change prototypes of
1447 auto-generated functions to take a gimple_match_op * instead of
1448 separate code_helper * and tree * parameters. Make the same
1449 change in the top-level overload and update calls to the
1450 gimple_resimplify routines. Update calls to the auto-generated
1451 functions and to maybe_push_res_to_seq in the publicly-facing
1452 operation-specific gimple_simplify overloads.
1453 (gimple_match_op::MAX_NUM_OPS): Define.
1454 (gimple_resimplify1): Replace rcode and ops with a single res_op
1455 parameter. Update call to gimple_simplify.
1456 (gimple_resimplify2): Likewise.
1457 (gimple_resimplify3): Likewise.
1458 (mprts_hook): Replace parameters with a gimple_match_op *.
1459 (maybe_build_generic_op): Likewise.
1460 (build_call_internal): Replace type, nargs and ops with
1461 a gimple_match_op *.
1462 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1463 with a single gimple_match_op *. Update calls to mprts_hook,
1464 build_call_internal and gimple_simplified_result_is_gimple_val.
1465 Factor out code that is common to the tree_code and combined_fn cases.
1466 * genmatch.c (expr::gen_transform): Replace tem_code and
1467 tem_ops with a gimple_match_op called tem_op. Update calls
1468 to the gimple_resimplify functions and maybe_push_res_to_seq.
1469 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1470 res_ops. Update call to the gimple_resimplify functions.
1471 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1472 (decision_tree::gen): Make the functions take a gimple_match_op *
1473 called res_op instead of separate res_code and res_ops parameters.
1474 Update call accordingly.
1475 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1476 and ops with a single res_op parameter. Update calls to
1477 maybe_build_generic_op and maybe_push_res_to_seq.
1478 (fold_stmt_1): Update calls to gimple_simplify and
1479 replace_stmt_with_simplification.
1480 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1481 and gimple_simplified_result_is_gimple_val.
1482 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1484 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1485 with a gimple_match_op *.
1486 (vn_nary_build_or_lookup): Likewise. Update call to
1487 vn_nary_build_or_lookup_1.
1488 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1489 gimple_match_op *. Update calls to the gimple_resimplify routines
1490 and to gimple_simplified_result_is_gimple_val.
1491 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1492 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1493 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1494 (visit_nary_op): Likewise.
1495 (visit_reference_op_load): Likewise.
1497 2018-05-23 Luis Machado <luis.machado@linaro.org>
1499 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1500 modifier for printing the step amount.
1502 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1505 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1508 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1510 * doc/sourcebuild.texi (Endianness): New subsubsection.
1512 2018-05-23 Luis Machado <luis.machado@linaro.org>
1514 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1515 <prefetch_dynamic_strides>: New const bool field.
1516 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1517 prefetch_dynamic_strides.
1518 (exynosm1_prefetch_tune): Likewise.
1519 (thunderxt88_prefetch_tune): Likewise.
1520 (thunderx_prefetch_tune): Likewise.
1521 (thunderx2t99_prefetch_tune): Likewise.
1522 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1524 (aarch64_override_options_internal): Update to set
1525 PARAM_PREFETCH_DYNAMIC_STRIDES.
1526 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1527 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1528 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1529 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1530 prefetch-dynamic-strides setting.
1532 2018-05-23 Luis Machado <luis.machado@linaro.org>
1534 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1535 <minimum_stride>: New const int field.
1536 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1537 minimum_stride field defaulting to -1.
1538 (exynosm1_prefetch_tune): Likewise.
1539 (thunderxt88_prefetch_tune): Likewise.
1540 (thunderx_prefetch_tune): Likewise.
1541 (thunderx2t99_prefetch_tune): Likewise.
1542 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1543 <default_opt_level>: Set to 3.
1544 (aarch64_override_options_internal): Update to set
1545 PARAM_PREFETCH_MINIMUM_STRIDE.
1546 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1547 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1548 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1549 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1550 stride is constant and is below the minimum stride threshold.
1552 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1554 * config/arm/arm-cpus.in (mode26): Delete.
1555 (armv4): Delete mode26 reference.
1556 * config/arm/arm.c (arm_configure_build_target): Delete use of
1559 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1561 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1563 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1564 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1566 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1567 (floatunsdidf2): Ditto.
1569 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1571 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1572 (fixuns_trunc<mode>si2_avx512f): Ditto.
1573 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1574 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1575 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1577 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1579 PR rtl-optimization/79985
1580 * df-scan.c (df_insn_refs_collect): Remove special case for
1581 global registers and asm statements.
1583 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1585 * extend.texi (Global Register Variables): Rewrite the bullet list.
1586 Note that the register is available for allocation. Note that access
1587 via inline asm must use constraints. Add note about async-signal
1588 handlers. Remove paragraph about automagic register selection.
1590 2018-05-23 Richard Biener <rguenther@suse.de>
1592 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1593 of fixed offset from memset VN.
1595 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1597 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1599 (alloc_cand_and_find_basis): Initialize first_interp field.
1600 (slsr_process_mul): Modify first_interp field.
1601 (slsr_process_add): Likewise.
1602 (slsr_process_cast): Modify first_interp field for each new
1604 (slsr_process_copy): Likewise.
1605 (dump_candidate): Dump first_interp field.
1606 (replace_mult_candidate): Process all interpretations, not just
1608 (replace_rhs_if_not_dup): Likewise.
1609 (replace_one_candidate): Likewise.
1611 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1613 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1615 (aarch64_needs_frame_chain): New function.
1616 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1618 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1621 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1622 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1623 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1624 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1625 as true for strict-align.
1626 (aarch64_can_inline_p): Perform checks even when callee has no
1627 attributes to check for strict alignment.
1628 * doc/extend.texi (AArch64 Function Attributes): Document
1630 * doc/invoke.texi: (AArch64 Options): Likewise.
1632 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1634 PR tree-optimization/85853
1635 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1636 the handling of the root of the node to...
1637 (vect_slp_analyze_node_operations_1): ...this new function,
1638 and run the whole thing with the child nodes' def types
1639 set according to their SLP node's def type.
1641 2018-05-23 Richard Biener <rguenther@suse.de>
1644 * tree-data-ref.c (create_runtime_alias_checks): Defer
1645 and ignore overflow warnings.
1647 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1649 PR tree-optimization/85822
1650 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1653 2018-05-23 Richard Biener <rguenther@suse.de>
1655 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1656 memset constants via native_interpret_expr.
1658 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1661 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1663 (cgraph_node::create_alias): Likewise.
1664 (cgraph_node::get_availability): Check ifunc_resolver instead
1665 of looking up ifunc attribute.
1666 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1667 * varasm.c (do_assemble_alias): Likewise.
1668 (assemble_alias): Likewise.
1669 (default_binds_local_p_3): Likewise.
1670 * cgraph.h (cgraph_node): Add ifunc_resolver.
1671 (cgraph_node::only_called_directly_or_aliased_p): Return false
1673 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1675 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1676 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1677 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1678 instead of looking up ifunc attribute.
1680 2018-05-22 Luis Machado <luis.machado@linaro.org>
1682 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1684 2018-05-22 Martin Sebor <msebor@redhat.com>
1687 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1688 only when expasion succeeds.
1689 (expand_builtin_strcmp): Same.
1690 (expand_builtin_strncmp): Same.
1692 2018-05-22 Martin Sebor <msebor@redhat.com>
1694 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1696 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1697 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1699 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1700 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1702 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1703 Add check for legitimate_address.
1704 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1705 (aarch64_swap_ldrstr_operands): New.
1706 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1709 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1710 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1712 * config/aarch64/aarch64.md: New patterns to generate stp
1714 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1715 single words and double words.
1716 (load_pair_sw, load_pair_dw): Likewise.
1717 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1719 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1721 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1722 for different mode ldpstp and add peephole for merged zero stores.
1724 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1726 (aarch64_gen_store_pair): Rename calls to match new patterns.
1727 (aarch64_gen_load_pair): Rename calls to match new patterns.
1728 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1729 (load_pair<DREG:mode><DREG2:mode>): ... This.
1730 (store_pair<mode>): Rename to...
1731 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1732 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1734 (V_INT_EQUIV): Handle SImode.
1735 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1738 2018-05-22 Martin Sebor <msebor@redhat.com>
1741 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1742 or ajust the presumed bound on an operation to avoid unnecessary
1745 2018-05-22 Martin Sebor <msebor@redhat.com>
1747 PR tree-optimization/85826
1748 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1749 assuming that a DECL necesarily has a constant size.
1751 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1754 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1756 2018-05-22 Richard Biener <rguenther@suse.de>
1758 PR tree-optimization/85834
1759 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1760 non-constant and non-zero memset arguments.
1762 2018-05-22 Martin Liska <mliska@suse.cz>
1765 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1767 2018-05-22 Richard Biener <rguenther@suse.de>
1769 PR tree-optimization/85863
1770 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1771 comparisons when vectype is specified.
1772 (vectorizable_condition): Do not specify vectype for
1773 vect_is_simple_cond when SLP vectorizing.
1775 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1778 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1779 define __ibm128 as long double.
1780 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1781 as a distinct type when IEEE 128-bit support is enabled.
1782 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1783 128-bit types to use the correct functions.
1784 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1785 convert between 128-bit floating point types that have different
1786 modes but the same representation, instead of using gen_lowpart to
1788 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1790 (IFKF_reg): New attributes to give the register constraints for
1792 (extend<mode>tf2_internal): New insns to mark an explicit
1793 conversion between 128-bit floating point types that have a
1794 different mode but share the same representation.
1796 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1798 PR tree-optimization/85814
1799 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1800 a null return from get_strinfo when unsharing the next
1801 strinfo in the chain.
1803 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1806 * varasm.c (weak_finish): Clean up weak_decls.
1808 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1810 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1811 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1812 UNSPEC_UADALP values.
1813 * config/aarch64/iterators.md (ABAL): New int iterator.
1816 (sur): Add mappings for the above.
1817 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1819 (aarch64_<sur>abal<mode>_4): Likewise.
1820 (aarch64_<sur>adalp<mode>_3): Likewise.
1821 (<sur>sadv16qi): New define_expand.
1823 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1825 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1826 (*movdf_internal): Ditto.
1827 (*rcpsf2_sse): Ditto.
1828 (*rsqrtsf2_sse): Ditto.
1829 (*sqrt<mode>2_sse): Ditto.
1831 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1833 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1835 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1836 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1837 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1838 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1840 * config/aarch64/arm_neon.h: Likewise.
1841 * config/aarch64/iterators.md (VQ_I): New.
1843 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1845 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1847 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1850 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1852 * config/nds32/constraints.md (S): New constraint.
1853 * config/nds32/nds32.md (call_internal): Use constraint S.
1854 (call_value_internal): Likewise.
1855 (sibcall_internal): Likewise.
1856 (sibcall_value_internal): Likewise.
1858 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1859 Chung-Ju Wu <jasonwucj@gmail.com>
1861 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1864 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1865 Chung-Ju Wu <jasonwucj@gmail.com>
1867 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1868 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1869 (nds32_rtx_costs_impl): Simplify.
1870 (nds32_address_cost_impl): Simplify.
1871 (nds32_init_rtx_costs): New function.
1872 (nds32_rtx_costs_speed_prefer): Likewise.
1873 (nds32_rtx_costs_size_prefer): Likewise.
1874 (nds32_address_cost_speed_prefer): Likewise.
1875 (nds32_address_cost_speed_fwprop): Likewise.
1876 (nds32_address_cost_size_prefer): Likewise.
1877 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1878 * config/nds32/nds32.c (nds32_option_override): Use
1879 nds32_init_rtx_costs function.
1881 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1883 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1884 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1885 (TARGET_PIPELINE_N8): Likewise.
1886 (TARGET_PIPELINE_N10): Likewise.
1887 (TARGET_PIPELINE_N13): Likewise.
1888 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1890 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1892 * config/nds32/nds32-fpu.md: Update copyright year.
1894 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1896 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1898 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1900 * config/nds32/nds32.c
1901 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1902 * config/nds32/nds32.opt (minline-asm-r15): New option.
1904 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1906 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1908 * config/nds32/nds32.md (abssi2): New pattern.
1910 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1912 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1913 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1914 Merge insn pattern from sse_cvtsi2ss<round_name> and
1915 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1916 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1917 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1918 using SWI48 mode iterator.
1919 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1920 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1921 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1922 pattern from sse_cvttss2si<round_saeonly_name>
1923 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1924 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1925 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1926 using SWI48 mode iterator.
1927 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1928 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1929 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1930 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1931 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1932 using SWI48 mode iterator.
1933 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1934 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1935 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1936 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1937 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1938 SWI48 mode iterator.
1939 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1940 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1941 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1942 pattern from sse_cvttsd2si<round_saeonly_name>
1943 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1945 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1947 * config/nds32/nds32-md-auxiliary.c
1948 (nds32_valid_smw_lwm_base_p): Refine.
1949 (nds32_output_smw_single_word): Refine.
1950 (nds32_output_smw_double_word): New.
1951 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1953 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1955 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1956 (nds32_output_stack_pop): Refine.
1957 (nds32_expand_unaligned_load): Refine.
1958 (nds32_expand_unaligned_store): Refine.
1960 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1961 Chung-Ju Wu <jasonwucj@gmail.com>
1963 * config/nds32/constants.md: Add TP_REGNUM constant.
1964 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1965 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1967 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1968 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1969 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1970 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1971 and PIC code generation.
1972 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1974 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1976 * config/nds32/nds32.md: Support TLS and PIC.
1977 * config/nds32/nds32.c: Support TLS and PIC.
1978 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1979 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1982 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1984 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1985 mode with E_ prefix.
1987 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1988 Chung-Ju Wu <jasonwucj@gmail.com>
1990 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1991 * config/nds32/nds32-md-auxiliary.c
1992 (symbolic_reference_mentioned_p): New.
1993 (nds32_legitimize_ict_address): New.
1994 (nds32_expand_ict_move): New.
1995 (nds32_indirect_call_referenced_p): New.
1996 (nds32_symbol_binds_local_p): Delete.
1997 (nds32_long_call_p): Modify.
1998 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1999 * config/nds32/nds32-protos.h
2000 (symbolic_reference_mentioned_p): Declare.
2001 (nds32_legitimize_ict_address): Declare.
2002 (nds32_expand_ict_move): Declare.
2003 (nds32_indirect_call_referenced_p): Declare.
2004 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2005 (nds32_relax_group): Use nds32_ict_const_p as condition.
2006 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2007 (nds32_asm_file_start): Output ict_model directive in asm code.
2008 (nds32_legitimate_address_p): Consider indirect call.
2009 (nds32_print_operand): Consider indirect call.
2010 (nds32_print_operand_address): Consider indirect call.
2011 (nds32_insert_attributes): Handle "indirect_call" attribute.
2012 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2013 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2014 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2015 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2016 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2017 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2018 (TARGET_ICT_MODEL_SMALL): Define.
2019 (TARGET_ICT_MODEL_LARGE): Define.
2020 * config/nds32/nds32.md (movsi): Consider ict model.
2021 (call, call_value): Consider ict model.
2022 (sibcall, sibcall_value): Consider ict model.
2023 * config/nds32/nds32.opt (mict-model): New option.
2024 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2027 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2028 Monk Chiang <sh.chiang04@gmail.com>
2029 Jim Wilson <jimw@sifive.com>
2031 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2032 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2033 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2034 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2035 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2036 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2037 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2038 compute save_libcall_adjustment properly.
2039 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2040 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2041 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2042 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2043 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2044 (ABI_SPEC): Handle mabi=ilp32e.
2045 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2046 (RVE): Add RVE mask.
2047 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2048 <-march>: Add rv32e as an example.
2050 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2053 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2054 (intra_create_variable_infos): Handle C++ constructors.
2056 2018-05-18 Martin Liska <mliska@suse.cz>
2058 * passes.def: Remove a redundant pass.
2060 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2063 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2065 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2067 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2069 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2071 (armv2, armv2a, armv3, armv3m): Delete architectures.
2072 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2073 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2074 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2076 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2077 (*mulsidi3adddi): Likewise.
2078 (mulsidi3): Likewise.
2079 (*mulsidi3_nov6): Likewise.
2080 (umulsidi3): Likewise.
2081 (umulsidi3_nov6): Likewise.
2082 (umaddsidi4): Likewise.
2083 (*umulsidi3adddi): Likewise.
2084 (smulsi3_highpart): Likewise.
2085 (*smulsi3_highpart_nov6): Likewise.
2086 (umulsi3_highpart): Likewise.
2087 (*umulsi3_highpart_nov6): Likewise.
2088 * config/arm/arm.h (arm_arch3m): Delete.
2089 * config/arm/arm.c (arm_arch3m): Delete.
2090 (arm_option_override_internal): Update armv3-related comment.
2091 (arm_configure_build_target): Delete use of isa_bit_mode32.
2092 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2093 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2094 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2096 (mulusa3): Likewise.
2097 * config/arm/arm-protos.h (arm_arch3m): Delete.
2098 * config/arm/arm-tables.opt: Regenerate.
2099 * config/arm/arm-tune.md: Likewise.
2100 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2101 deleted architectures.
2103 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2105 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2106 (armv5t, armv5te): New features.
2107 (ARMv5, ARMv5e): Delete fgroups.
2108 (ARMv5t, ARMv5te): Adjust for above changes.
2110 (armv5, armv5e): Delete arches.
2111 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2113 (*call_reg_arm): Likewise.
2114 (*call_value_reg_armv5): Likewise.
2115 (*call_value_reg_arm): Likewise.
2116 (*call_symbol): Likewise.
2117 (*call_value_symbol): Likewise.
2118 (*sibcall_insn): Likewise.
2119 (*sibcall_value_insn): Likewise.
2121 (prefetch): Likewise.
2122 (define_split and define_peephole2 dependent on arm_arch5):
2124 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2126 (TARGET_ARM_QBIT): Likewise.
2127 (TARGET_DSP_MULTIPLY): Likewise.
2128 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2129 (arm_arch5, arm_arch5e): Delete.
2130 (arm_arch5t, arm_arch5te): Declare.
2131 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2132 (arm_arch5t): Declare.
2133 (arm_option_reconfigure_globals): Update for the above.
2134 (arm_options_perform_arch_sanity_checks): Update comment, replace
2135 use of arm_arch5 with arm_arch5t.
2136 (use_return_insn): Likewise.
2137 (arm_emit_call_insn): Likewise.
2138 (output_return_instruction): Likewise.
2139 (arm_final_prescan_insn): Likewise.
2140 (arm_coproc_builtin_available): Likewise.
2141 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2142 arm_arch5e with arm_arch5t and arm_arch5te.
2143 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2144 (arm_arch5t, arm_arch5te): Declare.
2145 * config/arm/arm-tables.opt: Regenerate.
2146 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2147 * config/arm/t-multilib: Likewise.
2148 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2149 instead of arm_arch5.
2150 (*call_reg_thumb1): Likewise.
2151 (*call_value_reg_thumb1_v5): Likewise.
2152 (*call_value_reg_thumb1): Likewise.
2153 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2155 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2157 2018-05-18 Martin Liska <mliska@suse.cz>
2159 PR gcov-profile/84846
2160 * doc/gcov.texi: Document -t option of gcov tool.
2162 2018-05-18 Martin Liska <mliska@suse.cz>
2164 PR gcov-profile/84846
2165 * gcov.c (print_usage): Add new -t option.
2166 (process_args): Handle the option.
2167 (generate_results): Use stdout as output when requested by
2170 2018-05-18 Martin Liska <mliska@suse.cz>
2172 PR gcov-profile/84846
2173 * coverage.c (coverage_init): Write PWD to .gcno file.
2174 * doc/gcov.texi: Document how working directory is printed.
2175 * gcov-dump.c (dump_gcov_file): Print PWD.
2176 * gcov.c (output_intermediate_file): Likewise.
2177 (read_graph_file): Read PWD string.
2178 (output_lines): Print PWD.
2180 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2183 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2184 for retval and return false if all args to phi are zero.
2186 2018-05-18 Richard Biener <rguenther@suse.de>
2188 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2190 (evrp_dom_walker::before_dom_children): Call it.
2192 2018-05-18 Richard Biener <rguenther@suse.de>
2194 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2195 results when processing array refs with variable index.
2197 2018-05-18 Toon Moene <toon@moene.org>
2199 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2200 directly after that of -floop-interchange. Indicate that both
2201 options are enabled by default when specifying -O3.
2203 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2205 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2206 iterator. Delete separate integer-mode vec_set<mode> expander.
2207 (aarch64_simd_vec_setv2di): Delete.
2208 (vec_setv2di): Delete.
2209 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2210 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2211 the "w, r" alternative.
2213 2018-05-18 Martin Liska <mliska@suse.cz>
2215 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2216 * tree-pass.h (make_pass_lower_switch_O0): New function.
2217 * tree-switch-conversion.c (node_has_low_bound): Remove.
2218 (node_has_high_bound): Likewise.
2219 (node_is_bounded): Likewise.
2220 (class pass_lower_switch): Make it a template type and create
2222 (pass_lower_switch::execute): Add template argument.
2223 (make_pass_lower_switch): New function.
2224 (make_pass_lower_switch_O0): New function.
2225 (do_jump_if_equal): Remove.
2226 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2227 all the hard work to tree optimization passes.
2229 2018-05-18 Martin Liska <mliska@suse.cz>
2231 * dbgcnt.c (limit_low): Renamed from limit.
2232 (limit_high): New variable.
2233 (dbg_cnt_is_enabled): Check for upper limit.
2234 (dbg_cnt): Adjust dumping.
2235 (dbg_cnt_set_limit_by_index): Add new argument for high
2237 (dbg_cnt_set_limit_by_name): Likewise.
2238 (dbg_cnt_process_single_pair): Parse new format.
2239 (dbg_cnt_process_opt): Use strtok.
2240 (dbg_cnt_list_all_counters): Remove 'value' and add
2242 * doc/invoke.texi: Document changes.
2244 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2246 * doc/sourcebuild.texi (scalar_all_fma): Document.
2247 * tree.def (FMA_EXPR): Delete.
2248 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2249 * internal-fn.c (ternary_direct): New macro.
2250 (expand_ternary_optab_fn): Likewise.
2251 (direct_ternary_optab_supported_p): Likewise.
2252 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2253 * builtins.c (fold_builtin_fma): Delete.
2254 (fold_builtin_3): Don't call it.
2255 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2256 * expr.c (expand_expr_real_2): Likewise.
2257 * fold-const.c (operand_equal_p): Likewise.
2258 (fold_ternary_loc): Likewise.
2259 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2260 * gimple.c (DEFTREECODE): Likewise.
2261 * gimplify.c (gimplify_expr): Likewise.
2262 * optabs-tree.c (optab_for_tree_code): Likewise.
2263 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2264 * tree-eh.c (operation_could_trap_p): Likewise.
2265 (stmt_could_throw_1_p): Likewise.
2266 * tree-inline.c (estimate_operator_cost): Likewise.
2267 * tree-pretty-print.c (dump_generic_node): Likewise.
2268 (op_code_prio): Likewise.
2269 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2270 * tree-ssa-operands.c (get_expr_operands): Likewise.
2271 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2272 * fold-const-call.h (fold_fma): Delete.
2273 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2274 CFN_FNMA and CFN_FNMS.
2276 * genmatch.c (combined_fn): New enum.
2277 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2278 (commutative_op): New function.
2279 (commutate): Use it. Handle more than 2 operands.
2280 (dt_operand::gen_gimple_expr): Use commutative_op.
2281 (parser::parse_expr): Allow :c to be used with non-binary
2282 operators if the commutative operand is known.
2283 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2284 CFN_FMS, CFN_FNMA and CFN_FNMS.
2285 (backprop::process_assign_use): Remove FMA_EXPR handling.
2286 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2287 (gen_hsa_fma): New function.
2288 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2289 IFN_FNMA and IFN_FNMS.
2290 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2291 * gimple-fold.h (follow_all_ssa_edges): Declare.
2292 * gimple-fold.c (follow_all_ssa_edges): New function.
2293 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2294 gimple_build interface and use follow_all_ssa_edges to fold the result.
2295 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2296 instead of checking for optabs directly.
2297 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2298 rather than FMA_EXPRs.
2299 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2300 call to IFN_FMA instead of an FMA_EXPR.
2302 2018-05-17 Jim Wilson <jimw@sifive.com>
2304 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2305 sign extended promoted subreg, and the range does not have the sign bit
2306 set, then do a sign extend.
2308 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2309 test, check for sign extended subreg and/or constant operands, and
2310 do a sign extend in that case.
2312 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2314 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2315 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2317 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2318 Change logics_shift_reg to logics_shift_imm.
2319 (thunderx2t99_fp_loadpair_basic): Delete.
2320 (thunderx2t99_fp_storepair_basic): Delete.
2321 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2322 (thunderx2t99_asimd_polynomial): Delete.
2323 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2324 and neon_fp_mul_d_scalar_q.
2325 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2326 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2327 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2328 (thunderx2t99_asimd_lut): Add missing tbl types.
2329 (thunderx2t99_asimd_ext): Delete.
2330 (thunderx2t99_asimd_load1_1_mult): Delete.
2331 (thunderx2t99_asimd_load1_2_mult): Delete.
2332 (thunderx2t99_asimd_load1_ldp): New.
2333 (thunderx2t99_asimd_load1): New.
2334 (thunderx2t99_asimd_load2): Add missing *load2* types.
2335 (thunderx2t99_asimd_load3): New.
2336 (thunderx2t99_asimd_load4): New.
2337 (thunderx2t99_asimd_store1_1_mult): Delete.
2338 (thunderx2t99_asimd_store1_2_mult): Delete.
2339 (thunderx2t99_asimd_store2_mult): Delete.
2340 (thunderx2t99_asimd_store2_onelane): Delete.
2341 (thunderx2t99_asimd_store_stp): New.
2342 (thunderx2t99_asimd_store1): New.
2343 (thunderx2t99_asimd_store2): New.
2344 (thunderx2t99_asimd_store3): New.
2345 (thunderx2t99_asimd_store4): New.
2347 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2349 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2350 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2352 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2353 Segher Boessenkool <segher@kernel.crashing.org>
2356 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2359 2018-05-17 Richard Biener <rguenther@suse.de>
2361 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2362 for pruning loop and prune defs feeding only already visited PHIs.
2364 2018-05-17 Richard Biener <rguenther@suse.de>
2366 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2368 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2369 Richard Biener <rguenther@suse.de>
2371 PR tree-optimization/85793
2372 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2373 for VMAT_ELEMENTWISE.
2375 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2377 * internal-fn.h (lookup_internal_fn): Declare
2378 * internal-fn.c (lookup_internal_fn): New function.
2379 * gimple.c (gimple_build_call_from_tree): Handle calls to
2381 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2382 internal function names.
2383 * tree-pretty-print.c (dump_generic_node): Likewise.
2384 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2386 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2388 * gimple-fold.h (gimple_build): Make the function forms take
2389 combined_fn rather than built_in_function.
2390 (gimple_simplify): Likewise.
2391 * gimple-match-head.c (gimple_simplify): Likewise.
2392 * gimple-fold.c (gimple_build): Likewise.
2393 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2394 rather than gimple_build_call_internal.
2395 (get_initial_defs_for_reduction): Likewise.
2396 (vect_create_epilog_for_reduction): Likewise.
2397 (vectorizable_live_operation): Likewise.
2399 2018-05-17 Martin Liska <mliska@suse.cz>
2401 * gimple-ssa-sprintf.c (format_directive): Do not use
2402 space in between 'G_' and '('.
2404 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2407 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2408 even if the mask is not all ones.
2411 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2413 (ix86_gimple_fold_builtin): Likewise.
2416 * config/i386/i386.c: Include tree-vector-builder.h.
2417 (ix86_vector_shift_count): New function.
2418 (ix86_fold_builtin): Fold shift builtins by scalar count.
2419 (ix86_gimple_fold_builtin): Likewise.
2421 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2422 _mm512_setzero): New intrinsics.
2424 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2425 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2427 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2428 code generation for cases where splatting a value is not useful.
2429 * simplify-rtx.c (simplify_ternary_operation): Simplify
2430 vec_merge across a vec_duplicate and a paradoxical subreg forming
2431 a vector mode to a vec_concat.
2433 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2435 * config.gcc: Support "goldmont-plus".
2436 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2438 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2439 PROCESSOR_GOLDMONT_PLUS.
2440 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2441 (processor_target_table): Add "goldmont-plus".
2442 (PTA_GOLDMONT_PLUS): Define.
2443 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2444 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2445 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2446 (fold_builtin_cpu): Add "goldmont-plus".
2447 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2448 (ix86_option_override_internal): Add "goldmont-plus".
2449 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2450 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2451 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2452 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2454 2018-05-17 Richard Biener <rguenther@suse.de>
2456 PR tree-optimization/85757
2457 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2458 remove defs that only feed that PHI from further processing.
2460 2018-05-16 Jim Wilson <jimw@sifive.com>
2462 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2464 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2466 2018-05-16 Mark Wielaard <mark@klomp.org>
2468 * dwarf2out.c (count_index_strings): New function.
2469 (output_indirect_strings): Call count_index_strings and generate
2470 header for dwarf_version >= 5.
2472 2018-05-16 Mark Wielaard <mark@klomp.org>
2474 * dwarf2out.c (dwarf_FORM): New function.
2475 (set_indirect_string): Use dwarf_FORM.
2476 (reset_indirect_string): Likewise.
2477 (size_of_die): Likewise.
2478 (value_format): Likewise.
2479 (output_die): Likewise.
2480 (add_skeleton_AT_string): Likewise.
2481 (output_macinfo_op): Likewise.
2482 (index_string): Likewise.
2483 (output_index_string_offset): Likewise.
2484 (output_index_string): Likewise.
2485 (count_index_strings): Likewise.
2487 2018-05-16 Carl Love <cel@us.ibm.com>
2489 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2490 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2492 2018-05-16 Martin Jambor <mjambor@suse.cz>
2494 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2495 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2497 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2499 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2500 (fnma<mode>4): Likewise.
2501 (fms<mode>4): Likewise.
2502 (fnms<mode>4): Likewise.
2503 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2504 (aarch64_fnma<mode>4): Likewise.
2505 (aarch64_fms<mode>4): Likewise.
2506 (aarch64_fnms<mode>4): Likewise.
2507 (aarch64_fnmadd<mode>4): Likewise.
2509 2018-05-16 Jason Merrill <jason@redhat.com>
2511 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2513 2018-05-16 Richard Biener <rguenther@suse.de>
2515 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2516 (dump_stmt_cost): Declare.
2517 (add_stmt_cost): Dump cost we add.
2518 (add_stmt_costs): New function.
2519 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2521 (vect_analyze_stmt): Adjust prototype.
2522 (vectorizable_condition): Likewise.
2523 (vectorizable_live_operation): Likewise.
2524 (vectorizable_reduction): Likewise.
2525 (vectorizable_induction): Likewise.
2526 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2527 cost vector to pass to vectorizable_ and record afterwards.
2528 (vect_model_reduction_cost): Take cost vector argument and adjust.
2529 (vect_model_induction_cost): Likewise.
2530 (vectorizable_reduction): Likewise.
2531 (vectorizable_induction): Likewise.
2532 (vectorizable_live_operation): Likewise.
2533 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2534 SLP_TREE_NUMBER_OF_VEC_STMTS.
2535 (vect_analyze_slp_cost_1): Remove.
2536 (vect_analyze_slp_cost): Likewise.
2537 (vect_slp_analyze_node_operations): Take visited args and
2538 a target cost vector. Avoid processing already visited stmt sets.
2539 (vect_slp_analyze_operations): Use a local cost vector to gather
2540 costs and register those of non-discarded instances.
2541 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2542 (vect_schedule_slp_instance): Remove copying of
2543 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2545 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2546 adding cost. Record cost entry location.
2547 (vect_prologue_cost_for_slp_op): Function to compute cost of
2548 a constant or invariant generated for SLP vect in the prologue,
2549 split out from vect_analyze_slp_cost_1.
2550 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2551 (vect_model_promotion_demotion_cost): Likewise.
2552 (vect_model_store_cost): Likewise, make static.
2553 (vect_model_load_cost): Likewise.
2554 (vectorizable_bswap): Add cost vector arg and adjust.
2555 (vectorizable_call): Likewise.
2556 (vectorizable_simd_clone_call): Likewise.
2557 (vectorizable_conversion): Likewise.
2558 (vectorizable_assignment): Likewise.
2559 (vectorizable_shift): Likewise.
2560 (vectorizable_operation): Likewise.
2561 (vectorizable_store): Likewise.
2562 (vectorizable_load): Likewise.
2563 (vectorizable_condition): Likewise.
2564 (vectorizable_comparison): Likewise.
2565 (can_vectorize_live_stmts): Likewise.
2566 (vect_analyze_stmt): Likewise.
2567 (vect_transform_stmt): Adjust calls to vectorizable_*.
2568 * tree-vectorizer.c: Include gimple-pretty-print.h.
2569 (dump_stmt_cost): New function.
2571 2018-05-16 Richard Biener <rguenther@suse.de>
2573 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2574 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2575 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2576 (check_name): New callback.
2577 (dse_classify_store): Track cycles via a visited bitmap of PHI
2578 defs and simplify handling of in-loop and across loop dead stores
2579 and properly fail for loop-variant refs. Handle byte-tracking with
2580 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2583 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2585 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2586 (vect_get_mask_type_for_stmt): Likewise.
2587 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2589 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2590 to determine the statement's vector type and the vector type that
2591 should be used for calculating nunits. Deal with cases in which
2592 the type has to be deferred.
2593 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2594 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2595 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2596 (vect_determine_vf_for_stmt): New functions, split out from...
2597 (vect_determine_vectorization_factor): ...here.
2598 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2599 (vect_get_mask_type_for_stmt): New functions, split out from
2600 vect_determine_vectorization_factor.
2602 2018-05-16 Richard Biener <rguenther@suse.de>
2604 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2605 verify the [VEC_]COND_EXPR embedded comparison.
2607 2018-05-15 Martin Sebor <msebor@redhat.com>
2609 PR tree-optimization/85753
2610 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2611 RECORD_TYPE in addition to ARRAY_TYPE.
2613 2018-05-15 Martin Sebor <msebor@redhat.com>
2616 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2618 2018-05-15 Richard Biener <rguenther@suse.de>
2620 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2621 add by_clobber_p one. Change algorithm to collect all defs
2622 representing uses we need to walk and try reducing them to
2623 a single one before failing.
2624 (dse_dom_walker::dse_optimize_stmt): Adjust.
2626 2018-05-13 Mark Wielaard <mark@klomp.org>
2628 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2629 (size_of_loc_descr): Likewise.
2630 (output_loc_operands): Likewise.
2631 (output_loc_operands_raw): Likewise.
2632 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2633 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2634 (hash_loc_operands): Likewise.
2635 (compare_loc_operands): Likewise.
2637 2018-05-14 Mark Wielaard <mark@klomp.org>
2639 * dwarf2out.c (count_index_addrs): New function.
2640 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2642 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2644 PR tree-optimization/83648
2645 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2646 return value as malloc candidate.
2648 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2651 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2652 param as true in call to suggest_attribute.
2654 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2656 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2657 -mreadonly-in-sdata.
2659 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2661 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2663 (aarch64_crypto_aesd_fused): Likewise.
2665 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2667 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2668 (movsi_aarch64): Likewise.
2669 (load_pairsi): Likewise.
2670 (load_pairdi): Likewise.
2671 (store_pairsi): Likewise.
2672 (store_pairdi): Likewise.
2673 (load_pairsf): Likewise.
2674 (load_pairdf): Likewise.
2675 (store_pairsf): Likewise.
2676 (store_pairdf): Likewise.
2677 (zero_extend): Likewise.
2678 (trunc): Swap alternatives.
2679 (fcvt_target): Add '?' to prefer w over r.
2681 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2684 * config/i386/i386.md: Disallow non-commutative arithmetics in
2685 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2686 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2687 in the peephole2 before it.
2689 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2691 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2692 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2693 (ix86_handle_option): Handle -mcldemote.
2694 * config.gcc: New header.
2695 * config/i386/cldemoteintrin.h: New file.
2696 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2697 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2699 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2700 OPTION_MASK_ISA_CLDEMOTE.
2701 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2702 (ix86_valid_target_attribute_inner_p): Ditto.
2703 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2704 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2705 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2706 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2707 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2709 * config/i386/i386.opt: Add -mcldemote.
2710 * config/i386/x86intrin.h: New header.
2711 * doc/invoke.texi: Add -mcldemote.
2713 2018-05-14 Richard Biener <rguenther@suse.de>
2715 * doc/match-and-simplify.texi: Adjust :s documentation.
2717 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2719 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2720 intended memcpy size.
2721 (REORDER_45): Likewise.
2723 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2725 * sort.cc: New file.
2726 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2727 * vec.c (qsort_chk): Use gcc_qsort.
2728 * Makefile.in (OBJS-libcommon): Add sort.o.
2729 (build/sort.o): New target. Use it...
2730 (BUILD_RTL): ... here, and...
2731 (build/gencfn-macros): ... here, and...
2732 (build/genmatch): ... here.
2734 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2735 Chung-Ju Wu <jasonwucj@gmail.com>
2737 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2738 * config/nds32/nds32-graywolf.md: New file.
2739 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2740 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2742 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2743 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2744 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2745 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2746 * config/nds32/pipelines.md: Include n15 settings.
2748 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2749 Chung-Ju Wu <jasonwucj@gmail.com>
2751 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2752 * config/nds32/nds32-n13.md: New file.
2753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2756 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2757 * config/nds32/nds32.md (pipeline_model): Add n13.
2758 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2759 * config/nds32/pipelines.md: Include n13 settings.
2761 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2762 Chung-Ju Wu <jasonwucj@gmail.com>
2764 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2765 * config/nds32/nds32-n10.md: New file.
2766 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2767 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2769 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2770 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2771 * config/nds32/nds32.md (pipeline_model): Add n10.
2772 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2773 * config/nds32/pipelines.md: Include n10 settings.
2775 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2776 Kito Cheng <kito.cheng@gmail.com>
2777 Chung-Ju Wu <jasonwucj@gmail.com>
2779 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2780 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2781 Add enum values for DSP extension instructions.
2782 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2784 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2785 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2787 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2788 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2789 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2790 * config/nds32/nds32-intrinsic.md: Likewise.
2791 * config/nds32/nds32_intrinsic.h: Likewise.
2792 * config/nds32/nds32-md-auxiliary.c: Likewise.
2793 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2794 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2795 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2796 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2797 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2798 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2799 TYPE_DMAC in switch statement.
2800 * config/nds32/nds32.c: New checking and implementation for DSP
2801 extension instructions.
2802 * config/nds32/nds32.h: Likewise.
2803 * config/nds32/nds32.md: Likewise.
2804 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2805 * config/nds32/predicates.md: Implement new predicates for DSP
2808 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2810 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2811 Reformat alternatives and attributes so it is easier to identify
2812 which constraints/attributes go with which instruction.
2813 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2814 (mov<mode>_softfloat32, FMOVE64): Likewise.
2815 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2816 (mov<mode>_softfloat64, FMOVE64): Likewise.
2818 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2820 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2822 (Basic PowerPC Built-in Functions): The new name of the
2823 subsection previously known as "PowerPC Built-in Functions".
2824 (Basic PowerPC Built-in Functions Available on all Configurations):
2826 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2827 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2828 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2829 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2831 2018-05-11 Martin Jambor <mjambor@suse.cz>
2834 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2837 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2840 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2842 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2844 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2845 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2846 (ix86_handle_option): Handle -mwaitpkg.
2847 * config.gcc: New header.
2848 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2849 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2850 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2852 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2853 OPTION_MASK_ISA_WAITPKG.
2854 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2855 (ix86_option_override_internal): Add PTA_WAITPKG.
2856 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2857 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2858 IX86_BUILTIN_TPAUSE.
2859 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2860 __builtin_ia32_umwait and __builtin_ia32_tpause.
2861 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2862 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2863 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2864 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2865 UNSPECV_TPAUSE): New.
2866 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2867 * config/i386/i386.opt: Add -mwaitpkg.
2868 * config/i386/waitpkgintrin.h: New file.
2869 * config/i386/x86intrin.h: New header.
2870 * doc/invoke.texi: Add -mwaitpkg.
2872 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2875 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2877 (cortex-m0): Use armv6s-m isa.
2878 (cortex-m0plus): Likewise.
2879 (cortex-m1): Likewise.
2880 (cortex-m0.small-multiply): Likewise.
2881 (cortex-m0plus.small-multiply): Likewise.
2882 (cortex-m1.small-multiply): Likewise.
2884 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2885 Jakub Jelinek <jakub@redhat.com>
2887 PR tree-optimization/85692
2888 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2889 source permute as well.
2891 2018-05-11 Martin Liska <mliska@suse.cz>
2894 * doc/extend.texi: Document LLVM style format for no_sanitize
2897 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2899 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2900 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2901 (mode_supports_vsx_dform_quad): Likewise.
2902 (mode_supports_vmx_dform): Move these functions to be next to the
2903 other mode_supports functions.
2904 (mode_supports_dq_form): Likewise.
2905 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2906 mode_supports_dq_form.
2907 (reg_offset_addressing_ok_p): Likewise.
2908 (offsettable_ok_by_alignment): Likewise.
2909 (rs6000_legitimate_offset_address_p): Likewise.
2910 (legitimate_lo_sum_address_p): Likewise.
2911 (rs6000_legitimize_address): Likewise.
2912 (rs6000_legitimize_reload_address): Likewise.
2913 (rs6000_secondary_reload_inner): Likewise.
2914 (rs6000_preferred_reload_class): Likewise.
2915 (rs6000_output_move_128bit): Likewise.
2917 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2919 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2920 Generate SImode target register for null target.
2921 <case IX86_BUILTIN_XGETBV>: Ditto.
2922 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2923 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2925 2018-05-10 Carl Love <cel@us.ibm.com>
2927 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2928 dcbtt and dcbtstt if operands[2] is 0.
2930 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2933 * config/i386/sse.md (usadv64qi): New expander.
2935 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2937 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2938 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2939 -maltivec=be support.
2940 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2941 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2942 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2943 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2944 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2945 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2946 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2947 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2948 altivec_vsumsws): Adjust.
2949 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2950 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2951 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2953 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2954 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2955 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2956 (altivec_lve<VI_char>x): Delete expand.
2957 (*altivec_lve<VI_char>x_internal): Rename to...
2958 (altivec_lve<VI_char>x): ... this.
2959 (altivec_lvxl_<mode>): Delete expand.
2960 (*altivec_lvxl_<mode>_internal): Rename to ...
2961 (altivec_lvxl_<mode>): ... this.
2962 (altivec_stvxl_<mode>): Delete expand.
2963 (*altivec_stvxl_<mode>_internal): Rename to ...
2964 (altivec_stvxl_<mode>): ... this.
2965 (altivec_stve<VI_char>x): Delete expand.
2966 (*altivec_stve<VI_char>x_internal): Rename to ...
2967 (altivec_stve<VI_char>x): ... this.
2968 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2969 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2970 reduc_plus_scal_<mode>): Adjust.
2971 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2973 (rs6000_cpu_cpp_builtins): Adjust.
2974 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2975 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2976 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2977 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2978 -maltivec=be support.
2979 (rs6000_split_vec_extract_var): Adjust.
2980 (rs6000_split_v4si_init): Adjust.
2981 (swap_selector_for_mode): Delete.
2982 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2983 altivec_expand_stvex_be): Delete.
2984 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2985 -maltivec=be support.
2986 (rs6000_gimple_fold_builtin): Ditto.
2987 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2989 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2990 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2991 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2992 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2993 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2994 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2995 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2996 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2997 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2998 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2999 anonymous split): Adjust.
3000 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3001 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3003 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3005 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3006 when --with-gxx-include-dir is also specified.
3007 * configure: Regenerate.
3009 2018-05-09 Jim Wilson <jimw@sifive.com>
3012 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3013 * config/riscv/t-withmultilib: New.
3014 * config/riscv/withmultilib.h: New.
3015 * doc/install.texi: Document RISC-V --with-multilib-list support.
3017 2018-05-09 Richard Biener <rguenther@suse.de>
3019 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3021 (vect_bb_vectorization_profitable_p): Adjust. Compute
3022 actual scalar cost using the cost vector and the add_stmt_cost
3025 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3027 PR rtl-optimization/85645
3028 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3029 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3031 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3033 PR rtl-optimization/85645
3034 * shrink-wrap.c (spread_components): Return a boolean saying if
3035 anything was changed.
3036 (try_shrink_wrapping_separate): Iterate spread_components until
3037 nothing changes anymore.
3039 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3041 PR rtl-optimization/85645
3042 * regrename.c (build_def_use): Also kill the chains that include the
3043 destination of a REG_CFA_REGISTER note.
3045 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3047 PR rtl-optimization/85645
3048 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3049 insn that has a REG_CFA_REGISTER note.
3051 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3053 * cfgexpand.c (expand_clobber): New function.
3054 (expand_gimple_stmt_1): Use it.
3055 * tree-vect-stmts.c (vect_clobber_variable): New function,
3057 (vectorizable_simd_clone_call): ...here.
3058 (vectorizable_store): Emit a clobber either side of an
3059 IFN_STORE_LANES sequence.
3060 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3062 2018-05-09 Tom de Vries <tom@codesourcery.com>
3065 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3066 (define_insn "trap_if_false"): Add exit after trap.
3068 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3070 PR rtl-optimization/85638
3071 * bb-reorder.c: Include common/common-target.h.
3072 (create_forwarder_block): New function extracted from...
3073 (fix_up_crossing_landing_pad): ...here. Rename into...
3074 (dw2_fix_up_crossing_landing_pad): ...this.
3075 (sjlj_fix_up_crossing_landing_pad): New function.
3076 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3077 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3078 from both partitions and exit the loop after one iteration.
3080 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3083 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3085 (Basic PowerPC Built-in Functions): The new name of the
3086 subsection previously known as "PowerPC Built-in Functions".
3087 (Basic PowerPC Built-in Functions Available on all Configurations):
3089 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3091 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3093 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3095 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3098 2018-05-08 Jim Wilson <jimw@sifive.com>
3100 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3101 (LD_EMUL_SUFFIX): New.
3102 (LINK_SPEC): Use it.
3104 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3106 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3108 (Basic PowerPC Built-in Functions): The new name of the
3109 subsection previously known as "PowerPC Built-in Functions".
3110 (Basic PowerPC Built-in Functions Available on all Configurations):
3112 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3114 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3116 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3118 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3121 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3124 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3125 after cmpelim optimization.
3127 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3129 * config.gcc: Support "goldmont".
3130 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3131 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3133 * config/i386/i386.c (m_GOLDMONT): Define.
3134 (processor_target_table): Add "goldmont".
3135 (PTA_GOLDMONT): Define.
3136 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3137 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3138 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3139 (fold_builtin_cpu): Add "goldmont".
3140 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3141 (ix86_option_override_internal): Add "goldmont".
3142 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3143 (processor_type): Add PROCESSOR_GOLDMONT.
3144 * config/i386/i386.md: Add CPU "glm".
3145 * config/i386/glm.md: New file.
3146 * config/i386/x86-tune.def: Add m_GOLDMONT.
3147 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3149 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3152 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3154 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3155 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3156 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3159 * config/i386/i386.c (ix86_fold_builtin): Handle
3160 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3163 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3164 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3166 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3169 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3170 (check_arch): Likewise.
3171 (check_fpu): Return the result rather than printing it.
3172 (end arch): Fix operator precedence.
3173 (end cpu): Likewise.
3174 (END): Print the result from check_fpu.
3176 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3177 Alan Hayward <alan.hayward@arm.com>
3178 David Sherwood <david.sherwood@arm.com>
3180 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3181 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3182 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3183 (*fcmuo<mode>_and): New patterns.
3185 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3187 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3188 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3189 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3190 (cmp_op, sve_imm_con): New code attributes.
3191 (SVE_COND_INT_CMP, imm_con): Delete.
3192 (cmp_op): Remove above unspecs from int attribute.
3193 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3195 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3196 comparison-specific unspecs.
3197 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3198 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3199 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3200 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3201 (*vec_fcm<cmp_op><mode>): Rename to...
3202 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3203 (*vec_fcmuo<mode>): Rename to...
3204 (*fcmuo<mode>): ...this and adjust likewise.
3205 (*pred_fcm<cmp_op><mode>): New pattern.
3206 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3207 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3209 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3211 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3212 (aarch64_emit_sve_predicated_cond): New function.
3213 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3214 (aarch64_emit_unspec_cond_or): Replace with...
3215 (aarch64_emit_sve_or_conds): ...this new function. Use
3216 aarch64_emit_sve_ptrue_op for the individual comparisons and
3217 aarch64_emit_binop to OR them together.
3218 (aarch64_emit_inverted_unspec_cond): Replace with...
3219 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3220 aarch64_emit_sve_ptrue_op for the comparison and
3221 aarch64_emit_unop to invert the result.
3222 (aarch64_expand_sve_vec_cmp_float): Update after the above
3223 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3225 2018-05-07 Nathan Sidwell <nathan@acm.org>
3227 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3228 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3229 (Backwards Compatibility): Likewise.
3231 2018-05-07 Luis Machado <luis.machado@linaro.org>
3235 2018-05-07 Luis Machado <luis.machado@linaro.org>
3237 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3238 <prefetch_dynamic_strides>: New const bool field.
3239 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3240 prefetch_dynamic_strides.
3241 (exynosm1_prefetch_tune): Likewise.
3242 (thunderxt88_prefetch_tune): Likewise.
3243 (thunderx_prefetch_tune): Likewise.
3244 (thunderx2t99_prefetch_tune): Likewise.
3245 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3247 (aarch64_override_options_internal): Update to set
3248 PARAM_PREFETCH_DYNAMIC_STRIDES.
3249 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3250 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3251 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3252 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3253 prefetch-dynamic-strides setting.
3255 2018-05-07 Luis Machado <luis.machado@linaro.org>
3257 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3258 <minimum_stride>: New const int field.
3259 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3260 minimum_stride field.
3261 (exynosm1_prefetch_tune): Likewise.
3262 (thunderxt88_prefetch_tune): Likewise.
3263 (thunderx_prefetch_tune): Likewise.
3264 (thunderx2t99_prefetch_tune): Likewise.
3265 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3266 (aarch64_override_options_internal): Update to set
3267 PARAM_PREFETCH_MINIMUM_STRIDE.
3268 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3269 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3270 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3271 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3272 stride is constant and is below the minimum stride threshold.
3274 2018-05-07 Luis Machado <luis.machado@linaro.org>
3276 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3279 2018-05-07 Luis Machado <luis.machado@linaro.org>
3281 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3282 <prefetch_dynamic_strides>: New const bool field.
3283 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3284 prefetch_dynamic_strides.
3285 (exynosm1_prefetch_tune): Likewise.
3286 (thunderxt88_prefetch_tune): Likewise.
3287 (thunderx_prefetch_tune): Likewise.
3288 (thunderx2t99_prefetch_tune): Likewise.
3289 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3291 (aarch64_override_options_internal): Update to set
3292 PARAM_PREFETCH_DYNAMIC_STRIDES.
3293 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3294 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3295 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3296 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3297 prefetch-dynamic-strides setting.
3299 2018-05-07 Luis Machado <luis.machado@linaro.org>
3301 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3302 <minimum_stride>: New const int field.
3303 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3304 minimum_stride field.
3305 (exynosm1_prefetch_tune): Likewise.
3306 (thunderxt88_prefetch_tune): Likewise.
3307 (thunderx_prefetch_tune): Likewise.
3308 (thunderx2t99_prefetch_tune): Likewise.
3309 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3310 (aarch64_override_options_internal): Update to set
3311 PARAM_PREFETCH_MINIMUM_STRIDE.
3312 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3313 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3314 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3315 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3316 stride is constant and is below the minimum stride threshold.
3318 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3321 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3322 the type is addressable. Don't force op into register if it has
3325 2018-05-05 Roland McGrath <mcgrathr@google.com>
3328 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3329 any section for which we don't know a specific type it should have,
3330 regardless of name. Previously this was done only for the exact
3331 names ".init_array", ".fini_array", and ".preinit_array".
3332 (default_elf_asm_named_section): Add comment about
3333 relationship with default_section_type_flags and SECTION_NOTYPE.
3334 (get_section): Don't consider it a type conflict if one side has
3335 SECTION_NOTYPE and the other doesn't, as long as neither has the
3336 SECTION_BSS et al used in the default_section_type_flags logic.
3338 2018-05-05 Tom de Vries <tom@codesourcery.com>
3341 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3342 (workaround_barsyncs): New function.
3343 (nvptx_reorg): Use workaround_barsyncs.
3344 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3345 (define_expand "nvptx_membar_cta"): New define_expand.
3346 (define_insn "*nvptx_membar_cta"): New insn.
3348 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3350 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3351 To improve optimization opportunities.
3352 * builtin-types.def: The new needed builtin types for the above.
3354 2018-05-04 Richard Biener <rguenther@suse.de>
3356 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3357 * gimple-ssa-store-merging.c
3358 (imm_store_chain_info::output_merged_store): Remove redundant create,
3359 release split_store vector contents on failure.
3360 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3361 scalar stmt vector on cache hit.
3363 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3365 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3367 * config.gcc (powerpc-xilinx-eabi*): Remove.
3368 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3370 (fusion_addis_mem_combo_load): Ditto.
3371 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3373 (rs6000_cpu_cpp_builtins): Ditto.
3374 * config/rs6000/rs6000-linux.c
3375 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3376 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3377 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3379 (rs6000_setup_reg_addr_masks): Ditto.
3380 (rs6000_init_hard_regno_mode_ok): Ditto.
3381 (rs6000_option_override_internal): Ditto.
3382 (legitimate_lo_sum_address_p): Ditto.
3383 (rs6000_legitimize_address): Ditto.
3384 (rs6000_legitimize_reload_address): Ditto.
3385 (rs6000_legitimate_address_p): Ditto.
3386 (abi_v4_pass_in_fpr): Ditto.
3387 (setup_incoming_varargs): Ditto.
3388 (rs6000_gimplify_va_arg): Ditto.
3389 (rs6000_split_multireg_move): Ditto.
3390 (rs6000_savres_strategy): Ditto.
3391 (rs6000_emit_prologue_components): Ditto.
3392 (rs6000_emit_epilogue_components): Ditto.
3393 (rs6000_emit_prologue): Ditto.
3394 (rs6000_emit_epilogue): Ditto.
3395 (rs6000_elf_file_end): Ditto.
3396 (rs6000_function_value): Ditto.
3397 (rs6000_libcall_value): Ditto.
3398 * config/rs6000/rs6000.h: Ditto.
3399 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3400 (TARGET_MINMAX): ... this. New.
3401 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3402 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3403 (*movsi_internal1_single): Delete.
3404 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3405 mfpu=, mxilinx-fpu): Delete.
3406 * config/rs6000/singlefp.h: Delete.
3407 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3408 * config/rs6000/t-rs6000: Ditto.
3409 * config/rs6000/t-xilinx: Delete.
3410 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3411 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3412 (VStype_simple): Delete.
3413 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3414 * config/rs6000/xfpu.h: Delete.
3415 * config/rs6000/xfpu.md: Delete.
3416 * config/rs6000/xilinx.h: Delete.
3417 * config/rs6000/xilinx.opt: Delete.
3418 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3419 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3421 2018-05-04 Tom de Vries <tom@codesourcery.com>
3424 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3427 2018-05-04 Richard Biener <rguenther@suse.de>
3430 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3431 (expand_complex_div_wide): Likewise.
3432 (expand_complex_operations_1): Likewise.
3433 (expand_complex_libcall): Preserve EH info of the original stmt.
3434 (tree_lower_complex): Handle removed blocks.
3435 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3436 on complex multiplication and division libcall builtins.
3438 2018-05-04 Richard Biener <rguenther@suse.de>
3441 * fold-const.c (negate_expr_p): Restrict negation of operand
3442 zero of a division to when we know that can happen without
3444 (fold_negate_expr_1): Likewise.
3446 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3449 * real.h (real_nextafter): Declare.
3450 * real.c (real_nextafter): New function.
3451 * fold-const-call.c (fold_const_nextafter): New function.
3452 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3453 CASE_CFN_NEXTTOWARD.
3454 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3455 even when arg1_mode is different from arg0_mode.
3457 2018-05-03 Nathan Sidwell <nathan@acm.org>
3459 * doc/extend.texi (Deprecated Features): Remove
3461 (Backwards Compatibility): Likewise.
3462 * doc/invoke.texi (C++ Language Options): Likewise.
3463 (C++ Dialect Options): Likewise.
3465 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3468 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3469 _mm512_mask_mullox_epi64): New intrinsics.
3471 2018-05-03 Tom de Vries <tom@codesourcery.com>
3474 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3475 dump files): Add offload-tree.
3477 2018-05-03 Richard Biener <rguenther@suse.de>
3479 PR tree-optimization/85615
3480 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3481 to loops not nested in BBs loop father to avoid creating multi-entry
3484 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3486 PR tree-optimization/70291
3487 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3488 arguments. Change return type to tree. Emit libcall as a new
3489 statement rather than replacing existing one when inplace_p is true.
3490 (expand_complex_multiplication_components): New function.
3491 (expand_complex_multiplication): Expand floating-point complex
3492 multiplication using the above.
3493 (expand_complex_division): Rename inner_type parameter to type.
3494 Update expand_complex_libcall call-site.
3495 (expand_complex_operations_1): Update expand_complex_multiplication
3496 and expand_complex_division call-sites.
3498 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3501 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3502 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3503 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3504 the highest significant bit of the shift count mask is clear. In
3505 check whether and[sq]i3 is needed verify that all significant bits
3506 of the shift count other than the highest are set.
3508 2018-05-02 Tom de Vries <tom@codesourcery.com>
3511 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3512 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3513 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3514 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3515 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3516 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3517 __builtin_goacc_parlevel_size.
3519 2018-05-02 Richard Biener <rguenther@suse.de>
3521 PR tree-optimization/85597
3522 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3523 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3526 2018-05-02 Tom de Vries <tom@codesourcery.com>
3529 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3530 dump files): Add ltrans-tree.
3532 2018-05-02 Tom de Vries <tom@codesourcery.com>
3535 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3536 dump files): Add wpa-ipa.
3538 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3540 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3541 powerpc*-*-linux*paired* target.
3542 * config/rs6000/750cl.h: Delete.
3543 * config/rs6000/paired.h: Delete.
3544 * config/rs6000/paired.md: Delete.
3545 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3547 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3548 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3549 comment. Remove paired float support.
3550 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3551 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3553 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3554 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3556 * config/rs6000/rs6000.c: Remove paired float support.
3557 (paired_expand_vector_init, paired_expand_vector_move,
3558 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3559 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3560 paired_expand_builtin, paired_expand_predicate_builtin,
3561 paired_init_builtins): Delete.
3562 * config/rs6000/rs6000.h: Remove paired float support.
3563 * config/rs6000/rs6000.md: Remove paired float support.
3564 (move_from_CR_ov_bit): Delete.
3565 * config/rs6000/rs6000.opt (mpaired): Delete.
3566 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3567 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3569 2018-05-02 Richard Biener <rguenther@suse.de>
3572 * gimplify.c (gimplify_save_expr): When in SSA form allow
3573 SAVE_EXPRs to compute to SSA vars.
3575 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3578 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3579 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3580 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3581 clobber operands[2], instead use a new pseudo. Formatting fixes.
3583 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3585 PR tree-optimization/85586
3586 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3587 exit early for statements in the same group if the accesses are
3590 2018-05-02 Tom de Vries <tom@codesourcery.com>
3593 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3596 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3598 PR tree-optimization/85143
3599 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3601 2018-05-01 Tom de Vries <tom@codesourcery.com>
3604 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3605 not found" error message.
3607 2018-05-01 Tom de Vries <tom@codesourcery.com>
3610 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3611 * vec.c (test_ordered_remove_if): New function.
3612 (vec_c_tests): Call test_ordered_remove_if.
3613 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3614 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3615 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3616 VEC_ORDERED_REMOVE_IF.
3618 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3620 PR tree-optimization/82665
3621 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3622 pointer subtraction where arguments come from a memchr call.
3624 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3626 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3627 --push-state --as-needed and --pop-state instead of --as-needed and
3628 --no-as-needed if ld supports it.
3629 * configure: Regenerated.
3632 * doc/install.texi2html: Replace _002d with - and _002a with * in
3633 generated html files using sed.
3635 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3638 * gcc-rich-location.c (blank_line_before_p): New function.
3639 (use_new_line): New function.
3640 (gcc_rich_location::add_fixit_insert_formatted): New function.
3641 * gcc-rich-location.h
3642 (gcc_rich_location::add_fixit_insert_formatted): New function.
3644 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3646 * selftest.c (assert_streq): Rename "expected" and "actual" to
3647 "val1" and "val2". Extend NULL-handling to cover both inputs
3648 symmetrically, while still requiring both to be non-NULL for a pass.
3649 * selftest.h (assert_streq): Rename "expected" and "actual" to
3651 (ASSERT_EQ): Likewise.
3652 (ASSERT_EQ_AT): Likewise.
3653 (ASSERT_KNOWN_EQ): Likewise.
3654 (ASSERT_KNOWN_EQ_AT): Likewise.
3655 (ASSERT_NE): Likewise.
3656 (ASSERT_MAYBE_NE): Likewise.
3657 (ASSERT_MAYBE_NE_AT): Likewise.
3658 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3659 the assertion to pass.
3660 (ASSERT_STREQ_AT): Likewise.
3662 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3664 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3665 interaction with -pie.
3667 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3669 * selftest.h: Fix alphabetization of per-source-file selftest
3672 2018-04-30 Jason Merrill <jason@redhat.com>
3674 PR c++/61982 - dead stores to destroyed objects.
3675 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3678 2018-04-30 Jason Merrill <jason@redhat.com>
3680 * tree.c (build_clobber): New.
3681 * tree.h: Declare it.
3682 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3684 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3686 * diagnostic-show-locus.c (layout::layout): Update for
3687 location_get_source_line returning a char_span.
3688 (struct char_span): Move to input.h.
3689 (struct correction): Update for fields in char_span becoming
3691 (struct source_line): Update for location_get_source_line
3692 returning a char_span.
3693 (layout::print_line): Likewise.
3694 * edit-context.c (edited_file::print_content): Likewise.
3695 (edited_file::print_diff_hunk): Likewise.
3696 (edited_file::print_run_of_changed_lines): Likewise.
3697 (edited_file::get_num_lines): Likewise.
3698 (edited_line::edited_line): Likewise.
3699 * final.c (asm_show_source): Likewise.
3700 * input.c (location_get_source_line): Convert return type
3701 from const char * to char_span, losing the final "line_len"
3703 (dump_location_info): Update for the above.
3704 (get_substring_ranges_for_loc): Likewise. Use a char_span
3705 when handling the literal within the line.
3706 (test_reading_source_line): Update for location_get_source_line
3707 returning a char_span.
3708 * input.h (class char_span): Move here from
3709 diagnostic-show-locus.c, converting from a struct to a class.
3710 Make data members private.
3711 (char_span::operator bool): New.
3712 (char_span::length): New.
3713 (char_span::get_buffer): New.
3714 (char_span::operator[]): New.
3715 (char_span::subspan): Make const.
3716 (char_span::xstrdup): New.
3717 (location_get_source_line): Convert return type from const char *
3718 to char_span, losing the final "line_size" param.
3720 2018-04-30 Jan Hubicka <jh@suse.cz>
3722 * lto-wrapper.c (ltrans_priorities): New static var.
3723 (cmp_priority): New.
3724 (run_gcc): Read priorities and if doing parallel build order
3725 the Makefile by them.
3727 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3729 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3731 2018-04-30 Richard Biener <rguenther@suse.de>
3733 * tree-cfg.c (verify_address): Remove base argument, add
3734 flag whether to check TREE_ADDRESSABLE and do that.
3735 (verify_expr): Remove.
3736 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3737 (verify_gimple_assign_single): Likewise.
3738 (verify_gimple_switch): Likewise.
3739 (verify_expr_location_1): Dereference tp once. Add (disabled)
3740 piece from verify_expr.
3741 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3743 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3745 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3747 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3749 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3750 (small_data_pattern): Likewise.
3751 (arc_rewrite_small_data): Likewise.
3752 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3753 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3754 (get_symbol_alignment): New function.
3755 (legitimate_small_data_address_p): Likewise.
3756 (legitimate_scaled_address): Update, call
3757 legitimate_small_data_address_p.
3758 (output_sdata): New static variable.
3759 (arc_print_operand): Update how we handle small data operands.
3760 (arc_print_operand_address): Likewise.
3761 (arc_legitimate_address_p): Update, use
3762 legitimate_small_data_address_p.
3763 (arc_rewrite_small_data_p): Remove.
3764 (arc_rewrite_small_data_1): Likewise.
3765 (arc_rewrite_small_data): Likewise.
3766 (small_data_pattern): Likewise.
3767 (compact_sda_memory_operand): Update to use
3768 legitimate_small_data_address_p and get_symbol_alignment.
3769 (prepare_move_operands): Don't rewite sdata pattern.
3770 (prepare_extend_operands): Remove.
3771 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3773 (zero_extendqisi2): Likewise.
3774 (zero_extendhisi2): Likewise.
3775 (extendqihi2): Likewise.
3776 (extendqisi2): Likewise.
3777 (extendhisi2): Likewise.
3781 * config/arc/constraints.md (Usd): Change it to memory constraint.
3783 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3785 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3786 as source of std instructions.
3787 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3788 allow 6-bit constants as source for store instructions.
3789 (movdi_insn): Update instruction pattern to allow 6-bit constants
3790 as source for store instructions.
3792 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3794 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3796 2018-04-30 Nathan Sidwell <nathan@acm.org>
3797 Sandra Loosemore <sandra@codesourcery.com>
3799 * dumpfile.c (dump_open): Allow '-' for stdout.
3800 * doc/invoke.texi (Developer Options): Document dump filename
3801 determination early. Document stdin/stdout selection.
3803 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3805 Microblaze Target: PIC data text relative
3807 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3808 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3810 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3811 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3812 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3813 New addressing mode for data-text relative position indepenedent code.
3814 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3815 'ADDRESS_SYMBOLIC_TXT_REL'.
3816 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3817 (microblaze_legitimate_pic_operand): Exclude function calls from
3818 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3819 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3821 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3822 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3823 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3824 for 'address + offset'.
3825 (microblaze_expand_prologue): Add new function prologue call for
3827 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3828 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3829 table in case of TARGET_PIC_DATA_TEXT_REL.
3830 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3831 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3832 Add new macros 'UNSPEC_TEXT',
3833 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3834 + exclude function calls from 'UNSPEC_PLT' in case of data text
3836 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3837 new target hook for generating address diff vector tables in case of
3839 * doc/tm.texi : Regenerate.
3840 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3841 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3842 of addr diff vector generation.
3843 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3844 target hook definition.
3845 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3846 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3847 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3848 option for data text relative.
3850 2018-04-30 Richard Biener <rguenther@suse.de>
3852 * tree-chrec.h (evolution_function_is_constant_p): Remove
3854 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3856 2018-04-30 Richard Biener <rguenther@suse.de>
3859 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3861 2018-04-30 Richard Biener <rguenther@suse.de>
3863 PR tree-optimization/28364
3864 PR tree-optimization/85275
3865 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3866 copying first exit test.
3868 2018-04-28 Mark Wielaard <mark@klomp.org>
3870 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3872 (dwarf_AT): Handle DW_AT_addr_base.
3873 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3875 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3878 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3879 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3880 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3881 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3883 2018-04-28 Richard Biener <rguenther@suse.de>
3885 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3886 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3887 to reflect use. Only add interesting stmts.
3889 2018-04-27 Martin Jambor <mjambor@suse.cz>
3892 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3893 the jump function allows for passing through aggregate values.
3895 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3897 * input.h (in_system_header_at): Convert from macro to inline
3899 (from_macro_expansion_at): Likewise.
3900 (from_macro_definition_at): Likewise.
3902 2018-04-27 Jeff Law <law@redhat.com>
3904 * config.gcc: Mark tile* targets as deprecated/obsolete.
3906 2018-04-27 Richard Biener <rguenther@suse.de>
3908 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3911 2018-04-27 Richard Biener <rguenther@suse.de>
3913 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3915 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3917 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3918 with Yd constraint. Set "preferred_for_speed" attribute from
3919 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3921 (*movdi_internal): Ditto.
3922 (movti_interunit splitters): Remove
3923 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3924 (movdi_interunit splitters): Ditto.
3925 * config/i386/constraints.md (Ye): Remove.
3926 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3928 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3931 * config/aarch64/constraints.md (Usg): Limit to 31.
3934 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3936 PR tree-optimization/85529
3937 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3938 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3939 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3940 zero extension or masking of the MSB bit.
3941 (optimize_range_tests): Add FIRST_BB argument, pass it through
3942 to optimize_range_tests_var_bound.
3943 (maybe_optimize_range_tests, reassociate_bb): Adjust
3944 optimize_range_tests callers.
3946 2018-04-26 Richard Biener <rguenther@suse.de>
3947 Jakub Jelinek <jakub@redhat.com>
3949 * cgraph.h (symbol_table): Just declare debug method here.
3950 * symtab.c (symbol_table::debug): Define.
3952 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3954 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3956 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3958 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3959 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3960 (*movdi_internal): Substitute Yi and Yj constraint with x
3961 and Ym and Yn constraint with y constraint. Update "isa"
3962 attribute and set "preferred_for_speed" attribute from
3963 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3964 (*movsi_internal): Ditto.
3965 (*movdf_internal): Ditto.
3966 (*movsf_internal): Ditto.
3967 (*zero_extendsidi2): Ditto.
3968 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3969 (sse2_loadld): Ditto.
3970 (*vec_extract<ssevecmodelower>_0): Ditto.
3971 (*vec_extractv4si_0_zext_sse4): Ditto.
3972 (vec_concatv2di): Ditto.
3973 (*vec_dup<mode>): Ditto.
3974 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3975 * config/i386/constraints.md (Yi): Remove.
3980 2018-04-26 Nathan Sidwell <nathan@acm.org>
3982 * dumpfile.c (dump_open): New.
3983 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3984 (dump_finish): Detect stdio/stderr by value not name.
3986 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3988 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3990 2018-04-26 Tom de Vries <tom@codesourcery.com>
3993 * config/nvptx/nvptx.c (verify_neutering_jumps)
3994 (verify_neutering_labels): New function
3995 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3997 2018-04-26 Tom de Vries <tom@codesourcery.com>
4000 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4001 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4004 2018-04-26 Richard Biener <rguenther@suse.de>
4005 Tom de Vries <tom@codesourcery.com>
4008 * lto-streamer-out.c (output_function): Fixup loops if required to match
4009 discovery done in the reader.
4011 2018-04-26 Richard Biener <rguenther@suse.de>
4013 PR tree-optimization/85116
4014 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4015 have a loop exit from the single latch predecessor. Remove
4016 case of header with just condition.
4017 (ch_base::copy_headers): Exclude infinite loops from any
4019 (pass_ch::execute): Record exits.
4021 2018-04-26 Richard Biener <rguenther@suse.de>
4023 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4024 prologue cost vector and pass it to vect_get_load_cost.
4025 (vect_get_peeling_costs_all_drs): Likewise.
4026 (vect_peeling_hash_get_lowest_cost): Likewise.
4027 (vect_enhance_data_refs_alignment): Likewise.
4029 2018-04-26 Richard Biener <rguenther@suse.de>
4032 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4033 checking of integer<->pointer conversions.
4034 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4035 sign-/zero-extending pointer types.
4036 (expand_omp_for_static_chunk): Likewise.
4038 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4039 Jean Lee <xiaoyur347@gmail.com>
4041 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4042 (TARGET_ASAN_SHADOW_OFFSET): Define.
4043 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4044 true for -fsanitize=address.
4046 2018-04-25 Mark Wielaard <mark@klomp.org>
4048 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4051 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4053 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4054 than "alu", remove explicit "memory" and "imm_disp" attributes.
4055 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4058 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4059 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4060 gen_lowpart_no_emit.
4062 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4065 * config/i386/i386.c (ix86_expand_builtin): Change memory
4066 operand to XI, extend p0 to Pmode.
4067 * config/i386/i386.md: Change unspec volatile and operand
4068 1 mode to XI, change operand 0 mode to P.
4070 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4072 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4073 GET_MODE_MASK before any checking.
4074 (nds32_can_use_bset_p): Likewise.
4075 (nds32_can_use_btgl_p): Likewise.
4077 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4079 * config/nds32/nds32-doubleword.md: New define_split pattern for
4080 illegal register number.
4082 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4084 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4086 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4088 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4090 2018-04-25 Richard Biener <rguenther@suse.de>
4092 * lto-streamer.h (LTO_major_version): Bump to 8.
4094 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4096 * BASE-VER: Set to 9.0.0.
4098 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4100 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4101 in __abskf2 and __powikf2.
4103 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4106 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4107 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4108 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4109 Use the above on operand 2. Reindent.
4110 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4112 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4115 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4117 * config/i386/i386.opt (mcet): Removed.
4118 * doc/install.texi: Remove -mcet documentation.
4119 * doc/invoke.texi: Likewise.
4121 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4124 * doc/install.texi: Remove -mcet from bootstrap-cet.
4126 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4129 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4130 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4134 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4135 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4136 containing a CONST_VECTOR.
4138 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4140 * doc/install.texi: Update newlib dependency for nvptx.
4142 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4145 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4146 instead of INTVAL when shifting x left.
4148 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4150 PR tree-optimization/85478
4151 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4152 vect_grouped_store_supported for single element vectors.
4154 2018-04-24 Richard Biener <rguenther@suse.de>
4157 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4158 load cost increase to the case of non-constant step.
4160 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4163 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4164 destination if any_malformed_asm.
4166 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4169 * expr.c (store_field): In the bitfield case, if the value comes from
4170 a function call and is returned in registers by means of a PARALLEL,
4171 do not change the mode of the temporary unless BLKmode and VOIDmode.
4173 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4175 PR rtl-optimization/85423
4176 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4177 dependencies to debug insns when the previous insn is non-debug.
4179 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4181 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4182 enums into a single definition.
4183 (fls): Fix predicates and printing.
4186 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4188 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4189 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4190 and short u6 immediate.
4191 (check_if_valid_sleep_operand): Remove.
4192 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4194 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4196 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4197 flag_always_save_lp condition.
4198 * config/nds32/nds32.opt (malways-save-lp): New option.
4200 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4202 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4203 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4204 * config/nds32/nds32.h
4205 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4206 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4208 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4210 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4211 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4213 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4214 Chung-Ju Wu <jasonwucj@gmail.com>
4216 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4218 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4219 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4221 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4223 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4225 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4227 * config/nds32/nds32-protos.h (nds32_data_alignment,
4228 nds32_local_alignment): Declare.
4229 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4230 nds32_local_alignment): New functions.
4231 (TARGET_CONSTANT_ALIGNMENT): Define.
4232 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4234 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32.c
4237 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4238 (TARGET_MODES_TIEABLE_P): Likewise.
4240 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4242 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4245 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4246 Chung-Ju Wu <jasonwucj@gmail.com>
4248 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4249 for unaligned access.
4250 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4252 * config/nds32/nds32-intrinsic.md: Likewise.
4253 * config/nds32/nds32_intrinsic.h: Likewise.
4254 * config/nds32/nds32.h (nds32_builtins): Likewise.
4255 * config/nds32/nds32.opt (munaligned-access): New option.
4256 * config/nds32/nds32.c (nds32_asm_file_start): Display
4257 flag_unaligned_access status.
4259 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4261 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4262 -mno-relax is present.
4263 * config/riscv/linux.h (LINK_SPEC): Ditto.
4265 2018-04-20 Martin Sebor <msebor@redhat.com>
4268 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4270 (gimple_fold_builtin_stxcpy_chk): Same.
4271 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4273 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4276 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4277 __powikf2 when long double is IEEE 128-bit.
4279 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4281 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4282 step to make sure stack always aligned.
4284 2018-04-20 Carl Love <cel@us.ibm.com>
4287 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4288 size check for arg0.
4290 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4291 Tom de Vries <tom@codesourcery.com>
4294 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4295 Emit insns for calls too.
4296 (nvptx_find_par): Always look for worker-level predecessor insn.
4297 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4299 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4300 (nvptx_process_pars): Propagate frames for calls.
4302 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4305 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4307 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4308 (ix86_handle_option): Don't handle OPT_mibt.
4309 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4311 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4313 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4314 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4315 (ix86_target_macros): Define __CET__ with flag_cf_protection
4316 for -fcf-protection.
4317 * config/i386/i386.c (isa2_opts): Remove -mibt.
4318 * config/i386/i386.h (TARGET_IBT): Removed.
4319 (TARGET_IBT_P): Likewise.
4320 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4321 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4322 * config/i386/i386.opt (mcet): Update help message.
4325 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4326 -mcet as an alias for -mshstk.
4328 2018-04-20 Richard Biener <rguenther@suse.de>
4331 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4332 complexity by forcing a single use of the multiply operand.
4334 2018-04-20 Martin Jambor <mjambor@suse.cz>
4337 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4338 recursion dependency to only apply to non-clones.
4340 2018-04-20 Martin Jambor <mjambor@suse.cz>
4343 * ipa-cp.c (create_specialized_node): Check that clones of
4344 self-recursive edges exist during IPA-CP.
4346 2018-04-19 Toon Moene <toon@moene.org>
4348 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4351 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4353 PR tree-optimization/85467
4354 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4355 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4356 VECTOR_CST element to type.
4358 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4361 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4362 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4363 (builtin_longjmp): Likewise.
4364 (save_stack_nonlocal): New pattern.
4365 (restore_stack_nonlocal): Likewise.
4367 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4370 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4371 Replace ASM_OUTPUT_LABEL with fprintf.
4373 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4376 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4377 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4378 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4379 define __IBT__ and __SHSTK__ for -fcf-protection.
4380 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4382 (ix86_trampoline_init): Likewise.
4383 (x86_output_mi_thunk): Likewise.
4384 (ix86_notrack_prefixed_insn_p): Likewise.
4385 (ix86_option_override_internal): Don't disallow -fcf-protection.
4386 * config/i386/i386.md (rdssp<mode>): Also enable for
4388 (incssp<mode>): Likewise.
4389 (nop_endbr): Likewise.
4390 * config/i386/i386.opt (mcet): Change help message to built-in
4394 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4395 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4396 enable CET built-in functions.
4398 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4400 * common/config/i386/i386-common.c
4401 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4402 OPTION_MASK_ISA_MOVDIRI_UNSET,
4403 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4404 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4405 * config.gcc (movdirintrin.h): New header.
4406 * config/i386/cpuid.h (bit_MOVDIRI,
4407 bit_MOVDIR64B): New bits.
4408 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4410 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4411 (VOID, PVOID, PCVOID)): New function types.
4412 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4413 __builtin_ia32_directstoreu_u64,
4414 __builtin_ia32_movdir64b): New builtins.
4415 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4416 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4418 (ix86_valid_target_attribute_inner_p): Ditto.
4419 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4420 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4421 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4422 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4423 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4424 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4425 (movdiri<mode>, movdir64b_<mode>): New.
4426 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4427 * config/i386/immintrin.h: Include movdirintrin.h.
4428 * config/i386/movdirintrin.h: New file.
4429 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4431 2018-04-19 Richard Biener <rguenther@suse.de>
4434 * cfg.c (clear_bb_flags): When loop state says we have
4435 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4437 2018-04-19 Richard Biener <rguenther@suse.de>
4439 PR tree-optimization/84737
4440 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4441 copying restrict info.
4442 (vect_setup_realignment): Use it.
4443 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4444 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4445 the first DR to all generated stores.
4446 (vectorizable_load): Likewise for loads.
4448 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4450 PR tree-optimization/85446
4451 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4452 the integral and pointer types to have the same precision.
4454 * doc/install.texi: Document --disable-cet being the default and
4457 2018-04-18 Martin Liska <mliska@suse.cz>
4459 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4462 2018-04-18 Martin Liska <mliska@suse.cz>
4465 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4468 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4469 arguments if they are comparable.
4471 2018-04-18 Martin Liska <mliska@suse.cz>
4474 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4477 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4480 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4483 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4484 ENDBR after calling __morestack.
4486 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4489 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4490 by using gcc_base_ver to generate a gcc_driver_version, and use
4491 it when generating GCC_DRIVER_NAME.
4492 * configure: Regenerate.
4494 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4497 * config.gcc: Obsolete powerpc*-*-*spe*.
4499 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4502 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4503 (stabstr_D): Change type of unum from unsigned int to
4504 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4507 2018-04-17 Jim Wilson <jimw@sifive.com>
4510 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4511 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4512 Set arg_pointer_offset after using pretend_args_size.
4514 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4516 PR rtl-optimization/85431
4517 * dse.c (record_store): Ignore zero width stores.
4520 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4521 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4522 __builtin_stack_restore rather than after it.
4523 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4524 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4525 argument instead of virtual_dynamic_stack_rtx.
4527 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4529 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4531 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4532 Add note to error message to explain internal mapping of overloaded
4533 built-in function name to non-overloaded built-in function name.
4534 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4537 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4540 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4541 where the inputs overlap with the output.
4543 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4545 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4546 (=v, v) alternative and explicit "memory" attribute.
4547 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4548 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4550 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4551 "sselog1" type instead of "sselog".
4552 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4553 "sselog". Remove explicit "memory" attribute.
4554 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4555 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4557 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4558 "sselog1" type instead of "sselog". Remove explicit "memory"
4560 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4561 use "sselog1" type instead of "sselog". Remove explicit "memory"
4563 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4564 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4566 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4567 "sselog1" type instead of "sselog". Remove explicit "memory"
4569 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4570 use "sselog1" type instead of "sselog". Remove explicit "memory"
4574 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4577 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4580 2018-04-17 Martin Jambor <mjambor@suse.cz>
4583 * ipa-cp.c (create_specialized_node): Call
4584 expand_all_artificial_thunks if necessary.
4586 2018-04-17 Martin Liska <mliska@suse.cz>
4589 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4590 in message, remote space in between '_G' and '('.
4592 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4595 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4596 avx512f_vmcmp<mode>3<round_saeonly_name>,
4597 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4598 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4599 avx512f_rndscale<mode><round_saeonly_name>,
4600 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4601 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4602 Use %<iptr>2 instead of %2 for -masm=intel.
4603 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4604 avx512f_vcvttss2usi<round_saeonly_name>,
4605 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4607 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4608 avx512f_vcvttsd2usi<round_saeonly_name>,
4609 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4610 Use %q1 instead of %1 for -masm=intel.
4611 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4612 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4613 of %3 for -masm=intel.
4614 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4616 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4618 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4620 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4621 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4623 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4625 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4626 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4627 %g1 and one with %0 and %1.
4628 (avx512er_vmrcp28<mode><round_saeonly_name>,
4629 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4631 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4632 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4633 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4634 of %0 and %{%4%} for -masm=intel.
4635 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4636 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4637 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4638 order of %0 and %{%5%}%{z%} for -masm=intel.
4640 2018-04-17 Jan Hubicka <jh@suse.cz>
4643 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4645 2018-04-17 Martin Liska <mliska@suse.cz>
4648 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4649 for target_clone error message. Make default implementation
4650 clone to be a local declaration.
4651 (separate_attrs): Add new argument and check for an empty
4653 (expand_target_clones): Handle it.
4654 (ipa_target_clone): Make redirection just for target_clones
4657 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4658 Tom de Vries <tom@codesourcery.com>
4661 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4662 tiled basic blocks without omp continue statements.
4664 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4667 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4668 vec_extract expression as having side effects to make sure it gets
4671 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4674 * config/i386/i386.c (get_builtin_code_for_version): Check
4677 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4680 * gcc/config.gcc: Support "skylake".
4681 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4683 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4684 (processor_target_table): Add "skylake".
4685 (ix86_option_override_internal): Add "skylake".
4686 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4687 PROCESSOR_CANNONLAKE.
4688 (get_builtin_code_for_version): Fix priority for
4689 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4690 PROCESSOR_SKYLAKE-AVX512.
4691 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4692 (processor_type): Add PROCESSOR_SKYLAKE.
4694 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4695 Jason Merrill <jason@redhat.com>
4698 * convert.c (convert_to_integer_1): Use direct recursion for
4699 enumeral types and types with a precision less than the number
4700 of bits in their mode.
4702 2018-04-16 Julia Koval <julia.koval@intel.com>
4705 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4706 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4708 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4711 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4712 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4713 and -mno-direct-move.
4715 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4718 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4719 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4720 Ensure negative shifts result in {0}.
4722 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4724 PR rtl-optimization/79916
4725 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4726 regs (if any) to define how to gnerate SD moves when LRA is in
4729 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4731 PR rtl-optimization/85393
4732 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4733 * except.c (expand_dw2_landing_pad_for_region): Make static.
4734 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4735 a label and unconditional jump to old_bb, rather than
4736 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4739 PR rtl-optimization/85376
4740 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4741 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4742 instead of a specific value.
4744 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4745 Bin Cheng <bin.cheng@arm.com>
4747 PR tree-optimization/82965
4748 PR tree-optimization/83991
4749 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4750 by_profile_only parameter.
4751 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4752 information if the loop was predicted to iterate too many times.
4753 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4755 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4758 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4761 2018-04-13 Martin Liska <mliska@suse.cz>
4762 Jakub Jelinek <jakub@redhat.com>
4765 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4766 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4767 * builtins.c (expand_builtin_memory_copy_args): Use
4768 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4769 handle dest_addr == pc_rtx.
4771 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4774 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4775 asked to not generate direct moves.
4776 (fix_trunc<mode>si2_stfiwx): Similar.
4777 (fix_trunc<mode>si2_internal): Similar.
4779 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4782 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4783 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4784 lookup if dest in some wider mode is known to be const0_rtx and
4785 if so, record permanent equivalence for it to be ZERO_EXTEND of
4786 the narrower mode destination.
4788 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4790 * lto-streamer-out.c (output_function): Revert 259346.
4791 * omp-expand.c (expand_oacc_for): Likewise.
4793 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4795 PR rtl-optimization/85354
4796 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4797 * sel-sched.c (sel_global_init): ... here.
4799 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4802 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4803 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4804 mode for PE-COFF targets.
4805 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4806 (i386_pe_asm_lto_end): Likewise.
4807 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4808 (TARGET_ASM_LTO_END): Likewise.
4809 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4810 (i386_pe_asm_lto_start): New function.
4811 (i386_pe_asm_lto_end): Likewise.
4813 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4814 Richard Biener <rguenther@suse.de>
4817 * lto-streamer-out.c (output_function): Fix CFG loop state before
4819 * omp-expand.c (expand_oacc_for): Handle calls to internal
4820 functions like regular functions.
4822 2018-04-12 Richard Biener <rguenther@suse.de>
4825 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4826 for the early LTO debug to properly generate references to it
4827 during DIE emission. Do not re-use that for the skeleton for
4829 (dwarf2out_early_finish): Likewise.
4831 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4834 * config/i386/sse.md
4835 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4836 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4837 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4838 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4839 and output is a reg, avoid creating invalid lowpart subreg, but
4840 instead split into a 512-bit move. Don't split if not AVX512VL,
4841 input is xmm16+ reg and output is a mem.
4842 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4843 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4844 xmm16+ reg and output is a mem.
4846 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4848 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4849 also for flag_dwarf2_cfi_asm.
4851 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4853 PR rtl-optimization/85342
4854 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4855 a bool scalar var inside of the loop instead. Don't try to update
4856 recog_data.operand after failed apply_change_group.
4858 2018-04-12 Tom de Vries <tom@codesourcery.com>
4861 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4862 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4863 array with flexible array member as array without given dimension.
4864 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4865 argument for undefined param to true.
4867 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4870 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4871 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4872 from PowerPC section.
4873 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4874 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4875 help text that is too long.
4876 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4877 help text that is too long.
4878 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4879 help text that is too long.
4881 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4883 * config/alpha/alpha.md (stack_probe_internal): Rename
4884 from "probe_stack". Update all callers.
4886 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4888 PR rtl-optimization/84566
4889 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4890 sched_macro_fuse_insns.
4892 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4895 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4896 (compute_block_dependences): ... from here.
4898 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4900 PR tree-optimization/85331
4901 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4902 from int to HOST_WIDE_INT.
4904 2018-04-11 Martin Jambor <mjambor@suse.cz>
4907 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4908 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4909 not the same as the source val.
4910 (cgraph_edge_brings_value_p): New parameter.
4911 (gather_edges_for_value): Pass destination value to
4912 cgraph_edge_brings_value_p.
4913 (perhaps_add_new_callers): Likewise.
4914 (get_info_about_necessary_edges): Likewise and exclude values brought
4915 only by self-recursive edges.
4916 (create_specialized_node): Redirect only clones of self-calling edges.
4917 (+self_recursive_pass_through_p): New function.
4918 (find_more_scalar_values_for_callers_subset): Use it.
4919 (find_aggregate_values_for_callers_subset): Likewise.
4920 (known_aggs_to_agg_replacement_list): Removed.
4921 (decide_whether_version_node): Re-calculate known constants for all
4922 remaining context clones.
4924 2018-04-11 Richard Biener <rguenther@suse.de>
4927 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4928 from early DWARF output.
4929 (dwarf2out_early_finish): Output line info unconditionally into
4930 early DWARF and add reference to it.
4932 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4935 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4936 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4937 other than V2DFmode using iptr mode attribute.
4938 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4940 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4942 PR rtl-optimization/84659
4943 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4945 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4948 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4950 (output_loc_list): Pass address of a dummy size variable even in the
4951 locview handling loop.
4952 (index_location_lists): Add comment on why skip_loc_list_entry can't
4955 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4958 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4961 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4964 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4965 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4966 and -mstring-compare-inline-limit.
4968 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4971 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4972 for stack clash protection in a register whenever we need it to be in
4975 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4977 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4978 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4980 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4983 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4985 (mlong-double-): Ditto.
4986 * config/rs6000/sysv4.opt (msdata=): Ditto.
4987 (mtls-size=): Ditto.
4989 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4991 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4992 erroneous entries for
4993 "vector int vec_ldl (int, long int *)", and
4994 "vector unsigned int vec_ldl (int, unsigned long int *)".
4995 Add comments and entries for
4996 "vector bool char vec_ldl (int, bool char *)",
4997 "vector bool short vec_ldl (int, bool short *)",
4998 "vector bool int vec_ldl (int, bool int *)",
4999 "vector bool long long vec_ldl (int, bool long long *)",
5000 "vector pixel vec_ldl (int, pixel *)",
5001 "vector long long vec_ldl (int, long long *)",
5002 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5003 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5004 type tree bool_long_long_type_node and correct definition of
5005 bool_V2DI_type_node to make reference to this new type tree.
5006 (rs6000_mangle_type): Replace erroneous reference to
5007 bool_long_type_node with bool_long_long_type_node.
5008 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5009 comments to emphasize sign distinctions for char and int types and
5010 replace RS6000_BTI_bool_long constant with
5011 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5012 use of RS6000_BTI_pixel.
5013 (bool_long_type_node): Remove this macro definition.
5014 (bool_long_long_type_node): New macro definition
5016 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5018 PR rtl-optimization/85300
5019 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5020 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5021 simplify_unary_operation fails.
5023 2018-04-10 Martin Liska <mliska@suse.cz>
5025 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5026 cgraph_edge and ipa_ref.
5028 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5032 * config/i386/sse.md
5033 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5034 computation of the VEC_MERGE selector from mask.
5035 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5036 Fix decoding of the VEC_MERGE selector into mask.
5038 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5040 PR tree-optimization/85286
5041 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5043 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5045 * final.c (final_1): Set insn_last_address as well as
5046 insn_current_address.
5048 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5051 * explow.c (emit_stack_probe): Call validize_mem on memory location
5052 before passing it to gen_probe_stack. Create address operand and
5053 legitimize it for the probe_stack_address case.
5055 2018-04-09 Jan Hubicka <jh@suse.cz>
5058 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5059 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5060 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5062 (free_lang_data): Rebuild type inheritance graph.
5064 2018-04-09 Martin Sebor <msebor@redhat.com>
5066 * invoke.texi (-finline-small-functions): Mention other optimization
5068 (-findirect-inlining, -fpartial-inlining): Same.
5069 (-finline-functions-called-once): Same.
5070 (-freorder-blocks-and-partition): Same.
5072 2018-04-09 Jan Hubicka <jh@suse.cz>
5075 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5076 jumps; choose last target that matches the criteria (i.e.
5077 no partition changes for non-crossing jumps).
5078 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5079 support for redirecting crossing jumps to non-crossing.
5081 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5083 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5084 also for naked functions.
5086 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5088 * config/arc/arc.md (add_shift): New pattern.
5089 (add_shift2): Likewise.
5090 (sub_shift): Likewise.
5091 (sub_shift_cmp0_noout): Likewise.
5092 (compare_si_ashiftsi): Likewise.
5093 (xbfu_cmp0_noout): New combine pattern.
5094 (xbfu_cmp0"): Likewise.
5095 (movsi_set_cc_insn): Place the predicable variant first.
5096 (commutative_binary_cmp0_noout): Remove clobber.
5097 (commutative_binary_cmp0): New pattern.
5098 (noncommutative_binary_cmp0): Likewise.
5099 (noncommutative_binary_cmp0_noout): Likewise.
5100 (noncommutative_binary_comparison_result_used): Removed.
5101 (rsub_cmp0): New pattern.
5102 (rsub_cmp0_noout): Likewise.
5103 (extzvsi): Changed, keep only meaningful variants.
5104 (SQH, SEZ): New iterators.
5105 (SQH_postfix): New mode attribute.
5106 (SEZ_prefix): New code attribute.
5107 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5108 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5109 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5111 (noncommutative_operator): Check the availability of barrel
5114 2018-04-09 Richard Biener <rguenther@suse.de>
5116 PR tree-optimization/85284
5117 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5118 Only use the niter constraining form of simple_iv when the exit
5121 2018-04-09 Tom de Vries <tom@codesourcery.com>
5124 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5125 (define_expand "*memory_barrier"): New define_expand.
5126 (define_insn "memory_barrier"): New insn.
5128 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5130 PR rtl-optimization/80463
5131 PR rtl-optimization/83972
5132 PR rtl-optimization/83480
5134 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5135 correct producer for the insn.
5136 (tidy_control_flow): Fixup seqnos in case of debug insns.
5138 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5140 PR rtl-optimization/83913
5142 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5143 different sched-times when merging exprs.
5145 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5147 PR rtl-optimization/83962
5149 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5150 tidy_fallthru_edge and tidy_control_flow.
5152 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5154 PR rtl-optimization/83530
5156 * sel-sched.c (force_next_insn): New global variable.
5157 (remove_insn_for_debug): When force_next_insn is true, also leave only
5158 next insn in the ready list.
5159 (sel_sched_region): When the region wasn't scheduled, make another pass
5160 over it with force_next_insn set to 1.
5162 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5164 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5166 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5167 for interrupt control.
5168 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5169 functions for interrupt control.
5170 * config/nds32/nds32-intrinsic.md: Likewise.
5171 * config/nds32/nds32_intrinsic.h: Likewise.
5172 * config/nds32/nds32.h (nds32_builtins): Likewise.
5174 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5176 * config/nds32/nds32.c (nds32_init_machine_status,
5177 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5178 strict_aligned_p field.
5179 (nds32_expand_to_rtl_hook): New function.
5180 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5181 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5183 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5184 Chung-Ju Wu <jasonwucj@gmail.com>
5186 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5187 * config/nds32/nds32-n7.md: New file.
5188 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5189 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5191 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5192 * config/nds32/nds32.md (pipeline_model): Add n7.
5193 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5194 * config/nds32/pipelines.md: Include n7 settings.
5196 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5197 Chung-Ju Wu <jasonwucj@gmail.com>
5199 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5200 * config/nds32/nds32-e8.md: New file.
5201 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5202 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5204 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5205 * config/nds32/nds32.md (pipeline_model): Add e8.
5206 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5207 * config/nds32/pipelines.md: Include e8 settings.
5209 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5210 Chung-Ju Wu <jasonwucj@gmail.com>
5212 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5213 * config/nds32/nds32-n8.md: New file.
5214 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5215 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5217 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5218 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5219 * config/nds32/nds32.md (pipeline_model): Add n8.
5220 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5221 * config/nds32/pipelines.md: Include n8 settings.
5223 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5224 Chung-Ju Wu <jasonwucj@gmail.com>
5226 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5227 * config/nds32/nds32-n9-2r1w.md: New file.
5228 * config/nds32/nds32-n9-3r2w.md: New file.
5229 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5230 nds32_register_ports): New or modify for cpu n9.
5231 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5233 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5234 * config/nds32/nds32-utils.c: New file.
5235 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5236 TARGET_MUL_SLOW): Define.
5237 * config/nds32/nds32.md (pipeline_model): New attribute.
5238 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5239 New options that support cpu n9.
5240 * config/nds32/pipelines.md: Include n9 settings.
5241 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5243 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5245 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5246 information if necessary.
5247 (output_cond_branch_compare_zero): Likewise.
5248 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5249 (nds32_target_alignment): Refine for alignment.
5250 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5251 (FUNCTION_BOUNDARY): Modify.
5252 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5254 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5256 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5258 * config/nds32/constants.md (unspec_volatile_element): Add values for
5259 TLB operation and data prefetch.
5260 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5261 functions for TLB operation and data prefetch.
5262 * config/nds32/nds32-intrinsic.md: Likewise.
5263 * config/nds32/nds32_intrinsic.h: Likewise.
5264 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5265 (nds32_print_operand): Likewise.
5266 * config/nds32/nds32.h (nds32_builtins): Likewise.
5268 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5269 Andrew Pinski <pinsika@gcc.gnu.org>
5272 * match.pd: Use constant_boolean_node of correct type instead of
5273 boolean_true_node or boolean_false_node for simplifying
5274 pointer comparisons to zero.
5276 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5278 PR tree-optimization/80021
5279 * tree.c (verify_type_variant): Make error call in verify_variant_match
5280 translatable and remove final full stop.
5282 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5284 * config/nds32/constants.md (unspec_volatile_element): Add
5285 UNSPEC_VOLATILE_EH_RETURN.
5286 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5287 nds32_output_stack_pop): Support dwarf exception handling process.
5288 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5289 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5290 exception handling process.
5291 (nds32_compute_stack_frame): Likewise.
5292 (nds32_return_addr_rtx): Likewise.
5293 (nds32_initial_elimination_offset): Likewise.
5294 (nds32_expand_prologue): Likewise.
5295 (nds32_expand_epilogue): Likewise.
5296 (nds32_dynamic_chain_address): New function.
5297 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5299 (DYNAMIC_CHAIN_ADDRESS): Define.
5300 (EH_RETURN_DATA_REGNO): Define.
5301 (EH_RETURN_STACKADJ_RTX): Define.
5302 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5303 patterns for dwarf exception handling.
5305 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5307 * config/nds32/nds32.h: Clean up obsolete macros.
5309 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5311 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5312 Add enum values for particular instructions.
5313 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5314 particular intrinsic functions.
5315 * config/nds32/nds32-intrinsic.md: Likewise.
5316 * config/nds32/nds32_intrinsic.h: Likewise.
5317 * config/nds32/nds32.h (nds32_builtins): Likewise.
5318 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5319 (btst, ave): New patterns for particular instructions.
5321 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5323 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5324 Add enum values for atomic load/store and memory sync.
5325 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5327 * config/nds32/nds32-intrinsic.md: Likewise.
5328 * config/nds32/nds32_intrinsic.h: Likewise.
5329 * config/nds32/nds32.h (nds32_builtins): Likewise.
5331 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5333 PR tree-optimization/85257
5334 * fold-const.c (native_encode_vector): If not all elts could fit
5335 and off is -1, return 0 rather than offset.
5336 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5337 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5338 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5339 adjust buffer in native_interpret_expr call.
5341 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5343 * config/nds32/constants.md (unspec_volatile_element): Add cache
5344 control enum values.
5345 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5346 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5347 * config/nds32/nds32.c (nds32_cctl_names): New.
5348 (nds32_print_operand): Handle cache control register names.
5349 * config/nds32/nds32.h (nds32_builtins): New enum values.
5350 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5352 * config/nds32/nds32.md (type): Add mmu.
5353 * config/nds32/pipelines.md (simple_insn): Add mmu.
5355 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5357 * config/nds32/nds32.md (type): Remove call.
5358 * config/nds32/pipelines.md (simple_insn): Likewise.
5360 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5362 * config/nds32/constants.md (unspec_volatile_element): Add
5363 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5364 UNSPEC_VOLATILE_FMFCFG.
5365 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5366 description for fmfcfg and fmfcsr.
5367 (bdesc_1arg): Add fmtcsr.
5368 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5369 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5370 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5371 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5372 unspec_fmfcfg): New patterns.
5373 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5374 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5375 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5376 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5377 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5378 __nds32__fmfcfg): Define.
5380 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5382 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5383 intrinsic register names.
5384 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5385 intrinsic register enum values and macros.
5387 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5389 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5390 for load/store addressing form.
5391 (nds32_print_operand_address): Likewise.
5393 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5396 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5397 based on LABEL_REF. Remove useless assertion.
5398 (pic_address_needs_scratch): Fix formatting.
5399 (sparc_legitimize_pic_address): Minor tweaks.
5400 (sparc_delegitimize_address): Adjust assertion accordingly.
5401 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5402 into symbolic_operand.
5403 (movsi_high_pic_label_ref): Likewise.
5404 (movsi_lo_sum_pic_label_ref): Likewise.
5405 (movdi_pic_label_ref): Likewise.
5406 (movdi_high_pic_label_ref): Likewise.
5407 (movdi_lo_sum_pic_label_ref): Likewise.
5409 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5411 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5412 custom LIB_SPEC setup.
5414 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5415 Kito Cheng <kito.cheng@gmail.com>
5417 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5418 * config/riscv/freebsd.h: New.
5420 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5422 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5423 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5426 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5427 Kito Cheng <kito.cheng@gmail.com>
5429 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5430 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5431 * config/nds32/nds32-protos.h (nds32_output_call,
5432 nds32_output_return): Declare.
5433 * config/nds32/nds32.md: Refine all the call and return patterns.
5435 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5438 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5439 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5441 PR rtl-optimization/84872
5442 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5443 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5446 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5448 * expr.c (copy_blkmode_to_reg): Revert 254862.
5449 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5451 2018-04-06 Richard Biener <rguenther@suse.de>
5454 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5455 after seeing a component reference with an adjacent field. Treat
5456 refs to arrays at struct end of external decls similar to
5457 refs to unconstrained commons.
5459 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5462 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5463 look through SAVE_EXPRs with non-side-effects argument. Adjust
5465 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5468 2018-04-06 Richard Biener <rguenther@suse.de>
5471 * alias.c (find_base_term): New wrapper around find_base_term
5472 unwinding CSELIB_VAL_PTR changes.
5473 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5476 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5478 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5480 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5481 constant definitions.
5482 ("nop"): lr 0,0 -> nopr r0
5483 ("nop_lr0", "nop_lr1"): New insn definitions.
5485 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5487 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5488 NDS32_V3PUSH_AVAILABLE_P macro.
5490 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5491 Chung-Ju Wu <jasonwucj@gmail.com>
5493 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5494 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5495 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5496 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5497 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5498 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5499 * config/nds32/constraints.md: New constraints and checking for hard
5500 float configuration.
5501 * config/nds32/iterators.md: New mode iterator and attribute for hard
5502 float configuration.
5503 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5505 * config/nds32/nds32-fpu.md: New file.
5506 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5507 deal with hard float code generation.
5508 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5510 (abi_type, float_reg_number): New enum type.
5511 * config/nds32/nds32-predicates.c: New predicates for hard float.
5512 * config/nds32/nds32-protos.h: Declare functions for hard float.
5513 * config/nds32/nds32.c: Implementation for hard float configuration.
5514 * config/nds32/nds32.h: Definitions for hard float configuration.
5515 * config/nds32/nds32.md: Include hard float machine description and
5516 modify patterns for hard float configuration.
5517 * config/nds32/nds32.opt: New options for hard float configuration.
5518 * config/nds32/predicates.md: New predicates for hard float
5521 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5523 * common/config/nds32/nds32-common.c
5524 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5526 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5529 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5530 CONSTRUCTOR_ELT (ctor, ...)->value.
5532 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5535 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5537 2018-04-05 Tom de Vries <tom@codesourcery.com>
5540 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5543 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5544 Kito Cheng <kito.cheng@gmail.com>
5546 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5547 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5548 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5549 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5551 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5552 Kito Cheng <kito.cheng@gmail.com>
5554 * config/nds32/constraints.md (Ufe): New memory constraint.
5555 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5556 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5557 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5559 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5560 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5562 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5564 * config/nds32/nds32.md: Use optimize_size in the condition for
5565 alu-shift instructions.
5567 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5569 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5571 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5573 * config/nds32/nds32.md (negsi2): Refine pattern.
5575 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5576 Chung-Ju Wu <jasonwucj@gmail.com>
5578 * config/nds32/iterators.md (shift_rotate): New code iterator.
5579 (shift): New code attribute.
5580 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5581 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5582 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5583 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5584 bit-wise operations.
5585 (andsi3, *andsi3): Ditto.
5586 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5587 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5588 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5589 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5590 nds32_ior_operand, nds32_xor_operand): New predicates.
5592 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5594 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5595 (addsi3, subsi3): ... this.
5597 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5599 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5601 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5603 * config/nds32/nds32.md: Adjust indention.
5605 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5607 * config/nds32/nds32.md (feature): New attribute.
5609 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5611 * config/nds32/nds32.md (subtype): New attribute.
5613 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5616 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5617 expansion to perform a bitwise AND of the argument followed by a
5618 boolean negation of the result.
5620 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5622 PR rtl-optimization/84878
5623 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5624 the basic block. Assert the use reference is not artificial and that
5625 it has an associated insn.
5627 2018-04-04 Michael Matz <matz@suse.de>
5629 * builtins.c (compute_objsize): Pass correct operand
5630 to array_at_struct_end_p.
5632 2018-04-04 Richard Biener <rguenther@suse.de>
5635 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5636 from contexts for DINFO_LEVEL_TERSE and below.
5638 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5640 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5641 resiter_operand condition.
5642 * config/nds32/nds32.md (*move<mode>): Ditto.
5644 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5645 Monk Chiang <sh.chiang04@gmail.com>
5647 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5649 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5651 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5653 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5654 Kito Cheng <kito.cheng@gmail.com>
5656 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5657 nds32_cond_code_str, output_cond_branch,
5658 output_cond_branch_compare_zero, nds32_expand_cbranch,
5659 nds32_expand_cstore, nds32_expand_movcc,
5660 nds32_output_cbranchsi4_equality_zero,
5661 nds32_output_cbranchsi4_equality_reg,
5662 nds32_output_cbranchsi4_equality_reg_or_const_int,
5663 nds32_output_cbranchsi4_greater_less_zero: New functions.
5664 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5665 nds32_expand_cstore, nds32_expand_movcc,
5666 nds32_output_cbranchsi4_equality_zero,
5667 nds32_output_cbranchsi4_equality_reg,
5668 nds32_output_cbranchsi4_equality_reg_or_const_int,
5669 nds32_output_cbranchsi4_greater_less_zero): Declare.
5670 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5671 nds32_rimm11s_operand): New predicates.
5672 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5673 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5676 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5678 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5679 * config/nds32/nds32.md: Ditto.
5680 * config/nds32/pipelines.md: Ditto.
5682 2018-04-04 Richard Biener <rguenther@suse.de>
5684 PR tree-optimization/85168
5685 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5686 propagating abnormals.
5688 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5690 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5692 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5693 Kito Cheng <kito.cheng@gmail.com>
5695 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5696 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5697 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5698 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5699 * config/nds32/nds32.md (sibcall_internal): New.
5700 (sibcall_register): Remove.
5701 (sibcall_immediate): Remove.
5702 (sibcall_value_internal): New.
5703 (sibcall_value_register): Remove.
5704 (sibcall_value_immediate): Remove.
5705 * config/nds32/predicates.md (nds32_general_register_operand): New.
5706 (nds32_call_address_operand): New.
5708 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5710 PR rtl-optimization/85167
5711 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5712 bb_defs if *split_p, instead preinitialize it to NULL.
5714 PR tree-optimization/85156
5715 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5716 evaluating the argument multiple times.
5718 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5720 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5722 (_mm_cvtpd_ps): Likewise.
5723 (_mm_cvttpd_epi32): Likewise.
5724 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5725 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5726 vector, pixel, and bool following altivec.h include.
5728 2018-04-03 Martin Sebor <msebor@redhat.com>
5730 * doc/extend.texi (Common Function Attributes): Clarify.
5731 (const attribute): Likewise.
5732 (pure attribute): Likewise.
5734 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5737 * config/i386/i386.c (ix86_expand_vector_set): Use
5738 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5740 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5742 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5743 instructions when changing rounding bits to preserve precision bits
5744 in the x87 control word.
5746 2018-04-03 Martin Liska <mliska@suse.cz>
5748 PR tree-optimization/82491
5749 * rtl.h (strip_offset_and_add): Replace += suboffset with
5750 poly_uint64 () + suboffset.
5752 2018-03-29 Martin Liska <mliska@suse.cz>
5753 Martin Jambor <mjambor@suse.cz>
5756 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5757 param_type is not an integral or pointer type.
5759 2018-04-03 Richard Biener <rguenther@suse.de>
5761 * sese.h (recompute_all_dominators): Remove.
5763 2018-04-02 Martin Sebor <msebor@redhat.com>
5765 * doc/invoke.texi (-Wrestrict): Fix typos.
5767 2018-04-02 Jim Wilson <jimw@sifive.com>
5769 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5770 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5771 (<optab>di3, <optab>si3_extend): Likewise.
5772 (<optab>si3_mask, <optab>si3_mask_1): New.
5773 (<optab>di3_mask, <optab>di3_mask_1): New.
5774 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5775 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5776 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5778 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5780 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5783 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5785 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5786 (nds32_canonicalize_comparison): New function.
5788 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5789 Kito Cheng <kito.cheng@gmail.com>
5790 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5792 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5793 * config/nds32/constants.md (unspec_volatile_element): Add
5794 UNSPEC_VOLATILE_RELAX_GROUP.
5795 * config/nds32/nds32-relax-opt.c: New file.
5796 * config/nds32/nds32-predicates.c
5797 (nds32_symbol_load_store_p): New function.
5798 * config/nds32/nds32-protos.h
5799 (nds32_symbol_load_store_p): Declare function.
5800 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5801 * config/nds32/nds32.c
5802 (nds32_register_pass): New function to register pass.
5803 (nds32_register_passes): New function to register passes.
5804 * config/nds32/nds32.md (relax_group): New pattern.
5805 * config/nds32/nds32.opt (mrelax-hint): New option.
5806 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5808 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5810 * config/nds32/t-nds32: Modify files dependency.
5812 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5814 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5815 (PROFILE_HOOK): Define its implementation.
5817 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5819 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5820 type and 32-bit size.
5822 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5825 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5826 (V_128_256): New mode iterator.
5827 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5828 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5829 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5831 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5832 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5834 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5837 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5838 NaN inputs correctly.
5840 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5843 * config/rs6000/vsx.md (??r): New mode attribute.
5844 (*vsx_mov<mode>_64bit): Use it.
5845 (*vsx_mov<mode>_32bit): Likewise.
5847 2018-03-30 Martin Sebor <msebor@redhat.com>
5849 PR tree-optimization/84818
5850 * builtins.c (check_access): Use warning_n.
5852 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5855 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5857 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5860 2018-03-30 Julia Koval <julia.koval@intel.com>
5863 * x86-tune.def (movx, partial_reg_dependency): Enable for
5866 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5869 * lra-constraints.c (process_alt_operands): Move setting
5870 this_alternative_matches below.
5872 2018-03-29 Martin Liska <mliska@suse.cz>
5875 * doc/invoke.texi: Document how LTO works with debug info.
5876 Describe auto-load support of binutils. Mention 'x86-64'
5877 as valid option value of -march option.
5879 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5881 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5884 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5885 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5886 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5889 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5892 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5893 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5894 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5895 for RS6000_BTM_POWERPC64.
5896 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5897 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5898 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5903 2018-03-28 Carl Love <cel@us.ibm.com>
5906 2017-09-27 Carl Love <cel@us.ibm.com>
5908 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5909 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5910 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5913 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5915 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5916 instead of __vector bool.
5917 (_mm_max_pu8): Likewise.
5918 (_mm_min_pi16): Likewise.
5920 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5923 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5924 (DIVWEUO): Likewise.
5926 (DIVDEUO): Likewise.
5927 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5928 DIVWEUO and DIVDEUO.
5929 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5930 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5931 (div_extend): Likewise.
5932 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5934 (__builtin_divweuo): Likewise.
5935 (__builtin_divdeo): Likewise.
5936 (__builtin_divdeuo): Likewise.
5938 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5941 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5942 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5944 PR tree-optimization/82004
5945 * gimple-match-head.c (optimize_pow_to_exp): New function.
5946 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5947 Don't fold to exp if optimize_pow_to_exp is false.
5949 2018-03-28 Martin Liska <mliska@suse.cz>
5952 * calls.c (initialize_argument_information): Fix trailing space.
5953 * common.opt: Fix typo and provide better explanation for
5954 -fsanitize-coverage option.
5955 * config/i386/i386.opt: Fix typo.
5957 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5958 Martin Liska <mliska@suse.cz>
5961 * gimplify.c (asan_poison_variable): Don't do the check for
5962 gimplify_omp_ctxp here.
5963 (gimplify_decl_expr): Do it here.
5964 (gimplify_target_expr): Likewise.
5966 2018-03-28 Martin Liska <mliska@suse.cz>
5969 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5970 chkp_type_bounds_count if MPX is not enabled.
5972 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5974 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5976 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5979 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5980 function to create the function decl for complex long double
5981 multiply and divide for -mabi=ieeelongdouble.
5982 (init_float128_ieee): Call it.
5984 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5987 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5988 -fcf-protection=branch -mibt.
5989 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5991 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5994 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5995 arm_word_relocations.
5997 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6000 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6001 extern array declarations.
6003 2018-03-27 Richard Biener <rguenther@suse.de>
6006 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6007 explicit single_use checks.
6009 2018-03-27 Richard Biener <rguenther@suse.de>
6011 PR tree-optimization/85082
6012 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6015 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6017 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6018 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6019 Turn on fasynchronous-unwind-tables and funwind-tables.
6021 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6024 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6025 (*bmi_blsr_<mode>_ccz): Ditto.
6027 2018-03-26 Tom de Vries <tom@codesourcery.com>
6029 PR tree-optimization/85063
6030 * omp-general.c (offloading_function_p): New function. Factor out
6032 * omp-offload.c (pass_omp_target_link::gate): ... here.
6033 * omp-general.h (offloading_function_p): Declare.
6034 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6035 with attribute omp declare target for offloading functions.
6037 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6039 PR tree-optimization/84005
6040 * tree-data-ref.h (get_base_for_alignment): Declare.
6041 * tree-data-ref.c (get_base_for_alignment_1): New function.
6042 (get_base_for_alignment): Likewise.
6043 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6044 get_base_for_alignment to find a suitable base object, instead
6045 of always using drb->base_address.
6047 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6050 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6051 known size by default.
6053 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6056 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6057 and non BLKmode operands.
6059 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6062 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6063 Clean up attributes.
6065 2018-03-23 Richard Biener <rguenther@suse.de>
6068 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6069 we are going to emit early debug for LTO.
6071 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6074 * function.c (match_asm_constraints_1): Don't optimize if input
6075 doesn't satisfy general_operand predicate for output's mode.
6078 * alias.c (write_dependence_p): Don't require for x_canonicalized
6079 non-VOIDmode if x has VOIDmode.
6082 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6083 just don't try to optimize it rather than assert it never happens.
6085 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6087 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6088 macro expansions for definition of ST_INTERNAL_<mode> and
6089 LD_INTERNAL_<mode> builtins.
6090 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6092 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6094 (altivec_expand_st_builtin): Likewise.
6095 (altivec_expand_builtin): Remove calls to deleted functions.
6096 (rs6000_address_for_altivec): Delete this function.
6097 * config/rs6000/vector.md: Remove expands for
6098 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6100 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6103 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6104 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6105 re-computing once computed.
6106 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6107 (arm_init_machine_status): Initialize
6108 machine->static_chain_stack_bytes.
6110 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6113 * doc/extend.texi: Add four new prototypes for vec_ld.
6114 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6115 definitions for more logical presentation.
6116 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6117 entries for V1TI variants of __builtin_altivec_ld builtin.
6118 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6119 handling of V1TI variant of LVX icode pattern.
6120 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6121 (rs6000_gimple_fold_builtin): Likewise.
6122 (altivec_init_builtins): Add code to define
6123 __builtin_altivec_lvx_v1ti function.
6125 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6128 * function.c (match_asm_constraints_1): Don't do the optimization
6129 if input isn't a REG, SUBREG, MEM or constant.
6131 2018-03-22 Tom de Vries <tom@codesourcery.com>
6133 PR tree-optimization/84956
6134 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6135 bb_has_abnormal_pred.
6137 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6140 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6141 DECL_INITIAL (decl) to decl at the end.
6142 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6145 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6147 * doc/extend.texi (__builtin_tgmath): Document when complex
6148 integer types are treated as _Complex _Float64.
6150 2018-03-21 Tom de Vries <tom@codesourcery.com>
6152 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6154 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6156 PR tree-optimization/84960
6157 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6158 if it is ENTRY block, move them into single succ of ENTRY in that case.
6160 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6162 PR tree-optimization/84811
6163 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6164 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6165 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6166 (known_subrange_p): Update accordingly. Cast each value involved
6167 in the size comparison, rather than casting the result of the
6170 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6172 PR tree-optimization/84982
6173 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6174 by flipping the least significant bit rather than all bits from
6175 bitpos to bitpos + bitsize - 1.
6177 2018-03-21 Nathan Sidwell <nathan@acm.org>
6179 * doc/extend.texi (Deprecated Features): Remove mention of
6180 long-deleted deprecations.
6182 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6185 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6186 * configure: Regenerate.
6188 2018-03-21 Tom de Vries <tom@codesourcery.com>
6190 PR tree-optimization/83126
6191 * tree-parloops.c (num_phis): New function.
6192 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6194 2018-03-21 Nathan Sidwell <nathan@acm.org>
6196 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6197 mention anon-struct/union members and trailing attributes.
6199 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6201 PR tree-optimization/84969
6202 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6203 builtin memset partitions if they set different rhs values.
6205 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6207 PR rtl-optimization/84989
6208 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6209 VEC_DUPLICATE with scalar result mode.
6211 2018-03-21 Martin Liska <mliska@suse.cz>
6214 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6215 not intended return statement.
6217 2018-03-21 Martin Liska <mliska@suse.cz>
6220 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6221 (chkp_find_bound_slots_1): Limit number of iterations.
6223 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6226 * Minor grammar fixes for x86 options.
6228 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6231 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6232 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6234 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6237 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6238 (altivec_lvx_<mode>_1op): Likewise.
6239 (altivec_stvx_<mode>_2op): Likewise.
6240 (altivec_stvx_<mode>_1op): Likewise.
6241 (altivec_lvx_<VM2:mode>): New define_expand.
6242 (altivec_stvx_<VM2:mode>): Likewise.
6243 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6244 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6245 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6246 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6247 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6248 (rs6000_gen_lvx): Likewise.
6249 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6250 (altivec_expand_stv_builtin): Likewise.
6251 (altivec_expand_builtin): Likewise.
6252 * config/rs6000/vector.md: Likewise.
6254 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6257 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6260 2018-03-20 Richard Biener <rguenther@suse.de>
6263 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6264 sign-conversions as zero, fall back to standard scalar_stmt
6267 2018-03-20 Martin Liska <mliska@suse.cz>
6270 * predict.c (rebuild_frequencies): Handle case when we have
6271 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6273 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6276 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6277 flag_section_anchors.
6278 * varasm.c (use_blocks_for_decl_p): Remove hack for
6279 dw2_force_const_mem.
6282 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6284 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6285 be created, use lowpart_subreg of operands[0] rather than operands[0]
6287 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6288 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6289 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6290 and n constraint instead of aarch64_shift_imm_di and Usd.
6291 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6292 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6294 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6297 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6298 to favor GPR over NEON registers.
6299 (<shift>di3_neon): Likewise.
6301 2018-03-20 Tom de Vries <tom@codesourcery.com>
6304 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6305 (nvptx_process_pars): Emit bar.sync asap and alap.
6307 2018-03-20 Tom de Vries <tom@codesourcery.com>
6310 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6311 seen_label if seen_label is already set.
6313 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6316 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6317 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6318 Use 1U instead of 1. Formatting fixes.
6321 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6322 instead of TREE_TYPE (s1) for the return value.
6324 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6326 PR tree-optimization/84946
6327 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6328 bitsize + bitsize in poly_uint64 rather than poly_int64.
6331 * dwarf2asm.c: Include fold-const.c.
6332 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6333 of decl rather than decl itself.
6335 PR rtl-optimization/84643
6336 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6338 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6341 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6342 calling assemble_variable.
6344 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6347 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6348 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6350 2018-03-19 Jim Wilson <jimw@sifive.com>
6353 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6354 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6355 (riscv_first_stack_step): Likewise.
6356 (riscv_option_override): Use STACK_BOUNDARY instead of
6358 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6360 (BIGGEST_ALIGNMENT): Set to 128.
6361 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6362 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6365 2018-03-19 Richard Biener <rguenther@suse.de>
6367 PR tree-optimization/84933
6368 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6369 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6371 2018-03-19 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/84859
6374 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6375 (cond_if_else_store_replacement): Perform sinking operation on
6376 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6377 Generalize what a BB with a single eligible store is.
6379 2018-03-19 Richard Biener <rguenther@suse.de>
6381 PR tree-optimization/84929
6382 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6383 chrec_is_positive against non-chrec arg.
6385 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6388 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6390 2018-03-18 Martin Liska <mliska@suse.cz>
6392 PR rtl-optimization/84635
6393 * regrename.c (build_def_use): Use matches_mode only when
6396 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6398 PR tree-optimization/84913
6399 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6400 vectorize chains of COND_EXPRs.
6402 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6404 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6406 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6408 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6410 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6412 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6414 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6415 Kito Cheng <kito.cheng@gmail.com>
6417 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6418 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6419 (nds32_adjust_reg_alloc_order): New function.
6420 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6422 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6424 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6425 nds32_print_operand, nds32_print_operand_address): Use
6426 HOST_WIDE_INT_PRINT_DEC instead.
6428 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6430 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6432 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6435 * config/i386/i386.c (initial_ix86_tune_features,
6436 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6438 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6439 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6440 rather than 1u << ix86_tune. Formatting fix.
6441 (ix86_option_override_internal): Change ix86_arch_mask from
6442 unsigned int to unsigned HOST_WIDE_INT, initialize to
6443 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6444 (ix86_function_specific_restore): Likewise.
6446 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6449 * postreload.c (reload_combine_recognize_pattern): Perform
6450 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6451 truncate_int_for_mode the result for the destination's mode.
6454 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6455 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6457 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6460 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6463 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6464 regno_allocno_class_array and sorted_pseudos.
6465 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6466 insns where regno is used.
6468 2018-03-16 Martin Liska <mliska@suse.cz>
6471 * multiple_target.c (create_dispatcher_calls): Redirect
6472 reference in the symbol table.
6474 2018-03-16 Martin Liska <mliska@suse.cz>
6477 * multiple_target.c (create_dispatcher_calls): Redirect also
6480 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6484 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6485 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6486 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6488 2018-03-16 Julia Koval <julia.koval@intel.com>
6490 * doc/invoke.texi (Skylake Server): Add CLWB.
6491 Cannonlake): Remove CLWB.
6493 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6495 PR tree-optimization/84841
6496 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6498 (FLOAT_ONE_CONST_TYPE): Define.
6499 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6500 (sort_by_operand_rank): Put entries with higher constant_type last
6501 rather than first to match comments.
6503 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6505 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6508 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6511 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6512 check and use address of target always.
6514 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6517 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6518 (indirect_thunk_bnd_needed): Likewise.
6519 (indirect_thunks_used): Likewise.
6520 (indirect_thunks_bnd_used): Likewise.
6521 (indirect_return_needed): New.
6522 (indirect_return_bnd_needed): Likewise.
6523 (output_indirect_thunk_function): Add a bool argument for
6525 (output_indirect_thunk_function): Don't generate alias for
6526 function return thunk.
6527 (ix86_code_end): Call output_indirect_thunk_function to generate
6528 function return thunks.
6529 (ix86_output_function_return): Set indirect_return_bnd_needed
6530 and indirect_return_needed instead of indirect_thunk_bnd_needed
6531 and indirect_thunk_needed.
6533 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6535 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6536 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6537 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6539 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6540 Paul Hua <paul.hua.gm@gmail.com>
6543 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6545 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6547 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6548 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6551 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6554 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6555 instead of GET_MODE_SIZE when comparing Units.
6557 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6560 * varasm.c (hash_section): Return an unchangeble hash value
6561 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6562 Return !aarch64_can_use_per_function_literal_pools_p ().
6564 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6567 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6568 rather than address of cmode as last argument to prepare_cmp_insn.
6570 2018-03-15 Julia Koval <julia.koval@intel.com>
6572 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6573 F_AVX512VNNI, F_AVX512BITALG): New.
6575 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6578 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6579 insn for floating-point loads and stores.
6581 2018-03-14 Carl Love <cel@us.ibm.com>
6583 * config/rs6000/rs6000-c.c: Add macro definitions for
6584 ALTIVEC_BUILTIN_VEC_PERMXOR.
6585 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6586 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6587 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6588 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6590 * config/doc/extend.texi: Add prototypes for vec_permxor.
6592 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6595 * diagnostic-show-locus.c (class layout_point): Convert m_line
6596 from int to linenum_type.
6597 (line_span::comparator): Use linenum "compare" function when
6598 comparing line numbers.
6599 (test_line_span): New function.
6600 (layout_range::contains_point): Convert param "row" from int to
6602 (layout_range::intersects_line_p): Likewise.
6603 (layout::will_show_line_p): Likewise.
6604 (layout::print_source_line): Likewise.
6605 (layout::should_print_annotation_line_p): Likewise.
6606 (layout::print_annotation_line): Likewise.
6607 (layout::print_leading_fixits): Likewise.
6608 (layout::annotation_line_showed_range_p): Likewise.
6609 (struct line_corrections): Likewise for field m_row.
6610 (line_corrections::line_corrections): Likewise for param "row".
6611 (layout::print_trailing_fixits): Likewise.
6612 (layout::get_state_at_point): Likewise.
6613 (layout::get_x_bound_for_row): Likewise.
6614 (layout::print_line): Likewise.
6615 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6616 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6617 * input.c (selftest::test_linenum_comparisons): New function.
6618 (selftest::input_c_tests): Call it.
6619 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6620 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6621 * selftest.h (ASSERT_GT): New macro.
6622 (ASSERT_GT_AT): New macro.
6623 (ASSERT_LT): New macro.
6624 (ASSERT_LT_AT): New macro.
6626 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6628 PR rtl-optimization/84780
6629 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6631 2018-03-14 Martin Liska <mliska@suse.cz>
6633 * tree.c (record_node_allocation_statistics): Use
6634 get_stats_node_kind.
6635 (get_stats_node_kind): New function extracted from
6636 record_node_allocation_statistics.
6637 (free_node): Use get_stats_node_kind.
6639 2018-03-14 Richard Biener <rguenther@suse.de>
6641 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6642 that the value-set of ANTIC_IN doesn't grow.
6645 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6647 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6648 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6650 2018-03-14 Julia Koval <julia.koval@intel.com>
6652 * config.gcc (icelake-client, icelake-server): New.
6654 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6655 (initial_ix86_arch_features): Ditto.
6656 (PTA_SKYLAKE): Add SGX.
6657 (PTA_ICELAKE): Remove.
6658 (PTA_ICELAKE_CLIENT): New.
6659 (PTA_ICELAKE_SERVER): New.
6660 (ix86_option_override_internal): Split up icelake on icelake client and
6662 (get_builtin_code_for_version): Ditto.
6663 (fold_builtin_cpu): Ditto.
6664 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6665 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6666 * config/i386/i386.h (processor_type): Ditto.
6667 * doc/invoke.texi: Ditto.
6669 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6672 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6673 INTEGER_CST offset, add it together with bitpos / 8 and
6674 sign extend based on POINTER_SIZE.
6678 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6681 * config/i386/constraints.md (Yc): New register constraint.
6682 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6683 Use Yc constraint for alternative 2 of operand 0. Remove
6684 preferred_for_speed attribute.
6686 2018-03-14 Richard Biener <rguenther@suse.de>
6688 PR tree-optimization/84830
6689 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6690 with the old one to avoid oscillations.
6692 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6695 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6697 (assign_by_spills): Return a flag of reload assignment failure.
6698 Do not process the reload assignment failures. Do not spill other
6699 reload pseudos if they has the same reg class. Update n if
6701 (lra_assign): Add a return arg. Set up from the result of
6702 assign_by_spills call.
6703 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6704 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6705 usage_insns if it is not NULL.
6706 (spill_hard_reg_in_range): New function.
6707 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6708 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6709 function prototypes.
6710 (lra_assign): Change prototype.
6711 * lra.c (lra): Add code to deal with fails by splitting hard reg
6714 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6716 * config/riscv/riscv.opt (mrelax): New option.
6717 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6718 "norelax" when riscv_mrelax is disabled.
6719 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6721 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6724 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6725 reassociation for int modes.
6727 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6729 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6730 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6732 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6733 * config/aarch64/aarch64-sve.md
6734 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6735 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6736 (*extend<mode><Vwide>2): Rename to...
6737 (aarch64_sve_extend<mode><Vwide>2): ...this.
6738 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6739 renaming the old pattern to...
6740 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6742 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6743 define_expand, renaming the old pattern to...
6744 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6745 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6746 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6747 account when deciding which SVE instruction the optab should use.
6748 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6750 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6752 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6753 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6754 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6755 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6756 (tlsdesc_small_<mode>): Turn a define_expand and use
6757 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6758 (tlsdesc_small_advsimd_<mode>): ...this.
6759 (tlsdesc_small_sve_<mode>): New pattern.
6761 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6763 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6764 (UNSPEC_UMUL_HIGHPART): New constants.
6765 (MUL_HIGHPART): New int iteraor.
6766 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6767 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6769 (*<su>mul<mode>3_highpart): New define_insn.
6771 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6774 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6777 2018-03-13 Martin Liska <mliska@suse.cz>
6780 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6782 (sem_item_optimizer::~sem_item_optimizer): Release it.
6783 (sem_item_optimizer::merge_classes): Register variable aliases.
6784 (sem_item_optimizer::fixup_pt_set): New function.
6785 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6786 * ipa-icf.h: Declare new variables and functions.
6788 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6791 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6792 integer_pow2p@2 and test integer_pow2p in condition.
6793 (A < 0 ? C : 0): Similarly for @1.
6796 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6797 characters starting at p contain '\0' character, don't look beyond
6801 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6802 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6805 * reg-stack.c (change_stack): Change update_end var from int to
6806 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6807 also call set_block_for_insn on the newly added insns and rescan.
6810 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6811 on the last operand.
6814 * tree.c (stabilize_reference_1): Return save_expr (e) for
6815 STATEMENT_LIST even if it doesn't have side-effects.
6817 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6819 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6821 2018-03-12 Renlin Li <renlin.li@arm.com>
6823 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6824 aarch64_output_scalar_simd_mov_immediate.
6826 2018-03-12 Martin Sebor <msebor@redhat.com>
6828 PR tree-optimization/83456
6829 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6830 for perfectly overlapping calls to memcpy.
6831 (gimple_fold_builtin_memory_chk): Same.
6832 (gimple_fold_builtin_strcpy): Handle no-warning.
6833 (gimple_fold_builtin_stxcpy_chk): Same.
6834 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6836 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6838 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6839 parameter. Use it for SFmode.
6840 (rs6000_function_arg_advance_1): Adjust.
6841 (rs6000_function_arg): Adjust.
6842 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6844 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6846 PR rtl-optimization/84169
6847 PR rtl-optimization/84780
6848 * combine.c (can_combine_p): Check for a 2-insn combination whether
6849 the destination register is used between the two insns, too.
6851 2018-03-12 Richard Biener <rguenther@suse.de>
6853 PR tree-optimization/84803
6854 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6855 for refs DR analysis didn't process.
6857 2018-03-12 Richard Biener <rguenther@suse.de>
6859 PR tree-optimization/84777
6860 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6861 force-vectorize loops ignore whether we are optimizing for size.
6863 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6865 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6866 (TARGET_MD_ASM_ADJUST): Define.
6868 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6869 Kito Cheng <kito.cheng@gmail.com>
6870 Chung-Ju Wu <jasonwucj@gmail.com>
6872 * config/nds32/nds32.c (nds32_compute_stack_frame,
6873 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6874 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6875 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6876 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6877 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6878 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6879 * config/nds32/nds32.md (prologue, epilogue): Use macro
6880 NDS32_V3PUSH_AVAILABLE_P to do checking.
6882 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6885 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6886 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6887 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6888 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6889 addition of most attributes on !orig_type_die or the attribute not
6890 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6892 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6893 Chung-Ju Wu <jasonwucj@gmail.com>
6895 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6897 * config/nds32/nds32.opt (mvh): New option.
6899 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6900 Chung-Ju Wu <jasonwucj@gmail.com>
6902 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6904 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6905 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6908 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6909 Chung-Ju Wu <jasonwucj@gmail.com>
6911 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6913 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6914 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6916 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6917 Kito Cheng <kito.cheng@gmail.com>
6918 Chung-Ju Wu <jasonwucj@gmail.com>
6920 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6921 UNSPEC_FFMISM and UNSPEC_FLMISM.
6922 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6923 for ffb, ffmism and flmism.
6924 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6925 (unspec_ffmism): Ditto.
6926 (unspec_flmism): Ditto.
6927 (nds32_expand_builtin_impl): Check if string extension is available.
6928 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6929 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6931 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6934 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6937 * lra-assigns.c (assign_by_spills): Return a flag of reload
6938 assignment failure. Do not process the reload assignment
6939 failures. Do not spill other reload pseudos if they has the same
6941 (lra_assign): Add a return arg. Set up from the result of
6942 assign_by_spills call.
6943 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6944 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6945 usage_insns if it is not NULL.
6946 (spill_hard_reg_in_range): New function.
6947 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6948 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6949 function prototypes.
6950 (lra_assign): Change prototype.
6951 * lra.c (lra): Add code to deal with fails by splitting hard reg
6954 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6957 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6959 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6962 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6963 (dw_val_node): Add val_symbolic_view.
6964 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6965 (symview_upper_bound): New.
6966 (new_line_info_table): Initialize symviews_since_reset.
6967 (dwarf2out_source_line): Count symviews_since_reset and set
6968 symview_upper_bound.
6969 (dw_val_equal_p): Handle symview.
6970 (add_AT_symview): New.
6971 (print_dw_val): Handle symview.
6972 (attr_checksum, attr_checksum_ordered): Likewise.
6973 (same_dw_val_p, size_of_die): Likewise.
6974 (value_format, output_die): Likewise.
6975 (add_high_low_attributes): Use add_AT_symview for entry_view.
6976 (dwarf2out_finish): Reset symview_upper_bound, clear
6979 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6982 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6983 Add strict argument and use it.
6984 (rs6000_split_multireg_move): Update for new strict argument.
6985 (mem_operand_gpr): Disallow all non-offsettable addresses.
6986 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6988 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6991 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6992 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6993 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6996 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6997 decl, use remap_type if we want to use the type.
6999 2018-03-09 Martin Sebor <msebor@redhat.com>
7001 PR tree-optimization/84526
7002 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7004 (builtin_access::generic_overlap): Be prepared to handle non-array
7007 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7009 PR rtl-optimization/84682
7010 * lra-constraints.c (process_address_1): Check is_address flag
7011 for address constraints.
7012 (process_alt_operands): Likewise.
7013 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7014 preprocess_constraints.
7015 * recog.h (preprocess_constraints): Add oploc parameter.
7017 * recog.c (preprocess_constraints): Test address_operand for
7018 CT_ADDRESS constraints.
7020 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7023 * lra-assigns.c (assign_by_spills): Return a flag of reload
7024 assignment failure. Do not process the reload assignment
7025 failures. Do not spill other reload pseudos if they has the same
7027 (lra_assign): Add a return arg. Set up from the result of
7028 assign_by_spills call.
7029 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7030 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7031 usage_insns if it is not NULL.
7032 (spill_hard_reg_in_range): New function.
7033 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7034 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7035 function prototypes.
7036 (lra_assign): Change prototype.
7037 * lra.c (lra): Add code to deal with fails by splitting hard reg
7040 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7043 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7044 Accept complain bool parameter. Only emit errors if it is true.
7045 (arm_parse_cpu_option_name): Likewise.
7046 (arm_target_thumb_only): Adjust callers of the above.
7047 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7048 prototype to take a default true bool parameter.
7049 (arm_parse_arch_option_name): Likewise.
7051 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7052 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7056 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7057 * configure: Regenerate.
7058 * configure.ac ("linker --version-script option"): New.
7059 ("linker soname option"): New.
7061 2018-03-09 Richard Biener <rguenther@suse.de>
7063 PR tree-optimization/84775
7064 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7065 immediate uses of predicate stmts and mark them modified.
7068 PR tree-optimization/84178
7069 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7071 (version_loop_for_if_conversion): Delay update_ssa call.
7072 (tree_if_conversion): Delay update_ssa until after predicate
7075 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7078 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7079 when the function accesses prior frames.
7081 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7084 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7085 gen_llsym, otherwise call maybe_gen_llsym.
7088 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7089 has ',' character inside of it.
7091 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7094 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7095 as clobbering CC_REGNUM.
7097 2018-03-08 Richard Biener <rguenther@suse.de>
7100 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7101 (follow_copies_to_constant): Do not follow SSA names registered
7104 2018-03-08 Richard Biener <rguenther@suse.de>
7106 PR tree-optimization/84178
7107 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7109 (version_loop_for_if_conversion): Delay update_ssa call.
7110 (tree_if_conversion): Delay update_ssa until after predicate
7113 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7115 PR tree-optimization/84178
7116 * tree-if-conv.c (release_bb_predicate): Remove the
7117 the assertion that the stmts have NULL use_ops.
7118 Discard the statements, asserting that they haven't
7119 yet been added to a BB.
7121 2018-03-08 Richard Biener <rguenther@suse.de>
7123 PR tree-optimization/84746
7124 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7125 (phi_translate): Pass in destination ANTIC_OUT set.
7126 (phi_translate_1): Likewise. For a simplified result lookup
7127 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7128 (phi_translate_set): Adjust.
7129 (do_pre_regular_insertion): Likewise.
7130 (do_pre_partial_partial_insertion): Likewise.
7132 2018-03-08 Martin Liska <mliska@suse.cz>
7134 PR gcov-profile/84735
7135 * doc/gcov.texi: Document usage of profile files.
7136 * gcov-io.h: Document changes in the format.
7138 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7142 * dwarf2out.c (struct dw_line_info_table): Update comments for
7144 (FORCE_RESET_NEXT_VIEW): New.
7145 (FORCE_RESETTING_VIEW_P): New.
7146 (RESETTING_VIEW_P): Check for -1 too.
7147 (ZERO_VIEW_P): Likewise.
7148 (new_line_info_table): Force-reset next view.
7149 (dwarf2out_begin_function): Likewise.
7150 (dwarf2out_source_line): Simplify zero_view_p initialization.
7151 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7152 view directly. Omit view when omitting .loc at line 0.
7154 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7156 PR tree-optimization/84740
7157 * tree-switch-conversion.c (process_switch): Call build_constructors
7158 only if info.phi_count is non-zero.
7160 PR tree-optimization/84739
7161 * tree-tailcall.c (find_tail_calls): Check call arguments against
7162 DECL_ARGUMENTS (current_function_decl) rather than
7163 DECL_ARGUMENTS (func) when checking for tail recursion.
7165 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7167 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7168 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7169 Volker Reichelt's entry and add entries for people that perform
7170 GCC fuzzy testing and report numerous bugs.
7172 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7175 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7176 readonly data in sdata, if that is disabled.
7177 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7178 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7179 -mreadonly-in-sdata option.
7181 2018-03-07 Martin Sebor <msebor@redhat.com>
7183 PR tree-optimization/84468
7184 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7185 basic block when looking for nul assignment.
7187 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7190 * except.h (output_function_exception_table): Adjust prototype.
7191 * except.c (output_function_exception_table): Remove FNNAME parameter
7192 and add SECTION parameter. Ouput one part of the table at a time.
7193 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7194 the first part of the exception table and emit unwind directives.
7195 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7196 (i386_pe_seh_cold_init): Likewise.
7197 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7198 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7199 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7200 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7201 * config/i386/winnt.c: Include except.h.
7202 (struct seh_frame_state): Add reg_offset, after_prologue and
7203 in_cold_section fields.
7204 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7205 (i386_pe_seh_cold_init): New function.
7206 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7207 to seh->in_cold_section.
7208 (seh_emit_push): Record the offset of the push.
7209 (seh_emit_save): Record the offet of the save.
7210 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7211 Test seh->after_prologue to disregard the epilogue.
7212 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7213 (i386_pe_end_cold_function): New function.
7215 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7218 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7219 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7222 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7224 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7225 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7228 * multiple_target.c: Include tree-inline.h and intl.h.
7229 (expand_target_clones): Diagnose and fail if node->definition and
7230 !tree_versionable_function_p (node->decl).
7232 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7234 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7236 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7237 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7238 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7240 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7243 * combine.c (try_combine): Use reg_or_subregno instead of handling
7244 just paradoxical SUBREGs and REGs.
7246 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7248 * config/arc/arc.c (arc_finalize_pic): Remove function.
7249 (arc_must_save_register): We use single base PIC register, remove
7250 checks to save/restore the PIC register.
7251 (arc_expand_prologue): Likewise.
7252 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7254 (arc_verify_short): Likewise.
7255 (arc_attr_type): Likewise.
7256 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7257 (walk_stores): Likewise.
7258 (arc_address_cost): Make it static.
7259 (arc_verify_short): Likewise.
7260 (branch_dest): Likewise.
7261 (arc_attr_type): Likewise.
7262 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7263 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7264 (arc_final_prescan_insn): Remove inserting the nops due to
7265 hardware hazards. It is done in reorg step.
7266 (insn_length_variant_t): Remove.
7267 (insn_length_parameters_t): Likewise.
7268 (arc_insn_length_parameters): Likewise.
7269 (arc_get_insn_variants): Likewise.
7270 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7272 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7275 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7278 PR tree-optimization/84687
7279 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7281 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7283 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7285 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7286 Rename to ppc_speculation_barrier.
7287 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7288 __builtin_ppc_speculation_barrier.
7290 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7293 * combine.c (combine_simplify_rtx): Don't try to simplify if
7294 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7297 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7299 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7300 to 32 bytes when compiling for POWER9.
7302 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7305 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7306 regparm >= 3 with no arg reg available also for calls with
7307 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7310 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7312 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7314 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7317 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7319 2018-03-05 Richard Biener <rguenther@suse.de>
7321 PR tree-optimization/84486
7322 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7323 When inserting a __builtin_assume_aligned call set the LHS
7324 SSA name alignment info accordingly.
7326 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7328 PR tree-optimization/84114
7329 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7330 Avoid reassociation of FLOAT_MODE addition.
7332 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7334 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7335 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7336 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7337 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7338 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7339 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7340 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7342 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7343 __builtin_ia32_wbinvd): New builtins.
7344 (SPECIAL_ARGS2): New.
7345 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7346 (SPECIAL_ARGS2): New.
7347 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7348 (ix86_valid_target_attribute_inner_p): Ditto.
7349 (ix86_init_mmx_sse_builtins): Add special_args2.
7350 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7351 TARGET_WBNOINVD_P): New.
7352 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7353 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7354 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7355 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7356 * config/i386/pconfigintrin.h: New file.
7357 * config/i386/wbnoinvdintrin.h: Ditto.
7358 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7360 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7362 2018-03-05 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/84670
7365 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7367 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7368 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7369 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7370 doesn't grow if all successors (recursively) were visited at least
7373 2018-03-05 Richard Biener <rguenther@suse.de>
7375 PR tree-optimization/84650
7376 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7377 if executed in the loop pipeline.
7379 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7381 * doc/configfiles.texi (Configuration Files): Move info about
7382 conditionalizing $target-protos.h to...
7383 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7384 differs from $target-protos.h.
7386 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7387 Chung-Ju Wu <jasonwucj@gmail.com>
7389 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7390 * config/nds32/nds32-multiple.md (setmemsi): Define.
7391 * config/nds32/nds32-memory-manipulation.c
7392 (nds32_gen_dup_4_byte_to_word_value): New.
7393 (emit_setmem_word_loop): New.
7394 (emit_setmem_byte_loop): New.
7395 (nds32_expand_setmem_loop): New.
7396 (nds32_expand_setmem_loop_v3m): New.
7397 (nds32_expand_setmem_unroll): New.
7398 (nds32_expand_setmem): New.
7400 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7401 Chung-Ju Wu <jasonwucj@gmail.com>
7403 * config/nds32/nds32-memory-manipulation.c
7404 (nds32_emit_load_store): New.
7405 (nds32_emit_post_inc_load_store): New.
7406 (nds32_emit_mem_move): New.
7407 (nds32_emit_mem_move_block): New.
7408 (nds32_expand_movmemsi_loop_unknown_size): New.
7409 (nds32_expand_movmemsi_loop_known_size): New.
7410 (nds32_expand_movmemsi_loop): New.
7411 (nds32_expand_movmemsi_unroll): New.
7412 (nds32_expand_movmemqi): Rename ...
7413 (nds32_expand_movmemsi): ... to this.
7414 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7415 (movmemsi): ... to this.
7416 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7417 (nds32_expand_movmemsi): ... to this.
7419 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7420 Monk Chiang <sh.chiang04@gmail.com>
7421 Chung-Ju Wu <jasonwucj@gmail.com>
7423 * config/nds32/nds32-protos.h
7424 (nds32_expand_load_multiple): New arguments.
7425 (nds32_expand_store_multiple): Ditto.
7426 (nds32_valid_multiple_load_store): Rename ...
7427 (nds32_valid_multiple_load_store_p): ... to this.
7428 * config/nds32/nds32-memory-manipulation.c
7429 (nds32_expand_load_multiple): Refine implementation.
7430 (nds32_expand_store_multiple): Ditto.
7431 * config/nds32/nds32-multiple.md
7432 (load_multiple): Update nds32_expand_load_multiple interface.
7433 (store_multiple): Update nds32_expand_store_multiple interface.
7434 * config/nds32/nds32-predicates.c
7435 (nds32_valid_multiple_load_store): Rename ...
7436 (nds32_valid_multiple_load_store_p): ... to this and refine
7438 * config/nds32/predicates.md
7439 (nds32_load_multiple_and_update_address_operation): New predicate.
7440 (nds32_store_multiple_and_update_address_operation): New predicate.
7442 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7443 Chung-Ju Wu <jasonwucj@gmail.com>
7445 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7446 (combo): New attribute.
7447 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7449 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7451 * config/nds32/nds32.opt: Change -mcmodel= default value.
7453 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7454 Monk Chiang <sh.chiang04@gmail.com>
7455 Chung-Ju Wu <jasonwucj@gmail.com>
7457 * config/nds32/constants.md (unspec_element): New enum.
7458 * config/nds32/constraints.md (Umw): New constraint.
7459 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7460 * config/nds32/nds32-intrinsic.md: Likewise.
7461 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7462 (nds32_valid_smw_lwm_base_p): New.
7463 (nds32_output_smw_single_word): New.
7464 (nds32_output_lmw_single_word): New.
7465 (nds32_expand_unaligned_load): New.
7466 (nds32_expand_unaligned_store): New.
7467 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7468 (nds32_output_smw_single_word): Declare.
7469 (nds32_output_lmw_single_word): Declare.
7470 (nds32_expand_unaligned_load): Declare.
7471 (nds32_expand_unaligned_store): Declare.
7472 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7473 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7474 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7475 NDS32_BUILTIN_UASTORE_DW.
7476 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7479 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7480 Kito Cheng <kito.cheng@gmail.com>
7481 Chung-Ju Wu <jasonwucj@gmail.com>
7483 * config/nds32/nds32-intrinsic.c
7484 (nds32_expand_builtin_null_ftype_reg): Delete.
7485 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7486 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7487 (nds32_read_argument): New.
7488 (nds32_legitimize_target): Ditto.
7489 (nds32_legitimize_argument): Ditto.
7490 (nds32_check_constant_argument): Ditto.
7491 (nds32_expand_unop_builtin): Ditto.
7492 (nds32_expand_unopimm_builtin): Ditto.
7493 (nds32_expand_binop_builtin): Ditto.
7494 (nds32_builtin_decl_impl): Ditto.
7495 (builtin_description): Ditto.
7496 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7497 (nds32_init_builtins_impl): Ditto.
7498 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7499 (nds32_builtin_decl): New.
7500 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7501 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7503 2018-03-02 Jeff Law <law@redhat.com>
7505 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7506 (redundant_insn, fill_simple_delay_slots): Likewise.
7507 (fill_slots_from_thread): Likewise.
7508 * resource.c (mark_referenced_resources): Likewise.
7509 (mark_set_resources, find_dead_or_set_registers): Likewise.
7511 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7513 * substring-locations.h (format_warning_va): Formatting fix for
7515 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7517 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7518 * substring-locations.c: Include intl.h.
7519 (format_warning_va): Turned into small wrapper around
7520 format_warning_n_va, renamed to ...
7521 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7522 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7524 (format_warning_at_substring_n): New function.
7525 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7526 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7527 format_warning_at_substring with just a shorter name instead of
7528 const function pointer.
7529 (fmtwarn_n): New function.
7530 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7531 appropriate, get rid of all the fmtstr temporaries, move conditionals
7532 with G_() wrapped string literals directly into fmtwarn arguments,
7533 cast dir.len to (int), formatting fixes.
7535 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7537 * doc/invoke.texi: Remove "Cilk Plus" references.
7539 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7540 Richard Biener <rguenther@suse.de>
7543 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7544 for error or warning attributes if CALL_FROM_THUNK_P is set.
7547 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7550 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7551 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7554 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7555 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7557 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7558 instead of -1U in last predictors element's probability member.
7560 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7563 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7564 arguments if they are comparable.
7566 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7568 PR tree-optimization/84634
7569 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7570 masks and masked_loop_p with a single loop_masks, making sure it's
7571 null for bb vectorization.
7573 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7575 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7576 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7577 loop->force_vectorize to check whether there is no alias.
7579 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7582 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7584 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7586 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7587 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7588 instead of a loop around prev_real_insn.
7589 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7593 * config/i386/i386.c (ix86_print_operand): Use conditional
7594 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7597 2018-03-02 Richard Biener <rguenther@suse.de>
7599 PR tree-optimization/84427
7600 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7601 (bitmap_set_subtract_values): Rewrite to handle multiple
7604 (prune_clobbered_mems): Likewise.
7605 (phi_translate): Take edge instead of pred/phiblock.
7606 (phi_translate_1): Likewise.
7607 (phi_translate_set): Likewise. Insert all translated
7608 exprs for a value into the set, keeping possibly multiple
7609 expressions per value.
7610 (compute_antic_aux): Adjust for phi_translate changes.
7611 When intersecting union the expressions and prune those
7612 not in the final value set, keeping possibly multiple
7613 expressions per value. Do not use value-insertion
7614 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7615 all expressions. Add verification that the value-sets
7616 only shrink during iteration.
7617 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7618 (do_pre_regular_insertion): Likewise.
7619 (do_pre_partial_partial_insertion): Likewise.
7621 2018-03-02 Richard Biener <rguenther@suse.de>
7624 * config/darwin.c (saved_debug_info_level): New static global.
7625 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7626 (darwin_asm_lto_end): Restore debug info generation settings.
7628 2018-03-01 Martin Liska <mliska@suse.cz>
7631 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7634 2018-03-01 Richard Biener <rguenther@suse.de>
7637 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7638 type annotation with LTO when debug was disabled at compile-time.
7640 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7642 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7644 (mips_final_postscan_insn): Likewise.
7646 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7648 PR rtl-optimization/84528
7649 * alias.c (init_alias_target): Add commentary.
7650 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7651 a unique base value if the frame pointer is not eliminated
7652 to the stack pointer.
7654 2018-03-01 Tom de Vries <tom@codesourcery.com>
7656 PR rtl-optimization/83327
7657 * lra-int.h (hard_regs_spilled_into): Declare.
7658 * lra.c (hard_regs_spilled_into): Define.
7659 (init_reg_info): Init hard_regs_spilled_into.
7660 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7661 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7662 (process_bb_lives): Handle hard_regs_spilled_into.
7663 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7664 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7666 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7668 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7669 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7670 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7671 * config/rs6000/aix72.h: New file.
7673 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7675 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7676 instead of warning_at with conditional singular and plural messages
7680 * stor-layout.c (update_alignment_for_field): For
7681 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7682 && !DECL_PACKED (field), do the alignment update, just use
7683 only desired_align instead of MAX (type_align, desired_align)
7685 (place_field): Don't do known_align < desired_align handling
7686 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7687 is non-NULL, instead do it after rli->prev_field handling and
7688 only if not within a bitfield word. For DECL_PACKED (field)
7689 use type_align of BITS_PER_UNIT.
7691 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7693 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7694 superfluous parentheses and trailing spaces.
7696 2018-02-28 Richard Biener <rguenther@suse.de>
7698 PR tree-optimization/84584
7699 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7700 SCoPs with fake exit edge.
7702 2018-02-28 Martin Liska <mliska@suse.cz>
7705 * timevar.c (timer::print): Fix format to properly print 100%
7708 2018-02-28 Richard Biener <rguenther@suse.de>
7711 * genmatch.c (capture_info::walk_match): Do not mark
7712 captured expressions without operands as expr_p given
7713 they act more like predicates and should be subject to
7714 "lost tail" side-effect preserving.
7716 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7718 PR rtl-optimization/81611
7719 * auto-inc-dec.c (attempt_change): Move dead note from
7720 mem_insn if it's the next use of regno
7721 (find_address): Take address use of reg holding
7722 non-incremented value. Add parm to limit search to the named
7724 (merge_in_block): Attempt to use a mem insn that is the next
7725 use of the original regno.
7727 2018-02-27 Martin Sebor <msebor@redhat.com>
7730 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7731 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7733 2018-02-27 Martin Sebor <msebor@redhat.com>
7735 PR translation/84207
7736 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7737 n argument to unsigned HOST_WIDE_INT.
7738 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7739 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7740 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7741 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7743 2018-02-27 Richard Biener <rguenther@suse.de>
7745 PR tree-optimization/84512
7746 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7747 Do not use the estimate returned from record_stmt_cost for
7748 the scalar iteration cost but sum properly using add_stmt_cost.
7750 2018-02-27 Richard Biener <rguenther@suse.de>
7752 PR tree-optimization/84466
7753 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7754 Adjust last change to less strictly validate use operands.
7756 2018-02-27 Martin Liska <mliska@suse.cz>
7758 PR gcov-profile/84548
7759 * gcov.c (process_file): Allow partial overlap and consider it
7760 also as group functions.
7761 (output_lines): Properly calculate range of lines for a group.
7763 2018-02-27 Martin Liska <mliska@suse.cz>
7765 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7766 'ggc' suffixes. Change first column width.
7767 (timer::print): Fix formatting of the column.
7769 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7771 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7772 preserve inline entry blocks for the sake of debug inline
7773 entry point markers alone.
7774 (remove_unused_locals): Suggest in comments a better place to
7775 force the preservation of inline entry blocks that are
7776 otherwise unused, but do not preserve them.
7778 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7780 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7782 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7785 * config/i386/constraints.md (Bs): Replace
7786 ix86_indirect_branch_register with
7787 TARGET_INDIRECT_BRANCH_REGISTER.
7789 * config/i386/i386.md (indirect_jump): Likewise.
7790 (tablejump): Likewise.
7791 (*sibcall_memory): Likewise.
7792 (*sibcall_value_memory): Likewise.
7793 Peepholes of indirect call and jump via memory: Likewise.
7794 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7795 (*sibcall_value_GOT_32): Likewise.
7796 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7797 (GOT_memory_operand): Likewise.
7798 (call_insn_operand): Likewise.
7799 (sibcall_insn_operand): Likewise.
7800 (GOT32_symbol_operand): Likewise.
7801 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7803 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7805 PR rtl-optimization/83496
7806 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7807 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7808 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7809 redundant insn, if any.
7810 (relax_delay_slots): Likewise.
7811 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7813 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7815 PR tree-optimization/83965
7816 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7817 that grouped statements are part of a reduction chain. Return
7818 true if the statement is not marked as a reduction itself but
7820 (vect_recog_dot_prod_pattern): Don't check whether the statement
7821 is part of a group here.
7822 (vect_recog_sad_pattern): Likewise.
7823 (vect_recog_widen_sum_pattern): Likewise.
7825 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7828 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7830 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7833 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7835 (ix86_output_indirect_function_return): New prototype.
7836 (ix86_split_simple_return_pop_internal): Likewise.
7837 * config/i386/i386.c (indirect_return_via_cx): New.
7838 (indirect_return_via_cx_bnd): Likewise.
7839 (indirect_thunk_name): Handle return va CX_REG.
7840 (output_indirect_thunk_function): Create alias for
7841 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7842 (ix86_output_indirect_jmp): Remove the bool argument.
7843 (ix86_output_indirect_function_return): New function.
7844 (ix86_split_simple_return_pop_internal): Likewise.
7845 * config/i386/i386.md (*indirect_jump): Don't pass false
7846 to ix86_output_indirect_jmp.
7847 (*tablejump_1): Likewise.
7848 (simple_return_pop_internal): Change it to define_insn_and_split.
7849 Call ix86_split_simple_return_pop_internal to split it for
7851 (simple_return_indirect_internal): Call
7852 ix86_output_indirect_function_return instead of
7853 ix86_output_indirect_jmp.
7855 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7858 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7859 memset and value initialization afterwards.
7861 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7863 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7865 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7868 * common/config/aarch64/aarch64-common.c
7869 (aarch_option_optimization_table[]): Switch
7870 off fomit-frame-pointer
7872 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7873 Chung-Ju Wu <jasonwucj@gmail.com>
7875 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7877 (store_multiple): Ditto.
7879 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7881 * config.gcc: Add --with-cpu support for nds32 target.
7882 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7883 * config/nds32/nds32.opt: Add -mcpu= option.
7885 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7887 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7888 isel=yes): Warn for these deprecated options.
7890 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7892 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7893 ISA_2_5_MASKS_EMBEDDED.
7895 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7897 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7898 p->max as pointers rather than using iterative_hash_expr.
7900 2018-02-23 Carl Love <cel@us.ibm.com>
7902 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7903 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7905 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7906 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7907 P8V_BUILTIN_VEC_VUNSIGNED2.
7909 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7912 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7913 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7915 * lra-constraints.c (curr_insn_transform): Set up
7916 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7918 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7919 * lra-eliminations.c (spill_pseudos): Ditto.
7920 (process_insn_for_elimination): Ditto.
7921 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7922 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7923 LRA_NON_CLOBBERED_ALT.
7925 2018-02-22 Martin Sebor <msebor@redhat.com>
7927 PR tree-optimization/84480
7928 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7929 to maybe_diag_stxncpy_trunc. Call it.
7930 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7931 from gimple_fold_builtin_strcpy. Print inlining stack.
7932 (handle_builtin_stxncpy): Print inlining stack.
7933 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7935 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7938 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7939 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7940 and -fcheck-pointer-bounds are used together.
7941 (indirect_thunk_prefix): New enum.
7942 (indirect_thunk_need_prefix): New function.
7943 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7944 "_nt" instead of "_bnd" for NOTRACK prefix.
7945 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7946 (output_indirect_thunk_function): Likewise.
7948 (ix86_code_end): Update output_indirect_thunk_function calls.
7949 (ix86_output_indirect_branch_via_reg): Replace
7950 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7951 (ix86_output_indirect_branch_via_push): Likewise.
7952 (ix86_output_function_return): Likewise.
7953 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7954 incompatible with -fcf-protection=branch and
7955 -fcheck-pointer-bounds.
7957 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7960 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7961 Change gcc_assert call to output_operand_lossage.
7963 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7965 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7967 2018-02-22 DJ Delorie <dj@redhat.com>
7968 Sebastian Perta <sebastian.perta@renesas.com>
7969 Oleg Endo <olegendo@gcc.gnu.org>
7971 * config/rx/rx.c (rx_rtx_costs): New function.
7972 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7974 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7976 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7978 2018-02-22 Martin Liska <mliska@suse.cz>
7981 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7982 Add "native" as a possible value.
7984 2018-02-22 Martin Liska <mliska@suse.cz>
7987 * config/i386/i386.c (ix86_option_override_internal):
7988 Add "native" as a possible value for -march and -mtune.
7990 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7993 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7994 to all type variants.
7996 PR tree-optimization/84503
7997 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7998 width as info->bitpos + info->bitsize - start.
7999 (merged_store_group::merge_overlapping): Simplify width computation.
8000 (check_no_overlap): New function.
8001 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8002 start + width and last_order of the group, fail if check_no_overlap
8004 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8005 to group if check_no_overlap fails.
8007 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8009 * config/rs6000/altivec.md: Delete contraint arguments to
8010 define_expand, define_split, and define_peephole2, and in
8011 define_insn_and_split if always unused.
8012 * config/rs6000/darwin.md: Ditto.
8013 * config/rs6000/dfp.md: Ditto.
8014 * config/rs6000/rs6000.md: Ditto.
8015 * config/rs6000/sync.md: Ditto.
8016 * config/rs6000/vector.md: Ditto.
8017 * config/rs6000/vsx.md: Ditto.
8019 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8021 * config/rs6000/altivec.md: Write output control strings as braced
8022 blocks instead of double-quoted strings.
8023 * config/rs6000/darwin.md: Ditto.
8024 * config/rs6000/rs6000.md: Ditto.
8025 * config/rs6000/vector.md: Ditto.
8026 * config/rs6000/vsx.md: Ditto.
8028 2018-02-21 Jason Merrill <jason@redhat.com>
8030 PR c++/84314 - ICE with templates and fastcall attribute.
8031 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8033 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8035 * ipa-cp.c (determine_versionability): Fix comment typos.
8037 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8040 * ipa-cp.c (determine_versionability): Do not version functions caling
8043 2018-02-21 Martin Liska <mliska@suse.cz>
8046 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8047 Add "native" as a possible value.
8048 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8049 the macro when native cpu detection is available.
8051 2018-02-21 Martin Liska <mliska@suse.cz>
8054 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8055 Add "native" as a possible value.
8056 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8057 when native cpu detection is available.
8059 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8060 Martin Sebor <msebor@redhat.com>
8062 PR tree-optimization/84478
8063 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8065 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8066 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8067 support which is conservatively correct, for 2 only stay conservative
8068 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8069 argument to the 2 argument get_range_strlen, adjust 6 arg
8070 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8072 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8073 (gimple_fold_builtin_strlen): Pass true as last argument to
8076 2018-02-20 Martin Sebor <msebor@redhat.com>
8079 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8080 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8081 (builtin_memref::builtin_memref): Factor out parts into
8082 set_base_and_offset and call it.
8084 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8087 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8088 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8089 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8090 search at the associated MODE_INT.
8092 2018-02-20 Jeff Law <law@redhat.com>
8095 PR tree-optimization/81592
8097 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8098 for range data rather than using global data.
8099 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8100 range data rather than using global data.
8101 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8102 pass it to children as needed.
8103 (struct directive::fmtresult): Similarly.
8104 (struct directive::set_width): Similarly.
8105 (struct directive::set_precision): Similarly.
8106 (format_integer, format_directive, parse_directive): Similarly.
8107 (format_none): Accept unnamed vr_values parameter.
8108 (format_percent, format_floating, format_character): Similarly.
8109 (format_string, format_plain): Similarly.
8110 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8111 the EVRP range analyzer for range data rather than using global data.
8112 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8113 gimple-ssa-evrp-analyze.h
8114 (class sprintf_dom_walker): Add after_dom_children member function.
8115 Add evrp_range_analyzer member.
8116 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8117 range analyzer as needed.
8118 (sprintf_dom_walker::after_dom_children): New member function.
8119 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8121 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8122 (evrp_range_analyzer::pop_to_marker): Likewise.
8124 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8126 PR tree-optimization/84419
8127 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8128 with the required type if its current type is compatible but
8131 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8134 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8135 after vectorization.
8137 2018-02-20 Martin Liska <mliska@suse.cz>
8140 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8141 possible values if we don't have a hint.
8143 2018-02-20 Martin Liska <mliska@suse.cz>
8147 * final.c (shorten_branches): Build align_tab array with one
8149 * opts.c (finish_options): Add alignment option limit check.
8150 (MAX_CODE_ALIGN): Likewise.
8151 (MAX_CODE_ALIGN_VALUE): Likewise.
8152 * doc/invoke.texi: Document maximum allowed option value for
8153 all -falign-* options.
8155 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8158 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8159 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8160 * var-tracking.c (emit_note_insn_var_location): Remove all references
8161 to NOTE_INSN_CALL_ARG_LOCATION.
8162 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8163 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8164 Use copy_rtx_if_shared.
8165 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8166 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8167 (dwarf2out_var_location): Remove handling of
8168 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8170 * final.c (final_scan_insn): Remove all references to
8171 NOTE_INSN_CALL_ARG_LOCATION.
8172 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8173 before dumping final insns.
8174 * except.c (emit_note_eh_region_end): Remove all references to
8175 NOTE_INSN_CALL_ARG_LOCATION.
8176 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8177 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8178 * config/arc/arc.c (hwloop_optimize): Likewise.
8179 * config/arm/arm.c (create_fix_barrier): Likewise.
8180 * config/s390/s390.c (s390_chunkify_start): Likewise.
8181 * config/sh/sh.c (find_barrier): Likewise.
8182 * config/i386/i386.c (rest_of_insert_endbranch,
8183 ix86_seh_fixup_eh_fallthru): Likewise.
8184 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8185 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8186 * config/frv/frv.c (frv_function_prologue): Likewise.
8187 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8189 (note_outside_basic_block_p): Remove all references to
8190 NOTE_INSN_CALL_ARG_LOCATION.
8191 * gengtype.c (adjust_field_rtx_def): Likewise.
8192 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8194 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8195 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8198 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8201 PR tree-optimization/84452
8202 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8203 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8206 2018-02-19 Martin Liska <mliska@suse.cz>
8209 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8211 2018-02-19 Martin Liska <mliska@suse.cz>
8212 Richard Sandiford <richard.sandiford@linaro.org>
8214 PR tree-optimization/82491
8215 * gimple-fold.c (get_base_constructor): Make earlier bail out
8218 2018-02-19 Carl Love <cel@us.ibm.com>
8220 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8221 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8223 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8224 P8V_BUILTIN_VEC_NEG.
8226 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8228 * config/rl78/rl78.md (movdf): New define expand.
8230 2018-02-19 Martin Liska <mliska@suse.cz>
8233 * doc/invoke.texi: Fix typo.
8234 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8236 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8238 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8239 handle rs6000_single_float and rs6000_double_float specially for
8242 2018-02-16 Jeff Law <law@redhat.com>
8244 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8247 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8250 * ipa-inline.c (inline_small_functions): Fix a typo.
8252 2018-02-16 Nathan Sidwell <nathan@acm.org>
8254 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8256 2018-02-16 Carl Love <cel@us.ibm.com>
8258 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8259 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8260 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8261 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8262 expansion to P8V_BUILTIN_VEC_FLOAT2.
8264 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8266 PR rtl-optimization/70023
8267 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8268 src_regno into account.
8270 2018-02-16 Carl Love <cel@us.ibm.com>
8272 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8273 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8274 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8275 * config/rs6000/rs6000.c: Remove case statements for
8276 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8277 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8278 and P9V_BUILTIN_VEC_VINSERT4B.
8279 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8280 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8281 * config/rs6000/vsx.md:
8282 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8285 2018-02-16 Carl Love <cel@us.ibm.com>
8287 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8289 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8291 * config/rs6000/rs6000-c.c: Add the definitions for
8292 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8293 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8294 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8295 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8296 definition for insert4b and define insn *insert3b_internal.
8297 * doc/extend.texi: Add documentation for vec_extract4b.
8299 2018-02-16 Nathan Sidwell <nathan@acm.org>
8301 * doc/extend.texi (Backwards Compatibility): Mention friend
8302 injection. Note for-scope is deprecated.
8303 * doc/invoke.texi (-ffriend-injection): Deprecate.
8305 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8307 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8308 that moved to I2, also allow destinations that are a paradoxical
8309 subreg (instead of a normal reg).
8311 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8314 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8317 2018-02-16 Richard Biener <rguenther@suse.de>
8319 PR tree-optimization/84037
8320 PR tree-optimization/84016
8322 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8323 Adjust vec_construct for the fact we need additional higher latency
8324 128bit inserts for AVX256 and AVX512 vector builds.
8325 (ix86_add_stmt_cost): Scale vector construction cost for
8328 2018-02-16 Richard Biener <rguenther@suse.de>
8330 PR tree-optimization/84417
8331 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8332 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8333 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8335 2018-02-16 Martin Liska <mliska@suse.cz>
8338 * internal-fn.def (ASAN_CHECK): Set proper flags.
8339 (ASAN_MARK): Likewise.
8341 2018-02-16 Julia Koval <julia.koval@intel.com>
8343 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8344 from PTA_CANNONLAKE.
8346 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8349 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8350 Use ++iter rather than iter++ for std::list iterators.
8351 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8352 defer deleting them until all nodes in the forest are processed. Do
8353 free even leaf nodes. Change to_process into auto_vec.
8356 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8357 * vec.h (vec_default_construct): Use memset instead of placement new
8358 if BROKEN_VALUE_INITIALIZATION is defined.
8359 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8360 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8363 PR rtl-optimization/83723
8364 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8365 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8366 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8368 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8370 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8372 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8374 PR rtl-optimization/81443
8375 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8376 from inner REGs to paradoxical SUBREGs.
8378 2018-02-16 Richard Biener <rguenther@suse.de>
8380 PR tree-optimization/84399
8381 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8382 For operands we can analyze at their definition make sure we can
8383 analyze them at each use as well.
8385 2018-02-16 Richard Biener <rguenther@suse.de>
8387 PR tree-optimization/84190
8388 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8389 volatile accesses if the decl isn't volatile.
8391 2018-02-15 Jason Merrill <jason@redhat.com>
8393 PR c++/84314 - ICE with templates and fastcall attribute.
8394 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8395 TYPE_CANONICAL on an existing type.
8397 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8399 PR tree-optimization/84383
8400 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8401 dstoff nor call operand_equal_p if dstbase is NULL.
8403 PR tree-optimization/84334
8404 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8405 also a CONSTANT_CLASS_P, punt.
8407 2018-02-14 Jim Wilson <jimw@sifive.com>
8409 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8410 first SMALL_OPERAND check. New local min_second_step. Move assert
8411 to where locals are set. Add TARGET_RVC support.
8412 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8414 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8416 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8418 2018-02-14 Martin Sebor <msebor@redhat.com>
8420 PR tree-optimization/83698
8421 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8422 arrays constrain the offset range to their bounds.
8423 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8424 (builtin_access::overlap): Avoid setting the size of overlap if it's
8426 (maybe_diag_overlap): Also consider arrays when deciding what values
8427 of offsets to include in diagnostics.
8429 2018-02-14 Martin Sebor <msebor@redhat.com>
8432 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8433 that correspond to the kind of a declaration.
8435 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8438 * config/pa/pa.md: Load address of PIC label using the linkage table
8439 if the label is nonlocal.
8441 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8443 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8444 warning message if user requests -maltivec=be.
8445 * doc/invoke.texi: Document deprecation of -maltivec=be.
8447 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8450 * config/rs6000/rs6000-c.c: Update definitions for
8451 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8452 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8454 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8457 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8458 add _get_ssp intrinsics. Remove argument from
8459 __builtin_ia32_rdssp[d|q].
8460 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8461 * config/i386/i386-builtin.def: Remove argument from
8462 __builtin_ia32_rdssp[d|q].
8463 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8464 ix86_expand_special_args_builtin for _rdssp[d|q].
8465 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8466 Clear register before usage.
8467 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8468 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8470 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8472 PR tree-optimization/84357
8473 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8474 operand 1 of an ARRAY_REF too.
8476 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8479 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8480 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8482 (set_of_reg): New struct.
8483 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8484 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8485 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8487 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8488 Split into bitclr, bitset, bitinvert patterns if appropriate.
8489 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8490 use rx_fuse_in_memory_bitop.
8491 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8492 to named insn, correct maximum insn length.
8494 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8497 * machmode.def: Define a complex mode for PARTIAL_INT.
8498 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8500 * doc/rtl.texi: Document CSPImode.
8501 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8503 (msp430_hard_regno_nregs_with_padding): Likewise.
8505 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8508 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8510 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8512 PR rtl-optimization/84169
8513 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8514 we generated a parallel as new i3 and we split that to new i2 and i3
8515 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8516 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8517 those to i2, not i1. Partially rewrite this scan code.
8519 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8522 * stor-layout.c (place_field): For variable length fields, adjust
8523 offset_align afterwards not just based on the field's alignment,
8524 but also on the size.
8527 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8528 of exps and logs in the use_exp2 case.
8530 2018-02-13 Jeff Law <law@redhat.com>
8532 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8535 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8538 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8542 * common.opt (gas-loc-support, gas-locview-support): New.
8543 (ginline-points, ginternal-reset-location-views): New.
8544 * doc/invoke.texi: Document them. Use @itemx where intended.
8545 (gvariable-location-views): Adjust.
8546 * target.def (reset_location_view): New.
8547 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8548 (TARGET_RESET_LOCATION_VIEW): New.
8549 * doc/tm.texi: Rebuilt.
8550 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8551 (dwarf2out_default_as_locview_support): New.
8552 (output_asm_line_debug_info): Use option variables.
8553 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8554 (output_loc_list): Likewise.
8555 (add_high_low_attributes): Check option variables.
8556 Don't output entry view attribute in strict mode.
8557 (gen_inlined_subroutine_die): Check option variables.
8558 (dwarf2out_inline_entry): Likewise.
8559 (init_sections_and_labels): Likewise.
8560 (dwarf2out_early_finish): Likewise.
8561 (maybe_reset_location_view): New, from...
8562 (dwarf2out_var_location): ... here. Call it.
8563 * debug.h (dwarf2out_default_as_loc_support): Declare.
8564 (dwarf2out_default_as_locview_support): Declare.
8565 * hooks.c (hook_int_rtx_insn_0): New.
8566 * hooks.h (hook_int_rtx_insn_0): Declare.
8567 * toplev.c (process_options): Take -gas-loc-support and
8568 -gas-locview-support from dwarf2out. Enable
8569 -gvariable-location-views by default only with locview
8570 assembler support. Enable -ginternal-reset-location-views by
8571 default only if the target defines the corresponding hook.
8572 Enable -ginline-points by default if location views are
8573 enabled; force it disabled if statement frontiers are
8575 * tree-inline.c (expand_call_inline): Check option variables.
8576 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8578 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8580 PR tree-optimization/84321
8581 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8582 handling. Also check whether the anti-range contains any values
8583 that satisfy the mask; switch to a VR_RANGE if not.
8585 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8588 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8590 2018-02-13 Martin Jambor <mjambor@suse.cz>
8593 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8594 of call statements, also set location of a load to a temporary.
8596 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8598 * config/rl78/rl78.c (add_vector_labels): New function.
8599 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8600 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8601 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8602 which checks that no arguments are passed.
8603 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8604 * doc/extend.texi: Documentation for the new attribute.
8606 2018-02-13 Andreas Schwab <schwab@suse.de>
8608 * config/riscv/linux.h (CPP_SPEC): Define.
8610 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8613 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8614 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8615 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8616 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8617 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8618 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8619 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8620 temporarily for AES and PCLMUL builtins.
8622 PR tree-optimization/84339
8623 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8624 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8628 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8629 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8630 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8632 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8634 * omp-simd-clone.h: New file.
8635 * omp-simd-clone.c: Include omp-simd-clone.h.
8636 (expand_simd_clones): No longer static.
8637 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8638 cgraph.h and omp-simd-clone.h.
8639 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8640 (vect_recog_widen_shift_pattern): Formatting fix.
8641 (vect_pattern_recog_1): Don't check optab for calls.
8644 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8645 operands[2] into a REG before using gen_lowpart on it.
8647 2018-02-12 Jeff Law <law@redhat.com>
8650 * config/sh/sh.c (find_barrier): Consider a sibling call
8653 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8654 successfully back substituting a reg.
8656 2018-02-12 Richard Biener <rguenther@suse.de>
8658 PR tree-optimization/84037
8659 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8660 parameter, move visited init to caller.
8661 (vect_slp_analyze_operations): Separate cost from validity
8662 check, initialize visited once for all instances.
8663 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8665 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8667 (vect_model_promotion_demotion_cost): Likewise.
8668 (vectorizable_bswap): Guard cost modeling with !slp_node
8669 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8671 (vectorizable_call): Likewise.
8672 (vectorizable_conversion): Likewise.
8673 (vectorizable_assignment): Likewise.
8674 (vectorizable_shift): Likewise.
8675 (vectorizable_operation): Likewise.
8676 (vectorizable_store): Likewise.
8677 (vectorizable_load): Likewise.
8678 (vectorizable_condition): Likewise.
8679 (vectorizable_comparison): Likewise.
8681 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8684 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8685 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8686 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8688 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8691 * params.def (inline-min-speedup): Increase from 8 to 15.
8692 (max-inline-insns-auto): Decrease from 40 to 30.
8693 * ipa-split.c (consider_split): Add some buffer for function to
8694 be considered inlining candidate.
8695 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8698 2018-02-12 Richard Biener <rguenther@suse.de>
8700 PR tree-optimization/84037
8701 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8702 matched stmts if we cannot swap the non-matched ones.
8704 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8706 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8707 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8708 _mm_maskz_scalef_round_ss): New intrinsics.
8709 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8710 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8711 __builtin_ia32_scalefss_round): Remove.
8712 (__builtin_ia32_scalefsd_mask_round,
8713 __builtin_ia32_scalefss_mask_round): New intrinsics.
8714 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8715 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8716 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8717 "<round_constraint>")): Changed to ...
8718 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8719 "<round_scalar_constraint>")): ... this.
8720 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8721 %0, %1, %2<round_op3>}"): Changed to ...
8722 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8723 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8724 %2<round_scalar_mask_op3>}"): ... this.
8725 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8727 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8729 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8730 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8731 (_mm_maskz_sqrt_round_ss): New intrinsics.
8732 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8733 (__builtin_ia32_sqrtsd_mask_round)
8734 (__builtin_ia32_sqrtss_mask_round): New builtins.
8735 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8736 (__builtin_ia32_sqrtss_round): Remove.
8737 (__builtin_ia32_sqrtsd_mask_round)
8738 (__builtin_ia32_sqrtss_mask_round): New builtins.
8739 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8740 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8741 ((match_operand:VF_128 1 "vector_operand"
8742 "xBm,<round_constraint>")): Changed to ...
8743 ((match_operand:VF_128 1 "vector_operand"
8744 "xBm,<round_scalar_constraint>")): ... this.
8745 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8746 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8747 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8748 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8749 %<iptr>1<round_scalar_mask_op3>}): ... this.
8750 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8751 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8753 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8756 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8757 Cast vec_cmpeq result to correct type.
8758 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8759 Cast vec_cmpgt result to correct type.
8761 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8763 * final.c (final_scan_insn_1): Renamed from...
8764 (final_scan_insn): ... this. New wrapper, to recover
8765 seen from the outermost call in recursive ones.
8766 * config/sparc/sparc.c (output_return): Drop seen from call.
8767 (output_sibcall): Likewise.
8768 * config/visium/visium.c (output_branch): Likewise.
8770 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8772 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8775 2018-02-10 Alan Modra <amodra@gmail.com>
8778 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8779 Specify LR as an input.
8781 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8784 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8785 remove_member_access_dummy_vars): New functions.
8786 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8787 lower_omp_1, execute_lower_omp): Use them.
8789 PR rtl-optimization/84308
8790 * shrink-wrap.c (spread_components): Release todo vector.
8792 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8794 PR rtl-optimization/57193
8795 * ira-color.c (struct allocno_color_data): Add member
8796 conflict_allocno_hard_prefs.
8797 (update_conflict_allocno_hard_prefs): New.
8798 (bucket_allocno_compare_func): Add a preference based on
8799 conflict_allocno_hard_prefs.
8800 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8801 (color_allocnos): Remove a dead code. Initiate
8802 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8804 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8807 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8808 constraint from =wa to wa. Avoid a subreg on the output operand,
8809 instead use a pseudo and subreg it in a move.
8810 (p9_xxbrd_<mode>): Changed to ...
8811 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8812 (p9_xxbrd_v2df): New expander.
8813 (p9_xxbrw_<mode>): Changed to ...
8814 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8815 (p9_xxbrw_v4sf): New expander.
8817 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8819 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8821 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8824 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8825 multiply in 32-bit mode.
8826 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8827 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8830 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8832 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8833 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8834 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8835 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8837 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8840 * dwarf2out.c (is_trivial_indirect_ref): New function.
8841 (dwarf2out_late_global_decl): Do not generate a location
8842 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8843 and that are not defined in the current unit.
8845 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8847 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8848 instead of a libcall for UNORDERED.
8850 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8853 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8854 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8856 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8859 * config/s390/s390.c (s390_set_current_function): Invoke
8860 s390_indirect_branch_settings also if fndecl didn't change.
8862 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8864 * config/rs6000/rs6000.md (blockage): Set length to zero.
8866 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8868 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8870 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8873 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8874 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8878 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8879 PARALLEL incoming that failed vt_get_decl_and_offset check.
8882 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8883 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8885 (get_variable_section): For decls in named .bss* sections pass true as
8886 second argument to bss_initializer_p.
8888 2018-02-09 Marek Polacek <polacek@redhat.com>
8889 Jakub Jelinek <jakub@redhat.com>
8892 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8893 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8894 Sync some changes from cxx_fold_indirect_ref.
8896 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8898 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8900 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8901 (BLOCK_INLINE_ENTRY_LABEL): New.
8902 (dwarf2out_var_location): Disregard inline entry markers.
8903 (inline_entry_data): New struct.
8904 (inline_entry_data_hasher): New hashtable type.
8905 (inline_entry_data_hasher::hash): New.
8906 (inline_entry_data_hasher::equal): New.
8907 (inline_entry_data_table): New variable.
8908 (add_high_low_attributes): Add DW_AT_entry_pc and
8909 DW_AT_GNU_entry_view attributes if a pending entry is found
8910 in inline_entry_data_table. Add old entry_pc attribute only
8911 if debug nonbinding markers are disabled.
8912 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8913 markers are enabled.
8914 (block_within_block_p, dwarf2out_inline_entry): New.
8915 (dwarf2out_finish): Check that no entries remained in
8916 inline_entry_data_table.
8917 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8918 (final_scan_insn, notice_source_line): Likewise.
8919 (rest_of_clean_state): Skip inline entry markers.
8920 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8922 * gimple.c (gimple_build_debug_inline_entry): New.
8923 * gimple.h (enum gimple_debug_subcode): Add
8924 GIMPLE_DEBUG_INLINE_ENTRY.
8925 (gimple_build_debug_inline_entry): Declare.
8926 (gimple_debug_inline_entry_p): New.
8927 (gimple_debug_nonbind_marker_p): Adjust.
8928 * insn-notes.def (INLINE_ENTRY): New.
8929 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8930 inline entry marker notes.
8931 (print_insn): Likewise.
8932 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8933 (INSN_DEBUG_MARKER_KIND): Likewise.
8934 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8935 * tree-inline.c (expand_call_inline): Build and insert
8936 debug_inline_entry stmt.
8937 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8938 inline entry blocks early, if nonbind markers are enabled.
8939 (dump_scope_block): Dump fragment info.
8940 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8941 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8942 (gimple_build_debug_inline_entry): New.
8943 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8944 Enable/disable inline entry points too.
8945 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8946 (DEBUG_INSN): Describe inline entry markers.
8948 * common.opt (gvariable-location-views): New.
8949 (gvariable-location-views=incompat5): New.
8950 * config.in: Rebuilt.
8951 * configure: Rebuilt.
8952 * configure.ac: Test assembler for view support.
8953 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8954 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8955 * dwarf2out.c (var_loc_view): New typedef.
8956 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8957 (dwarf2out_locviews_in_attribute): New.
8958 (dwarf2out_locviews_in_loclist): New.
8959 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8960 (enum dw_line_info_opcode): Add LI_adv_address.
8961 (struct dw_line_info_table): Add view.
8962 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8963 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8964 (zero_view_p): New variable.
8965 (ZERO_VIEW_P): New macro.
8966 (output_asm_line_debug_info): New.
8967 (struct var_loc_node): Add view.
8968 (add_AT_view_list, AT_loc_list): New.
8969 (add_var_loc_to_decl): Add view param. Test it against last.
8970 (new_loc_list): Add view params. Record them.
8971 (AT_loc_list_ptr): Handle loc and view lists.
8972 (view_list_to_loc_list_val_node): New.
8973 (print_dw_val): Handle dw_val_class_view_list.
8974 (size_of_die): Likewise.
8975 (value_format): Likewise.
8976 (loc_list_has_views): New.
8977 (gen_llsym): Set vl_symbol too.
8978 (maybe_gen_llsym, skip_loc_list_entry): New.
8979 (dwarf2out_maybe_output_loclist_view_pair): New.
8980 (output_loc_list): Output view list or entries too.
8981 (output_view_list_offset): New.
8982 (output_die): Handle dw_val_class_view_list.
8983 (output_dwarf_version): New.
8984 (output_compilation_unit_header): Use it.
8985 (output_skeleton_debug_sections): Likewise.
8986 (output_rnglists, output_line_info): Likewise.
8987 (output_pubnames, output_aranges): Update version comments.
8988 (output_one_line_info_table): Output view numbers in asm comments.
8989 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8990 Call maybe_gen_llsym.
8991 (loc_list_from_tree_1): Adjust.
8992 (add_AT_location_description): Create view list attribute if
8993 needed, check it's absent otherwise.
8994 (convert_cfa_to_fb_loc_list): Adjust.
8995 (maybe_emit_file): Call output_asm_line_debug_info for test.
8996 (dwarf2out_var_location): Reset views as needed. Precompute
8997 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8998 attribute. Set view.
8999 (new_line_info_table): Reset next view.
9000 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9001 (dwarf2out_source_line): Likewise. Output view resets and labels to
9002 the assembler, or select appropriate line info opcodes.
9003 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9004 (optimize_string_length): Catch it. Adjust.
9005 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9006 dw_val_class_view_list, and remove it if no longer needed.
9007 (hash_loc_list): Hash view numbers.
9008 (loc_list_hasher::equal): Compare them.
9009 (optimize_location_lists): Check whether a view list symbol is
9010 needed, and whether the locview attribute is present, and
9011 whether they match. Remove the locview attribute if no longer
9013 (index_location_lists): Call skip_loc_list_entry for test.
9014 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9015 Use output_dwarf_version.
9016 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9017 (struct dw_val_node): Add val_view_list.
9018 * final.c (SEEN_NEXT_VIEW): New.
9019 (set_next_view_needed): New.
9020 (clear_next_view_needed): New.
9021 (maybe_output_next_view): New.
9022 (final_start_function): Rename to...
9023 (final_start_function_1): ... this. Take pointer to FIRST,
9024 add SEEN parameter. Emit param bindings in the initial view.
9025 (final_start_function): Reintroduce SEEN-less interface.
9026 (final): Rename to...
9027 (final_1): ... this. Take SEEN parameter. Output final pending
9028 next view at the end.
9029 (final): Reintroduce seen-less interface.
9030 (final_scan_insn): Output pending next view before switching
9031 sections or ending a block. Mark the next view as needed when
9032 outputting variable locations. Notify debug backend of section
9033 changes, and of location view changes.
9034 (rest_of_handle_final): Adjust.
9035 * toplev.c (process_options): Autodetect value for debug variable
9036 location views option. Warn on incompat5 without -gdwarf-5.
9037 * doc/invoke.texi (gvariable-location-views): New.
9038 (gvariable-location-views=incompat5): New.
9039 (gno-variable-location-views): New.
9041 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9043 PR tree-optimization/84136
9044 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9045 that the result of find_edge is non-NULL.
9047 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9050 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9051 storing integer register in SImode. Fix cost of 256 and 512
9052 byte aligned SSE register store.
9054 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9056 * config/i386/i386.c (ix86_multiplication_cost): Fix
9057 multiplication cost for TARGET_AVX512DQ.
9059 2018-02-08 Marek Polacek <polacek@redhat.com>
9061 PR tree-optimization/84238
9062 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9065 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9067 PR tree-optimization/84265
9068 * tree-vect-stmts.c (vectorizable_store): Don't treat
9069 VMAT_CONTIGUOUS accesses as grouped.
9070 (vectorizable_load): Likewise.
9072 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9074 PR tree-optimization/81635
9075 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9076 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9077 (test_round_for_mask): New functions.
9078 (wide_int_cc_tests): Call test_round_for_mask.
9079 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9080 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9081 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9082 range returned by get_range_info.
9084 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9087 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9088 * symtab.c: Include builtins.h
9089 (symtab_node::output_to_lto_symbol_table_p): Move here
9090 from lto-streamer-out.c:output_symbol_p.
9091 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9092 (output_symbol_p): Move all logic to symtab.c
9093 (produce_symtab): Update.
9095 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9097 * config/s390/s390-opts.h (enum indirect_branch): Define.
9098 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9099 (s390_indirect_branch_via_thunk)
9100 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9101 (enum s390_indirect_branch_type): Define.
9102 * config/s390/s390.c (struct s390_frame_layout, struct
9103 machine_function): Remove.
9104 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9105 (indirect_branch_table_label_no, indirect_branch_table_name):
9107 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9108 (enum s390_indirect_branch_option): Define.
9109 (s390_return_addr_from_memory): New function.
9110 (s390_handle_string_attribute): New function.
9111 (s390_attribute_table): Add new attribute handler.
9112 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9113 (s390_indirect_branch_via_thunk): New function.
9114 (s390_indirect_branch_via_inline_thunk): New function.
9115 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9116 sibling call optimization for non z10 compiles.
9117 (s390_emit_call): Force indirect branch target to be a single
9118 register. Add r1 clobber for non-z10 compiles.
9119 (s390_emit_epilogue): Emit return jump via return_use expander.
9120 (s390_reorg): Handle JUMP_INSNs as execute targets.
9121 (s390_option_override_internal): Perform validity checks for the
9122 new command line options.
9123 (s390_indirect_branch_attrvalue): New function.
9124 (s390_indirect_branch_settings): New function.
9125 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9126 (s390_output_indirect_thunk_function): New function.
9127 (s390_code_end): Implement target hook.
9128 (s390_case_values_threshold): Implement target hook.
9129 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9131 * config/s390/s390.h (struct s390_frame_layout)
9132 (struct machine_function): Move here from s390.c.
9133 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9134 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9135 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9136 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9137 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9138 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9139 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9140 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9141 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9142 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9143 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9144 (mnemonic attribute): Add values which aren't recognized
9146 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9147 pattern for branch conversion. Fix mnemonic attribute.
9148 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9149 indirect branch via thunk if requested.
9150 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9151 ("*indirect_jump"): Disable for branch conversion using out of
9153 ("indirect_jump_via_thunk<mode>_z10")
9154 ("indirect_jump_via_thunk<mode>")
9155 ("indirect_jump_via_inlinethunk<mode>_z10")
9156 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9157 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9158 ("casesi_jump_via_inlinethunk<mode>_z10")
9159 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9160 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9161 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9162 ("*indirect2_jump"): Disable for branch conversion.
9163 ("casesi_jump"): Turn into expander and expand patterns for branch
9165 ("return_use"): New expander.
9166 ("*return"): Emit return via thunk and rename it to ...
9167 ("*return<mode>"): ... this one.
9168 * config/s390/s390.opt: Add new options and and enum for the
9171 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9173 * lra-constraints.c (match_reload): Unconditionally use
9174 gen_lowpart_SUBREG, rather than selecting between that
9175 and equivalent gen_rtx_SUBREG code.
9177 2018-02-08 Richard Biener <rguenther@suse.de>
9179 PR tree-optimization/84233
9180 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9181 changed flag instead of boguously re-using phi_inserted.
9183 2018-02-08 Martin Jambor <mjambor@suse.cz>
9185 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9186 static local variables.
9188 2018-02-08 Richard Biener <rguenther@suse.de>
9190 PR tree-optimization/84278
9191 * tree-vect-stmts.c (vectorizable_store): When looking for
9192 smaller vector types to perform grouped strided loads/stores
9193 make sure the mode is supported by the target.
9194 (vectorizable_load): Likewise.
9196 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9198 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9199 Increase LDP/STP opportunities by adding adjacent callee-saves.
9201 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9203 PR rtl-optimization/84068
9204 PR rtl-optimization/83459
9205 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9207 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9209 PR tree-optimization/84224
9210 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9211 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9214 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9217 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9218 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9219 count, remove one USE.
9221 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9223 * doc/install.texi (Configuration): Document the
9224 --with-long-double-format={ibm,ieee} PowerPC configuration
9228 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9229 Convert from define_expand to be define_insn_and_split. Rework
9230 float/double/_Float128 conversions to QI/HI/SImode to work with
9231 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9232 conversions to QI/HImode types did a store and then a load to
9233 truncate the value. For conversions to VSX registers, don't split
9234 the insn, instead emit the code directly. Use the code iterator
9235 any_fix to combine signed and unsigned conversions.
9236 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9237 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9238 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9239 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9240 (fix_<mode>di2_hw): Likewise.
9241 (fixuns_<mode>di2_hw): Likewise.
9242 (fix_<mode>si2_hw): Likewise.
9243 (fixuns_<mode>si2_hw): Likewise.
9244 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9245 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9246 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9247 fix<uns>_trunc<SFDF:mode>si2_p8.
9248 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9249 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9250 (fix<uns>_<mode>_mem): Likewise.
9251 (fctiw<u>z_<mode>_mem): Likewise.
9252 (fix<uns>_<mode>_mem): Likewise.
9253 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9254 the register allocator from doing a direct move to the GPRs to do
9255 a store, and instead use the ISA 3.0 store byte/half-word from
9256 vector register instruction. For IEEE 128-bit floating point,
9257 also optimize stores of 32-bit ints.
9258 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9260 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9262 * genextract.c (push_pathstr_operand): New function to support
9264 (walk_rtx): Call push_pathstr_operand.
9265 (print_path): Support [a-zA-Z].
9267 2018-02-07 Richard Biener <rguenther@suse.de>
9269 PR tree-optimization/84037
9270 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9271 (cse_and_gimplify_to_preheader): Declare.
9272 (vect_get_place_in_interleaving_chain): Likewise.
9273 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9275 (_loop_vec_info::~_loop_vec_info): Delete it.
9276 (cse_and_gimplify_to_preheader): New function.
9277 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9278 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9279 (vectorizable_load): Likewise. For grouped stores always base
9280 the IV on the first element.
9281 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9282 condition before gimplifying.
9284 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9286 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9287 *DIV_EXPR and *MOD_EXPR.
9289 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9292 * config/i386/i386.c (ix86_option_override_internal): Mask out
9293 the CF_SET bit when checking -fcf-protection.
9295 2018-02-07 Tom de Vries <tom@codesourcery.com>
9298 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9301 2018-02-07 Richard Biener <rguenther@suse.de>
9303 PR tree-optimization/84204
9304 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9307 PR tree-optimization/84205
9308 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9309 special-case isl_ast_op_zdiv_r.
9311 PR tree-optimization/84223
9312 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9313 Only add conditions from within the region.
9314 (gather_bbs::after_dom_children): Adjust.
9316 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9319 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9320 * config/avr/avr.md: Only post-reload split REG-REG moves if
9321 either register is GENERAL_REG_P.
9323 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9325 PR tree-optimization/84235
9326 * tree-ssa-scopedtables.c
9327 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9328 if the subtraction is performed in floating point type where NaNs are
9329 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9330 build 1. Formatting fix.
9332 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9335 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9336 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9337 and skip it regardless of bb boundaries. Use CALL_P macro,
9338 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9339 unnecessarily, formatting fix.
9341 2018-02-06 Michael Collison <michael.collison@arm.com>
9343 * config/arm/thumb2.md:
9344 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9345 (*thumb_mov_notscc): Ditto.
9347 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9350 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9351 unsigned_fix, not "s".
9353 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9355 * configure.ac (gcc_fn_eh_frame_ro): New function.
9356 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9357 correct .eh_frame permissions.
9358 * configure: Regenerate.
9360 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9362 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9365 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9367 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9368 Display warning message for -mno-speculate-indirect-jumps.
9370 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9372 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9374 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9376 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9378 PR tree-optimization/84225
9379 * tree-eh.c (find_trapping_overflow): Only call
9380 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9382 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9385 * config/i386/i386.c: Reimplement the check of possible options
9386 -mibt/-mshstk conbination. Change error messages.
9387 * doc/invoke.texi: Fix a typo: remove extra '='.
9389 2018-02-06 Marek Polacek <polacek@redhat.com>
9391 PR tree-optimization/84228
9392 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9394 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9397 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9398 emitted arch directives.
9399 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9400 __ARM_FEATURE_COPROC before changing architectures.
9402 2018-02-06 Richard Biener <rguenther@suse.de>
9404 * config/i386/i386.c (print_reg): Fix typo.
9405 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9407 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9409 * configure: Regenerate.
9411 2018-02-05 Martin Sebor <msebor@redhat.com>
9413 PR tree-optimization/83369
9414 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9417 2018-02-05 Martin Liska <mliska@suse.cz>
9419 * doc/invoke.texi: Cherry-pick upstream r323995.
9421 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9423 * ira.c (ira_init_register_move_cost): Adjust comment.
9425 2018-02-05 Martin Liska <mliska@suse.cz>
9427 PR gcov-profile/84137
9428 * doc/gcov.texi: Fix typo in documentation.
9430 2018-02-05 Martin Liska <mliska@suse.cz>
9432 PR gcov-profile/83879
9433 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9434 using dlopen functionality.
9436 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9438 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9439 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9440 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9441 _mm_maskz_range_round_ss): New intrinsics.
9442 (__builtin_ia32_rangesd128_round)
9443 (__builtin_ia32_rangess128_round): Remove.
9444 (__builtin_ia32_rangesd128_mask_round,
9445 __builtin_ia32_rangess128_mask_round): New builtins.
9446 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9447 __builtin_ia32_rangess128_round): Remove.
9448 (__builtin_ia32_rangesd128_mask_round,
9449 __builtin_ia32_rangess128_mask_round): New builtins.
9450 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9451 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9452 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9453 "<round_saeonly_constraint>")): Changed to ...
9454 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9455 "<round_saeonly_scalar_constraint>")): ... this.
9456 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9457 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9458 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9459 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9460 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9462 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9464 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9466 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9467 Remove all values except native, 8540 and 8548.
9469 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9471 * config/i386/i386.c (ix86_output_function_return): Pass
9472 INVALID_REGNUM, instead of -1, as invalid register number to
9473 indirect_thunk_name and output_indirect_thunk.
9475 2018-02-02 Julia Koval <julia.koval@intel.com>
9477 * config.gcc: Add -march=icelake.
9478 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9479 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9480 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9481 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9482 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9483 (processor_target_table): Add icelake.
9484 (ix86_option_override_internal): Handle new PTAs.
9485 (get_builtin_code_for_version): Handle icelake.
9486 (M_INTEL_COREI7_ICELAKE): New.
9487 (fold_builtin_cpu): Handle icelake.
9488 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9489 * doc/invoke.texi: Add -march=icelake.
9491 2018-02-02 Julia Koval <julia.koval@intel.com>
9493 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9494 to wide_int_bitmask.
9495 * wide-int-bitmask.h: New.
9497 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9500 * config/i386/i386.md: Replace Pmode with word_mode in
9501 builtin_setjmp_setup and builtin_longjmp to support x32.
9503 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9507 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9509 (rs6000_supported_cpu_names): New static variable.
9510 (linux_cpu_translation_table): Likewise.
9511 (elf_platform) <cpu>: Define new static variable and use it.
9512 Translate kernel AT_PLATFORM name to canonical name if needed.
9513 Error if platform name is unknown.
9515 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9518 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9520 2018-02-01 Jeff Law <law@redhat.com>
9523 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9524 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9525 the scratch if RELEASE_VIA_POP is false.
9526 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9527 If we have to save a temporary register, decrement SIZE appropriately.
9528 Pass new arguments to release_scratch_register_on_entry.
9529 (ix86_adjust_stack_and_probe): Likewise.
9530 (ix86_emit_probe_stack_range): Pass new arguments to
9531 release_scratch_register_on_entry.
9533 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9535 PR rtl-optimization/84157
9536 * combine.c (change_zero_ext): Use REG_P predicate in
9537 front of HARD_REGISTER_P predicate.
9539 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9541 * config/avr/avr.c (avr_option_override): Move disabling of
9542 -fdelete-null-pointer-checks to...
9543 * common/config/avr/avr-common.c (avr_option_optimization_table):
9546 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9548 PR tree-optimization/81635
9549 * tree-data-ref.c (split_constant_offset_1): For types that
9550 wrap on overflow, try to use range info to prove that wrapping
9553 2018-02-01 Renlin Li <renlin.li@arm.com>
9556 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9558 (aarch64_register_move_cost): Likewise.
9559 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9561 (REG_CLASS_NAMES): Likewise.
9562 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9563 TAILCALL_ADDR_REGS. Remove IP registers.
9564 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9566 2018-02-01 Richard Biener <rguenther@suse.de>
9568 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9569 for specifying RPO order and allow NULL for that.
9570 * domwalk.c (dom_walker::dom_walker): Likewise.
9571 (dom_walker::walk): Handle NULL RPO order.
9572 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9574 (rewrite_update_dom_walker): Likewise.
9575 (mark_def_dom_walker): Likewise.
9577 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9579 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9580 (aarch64_maybe_expand_sve_subreg_move): Declare.
9581 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9582 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9584 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9585 that are semantically a reverse operation.
9586 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9587 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9588 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9590 (aarch64_can_change_mode_class): For big-endian, forbid changes
9591 between two SVE modes if they have different element sizes.
9593 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9595 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9596 the TImode handling for big-endian targets.
9598 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9600 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9601 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9603 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9604 Remove BSWAP handing for big-endian targets and use the form of
9605 LD1RQ appropariate for the mode.
9607 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9609 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9610 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9613 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9616 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9617 check for operands that need to go through aarch64_sve_reload_be.
9619 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9621 PR tree-optimization/81661
9622 PR tree-optimization/84117
9623 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9624 * tree-eh.c: Include gimplify.h.
9625 (find_trapping_overflow, replace_trapping_overflow,
9626 rewrite_to_non_trapping_overflow): New functions.
9627 * tree-vect-loop.c: Include tree-eh.h.
9628 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9629 * tree-data-ref.c: Include tree-eh.h.
9630 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9632 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9634 PR rtl-optimization/84123
9635 * combine.c (change_zero_ext): Check if hard register satisfies
9636 can_change_dest_mode before calling gen_lowpart_SUBREG.
9638 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9641 * ira.c (ira_init_register_move_cost): Remove assert.
9643 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9645 PR rtl-optimization/84071
9646 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9647 * doc/tm.texi: Regenerate.
9649 2018-01-31 Richard Biener <rguenther@suse.de>
9651 PR tree-optimization/84132
9652 * tree-data-ref.c (analyze_miv_subscript): Properly
9653 check whether evolution_function_is_affine_multivariate_p
9654 before calling gcd_of_steps_may_divide_p.
9656 2018-01-31 Julia Koval <julia.koval@intel.com>
9659 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9660 * config/i386/i386.md (rdpid_rex64) New.
9661 (rdpid): Make 32bit only.
9663 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9666 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9667 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9669 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9672 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9674 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9676 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9678 PR rtl-optimization/84071
9679 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9680 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9682 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9684 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9685 (arc_attribute_table): Add 'aux' attribute.
9686 (arc_in_small_data_p): Consider aux like variables.
9687 (arc_is_aux_reg_p): New function.
9688 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9689 (arc_get_aux_arg): New function.
9690 (prepare_move_operands): Handle aux-register access.
9691 (arc_handle_aux_attribute): New function.
9692 * doc/extend.texi (ARC Variable attributes): Add subsection.
9694 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9696 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9697 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9698 (arc_attribute_table): Add 'uncached' attribute.
9699 (arc_print_operand): Print '.di' flag for uncached memory
9701 (arc_in_small_data_p): Do not consider for small data the uncached
9703 (arc_is_uncached_mem_p): New function.
9704 * config/arc/predicates.md (compact_store_memory_operand): Check
9705 for uncached memory accesses.
9706 (nonvol_nonimm_operand): Likewise.
9707 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9709 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9712 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9713 falign-loops=): Add Optimization flag.
9715 2018-01-30 Jeff Law <law@redhat.com>
9718 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9719 INT_REGISTERS_SAVED. Check it prior to calling
9720 get_scratch_register_on_entry.
9721 (ix86_adjust_stack_and_probe): Similarly.
9722 (ix86_emit_probe_stack_range): Similarly.
9723 (ix86_expand_prologue): Corresponding changes.
9725 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9728 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9729 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9731 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9734 * lra-constraints.c (curr_insn_transform): Process AND in the
9737 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9739 PR rtl-optimization/83986
9740 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9741 dependence against last_pending_memory_flush in addition to
9744 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9746 PR tree-optimization/81611
9747 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9750 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9753 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9756 2018-01-30 Richard Biener <rguenther@suse.de>
9757 Jakub Jelinek <jakub@redhat.com>
9759 PR tree-optimization/84111
9760 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9761 inner loops added during recursion, as they don't have up-to-date
9764 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9767 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9768 (can_inline_edge_by_limits_p): ... here.
9769 (can_early_inline_edge_p, check_callers,
9770 update_caller_keys, update_callee_keys, recursive_inlining,
9771 add_new_edges_to_heap, speculation_useful_p,
9772 inline_small_functions,
9773 inline_small_functions, flatten_function,
9774 inline_to_all_callers_1): Update.
9776 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9778 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9781 2018-01-30 Richard Biener <rguenther@suse.de>
9783 PR tree-optimization/83008
9784 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9785 invariant and constant vector uses in stmts when they need
9788 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9791 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9792 * configure: Regenerate.
9794 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9796 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9798 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9799 Use gen_rtx_REG rather than gen_lowpart.
9801 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9803 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9804 rather than 0 when creating partial subregs.
9806 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9808 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9811 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9814 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9815 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9816 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9817 flags. This restores the settings used before the 2017-07-24.
9818 Turning off pre increment/decrement/modify allows IVOPTS to
9819 optimize DF/SF loops where the index is an int.
9821 2018-01-29 Richard Biener <rguenther@suse.de>
9822 Kelvin Nilsen <kelvin@gcc.gnu.org>
9825 * tree-vect-stmts.c (vectorizable_call): Don't call
9826 targetm.vectorize_builtin_md_vectorized_function if callee is
9829 2018-01-22 Carl Love <cel@us.ibm.com>
9831 * doc/extend.tex: Fix typo in second arg in
9832 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9834 2018-01-29 Richard Biener <rguenther@suse.de>
9836 PR tree-optimization/84086
9837 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9838 (flush_ssaname_freelist): When SSA names were released reset
9839 the SCEV hash table.
9841 2018-01-29 Richard Biener <rguenther@suse.de>
9843 PR tree-optimization/84057
9844 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9845 removed paths when removing edges.
9847 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9849 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9850 -mfunction-return=@var{choice}.
9852 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9855 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9857 (layout::print_source_line): Likewise.
9858 (test_get_line_width_without_trailing_whitespace): Add test cases.
9860 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9863 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9866 2018-01-26 Jim Wilson <jimw@sifive.com>
9868 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9870 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9873 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9875 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9876 and CMP + SUB-immediate -> SUBS.
9878 2018-01-26 Martin Sebor <msebor@redhat.com>
9880 PR tree-optimization/83896
9881 * tree-ssa-strlen.c (get_string_len): Rename...
9882 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9883 Avoid assuming length is constant.
9884 (handle_char_store): Use HOST_WIDE_INT for string length.
9886 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9889 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9890 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9892 2018-01-26 Richard Biener <rguenther@suse.de>
9894 PR rtl-optimization/84003
9895 * dse.c (record_store): Only record redundant stores when
9896 the earlier store aliases at least all accesses the later one does.
9898 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9900 PR rtl-optimization/83985
9901 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9902 REG_CFA_RESTORE insns.
9903 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9904 REG_CFA_RESTORE insns here.
9907 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9908 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9910 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9912 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9913 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9914 (arc_init): Likewise.
9915 (arc_override_options): Likewise.
9916 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9918 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9920 * config/arc/arc.h (TARGET_DBNZ): Define.
9921 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9922 properly set the tune attribute.
9923 (dbnz): Use TARGET_DBNZ guard.
9924 * config/arc/arc.opt (mtune): Add core3 option.
9926 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9928 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9929 recognize new pic like addresses.
9930 (arc_delegitimize_address): Clean up.
9932 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9934 * config/arc/arc-arches.def: Option mrf16 valid for all
9936 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9937 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9938 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9939 * config/arc/arc-tables.opt: Regenerate.
9940 * config/arc/arc.c (arc_conditional_register_usage): Handle
9941 reduced register file case.
9942 (arc_file_start): Set must have build attributes.
9943 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9945 * config/arc/arc.opt (mrf16): Add new option.
9946 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9947 * config/arc/genmultilib.awk: Handle new mrf16 option.
9948 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9949 * config/arc/t-multilib: Regenerate.
9950 * doc/invoke.texi (ARC Options): Document mrf16 option.
9952 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9954 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9955 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9956 (arc_attribute_table): Add 'secure_call' attribute.
9957 (arc_print_operand): Print secure call operand.
9958 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9960 (arc_is_secure_call_p): New function. * config/arc/arc.md
9961 (call_i): Add support for sjli instruction.
9962 (call_value_i): Likewise.
9963 * config/arc/constraints.md (Csc): New constraint.
9965 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9966 John Eric Martin <John.Martin@emmicro-us.com>
9968 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9969 * config/arc/arc.c (_arc_jli_section): New struct.
9970 (arc_jli_section): New type.
9971 (rc_jli_sections): New static variable.
9972 (arc_handle_jli_attribute): New function.
9973 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9974 (arc_file_end): New function.
9975 (TARGET_ASM_FILE_END): Define.
9976 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9977 (arc_add_jli_section): New function.
9978 (jli_call_scan): Likewise.
9979 (arc_reorg): Call jli_call_scan.
9980 (arc_output_addsi): Remove 'S' from printing asm operand.
9981 (arc_is_jli_call_p): New function.
9982 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9984 (movhi_insn): Likewise.
9985 (movsi_insn): Likewise.
9986 (movsi_set_cc_insn): Likewise.
9987 (loadqi_update): Likewise.
9988 (load_zeroextendqisi_update): Likewise.
9989 (load_signextendqisi_update): Likewise.
9990 (loadhi_update): Likewise.
9991 (load_zeroextendhisi_update): Likewise.
9992 (load_signextendhisi_update): Likewise.
9993 (loadsi_update): Likewise.
9994 (loadsf_update): Likewise.
9995 (movsicc_insn): Likewise.
9996 (bset_insn): Likewise.
9997 (bxor_insn): Likewise.
9998 (bclr_insn): Likewise.
9999 (bmsk_insn): Likewise.
10000 (bicsi3_insn): Likewise.
10001 (cmpsi_cc_c_insn): Likewise.
10002 (movsi_ne): Likewise.
10003 (movsi_cond_exec): Likewise.
10004 (clrsbsi2): Likewise.
10005 (norm_f): Likewise.
10015 (call_i): Remove 'S' asm letter, add jli instruction.
10016 (call_value_i): Likewise.
10017 * config/arc/arc.op (mjli-always): New option.
10018 * config/arc/constraints.md (Cji): New constraint.
10019 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10021 (subsf3_fpx): Likewise.
10022 (mulsf3_fpx): Likewise.
10023 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10025 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10026 function attrbutes.
10027 * doc/invoke.texi (ARC): Document mjli-always option.
10029 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10031 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10032 avoid addition with 0 and use incw and decw where possible.
10034 2018-01-26 Richard Biener <rguenther@suse.de>
10036 PR tree-optimization/81082
10037 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10038 association if it requires casting to unsigned.
10039 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10040 from fold_plusminus_mult_expr to catch important cases late when
10041 range info is available.
10043 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10045 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10046 * configure.ac (hidden_linkonce): New test.
10047 * configure: Regenerate.
10048 * config.in: Regenerate.
10050 2018-01-26 Julia Koval <julia.koval@intel.com>
10052 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10053 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10054 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10055 _mm_mask_bitshuffle_epi64_mask): Fix type.
10056 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10057 USI_FTYPE_V4DI_V4DI_USI): Remove.
10058 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10059 __builtin_ia32_vpshufbitqmb256_mask,
10060 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10061 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10062 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10064 2018-01-26 Alan Modra <amodra@gmail.com>
10067 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10068 UNSPEC_VBPERMQ. Sort other unspecs.
10070 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10072 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10074 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10076 PR middle-end/83055
10077 * predict.c (drop_profile): Do not push/pop cfun; update also
10079 (handle_missing_profiles): Fix logic looking for zero profiles.
10081 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10083 PR middle-end/83977
10084 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10085 on functions with #pragma omp declare simd or functions with simd
10087 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10088 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10089 Remove trailing \n from warning_at calls.
10091 2018-01-25 Tom de Vries <tom@codesourcery.com>
10094 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10095 for neutered workers.
10097 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10100 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10101 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10103 2018-01-24 Jeff Law <law@redhat.com>
10106 * i386.c (get_probe_interval): Move to earlier point.
10107 (ix86_compute_frame_layout): If -fstack-clash-protection and
10108 the frame is larger than the probe interval, then use pushes
10109 to save registers rather than reg->mem moves.
10110 (ix86_expand_prologue): Remove conditional for int_registers_saved
10113 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10116 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10117 min/max for never referenced object.
10119 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10121 PR middle-end/83977
10122 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10124 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10125 attributes from DECL_ATTRIBUTES (decl) without affecting
10126 DECL_ATTRIBUTES (current_function_decl).
10127 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10128 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10130 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10132 PR tree-optimization/83979
10133 * fold-const.c (fold_comparison): Use constant_boolean_node
10134 instead of boolean_{true,false}_node.
10136 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10138 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10141 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10144 Simplify the clause that sets the length attribute.
10145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10146 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10147 clause that sets the length attribute.
10148 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10150 2018-01-24 Tom de Vries <tom@codesourcery.com>
10153 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10154 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10155 Add strict parameter.
10156 (prevent_branch_around_nothing): Insert dummy insn between branch to
10157 label and label with no ptx insn inbetween.
10158 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10160 2018-01-24 Tom de Vries <tom@codesourcery.com>
10163 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10164 for neutered threads in warp.
10165 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10167 2018-01-24 Richard Biener <rguenther@suse.de>
10169 PR tree-optimization/83176
10170 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10173 2018-01-24 Richard Biener <rguenther@suse.de>
10175 PR tree-optimization/82819
10176 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10177 code generating pluses that are no-ops in the target precision.
10179 2018-01-24 Richard Biener <rguenther@suse.de>
10181 PR middle-end/84000
10182 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10184 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10186 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10187 to merge probabilities.
10188 * predict.c (probably_never_executed): Also mark as cold functions
10189 with global 0 profile and guessed local profile.
10190 * profile-count.c (profile_probability::combine_with_count): New
10192 * profile-count.h (profile_probability::operator*,
10193 profile_probability::operator*=, profile_probability::operator/,
10194 profile_probability::operator/=): Reduce precision to adjusted
10195 and set value to guessed on contradictory divisions.
10196 (profile_probability::combine_with_freq): Remove.
10197 (profile_probability::combine_wiht_count): Declare.
10198 (profile_count::force_nonzero):: Set to adjusted.
10199 (profile_count::probability_in):: Set quality to adjusted.
10200 * tree-ssa-tail-merge.c (replace_block_by): Use
10201 combine_with_count.
10203 2018-01-23 Andrew Waterman <andrew@sifive.com>
10204 Jim Wilson <jimw@sifive.com>
10206 * config/riscv/riscv.c (riscv_stack_boundary): New.
10207 (riscv_option_override): Set riscv_stack_boundary. Handle
10208 riscv_preferred_stack_boundary_arg.
10209 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10210 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10211 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10212 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10213 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10214 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10216 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10219 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10220 of struct ix86_frame.
10221 (ix86_expand_epilogue): Likewise. Add a local variable for
10222 the reg_save_offset field in struct ix86_frame.
10224 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10226 PR tree-optimization/82604
10227 * tree-loop-distribution.c (enum partition_kind): New enum item
10228 PKIND_PARTIAL_MEMSET.
10229 (partition_builtin_p): Support above new enum item.
10230 (generate_code_for_partition): Ditto.
10231 (compute_access_range): Differentiate cases that equality can be
10232 proven at all loops, the innermost loops or no loops.
10233 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10234 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10235 (finalize_partitions, distribute_loop): Don't fuse partition of
10236 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10237 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10238 parloop is enabled.
10240 2018-01-23 Martin Liska <mliska@suse.cz>
10242 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10243 order to ignore the predictor.
10244 (PRED_POLYMORPHIC_CALL): Likewise.
10245 (PRED_RECURSIVE_CALL): Likewise.
10247 2018-01-23 Martin Liska <mliska@suse.cz>
10249 * tree-profile.c (tree_profiling): Print function header to
10250 aware reader which function we are working on.
10251 * value-prof.c (gimple_find_values_to_profile): Do not print
10252 not interesting value histograms.
10254 2018-01-23 Martin Liska <mliska@suse.cz>
10256 * profile-count.h (enum profile_quality): Add
10257 profile_uninitialized as the first value. Do not number values
10258 as they are zero based.
10259 (profile_count::verify): Update sanity check.
10260 (profile_probability::verify): Likewise.
10262 2018-01-23 Nathan Sidwell <nathan@acm.org>
10264 * doc/invoke.texi (ffor-scope): Deprecate.
10266 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10268 PR tree-optimization/83510
10269 * domwalk.c (set_all_edges_as_executable): New function.
10270 (dom_walker::dom_walker): Convert bool param
10271 "skip_unreachable_blocks" to enum reachability. Move setup of
10272 edge flags to set_all_edges_as_executable and only do it when
10273 reachability is REACHABLE_BLOCKS.
10274 * domwalk.h (enum dom_walker::reachability): New enum.
10275 (dom_walker::dom_walker): Convert bool param
10276 "skip_unreachable_blocks" to enum reachability.
10277 (set_all_edges_as_executable): New decl.
10278 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10279 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10281 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10282 but converting true to REACHABLE_BLOCKS.
10283 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10285 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10286 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10287 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10289 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10290 if check_all_array_refs will be called.
10292 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10294 * tree.c (selftest::test_location_wrappers): Add more test
10297 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10299 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10300 (selftest::test_bit_in_range): Likewise.
10302 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10305 * doc/sourcebuild.texi (vect_float): Say that the selector
10306 only describes the situation when -funsafe-math-optimizations is on.
10307 (vect_float_strict): Document.
10309 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10311 PR tree-optimization/83965
10312 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10313 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10314 instead of checking only for a reduction.
10315 (vect_recog_widen_sum_pattern): Likewise.
10317 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10319 * predict.c (probably_never_executed): Only use precise profile info.
10320 (compute_function_frequency): Skip after inlining hack since we now
10321 have quality checking.
10323 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10325 * profile-count.h (profile_probability::very_unlikely,
10326 profile_probability::unlikely, profile_probability::even): Set
10327 precision to guessed.
10329 2018-01-23 Richard Biener <rguenther@suse.de>
10331 PR tree-optimization/83963
10332 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10333 Properly terminate dominator walk when crossing the exit edge not
10334 when visiting its source block.
10336 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10339 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10340 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10342 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10344 PR tree-optimization/83957
10345 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10346 semicolon after for body surrounded by braces.
10348 PR tree-optimization/83081
10349 * profile-count.h (profile_probability::split): New method.
10350 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10351 Use profile_probability::split.
10352 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10353 when splitting a single conditional jump into 2.
10355 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10357 PR tree-optimization/69452
10358 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10361 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10363 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10364 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10365 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10367 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10369 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10371 * config/rl78/rl78.md (movdi): New define_expand.
10372 * config/rl78/rl78.c (rl78_split_movdi): New function.
10374 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10377 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10379 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10380 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10381 128-bit to produce an UNSPEC move to get the double word with the
10382 signbit and then a shift directly to do signbit.
10383 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10384 implementation with a new version that just does either a direct
10385 move or a regular move. Move memory interface to separate insns.
10386 Move insns so they are next to the expander.
10387 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10388 with signbit move. Split big and little endian case.
10389 (signbit<mode>2_dm_mem_le): Likewise.
10390 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10391 (signbit<mode>2_dm2): Likewise.
10393 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10395 * config/rl78/rl78.md (anddi3): New define_expand.
10397 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10399 * config/rl78/rl78.md (umindi3): New define_expand.
10401 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10403 * config/rl78/rl78.md (smindi3): New define_expand.
10405 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10407 * config/rl78/rl78.md (smaxdi3): New define_expand.
10409 2018-01-22 Carl Love <cel@us.ibm.com>
10411 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10412 LVX_V1TI): Add macro expansion.
10413 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10414 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10415 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10416 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10417 Change check to determine if the instruction is a byte reversing
10418 entry. Fix typo in comment.
10419 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10420 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10421 Add def_builtin calls for new builtins.
10422 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10423 Add define_insn expansion.
10425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10427 * config/rl78/rl78.md (umaxdi3): New define_expand.
10429 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10431 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10432 for non-QImode registers.
10434 2018-01-22 Richard Biener <rguenther@suse.de>
10436 PR tree-optimization/83963
10437 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10438 including the loop exit block.
10439 (scop_detection::merge_sese): Likewise.
10440 (scop_detection::add_scop): Do it here instead.
10442 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10444 * doc/sourcebuild.texi (arm_softfloat): Document.
10446 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10449 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10450 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10451 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10453 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10454 David Edelsohn <dje.gcc@gmail.com>
10457 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10458 Change "crset eq" to "crset 2".
10459 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10460 (*call_indirect_aix<mode>_nospec): Likewise.
10461 (*call_value_indirect_aix<mode>_nospec): Likewise.
10462 (*call_indirect_elfv2<mode>_nospec): Likewise.
10463 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10464 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10465 change assembly output from . to $.
10466 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10467 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10468 (*tablejump<mode>_internal1_nospec): Likewise.
10470 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10473 * config/sh/sh_optimize_sett_clrt.cc:
10474 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10476 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10478 PR tree-optimization/83940
10479 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10480 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10481 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10482 use it to pass back the definition type.
10483 (vect_check_store_rhs): Likewise rhs_dt_out.
10484 (vect_build_gather_load_calls): Add a mask_dt argument and use
10485 it instead of a call to vect_is_simple_use.
10486 (vectorizable_store): Update calls to vect_check_load_store_mask
10487 and vect_check_store_rhs. Use the dt returned by the latter instead
10488 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10489 instead of calls to vect_is_simple_use. Pass the scalar rather
10490 than the vector operand to vect_is_simple_use when handling
10491 second and subsequent copies of an rhs value.
10492 (vectorizable_load): Update calls to vect_check_load_store_mask
10493 and vect_build_gather_load_calls. Use the cached mask_dt and
10494 gs_info.offset_dt instead of calls to vect_is_simple_use.
10496 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10498 PR middle-end/83945
10499 * tree-emutls.c: Include gimplify.h.
10500 (lower_emutls_2): New function.
10501 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10502 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10503 it before further processing.
10506 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10507 UINTVAL (trueop1) instead of INTVAL (op1).
10509 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10513 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10514 INCOMING_FRAME_SP_OFFSET if not defined.
10515 (scan_trace): Add ENTRY argument. If true and
10516 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10517 emit a note to adjust the CFA offset.
10518 (create_cfi_notes): Adjust scan_trace callers.
10519 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10520 INCOMING_FRAME_SP_OFFSET in the CIE.
10521 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10522 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10524 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10525 * doc/tm.texi: Regenerated.
10527 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10529 PR rtl-optimization/83147
10530 * lra-constraints.c (remove_inheritance_pseudos): Use
10531 lra_substitute_pseudo_within_insn.
10533 2018-01-19 Tom de Vries <tom@codesourcery.com>
10534 Cesar Philippidis <cesar@codesourcery.com>
10537 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10539 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10542 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10543 spaces for function labels.
10545 2018-01-19 Martin Liska <mliska@suse.cz>
10547 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10548 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10549 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10550 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10551 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10552 (PRED_CONST_RETURN): Change from 69 to 65.
10553 (PRED_NULL_RETURN): Change from 91 to 71.
10554 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10555 (PRED_LOOP_GUARD): Change from 66 to 73.
10557 2018-01-19 Martin Liska <mliska@suse.cz>
10559 * predict.c (predict_insn_def): Add new assert.
10560 (struct branch_predictor): Change type to signed integer.
10561 (test_prediction_value_range): Amend test to cover
10562 PROB_UNINITIALIZED.
10563 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10564 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10565 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10566 (PRED_LOOP_IV_COMPARE): Likewise.
10567 * predict.h (PROB_UNINITIALIZED): Define new constant.
10569 2018-01-19 Martin Liska <mliska@suse.cz>
10571 * predict.c (dump_prediction): Add new format for
10572 analyze_brprob.py script which is enabled with -details
10574 * profile-count.h (precise_p): New function.
10576 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10578 PR tree-optimization/83922
10579 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10580 there are no statements that need masking.
10581 (vect_active_double_reduction_p): New function.
10582 (vect_analyze_loop_operations): Use it when handling phis that
10583 are not in the loop header.
10585 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10587 PR tree-optimization/83914
10588 * tree-vect-loop.c (vectorizable_induction): Don't convert
10589 init_expr or apply the peeling adjustment for inductions
10590 that are nested within the vectorized loop.
10592 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10594 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10597 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10601 * function.h (gimplify_parameters): Add gimple_seq * argument.
10602 * function.c: Include gimple.h and options.h.
10603 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10604 for the added local temporaries if needed.
10605 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10606 if there are any parameter cleanups, wrap whole body into a
10607 try/finally with the cleanups.
10609 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10612 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10613 Use GET_MODE_CLASS for scalar floating point.
10615 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10619 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10620 Fix call of call_cgraph_insertion_hooks.
10622 2018-01-18 Martin Sebor <msebor@redhat.com>
10624 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10626 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10629 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10632 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10635 * common.opt: (-ffile-prefix-map): New option.
10636 * opts.c (common_handle_option): Defer it.
10637 * opts-global.c (handle_common_deferred_options): Handle it.
10638 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10639 * file-prefix-map.h: New file.
10640 (remap_debug_filename, add_debug_prefix_map): ...here.
10641 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10642 * final.c (debug_prefix_map, add_debug_prefix_map
10643 remap_debug_filename): Move to...
10644 * file-prefix-map.c: New file.
10645 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10646 generalize, get rid of alloca(), use strrchr() instead of strchr().
10647 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10648 Implement in terms of add_prefix_map().
10649 (remap_macro_filename, remap_debug_filename): Implement in term of
10651 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10652 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10653 * dbxout.c: Include file-prefix-map.h.
10654 * varasm.c: Likewise.
10655 * vmsdbgout.c: Likewise.
10656 * xcoffout.c: Likewise.
10657 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10658 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10659 * doc/invoke.texi (-ffile-prefix-map): Document.
10660 (-fdebug-prefix-map): Update description.
10662 2018-01-18 Martin Liska <mliska@suse.cz>
10664 * config/i386/i386.c (indirect_thunk_name): Document that also
10666 (output_indirect_thunk): Document why both instructions
10667 (pause and lfence) are generated.
10669 2018-01-18 Richard Biener <rguenther@suse.de>
10671 PR tree-optimization/83887
10672 * graphite-scop-detection.c
10673 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10674 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10675 (scop_detection::merge_sese): Re-implement with a flood-fill
10676 algorithm that properly finds a SESE region if it exists.
10678 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10681 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10682 pointer_diff optimizations use view_convert instead of convert.
10684 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10686 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10687 Generate different code for -mno-speculate-indirect-jumps.
10688 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10689 (*call_indirect_aix<mode>): Disable for
10690 -mno-speculate-indirect-jumps.
10691 (*call_indirect_aix<mode>_nospec): New define_insn.
10692 (*call_value_indirect_aix<mode>): Disable for
10693 -mno-speculate-indirect-jumps.
10694 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10695 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10696 -mno-speculate-indirect-jumps.
10697 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10699 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10701 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10702 long double type, set the flags for noting the default long double
10703 type, even if we don't pass or return a long double type.
10705 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10708 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10711 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10713 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10714 support for merge[hl].
10715 (fold_mergehl_helper): New helper function.
10716 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10717 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10718 (altivec_vmrglw_direct): Add xxmrglw insn.
10720 2018-01-17 Andrew Waterman <andrew@sifive.com>
10722 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10723 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10725 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10728 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10729 call the lto_location_cache before reading the
10730 DECL_SOURCE_LOCATION of the types.
10732 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10733 Richard Sandiford <richard.sandiford@linaro.org>
10735 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10736 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10737 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10738 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10739 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10741 * config/aarch64/constraints.md (aarch64_movti_operand):
10743 * config/aarch64/predicates.md (Uti): Add new constraint.
10745 2018-01-17 Carl Love <cel@us.ibm.com>
10747 * config/rs6000/vsx.md (define_expand xl_len_r,
10748 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10749 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10751 (define_expand, define_insn): Move the shift left from the
10752 define_insn to the define_expand for lxvl and stxvl instructions.
10753 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10754 and XL_LEN_R definitions to PURE.
10756 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10758 * config/i386/i386.c (indirect_thunk_name): Declare regno
10759 as unsigned int. Compare regno with INVALID_REGNUM.
10760 (output_indirect_thunk): Ditto.
10761 (output_indirect_thunk_function): Ditto.
10762 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10763 in the call to output_indirect_thunk_function.
10765 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10767 PR middle-end/83884
10768 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10769 rather than the size of inner_type to determine the stack slot size
10770 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10772 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10775 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10778 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10780 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10781 endian Linux systems to optionally enable multilibs for selecting
10782 the long double type if the user configured an explicit type.
10783 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10784 have no long double multilibs if not defined.
10785 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10786 warn if the user used -mabi={ieee,ibm}longdouble and we built
10787 multilibs for long double.
10788 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10789 appropriate multilib option.
10790 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10792 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10793 for building long double multilibs.
10794 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10796 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10798 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10801 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10803 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10806 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10809 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10812 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10814 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10815 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10817 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10819 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10820 different rtl trees depending on TARGET_64BIT.
10821 (rs6000_gen_lvx): Likewise.
10823 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10825 * config/visium/visium.md (nop): Tweak comment.
10826 (hazard_nop): Likewise.
10828 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10830 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10831 -mspeculate-indirect-jumps.
10832 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10833 for -mno-speculate-indirect-jumps.
10834 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10835 (*call_value_indirect_elfv2<mode>): Disable for
10836 -mno-speculate-indirect-jumps.
10837 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10838 (indirect_jump): Emit different RTL for
10839 -mno-speculate-indirect-jumps.
10840 (*indirect_jump<mode>): Disable for
10841 -mno-speculate-indirect-jumps.
10842 (*indirect_jump<mode>_nospec): New define_insn.
10843 (tablejump): Emit different RTL for
10844 -mno-speculate-indirect-jumps.
10845 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10846 (tablejumpsi_nospec): New define_expand.
10847 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10848 (tablejumpdi_nospec): New define_expand.
10849 (*tablejump<mode>_internal1): Disable for
10850 -mno-speculate-indirect-jumps.
10851 (*tablejump<mode>_internal1_nospec): New define_insn.
10852 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10855 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10857 * caller-save.c (insert_save): Drop unnecessary parameter. All
10860 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10861 Richard Biener <rguenth@suse.de>
10864 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10865 return early, inline manually is_gimple_sizepos. Make sure if we
10866 call gimplify_expr we don't end up with a gimple constant.
10867 * tree.c (variably_modified_type_p): Don't return true for
10868 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10869 * gimplify.h (is_gimple_sizepos): Remove.
10871 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10873 PR tree-optimization/83857
10874 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10875 vectorizable_live_operation for pure SLP statements.
10876 (vectorizable_live_operation): Handle PHIs.
10878 2018-01-16 Richard Biener <rguenther@suse.de>
10880 PR tree-optimization/83867
10881 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10882 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10884 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10887 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10888 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10889 If off is not INTEGER_CST, issue a may not be aligned warning
10890 rather than isn't aligned. Use isn%'t rather than isn't.
10891 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10893 <case MULT_EXPR>: Improve the case when bottom and one of the
10894 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10895 operand, in that case check if the other operand is multiple of
10896 bottom divided by the INTEGER_CST operand.
10898 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10901 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10902 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10903 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10904 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10905 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10906 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10907 (pa_function_arg_size): New function.
10909 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10911 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10912 in a separate statement.
10914 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10916 PR tree-optimization/83847
10917 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10918 group gathers and scatters.
10920 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10922 PR rtl-optimization/86620
10923 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10925 PR rtl-optimization/83213
10926 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10927 to last if both are JUMP_INSNs.
10929 PR tree-optimization/83843
10930 * gimple-ssa-store-merging.c
10931 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10932 store_immediate_info for bswap/nop orig_stores.
10934 2018-01-15 Andrew Waterman <andrew@sifive.com>
10936 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10938 <UDIV>: Increase cost if !TARGET_DIV.
10940 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10942 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10943 (define_attr "cr_logical_3op"): New.
10944 (cceq_ior_compare): Adjust.
10945 (cceq_ior_compare_complement): Adjust.
10946 (*cceq_rev_compare): Adjust.
10947 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10948 (is_cracked_insn): Adjust.
10949 (insn_must_be_first_in_group): Adjust.
10950 * config/rs6000/40x.md: Adjust.
10951 * config/rs6000/440.md: Adjust.
10952 * config/rs6000/476.md: Adjust.
10953 * config/rs6000/601.md: Adjust.
10954 * config/rs6000/603.md: Adjust.
10955 * config/rs6000/6xx.md: Adjust.
10956 * config/rs6000/7450.md: Adjust.
10957 * config/rs6000/7xx.md: Adjust.
10958 * config/rs6000/8540.md: Adjust.
10959 * config/rs6000/cell.md: Adjust.
10960 * config/rs6000/e300c2c3.md: Adjust.
10961 * config/rs6000/e500mc.md: Adjust.
10962 * config/rs6000/e500mc64.md: Adjust.
10963 * config/rs6000/e5500.md: Adjust.
10964 * config/rs6000/e6500.md: Adjust.
10965 * config/rs6000/mpc.md: Adjust.
10966 * config/rs6000/power4.md: Adjust.
10967 * config/rs6000/power5.md: Adjust.
10968 * config/rs6000/power6.md: Adjust.
10969 * config/rs6000/power7.md: Adjust.
10970 * config/rs6000/power8.md: Adjust.
10971 * config/rs6000/power9.md: Adjust.
10972 * config/rs6000/rs64.md: Adjust.
10973 * config/rs6000/titan.md: Adjust.
10975 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10977 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10978 ix86_indirect_branch_register logic.
10980 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10982 * config/i386/constraints.md (Bs): Update
10983 ix86_indirect_branch_register check. Don't check
10984 ix86_indirect_branch_register with GOT_memory_operand.
10986 * config/i386/predicates.md (GOT_memory_operand): Don't check
10987 ix86_indirect_branch_register here.
10988 (GOT32_symbol_operand): Likewise.
10990 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10992 * config/i386/predicates.md (constant_call_address_operand):
10993 Rewrite ix86_indirect_branch_register logic.
10994 (sibcall_insn_operand): Likewise.
10996 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10998 * config/i386/constraints.md (Bs): Replace
10999 ix86_indirect_branch_thunk_register with
11000 ix86_indirect_branch_register.
11002 * config/i386/i386.md (indirect_jump): Likewise.
11003 (tablejump): Likewise.
11004 (*sibcall_memory): Likewise.
11005 (*sibcall_value_memory): Likewise.
11006 Peepholes of indirect call and jump via memory: Likewise.
11007 * config/i386/i386.opt: Likewise.
11008 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11009 (GOT_memory_operand): Likewise.
11010 (call_insn_operand): Likewise.
11011 (sibcall_insn_operand): Likewise.
11012 (GOT32_symbol_operand): Likewise.
11014 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11016 PR middle-end/83837
11017 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11018 type rather than type addr's type points to.
11019 (expand_omp_atomic_mutex): Likewise.
11020 (expand_omp_atomic): Likewise.
11022 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11025 * config/i386/i386.c (output_indirect_thunk_function): Use
11026 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11027 for __x86_return_thunk.
11029 2018-01-15 Richard Biener <rguenther@suse.de>
11031 PR middle-end/83850
11032 * expmed.c (extract_bit_field_1): Fix typo.
11034 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11037 * config/arm/iterators.md (VF): New mode iterator.
11038 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11039 Remove integer-related logic from pattern.
11040 (neon_vabd<mode>_3): Likewise.
11042 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11044 PR middle-end/82694
11045 * common.opt (fstrict-overflow): No longer an alias.
11046 (fwrapv-pointer): New option.
11047 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11048 also for pointer types based on flag_wrapv_pointer.
11049 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11050 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11051 opts->x_flag_wrapv got set.
11052 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11053 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11054 POINTER_TYPE_OVERFLOW_UNDEFINED.
11055 * match.pd: Likewise in address comparison pattern.
11056 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11058 2018-01-15 Richard Biener <rguenther@suse.de>
11061 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11062 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11063 Reset type names to their identifier if their TYPE_DECL doesn't
11064 have linkage (and thus is used for ODR and devirt).
11065 (save_debug_info_for_decl): Remove.
11066 (save_debug_info_for_type): Likewise.
11067 (add_tree_to_fld_list): Adjust.
11068 * tree-pretty-print.c (dump_generic_node): Make dumping of
11069 type names more robust.
11071 2018-01-15 Richard Biener <rguenther@suse.de>
11073 * BASE-VER: Bump to 8.0.1.
11075 2018-01-14 Martin Sebor <msebor@redhat.com>
11078 * builtins.c (check_access): Avoid warning when the no-warning bit
11081 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11083 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11084 * ira-color (allocno_hard_regs_compare): Likewise.
11086 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11089 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11090 Use .pushsection/.popsection.
11092 2018-01-14 Martin Sebor <msebor@redhat.com>
11095 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11097 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11099 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11100 entry from extra_headers.
11101 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11102 extra_headers, make the list bitwise identical to the i?86-*-* one.
11104 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11106 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11107 -mcmodel=large with -mindirect-branch=thunk,
11108 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11109 -mfunction-return=thunk-extern.
11110 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11111 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11112 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11114 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11116 * config/i386/i386.c (print_reg): Print the name of the full
11117 integer register without '%'.
11118 (ix86_print_operand): Handle 'V'.
11119 * doc/extend.texi: Document 'V' modifier.
11121 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11123 * config/i386/constraints.md (Bs): Disallow memory operand for
11124 -mindirect-branch-register.
11126 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11127 (GOT_memory_operand): Likewise.
11128 (call_insn_operand): Likewise.
11129 (sibcall_insn_operand): Likewise.
11130 (GOT32_symbol_operand): Likewise.
11131 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11132 for -mindirect-branch-register.
11133 (tablejump): Likewise.
11134 (*sibcall_memory): Likewise.
11135 (*sibcall_value_memory): Likewise.
11136 Disallow peepholes of indirect call and jump via memory for
11137 -mindirect-branch-register.
11138 (*call_pop): Replace m with Bw.
11139 (*call_value_pop): Likewise.
11140 (*sibcall_pop_memory): Replace m with Bs.
11141 * config/i386/i386.opt (mindirect-branch-register): New option.
11142 * doc/invoke.texi: Document -mindirect-branch-register option.
11144 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11146 * config/i386/i386-protos.h (ix86_output_function_return): New.
11147 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11148 set function_return_type.
11149 (indirect_thunk_name): Add ret_p to indicate thunk for function
11151 (output_indirect_thunk_function): Pass false to
11152 indirect_thunk_name.
11153 (ix86_output_indirect_branch_via_reg): Likewise.
11154 (ix86_output_indirect_branch_via_push): Likewise.
11155 (output_indirect_thunk_function): Create alias for function
11156 return thunk if regno < 0.
11157 (ix86_output_function_return): New function.
11158 (ix86_handle_fndecl_attribute): Handle function_return.
11159 (ix86_attribute_table): Add function_return.
11160 * config/i386/i386.h (machine_function): Add
11161 function_return_type.
11162 * config/i386/i386.md (simple_return_internal): Use
11163 ix86_output_function_return.
11164 (simple_return_internal_long): Likewise.
11165 * config/i386/i386.opt (mfunction-return=): New option.
11166 (indirect_branch): Mention -mfunction-return=.
11167 * doc/extend.texi: Document function_return function attribute.
11168 * doc/invoke.texi: Document -mfunction-return= option.
11170 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11172 * config/i386/i386-opts.h (indirect_branch): New.
11173 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11174 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11175 with local indirect jump when converting indirect call and jump.
11176 (ix86_set_indirect_branch_type): New.
11177 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11178 (indirectlabelno): New.
11179 (indirect_thunk_needed): Likewise.
11180 (indirect_thunk_bnd_needed): Likewise.
11181 (indirect_thunks_used): Likewise.
11182 (indirect_thunks_bnd_used): Likewise.
11183 (INDIRECT_LABEL): Likewise.
11184 (indirect_thunk_name): Likewise.
11185 (output_indirect_thunk): Likewise.
11186 (output_indirect_thunk_function): Likewise.
11187 (ix86_output_indirect_branch_via_reg): Likewise.
11188 (ix86_output_indirect_branch_via_push): Likewise.
11189 (ix86_output_indirect_branch): Likewise.
11190 (ix86_output_indirect_jmp): Likewise.
11191 (ix86_code_end): Call output_indirect_thunk_function if needed.
11192 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11194 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11195 (ix86_attribute_table): Add indirect_branch.
11196 * config/i386/i386.h (machine_function): Add indirect_branch_type
11197 and has_local_indirect_jump.
11198 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11200 (tablejump): Likewise.
11201 (*indirect_jump): Use ix86_output_indirect_jmp.
11202 (*tablejump_1): Likewise.
11203 (simple_return_indirect_internal): Likewise.
11204 * config/i386/i386.opt (mindirect-branch=): New option.
11205 (indirect_branch): New.
11208 (thunk-inline): Likewise.
11209 (thunk-extern): Likewise.
11210 * doc/extend.texi: Document indirect_branch function attribute.
11211 * doc/invoke.texi: Document -mindirect-branch= option.
11213 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11216 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11218 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11220 * ipa-inline.c (want_inline_small_function_p): Return false if
11221 inlining has already failed with CIF_FINAL_ERROR.
11222 (update_caller_keys): Call want_inline_small_function_p before
11224 (update_callee_keys): Likewise.
11226 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11228 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11230 (rs6000_quadword_masked_address_p): Likewise.
11231 (quad_aligned_load_p): Likewise.
11232 (quad_aligned_store_p): Likewise.
11233 (const_load_sequence_p): Add comment to describe the outer-most loop.
11234 (mimic_memory_attributes_and_flags): New function.
11235 (rs6000_gen_stvx): Likewise.
11236 (replace_swapped_aligned_store): Likewise.
11237 (rs6000_gen_lvx): Likewise.
11238 (replace_swapped_aligned_load): Likewise.
11239 (replace_swapped_load_constant): Capitalize argument name in
11240 comment describing this function.
11241 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11242 and stores that access quad-word aligned addresses and replace
11243 with stvx or lvx instructions when appropriate.
11244 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11245 New function prototype.
11246 (rs6000_quadword_masked_address_p): Likewise.
11247 (rs6000_gen_lvx): Likewise.
11248 (rs6000_gen_stvx): Likewise.
11249 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11250 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11251 when memory address is aligned.
11252 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11253 this split to select lvx instruction when memory address is aligned.
11254 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11255 instruction when memory address is aligned.
11256 (*vsx_le_perm_load_v16qi): Likewise.
11257 (four unnamed splitters): Modify to select the stvx instruction
11258 when memory is aligned.
11260 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11262 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11263 which appears in the queue multiple times.
11265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11266 Alan Hayward <alan.hayward@arm.com>
11267 David Sherwood <david.sherwood@arm.com>
11269 * tree-vectorizer.h (vec_lower_bound): New structure.
11270 (_loop_vec_info): Add check_nonzero and lower_bounds.
11271 (LOOP_VINFO_CHECK_NONZERO): New macro.
11272 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11273 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11274 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11275 fields. Make seg_len the distance travelled, not including the
11277 (dr_direction_indicator): Declare.
11278 (dr_zero_step_indicator): Likewise.
11279 (dr_known_forward_stride_p): Likewise.
11280 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11282 (runtime_alias_check_p): Allow runtime alias checks with
11284 (operator ==): Compare access_size and align.
11285 (prune_runtime_alias_test_list): Rework for new distinction between
11286 the access_size and seg_len.
11287 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11289 (get_segment_min_max): New function.
11290 (create_intersect_range_checks): Use it.
11291 (dr_step_indicator): New function.
11292 (dr_direction_indicator): Likewise.
11293 (dr_zero_step_indicator): Likewise.
11294 (dr_known_forward_stride_p): Likewise.
11295 * tree-loop-distribution.c (data_ref_segment_size): Return
11296 DR_STEP * (niters - 1).
11297 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11299 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11300 (vect_preserves_scalar_order_p): New function, split out from...
11301 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11302 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11303 (vect_vfa_access_size): New function.
11304 (vect_vfa_align): Likewise.
11305 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11306 (dump_lower_bound): New function.
11307 (vect_check_lower_bound): Likewise.
11308 (vect_small_gap_p): Likewise.
11309 (vectorizable_with_step_bound_p): Likewise.
11310 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11311 depencies if the vectorization factor is 1. Convert the checks
11312 for nonzero steps into checks on the bounds of DR_STEP. Try using
11313 a bunds check for variable steps if the minimum required step is
11314 relatively small. Update calls to the dr_with_seg_len
11315 constructor and to vect_compile_time_alias.
11316 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11318 (vect_loop_versioning): Call it.
11319 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11321 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11324 Alan Hayward <alan.hayward@arm.com>
11325 David Sherwood <david.sherwood@arm.com>
11327 * doc/sourcebuild.texi (vect_scatter_store): Document.
11328 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11330 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11332 * genopinit.c (main): Add supports_vec_scatter_store and
11333 supports_vec_scatter_store_cached to target_optabs.
11334 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11335 IFN_MASK_SCATTER_STORE.
11336 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11338 * internal-fn.h (internal_store_fn_p): Declare.
11339 (internal_fn_stored_value_index): Likewise.
11340 * internal-fn.c (scatter_store_direct): New macro.
11341 (expand_scatter_store_optab_fn): New function.
11342 (direct_scatter_store_optab_supported_p): New macro.
11343 (internal_store_fn_p): New function.
11344 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11345 IFN_MASK_SCATTER_STORE.
11346 (internal_fn_mask_index): Likewise.
11347 (internal_fn_stored_value_index): New function.
11348 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11349 for scatter stores.
11350 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11351 * optabs-query.c (supports_vec_scatter_store_p): New function.
11352 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11353 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11354 true for scatter stores.
11355 (vect_gather_scatter_fn_p): Handle scatter stores too.
11356 (vect_check_gather_scatter): Consider using scatter stores if
11357 supports_vec_scatter_store_p.
11358 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11359 scatter stores too.
11360 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11361 internal_fn_stored_value_index.
11362 (check_load_store_masking): Handle scatter stores too.
11363 (vect_get_store_rhs): Make public.
11364 (vectorizable_call): Use internal_store_fn_p.
11365 (vectorizable_store): Handle scatter store internal functions.
11366 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11367 when deciding whether the end of the group has been reached.
11368 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11369 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11370 (mask_scatter_store<mode>): New insns.
11372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11373 Alan Hayward <alan.hayward@arm.com>
11374 David Sherwood <david.sherwood@arm.com>
11376 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11377 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11378 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11380 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11381 Use vect_truncate_gather_scatter_offset if we can't treat the
11382 operation as a normal gather load or scatter store.
11383 (get_group_load_store_type): Take the gather_scatter_info
11384 as argument. Try using a gather load or scatter store for
11385 single-element groups.
11386 (get_load_store_type): Update calls to get_group_load_store_type
11387 and vect_use_strided_gather_scatters_p.
11389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11390 Alan Hayward <alan.hayward@arm.com>
11391 David Sherwood <david.sherwood@arm.com>
11393 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11394 optional tree argument.
11395 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11397 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11398 but continue to use the current value as a fallback.
11399 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11400 to compare the updates.
11401 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11402 (get_load_store_type): Use it when handling a strided access.
11403 (vect_get_strided_load_store_ops): New function.
11404 (vect_get_data_ptr_increment): Likewise.
11405 (vectorizable_load): Handle strided gather loads. Always pass
11406 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11409 Alan Hayward <alan.hayward@arm.com>
11410 David Sherwood <david.sherwood@arm.com>
11412 * doc/md.texi (gather_load@var{m}): Document.
11413 (mask_gather_load@var{m}): Likewise.
11414 * genopinit.c (main): Add supports_vec_gather_load and
11415 supports_vec_gather_load_cached to target_optabs.
11416 * optabs-tree.c (init_tree_optimization_optabs): Use
11417 ggc_cleared_alloc to allocate target_optabs.
11418 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11419 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11421 * internal-fn.h (internal_load_fn_p): Declare.
11422 (internal_gather_scatter_fn_p): Likewise.
11423 (internal_fn_mask_index): Likewise.
11424 (internal_gather_scatter_fn_supported_p): Likewise.
11425 * internal-fn.c (gather_load_direct): New macro.
11426 (expand_gather_load_optab_fn): New function.
11427 (direct_gather_load_optab_supported_p): New macro.
11428 (direct_internal_fn_optab): New function.
11429 (internal_load_fn_p): Likewise.
11430 (internal_gather_scatter_fn_p): Likewise.
11431 (internal_fn_mask_index): Likewise.
11432 (internal_gather_scatter_fn_supported_p): Likewise.
11433 * optabs-query.c (supports_at_least_one_mode_p): New function.
11434 (supports_vec_gather_load_p): Likewise.
11435 * optabs-query.h (supports_vec_gather_load_p): Declare.
11436 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11437 and memory_type field.
11438 (NUM_PATTERNS): Bump to 15.
11439 * tree-vect-data-refs.c: Include internal-fn.h.
11440 (vect_gather_scatter_fn_p): New function.
11441 (vect_describe_gather_scatter_call): Likewise.
11442 (vect_check_gather_scatter): Try using internal functions for
11443 gather loads. Recognize existing calls to a gather load function.
11444 (vect_analyze_data_refs): Consider using gather loads if
11445 supports_vec_gather_load_p.
11446 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11447 (vect_get_gather_scatter_offset_type): Likewise.
11448 (vect_convert_mask_for_vectype): Likewise.
11449 (vect_add_conversion_to_patterm): Likewise.
11450 (vect_try_gather_scatter_pattern): Likewise.
11451 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11452 (vect_vect_recog_func_ptrs): Add it.
11453 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11454 internal_fn_mask_index and internal_gather_scatter_fn_p.
11455 (check_load_store_masking): Take the gather_scatter_info as an
11456 argument and handle gather loads.
11457 (vect_get_gather_scatter_ops): New function.
11458 (vectorizable_call): Check internal_load_fn_p.
11459 (vectorizable_load): Likewise. Handle gather load internal
11461 (vectorizable_store): Update call to check_load_store_masking.
11462 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11463 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11464 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11465 (aarch64_gather_scale_operand_d): New predicates.
11466 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11467 (mask_gather_load<mode>): New insns.
11469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11470 Alan Hayward <alan.hayward@arm.com>
11471 David Sherwood <david.sherwood@arm.com>
11473 * optabs.def (fold_left_plus_optab): New optab.
11474 * doc/md.texi (fold_left_plus_@var{m}): Document.
11475 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11476 * internal-fn.c (fold_left_direct): Define.
11477 (expand_fold_left_optab_fn): Likewise.
11478 (direct_fold_left_optab_supported_p): Likewise.
11479 * fold-const-call.c (fold_const_fold_left): New function.
11480 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11481 * tree-parloops.c (valid_reduction_p): New function.
11482 (gather_scalar_reductions): Use it.
11483 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11484 (vect_finish_replace_stmt): Declare.
11485 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11486 (needs_fold_left_reduction_p): New function, split out from...
11487 (vect_is_simple_reduction): ...here. Accept reductions that
11488 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11489 (vect_force_simple_reduction): Also store the reduction type in
11490 the assignment's STMT_VINFO_REDUC_TYPE.
11491 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11492 (merge_with_identity): New function.
11493 (vect_expand_fold_left): Likewise.
11494 (vectorize_fold_left_reduction): Likewise.
11495 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11496 scalar phi in place for it. Check for target support and reject
11497 cases that would reassociate the operation. Defer the transform
11498 phase to vectorize_fold_left_reduction.
11499 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11500 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11501 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11505 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11506 call to ifc_temp_var.
11508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11509 Alan Hayward <alan.hayward@arm.com>
11510 David Sherwood <david.sherwood@arm.com>
11512 * target.def (legitimize_address_displacement): Take the original
11513 offset as a poly_int.
11514 * targhooks.h (default_legitimize_address_displacement): Update
11516 * targhooks.c (default_legitimize_address_displacement): Likewise.
11517 * doc/tm.texi: Regenerate.
11518 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11519 as an argument, moving assert of ad->disp == ad->disp_term to...
11520 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11521 Try calling targetm.legitimize_address_displacement before expanding
11522 the address rather than afterwards, and adjust for the new interface.
11523 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11524 Match the new hook interface. Handle SVE addresses.
11525 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11526 new hook interface.
11528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11530 * Makefile.in (OBJS): Add early-remat.o.
11531 * target.def (select_early_remat_modes): New hook.
11532 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11533 * doc/tm.texi: Regenerate.
11534 * targhooks.h (default_select_early_remat_modes): Declare.
11535 * targhooks.c (default_select_early_remat_modes): New function.
11536 * timevar.def (TV_EARLY_REMAT): New timevar.
11537 * passes.def (pass_early_remat): New pass.
11538 * tree-pass.h (make_pass_early_remat): Declare.
11539 * early-remat.c: New file.
11540 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11542 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11545 Alan Hayward <alan.hayward@arm.com>
11546 David Sherwood <david.sherwood@arm.com>
11548 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11549 vfm1 with a bound_epilog parameter.
11550 (vect_do_peeling): Update calls accordingly, and move the prologue
11551 call earlier in the function. Treat the base bound_epilog as 0 for
11552 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11553 this base when peeling for gaps.
11554 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11555 with fully-masked loops.
11556 (vect_estimate_min_profitable_iters): Handle the single peeled
11557 iteration in that case.
11559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11560 Alan Hayward <alan.hayward@arm.com>
11561 David Sherwood <david.sherwood@arm.com>
11563 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11564 single-element interleaving even if the size is not a power of 2.
11565 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11566 accesses for single-element interleaving if the group size is
11569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11570 Alan Hayward <alan.hayward@arm.com>
11571 David Sherwood <david.sherwood@arm.com>
11573 * doc/md.texi (fold_extract_last_@var{m}): Document.
11574 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11575 * optabs.def (fold_extract_last_optab): New optab.
11576 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11577 * internal-fn.c (fold_extract_direct): New macro.
11578 (expand_fold_extract_optab_fn): Likewise.
11579 (direct_fold_extract_optab_supported_p): Likewise.
11580 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11581 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11582 EXTRACT_LAST_REDUCTION.
11583 (get_initial_def_for_reduction): Do not create an initial vector
11584 for EXTRACT_LAST_REDUCTION reductions.
11585 (vectorizable_reduction): Leave the scalar phi in place for
11586 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11587 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11588 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11589 transform phase to vectorizable_condition.
11590 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11592 (vect_finish_stmt_generation): ...here.
11593 (vect_finish_replace_stmt): New function.
11594 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11595 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11597 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11600 Alan Hayward <alan.hayward@arm.com>
11601 David Sherwood <david.sherwood@arm.com>
11603 * doc/md.texi (extract_last_@var{m}): Document.
11604 * optabs.def (extract_last_optab): New optab.
11605 * internal-fn.def (EXTRACT_LAST): New internal function.
11606 * internal-fn.c (cond_unary_direct): New macro.
11607 (expand_cond_unary_optab_fn): Likewise.
11608 (direct_cond_unary_optab_supported_p): Likewise.
11609 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11610 loops using EXTRACT_LAST.
11611 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11612 (extract_last_<mode>): ...this optab.
11613 (vec_extract<mode><Vel>): Update accordingly.
11615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11616 Alan Hayward <alan.hayward@arm.com>
11617 David Sherwood <david.sherwood@arm.com>
11619 * target.def (empty_mask_is_expensive): New hook.
11620 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11621 * doc/tm.texi: Regenerate.
11622 * targhooks.h (default_empty_mask_is_expensive): Declare.
11623 * targhooks.c (default_empty_mask_is_expensive): New function.
11624 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11625 if the target says that empty masks are expensive.
11626 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11628 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11631 Alan Hayward <alan.hayward@arm.com>
11632 David Sherwood <david.sherwood@arm.com>
11634 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11635 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11636 (vect_use_loop_mask_for_alignment_p): New function.
11637 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11638 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11639 niters_skip argument. Make sure that the first niters_skip elements
11640 of the first iteration are inactive.
11641 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11642 Update call to vect_set_loop_masks_directly.
11643 (get_misalign_in_elems): New function, split out from...
11644 (vect_gen_prolog_loop_niters): ...here.
11645 (vect_update_init_of_dr): Take a code argument that specifies whether
11646 the adjustment should be added or subtracted.
11647 (vect_update_init_of_drs): Likewise.
11648 (vect_prepare_for_masked_peels): New function.
11649 (vect_do_peeling): Skip prologue peeling if we're using a mask
11650 instead. Update call to vect_update_inits_of_drs.
11651 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11653 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11654 alignment. Do not include the number of peeled iterations in
11655 the minimum threshold in that case.
11656 (vectorizable_induction): Adjust the start value down by
11657 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11658 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11659 Take the number of skipped iterations into account when calculating
11661 * tree-vect-stmts.c (vect_gen_while_not): New function.
11663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11664 Alan Hayward <alan.hayward@arm.com>
11665 David Sherwood <david.sherwood@arm.com>
11667 * doc/sourcebuild.texi (vect_fully_masked): Document.
11668 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11669 default value to 0.
11670 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11672 (vect_analyze_loop_2): ...here. Don't check the vectorization
11673 factor against the number of loop iterations if the loop is
11676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11677 Alan Hayward <alan.hayward@arm.com>
11678 David Sherwood <david.sherwood@arm.com>
11680 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11681 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11682 (dump_groups): Update accordingly.
11683 (iv_use::mem_type): New member variable.
11684 (address_p): New function.
11685 (record_use): Add a mem_type argument and initialize the new
11687 (record_group_use): Add a mem_type argument. Use address_p.
11688 Remove obsolete null checks of base_object. Update call to record_use.
11689 (find_interesting_uses_op): Update call to record_group_use.
11690 (find_interesting_uses_cond): Likewise.
11691 (find_interesting_uses_address): Likewise.
11692 (get_mem_type_for_internal_fn): New function.
11693 (find_address_like_use): Likewise.
11694 (find_interesting_uses_stmt): Try find_address_like_use before
11695 calling find_interesting_uses_op.
11696 (addr_offset_valid_p): Use the iv mem_type field as the type
11697 of the addressed memory.
11698 (add_autoinc_candidates): Likewise.
11699 (get_address_cost): Likewise.
11700 (split_small_address_groups_p): Use address_p.
11701 (split_address_groups): Likewise.
11702 (add_iv_candidate_for_use): Likewise.
11703 (autoinc_possible_for_pair): Likewise.
11704 (rewrite_groups): Likewise.
11705 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11706 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11707 (get_alias_ptr_type_for_ptr_address): New function.
11708 (rewrite_use_address): Rewrite address uses in calls that were
11709 identified by find_address_like_use.
11711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11712 Alan Hayward <alan.hayward@arm.com>
11713 David Sherwood <david.sherwood@arm.com>
11715 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11717 * gimple-expr.h (is_gimple_addressable: Likewise.
11718 * gimple-expr.c (is_gimple_address): Likewise.
11719 * internal-fn.c (expand_call_mem_ref): New function.
11720 (expand_mask_load_optab_fn): Use it.
11721 (expand_mask_store_optab_fn): Likewise.
11723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11724 Alan Hayward <alan.hayward@arm.com>
11725 David Sherwood <david.sherwood@arm.com>
11727 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11728 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11729 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11730 (cond_umax@var{mode}): Document.
11731 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11732 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11733 (cond_umin_optab, cond_umax_optab): New optabs.
11734 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11735 (COND_IOR, COND_XOR): New internal functions.
11736 * internal-fn.h (get_conditional_internal_fn): Declare.
11737 * internal-fn.c (cond_binary_direct): New macro.
11738 (expand_cond_binary_optab_fn): Likewise.
11739 (direct_cond_binary_optab_supported_p): Likewise.
11740 (get_conditional_internal_fn): New function.
11741 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11742 Cope with reduction statements that are vectorized as calls rather
11744 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11745 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11746 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11747 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11748 (UNSPEC_COND_EOR): New unspecs.
11749 (optab): Add mappings for them.
11750 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11751 (sve_int_op, sve_fp_op): New int attributes.
11753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11754 Alan Hayward <alan.hayward@arm.com>
11755 David Sherwood <david.sherwood@arm.com>
11757 * optabs.def (while_ult_optab): New optab.
11758 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11759 * internal-fn.def (WHILE_ULT): New internal function.
11760 * internal-fn.h (direct_internal_fn_supported_p): New override
11761 that takes two types as argument.
11762 * internal-fn.c (while_direct): New macro.
11763 (expand_while_optab_fn): New function.
11764 (convert_optab_supported_p): Likewise.
11765 (direct_while_optab_supported_p): New macro.
11766 * wide-int.h (wi::udiv_ceil): New function.
11767 * tree-vectorizer.h (rgroup_masks): New structure.
11768 (vec_loop_masks): New typedef.
11769 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11770 and fully_masked_p.
11771 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11772 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11773 (vect_max_vf): New function.
11774 (slpeel_make_loop_iterate_ntimes): Delete.
11775 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11776 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11777 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11778 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11779 internal-fn.h, stor-layout.h and optabs-query.h.
11780 (vect_set_loop_mask): New function.
11781 (add_preheader_seq): Likewise.
11782 (add_header_seq): Likewise.
11783 (interleave_supported_p): Likewise.
11784 (vect_maybe_permute_loop_masks): Likewise.
11785 (vect_set_loop_masks_directly): Likewise.
11786 (vect_set_loop_condition_masked): Likewise.
11787 (vect_set_loop_condition_unmasked): New function, split out from
11788 slpeel_make_loop_iterate_ntimes.
11789 (slpeel_make_loop_iterate_ntimes): Rename to..
11790 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11791 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11792 (vect_do_peeling): Update call accordingly.
11793 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11795 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11796 mask_compare_type, can_fully_mask_p and fully_masked_p.
11797 (release_vec_loop_masks): New function.
11798 (_loop_vec_info): Use it to free the loop masks.
11799 (can_produce_all_loop_masks_p): New function.
11800 (vect_get_max_nscalars_per_iter): Likewise.
11801 (vect_verify_full_masking): Likewise.
11802 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11803 retries, and free the mask rgroups before retrying. Check loop-wide
11804 reasons for disallowing fully-masked loops. Make the final decision
11805 about whether use a fully-masked loop or not.
11806 (vect_estimate_min_profitable_iters): Do not assume that peeling
11807 for the number of iterations will be needed for fully-masked loops.
11808 (vectorizable_reduction): Disable fully-masked loops.
11809 (vectorizable_live_operation): Likewise.
11810 (vect_halve_mask_nunits): New function.
11811 (vect_double_mask_nunits): Likewise.
11812 (vect_record_loop_mask): Likewise.
11813 (vect_get_loop_mask): Likewise.
11814 (vect_transform_loop): Handle the case in which the final loop
11815 iteration might handle a partial vector. Call vect_set_loop_condition
11816 instead of slpeel_make_loop_iterate_ntimes.
11817 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11818 (check_load_store_masking): New function.
11819 (prepare_load_store_mask): Likewise.
11820 (vectorizable_store): Handle fully-masked loops.
11821 (vectorizable_load): Likewise.
11822 (supportable_widening_operation): Use vect_halve_mask_nunits for
11824 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11825 (vect_gen_while): New function.
11826 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11827 (aarch64_uqdec<mode>): New insn.
11829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11830 Alan Hayward <alan.hayward@arm.com>
11831 David Sherwood <david.sherwood@arm.com>
11833 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11834 (reduc_xor_scal_optab): New optabs.
11835 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11836 (reduc_xor_scal_@var{m}): Document.
11837 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11838 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11839 internal functions.
11840 * fold-const-call.c (fold_const_call): Handle them.
11841 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11842 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11843 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11844 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11845 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11846 (UNSPEC_XORV): New unspecs.
11847 (optab): Add entries for them.
11848 (BITWISEV): New int iterator.
11849 (bit_reduc_op): New int attributes.
11851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11852 Alan Hayward <alan.hayward@arm.com>
11853 David Sherwood <david.sherwood@arm.com>
11855 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11856 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11857 * optabs.def (vec_shl_insert_optab): New optab.
11858 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11859 (duplicate_and_interleave): Likewise.
11860 * tree-vect-loop.c: Include internal-fn.h.
11861 (neutral_op_for_slp_reduction): New function, split out from
11862 get_initial_defs_for_reduction.
11863 (get_initial_def_for_reduction): Handle option 2 for variable-length
11864 vectors by loading the neutral value into a vector and then shifting
11865 the initial value into element 0.
11866 (get_initial_defs_for_reduction): Replace the code argument with
11867 the neutral value calculated by neutral_op_for_slp_reduction.
11868 Use gimple_build_vector for constant-length vectors.
11869 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11870 but the first group_size elements have a neutral value.
11871 Use duplicate_and_interleave otherwise.
11872 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11873 Update call to get_initial_defs_for_reduction. Handle SLP
11874 reductions for variable-length vectors by creating one vector
11875 result for each scalar result, with the elements associated
11876 with other scalar results stubbed out with the neutral value.
11877 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11878 Require IFN_VEC_SHL_INSERT for double reductions on
11879 variable-length vectors, or SLP reductions that have
11880 a neutral value. Require can_duplicate_and_interleave_p
11881 support for variable-length unchained SLP reductions if there
11882 is no neutral value, such as for MIN/MAX reductions. Also require
11883 the number of vector elements to be a multiple of the number of
11884 SLP statements when doing variable-length unchained SLP reductions.
11885 Update call to vect_create_epilog_for_reduction.
11886 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11887 and remove initial values.
11888 (duplicate_and_interleave): Make public.
11889 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11890 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11893 Alan Hayward <alan.hayward@arm.com>
11894 David Sherwood <david.sherwood@arm.com>
11896 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11897 (can_duplicate_and_interleave_p): New function.
11898 (vect_get_and_check_slp_defs): Take the vector of statements
11899 rather than just the current one. Remove excess parentheses.
11900 Restriction rejectinon of vect_constant_def and vect_external_def
11901 for variable-length vectors to boolean types, or types for which
11902 can_duplicate_and_interleave_p is false.
11903 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11904 (duplicate_and_interleave): New function.
11905 (vect_get_constant_vectors): Use gimple_build_vector for
11906 constant-length vectors and suitable variable-length constant
11907 vectors. Use duplicate_and_interleave for other variable-length
11908 vectors. Don't defer the update when inserting new statements.
11910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11911 Alan Hayward <alan.hayward@arm.com>
11912 David Sherwood <david.sherwood@arm.com>
11914 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11915 min_profitable_iters doesn't go negative.
11917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11922 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11923 * optabs.def (vec_mask_load_lanes_optab): New optab.
11924 (vec_mask_store_lanes_optab): Likewise.
11925 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11926 (MASK_STORE_LANES): Likewise.
11927 * internal-fn.c (mask_load_lanes_direct): New macro.
11928 (mask_store_lanes_direct): Likewise.
11929 (expand_mask_load_optab_fn): Handle masked operations.
11930 (expand_mask_load_lanes_optab_fn): New macro.
11931 (expand_mask_store_optab_fn): Handle masked operations.
11932 (expand_mask_store_lanes_optab_fn): New macro.
11933 (direct_mask_load_lanes_optab_supported_p): Likewise.
11934 (direct_mask_store_lanes_optab_supported_p): Likewise.
11935 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11937 (vect_load_lanes_supported): Likewise.
11938 * tree-vect-data-refs.c (strip_conversion): New function.
11939 (can_group_stmts_p): Likewise.
11940 (vect_analyze_data_ref_accesses): Use it instead of checking
11941 for a pair of assignments.
11942 (vect_store_lanes_supported): Take a masked_p parameter.
11943 (vect_load_lanes_supported): Likewise.
11944 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11945 vect_store_lanes_supported and vect_load_lanes_supported.
11946 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11947 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11948 parameter. Don't allow gaps for masked accesses.
11949 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11950 and vect_load_lanes_supported.
11951 (get_load_store_type): Take a masked_p parameter and update
11952 call to get_group_load_store_type.
11953 (vectorizable_store): Update call to get_load_store_type.
11954 Handle IFN_MASK_STORE_LANES.
11955 (vectorizable_load): Update call to get_load_store_type.
11956 Handle IFN_MASK_LOAD_LANES.
11958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11959 Alan Hayward <alan.hayward@arm.com>
11960 David Sherwood <david.sherwood@arm.com>
11962 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11964 * config/aarch64/aarch64-protos.h
11965 (aarch64_sve_struct_memory_operand_p): Declare.
11966 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11967 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11968 (VPRED, vpred): Handle SVE structure modes.
11969 * config/aarch64/constraints.md (Utx): New constraint.
11970 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11971 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11972 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11973 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11974 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11975 structure modes. Split into pieces after RA.
11976 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11977 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11979 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11980 SVE structure modes.
11981 (aarch64_classify_address): Likewise.
11982 (sizetochar): Move earlier in file.
11983 (aarch64_print_operand): Handle SVE register lists.
11984 (aarch64_array_mode): New function.
11985 (aarch64_sve_struct_memory_operand_p): Likewise.
11986 (TARGET_ARRAY_MODE): Redefine.
11988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11989 Alan Hayward <alan.hayward@arm.com>
11990 David Sherwood <david.sherwood@arm.com>
11992 * target.def (array_mode): New target hook.
11993 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11994 * doc/tm.texi: Regenerate.
11995 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11996 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11997 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11998 targetm.array_mode.
11999 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12003 Alan Hayward <alan.hayward@arm.com>
12004 David Sherwood <david.sherwood@arm.com>
12006 * fold-const.c (fold_binary_loc): Check the argument types
12007 rather than the result type when testing for a vector operation.
12009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12011 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12012 * doc/tm.texi: Regenerate.
12014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12015 Alan Hayward <alan.hayward@arm.com>
12016 David Sherwood <david.sherwood@arm.com>
12018 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12019 (sve): Document new AArch64 extension.
12020 * doc/md.texi (w): Extend the description of the AArch64
12021 constraint to include SVE vectors.
12022 (Upl, Upa): Document new AArch64 predicate constraints.
12023 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12025 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12026 (msve-vector-bits=): New option.
12027 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12028 SVE when these are disabled.
12029 (sve): New extension.
12030 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12031 modes. Adjust their number of units based on aarch64_sve_vg.
12032 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12033 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12034 aarch64_addr_query_type.
12035 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12036 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12037 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12038 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12039 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12040 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12041 (aarch64_simd_imm_zero_p): Delete.
12042 (aarch64_check_zero_based_sve_index_immediate): Declare.
12043 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12044 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12045 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12046 (aarch64_sve_float_mul_immediate_p): Likewise.
12047 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12048 rather than an rtx.
12049 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12050 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12051 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12052 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12053 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12054 (aarch64_regmode_natural_size): Likewise.
12055 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12056 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12058 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12059 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12060 for VG and the SVE predicate registers.
12061 (V_ALIASES): Add a "z"-prefixed alias.
12062 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12063 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12064 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12065 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12066 (REG_CLASS_NAMES): Add entries for them.
12067 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12068 and the predicate registers.
12069 (aarch64_sve_vg): Declare.
12070 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12071 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12072 (REGMODE_NATURAL_SIZE): Define.
12073 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12075 * config/aarch64/aarch64.c: Include cfgrtl.h.
12076 (simd_immediate_info): Add a constructor for series vectors,
12077 and an associated step field.
12078 (aarch64_sve_vg): New variable.
12079 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12080 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12081 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12082 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12083 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12084 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12085 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12086 (aarch64_get_mask_mode): New functions.
12087 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12088 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12089 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12090 predicate modes and predicate registers. Explicitly restrict
12091 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12092 to store a vector mode if it is recognized by
12093 aarch64_classify_vector_mode.
12094 (aarch64_regmode_natural_size): New function.
12095 (aarch64_hard_regno_caller_save_mode): Return the original mode
12097 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12098 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12099 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12100 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12102 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12103 does not overlap dest if the function is frame-related. Handle
12105 (aarch64_split_add_offset): New function.
12106 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12107 them aarch64_add_offset.
12108 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12109 and update call to aarch64_sub_sp.
12110 (aarch64_add_cfa_expression): New function.
12111 (aarch64_expand_prologue): Pass extra temporary registers to the
12112 functions above. Handle the case in which we need to emit new
12113 DW_CFA_expressions for registers that were originally saved
12114 relative to the stack pointer, but now have to be expressed
12115 relative to the frame pointer.
12116 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12118 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12119 IP0 and IP1 values for SVE frames.
12120 (aarch64_expand_vec_series): New function.
12121 (aarch64_expand_sve_widened_duplicate): Likewise.
12122 (aarch64_expand_sve_const_vector): Likewise.
12123 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12124 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12125 into the register, rather than emitting a SET directly.
12126 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12127 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12128 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12129 (offset_9bit_signed_scaled_p): New functions.
12130 (aarch64_replicate_bitmask_imm): New function.
12131 (aarch64_bitmask_imm): Use it.
12132 (aarch64_cannot_force_const_mem): Reject expressions involving
12133 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12134 (aarch64_classify_index): Handle SVE indices, by requiring
12135 a plain register index with a scale that matches the element size.
12136 (aarch64_classify_address): Handle SVE addresses. Assert that
12137 the mode of the address is VOIDmode or an integer mode.
12138 Update call to aarch64_classify_symbol.
12139 (aarch64_classify_symbolic_expression): Update call to
12140 aarch64_classify_symbol.
12141 (aarch64_const_vec_all_in_range_p): New function.
12142 (aarch64_print_vector_float_operand): Likewise.
12143 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12144 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12145 and the FP immediates 1.0 and 0.5.
12146 (aarch64_print_address_internal): Handle SVE addresses.
12147 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12148 (aarch64_regno_regclass): Handle predicate registers.
12149 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12151 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12152 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12153 (aarch64_convert_sve_vector_bits): New function.
12154 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12155 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12156 rather than an rtx.
12157 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12158 Handle SVE vector and predicate modes. Accept VL-based constants
12159 that need only one temporary register, and VL offsets that require
12160 no temporary registers.
12161 (aarch64_conditional_register_usage): Mark the predicate registers
12162 as fixed if SVE isn't available.
12163 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12164 Return true for SVE vector and predicate modes.
12165 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12166 rather than an unsigned int. Handle SVE modes.
12167 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12169 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12171 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12172 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12173 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12174 (aarch64_sve_float_mul_immediate_p): New functions.
12175 (aarch64_sve_valid_immediate): New function.
12176 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12177 Explicitly reject structure modes. Check for INDEX constants.
12178 Handle PTRUE and PFALSE constants.
12179 (aarch64_check_zero_based_sve_index_immediate): New function.
12180 (aarch64_simd_imm_zero_p): Delete.
12181 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12182 vector modes. Accept constants in the range of CNT[BHWD].
12183 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12184 ask for an Advanced SIMD mode.
12185 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12186 (aarch64_simd_vector_alignment): Handle SVE predicates.
12187 (aarch64_vectorize_preferred_vector_alignment): New function.
12188 (aarch64_simd_vector_alignment_reachable): Use it instead of
12190 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12191 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12193 (MAX_VECT_LEN): Delete.
12194 (expand_vec_perm_d): Add a vec_flags field.
12195 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12196 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12197 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12199 (aarch64_evpc_rev): Rename to...
12200 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12201 (aarch64_evpc_rev_global): New function.
12202 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12203 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12205 (aarch64_evpc_sve_tbl): New function.
12206 (aarch64_expand_vec_perm_const_1): Update after rename of
12207 aarch64_evpc_rev. Handle SVE permutes too, trying
12208 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12209 than aarch64_evpc_tbl.
12210 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12211 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12212 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12213 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12214 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12215 (aarch64_expand_sve_vcond): New functions.
12216 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12217 of aarch64_vector_mode_p.
12218 (aarch64_dwarf_poly_indeterminate_value): New function.
12219 (aarch64_compute_pressure_classes): Likewise.
12220 (aarch64_can_change_mode_class): Likewise.
12221 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12222 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12223 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12224 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12225 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12226 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12227 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12228 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12230 (Dn, Dl, Dr): Accept const as well as const_vector.
12231 (Dz): Likewise. Compare against CONST0_RTX.
12232 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12233 of "vector" where appropriate.
12234 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12235 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12236 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12237 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12238 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12239 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12240 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12241 (v_int_equiv): Extend to SVE modes.
12242 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12244 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12245 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12246 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12247 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12248 (SVE_COND_FP_CMP): New int iterators.
12249 (perm_hilo): Handle the new unpack unspecs.
12250 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12252 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12253 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12254 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12255 (aarch64_equality_operator, aarch64_constant_vector_operand)
12256 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12257 (aarch64_sve_nonimmediate_operand): Likewise.
12258 (aarch64_sve_general_operand): Likewise.
12259 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12260 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12261 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12262 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12263 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12264 (aarch64_sve_float_arith_immediate): Likewise.
12265 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12266 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12267 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12268 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12269 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12270 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12271 (aarch64_sve_float_arith_operand): Likewise.
12272 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12273 (aarch64_sve_float_mul_operand): Likewise.
12274 (aarch64_sve_vec_perm_operand): Likewise.
12275 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12276 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12277 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12278 as well as const_vector.
12279 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12280 in file. Use CONST0_RTX and CONSTM1_RTX.
12281 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12282 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12283 Use aarch64_simd_imm_zero.
12284 * config/aarch64/aarch64-sve.md: New file.
12285 * config/aarch64/aarch64.md: Include it.
12286 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12287 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12288 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12289 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12290 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12291 (sve): New attribute.
12292 (enabled): Disable instructions with the sve attribute unless
12294 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12295 aarch64_expand_mov_immediate.
12296 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12297 CNT[BHSD] immediates.
12298 (movti): Split CONST_POLY_INT moves into two halves.
12299 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12300 Split additions that need a temporary here if the destination
12301 is the stack pointer.
12302 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12303 (*add<mode>3_poly_1): New instruction.
12304 (set_clobber_cc): New expander.
12306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12308 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12309 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12310 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12311 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12312 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12313 Change innermode from fixed_mode_size to machine_mode.
12314 (simplify_subreg): Update call accordingly. Handle a constant-sized
12315 subreg of a variable-length CONST_VECTOR.
12317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12318 Alan Hayward <alan.hayward@arm.com>
12319 David Sherwood <david.sherwood@arm.com>
12321 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12322 (add_offset_to_base): New function, split out from...
12323 (create_mem_ref): ...here. When handling a scale other than 1,
12324 check first whether the address is valid without the offset.
12325 Add it into the base if so, leaving the index and scale as-is.
12327 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12330 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12331 fold_for_warn before checking if arg2 is INTEGER_CST.
12333 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12335 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12336 (store_multiple_operation): Delete.
12337 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12338 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12339 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12340 guarded by TARGET_STRING.
12341 (rs6000_output_load_multiple): Delete.
12342 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12343 OPTION_MASK_STRING / TARGET_STRING handling.
12344 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12345 (const rs6000_opt_masks) <"string">: Change mask to 0.
12346 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12347 (MASK_STRING): Delete.
12348 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12350 (load_multiple): Delete.
12357 (store_multiple): Delete.
12364 (movmemsi_8reg): Delete.
12365 (corresponding unnamed define_insn): Delete.
12366 (movmemsi_6reg): Delete.
12367 (corresponding unnamed define_insn): Delete.
12368 (movmemsi_4reg): Delete.
12369 (corresponding unnamed define_insn): Delete.
12370 (movmemsi_2reg): Delete.
12371 (corresponding unnamed define_insn): Delete.
12372 (movmemsi_1reg): Delete.
12373 (corresponding unnamed define_insn): Delete.
12374 * config/rs6000/rs6000.opt (mno-string): New.
12375 (mstring): Replace by deprecation warning stub.
12376 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12378 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12380 * regrename.c (regrename_do_replace): If replacing the same
12381 reg multiple times, try to reuse last created gen_raw_REG.
12384 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12385 main to workaround a bug in GDB.
12387 2018-01-12 Tom de Vries <tom@codesourcery.com>
12390 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12392 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12394 PR rtl-optimization/80481
12395 * ira-color.c (get_cap_member): New function.
12396 (allocnos_conflict_by_live_ranges_p): Use it.
12397 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12398 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12400 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12403 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12404 (*saddl_se_1): Ditto.
12405 (*ssubsi_1): Ditto.
12406 (*ssubl_se_1): Ditto.
12408 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12410 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12411 rather than wi::to_widest for DR_INITs.
12412 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12413 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12414 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12416 (vect_analyze_group_access_1): Note that here.
12418 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12420 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12421 polynomial type sizes.
12423 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12425 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12426 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12427 (gimple_add_tmp_var): Likewise.
12429 2018-01-12 Martin Liska <mliska@suse.cz>
12431 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12432 (gimple_alloc_sizes): Likewise.
12433 (dump_gimple_statistics): Use PRIu64 in printf format.
12434 * gimple.h: Change uint64_t to int.
12436 2018-01-12 Martin Liska <mliska@suse.cz>
12438 * tree-core.h: Use uint64_t instead of int.
12439 * tree.c (tree_node_counts): Likewise.
12440 (tree_node_sizes): Likewise.
12441 (dump_tree_statistics): Use PRIu64 in printf format.
12443 2018-01-12 Martin Liska <mliska@suse.cz>
12445 * Makefile.in: As qsort_chk is implemented in vec.c, add
12446 vec.o to linkage of gencfn-macros.
12447 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12448 passing the info to record_node_allocation_statistics.
12449 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12451 * ggc-common.c (struct ggc_usage): Add operator== and use
12452 it in operator< and compare function.
12453 * mem-stats.h (struct mem_usage): Likewise.
12454 * vec.c (struct vec_usage): Remove operator< and compare
12455 function. Can be simply inherited.
12457 2018-01-12 Martin Jambor <mjambor@suse.cz>
12460 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12461 * tree-ssa-math-opts.c: Include domwalk.h.
12462 (convert_mult_to_fma_1): New function.
12463 (fma_transformation_info): New type.
12464 (fma_deferring_state): Likewise.
12465 (cancel_fma_deferring): New function.
12466 (result_of_phi): Likewise.
12467 (last_fma_candidate_feeds_initial_phi): Likewise.
12468 (convert_mult_to_fma): Added deferring logic, split actual
12469 transformation to convert_mult_to_fma_1.
12470 (math_opts_dom_walker): New type.
12471 (math_opts_dom_walker::after_dom_children): New method, body moved
12472 here from pass_optimize_widening_mul::execute, added deferring logic
12474 (pass_optimize_widening_mul::execute): Moved most of code to
12475 math_opts_dom_walker::after_dom_children.
12476 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12477 * config/i386/i386.c (ix86_option_override_internal): Added
12478 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12480 2018-01-12 Richard Biener <rguenther@suse.de>
12483 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12484 inline instance vars.
12486 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12489 * config/rx/rx.c (rx_is_restricted_memory_address):
12490 Handle SUBREG case.
12492 2018-01-12 Richard Biener <rguenther@suse.de>
12494 PR tree-optimization/80846
12495 * target.def (split_reduction): New target hook.
12496 * targhooks.c (default_split_reduction): New function.
12497 * targhooks.h (default_split_reduction): Declare.
12498 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12499 target requests first reduce vectors by combining low and high
12501 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12502 (get_vectype_for_scalar_type_and_size): Export.
12503 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12504 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12505 * doc/tm.texi: Regenerate.
12506 * config/i386/i386.c (ix86_split_reduction): Implement
12507 TARGET_VECTORIZE_SPLIT_REDUCTION.
12509 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12512 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12513 in PIC mode except for TARGET_VXWORKS_RTP.
12514 * config/sparc/sparc.c: Include cfgrtl.h.
12515 (TARGET_INIT_PIC_REG): Define.
12516 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12517 (sparc_pic_register_p): New predicate.
12518 (sparc_legitimate_address_p): Use it.
12519 (sparc_legitimize_pic_address): Likewise.
12520 (sparc_delegitimize_address): Likewise.
12521 (sparc_mode_dependent_address_p): Likewise.
12522 (gen_load_pcrel_sym): Remove 4th parameter.
12523 (load_got_register): Adjust call to above. Remove obsolete stuff.
12524 (sparc_expand_prologue): Do not call load_got_register here.
12525 (sparc_flat_expand_prologue): Likewise.
12526 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12527 (sparc_use_pseudo_pic_reg): New function.
12528 (sparc_init_pic_reg): Likewise.
12529 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12530 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12532 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12534 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12535 Add item for branch_cost.
12537 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12539 PR rtl-optimization/83565
12540 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12541 not extend the result to a larger mode for rotate operations.
12542 (num_sign_bit_copies1): Likewise.
12544 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12547 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12549 Use values-Xc.o for -pedantic.
12550 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12552 2018-01-12 Martin Liska <mliska@suse.cz>
12555 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12557 (possible_polymorphic_call_targets): Use it.
12558 (ipa_devirt): Likewise.
12560 2018-01-12 Martin Liska <mliska@suse.cz>
12562 * profile-count.h (enum profile_quality): Use 0 as invalid
12563 enum value of profile_quality.
12565 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12567 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12568 -mext-string options.
12570 2018-01-12 Richard Biener <rguenther@suse.de>
12572 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12573 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12574 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12576 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12578 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12580 * configure.ac (--with-long-double-format): Add support for the
12581 configuration option to change the default long double format on
12583 * config.gcc (powerpc*-linux*-*): Likewise.
12584 * configure: Regenerate.
12585 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12586 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12587 used without modification.
12589 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12591 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12592 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12593 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12594 MISC_BUILTIN_SPEC_BARRIER.
12595 (rs6000_init_builtins): Likewise.
12596 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12598 (speculation_barrier): New define_insn.
12599 * doc/extend.texi: Document __builtin_speculation_barrier.
12601 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12604 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12605 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12606 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12608 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12609 integral modes instead of "ss" and "sd".
12610 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12611 vectors with 32-bit and 64-bit elements.
12612 (vecdupssescalarmodesuffix): New mode attribute.
12613 (vec_dup<mode>): Use it.
12615 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12618 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12619 frame if argument is passed on stack.
12621 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12624 * ree.c (combine_reaching_defs): Optimize also
12625 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12626 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12628 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12630 PR middle-end/83189
12631 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12633 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12635 PR middle-end/83718
12636 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12637 after they are computed.
12639 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12641 PR tree-optimization/83695
12642 * gimple-loop-linterchange.cc
12643 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12644 reset cached scev information after interchange.
12645 (pass_linterchange::execute): Remove call to scev_reset_htab.
12647 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12649 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12650 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12651 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12652 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12653 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12654 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12655 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12656 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12657 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12658 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12659 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12660 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12661 (V_lane_reg): Likewise.
12662 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12664 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12665 (vfmal_lane_low<mode>_intrinsic,
12666 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12667 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12668 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12669 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12670 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12671 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12673 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12675 * config/arm/arm-cpus.in (fp16fml): New feature.
12676 (ALL_SIMD): Add fp16fml.
12677 (armv8.2-a): Add fp16fml as an option.
12678 (armv8.3-a): Likewise.
12679 (armv8.4-a): Add fp16fml as part of fp16.
12680 * config/arm/arm.h (TARGET_FP16FML): Define.
12681 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12683 * config/arm/arm-modes.def (V2HF): Define.
12684 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12685 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12686 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12687 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12688 vfmsl_low, vfmsl_high): New set of builtins.
12689 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12690 (vfml_op): New code attribute.
12691 (VFMLHALVES): New int iterator.
12692 (VFML, VFMLSEL): New mode attributes.
12693 (V_reg): Define mapping for V2HF.
12694 (V_hi, V_lo): New mode attributes.
12695 (VF_constraint): Likewise.
12696 (vfml_half, vfml_half_selector): New int attributes.
12697 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12699 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12700 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12702 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12703 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12704 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12705 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12707 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12708 Document new effective target and option set.
12710 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12712 * config/arm/arm-cpus.in (armv8_4): New feature.
12713 (ARMv8_4a): New fgroup.
12714 (armv8.4-a): New arch.
12715 * config/arm/arm-tables.opt: Regenerate.
12716 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12717 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12718 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12719 Add matching rules for -march=armv8.4-a and extensions.
12720 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12722 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12725 * config/rx/rx.md (BW): New mode attribute.
12726 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12728 2018-01-11 Richard Biener <rguenther@suse.de>
12730 PR tree-optimization/83435
12731 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12732 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12733 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12735 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12741 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12742 (aarch64_print_address_internal): Use it to check for a zero offset.
12744 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12745 Alan Hayward <alan.hayward@arm.com>
12746 David Sherwood <david.sherwood@arm.com>
12748 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12749 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12750 Return a poly_int64 rather than a HOST_WIDE_INT.
12751 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12752 rather than a HOST_WIDE_INT.
12753 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12754 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12755 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12756 final_offset from HOST_WIDE_INT to poly_int64.
12757 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12758 to_constant when getting the number of units in an Advanced SIMD
12760 (aarch64_builtin_vectorized_function): Check for a constant number
12762 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12764 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12765 attribute instead of GET_MODE_NUNITS.
12766 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12767 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12768 GET_MODE_SIZE for fixed-size registers.
12769 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12770 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12771 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12772 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12773 (aarch64_print_operand, aarch64_print_address_internal)
12774 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12775 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12776 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12777 Handle polynomial GET_MODE_SIZE.
12778 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12779 wider than SImode without modification.
12780 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12781 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12782 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12783 passing and returning SVE modes.
12784 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12785 rather than GEN_INT.
12786 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12787 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12788 (aarch64_allocate_and_probe_stack_space): Likewise.
12789 (aarch64_layout_frame): Cope with polynomial offsets.
12790 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12791 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12792 polynomial offsets.
12793 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12794 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12795 poly_int64 rather than a HOST_WIDE_INT.
12796 (aarch64_get_separate_components, aarch64_process_components)
12797 (aarch64_expand_prologue, aarch64_expand_epilogue)
12798 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12799 (aarch64_anchor_offset): New function, split out from...
12800 (aarch64_legitimize_address): ...here.
12801 (aarch64_builtin_vectorization_cost): Handle polynomial
12802 TYPE_VECTOR_SUBPARTS.
12803 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12805 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12806 number of elements from the PARALLEL rather than the mode.
12807 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12808 rather than GET_MODE_BITSIZE.
12809 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12810 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12811 (aarch64_expand_vec_perm_const_1): Handle polynomial
12812 d->perm.length () and d->perm elements.
12813 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12814 Apply to_constant to d->perm elements.
12815 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12816 polynomial CONST_VECTOR_NUNITS.
12817 (aarch64_move_pointer): Take amount as a poly_int64 rather
12819 (aarch64_progress_pointer): Avoid temporary variable.
12820 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12821 the mode attribute instead of GET_MODE.
12823 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12824 Alan Hayward <alan.hayward@arm.com>
12825 David Sherwood <david.sherwood@arm.com>
12827 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12828 x exists before using it.
12829 (aarch64_add_constant_internal): Rename to...
12830 (aarch64_add_offset_1): ...this. Replace regnum with separate
12831 src and dest rtxes. Handle the case in which they're different,
12832 including when the offset is zero. Replace scratchreg with an rtx.
12833 Use 2 additions if there is no spare register into which we can
12834 move a 16-bit constant.
12835 (aarch64_add_constant): Delete.
12836 (aarch64_add_offset): Replace reg with separate src and dest
12837 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12838 Use aarch64_add_offset_1.
12839 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12840 an rtx rather than an int. Take the delta as a poly_int64
12841 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12842 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12843 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12844 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12845 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12846 and aarch64_add_sp.
12847 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12848 aarch64_add_constant.
12850 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12852 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12853 Use scalar_float_mode.
12855 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12857 * config/aarch64/aarch64-simd.md
12858 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12859 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12860 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12861 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12862 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12863 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12864 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12865 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12866 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12867 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12869 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12872 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12873 targ_options->x_arm_arch_string is non NULL.
12875 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12877 * config/aarch64/aarch64.h
12878 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12880 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12883 * expmed.c (emit_store_flag_force): Swap if const op0
12884 and change VOIDmode to mode of op0.
12886 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12888 PR rtl-optimization/83761
12889 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12890 than bytes to mode_for_size.
12892 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12894 PR middle-end/83189
12895 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12896 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12899 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12901 PR middle-end/83575
12902 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12903 when in layout mode.
12904 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12905 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12908 2018-01-10 Michael Collison <michael.collison@arm.com>
12910 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12911 * config/aarch64/aarch64-option-extension.def: Add
12912 AARCH64_OPT_EXTENSION of 'fp16fml'.
12913 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12914 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12915 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12916 * config/aarch64/constraints.md (Ui7): New constraint.
12917 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12918 (VFMLA_SEL_W): Ditto.
12921 (VFMLA16_LOW): New int iterator.
12922 (VFMLA16_HIGH): Ditto.
12923 (UNSPEC_FMLAL): New unspec.
12924 (UNSPEC_FMLSL): Ditto.
12925 (UNSPEC_FMLAL2): Ditto.
12926 (UNSPEC_FMLSL2): Ditto.
12927 (f16mac): New code attribute.
12928 * config/aarch64/aarch64-simd-builtins.def
12929 (aarch64_fmlal_lowv2sf): Ditto.
12930 (aarch64_fmlsl_lowv2sf): Ditto.
12931 (aarch64_fmlalq_lowv4sf): Ditto.
12932 (aarch64_fmlslq_lowv4sf): Ditto.
12933 (aarch64_fmlal_highv2sf): Ditto.
12934 (aarch64_fmlsl_highv2sf): Ditto.
12935 (aarch64_fmlalq_highv4sf): Ditto.
12936 (aarch64_fmlslq_highv4sf): Ditto.
12937 (aarch64_fmlal_lane_lowv2sf): Ditto.
12938 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12939 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12940 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12941 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12942 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12943 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12944 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12945 (aarch64_fmlal_lane_highv2sf): Ditto.
12946 (aarch64_fmlsl_lane_highv2sf): Ditto.
12947 (aarch64_fmlal_laneq_highv2sf): Ditto.
12948 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12949 (aarch64_fmlalq_lane_highv4sf): Ditto.
12950 (aarch64_fmlsl_lane_highv4sf): Ditto.
12951 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12952 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12953 * config/aarch64/aarch64-simd.md:
12954 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12955 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12956 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12957 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12958 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12959 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12960 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12961 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12962 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12963 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12964 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12965 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12966 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12967 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12968 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12969 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12970 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12971 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12972 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12973 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12974 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12975 (vfmlsl_low_u32): Ditto.
12976 (vfmlalq_low_u32): Ditto.
12977 (vfmlslq_low_u32): Ditto.
12978 (vfmlal_high_u32): Ditto.
12979 (vfmlsl_high_u32): Ditto.
12980 (vfmlalq_high_u32): Ditto.
12981 (vfmlslq_high_u32): Ditto.
12982 (vfmlal_lane_low_u32): Ditto.
12983 (vfmlsl_lane_low_u32): Ditto.
12984 (vfmlal_laneq_low_u32): Ditto.
12985 (vfmlsl_laneq_low_u32): Ditto.
12986 (vfmlalq_lane_low_u32): Ditto.
12987 (vfmlslq_lane_low_u32): Ditto.
12988 (vfmlalq_laneq_low_u32): Ditto.
12989 (vfmlslq_laneq_low_u32): Ditto.
12990 (vfmlal_lane_high_u32): Ditto.
12991 (vfmlsl_lane_high_u32): Ditto.
12992 (vfmlal_laneq_high_u32): Ditto.
12993 (vfmlsl_laneq_high_u32): Ditto.
12994 (vfmlalq_lane_high_u32): Ditto.
12995 (vfmlslq_lane_high_u32): Ditto.
12996 (vfmlalq_laneq_high_u32): Ditto.
12997 (vfmlslq_laneq_high_u32): Ditto.
12998 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12999 (AARCH64_FL_FOR_ARCH8_4): New.
13000 (AARCH64_ISA_F16FML): New ISA flag.
13001 (TARGET_F16FML): New feature flag for fp16fml.
13002 (doc/invoke.texi): Document new fp16fml option.
13004 2018-01-10 Michael Collison <michael.collison@arm.com>
13006 * config/aarch64/aarch64-builtins.c:
13007 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13008 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13009 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13010 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13011 (AARCH64_ISA_SHA3): New ISA flag.
13012 (TARGET_SHA3): New feature flag for sha3.
13013 * config/aarch64/iterators.md (sha512_op): New int attribute.
13014 (CRYPTO_SHA512): New int iterator.
13015 (UNSPEC_SHA512H): New unspec.
13016 (UNSPEC_SHA512H2): Ditto.
13017 (UNSPEC_SHA512SU0): Ditto.
13018 (UNSPEC_SHA512SU1): Ditto.
13019 * config/aarch64/aarch64-simd-builtins.def
13020 (aarch64_crypto_sha512hqv2di): New builtin.
13021 (aarch64_crypto_sha512h2qv2di): Ditto.
13022 (aarch64_crypto_sha512su0qv2di): Ditto.
13023 (aarch64_crypto_sha512su1qv2di): Ditto.
13024 (aarch64_eor3qv8hi): Ditto.
13025 (aarch64_rax1qv2di): Ditto.
13026 (aarch64_xarqv2di): Ditto.
13027 (aarch64_bcaxqv8hi): Ditto.
13028 * config/aarch64/aarch64-simd.md:
13029 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13030 (aarch64_crypto_sha512su0qv2di): Ditto.
13031 (aarch64_crypto_sha512su1qv2di): Ditto.
13032 (aarch64_eor3qv8hi): Ditto.
13033 (aarch64_rax1qv2di): Ditto.
13034 (aarch64_xarqv2di): Ditto.
13035 (aarch64_bcaxqv8hi): Ditto.
13036 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13037 (vsha512h2q_u64): Ditto.
13038 (vsha512su0q_u64): Ditto.
13039 (vsha512su1q_u64): Ditto.
13040 (veor3q_u16): Ditto.
13041 (vrax1q_u64): Ditto.
13042 (vxarq_u64): Ditto.
13043 (vbcaxq_u16): Ditto.
13044 * config/arm/types.md (crypto_sha512): New type attribute.
13045 (crypto_sha3): Ditto.
13046 (doc/invoke.texi): Document new sha3 option.
13048 2018-01-10 Michael Collison <michael.collison@arm.com>
13050 * config/aarch64/aarch64-builtins.c:
13051 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13052 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13053 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13054 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13055 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13056 (AARCH64_ISA_SM4): New ISA flag.
13057 (TARGET_SM4): New feature flag for sm4.
13058 * config/aarch64/aarch64-simd-builtins.def
13059 (aarch64_sm3ss1qv4si): Ditto.
13060 (aarch64_sm3tt1aq4si): Ditto.
13061 (aarch64_sm3tt1bq4si): Ditto.
13062 (aarch64_sm3tt2aq4si): Ditto.
13063 (aarch64_sm3tt2bq4si): Ditto.
13064 (aarch64_sm3partw1qv4si): Ditto.
13065 (aarch64_sm3partw2qv4si): Ditto.
13066 (aarch64_sm4eqv4si): Ditto.
13067 (aarch64_sm4ekeyqv4si): Ditto.
13068 * config/aarch64/aarch64-simd.md:
13069 (aarch64_sm3ss1qv4si): Ditto.
13070 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13071 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13072 (aarch64_sm4eqv4si): Ditto.
13073 (aarch64_sm4ekeyqv4si): Ditto.
13074 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13075 (sm3part_op): Ditto.
13076 (CRYPTO_SM3TT): Ditto.
13077 (CRYPTO_SM3PART): Ditto.
13078 (UNSPEC_SM3SS1): New unspec.
13079 (UNSPEC_SM3TT1A): Ditto.
13080 (UNSPEC_SM3TT1B): Ditto.
13081 (UNSPEC_SM3TT2A): Ditto.
13082 (UNSPEC_SM3TT2B): Ditto.
13083 (UNSPEC_SM3PARTW1): Ditto.
13084 (UNSPEC_SM3PARTW2): Ditto.
13085 (UNSPEC_SM4E): Ditto.
13086 (UNSPEC_SM4EKEY): Ditto.
13087 * config/aarch64/constraints.md (Ui2): New constraint.
13088 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13089 * config/arm/types.md (crypto_sm3): New type attribute.
13090 (crypto_sm4): Ditto.
13091 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13092 (vsm3tt1aq_u32): Ditto.
13093 (vsm3tt1bq_u32): Ditto.
13094 (vsm3tt2aq_u32): Ditto.
13095 (vsm3tt2bq_u32): Ditto.
13096 (vsm3partw1q_u32): Ditto.
13097 (vsm3partw2q_u32): Ditto.
13098 (vsm4eq_u32): Ditto.
13099 (vsm4ekeyq_u32): Ditto.
13100 (doc/invoke.texi): Document new sm4 option.
13102 2018-01-10 Michael Collison <michael.collison@arm.com>
13104 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13105 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13106 (AARCH64_FL_FOR_ARCH8_4): New.
13107 (AARCH64_FL_V8_4): New flag.
13108 (doc/invoke.texi): Document new armv8.4-a option.
13110 2018-01-10 Michael Collison <michael.collison@arm.com>
13112 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13113 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13114 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13115 * config/aarch64/aarch64-option-extension.def: Add
13116 AARCH64_OPT_EXTENSION of 'sha2'.
13117 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13118 (crypto): Disable sha2 and aes if crypto disabled.
13119 (crypto): Enable aes and sha2 if enabled.
13120 (simd): Disable sha2 and aes if simd disabled.
13121 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13123 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13124 (TARGET_SHA2): New feature flag for sha2.
13125 (TARGET_AES): New feature flag for aes.
13126 * config/aarch64/aarch64-simd.md:
13127 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13128 conditional on TARGET_AES.
13129 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13130 (aarch64_crypto_sha1hsi): Make pattern conditional
13132 (aarch64_crypto_sha1hv4si): Ditto.
13133 (aarch64_be_crypto_sha1hv4si): Ditto.
13134 (aarch64_crypto_sha1su1v4si): Ditto.
13135 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13136 (aarch64_crypto_sha1su0v4si): Ditto.
13137 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13138 (aarch64_crypto_sha256su0v4si): Ditto.
13139 (aarch64_crypto_sha256su1v4si): Ditto.
13140 (doc/invoke.texi): Document new aes and sha2 options.
13142 2018-01-10 Martin Sebor <msebor@redhat.com>
13144 PR tree-optimization/83781
13145 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13148 2018-01-11 Martin Sebor <msebor@gmail.com>
13149 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13151 PR tree-optimization/83501
13152 PR tree-optimization/81703
13154 * tree-ssa-strlen.c (get_string_cst): Rename...
13155 (get_string_len): ...to this. Handle global constants.
13156 (handle_char_store): Adjust.
13158 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13159 Jim Wilson <jimw@sifive.com>
13161 * config/riscv/riscv-protos.h (riscv_output_return): New.
13162 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13163 (riscv_attribute_table, riscv_output_return),
13164 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13165 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13166 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13168 (riscv_expand_prologue): Add early return for naked function.
13169 (riscv_expand_epilogue): Likewise.
13170 (riscv_function_ok_for_sibcall): Return false for naked function.
13171 (riscv_set_current_function): New.
13172 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13173 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13174 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13175 * doc/extend.texi (RISC-V Function Attributes): New.
13177 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13179 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13180 check for 128-bit long double before checking TCmode.
13181 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13182 128-bit long doubles before checking TFmode or TCmode.
13183 (FLOAT128_IBM_P): Likewise.
13185 2018-01-10 Martin Sebor <msebor@redhat.com>
13187 PR tree-optimization/83671
13188 * builtins.c (c_strlen): Unconditionally return zero for the empty
13190 Use -Warray-bounds for warnings.
13191 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13192 for non-constant array indices with COMPONENT_REF, arrays of
13193 arrays, and pointers to arrays.
13194 (gimple_fold_builtin_strlen): Determine and set length range for
13195 non-constant character arrays.
13197 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13199 PR middle-end/81897
13200 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13203 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13205 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13207 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13210 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13211 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13212 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13213 indexed_or_indirect_operand predicate.
13214 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13215 (*vsx_le_perm_load_v8hi): Likewise.
13216 (*vsx_le_perm_load_v16qi): Likewise.
13217 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13218 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13219 (*vsx_le_perm_store_v8hi): Likewise.
13220 (*vsx_le_perm_store_v16qi): Likewise.
13221 (eight unnamed splitters): Likewise.
13223 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13225 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13226 * config/rs6000/emmintrin.h: Likewise.
13227 * config/rs6000/mmintrin.h: Likewise.
13228 * config/rs6000/xmmintrin.h: Likewise.
13230 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13233 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13235 * tree.c (tree_nop_conversion): Return true for location wrapper
13237 (maybe_wrap_with_location): New function.
13238 (selftest::check_strip_nops): New function.
13239 (selftest::test_location_wrappers): New function.
13240 (selftest::tree_c_tests): Call it.
13241 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13242 (maybe_wrap_with_location): New decl.
13243 (EXPR_LOCATION_WRAPPER_P): New macro.
13244 (location_wrapper_p): New inline function.
13245 (tree_strip_any_location_wrapper): New inline function.
13247 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13250 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13251 stack_realign_offset for the largest alignment of stack slot
13253 (ix86_find_max_used_stack_alignment): New function.
13254 (ix86_finalize_stack_frame_flags): Use it. Set
13255 max_used_stack_alignment if we don't realign stack.
13256 * config/i386/i386.h (machine_function): Add
13257 max_used_stack_alignment.
13259 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13261 * config/arm/arm.opt (-mbranch-cost): New option.
13262 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13265 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13268 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13269 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13271 2018-01-10 Richard Biener <rguenther@suse.de>
13274 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13275 early out so it also covers the case where we have a non-NULL
13278 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13280 PR tree-optimization/83753
13281 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13282 for non-strided grouped accesses if the number of elements is 1.
13284 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13287 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13288 * i386.h (TARGET_USE_GATHER): Define.
13289 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13291 2018-01-10 Martin Liska <mliska@suse.cz>
13294 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13295 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13297 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13298 CLEANUP_NO_PARTITIONING is not set.
13300 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13302 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13303 for vectors, as a partial revert of r254296.
13304 * rtl.h (const_vec_p): Delete.
13305 (const_vec_duplicate_p): Don't test for vector CONSTs.
13306 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13307 * expmed.c (make_tree): Likewise.
13310 * common.md (E, F): Use CONSTANT_P instead of checking for
13312 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13313 checking for CONST_VECTOR.
13315 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13317 PR middle-end/83575
13318 * predict.c (force_edge_cold): Handle in more sane way edges
13319 with no prediction.
13321 2018-01-09 Carl Love <cel@us.ibm.com>
13323 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13325 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13326 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13327 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13328 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13329 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13330 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13331 * config/rs6000/rs6000-protos.h: Add extern defition for
13332 rs6000_generate_float2_double_code.
13333 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13335 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13336 (float2_v2df): Add define_expand.
13338 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13341 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13342 op_mode in the force_to_mode call.
13344 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13346 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13347 instead of checking each element individually.
13348 (aarch64_evpc_uzp): Likewise.
13349 (aarch64_evpc_zip): Likewise.
13350 (aarch64_evpc_ext): Likewise.
13351 (aarch64_evpc_rev): Likewise.
13352 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13353 instead of checking each element individually. Return true without
13355 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13356 whether all selected elements come from the same input, instead of
13357 checking each element individually. Remove calls to gen_rtx_REG,
13358 start_sequence and end_sequence and instead assert that no rtl is
13361 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13363 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13364 order of HIGH and CONST checks.
13366 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13368 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13369 if the destination isn't an SSA_NAME.
13371 2018-01-09 Richard Biener <rguenther@suse.de>
13373 PR tree-optimization/83668
13374 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13376 (canonicalize_loop_form): ... here, renamed from ...
13377 (canonicalize_loop_closed_ssa_form): ... this and amended to
13378 swap successor edges for loop exit blocks to make us use
13379 the RPO order we need for initial schedule generation.
13381 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13383 PR tree-optimization/64811
13384 * match.pd: When optimizing comparisons with Inf, avoid
13385 introducing or losing exceptions from comparisons with NaN.
13387 2018-01-09 Martin Liska <mliska@suse.cz>
13390 * asan.c (shadow_mem_size): Add gcc_assert.
13392 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13394 Don't save registers in main().
13397 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13398 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13399 * config/avr/avr.c (avr_set_current_function): Don't error if
13400 naked, OS_task or OS_main are specified at the same time.
13401 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13403 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13405 * common/config/avr/avr-common.c (avr_option_optimization_table):
13406 Switch on -mmain-is-OS_task for optimizing compilations.
13408 2018-01-09 Richard Biener <rguenther@suse.de>
13410 PR tree-optimization/83572
13411 * graphite.c: Include cfganal.h.
13412 (graphite_transform_loops): Connect infinite loops to exit
13413 and remove fake edges at the end.
13415 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13417 * ipa-inline.c (edge_badness): Revert accidental checkin.
13419 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13422 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13423 symbols; not inline clones.
13425 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13428 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13429 hard registers. Formatting fixes.
13431 PR preprocessor/83722
13432 * gcc.c (try_generate_repro): Pass
13433 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13434 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13437 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13438 Kito Cheng <kito.cheng@gmail.com>
13440 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13441 (riscv_leaf_function_p): Delete.
13442 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13444 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13446 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13448 (do_ifelse): New function.
13449 (do_isel): New function.
13450 (do_sub3): New function.
13451 (do_add3): New function.
13452 (do_load_mask_compare): New function.
13453 (do_overlap_load_compare): New function.
13454 (expand_compare_loop): New function.
13455 (expand_block_compare): Call expand_compare_loop() when appropriate.
13456 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13457 option description.
13458 (-mblock-compare-inline-loop-limit): New option.
13460 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13463 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13464 Reverse order of second and third operands in first alternative.
13465 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13466 of first and second elements in UNSPEC_VPERMR vector.
13467 (altivec_expand_vec_perm_le): Likewise.
13469 2018-01-08 Jeff Law <law@redhat.com>
13471 PR rtl-optimizatin/81308
13472 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13473 (process_switch): If group_case_labels makes a change, then set
13475 (pass_convert_switch::execute): If a switch is converted, then
13476 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13478 PR rtl-optimization/81308
13479 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13482 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13484 PR target/83663 - Revert r255946
13485 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13486 generation for cases where splatting a value is not useful.
13487 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13488 across a vec_duplicate and a paradoxical subreg forming a vector
13489 mode to a vec_concat.
13491 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13493 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13494 -march=armv8.3-a variants.
13495 * config/arm/t-multilib: Likewise.
13496 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13498 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13500 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13502 (cceq_ior_compare_complement): Give it a name so I can use it, and
13503 change boolean_or_operator predicate to boolean_operator so it can
13504 be used to generate a crand.
13505 (eqne): New code iterator.
13506 (bd/bd_neg): New code_attrs.
13507 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13508 a single define_insn.
13509 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13510 decrement (bdnzt/bdnzf/bdzt/bdzf).
13511 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13512 with the new names of the branch decrement patterns, and added the
13513 names of the branch decrement conditional patterns.
13515 2018-01-08 Richard Biener <rguenther@suse.de>
13517 PR tree-optimization/83563
13518 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13521 2018-01-08 Richard Biener <rguenther@suse.de>
13523 PR middle-end/83713
13524 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13526 2018-01-08 Richard Biener <rguenther@suse.de>
13528 PR tree-optimization/83685
13529 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13530 references to abnormals.
13532 2018-01-08 Richard Biener <rguenther@suse.de>
13535 * dwarf2out.c (output_indirect_strings): Handle empty
13536 skeleton_debug_str_hash.
13537 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13539 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13541 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13542 (emit_store_direct): Likewise.
13543 (arc_trampoline_adjust_address): Likewise.
13544 (arc_asm_trampoline_template): New function.
13545 (arc_initialize_trampoline): Use asm_trampoline_template.
13546 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13547 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13548 * config/arc/arc.md (flush_icache): Delete pattern.
13550 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13552 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13553 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13556 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13559 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13560 by not USED_FOR_TARGET.
13561 (make_pass_resolve_sw_modes): Likewise.
13563 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13565 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13568 2018-01-08 Richard Biener <rguenther@suse.de>
13570 PR middle-end/83580
13571 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13573 2018-01-08 Richard Biener <rguenther@suse.de>
13575 PR middle-end/83517
13576 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13578 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13580 PR middle-end/81897
13581 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13582 basic blocks with a small number of successors.
13583 (convert_control_dep_chain_into_preds): Improve handling of
13585 (dump_predicates): Split apart into...
13586 (dump_pred_chain): ...here...
13587 (dump_pred_info): ...and here.
13588 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13589 (can_chain_union_be_invalidated_p): Improve check for invalidation
13591 (uninit_uses_cannot_happen): Avoid unnecessary if
13592 convert_control_dep_chain_into_preds yielded nothing.
13594 2018-01-06 Martin Sebor <msebor@redhat.com>
13596 PR tree-optimization/83640
13597 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13598 subtracting negative offset from size.
13599 (builtin_access::overlap): Adjust offset bounds of the access to fall
13600 within the size of the object if possible.
13602 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13604 PR rtl-optimization/83699
13605 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13606 extract_bit_field_as_subreg to cases in which the extracted
13607 value is also a vector.
13609 * lra-constraints.c (process_alt_operands): Test for the equivalence
13610 substitutions when detecting a possible reload cycle.
13612 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13615 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13616 by default if flag_selective_schedling{,2}. Formatting fixes.
13618 PR rtl-optimization/83682
13619 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13620 if it has non-VECTOR_MODE element mode.
13621 (vec_duplicate_p): Likewise.
13623 PR middle-end/83694
13624 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13625 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13627 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13630 * config/i386/i386-builtin.def
13631 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13632 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13633 Require also OPTION_MASK_ISA_AVX512F in addition to
13634 OPTION_MASK_ISA_GFNI.
13635 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13636 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13637 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13638 to OPTION_MASK_ISA_GFNI.
13639 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13640 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13641 OPTION_MASK_ISA_AVX512BW.
13642 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13643 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13644 addition to OPTION_MASK_ISA_GFNI.
13645 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13646 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13647 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13648 to OPTION_MASK_ISA_GFNI.
13649 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13650 a requirement for all ISAs rather than any of them with a few
13652 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13654 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13655 bitmasks to be enabled with 3 exceptions, instead of requiring any
13656 enabled ISA with lots of exceptions.
13657 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13658 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13659 Change avx512bw in isa attribute to avx512f.
13660 * config/i386/sgxintrin.h: Add license boilerplate.
13661 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13662 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13663 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13664 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13666 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13667 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13668 temporarily sse2 rather than sse if not enabled already.
13671 * config/i386/sse.md (VI248_VLBW): Rename to ...
13672 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13673 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13674 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13675 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13676 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13677 mode iterator instead of VI248_VLBW.
13679 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13681 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13682 (record_modified): Skip clobbers; add debug output.
13683 (param_change_prob): Use sreal frequencies.
13685 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13687 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13688 punt for user-aligned variables.
13690 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13692 * tree-chrec.c (chrec_contains_symbols): Return true for
13695 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13698 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13699 of (x|y) == x for BICS pattern.
13701 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13703 PR tree-optimization/83605
13704 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13705 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13708 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13710 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13711 * config/epiphany/rtems.h: New file.
13713 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13714 Uros Bizjak <ubizjak@gmail.com>
13717 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13718 QIreg_operand instead of register_operand predicate.
13719 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13720 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13721 comments instead of -fmitigate[-_]rop.
13723 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13726 * cgraphunit.c (symbol_table::compile): Switch to text_section
13727 before calling assembly_start debug hook.
13728 * run-rtl-passes.c (run_rtl_passes): Likewise.
13731 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13733 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13734 range_int_cst_p rather than !symbolic_range_p before calling
13735 extract_range_from_multiplicative_op_1.
13737 2018-01-04 Jeff Law <law@redhat.com>
13739 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13740 redundant test in assertion.
13742 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13744 * doc/rtl.texi: Document machine_mode wrapper classes.
13746 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13748 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13749 using tree_to_uhwi.
13751 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13753 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13754 the VEC_PERM_EXPR fold to fail.
13756 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13759 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13760 to switched_sections.
13762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13765 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13766 test for d.testing.
13768 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13771 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13772 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13774 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13777 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13778 is BLKmode and bitpos not zero or mode change is needed.
13780 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13783 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13786 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13789 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13790 instead of MULT rtx. Update all corresponding splitters.
13791 (*saddl_se): Ditto.
13792 (*ssub<modesuffix>): Ditto.
13793 (*ssubl_se): Ditto.
13794 (*cmp_sadd_di): Update split patterns.
13795 (*cmp_sadd_si): Ditto.
13796 (*cmp_sadd_sidi): Ditto.
13797 (*cmp_ssub_di): Ditto.
13798 (*cmp_ssub_si): Ditto.
13799 (*cmp_ssub_sidi): Ditto.
13800 * config/alpha/predicates.md (const23_operand): New predicate.
13801 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13802 Look for ASHIFT, not MULT inner operand.
13803 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13805 2018-01-04 Martin Liska <mliska@suse.cz>
13807 PR gcov-profile/83669
13808 * gcov.c (output_intermediate_file): Add version to intermediate
13810 * doc/gcov.texi: Document new field 'version' in intermediate
13811 file format. Fix location of '-k' option of gcov command.
13813 2018-01-04 Martin Liska <mliska@suse.cz>
13816 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13818 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13820 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13822 2018-01-03 Martin Sebor <msebor@redhat.com>
13824 PR tree-optimization/83655
13825 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13826 checking calls with invalid arguments.
13828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13830 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13831 (vectorizable_mask_load_store): Delete.
13832 (vectorizable_call): Return false for masked loads and stores.
13833 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13834 instead of gimple_assign_rhs1.
13835 (vectorizable_load): Handle IFN_MASK_LOAD.
13836 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13840 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13842 (vectorizable_mask_load_store): ...here.
13843 (vectorizable_load): ...and here.
13845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13847 * tree-vect-stmts.c (vect_build_all_ones_mask)
13848 (vect_build_zero_merge_argument): New functions, split out from...
13849 (vectorizable_load): ...here.
13851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13853 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13855 (vectorizable_mask_load_store): ...here.
13856 (vectorizable_store): ...and here.
13858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13860 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13862 (vectorizable_mask_load_store): ...here.
13864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13866 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13867 (vect_model_store_cost): Take a vec_load_store_type instead of a
13869 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13870 (vect_model_store_cost): Take a vec_load_store_type instead of a
13872 (vectorizable_mask_load_store): Update accordingly.
13873 (vectorizable_store): Likewise.
13874 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13878 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13879 IFN_MASK_LOAD calls here rather than...
13880 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13883 Alan Hayward <alan.hayward@arm.com>
13884 David Sherwood <david.sherwood@arm.com>
13886 * expmed.c (extract_bit_field_1): For vector extracts,
13887 fall back to extract_bit_field_as_subreg if vec_extract
13890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13895 they are variable or constant sized.
13896 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13897 slots for constant-sized data.
13899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13900 Alan Hayward <alan.hayward@arm.com>
13901 David Sherwood <david.sherwood@arm.com>
13903 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13904 handling COND_EXPRs with boolean comparisons, try to find a better
13905 basis for the mask type than the boolean itself.
13907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13909 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13910 is calculated and how it can be overridden.
13911 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13912 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13914 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13918 Alan Hayward <alan.hayward@arm.com>
13919 David Sherwood <david.sherwood@arm.com>
13921 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13922 Remove the mode argument.
13923 (aarch64_simd_valid_immediate): Remove the mode and inverse
13925 * config/aarch64/iterators.md (bitsize): New iterator.
13926 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13927 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13928 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13929 aarch64_simd_valid_immediate.
13930 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13931 (aarch64_reg_or_bic_imm): Likewise.
13932 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13933 with an insn_type enum and msl with a modifier_type enum.
13934 Replace element_width with a scalar_mode. Change the shift
13935 to unsigned int. Add constructors for scalar_float_mode and
13936 scalar_int_mode elements.
13937 (aarch64_vect_float_const_representable_p): Delete.
13938 (aarch64_can_const_movi_rtx_p)
13939 (aarch64_simd_scalar_immediate_valid_for_move)
13940 (aarch64_simd_make_constant): Update call to
13941 aarch64_simd_valid_immediate.
13942 (aarch64_advsimd_valid_immediate_hs): New function.
13943 (aarch64_advsimd_valid_immediate): Likewise.
13944 (aarch64_simd_valid_immediate): Remove mode and inverse
13945 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13946 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13947 and aarch64_float_const_representable_p on the result.
13948 (aarch64_output_simd_mov_immediate): Remove mode argument.
13949 Update call to aarch64_simd_valid_immediate and use of
13950 simd_immediate_info.
13951 (aarch64_output_scalar_simd_mov_immediate): Update call
13954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13955 Alan Hayward <alan.hayward@arm.com>
13956 David Sherwood <david.sherwood@arm.com>
13958 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13959 (mode_nunits): Likewise CONST_MODE_NUNITS.
13960 * machmode.def (ADJUST_NUNITS): Document.
13961 * genmodes.c (mode_data::need_nunits_adj): New field.
13962 (blank_mode): Update accordingly.
13963 (adj_nunits): New variable.
13964 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13966 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13967 listed in adj_nunits.
13968 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13969 listed in adj_nunits. Don't emit case statements for such modes.
13970 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13971 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13972 nothing if adj_nunits is nonnull.
13973 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13974 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13975 (emit_mode_fbit): Update use of print_maybe_const_decl.
13976 (emit_move_size): Likewise. Treat the array as non-const
13978 (emit_mode_adjustments): Handle adj_nunits.
13980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13982 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13983 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13984 (VECTOR_MODES): Use it.
13985 (make_vector_modes): Take the prefix as an argument.
13987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13988 Alan Hayward <alan.hayward@arm.com>
13989 David Sherwood <david.sherwood@arm.com>
13991 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13992 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13993 for MODE_VECTOR_BOOL.
13994 * machmode.def (VECTOR_BOOL_MODE): Document.
13995 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13996 (make_vector_bool_mode): New function.
13997 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13999 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14000 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14002 * stor-layout.c (int_mode_for_mode): Likewise.
14003 * tree.c (build_vector_type_for_mode): Likewise.
14004 * varasm.c (output_constant_pool_2): Likewise.
14005 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14006 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14007 for MODE_VECTOR_BOOL.
14008 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14009 of mode class checks.
14010 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14011 instead of a list of mode class checks.
14012 (expand_vector_scalar_condition): Likewise.
14013 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14016 Alan Hayward <alan.hayward@arm.com>
14017 David Sherwood <david.sherwood@arm.com>
14019 * machmode.h (mode_size): Change from unsigned short to
14021 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14022 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14023 or if measurement_type is not polynomial.
14024 (fixed_size_mode::includes_p): Check for constant-sized modes.
14025 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14026 return a poly_uint16 rather than an unsigned short.
14027 (emit_mode_size): Change the type of mode_size from unsigned short
14028 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14029 (emit_mode_adjustments): Cope with polynomial vector sizes.
14030 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14032 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14034 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14035 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14036 * caller-save.c (setup_save_areas): Likewise.
14037 (replace_reg_with_saved_mem): Likewise.
14038 * calls.c (emit_library_call_value_1): Likewise.
14039 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14040 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14041 (gen_lowpart_for_combine): Likewise.
14042 * convert.c (convert_to_integer_1): Likewise.
14043 * cse.c (equiv_constant, cse_insn): Likewise.
14044 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14045 (cselib_subst_to_values): Likewise.
14046 * dce.c (word_dce_process_block): Likewise.
14047 * df-problems.c (df_word_lr_mark_ref): Likewise.
14048 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14049 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14050 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14051 (rtl_for_decl_location): Likewise.
14052 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14053 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14054 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14055 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14056 (expand_expr_real_1): Likewise.
14057 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14058 (pad_below): Likewise.
14059 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14060 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14061 * ira.c (get_subreg_tracking_sizes): Likewise.
14062 * ira-build.c (ira_create_allocno_objects): Likewise.
14063 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14064 (ira_sort_regnos_for_alter_reg): Likewise.
14065 * ira-costs.c (record_operand_costs): Likewise.
14066 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14067 (resolve_simple_move): Likewise.
14068 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14069 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14070 (lra_constraints): Likewise.
14071 (CONST_POOL_OK_P): Reject variable-sized modes.
14072 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14073 (add_pseudo_to_slot, lra_spill): Likewise.
14074 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14075 * optabs-query.c (get_best_extraction_insn): Likewise.
14076 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14077 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14078 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14079 * recog.c (offsettable_address_addr_space_p): Likewise.
14080 * regcprop.c (maybe_mode_change): Likewise.
14081 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14082 * regrename.c (build_def_use): Likewise.
14083 * regstat.c (dump_reg_info): Likewise.
14084 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14085 (find_reloads, find_reloads_subreg_address): Likewise.
14086 * reload1.c (eliminate_regs_1): Likewise.
14087 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14088 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14089 (simplify_binary_operation_1, simplify_subreg): Likewise.
14090 * targhooks.c (default_function_arg_padding): Likewise.
14091 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14092 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14093 (verify_gimple_assign_ternary): Likewise.
14094 * tree-inline.c (estimate_move_cost): Likewise.
14095 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14096 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14097 (get_address_cost_ainc): Likewise.
14098 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14099 (vect_supportable_dr_alignment): Likewise.
14100 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14101 (vectorizable_reduction): Likewise.
14102 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14103 (vectorizable_operation, vectorizable_load): Likewise.
14104 * tree.c (build_same_sized_truth_vector_type): Likewise.
14105 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14106 * var-tracking.c (emit_note_insn_var_location): Likewise.
14107 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14108 (ADDR_VEC_ALIGN): Likewise.
14110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14111 Alan Hayward <alan.hayward@arm.com>
14112 David Sherwood <david.sherwood@arm.com>
14114 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14116 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14117 or if measurement_type is polynomial.
14118 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14119 * combine.c (make_extraction): Likewise.
14120 * dse.c (find_shift_sequence): Likewise.
14121 * dwarf2out.c (mem_loc_descriptor): Likewise.
14122 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14123 (extract_bit_field, extract_low_bits): Likewise.
14124 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14125 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14126 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14127 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14128 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14129 * reload.c (find_reloads): Likewise.
14130 * reload1.c (alter_reg): Likewise.
14131 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14132 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14133 * tree-if-conv.c (predicate_mem_writes): Likewise.
14134 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14135 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14136 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14137 * valtrack.c (dead_debug_insert_temp): Likewise.
14138 * varasm.c (mergeable_constant_section): Likewise.
14139 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14142 Alan Hayward <alan.hayward@arm.com>
14143 David Sherwood <david.sherwood@arm.com>
14145 * expr.c (expand_assignment): Cope with polynomial mode sizes
14146 when assigning to a CONCAT.
14148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14149 Alan Hayward <alan.hayward@arm.com>
14150 David Sherwood <david.sherwood@arm.com>
14152 * machmode.h (mode_precision): Change from unsigned short to
14154 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14156 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14157 or if measurement_type is not polynomial.
14158 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14159 in which the mode is already known to be a scalar_int_mode.
14160 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14161 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14163 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14164 for GET_MODE_PRECISION.
14165 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14166 for GET_MODE_PRECISION.
14167 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14169 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14170 (expand_field_assignment, make_extraction): Likewise.
14171 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14172 (get_last_value): Likewise.
14173 * convert.c (convert_to_integer_1): Likewise.
14174 * cse.c (cse_insn): Likewise.
14175 * expr.c (expand_expr_real_1): Likewise.
14176 * lra-constraints.c (simplify_operand_subreg): Likewise.
14177 * optabs-query.c (can_atomic_load_p): Likewise.
14178 * optabs.c (expand_atomic_load): Likewise.
14179 (expand_atomic_store): Likewise.
14180 * ree.c (combine_reaching_defs): Likewise.
14181 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14182 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14183 * tree.h (type_has_mode_precision_p): Likewise.
14184 * ubsan.c (instrument_si_overflow): Likewise.
14186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14187 Alan Hayward <alan.hayward@arm.com>
14188 David Sherwood <david.sherwood@arm.com>
14190 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14191 polynomial numbers of units.
14192 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14193 (valid_vector_subparts_p): New function.
14194 (build_vector_type): Remove temporary shim and take the number
14195 of units as a poly_uint64 rather than an int.
14196 (build_opaque_vector_type): Take the number of units as a
14197 poly_uint64 rather than an int.
14198 * tree.c (build_vector_from_ctor): Handle polynomial
14199 TYPE_VECTOR_SUBPARTS.
14200 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14201 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14202 (build_vector_from_val): If the number of units is variable,
14203 use build_vec_duplicate_cst for constant operands and
14204 VEC_DUPLICATE_EXPR otherwise.
14205 (make_vector_type): Remove temporary is_constant ().
14206 (build_vector_type, build_opaque_vector_type): Take the number of
14207 units as a poly_uint64 rather than an int.
14208 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14210 * cfgexpand.c (expand_debug_expr): Likewise.
14211 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14212 (store_constructor, expand_expr_real_1): Likewise.
14213 (const_scalar_mask_from_tree): Likewise.
14214 * fold-const-call.c (fold_const_reduction): Likewise.
14215 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14216 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14217 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14218 (fold_relational_const): Likewise.
14219 (native_interpret_vector): Likewise. Change the size from an
14220 int to an unsigned int.
14221 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14222 TYPE_VECTOR_SUBPARTS.
14223 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14224 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14225 duplicating a non-constant operand into a variable-length vector.
14226 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14227 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14228 * ipa-icf.c (sem_variable::equals): Likewise.
14229 * match.pd: Likewise.
14230 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14231 * print-tree.c (print_node): Likewise.
14232 * stor-layout.c (layout_type): Likewise.
14233 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14234 * tree-cfg.c (verify_gimple_comparison): Likewise.
14235 (verify_gimple_assign_binary): Likewise.
14236 (verify_gimple_assign_ternary): Likewise.
14237 (verify_gimple_assign_single): Likewise.
14238 * tree-pretty-print.c (dump_generic_node): Likewise.
14239 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14240 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14241 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14242 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14243 (vect_shift_permute_load_chain): Likewise.
14244 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14245 (expand_vector_condition, optimize_vector_constructor): Likewise.
14246 (lower_vec_perm, get_compute_type): Likewise.
14247 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14248 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14249 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14250 (vect_recog_mask_conversion_pattern): Likewise.
14251 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14252 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14253 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14254 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14255 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14256 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14257 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14258 (supportable_widening_operation): Likewise.
14259 (supportable_narrowing_operation): Likewise.
14260 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14262 * varasm.c (output_constant): Likewise.
14264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14265 Alan Hayward <alan.hayward@arm.com>
14266 David Sherwood <david.sherwood@arm.com>
14268 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14269 so that both the length == 3 and length != 3 cases set up their
14270 own permute vectors. Add comments explaining why we know the
14271 number of elements is constant.
14272 (vect_permute_load_chain): Likewise.
14274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14275 Alan Hayward <alan.hayward@arm.com>
14276 David Sherwood <david.sherwood@arm.com>
14278 * machmode.h (mode_nunits): Change from unsigned char to
14280 (ONLY_FIXED_SIZE_MODES): New macro.
14281 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14282 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14283 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14285 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14286 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14287 or if measurement_type is not polynomial.
14288 * genmodes.c (ZERO_COEFFS): New macro.
14289 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14291 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14292 Use ZERO_COEFFS when emitting initializers.
14293 * data-streamer.h (bp_pack_poly_value): New function.
14294 (bp_unpack_poly_value): Likewise.
14295 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14296 for GET_MODE_NUNITS.
14297 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14298 for GET_MODE_NUNITS.
14299 * tree.c (make_vector_type): Remove temporary shim and make
14300 the real function take the number of units as a poly_uint64
14301 rather than an int.
14302 (build_vector_type_for_mode): Handle polynomial nunits.
14303 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14304 * emit-rtl.c (const_vec_series_p_1): Likewise.
14305 (gen_rtx_CONST_VECTOR): Likewise.
14306 * fold-const.c (test_vec_duplicate_folding): Likewise.
14307 * genrecog.c (validate_pattern): Likewise.
14308 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14309 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14310 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14311 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14312 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14313 * rtlanal.c (subreg_get_info): Likewise.
14314 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14315 (vect_grouped_load_supported): Likewise.
14316 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14317 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14318 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14319 (simplify_const_unary_operation, simplify_binary_operation_1)
14320 (simplify_const_binary_operation, simplify_ternary_operation)
14321 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14322 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14323 instead of CONST_VECTOR_NUNITS.
14324 * varasm.c (output_constant_pool_2): Likewise.
14325 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14326 explicit-encoded elements in the XVEC for variable-length vectors.
14328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14330 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14333 Alan Hayward <alan.hayward@arm.com>
14334 David Sherwood <david.sherwood@arm.com>
14336 * coretypes.h (fixed_size_mode): Declare.
14337 (fixed_size_mode_pod): New typedef.
14338 * builtins.h (target_builtins::x_apply_args_mode)
14339 (target_builtins::x_apply_result_mode): Change type to
14340 fixed_size_mode_pod.
14341 * builtins.c (apply_args_size, apply_result_size, result_vector)
14342 (expand_builtin_apply_args_1, expand_builtin_apply)
14343 (expand_builtin_return): Update accordingly.
14345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14347 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14348 * cselib.c (cselib_hash_rtx): Likewise.
14349 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14350 CONST_VECTOR encoding.
14352 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14353 Jeff Law <law@redhat.com>
14356 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14357 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14358 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14359 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14362 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14363 explicitly probe *sp in a noreturn function if there were any callee
14364 register saves or frame pointer is needed.
14366 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14369 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14370 BLKmode for ternary, binary or unary expressions.
14373 * var-tracking.c (delete_vta_debug_insn): New inline function.
14374 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14375 insns from get_insns () to NULL instead of each bb separately.
14376 Use delete_vta_debug_insn. No longer static.
14377 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14378 delete_vta_debug_insns callers.
14379 * rtl.h (delete_vta_debug_insns): Declare.
14380 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14381 instead of variable_tracking_main.
14383 2018-01-03 Martin Sebor <msebor@redhat.com>
14385 PR tree-optimization/83603
14386 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14387 arguments past the endof the argument list in functions declared
14388 without a prototype.
14389 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14390 Avoid checking when arguments are null.
14392 2018-01-03 Martin Sebor <msebor@redhat.com>
14395 * doc/extend.texi (attribute const): Fix a typo.
14396 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14397 issuing -Wsuggest-attribute for void functions.
14399 2018-01-03 Martin Sebor <msebor@redhat.com>
14401 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14402 offset_int::from instead of wide_int::to_shwi.
14403 (maybe_diag_overlap): Remove assertion.
14404 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14405 * gimple-ssa-sprintf.c (format_directive): Same.
14406 (parse_directive): Same.
14407 (sprintf_dom_walker::compute_format_length): Same.
14408 (try_substitute_return_value): Same.
14410 2018-01-03 Jeff Law <law@redhat.com>
14412 PR middle-end/83654
14413 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14414 non-constant residual for zero at runtime and avoid probing in
14415 that case. Reorganize code for trailing problem to mirror handling
14418 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14420 PR tree-optimization/83501
14421 * tree-ssa-strlen.c (get_string_cst): New.
14422 (handle_char_store): Call get_string_cst.
14424 2018-01-03 Martin Liska <mliska@suse.cz>
14426 PR tree-optimization/83593
14427 * tree-ssa-strlen.c: Include tree-cfg.h.
14428 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14429 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14430 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14432 (strlen_dom_walker::before_dom_children): Call
14433 gimple_purge_dead_eh_edges. Dump tranformation with details
14435 (strlen_dom_walker::before_dom_children): Update call by adding
14436 new argument cleanup_eh.
14437 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14439 2018-01-03 Martin Liska <mliska@suse.cz>
14442 * cif-code.def (VARIADIC_THUNK): New enum value.
14443 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14446 2018-01-03 Jan Beulich <jbeulich@suse.com>
14448 * sse.md (mov<mode>_internal): Tighten condition for when to use
14449 vmovdqu<ssescalarsize> for TI and OI modes.
14451 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14453 Update copyright years.
14455 2018-01-03 Martin Liska <mliska@suse.cz>
14458 * ipa-visibility.c (function_and_variable_visibility): Skip
14459 functions with noipa attribure.
14461 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14463 * gcc.c (process_command): Update copyright notice dates.
14464 * gcov-dump.c (print_version): Ditto.
14465 * gcov.c (print_version): Ditto.
14466 * gcov-tool.c (print_version): Ditto.
14467 * gengtype.c (create_file): Ditto.
14468 * doc/cpp.texi: Bump @copying's copyright year.
14469 * doc/cppinternals.texi: Ditto.
14470 * doc/gcc.texi: Ditto.
14471 * doc/gccint.texi: Ditto.
14472 * doc/gcov.texi: Ditto.
14473 * doc/install.texi: Ditto.
14474 * doc/invoke.texi: Ditto.
14476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14478 * vector-builder.h (vector_builder::m_full_nelts): Change from
14479 unsigned int to poly_uint64.
14480 (vector_builder::full_nelts): Update prototype accordingly.
14481 (vector_builder::new_vector): Likewise.
14482 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14483 (vector_builder::operator ==): Likewise.
14484 (vector_builder::finalize): Likewise.
14485 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14486 Take the number of elements as a poly_uint64 rather than an
14488 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14489 from unsigned int to poly_uint64.
14490 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14491 (vec_perm_indices::new_vector): Likewise.
14492 (vec_perm_indices::length): Likewise.
14493 (vec_perm_indices::nelts_per_input): Likewise.
14494 (vec_perm_indices::input_nelts): Likewise.
14495 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14496 number of elements per input as a poly_uint64 rather than an
14497 unsigned int. Use the original encoding for variable-length
14498 vectors, rather than clamping each individual element.
14499 For the second and subsequent elements in each pattern,
14500 clamp the step and base before clamping their sum.
14501 (vec_perm_indices::series_p): Handle polynomial element counts.
14502 (vec_perm_indices::all_in_range_p): Likewise.
14503 (vec_perm_indices_to_tree): Likewise.
14504 (vec_perm_indices_to_rtx): Likewise.
14505 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14506 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14507 (tree_vector_builder::new_binary_operation): Handle polynomial
14508 element counts. Return false if we need to know the number
14509 of elements at compile time.
14510 * fold-const.c (fold_vec_perm): Punt if the number of elements
14511 isn't known at compile time.
14513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14515 * vec-perm-indices.h (vec_perm_builder): Change element type
14516 from HOST_WIDE_INT to poly_int64.
14517 (vec_perm_indices::element_type): Update accordingly.
14518 (vec_perm_indices::clamp): Handle polynomial element_types.
14519 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14520 (vec_perm_indices::all_in_range_p): Likewise.
14521 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14523 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14524 polynomial vec_perm_indices element types.
14525 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14526 * fold-const.c (fold_vec_perm): Likewise.
14527 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14528 * tree-vect-generic.c (lower_vec_perm): Likewise.
14529 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14530 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14531 element type to HOST_WIDE_INT.
14533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14534 Alan Hayward <alan.hayward@arm.com>
14535 David Sherwood <david.sherwood@arm.com>
14537 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14538 rather than an int. Use plus_constant.
14539 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14540 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14543 Alan Hayward <alan.hayward@arm.com>
14544 David Sherwood <david.sherwood@arm.com>
14546 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14547 a HOST_WIDE_INT to a poly_int64.
14549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14550 Alan Hayward <alan.hayward@arm.com>
14551 David Sherwood <david.sherwood@arm.com>
14553 * calls.c (load_register_parameters): Cope with polynomial
14554 mode sizes. Require a constant size for BLKmode parameters
14555 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14556 forces a parameter to be padded at the lsb end in order to
14557 fill a complete number of words, require the parameter size
14558 to be ordered wrt UNITS_PER_WORD.
14560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * reload1.c (spill_stack_slot_width): Change element type
14565 from unsigned int to poly_uint64_pod.
14566 (alter_reg): Treat mode sizes as polynomial.
14568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14569 Alan Hayward <alan.hayward@arm.com>
14570 David Sherwood <david.sherwood@arm.com>
14572 * reload.c (complex_word_subreg_p): New function.
14573 (reload_inner_reg_of_subreg, push_reload): Use it.
14575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14576 Alan Hayward <alan.hayward@arm.com>
14577 David Sherwood <david.sherwood@arm.com>
14579 * lra-constraints.c (process_alt_operands): Reject matched
14580 operands whose sizes aren't ordered.
14581 (match_reload): Refer to this check here.
14583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14584 Alan Hayward <alan.hayward@arm.com>
14585 David Sherwood <david.sherwood@arm.com>
14587 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14588 that the mode size is in the set {1, 2, 4, 8, 16}.
14590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14591 Alan Hayward <alan.hayward@arm.com>
14592 David Sherwood <david.sherwood@arm.com>
14594 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14595 Use plus_constant instead of gen_rtx_PLUS.
14597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14598 Alan Hayward <alan.hayward@arm.com>
14599 David Sherwood <david.sherwood@arm.com>
14601 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14602 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14603 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14604 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14605 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14606 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14607 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14608 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14609 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14610 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14612 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14613 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14614 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14615 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14616 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14617 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14618 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14619 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14620 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14621 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14623 * expr.c (emit_move_resolve_push): Treat the input and result
14624 of PUSH_ROUNDING as a poly_int64.
14625 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14626 (emit_push_insn): Likewise.
14627 * lra-eliminations.c (mark_not_eliminable): Likewise.
14628 * recog.c (push_operand): Likewise.
14629 * reload1.c (elimination_effects): Likewise.
14630 * rtlanal.c (nonzero_bits1): Likewise.
14631 * calls.c (store_one_arg): Likewise. Require the padding to be
14632 known at compile time.
14634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14635 Alan Hayward <alan.hayward@arm.com>
14636 David Sherwood <david.sherwood@arm.com>
14638 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14639 Use plus_constant instead of gen_rtx_PLUS.
14641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14642 Alan Hayward <alan.hayward@arm.com>
14643 David Sherwood <david.sherwood@arm.com>
14645 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14646 rather than an int.
14648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14649 Alan Hayward <alan.hayward@arm.com>
14650 David Sherwood <david.sherwood@arm.com>
14652 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14653 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14654 via stack temporaries. Treat the mode size as polynomial too.
14656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14657 Alan Hayward <alan.hayward@arm.com>
14658 David Sherwood <david.sherwood@arm.com>
14660 * expr.c (expand_expr_real_2): When handling conversions involving
14661 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14662 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14663 as a poly_uint64 too.
14665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14666 Alan Hayward <alan.hayward@arm.com>
14667 David Sherwood <david.sherwood@arm.com>
14669 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14672 Alan Hayward <alan.hayward@arm.com>
14673 David Sherwood <david.sherwood@arm.com>
14675 * combine.c (can_change_dest_mode): Handle polynomial
14676 REGMODE_NATURAL_SIZE.
14677 * expmed.c (store_bit_field_1): Likewise.
14678 * expr.c (store_constructor): Likewise.
14679 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14680 and polynomial REGMODE_NATURAL_SIZE.
14681 (gen_lowpart_common): Likewise.
14682 * reginfo.c (record_subregs_of_mode): Likewise.
14683 * rtlanal.c (read_modify_subreg_p): Likewise.
14685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14686 Alan Hayward <alan.hayward@arm.com>
14687 David Sherwood <david.sherwood@arm.com>
14689 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14690 numbers of elements.
14692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14693 Alan Hayward <alan.hayward@arm.com>
14694 David Sherwood <david.sherwood@arm.com>
14696 * match.pd: Cope with polynomial numbers of vector elements.
14698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14699 Alan Hayward <alan.hayward@arm.com>
14700 David Sherwood <david.sherwood@arm.com>
14702 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14703 in a POINTER_PLUS_EXPR.
14705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14706 Alan Hayward <alan.hayward@arm.com>
14707 David Sherwood <david.sherwood@arm.com>
14709 * omp-simd-clone.c (simd_clone_subparts): New function.
14710 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14711 (ipa_simd_modify_function_body): Likewise.
14713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14714 Alan Hayward <alan.hayward@arm.com>
14715 David Sherwood <david.sherwood@arm.com>
14717 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14718 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14719 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14720 (expand_vector_condition, vector_element): Likewise.
14721 (subparts_gt): New function.
14722 (get_compute_type): Use subparts_gt.
14723 (count_type_subparts): Delete.
14724 (expand_vector_operations_1): Use subparts_gt instead of
14725 count_type_subparts.
14727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14728 Alan Hayward <alan.hayward@arm.com>
14729 David Sherwood <david.sherwood@arm.com>
14731 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14732 (vect_compile_time_alias): ...this new function. Do the calculation
14733 on poly_ints rather than trees.
14734 (vect_prune_runtime_alias_test_list): Update call accordingly.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14737 Alan Hayward <alan.hayward@arm.com>
14738 David Sherwood <david.sherwood@arm.com>
14740 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14742 (vect_schedule_slp_instance): Likewise.
14744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14745 Alan Hayward <alan.hayward@arm.com>
14746 David Sherwood <david.sherwood@arm.com>
14748 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14749 constant and extern definitions for variable-length vectors.
14750 (vect_get_constant_vectors): Note that the number of units
14751 is known to be constant.
14753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14754 Alan Hayward <alan.hayward@arm.com>
14755 David Sherwood <david.sherwood@arm.com>
14757 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14758 of units as polynomial. Choose between WIDE and NARROW based
14761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14762 Alan Hayward <alan.hayward@arm.com>
14763 David Sherwood <david.sherwood@arm.com>
14765 * tree-vect-stmts.c (simd_clone_subparts): New function.
14766 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14769 Alan Hayward <alan.hayward@arm.com>
14770 David Sherwood <david.sherwood@arm.com>
14772 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14773 vectors as polynomial. Use build_index_vector for
14774 IFN_GOMP_SIMD_LANE.
14776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14777 Alan Hayward <alan.hayward@arm.com>
14778 David Sherwood <david.sherwood@arm.com>
14780 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14781 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14782 for variable-length vectors.
14783 (vectorizable_mask_load_store): Treat the number of units as
14784 polynomial, asserting that it is constant if the condition has
14785 already been enforced.
14786 (vectorizable_store, vectorizable_load): Likewise.
14788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14789 Alan Hayward <alan.hayward@arm.com>
14790 David Sherwood <david.sherwood@arm.com>
14792 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14793 of units as polynomial. Punt if we can't tell at compile time
14794 which vector contains the final result.
14796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14797 Alan Hayward <alan.hayward@arm.com>
14798 David Sherwood <david.sherwood@arm.com>
14800 * tree-vect-loop.c (vectorizable_induction): Treat the number
14801 of units as polynomial. Punt on SLP inductions. Use an integer
14802 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14803 cast of such a series for variable-length floating-point
14806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14807 Alan Hayward <alan.hayward@arm.com>
14808 David Sherwood <david.sherwood@arm.com>
14810 * tree.h (build_index_vector): Declare.
14811 * tree.c (build_index_vector): New function.
14812 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14813 of units as polynomial, forcibly converting it to a constant if
14814 vectorizable_reduction has already enforced the condition.
14815 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14816 to create a {1,2,3,...} vector.
14817 (vectorizable_reduction): Treat the number of units as polynomial.
14818 Choose vectype_in based on the largest scalar element size rather
14819 than the smallest number of units. Enforce the restrictions
14822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14823 Alan Hayward <alan.hayward@arm.com>
14824 David Sherwood <david.sherwood@arm.com>
14826 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14827 number of units as polynomial.
14829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14830 Alan Hayward <alan.hayward@arm.com>
14831 David Sherwood <david.sherwood@arm.com>
14833 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14834 * target.def (autovectorize_vector_sizes): Return the vector sizes
14835 by pointer, using vector_sizes rather than a bitmask.
14836 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14837 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14838 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14840 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14841 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14842 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14843 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14844 * omp-general.c (omp_max_vf): Likewise.
14845 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14846 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14847 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14848 * tree-vect-slp.c (vect_slp_bb): Likewise.
14849 * doc/tm.texi: Regenerate.
14850 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14852 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14853 the vector size as a poly_uint64 rather than an unsigned int.
14854 (current_vector_size): Change from an unsigned int to a poly_uint64.
14855 (get_vectype_for_scalar_type): Update accordingly.
14856 * tree.h (build_truth_vector_type): Take the size and number of
14857 units as a poly_uint64 rather than an unsigned int.
14858 (build_vector_type): Add a temporary overload that takes
14859 the number of units as a poly_uint64 rather than an unsigned int.
14860 * tree.c (make_vector_type): Likewise.
14861 (build_truth_vector_type): Take the number of units as a poly_uint64
14862 rather than an unsigned int.
14864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14865 Alan Hayward <alan.hayward@arm.com>
14866 David Sherwood <david.sherwood@arm.com>
14868 * target.def (get_mask_mode): Take the number of units and length
14869 as poly_uint64s rather than unsigned ints.
14870 * targhooks.h (default_get_mask_mode): Update accordingly.
14871 * targhooks.c (default_get_mask_mode): Likewise.
14872 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14873 * doc/tm.texi: Regenerate.
14875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14876 Alan Hayward <alan.hayward@arm.com>
14877 David Sherwood <david.sherwood@arm.com>
14879 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14880 * omp-general.c (omp_max_vf): Likewise.
14881 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14882 (expand_omp_simd): Handle polynomial safelen.
14883 * omp-low.c (omplow_simd_context): Add a default constructor.
14884 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14885 (lower_rec_simd_input_clauses): Update accordingly.
14886 (lower_rec_input_clauses): Likewise.
14888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14889 Alan Hayward <alan.hayward@arm.com>
14890 David Sherwood <david.sherwood@arm.com>
14892 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14893 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14894 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14895 (vect_analyze_slp_cost): Likewise.
14896 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14897 (vect_model_load_cost): Likewise.
14899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14900 Alan Hayward <alan.hayward@arm.com>
14901 David Sherwood <david.sherwood@arm.com>
14903 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14904 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14905 from an unsigned int * to a poly_uint64_pod *.
14906 (calculate_unrolling_factor): New function.
14907 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14910 Alan Hayward <alan.hayward@arm.com>
14911 David Sherwood <david.sherwood@arm.com>
14913 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14914 from an unsigned int to a poly_uint64.
14915 (_loop_vec_info::slp_unrolling_factor): Likewise.
14916 (_loop_vec_info::vectorization_factor): Change from an int
14918 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14919 (vect_get_num_vectors): New function.
14920 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14921 (vect_get_num_copies): Use vect_get_num_vectors.
14922 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14923 to an unsigned int *.
14924 (vect_analyze_data_refs): Change min_vf from an int * to a
14926 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14927 than an unsigned HOST_WIDE_INT.
14928 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14929 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14930 to an unsigned int *.
14931 (vect_analyze_data_ref_dependences): Likewise.
14932 (vect_compute_data_ref_alignment): Handle polynomial vf.
14933 (vect_enhance_data_refs_alignment): Likewise.
14934 (vect_prune_runtime_alias_test_list): Likewise.
14935 (vect_shift_permute_load_chain): Likewise.
14936 (vect_supportable_dr_alignment): Likewise.
14937 (dependence_distance_ge_vf): Take the vectorization factor as a
14938 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14939 (vect_analyze_data_refs): Change min_vf from an int * to a
14941 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14942 vfm1 as a poly_uint64 rather than an int. Make the same change
14943 for the returned bound_scalar.
14944 (vect_gen_vector_loop_niters): Handle polynomial vf.
14945 (vect_do_peeling): Likewise. Update call to
14946 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14947 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14949 * tree-vect-loop.c (vect_determine_vectorization_factor)
14950 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14951 (vect_get_known_peeling_cost): Likewise.
14952 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14953 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14954 (vect_transform_loop): Likewise. Use the lowest possible VF when
14955 updating the upper bounds of the loop.
14956 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14957 rather than an int.
14958 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14959 polynomial unroll factors.
14960 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14961 (vect_make_slp_decision): Likewise.
14962 (vect_supported_load_permutation_p): Likewise, and polynomial
14964 (vect_analyze_slp_cost): Handle polynomial vf.
14965 (vect_slp_analyze_node_operations): Likewise.
14966 (vect_slp_analyze_bb_1): Likewise.
14967 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14968 than an unsigned HOST_WIDE_INT.
14969 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14970 (vectorizable_load): Handle polynomial vf.
14971 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14973 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14976 Alan Hayward <alan.hayward@arm.com>
14977 David Sherwood <david.sherwood@arm.com>
14979 * match.pd: Handle bit operations involving three constants
14980 and try to fold one pair.
14982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14984 * tree-vect-loop-manip.c: Include gimple-fold.h.
14985 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14986 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14987 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14988 Add a path that uses a step of VF instead of 1, but disable it
14990 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14991 and niters_no_overflow parameters. Update calls to
14992 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14993 Create a new SSA name if the latter choses to use a ste other
14994 than zero, and return it via niters_vector_mult_vf_var.
14995 * tree-vect-loop.c (vect_transform_loop): Update calls to
14996 vect_do_peeling, vect_gen_vector_loop_niters and
14997 slpeel_make_loop_iterate_ntimes.
14998 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14999 (vect_gen_vector_loop_niters): Update declarations after above changes.
15001 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15003 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15004 128-bit round to integer instructions.
15005 (ceil<mode>2): Likewise.
15006 (btrunc<mode>2): Likewise.
15007 (round<mode>2): Likewise.
15009 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15011 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15012 unaligned VSX load/store on P8/P9.
15013 (expand_block_clear): Allow the use of unaligned VSX
15014 load/store on P8/P9.
15016 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15018 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15020 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15021 swap associated with both a load and a store.
15023 2018-01-02 Andrew Waterman <andrew@sifive.com>
15025 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15026 * config/riscv/riscv.md (clear_cache): Use it.
15028 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15030 * web.c: Remove out-of-date comment.
15032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15034 * expr.c (fixup_args_size_notes): Check that any existing
15035 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15036 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15037 (emit_single_push_insn): ...here.
15039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15041 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15042 (const_vector_encoded_nelts): New function.
15043 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15044 (const_vector_int_elt, const_vector_elt): Declare.
15045 * emit-rtl.c (const_vector_int_elt_1): New function.
15046 (const_vector_elt): Likewise.
15047 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15048 of CONST_VECTOR_ELT.
15050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15052 * expr.c: Include rtx-vector-builder.h.
15053 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15054 directly on the tree encoding.
15055 (const_vector_from_tree): Likewise.
15056 * optabs.c: Include rtx-vector-builder.h.
15057 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15058 sequence of "u" values.
15059 * vec-perm-indices.c: Include rtx-vector-builder.h.
15060 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15061 directly on the vec_perm_indices encoding.
15063 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15065 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15066 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15067 * rtx-vector-builder.h: New file.
15068 * rtx-vector-builder.c: Likewise.
15069 * rtl.h (rtx_def::u2): Add a const_vector field.
15070 (CONST_VECTOR_NPATTERNS): New macro.
15071 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15072 (CONST_VECTOR_DUPLICATE_P): Likewise.
15073 (CONST_VECTOR_STEPPED_P): Likewise.
15074 (CONST_VECTOR_ENCODED_ELT): Likewise.
15075 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15076 (unwrap_const_vec_duplicate): Likewise.
15077 (const_vec_series_p): Check for a non-duplicated vector encoding.
15078 Say that the function only returns true for integer vectors.
15079 * emit-rtl.c: Include rtx-vector-builder.h.
15080 (gen_const_vec_duplicate_1): Delete.
15081 (gen_const_vector): Call gen_const_vec_duplicate instead of
15082 gen_const_vec_duplicate_1.
15083 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15084 (gen_const_vec_duplicate): Use rtx_vector_builder.
15085 (gen_const_vec_series): Likewise.
15086 (gen_rtx_CONST_VECTOR): Likewise.
15087 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15088 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15089 Build a new vector rather than modifying a CONST_VECTOR in-place.
15090 (handle_special_swappables): Update call accordingly.
15091 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15092 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15093 Build a new vector rather than modifying a CONST_VECTOR in-place.
15094 (handle_special_swappables): Update call accordingly.
15096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15098 * simplify-rtx.c (simplify_const_binary_operation): Use
15099 CONST_VECTOR_ELT instead of XVECEXP.
15101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15103 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15104 the selector elements to be different from the data elements
15105 if the selector is a VECTOR_CST.
15106 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15107 ssizetype for the selector.
15109 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15111 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15112 before testing each element individually.
15113 * tree-vect-generic.c (lower_vec_perm): Likewise.
15115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15117 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15118 * selftest-run-tests.c (selftest::run_tests): Call it.
15119 * vector-builder.h (vector_builder::operator ==): New function.
15120 (vector_builder::operator !=): Likewise.
15121 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15122 (vec_perm_indices::all_from_input_p): New function.
15123 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15124 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15125 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15126 instead of reading the VECTOR_CST directly. Detect whether both
15127 vector inputs are the same before constructing the vec_perm_indices,
15128 and update the number of inputs argument accordingly. Use the
15129 utility functions added above. Only construct sel2 if we need to.
15131 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15133 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15134 the broadcast of the low byte.
15135 (expand_mult_highpart): Use an explicit encoding for the permutes.
15136 * optabs-query.c (can_mult_highpart_p): Likewise.
15137 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15138 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15139 (vectorizable_bswap): Likewise.
15140 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15141 explicit encoding for the power-of-2 permutes.
15142 (vect_permute_store_chain): Likewise.
15143 (vect_grouped_load_supported): Likewise.
15144 (vect_permute_load_chain): Likewise.
15146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15148 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15149 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15150 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15151 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15152 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15153 (vect_gen_perm_mask_any): Likewise.
15155 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15157 * int-vector-builder.h: New file.
15158 * vec-perm-indices.h: Include int-vector-builder.h.
15159 (vec_perm_indices): Redefine as an int_vector_builder.
15160 (auto_vec_perm_indices): Delete.
15161 (vec_perm_builder): Redefine as a stand-alone class.
15162 (vec_perm_indices::vec_perm_indices): New function.
15163 (vec_perm_indices::clamp): Likewise.
15164 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15165 (vec_perm_indices::new_vector): New function.
15166 (vec_perm_indices::new_expanded_vector): Update for new
15167 vec_perm_indices class.
15168 (vec_perm_indices::rotate_inputs): New function.
15169 (vec_perm_indices::all_in_range_p): Operate directly on the
15170 encoded form, without computing elided elements.
15171 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15172 encoding. Update for new vec_perm_indices class.
15173 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15174 the given vec_perm_builder.
15175 (expand_vec_perm_var): Update vec_perm_builder constructor.
15176 (expand_mult_highpart): Use vec_perm_builder instead of
15177 auto_vec_perm_indices.
15178 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15179 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15180 or double series encoding as appropriate.
15181 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15182 vec_perm_indices instead of auto_vec_perm_indices.
15183 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15184 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15185 (vect_permute_store_chain): Likewise.
15186 (vect_grouped_load_supported): Likewise.
15187 (vect_permute_load_chain): Likewise.
15188 (vect_shift_permute_load_chain): Likewise.
15189 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15190 (vect_transform_slp_perm_load): Likewise.
15191 (vect_schedule_slp_instance): Likewise.
15192 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15193 (vectorizable_mask_load_store): Likewise.
15194 (vectorizable_bswap): Likewise.
15195 (vectorizable_store): Likewise.
15196 (vectorizable_load): Likewise.
15197 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15198 vec_perm_indices instead of auto_vec_perm_indices. Use
15199 tree_to_vec_perm_builder to read the vector from a tree.
15200 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15201 vec_perm_builder instead of a vec_perm_indices.
15202 (have_whole_vector_shift): Use vec_perm_builder and
15203 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15204 truncation to calc_vec_perm_mask_for_shift.
15205 (vect_create_epilog_for_reduction): Likewise.
15206 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15207 from auto_vec_perm_indices to vec_perm_indices.
15208 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15209 instead of changing individual elements.
15210 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15211 the vector in d.perm.
15212 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15213 from auto_vec_perm_indices to vec_perm_indices.
15214 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15215 instead of changing individual elements.
15216 (arm_vectorize_vec_perm_const): Use new_vector to install
15217 the vector in d.perm.
15218 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15219 Update vec_perm_builder constructor.
15220 (rs6000_expand_interleave): Likewise.
15221 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15222 (rs6000_expand_interleave): Likewise.
15224 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15226 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15227 to qimode could truncate the indices.
15228 * optabs.c (expand_vec_perm_var): Likewise.
15230 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15232 * Makefile.in (OBJS): Add vec-perm-indices.o.
15233 * vec-perm-indices.h: New file.
15234 * vec-perm-indices.c: Likewise.
15235 * target.h (vec_perm_indices): Replace with a forward class
15237 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15238 * optabs.h: Include vec-perm-indices.h.
15239 (expand_vec_perm): Delete.
15240 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15241 (expand_vec_perm_const): Declare.
15242 * target.def (vec_perm_const_ok): Replace with...
15243 (vec_perm_const): ...this new hook.
15244 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15245 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15246 * doc/tm.texi: Regenerate.
15247 * optabs.def (vec_perm_const): Delete.
15248 * doc/md.texi (vec_perm_const): Likewise.
15249 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15250 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15251 expand_vec_perm for constant permutation vectors. Assert that
15252 the mode of variable permutation vectors is the integer equivalent
15253 of the mode that is being permuted.
15254 * optabs-query.h (selector_fits_mode_p): Declare.
15255 * optabs-query.c: Include vec-perm-indices.h.
15256 (selector_fits_mode_p): New function.
15257 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15258 is defined, instead of checking whether the vec_perm_const_optab
15259 exists. Use targetm.vectorize.vec_perm_const instead of
15260 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15261 fit in the vector mode before using a variable permute.
15262 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15263 vec_perm_indices instead of an rtx.
15264 (expand_vec_perm): Replace with...
15265 (expand_vec_perm_const): ...this new function. Take the selector
15266 as a vec_perm_indices rather than an rtx. Also take the mode of
15267 the selector. Update call to shift_amt_for_vec_perm_mask.
15268 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15269 Use vec_perm_indices::new_expanded_vector to expand the original
15270 selector into bytes. Check whether the indices fit in the vector
15271 mode before using a variable permute.
15272 (expand_vec_perm_var): Make global.
15273 (expand_mult_highpart): Use expand_vec_perm_const.
15274 * fold-const.c: Includes vec-perm-indices.h.
15275 * tree-ssa-forwprop.c: Likewise.
15276 * tree-vect-data-refs.c: Likewise.
15277 * tree-vect-generic.c: Likewise.
15278 * tree-vect-loop.c: Likewise.
15279 * tree-vect-slp.c: Likewise.
15280 * tree-vect-stmts.c: Likewise.
15281 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15283 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15284 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15285 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15286 (aarch64_vectorize_vec_perm_const): ...this new function.
15287 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15288 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15289 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15290 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15291 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15292 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15293 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15295 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15296 check for NEON modes.
15297 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15298 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15299 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15300 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15302 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15303 the old VEC_PERM_CONST conditions.
15304 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15305 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15306 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15307 (ia64_vectorize_vec_perm_const_ok): Merge into...
15308 (ia64_vectorize_vec_perm_const): ...this new function.
15309 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15310 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15311 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15312 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15313 * config/mips/mips.c (mips_expand_vec_perm_const)
15314 (mips_vectorize_vec_perm_const_ok): Merge into...
15315 (mips_vectorize_vec_perm_const): ...this new function.
15316 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15317 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15318 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15319 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15320 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15321 (rs6000_expand_vec_perm_const): Delete.
15322 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15324 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15325 (altivec_expand_vec_perm_const_le): Take each operand individually.
15326 Operate on constant selectors rather than rtxes.
15327 (altivec_expand_vec_perm_const): Likewise. Update call to
15328 altivec_expand_vec_perm_const_le.
15329 (rs6000_expand_vec_perm_const): Delete.
15330 (rs6000_vectorize_vec_perm_const_ok): Delete.
15331 (rs6000_vectorize_vec_perm_const): New function.
15332 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15333 an element count and rtx array.
15334 (rs6000_expand_extract_even): Update call accordingly.
15335 (rs6000_expand_interleave): Likewise.
15336 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15337 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15338 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15339 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15340 (rs6000_expand_vec_perm_const): Delete.
15341 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15342 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15343 (altivec_expand_vec_perm_const_le): Take each operand individually.
15344 Operate on constant selectors rather than rtxes.
15345 (altivec_expand_vec_perm_const): Likewise. Update call to
15346 altivec_expand_vec_perm_const_le.
15347 (rs6000_expand_vec_perm_const): Delete.
15348 (rs6000_vectorize_vec_perm_const_ok): Delete.
15349 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15350 reference to the SPE evmerge intructions.
15351 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15352 an element count and rtx array.
15353 (rs6000_expand_extract_even): Update call accordingly.
15354 (rs6000_expand_interleave): Likewise.
15355 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15356 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15358 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15362 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15363 vector mode and that that mode matches the mode of the data
15365 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15366 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15367 directly using expand_vec_perm_1 when forcing selectors into
15369 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15373 * optabs-query.h (can_vec_perm_p): Delete.
15374 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15375 * optabs-query.c (can_vec_perm_p): Split into...
15376 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15377 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15378 particular selector is valid.
15379 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15380 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15381 (vect_grouped_load_supported): Likewise.
15382 (vect_shift_permute_load_chain): Likewise.
15383 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15384 (vect_transform_slp_perm_load): Likewise.
15385 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15386 (vectorizable_bswap): Likewise.
15387 (vect_gen_perm_mask_checked): Likewise.
15388 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15389 implementations of variable permutation vectors into account
15390 when deciding which selector to use.
15391 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15392 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15393 with a false third argument.
15394 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15395 to test whether the constant selector is valid and can_vec_perm_var_p
15396 to test whether a variable selector is valid.
15398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15400 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15401 * optabs-query.c (can_vec_perm_p): Likewise.
15402 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15403 instead of vec_perm_indices.
15404 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15405 (vect_gen_perm_mask_checked): Likewise,
15406 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15407 (vect_gen_perm_mask_checked): Likewise,
15409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15411 * optabs-query.h (qimode_for_vec_perm): Declare.
15412 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15413 (qimode_for_vec_perm): ...this new function.
15414 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15416 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15418 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15419 does not have a conditional at the top.
15421 2018-01-02 Richard Biener <rguenther@suse.de>
15423 * ipa-inline.c (big_speedup_p): Fix expression.
15425 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15428 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15431 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15435 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15436 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15437 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15438 cond_taken_branch_cost 3->4.
15440 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15442 PR tree-optimization/83581
15443 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15444 TODO_cleanup_cfg if any changes have been made.
15446 PR middle-end/83608
15447 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15448 convert_modes if target mode has the right side, but different mode
15451 PR middle-end/83609
15452 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15453 last argument when extracting from CONCAT. If either from_real or
15454 from_imag is NULL, use expansion through memory. If result is not
15455 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15456 the parts directly to inner mode, if even that fails, use expansion
15459 PR middle-end/83623
15460 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15461 check for bswap in mode rather than HImode and use that in expand_unop
15464 Copyright (C) 2018 Free Software Foundation, Inc.
15466 Copying and distribution of this file, with or without modification,
15467 are permitted in any medium without royalty provided the copyright
15468 notice and this notice are preserved.