| blob | 6d5a9b0dc9f1036d51a4fa50de659cba9b30e6f2 |
1 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4 track if we pass or return IEEE 128-bit floating point.
5 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6 C++ mangling that is compatible with GCC 8.1.
7 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9 floating point types.
10 (rs6000_function_arg_advance_1): Likewise.
11 (rs6000_mangle_type): Optionally generate mangled names that match
12 what GCC 8.1 generated for IEEE 128-bit floating point types.
13 (rs6000_globalize_decl_name): If we have an external function that
14 passes or returns IEEE 128-bit types, generate a weak reference
15 from the mangled name used in GCC 8.1 to the current mangled
16 name.
17 (rs6000_init_builtins): Make __ibm128 use the long double type if
18 long double is IBM extended double. Make __float128 use the long
19 double type if long double is IEEE 128-bit.
21 PR target/85657
22 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
23 macro for __ibm128 built-in functions.
24 (PACK_IF): Add __ibm128 pack/unpack functions.
25 (UNPACK_IF): Likewise.
26 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
27 enable long double built-in functions if long double is IEEE
28 128-bit floating point.
29 (rs6000_invalid_builtin): Update long double built-in function
30 error message.
31 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
32 functions, adjust the built-in function to use the long double
33 built-in function if __ibm128 and long double are the same type.
34 * doc/extend.texi (PowerPC builtins): Update documention for
35 __builtin_{,un}pack_longdouble. Add documentation for
36 __builtin_{,un}pack_ibm128.
38 2018-06-06 Jim Wilson <jimw@sifive.com>
40 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
41 (struct machine_function): New field interrupt_mode.
42 (riscv_handle_type_attribute): New function. Add forward declaration.
43 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
44 (riscv_expand_epilogue): Check interrupt_mode field.
45 (riscv_set_current_function): Check interrupt attribute args and
46 set interrupt_mode field.
47 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
48 (riscv_sret, riscv_uret): New.
49 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
50 new arguments to interrupt attribute.
52 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
54 PR target/63177
55 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
56 Don't handle -mcpu=power8 if -mpower9-vector is also used.
58 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
60 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
61 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
62 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
63 several redundant entries.
65 2018-06-06 David Malcolm <dmalcolm@redhat.com>
67 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
68 type from "rtx" to "rtx_insn *".
69 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
70 for local "call_insn", removing cast.
71 (ix86_expand_call): Likewise, introducing a "call_insn" local.
73 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
75 PR tree-optimization/86066
76 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
77 for BIT_INSERT_EXPR stores.
79 2018-06-06 Richard Biener <rguenther@suse.de>
81 PR tree-optimization/86062
82 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
83 component refs ontop
84 of to be offsetted base.
86 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
88 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
89 to be static and remove check on interrupt attribute name.
91 gcc/testsuite/gcc.target/msp430/
92 * function-attributes-4.c: New test.
93 * static-interrupts.c: New test.
95 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
97 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
98 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
100 2018-06-05 Steve Ellcey <sellcey@cavium.com>
102 PR target/79924
103 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
104 second argument.
105 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
106 Remove second argument, change how error is called.
107 (aarch64_layout_arg): Remove second argument from
108 aarch64_err_no_fpadvsimd call.
109 (aarch64_init_cumulative_args): Ditto.
110 (aarch64_gimplify_va_arg_expr): Ditto.
111 * config/aarch64/aarch64.md (mov<mode>): Ditto.
113 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
115 * config/i386/i386.md (simple_return_indirect_internal): New expander.
116 (*simple_return_indirect_internal<mode>): Rename from
117 simple_return_indirect_internal. Use W mode iterator.
118 (rstorssp): New expander.
119 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
120 (clrssbsy): New expander.
121 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
123 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
125 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
126 __typeof__.
127 (cmse_check_pointed_object): Likewise.
129 2018-06-05 Martin Liska <mliska@suse.cz>
131 PR gcov-profile/47618
132 * doc/invoke.texi: Document how -fprofile-dir format
133 is extended.
135 2018-06-05 Richard Biener <rguenther@suse.de>
137 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
138 removal pretend DOM info isn't available so we do not update
139 it and only remove edges, not dominated blocks. Actually free
140 DOM info in case we removed something. Remove unreachable blocks.
141 (mfb_keep_latches): Work with either DOM info or marked backedges.
142 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
143 first. Mark backedges if DOM info isn't available.
144 (Re-)compute DOM info after cleanup_control_flow_pre.
146 2018-06-05 Richard Biener <rguenther@suse.de>
148 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
149 (locus_discrim_hasher::hash): Adjust.
150 (locus_discrim_hasher::equal): Likewise.
151 (next_discriminator_for_locus): Work on line directly.
152 (same_line_p): Pass in expanded locus1 as well.
153 (assign_discriminators): Avoid redundant location expansions.
155 2018-06-05 Richard Biener <rguenther@suse.de>
157 PR tree-optimization/86046
158 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
159 if required after clearing TREE_ADDRESSABLE.
161 2018-06-05 Richard Biener <rguenther@suse.de>
163 PR tree-optimization/86047
164 * tree-ssa-loop.c (for_each_index): Glob handling of all
165 decls and constants and really handle all of them.
167 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
169 PR target/81497
170 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
171 qualifier_void_pointer and qualifier_const_void_pointer.
172 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
173 (arm_init_builtins): Handle the above.
174 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
175 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
176 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
177 void intrinsics.
179 2018-06-05 Martin Liska <mliska@suse.cz>
181 * auto-profile.c (read_autofdo_file): Do not use
182 gcov_ctr_summary struct.
183 (afdo_callsite_hot_enough_for_early_inline): Likewise.
184 * coverage.c (struct counts_entry): Likewise.
185 (read_counts_file): Read just single summary entry.
186 (get_coverage_counts): Use gcov_summary struct.
187 * coverage.h (get_coverage_counts): Likewise.
188 * gcov-dump.c (dump_working_sets): Likewise.
189 (tag_summary): Dump just single summary.
190 * gcov-io.c (gcov_write_summary): Write just histogram
191 summary.
192 (gcov_read_summary): Read just single summary.
193 (compute_working_sets): Use gcov_summary struct.
194 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
195 of GCOV_COUNTERS_SUMMABLE.
196 (GCOV_COUNTERS_SUMMABLE): Remove.
197 (GCOV_FIRST_VALUE_COUNTER): Replace with
198 GCOV_COUNTER_V_INTERVAL.
199 (struct gcov_ctr_summary): Remove.
200 (struct gcov_summary): Directly use fields of former
201 gcov_ctr_summary.
202 (compute_working_sets): Use gcov_summary struct.
203 * gcov.c (read_count_file): Do not use ctrs fields.
204 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
205 struct.
206 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
207 struct.
208 * profile.c: Likewise.
209 * profile.h: Likewise.
211 2018-06-05 Martin Liska <mliska@suse.cz>
213 PR gcov-profile/84846
214 * gcov.c (output_lines): Print working directory only
215 in intermediate format.
217 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
219 * config/s390/s390-builtin-types.def: Add void function type.
220 * config/s390/s390-builtins.def: Use the function type for the
221 tbeginc builtin.
223 2018-06-04 Jim Wilson <jimw@sifive.com>
225 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
226 to int.
227 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
228 and maybe_eh_return. Change regno to unsigned int. Use new args to
229 handle EH_RETURN_DATA_REGNO registers properly.
230 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
231 (riscv_expand_epilogue): Update comment. Change argument name and
232 type. Update code to use new name and type. Pass new args to
233 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
234 EXCEPTION_RETURN.
235 * config/riscv/riscv.md (NORMAL_RETURN): New.
236 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
237 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
238 (eh_return): Call gen_eh_return_internal and emit barrier.
239 (eh_return_internal): Call riscv_expand_epilogue.
241 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
243 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
244 bit_insertion field and declare can_be_merged_into method.
245 (merged_store_group::can_be_merged_into): New method.
246 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
247 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
248 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
250 2018-06-04 Richard Biener <rguenther@suse.de>
252 PR tree-optimization/85955
253 * builtins.c (fold_builtin_sincos): Convert pointers to
254 destination to appropriate type before dereferencing.
256 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
258 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
260 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
262 * expr.c (expand_expr_real_1): Force the operand into memory if
263 its TYPE_MODE is BLKmode and if there is no integer mode for
264 the number of bits being extracted.
266 2018-06-04 Jakub Jelinek <jakub@redhat.com>
268 PR target/85832
269 PR target/86036
270 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
271 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
273 2018-06-04 Richard Biener <rguenther@suse.de>
275 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
276 (cleanup_tree_cfg_noloop): ... single caller. Do
277 start_recording_case_labels later.
279 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
281 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
282 to _IMMINTRIN_H_INCLUDED.
283 * config/i386/pconfigintrin.h: Ditto.
284 * config/i386/waitpkgintrin.h: Ditto.
285 * config/i386/immintrin.h: Add includes for sgxintrin.h,
286 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
287 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
288 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
289 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
290 waitpkgintrin.h and cldemoteintrin.h.
292 2018-06-04 Richard Biener <rguenther@suse.de>
294 PR tree-optimization/86038
295 * tracer.c (find_best_successor): Check probability for
296 being initialized, bail out if not.
298 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
300 PR target/86003
301 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
302 of bits to ignore when comparing architectures.
304 2018-06-04 Jakub Jelinek <jakub@redhat.com>
306 PR tree-optimization/69615
307 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
308 maximum or minimum of the type, try to merge it also as if
309 range1 is + [-, x - 1] or + [x + 1, -].
311 PR c++/86025
312 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
314 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
316 PR tree-optimization/86034
317 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
318 the unsigned bitfield type in a bit insertion sequence if it does not
319 have a larger precision than the bitfield size.
320 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
322 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
324 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
326 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
328 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
329 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
330 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
331 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
333 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
335 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
336 Disable -fdelete-null-pointer-checks for ELF toolchain.
338 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
339 Kito Cheng <kito.cheng@gmail.com>
341 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
342 (nds32le-*-*, nds32be-*-*): Integrate checking process.
343 (nds32*-*-*): Add glibc and uclibc conditions.
344 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
345 (TARGET_EXCEPT_UNWIND_INFO): Define.
346 * config/nds32/elf.h: New file.
347 * config/nds32/linux.h: New file.
348 * config/nds32/nds32-elf.opt: New file.
349 * config/nds32/nds32-linux.opt: New file.
350 * config/nds32/nds32-fp-as-gp.c
351 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
352 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
353 TARGET_LINUX_ABI.
354 (nds32_asm_file_end): Ditto.
355 (nds32_print_operand): Ditto.
356 (nds32_insert_attributes): Ditto.
357 (nds32_init_libfuncs): New function.
358 (TARGET_HAVE_TLS): Define.
359 (TARGET_INIT_LIBFUNCS): Define.
360 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
361 spec content.
362 (TARGET_ELF): Apply different mcmodel setting.
363 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
364 been migrated into elf.h and linux.h files.
365 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
366 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
367 (mcmodel): The content has been migrated into nds32-elf.opt and
368 nds32-linux.opt files.
369 * config/nds32/t-elf: New file.
370 * config/nds32/t-linux: New file.
372 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
373 Shiva Chen <shiva0217@gmail.com>
375 * config/nds32/constants.md (unspec_volatile_element): Add
376 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
377 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
378 optimization.
379 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
380 (make_pass_nds32_fp_as_gp): Declare.
381 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
382 optmization pass.
383 (nds32_asm_function_end_prologue): Remove unused asm output.
384 (nds32_asm_function_begin_epilogue): Remove unused asm output.
385 (nds32_asm_file_start): Output necessary fp_as_gp information.
386 (nds32_option_override): Adjust register usage.
387 (nds32_expand_prologue): Consider fp_as_gp situation.
388 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
389 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
390 (epilogue): Ditto.
391 (return): Ditto.
392 (simple_return): Ditto.
393 (omit_fp_begin): Output special directive for fp_as_gp.
394 (omit_fp_end): Output special directive for fp_as_gp.
395 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
396 mforbid-fp-as-gp): New options.
398 2018-06-01 Mark Wielaard <mark@klomp.org>
400 * dwarf2out.c (dwarf2out_finish): Remove generation of
401 DW_AT_loclists_base.
403 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
405 * gimple-ssa-store-merging.c: Include gimple-fold.h.
406 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
407 (struct merged_store_group): Add bit_insertion field.
408 (dump_char_array): Use standard hexadecimal format.
409 (merged_store_group::merged_store_group): Set bit_insertion to false.
410 (merged_store_group::apply_stores): Use optimal buffer size. Deal
411 with BIT_INSERT_EXPR stores. Move up code updating the mask and
412 also print the mask in the dump file.
413 (pass_store_merging::gate): Minor tweak.
414 (imm_store_chain_info::coalesce_immediate): Fix wrong association
415 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
416 stores with INTEGER_CST stores.
417 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
418 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
419 and use it throughout. Generate bit insertion sequences if need be.
420 (pass_store_merging::process_store): Remove redundant condition.
421 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
423 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
425 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
426 the 128-bit floating point types. Fix function comment.
428 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
430 * config/aarch64/aarch64-simd.md
431 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
432 mnemonics.
433 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
434 mnemonics.
436 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
438 PR tree-optimization/85989
439 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
440 variable.
441 (backprop::intersect_uses): Check it when deciding whether this
442 is a backedge reference.
443 (backprop::process_block): Add each phi to m_visited_phis
444 after visiting it, then clear it at the end.
446 2018-06-01 Richard Biener <rguenther@suse.de>
448 * tree-vectorizer.h (vect_dr_stmt): New function.
449 (vect_get_load_cost): Adjust.
450 (vect_get_store_cost): Likewise.
451 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
452 Use vect_dr_stmt instead of DR_SMTT.
453 (vect_record_base_alignments): Likewise.
454 (vect_calculate_target_alignment): Likewise.
455 (vect_compute_data_ref_alignment): Likewise and make static.
456 (vect_update_misalignment_for_peel): Likewise.
457 (vect_verify_datarefs_alignment): Likewise.
458 (vector_alignment_reachable_p): Likewise.
459 (vect_get_data_access_cost): Likewise. Pass down
460 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
461 (vect_get_peeling_costs_all_drs): Likewise.
462 (vect_peeling_hash_get_lowest_cost): Likewise.
463 (vect_enhance_data_refs_alignment): Likewise.
464 (vect_find_same_alignment_drs): Likewise.
465 (vect_analyze_data_refs_alignment): Likewise.
466 (vect_analyze_group_access_1): Likewise.
467 (vect_analyze_group_access): Likewise.
468 (vect_analyze_data_ref_access): Likewise.
469 (vect_analyze_data_ref_accesses): Likewise.
470 (vect_vfa_segment_size): Likewise.
471 (vect_small_gap_p): Likewise.
472 (vectorizable_with_step_bound_p): Likewise.
473 (vect_prune_runtime_alias_test_list): Likewise.
474 (vect_analyze_data_refs): Likewise.
475 (vect_supportable_dr_alignment): Likewise.
476 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
477 (vect_gen_prolog_loop_niters): Likewise.
478 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
479 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
480 modify DR_STMT.
481 (vect_recog_mask_conversion_pattern): Likewise.
482 (vect_try_gather_scatter_pattern): Likewise.
483 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
484 to vect_get_store_cost.
485 (vect_get_store_cost): Get stmt_info instead of DR.
486 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
487 (vect_get_load_cost): Get stmt_info instead of DR.
489 2018-06-01 Richard Biener <rguenther@suse.de>
491 PR middle-end/86017
492 * gimple-fold.c (var_decl_component_p): Also allow offsetted
493 vars wrapped in MEM_REFs.
495 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
497 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
498 Fix subreg tests so that we only return a choice between
499 GENERAL_REGS and FP_REGS if the original classes included both.
501 2018-06-01 Richard Biener <rguenther@suse.de>
503 PR ipa/85960
504 * tree-ssa-structalias.c (get_function_part_constraint):
505 Handle NULL fi->decl.
506 (find_func_aliases_for_call): Properly handle indirect
507 fi from direct call.
508 (find_func_clobbers): Likewise.
509 (ipa_pta_execute): Likewise.
510 (create_variable_info_for): For functions that are ifunc_resolver
511 resolve to a varinfo that contains the result of the resolver call.
512 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
513 aliases.
515 2018-05-31 Michael Collison <michael.collison@arm.com>
517 * config/aarch64/aarch64.md:
518 (*fix_to_zero_extenddfdi2): New pattern.
519 * gcc.target/aarch64/fix_extend1.c: New testcase.
521 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
523 PR middle-end/78809
524 PR middle-end/83026
525 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
526 and BUILT_IN_STRNCMP_EQ.
527 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
528 BUILT_IN_STRNCMP_EQ.
529 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
530 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
531 (gimple_fold_builtin): Likewise.
532 * tree-ssa-strlen.c (compute_string_length): New function.
533 (determine_min_obsize): New function.
534 (handle_builtin_string_cmp): New function to handle calls to
535 string compare functions.
536 (strlen_optimize_stmt): Add handling to builtin string compare
537 calls.
538 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
539 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
540 * tree.c (build_common_builtin_nodes): Add new defines of
541 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
543 2018-05-31 Jakub Jelinek <jakub@redhat.com>
545 PR target/85984
546 * bb-reorder.c (pass_partition_blocks::gate): Return false for
547 functions with naked attribute.
549 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
551 * config/i386/sse.md (avx_vec_concat<mode>):
552 Substitute concat_tg_mode mode attribute with xtg_mode.
553 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
554 (concat_tg_mode): Remove mode attribute.
556 2018-05-31 Martin Sebor <msebor@redhat.com>
558 PR c/82063
559 * calls.c (alloc_max_size): Correct a logic error/typo.
560 Treat excessive arguments as infinite. Warn for invalid arguments.
561 * doc/invoke.texi (-Walloc-size-larger-than): Update.
563 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
565 PR target/85829
566 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
567 and movx for Haswell.
569 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
570 Cesar Philippidis <cesar@codesourcery.com>
572 PR middle-end/85879
573 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
574 when emitting error on private/firstprivate reductions.
575 * omp-low.c (lower_omp_target): Avoid reference-type processing
576 on pointers for firstprivate clause.
578 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
580 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
581 (st1x2): Likewise.
582 (st1x3): Likewise.
583 * config/aarch64/aarch64-simd.md
584 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
585 (aarch64_ld1_x3_<mode>): Likewise
586 (aarch64_st1x2<VALLDIF:mode>): Likewise
587 (aarch64_st1_x2_<mode>): Likewise
588 (aarch64_st1x3<VALLDIF:mode>): Likewise
589 (aarch64_st1_x3_<mode>): Likewise
590 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
591 (vld1_s8_x3): Likewise.
592 (vld1_u16_x3): Likewise.
593 (vld1_s16_x3): Likewise.
594 (vld1_u32_x3): Likewise.
595 (vld1_s32_x3): Likewise.
596 (vld1_u64_x3): Likewise.
597 (vld1_s64_x3): Likewise.
598 (vld1_f16_x3): Likewise.
599 (vld1_f32_x3): Likewise.
600 (vld1_f64_x3): Likewise.
601 (vld1_p8_x3): Likewise.
602 (vld1_p16_x3): Likewise.
603 (vld1_p64_x3): Likewise.
604 (vld1q_u8_x3): Likewise.
605 (vld1q_s8_x3): Likewise.
606 (vld1q_u16_x3): Likewise.
607 (vld1q_s16_x3): Likewise.
608 (vld1q_u32_x3): Likewise.
609 (vld1q_s32_x3): Likewise.
610 (vld1q_u64_x3): Likewise.
611 (vld1q_s64_x3): Likewise.
612 (vld1q_f16_x3): Likewise.
613 (vld1q_f32_x3): Likewise.
614 (vld1q_f64_x3): Likewise.
615 (vld1q_p8_x3): Likewise.
616 (vld1q_p16_x3): Likewise.
617 (vld1q_p64_x3): Likewise.
618 (vst1_s64_x2): Likewise.
619 (vst1_u64_x2): Likewise.
620 (vst1_f64_x2): Likewise.
621 (vst1_s8_x2): Likewise.
622 (vst1_p8_x2): Likewise.
623 (vst1_s16_x2): Likewise.
624 (vst1_p16_x2): Likewise.
625 (vst1_s32_x2): Likewise.
626 (vst1_u8_x2): Likewise.
627 (vst1_u16_x2): Likewise.
628 (vst1_u32_x2): Likewise.
629 (vst1_f16_x2): Likewise.
630 (vst1_f32_x2): Likewise.
631 (vst1_p64_x2): Likewise.
632 (vst1q_s8_x2): Likewise.
633 (vst1q_p8_x2): Likewise.
634 (vst1q_s16_x2): Likewise.
635 (vst1q_p16_x2): Likewise.
636 (vst1q_s32_x2): Likewise.
637 (vst1q_s64_x2): Likewise.
638 (vst1q_u8_x2): Likewise.
639 (vst1q_u16_x2): Likewise.
640 (vst1q_u32_x2): Likewise.
641 (vst1q_u64_x2): Likewise.
642 (vst1q_f16_x2): Likewise.
643 (vst1q_f32_x2): Likewise.
644 (vst1q_f64_x2): Likewise.
645 (vst1q_p64_x2): Likewise.
646 (vst1_s64_x3): Likewise.
647 (vst1_u64_x3): Likewise.
648 (vst1_f64_x3): Likewise.
649 (vst1_s8_x3): Likewise.
650 (vst1_p8_x3): Likewise.
651 (vst1_s16_x3): Likewise.
652 (vst1_p16_x3): Likewise.
653 (vst1_s32_x3): Likewise.
654 (vst1_u8_x3): Likewise.
655 (vst1_u16_x3): Likewise.
656 (vst1_u32_x3): Likewise.
657 (vst1_f16_x3): Likewise.
658 (vst1_f32_x3): Likewise.
659 (vst1_p64_x3): Likewise.
660 (vst1q_s8_x3): Likewise.
661 (vst1q_p8_x3): Likewise.
662 (vst1q_s16_x3): Likewise.
663 (vst1q_p16_x3): Likewise.
664 (vst1q_s32_x3): Likewise.
665 (vst1q_s64_x3): Likewise.
666 (vst1q_u8_x3): Likewise.
667 (vst1q_u16_x3): Likewise.
668 (vst1q_u32_x3): Likewise.
669 (vst1q_u64_x3): Likewise.
670 (vst1q_f16_x3): Likewise.
671 (vst1q_f32_x3): Likewise.
672 (vst1q_f64_x3): Likewise.
673 (vst1q_p64_x3): Likewise.
675 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
677 * config/msp430/msp430.c (msp430_output_labelref): Prepend
678 user_label_prefix to name.
680 * tree-core.h: Update comment about the format of NAME string
681 passed to handler in attribute_spec.
683 * config/msp430/msp430.md: Remove erroneous subreg expression from
684 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
685 zero_extend{q,h}isi2.
687 2018-05-30 Borislav Petkov <bp@suse.de>
689 * doc/extend.texi: Document some architecture specific
690 constraints and sort entries.
692 2018-05-30 Martin Sebor <msebor@redhat.com>
694 PR middle-end/85369
695 * builtins.c (expand_builtin_stpcpy_1): New function.
696 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
697 only if the former succeeds.
699 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
701 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
702 in saphira.
704 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
706 * doc/invoke.texi (-flinker-output): Document
708 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
710 * passes.c (ipa_write_summaries): Only modify statements if body
711 is in memory.
712 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
713 incrementally linking.
714 (ipa_passes): Likewise.
715 * lto-cgraph.c (lto_output_node): When incrementally linking do not
716 pass down resolution info.
717 * common.opt (flag_incremental_link): Update info.
718 * gcc.c (plugin specs): Turn flinker-output=* to
719 -plugin-opt=-linker-output-known
720 * toplev.c (compile_file): Also cut compilation when doing incremental
721 link.
722 * flag-types. (enum lto_partition_model): Add
723 LTO_LINKER_OUTPUT_NOLTOREL.
724 (invoke.texi): Add -flinker-output docs.
725 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
726 link same way as WPA; do not stream in dead initializers.
728 * dwarf2out.c (dwarf2out_die_ref_for_decl,
729 darf2out_register_external_decl): Support incremental link.
731 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
733 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
735 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
737 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
738 it down to simple_object_copy_lto_debug_sections.
739 (run_gcc): Determine incremental LTO link time and configure
740 lto1 into non-wpa mode, disable renaming of debug sections.
742 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
744 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
745 descriptions of various incorrectly documented functions.
747 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
749 Revert:
750 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
751 address check not strict.
753 2018-05-30 Richard Biener <rguenther@suse.de>
755 PR tree-optimization/85964
756 * tracer.c (better_p): Drop initialized count check, we only
757 call the function with initialized counts now.
758 (find_best_successor): Do find a best edge if one
759 has uninitialized count.
760 (find_best_predecessor): Likewise. Do BB frequency check only
761 if count is initialized.
763 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
765 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
766 (aarch64_ldrstr_offset_compare): New.
767 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
768 load/store orderings.
769 (aarch64_gen_adjusted_ldpstp): Likewise.
771 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
773 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
774 Check for subset of GENERAL_REGS and FP_REGS.
775 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
776 r=w alternative.
778 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
780 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
781 and wi::to_poly_offset. Add the current offset and then check
782 whether the sum fits, rather than using an unchecked addition of
783 a checked term. Check for a shwi rather than a uhwi.
784 * expr.c (get_bit_range): Use tree_to_poly_uint64.
785 (store_constructor): Use poly_int_tree_p.
786 (expand_expr_real_1): Likewise.
787 * function.c (assign_temp): Likewise.
788 * fold-const.c (const_binop): Use poly_int_tree_p and
789 wi::to_poly_offset.
790 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
791 division.
792 * ipa-icf-gimple.c (func_checker::compare_operand): Use
793 to_poly_offset for MEM offsets.
794 * ipa-icf.c (sem_variable::equals): Likewise.
795 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
796 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
797 wi::to_poly_offset for BIT_FIELD_REF offsets.
798 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
799 wi::to_poly_offset.
800 * var-tracking.c (emit_note_insn_var_location): Use
801 tree_to_poly_uint64.
803 2018-05-29 Jim Wilson <jimw@sifive.com>
805 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
807 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
809 PR target/85950
810 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
811 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
812 sequence.
813 (sse4_1_round<mode>2): Use nonimmediate_operand
814 for operand 1 predicate.
816 2018-05-29 Martin Sebor <msebor@redhat.com>
817 Richard Biener <rguenther@suse.de>
819 PR testsuite/85888
820 * calls.c (get_size_range): Call determine_value_range instead
821 of get_value_range..
822 * tree-vrp.h (determine_value_range): Declared new function.
823 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
825 2018-05-29 Richard Biener <rguenther@suse.de>
827 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
828 sure to use non-pattern stmts for get_earlier_stmt arguments.
829 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
830 called on pattern stmts.
831 (get_later_stmt): Likewise.
833 2018-05-29 Martin Liska <mliska@suse.cz>
835 PR gcov-profile/85759
836 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
837 env variables.
839 2018-05-29 Jakub Jelinek <jakub@redhat.com>
841 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
842 VEC_UNPACK_*_EXPR.
843 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
844 VEC_PACK_*_EXPR.
846 PR target/85918
847 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
848 VEC_PACK_FLOAT_EXPR): New tree codes.
849 * tree-pretty-print.c (op_code_prio): Handle
850 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
851 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
852 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
853 * tree-inline.c (estimate_operator_cost): Likewise.
854 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
855 * fold-const.c (const_binop): Likewise.
856 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
857 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
858 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
859 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
860 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
861 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
862 * expr.c (expand_expr_real_2): Likewise.
863 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
864 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
865 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
866 optabs.
867 * optabs.c (expand_widen_pattern_expr): For
868 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
869 sign from result type rather than operand's type.
870 (expand_binop_directly): For vec_packu_float_optab and
871 vec_packs_float_optab allow result type to be different from operand's
872 type.
873 * optabs-tree.c (optab_for_tree_code): Handle
874 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
875 VEC_PACK_FLOAT_EXPR. Formatting fixes.
876 * tree-vect-generic.c (expand_vector_operations_1): Handle
877 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
878 VEC_PACK_FLOAT_EXPR.
879 * tree-vect-stmts.c (supportable_widening_operation): Handle
880 FIX_TRUNC_EXPR.
881 (supportable_narrowing_operation): Handle FLOAT_EXPR.
882 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
883 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
884 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
885 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
886 mode attributes.
887 (vec_pack<floatprefix>_float_<mode>): New expander.
888 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
889 attributes.
890 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
891 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
892 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
893 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
894 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
895 Document.
896 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
897 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
898 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
899 VEC_PACK_FLOAT_EXPR): Document.
901 2018-05-29 Richard Biener <rguenther@suse.de>
903 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
904 member.
905 (stmt_vec_info_vec): Make pointer.
906 (init_stmt_vec_info_vec): Remove.
907 (free_stmt_vec_info_vec): Likewise.
908 (set_stmt_vec_info_vec): New function.
909 (free_stmt_vec_infos): Likewise.
910 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
911 (set_vinfo_for_stmt): Likewise.
912 (get_earlier_stmt): Likewise.
913 (get_later_stmt): Likewise.
914 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
915 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
916 (vec_info::~vec_info): Free stmt_vec_infos.
917 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
918 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
919 (pass_slp_vectorize::execute): Likewise.
920 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
921 (free_stmt_vec_info_vec): Likewise.
922 (set_stmt_vec_info_vec): New function.
923 (free_stmt_vec_infos): Likewise.
924 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
925 the global stmt_vec_info_vec.
926 * tree-parloops.c (gather_scalar_reductions): Use
927 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
928 vector.
930 2018-05-29 Richard Biener <rguenther@suse.de>
932 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
934 2018-05-29 Martin Liska <mliska@suse.cz>
935 David Malcolm <dmalcolm@redhat.com>
937 * vec.c (test_reverse): New.
938 (vec_c_tests): Add new test.
939 * vec.h (vl_ptr>::reverse): New function.
941 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
943 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
945 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
946 and later.
948 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
950 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
952 2018-05-28 Richard Biener <rguenther@suse.de>
954 PR tree-optimization/85933
955 * tree-vect-data-refs.c (vect_record_base_alignments): Only
956 look at stmts marked as vectorizable.
958 2018-05-28 Richard Biener <rguenther@suse.de>
960 PR tree-optimization/85934
961 * tree-vect-generic.c (expand_vector_operations_1): Hoist
962 vector boolean check before scalar optimization.
964 2018-05-28 Jakub Jelinek <jakub@redhat.com>
966 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
967 for armv5te.
969 2018-05-28 Mark Wielaard <mark@klomp.org>
971 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
972 if it is an expression containing a minus sign.
974 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
976 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
978 2018-05-27 Paul Koning <ni1d@arrl.net>
980 * config/pdp11/pdp11.md (truncsihi2): Remove.
982 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
983 Chung-Ju Wu <jasonwucj@gmail.com>
985 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
986 implementation.
987 (unaligned_store_dw): Ditto.
988 * config/nds32/nds32-memory-manipulation.c
989 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
990 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
991 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
992 (emit_setmem_word_loop): Rename to ...
993 (emit_setmem_doubleword_loop): ... this.
994 (nds32_gen_dup_4_byte_to_word_value): New function.
995 (nds32_gen_dup_8_byte_to_double_word_value): New function.
996 (nds32_expand_setmem_loop): Refine implementation.
997 (nds32_expand_setmem_loop_v3m): Ditto.
998 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
999 pattern.
1001 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1003 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1005 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1007 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1008 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1009 machine->attr_no_prologue_p.
1010 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1011 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1012 (nds32_expand_epilogue): Consider attr_naked_p.
1013 (nds32_expand_epilogue_v3pop): Likewise.
1014 (nds32_can_use_return_insn): Likewise.
1015 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1016 attr_no_prologue_p fields.
1017 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1019 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1021 PR target/85918
1022 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1023 attributes.
1024 * config/i386/sse.md
1025 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1026 Rename to ...
1027 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1028 ... this.
1029 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1030 Rename to ...
1031 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1032 ... this.
1033 (*<floatsuffix>floatv2div2sf2): Rename to ...
1034 (*float<floatunssuffix>v2div2sf2): ... this.
1035 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1036 (float<floatunssuffix>v2div2sf2_mask): ... this.
1037 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1038 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1039 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1040 to ...
1041 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1042 ... this.
1043 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1044 Rename to ...
1045 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1046 ... this.
1047 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1048 Rename to ...
1049 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1050 ... this.
1051 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1052 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1053 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1054 gen_ufix_truncv8dfv8si2.
1055 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1056 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1057 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1058 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1059 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1060 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1061 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1062 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1064 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1066 PR target/85900
1067 PR target/85345
1068 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1070 2018-05-25 Jim Wilson <jimw@sifive.com>
1072 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1073 * config/riscv/riscv.c (struct machine_function): Add
1074 interrupt_handler_p and attribute_checked_p fields.
1075 (riscv_attribute_table): Add interrupt.
1076 (riscv_interrupt_type_p): New.
1077 (riscv_save_reg_p): Save extra regs for interrupt handler.
1078 (riscv_use_save_libcall): Return false for interrupt handler.
1079 (riscv_first_stack_step): Add forward declaration.
1080 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1081 for interrupt handler with large frame. Use it for saved reg list.
1082 (riscv_expand_prologue): Move flag_stack_usage_info support to
1083 eliminate duplication.
1084 (riscv_expand_epilogue): Generate mret for interrupt handler.
1085 (riscv_epilogue_uses): New.
1086 (riscv_can_use_return_insn): Return false for interrupt handler.
1087 (riscv_function_ok_for_sibcall): Likewise.
1088 (riscv_set_current_function): Add interrupt handler support.
1089 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1090 * config/riscv/riscv.md (UNSPECV_MRET): New.
1091 (GP_REGNUM): New.
1092 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1093 (riscv_mret): New.
1094 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1096 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1098 PR tree-optimization/85712
1099 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1100 this candidate has already been replaced in-situ by a copy.
1102 2018-05-25 Jason Merrill <jason@redhat.com>
1104 PR c++/80485 - inline function non-zero address.
1105 * symtab.c (nonzero_address): Check DECL_COMDAT.
1107 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1109 PR target/83628
1110 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1111 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1112 extension of SImode operation. Use const123_operand predicate.
1113 (*saddsi_1): Remove.
1114 (*saddl_se_1): Ditto.
1115 (*ssubsi_1): Ditto.
1116 (*ssubl_se_1): Ditto.
1117 * config/alpha/predicates.md (const123_operand): New predicate.
1118 * config/alpha/constraints.md (P): Use IN_RANGE.
1120 2018-05-25 Richard Biener <rguenther@suse.de>
1122 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1123 defaulted to true.
1124 (ref_maybe_used_by_stmt_p): Likewise.
1125 (stmt_may_clobber_ref_p): Likewise.
1126 (stmt_may_clobber_ref_p_1): Likewise.
1127 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1128 and pass it along.
1129 (ref_maybe_used_by_stmt_p): Likewise.
1130 (stmt_may_clobber_ref_p): Likewise.
1131 (stmt_may_clobber_ref_p_1): Likewise.
1132 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1133 the alias oracle to disambiguate DRs with stmts DR analysis
1134 couldn't handle.
1135 (vect_analyze_data_refs): Do not give up on not analyzable
1136 DRs for BB vectorization. Remove code truncating the dataref
1137 vector.
1139 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1141 PR target/85832
1142 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1143 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1144 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1146 2018-05-25 Richard Biener <rguenther@suse.de>
1148 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1149 function, combining stmt data ref gathering and fatal analysis
1150 parts.
1151 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1152 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1153 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1154 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1155 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1157 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1159 PR tree-optimization/85720
1160 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1161 SCC if all partitions are builtins.
1162 (version_loop_by_alias_check): New parameter. Generate cancelable
1163 runtime alias check if all partitions are builtins.
1164 (distribute_loop): Update call to above function.
1166 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1168 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1169 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1170 (parm_default_def_partition_arg): Ditto.
1171 (set_parm_default_def_partition): Ditto.
1172 (get_parm_default_def_partitions): Ditto and make it static.
1173 (get_undefined_value_partitions): Ditto and make it static.
1174 (remove_ssa_form): Refactor call to init_var_map here.
1175 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1176 computation for loop region.
1177 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1178 (register_default_def): Delete.
1179 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1180 (parm_default_def_partition_arg): Ditto.
1181 (set_parm_default_def_partition): Ditto.
1182 (get_parm_default_def_partitions): Ditto and make it static.
1183 (get_undefined_value_partitions): Ditto and make it static.
1184 (coalesce_with_default, coalesce_with_default): Update comment.
1185 (create_coalesce_list_for_region): New func factored out from
1186 create_outofssa_var_map.
1187 (populate_coalesce_list_for_outofssa): New func factored out from
1188 create_outofssa_var_map and coalesce_ssa_name.
1189 (create_outofssa_var_map): Delete.
1190 (coalesce_ssa_name): Refactor to support live range computation.
1191 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1192 (get_parm_default_def_partitions): Delete.
1193 (get_undefined_value_partitions): Ditto.
1194 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1195 computation for loop region.
1196 (new_tree_live_info, loe_visit_block): Ditto.
1197 (live_worklist, set_var_live_on_entry): Ditto.
1198 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1199 * tree-ssa-live.h (struct _var_map): New fields.
1200 (init_var_map): Change decl.
1201 (region_contains_p): New.
1203 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1205 * tree-ssa-live.h (live_merge_and_clear): Delete.
1207 2018-05-25 Richard Biener <rguenther@suse.de>
1209 PR c++/85912
1210 * tree-dump.c (dequeue_and_dump): Remove access to removed
1211 operand 2 of a SWITCH_EXPR.
1213 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1215 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1216 multiplication and division.
1217 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1218 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1219 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1220 (cond_udiv_optab, cond_umod_optab): New optabs.
1221 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1222 (IFN_COND_RDIV): New internal functions.
1223 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1224 TRUNC_MOD_EXPR and RDIV_EXPR.
1225 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1226 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1227 New unspecs.
1228 (SVE_INT_BINARY): Include mult.
1229 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1230 (optab, sve_int_op): Handle mult.
1231 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1232 UNSPEC_COND_DIV.
1233 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1234 for SVE_INT_BINARY_SD.
1236 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1238 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1239 (optab, sve_int_op): Handle div and udiv.
1240 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1241 for SVE_INT_BINARY_SD.
1242 (*<optab><mode>3): New insn for the same.
1244 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1246 * tree-vect-patterns.c: Include predict.h.
1247 (vect_recog_divmod_pattern): Restrict check for division support
1248 to when optimizing for size.
1250 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1252 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1253 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1254 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1255 (gimple_match_op::set_op): Likewise.
1256 (gimple_resimplify4): Declare.
1257 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1258 (expr::gen_transform): Likewise.
1259 (decision_tree::gen): Generate a simplification routine for 4 operands.
1260 * gimple-match-head.c (gimple_simplify): Add an overload for
1261 4 operands. In the top-level function, handle up to 4 call
1262 arguments and call gimple_resimplify4.
1263 (gimple_resimplify4): New function.
1264 (build_call_internal): Pass a fourth operand.
1265 (maybe_push_to_seq): Likewise.
1266 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1267 Fold VEC_COND_EXPRs of an operation and a default value into
1268 an IFN_COND_* function if possible.
1269 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1270 New unspecs.
1271 (SVE_COND_FP_BINARY): Include them.
1272 (optab, sve_fp_op): Handle them.
1273 (SVE_INT_BINARY_REV): New code iterator.
1274 (SVE_COND_FP_BINARY_REV): New int iterator.
1275 (commutative): New int attribute.
1276 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1277 Declare.
1278 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1279 function.
1280 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1281 (*cond_<optab><mode>): New patterns for reversed operands.
1283 2018-05-25 Richard Biener <rguenther@suse.de>
1285 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1286 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1287 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1288 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1289 * tree-vect-data-refs.c (everywhere): Adjust users.
1290 * tree-vect-loop.c (everywhere): Likewise.
1291 * tree-vect-slp.c (everywhere): Likewise.
1292 * tree-vect-stmts.c (everywhere): Likewise.
1293 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1295 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1297 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1298 Rename to...
1299 (gcc_cv_as_section_exclude): ... this.
1300 Try Solaris as #exclude syntax.
1301 * configure: Regenerate.
1302 * config.in: Regenerate.
1303 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1304 SECTION_EXCLUDE.
1305 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1306 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1308 * varasm.c (default_elf_asm_named_section): Don't check if
1309 HAVE_GAS_SECTION_EXCLUDE is defined.
1311 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1313 * doc/md.texi: Update the documentation of the cond_* optabs
1314 to mention the new final operand. Fix GET_MODE_NUNITS call.
1315 Describe the scalar case too.
1316 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1317 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1318 instead of 2.
1319 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1320 (get_conditional_internal_fn): Update comment.
1321 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1322 accumulator value as a final argument to conditional functions.
1323 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1324 a define_expand and add an "else" operand. Assert for now that
1325 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1326 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1327 (*cond_<optab><mode>): New patterns.
1328 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1329 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1330 (UNSPEC_COND_EOR): Delete.
1331 (optab): Remove associated mappings.
1332 (SVE_INT_BINARY): New code iterator.
1333 (sve_int_op): Remove int attribute and add "minus" to the code
1334 attribute.
1335 (SVE_COND_INT_OP): Delete.
1336 (SVE_COND_FP_OP): Rename to...
1337 (SVE_COND_FP_BINARY): ...this.
1339 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1341 * optabs.c (can_reuse_operands_p): New function.
1342 (maybe_legitimize_operands): Try to reuse the results for
1343 earlier operands.
1345 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1347 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1348 Add {q} suffix to insn mnemonic.
1350 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1352 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1353 (msp430_warn_func_return): New.
1355 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1357 * fold-const.c (tree_nonzero_bits): New function.
1358 * fold-const.h (tree_nonzero_bits): Likewise.
1359 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1360 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1362 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1364 PR target/85900
1365 PR target/85345
1366 * varasm.c (assemble_alias): Check ifunc_resolver only on
1367 FUNCTION_DECL.
1369 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1371 PR target/85903
1372 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1373 when memory input operand is handled.
1375 2018-05-24 Luis Machado <luis.machado@linaro.org>
1377 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1378 global.
1379 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1381 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1383 * match.pd: Delay FMA folds until after vectorization.
1385 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1387 PR target/83009
1388 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1389 address check not strict.
1391 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1393 * gimple-match.h (gimple_match_op): New class.
1394 (mprts_hook): Replace parameters with a gimple_match_op *.
1395 (maybe_build_generic_op): Likewise.
1396 (gimple_simplified_result_is_gimple_val): Replace parameters with
1397 a const gimple_match_op *.
1398 (gimple_simplify): Replace code_helper * and tree * parameters with
1399 a gimple_match_op * parameter.
1400 (gimple_resimplify1): Replace code_helper *, tree and tree *
1401 parameters with a gimple_match_op * parameter.
1402 (gimple_resimplify2): Likewise.
1403 (gimple_resimplify3): Likewise.
1404 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1405 parameters with a gimple_match_op * parameter.
1406 * gimple-match-head.c (gimple_simplify): Change prototypes of
1407 auto-generated functions to take a gimple_match_op * instead of
1408 separate code_helper * and tree * parameters. Make the same
1409 change in the top-level overload and update calls to the
1410 gimple_resimplify routines. Update calls to the auto-generated
1411 functions and to maybe_push_res_to_seq in the publicly-facing
1412 operation-specific gimple_simplify overloads.
1413 (gimple_match_op::MAX_NUM_OPS): Define.
1414 (gimple_resimplify1): Replace rcode and ops with a single res_op
1415 parameter. Update call to gimple_simplify.
1416 (gimple_resimplify2): Likewise.
1417 (gimple_resimplify3): Likewise.
1418 (mprts_hook): Replace parameters with a gimple_match_op *.
1419 (maybe_build_generic_op): Likewise.
1420 (build_call_internal): Replace type, nargs and ops with
1421 a gimple_match_op *.
1422 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1423 with a single gimple_match_op *. Update calls to mprts_hook,
1424 build_call_internal and gimple_simplified_result_is_gimple_val.
1425 Factor out code that is common to the tree_code and combined_fn cases.
1426 * genmatch.c (expr::gen_transform): Replace tem_code and
1427 tem_ops with a gimple_match_op called tem_op. Update calls
1428 to the gimple_resimplify functions and maybe_push_res_to_seq.
1429 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1430 res_ops. Update call to the gimple_resimplify functions.
1431 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1432 (decision_tree::gen): Make the functions take a gimple_match_op *
1433 called res_op instead of separate res_code and res_ops parameters.
1434 Update call accordingly.
1435 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1436 and ops with a single res_op parameter. Update calls to
1437 maybe_build_generic_op and maybe_push_res_to_seq.
1438 (fold_stmt_1): Update calls to gimple_simplify and
1439 replace_stmt_with_simplification.
1440 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1441 and gimple_simplified_result_is_gimple_val.
1442 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1443 gimple_simplify.
1444 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1445 with a gimple_match_op *.
1446 (vn_nary_build_or_lookup): Likewise. Update call to
1447 vn_nary_build_or_lookup_1.
1448 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1449 gimple_match_op *. Update calls to the gimple_resimplify routines
1450 and to gimple_simplified_result_is_gimple_val.
1451 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1452 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1453 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1454 (visit_nary_op): Likewise.
1455 (visit_reference_op_load): Likewise.
1457 2018-05-23 Luis Machado <luis.machado@linaro.org>
1459 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1460 modifier for printing the step amount.
1462 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1464 PR target/78849
1465 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1466 types.
1468 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1470 * doc/sourcebuild.texi (Endianness): New subsubsection.
1472 2018-05-23 Luis Machado <luis.machado@linaro.org>
1474 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1475 <prefetch_dynamic_strides>: New const bool field.
1476 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1477 prefetch_dynamic_strides.
1478 (exynosm1_prefetch_tune): Likewise.
1479 (thunderxt88_prefetch_tune): Likewise.
1480 (thunderx_prefetch_tune): Likewise.
1481 (thunderx2t99_prefetch_tune): Likewise.
1482 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1483 false.
1484 (aarch64_override_options_internal): Update to set
1485 PARAM_PREFETCH_DYNAMIC_STRIDES.
1486 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1487 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1488 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1489 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1490 prefetch-dynamic-strides setting.
1492 2018-05-23 Luis Machado <luis.machado@linaro.org>
1494 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1495 <minimum_stride>: New const int field.
1496 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1497 minimum_stride field defaulting to -1.
1498 (exynosm1_prefetch_tune): Likewise.
1499 (thunderxt88_prefetch_tune): Likewise.
1500 (thunderx_prefetch_tune): Likewise.
1501 (thunderx2t99_prefetch_tune): Likewise.
1502 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1503 <default_opt_level>: Set to 3.
1504 (aarch64_override_options_internal): Update to set
1505 PARAM_PREFETCH_MINIMUM_STRIDE.
1506 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1507 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1508 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1509 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1510 stride is constant and is below the minimum stride threshold.
1512 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1514 * config/arm/arm-cpus.in (mode26): Delete.
1515 (armv4): Delete mode26 reference.
1516 * config/arm/arm.c (arm_configure_build_target): Delete use of
1517 isa_bit_mode26.
1519 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1521 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1522 New insn pattern.
1523 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1524 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1525 for non-SSE modes.
1526 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1527 (floatunsdidf2): Ditto.
1529 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1531 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1532 (fixuns_trunc<mode>si2_avx512f): Ditto.
1533 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1534 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1535 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1537 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1539 PR rtl-optimization/79985
1540 * df-scan.c (df_insn_refs_collect): Remove special case for
1541 global registers and asm statements.
1543 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1545 * extend.texi (Global Register Variables): Rewrite the bullet list.
1546 Note that the register is available for allocation. Note that access
1547 via inline asm must use constraints. Add note about async-signal
1548 handlers. Remove paragraph about automagic register selection.
1550 2018-05-23 Richard Biener <rguenther@suse.de>
1552 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1553 of fixed offset from memset VN.
1555 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1557 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1558 first_interp field.
1559 (alloc_cand_and_find_basis): Initialize first_interp field.
1560 (slsr_process_mul): Modify first_interp field.
1561 (slsr_process_add): Likewise.
1562 (slsr_process_cast): Modify first_interp field for each new
1563 interpretation.
1564 (slsr_process_copy): Likewise.
1565 (dump_candidate): Dump first_interp field.
1566 (replace_mult_candidate): Process all interpretations, not just
1567 subsequent ones.
1568 (replace_rhs_if_not_dup): Likewise.
1569 (replace_one_candidate): Likewise.
1571 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1573 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1574 Add new boolean.
1575 (aarch64_needs_frame_chain): New function.
1576 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1578 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1580 PR target/84882
1581 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1582 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1583 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1584 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1585 as true for strict-align.
1586 (aarch64_can_inline_p): Perform checks even when callee has no
1587 attributes to check for strict alignment.
1588 * doc/extend.texi (AArch64 Function Attributes): Document
1589 no-strict-align.
1590 * doc/invoke.texi: (AArch64 Options): Likewise.
1592 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1594 PR tree-optimization/85853
1595 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1596 the handling of the root of the node to...
1597 (vect_slp_analyze_node_operations_1): ...this new function,
1598 and run the whole thing with the child nodes' def types
1599 set according to their SLP node's def type.
1601 2018-05-23 Richard Biener <rguenther@suse.de>
1603 PR middle-end/85874
1604 * tree-data-ref.c (create_runtime_alias_checks): Defer
1605 and ignore overflow warnings.
1607 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1609 PR tree-optimization/85822
1610 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1611 constants.
1613 2018-05-23 Richard Biener <rguenther@suse.de>
1615 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1616 memset constants via native_interpret_expr.
1618 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1620 PR target/85345
1621 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1622 attribute.
1623 (cgraph_node::create_alias): Likewise.
1624 (cgraph_node::get_availability): Check ifunc_resolver instead
1625 of looking up ifunc attribute.
1626 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1627 * varasm.c (do_assemble_alias): Likewise.
1628 (assemble_alias): Likewise.
1629 (default_binds_local_p_3): Likewise.
1630 * cgraph.h (cgraph_node): Add ifunc_resolver.
1631 (cgraph_node::only_called_directly_or_aliased_p): Return false
1632 for IFUNC resolver.
1633 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1634 attribute.
1635 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1636 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1637 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1638 instead of looking up ifunc attribute.
1640 2018-05-22 Luis Machado <luis.machado@linaro.org>
1642 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1644 2018-05-22 Martin Sebor <msebor@redhat.com>
1646 PR middle-end/85359
1647 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1648 only when expasion succeeds.
1649 (expand_builtin_strcmp): Same.
1650 (expand_builtin_strncmp): Same.
1652 2018-05-22 Martin Sebor <msebor@redhat.com>
1654 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1656 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1657 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1659 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1660 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1661 code.
1662 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1663 Add check for legitimate_address.
1664 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1665 (aarch64_swap_ldrstr_operands): New.
1666 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1667 Define prototype.
1669 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1670 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1672 * config/aarch64/aarch64.md: New patterns to generate stp
1673 and ldp.
1674 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1675 single words and double words.
1676 (load_pair_sw, load_pair_dw): Likewise.
1677 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1678 Delete.
1679 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1680 Delete.
1681 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1682 for different mode ldpstp and add peephole for merged zero stores.
1683 Likewise for loads.
1684 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1685 Add size check.
1686 (aarch64_gen_store_pair): Rename calls to match new patterns.
1687 (aarch64_gen_load_pair): Rename calls to match new patterns.
1688 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1689 (load_pair<DREG:mode><DREG2:mode>): ... This.
1690 (store_pair<mode>): Rename to...
1691 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1692 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1693 New mode iterators.
1694 (V_INT_EQUIV): Handle SImode.
1695 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1696 New predicate.
1698 2018-05-22 Martin Sebor <msebor@redhat.com>
1700 PR c/85623
1701 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1702 or ajust the presumed bound on an operation to avoid unnecessary
1703 warnings.
1705 2018-05-22 Martin Sebor <msebor@redhat.com>
1707 PR tree-optimization/85826
1708 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1709 assuming that a DECL necesarily has a constant size.
1711 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1713 PR middle-end/85862
1714 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1716 2018-05-22 Richard Biener <rguenther@suse.de>
1718 PR tree-optimization/85834
1719 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1720 non-constant and non-zero memset arguments.
1722 2018-05-22 Martin Liska <mliska@suse.cz>
1724 PR ipa/85607
1725 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1727 2018-05-22 Richard Biener <rguenther@suse.de>
1729 PR tree-optimization/85863
1730 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1731 comparisons when vectype is specified.
1732 (vectorizable_condition): Do not specify vectype for
1733 vect_is_simple_cond when SLP vectorizing.
1735 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1737 PR target/85657
1738 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1739 define __ibm128 as long double.
1740 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1741 as a distinct type when IEEE 128-bit support is enabled.
1742 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1743 128-bit types to use the correct functions.
1744 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1745 convert between 128-bit floating point types that have different
1746 modes but the same representation, instead of using gen_lowpart to
1747 makean alias.
1748 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1749 KFmode.
1750 (IFKF_reg): New attributes to give the register constraints for
1751 IFmode and KFmode.
1752 (extend<mode>tf2_internal): New insns to mark an explicit
1753 conversion between 128-bit floating point types that have a
1754 different mode but share the same representation.
1756 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1758 PR tree-optimization/85814
1759 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1760 a null return from get_strinfo when unsharing the next
1761 strinfo in the chain.
1763 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1765 PR gcc/84923
1766 * varasm.c (weak_finish): Clean up weak_decls.
1768 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1770 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1771 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1772 UNSPEC_UADALP values.
1773 * config/aarch64/iterators.md (ABAL): New int iterator.
1774 (ABDL2): Likewise.
1775 (ADALP): Likewise.
1776 (sur): Add mappings for the above.
1777 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1778 New define_insn.
1779 (aarch64_<sur>abal<mode>_4): Likewise.
1780 (aarch64_<sur>adalp<mode>_3): Likewise.
1781 (<sur>sadv16qi): New define_expand.
1783 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1785 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1786 (*movdf_internal): Ditto.
1787 (*rcpsf2_sse): Ditto.
1788 (*rsqrtsf2_sse): Ditto.
1789 (*sqrt<mode>2_sse): Ditto.
1791 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1793 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1794 eor3q<mode>4.
1795 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1796 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1797 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1798 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1799 vbcaxq_s64): New.
1800 * config/aarch64/arm_neon.h: Likewise.
1801 * config/aarch64/iterators.md (VQ_I): New.
1803 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1805 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1806 arc*-*-linux*.
1807 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1808 MULTILIB_DIRNAMES
1810 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1812 * config/nds32/constraints.md (S): New constraint.
1813 * config/nds32/nds32.md (call_internal): Use constraint S.
1814 (call_value_internal): Likewise.
1815 (sibcall_internal): Likewise.
1816 (sibcall_value_internal): Likewise.
1818 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1819 Chung-Ju Wu <jasonwucj@gmail.com>
1821 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1822 into consideration.
1824 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1825 Chung-Ju Wu <jasonwucj@gmail.com>
1827 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1828 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1829 (nds32_rtx_costs_impl): Simplify.
1830 (nds32_address_cost_impl): Simplify.
1831 (nds32_init_rtx_costs): New function.
1832 (nds32_rtx_costs_speed_prefer): Likewise.
1833 (nds32_rtx_costs_size_prefer): Likewise.
1834 (nds32_address_cost_speed_prefer): Likewise.
1835 (nds32_address_cost_speed_fwprop): Likewise.
1836 (nds32_address_cost_size_prefer): Likewise.
1837 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1838 * config/nds32/nds32.c (nds32_option_override): Use
1839 nds32_init_rtx_costs function.
1841 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1844 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1845 (TARGET_PIPELINE_N8): Likewise.
1846 (TARGET_PIPELINE_N10): Likewise.
1847 (TARGET_PIPELINE_N13): Likewise.
1848 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1850 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1852 * config/nds32/nds32-fpu.md: Update copyright year.
1854 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1856 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1858 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1860 * config/nds32/nds32.c
1861 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1862 * config/nds32/nds32.opt (minline-asm-r15): New option.
1864 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1866 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1867 MASK_HW_ABS.
1868 * config/nds32/nds32.md (abssi2): New pattern.
1870 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1872 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1873 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1874 Merge insn pattern from sse_cvtsi2ss<round_name> and
1875 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1876 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1877 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1878 using SWI48 mode iterator.
1879 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1880 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1881 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1882 pattern from sse_cvttss2si<round_saeonly_name>
1883 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1884 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1885 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1886 using SWI48 mode iterator.
1887 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1888 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1889 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1890 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1891 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1892 using SWI48 mode iterator.
1893 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1894 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1895 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1896 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1897 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1898 SWI48 mode iterator.
1899 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1900 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1901 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1902 pattern from sse_cvttsd2si<round_saeonly_name>
1903 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1905 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1907 * config/nds32/nds32-md-auxiliary.c
1908 (nds32_valid_smw_lwm_base_p): Refine.
1909 (nds32_output_smw_single_word): Refine.
1910 (nds32_output_smw_double_word): New.
1911 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1913 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1915 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1916 (nds32_output_stack_pop): Refine.
1917 (nds32_expand_unaligned_load): Refine.
1918 (nds32_expand_unaligned_store): Refine.
1920 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1921 Chung-Ju Wu <jasonwucj@gmail.com>
1923 * config/nds32/constants.md: Add TP_REGNUM constant.
1924 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1925 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1926 UNSPEC_ADD32.
1927 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1928 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1929 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1930 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1931 and PIC code generation.
1932 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1933 code generation.
1934 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1935 optimization.
1936 * config/nds32/nds32.md: Support TLS and PIC.
1937 * config/nds32/nds32.c: Support TLS and PIC.
1938 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1939 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1940 predicate.
1942 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1944 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1945 mode with E_ prefix.
1947 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1948 Chung-Ju Wu <jasonwucj@gmail.com>
1950 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1951 * config/nds32/nds32-md-auxiliary.c
1952 (symbolic_reference_mentioned_p): New.
1953 (nds32_legitimize_ict_address): New.
1954 (nds32_expand_ict_move): New.
1955 (nds32_indirect_call_referenced_p): New.
1956 (nds32_symbol_binds_local_p): Delete.
1957 (nds32_long_call_p): Modify.
1958 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1959 * config/nds32/nds32-protos.h
1960 (symbolic_reference_mentioned_p): Declare.
1961 (nds32_legitimize_ict_address): Declare.
1962 (nds32_expand_ict_move): Declare.
1963 (nds32_indirect_call_referenced_p): Declare.
1964 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1965 (nds32_relax_group): Use nds32_ict_const_p as condition.
1966 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1967 (nds32_asm_file_start): Output ict_model directive in asm code.
1968 (nds32_legitimate_address_p): Consider indirect call.
1969 (nds32_print_operand): Consider indirect call.
1970 (nds32_print_operand_address): Consider indirect call.
1971 (nds32_insert_attributes): Handle "indirect_call" attribute.
1972 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1973 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1974 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1975 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1976 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1977 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1978 (TARGET_ICT_MODEL_SMALL): Define.
1979 (TARGET_ICT_MODEL_LARGE): Define.
1980 * config/nds32/nds32.md (movsi): Consider ict model.
1981 (call, call_value): Consider ict model.
1982 (sibcall, sibcall_value): Consider ict model.
1983 * config/nds32/nds32.opt (mict-model): New option.
1984 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1985 model.
1987 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1988 Monk Chiang <sh.chiang04@gmail.com>
1989 Jim Wilson <jimw@sifive.com>
1991 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1992 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1993 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1994 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1995 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1996 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1997 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1998 compute save_libcall_adjustment properly.
1999 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2000 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2001 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2002 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2003 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2004 (ABI_SPEC): Handle mabi=ilp32e.
2005 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2006 (RVE): Add RVE mask.
2007 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2008 <-march>: Add rv32e as an example.
2010 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2012 PR c++/82899
2013 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2014 (intra_create_variable_infos): Handle C++ constructors.
2016 2018-05-18 Martin Liska <mliska@suse.cz>
2018 * passes.def: Remove a redundant pass.
2020 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2022 PR bootstrap/85838
2023 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2025 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2027 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2028 (ARMv4): Update.
2029 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2030 (ARMv6m): Update.
2031 (armv2, armv2a, armv3, armv3m): Delete architectures.
2032 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2033 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2034 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2035 Delete cpus.
2036 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2037 (*mulsidi3adddi): Likewise.
2038 (mulsidi3): Likewise.
2039 (*mulsidi3_nov6): Likewise.
2040 (umulsidi3): Likewise.
2041 (umulsidi3_nov6): Likewise.
2042 (umaddsidi4): Likewise.
2043 (*umulsidi3adddi): Likewise.
2044 (smulsi3_highpart): Likewise.
2045 (*smulsi3_highpart_nov6): Likewise.
2046 (umulsi3_highpart): Likewise.
2047 (*umulsi3_highpart_nov6): Likewise.
2048 * config/arm/arm.h (arm_arch3m): Delete.
2049 * config/arm/arm.c (arm_arch3m): Delete.
2050 (arm_option_override_internal): Update armv3-related comment.
2051 (arm_configure_build_target): Delete use of isa_bit_mode32.
2052 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2053 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2054 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2055 (mulsa3): Likewise.
2056 (mulusa3): Likewise.
2057 * config/arm/arm-protos.h (arm_arch3m): Delete.
2058 * config/arm/arm-tables.opt: Regenerate.
2059 * config/arm/arm-tune.md: Likewise.
2060 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2061 deleted architectures.
2063 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2065 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2066 (armv5t, armv5te): New features.
2067 (ARMv5, ARMv5e): Delete fgroups.
2068 (ARMv5t, ARMv5te): Adjust for above changes.
2069 (ARMv6m): Likewise.
2070 (armv5, armv5e): Delete arches.
2071 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2072 arm_arch5.
2073 (*call_reg_arm): Likewise.
2074 (*call_value_reg_armv5): Likewise.
2075 (*call_value_reg_arm): Likewise.
2076 (*call_symbol): Likewise.
2077 (*call_value_symbol): Likewise.
2078 (*sibcall_insn): Likewise.
2079 (*sibcall_value_insn): Likewise.
2080 (clzsi2): Likewise.
2081 (prefetch): Likewise.
2082 (define_split and define_peephole2 dependent on arm_arch5):
2083 Likewise.
2084 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2085 arm_arch5e.
2086 (TARGET_ARM_QBIT): Likewise.
2087 (TARGET_DSP_MULTIPLY): Likewise.
2088 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2089 (arm_arch5, arm_arch5e): Delete.
2090 (arm_arch5t, arm_arch5te): Declare.
2091 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2092 (arm_arch5t): Declare.
2093 (arm_option_reconfigure_globals): Update for the above.
2094 (arm_options_perform_arch_sanity_checks): Update comment, replace
2095 use of arm_arch5 with arm_arch5t.
2096 (use_return_insn): Likewise.
2097 (arm_emit_call_insn): Likewise.
2098 (output_return_instruction): Likewise.
2099 (arm_final_prescan_insn): Likewise.
2100 (arm_coproc_builtin_available): Likewise.
2101 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2102 arm_arch5e with arm_arch5t and arm_arch5te.
2103 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2104 (arm_arch5t, arm_arch5te): Declare.
2105 * config/arm/arm-tables.opt: Regenerate.
2106 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2107 * config/arm/t-multilib: Likewise.
2108 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2109 instead of arm_arch5.
2110 (*call_reg_thumb1): Likewise.
2111 (*call_value_reg_thumb1_v5): Likewise.
2112 (*call_value_reg_thumb1): Likewise.
2113 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2114 unreachable path.
2115 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2117 2018-05-18 Martin Liska <mliska@suse.cz>
2119 PR gcov-profile/84846
2120 * doc/gcov.texi: Document -t option of gcov tool.
2122 2018-05-18 Martin Liska <mliska@suse.cz>
2124 PR gcov-profile/84846
2125 * gcov.c (print_usage): Add new -t option.
2126 (process_args): Handle the option.
2127 (generate_results): Use stdout as output when requested by
2128 the option.
2130 2018-05-18 Martin Liska <mliska@suse.cz>
2132 PR gcov-profile/84846
2133 * coverage.c (coverage_init): Write PWD to .gcno file.
2134 * doc/gcov.texi: Document how working directory is printed.
2135 * gcov-dump.c (dump_gcov_file): Print PWD.
2136 * gcov.c (output_intermediate_file): Likewise.
2137 (read_graph_file): Read PWD string.
2138 (output_lines): Print PWD.
2140 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2142 PR middle-end/85817
2143 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2144 for retval and return false if all args to phi are zero.
2146 2018-05-18 Richard Biener <rguenther@suse.de>
2148 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2149 method.
2150 (evrp_dom_walker::before_dom_children): Call it.
2152 2018-05-18 Richard Biener <rguenther@suse.de>
2154 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2155 results when processing array refs with variable index.
2157 2018-05-18 Toon Moene <toon@moene.org>
2159 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2160 directly after that of -floop-interchange. Indicate that both
2161 options are enabled by default when specifying -O3.
2163 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2165 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2166 iterator. Delete separate integer-mode vec_set<mode> expander.
2167 (aarch64_simd_vec_setv2di): Delete.
2168 (vec_setv2di): Delete.
2169 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2170 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2171 the "w, r" alternative.
2173 2018-05-18 Martin Liska <mliska@suse.cz>
2175 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2176 * tree-pass.h (make_pass_lower_switch_O0): New function.
2177 * tree-switch-conversion.c (node_has_low_bound): Remove.
2178 (node_has_high_bound): Likewise.
2179 (node_is_bounded): Likewise.
2180 (class pass_lower_switch): Make it a template type and create
2181 two instances.
2182 (pass_lower_switch::execute): Add template argument.
2183 (make_pass_lower_switch): New function.
2184 (make_pass_lower_switch_O0): New function.
2185 (do_jump_if_equal): Remove.
2186 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2187 all the hard work to tree optimization passes.
2189 2018-05-18 Martin Liska <mliska@suse.cz>
2191 * dbgcnt.c (limit_low): Renamed from limit.
2192 (limit_high): New variable.
2193 (dbg_cnt_is_enabled): Check for upper limit.
2194 (dbg_cnt): Adjust dumping.
2195 (dbg_cnt_set_limit_by_index): Add new argument for high
2196 value.
2197 (dbg_cnt_set_limit_by_name): Likewise.
2198 (dbg_cnt_process_single_pair): Parse new format.
2199 (dbg_cnt_process_opt): Use strtok.
2200 (dbg_cnt_list_all_counters): Remove 'value' and add
2201 'limit_high'.
2202 * doc/invoke.texi: Document changes.
2204 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2206 * doc/sourcebuild.texi (scalar_all_fma): Document.
2207 * tree.def (FMA_EXPR): Delete.
2208 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2209 * internal-fn.c (ternary_direct): New macro.
2210 (expand_ternary_optab_fn): Likewise.
2211 (direct_ternary_optab_supported_p): Likewise.
2212 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2213 * builtins.c (fold_builtin_fma): Delete.
2214 (fold_builtin_3): Don't call it.
2215 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2216 * expr.c (expand_expr_real_2): Likewise.
2217 * fold-const.c (operand_equal_p): Likewise.
2218 (fold_ternary_loc): Likewise.
2219 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2220 * gimple.c (DEFTREECODE): Likewise.
2221 * gimplify.c (gimplify_expr): Likewise.
2222 * optabs-tree.c (optab_for_tree_code): Likewise.
2223 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2224 * tree-eh.c (operation_could_trap_p): Likewise.
2225 (stmt_could_throw_1_p): Likewise.
2226 * tree-inline.c (estimate_operator_cost): Likewise.
2227 * tree-pretty-print.c (dump_generic_node): Likewise.
2228 (op_code_prio): Likewise.
2229 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2230 * tree-ssa-operands.c (get_expr_operands): Likewise.
2231 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2232 * fold-const-call.h (fold_fma): Delete.
2233 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2234 CFN_FNMA and CFN_FNMS.
2235 (fold_fma): Delete.
2236 * genmatch.c (combined_fn): New enum.
2237 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2238 (commutative_op): New function.
2239 (commutate): Use it. Handle more than 2 operands.
2240 (dt_operand::gen_gimple_expr): Use commutative_op.
2241 (parser::parse_expr): Allow :c to be used with non-binary
2242 operators if the commutative operand is known.
2243 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2244 CFN_FMS, CFN_FNMA and CFN_FNMS.
2245 (backprop::process_assign_use): Remove FMA_EXPR handling.
2246 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2247 (gen_hsa_fma): New function.
2248 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2249 IFN_FNMA and IFN_FNMS.
2250 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2251 * gimple-fold.h (follow_all_ssa_edges): Declare.
2252 * gimple-fold.c (follow_all_ssa_edges): New function.
2253 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2254 gimple_build interface and use follow_all_ssa_edges to fold the result.
2255 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2256 instead of checking for optabs directly.
2257 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2258 rather than FMA_EXPRs.
2259 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2260 call to IFN_FMA instead of an FMA_EXPR.
2262 2018-05-17 Jim Wilson <jimw@sifive.com>
2264 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2265 sign extended promoted subreg, and the range does not have the sign bit
2266 set, then do a sign extend.
2268 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2269 test, check for sign extended subreg and/or constant operands, and
2270 do a sign extend in that case.
2272 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2274 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2275 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2276 Add untyped.
2277 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2278 Change logics_shift_reg to logics_shift_imm.
2279 (thunderx2t99_fp_loadpair_basic): Delete.
2280 (thunderx2t99_fp_storepair_basic): Delete.
2281 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2282 (thunderx2t99_asimd_polynomial): Delete.
2283 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2284 and neon_fp_mul_d_scalar_q.
2285 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2286 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2287 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2288 (thunderx2t99_asimd_lut): Add missing tbl types.
2289 (thunderx2t99_asimd_ext): Delete.
2290 (thunderx2t99_asimd_load1_1_mult): Delete.
2291 (thunderx2t99_asimd_load1_2_mult): Delete.
2292 (thunderx2t99_asimd_load1_ldp): New.
2293 (thunderx2t99_asimd_load1): New.
2294 (thunderx2t99_asimd_load2): Add missing *load2* types.
2295 (thunderx2t99_asimd_load3): New.
2296 (thunderx2t99_asimd_load4): New.
2297 (thunderx2t99_asimd_store1_1_mult): Delete.
2298 (thunderx2t99_asimd_store1_2_mult): Delete.
2299 (thunderx2t99_asimd_store2_mult): Delete.
2300 (thunderx2t99_asimd_store2_onelane): Delete.
2301 (thunderx2t99_asimd_store_stp): New.
2302 (thunderx2t99_asimd_store1): New.
2303 (thunderx2t99_asimd_store2): New.
2304 (thunderx2t99_asimd_store3): New.
2305 (thunderx2t99_asimd_store4): New.
2307 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2309 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2310 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2312 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2313 Segher Boessenkool <segher@kernel.crashing.org>
2315 PR target/85698
2316 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2317 operand.
2319 2018-05-17 Richard Biener <rguenther@suse.de>
2321 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2322 for pruning loop and prune defs feeding only already visited PHIs.
2324 2018-05-17 Richard Biener <rguenther@suse.de>
2326 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2328 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2329 Richard Biener <rguenther@suse.de>
2331 PR tree-optimization/85793
2332 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2333 for VMAT_ELEMENTWISE.
2335 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2337 * internal-fn.h (lookup_internal_fn): Declare
2338 * internal-fn.c (lookup_internal_fn): New function.
2339 * gimple.c (gimple_build_call_from_tree): Handle calls to
2340 internal functions.
2341 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2342 internal function names.
2343 * tree-pretty-print.c (dump_generic_node): Likewise.
2344 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2346 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2348 * gimple-fold.h (gimple_build): Make the function forms take
2349 combined_fn rather than built_in_function.
2350 (gimple_simplify): Likewise.
2351 * gimple-match-head.c (gimple_simplify): Likewise.
2352 * gimple-fold.c (gimple_build): Likewise.
2353 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2354 rather than gimple_build_call_internal.
2355 (get_initial_defs_for_reduction): Likewise.
2356 (vect_create_epilog_for_reduction): Likewise.
2357 (vectorizable_live_operation): Likewise.
2359 2018-05-17 Martin Liska <mliska@suse.cz>
2361 * gimple-ssa-sprintf.c (format_directive): Do not use
2362 space in between 'G_' and '('.
2364 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2366 PR target/85323
2367 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2368 even if the mask is not all ones.
2370 PR target/85323
2371 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2372 vector.
2373 (ix86_gimple_fold_builtin): Likewise.
2375 PR target/85323
2376 * config/i386/i386.c: Include tree-vector-builder.h.
2377 (ix86_vector_shift_count): New function.
2378 (ix86_fold_builtin): Fold shift builtins by scalar count.
2379 (ix86_gimple_fold_builtin): Likewise.
2381 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2382 _mm512_setzero): New intrinsics.
2384 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2385 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2387 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2388 code generation for cases where splatting a value is not useful.
2389 * simplify-rtx.c (simplify_ternary_operation): Simplify
2390 vec_merge across a vec_duplicate and a paradoxical subreg forming
2391 a vector mode to a vec_concat.
2393 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2395 * config.gcc: Support "goldmont-plus".
2396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2397 "goldmont-plus".
2398 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2399 PROCESSOR_GOLDMONT_PLUS.
2400 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2401 (processor_target_table): Add "goldmont-plus".
2402 (PTA_GOLDMONT_PLUS): Define.
2403 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2404 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2405 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2406 (fold_builtin_cpu): Add "goldmont-plus".
2407 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2408 (ix86_option_override_internal): Add "goldmont-plus".
2409 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2410 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2411 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2412 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2414 2018-05-17 Richard Biener <rguenther@suse.de>
2416 PR tree-optimization/85757
2417 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2418 remove defs that only feed that PHI from further processing.
2420 2018-05-16 Jim Wilson <jimw@sifive.com>
2422 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2423 asterisk to name.
2424 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2426 2018-05-16 Mark Wielaard <mark@klomp.org>
2428 * dwarf2out.c (count_index_strings): New function.
2429 (output_indirect_strings): Call count_index_strings and generate
2430 header for dwarf_version >= 5.
2432 2018-05-16 Mark Wielaard <mark@klomp.org>
2434 * dwarf2out.c (dwarf_FORM): New function.
2435 (set_indirect_string): Use dwarf_FORM.
2436 (reset_indirect_string): Likewise.
2437 (size_of_die): Likewise.
2438 (value_format): Likewise.
2439 (output_die): Likewise.
2440 (add_skeleton_AT_string): Likewise.
2441 (output_macinfo_op): Likewise.
2442 (index_string): Likewise.
2443 (output_index_string_offset): Likewise.
2444 (output_index_string): Likewise.
2445 (count_index_strings): Likewise.
2447 2018-05-16 Carl Love <cel@us.ibm.com>
2449 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2450 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2452 2018-05-16 Martin Jambor <mjambor@suse.cz>
2454 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2455 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2457 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2459 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2460 (fnma<mode>4): Likewise.
2461 (fms<mode>4): Likewise.
2462 (fnms<mode>4): Likewise.
2463 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2464 (aarch64_fnma<mode>4): Likewise.
2465 (aarch64_fms<mode>4): Likewise.
2466 (aarch64_fnms<mode>4): Likewise.
2467 (aarch64_fnmadd<mode>4): Likewise.
2469 2018-05-16 Jason Merrill <jason@redhat.com>
2471 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2473 2018-05-16 Richard Biener <rguenther@suse.de>
2475 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2476 (dump_stmt_cost): Declare.
2477 (add_stmt_cost): Dump cost we add.
2478 (add_stmt_costs): New function.
2479 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2480 No longer exported.
2481 (vect_analyze_stmt): Adjust prototype.
2482 (vectorizable_condition): Likewise.
2483 (vectorizable_live_operation): Likewise.
2484 (vectorizable_reduction): Likewise.
2485 (vectorizable_induction): Likewise.
2486 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2487 cost vector to pass to vectorizable_ and record afterwards.
2488 (vect_model_reduction_cost): Take cost vector argument and adjust.
2489 (vect_model_induction_cost): Likewise.
2490 (vectorizable_reduction): Likewise.
2491 (vectorizable_induction): Likewise.
2492 (vectorizable_live_operation): Likewise.
2493 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2494 SLP_TREE_NUMBER_OF_VEC_STMTS.
2495 (vect_analyze_slp_cost_1): Remove.
2496 (vect_analyze_slp_cost): Likewise.
2497 (vect_slp_analyze_node_operations): Take visited args and
2498 a target cost vector. Avoid processing already visited stmt sets.
2499 (vect_slp_analyze_operations): Use a local cost vector to gather
2500 costs and register those of non-discarded instances.
2501 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2502 (vect_schedule_slp_instance): Remove copying of
2503 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2504 zero.
2505 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2506 adding cost. Record cost entry location.
2507 (vect_prologue_cost_for_slp_op): Function to compute cost of
2508 a constant or invariant generated for SLP vect in the prologue,
2509 split out from vect_analyze_slp_cost_1.
2510 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2511 (vect_model_promotion_demotion_cost): Likewise.
2512 (vect_model_store_cost): Likewise, make static.
2513 (vect_model_load_cost): Likewise.
2514 (vectorizable_bswap): Add cost vector arg and adjust.
2515 (vectorizable_call): Likewise.
2516 (vectorizable_simd_clone_call): Likewise.
2517 (vectorizable_conversion): Likewise.
2518 (vectorizable_assignment): Likewise.
2519 (vectorizable_shift): Likewise.
2520 (vectorizable_operation): Likewise.
2521 (vectorizable_store): Likewise.
2522 (vectorizable_load): Likewise.
2523 (vectorizable_condition): Likewise.
2524 (vectorizable_comparison): Likewise.
2525 (can_vectorize_live_stmts): Likewise.
2526 (vect_analyze_stmt): Likewise.
2527 (vect_transform_stmt): Adjust calls to vectorizable_*.
2528 * tree-vectorizer.c: Include gimple-pretty-print.h.
2529 (dump_stmt_cost): New function.
2531 2018-05-16 Richard Biener <rguenther@suse.de>
2533 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2534 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2535 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2536 (check_name): New callback.
2537 (dse_classify_store): Track cycles via a visited bitmap of PHI
2538 defs and simplify handling of in-loop and across loop dead stores
2539 and properly fail for loop-variant refs. Handle byte-tracking with
2540 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2541 limiting the walk.
2543 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2545 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2546 (vect_get_mask_type_for_stmt): Likewise.
2547 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2548 split out from...
2549 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2550 to determine the statement's vector type and the vector type that
2551 should be used for calculating nunits. Deal with cases in which
2552 the type has to be deferred.
2553 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2554 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2555 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2556 (vect_determine_vf_for_stmt): New functions, split out from...
2557 (vect_determine_vectorization_factor): ...here.
2558 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2559 (vect_get_mask_type_for_stmt): New functions, split out from
2560 vect_determine_vectorization_factor.
2562 2018-05-16 Richard Biener <rguenther@suse.de>
2564 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2565 verify the [VEC_]COND_EXPR embedded comparison.
2567 2018-05-15 Martin Sebor <msebor@redhat.com>
2569 PR tree-optimization/85753
2570 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2571 RECORD_TYPE in addition to ARRAY_TYPE.
2573 2018-05-15 Martin Sebor <msebor@redhat.com>
2575 PR middle-end/85643
2576 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2578 2018-05-15 Richard Biener <rguenther@suse.de>
2580 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2581 add by_clobber_p one. Change algorithm to collect all defs
2582 representing uses we need to walk and try reducing them to
2583 a single one before failing.
2584 (dse_dom_walker::dse_optimize_stmt): Adjust.
2586 2018-05-13 Mark Wielaard <mark@klomp.org>
2588 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2589 (size_of_loc_descr): Likewise.
2590 (output_loc_operands): Likewise.
2591 (output_loc_operands_raw): Likewise.
2592 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2593 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2594 (hash_loc_operands): Likewise.
2595 (compare_loc_operands): Likewise.
2597 2018-05-14 Mark Wielaard <mark@klomp.org>
2599 * dwarf2out.c (count_index_addrs): New function.
2600 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2602 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2604 PR tree-optimization/83648
2605 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2606 return value as malloc candidate.
2608 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2610 PR ipa/85734
2611 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2612 param as true in call to suggest_attribute.
2614 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2616 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2617 -mreadonly-in-sdata.
2619 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2621 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2622 New pattern.
2623 (aarch64_crypto_aesd_fused): Likewise.
2625 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2627 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2628 (movsi_aarch64): Likewise.
2629 (load_pairsi): Likewise.
2630 (load_pairdi): Likewise.
2631 (store_pairsi): Likewise.
2632 (store_pairdi): Likewise.
2633 (load_pairsf): Likewise.
2634 (load_pairdf): Likewise.
2635 (store_pairsf): Likewise.
2636 (store_pairdf): Likewise.
2637 (zero_extend): Likewise.
2638 (trunc): Swap alternatives.
2639 (fcvt_target): Add '?' to prefer w over r.
2641 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2643 PR target/85756
2644 * config/i386/i386.md: Disallow non-commutative arithmetics in
2645 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2646 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2647 in the peephole2 before it.
2649 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2651 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2652 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2653 (ix86_handle_option): Handle -mcldemote.
2654 * config.gcc: New header.
2655 * config/i386/cldemoteintrin.h: New file.
2656 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2657 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2658 -mcldemote.
2659 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2660 OPTION_MASK_ISA_CLDEMOTE.
2661 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2662 (ix86_valid_target_attribute_inner_p): Ditto.
2663 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2664 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2665 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2666 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2667 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2668 (cldemote): New.
2669 * config/i386/i386.opt: Add -mcldemote.
2670 * config/i386/x86intrin.h: New header.
2671 * doc/invoke.texi: Add -mcldemote.
2673 2018-05-14 Richard Biener <rguenther@suse.de>
2675 * doc/match-and-simplify.texi: Adjust :s documentation.
2677 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2679 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2680 intended memcpy size.
2681 (REORDER_45): Likewise.
2683 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2685 * sort.cc: New file.
2686 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2687 * vec.c (qsort_chk): Use gcc_qsort.
2688 * Makefile.in (OBJS-libcommon): Add sort.o.
2689 (build/sort.o): New target. Use it...
2690 (BUILD_RTL): ... here, and...
2691 (build/gencfn-macros): ... here, and...
2692 (build/genmatch): ... here.
2694 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2695 Chung-Ju Wu <jasonwucj@gmail.com>
2697 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2698 * config/nds32/nds32-graywolf.md: New file.
2699 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2700 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2701 pipeline.
2702 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2703 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2704 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2705 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2706 * config/nds32/pipelines.md: Include n15 settings.
2708 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2709 Chung-Ju Wu <jasonwucj@gmail.com>
2711 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2712 * config/nds32/nds32-n13.md: New file.
2713 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2714 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2715 pipeline.
2716 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2717 * config/nds32/nds32.md (pipeline_model): Add n13.
2718 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2719 * config/nds32/pipelines.md: Include n13 settings.
2721 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2722 Chung-Ju Wu <jasonwucj@gmail.com>
2724 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2725 * config/nds32/nds32-n10.md: New file.
2726 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2727 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2728 pipeline.
2729 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2730 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2731 * config/nds32/nds32.md (pipeline_model): Add n10.
2732 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2733 * config/nds32/pipelines.md: Include n10 settings.
2735 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2736 Kito Cheng <kito.cheng@gmail.com>
2737 Chung-Ju Wu <jasonwucj@gmail.com>
2739 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2740 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2741 Add enum values for DSP extension instructions.
2742 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2743 New constraints.
2744 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2745 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2746 New code iterators.
2747 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2748 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2749 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2750 * config/nds32/nds32-intrinsic.md: Likewise.
2751 * config/nds32/nds32_intrinsic.h: Likewise.
2752 * config/nds32/nds32-md-auxiliary.c: Likewise.
2753 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2754 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2755 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2756 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2757 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2758 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2759 TYPE_DMAC in switch statement.
2760 * config/nds32/nds32.c: New checking and implementation for DSP
2761 extension instructions.
2762 * config/nds32/nds32.h: Likewise.
2763 * config/nds32/nds32.md: Likewise.
2764 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2765 * config/nds32/predicates.md: Implement new predicates for DSP
2766 extension.
2768 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2770 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2771 Reformat alternatives and attributes so it is easier to identify
2772 which constraints/attributes go with which instruction.
2773 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2774 (mov<mode>_softfloat32, FMOVE64): Likewise.
2775 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2776 (mov<mode>_softfloat64, FMOVE64): Likewise.
2778 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2780 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2781 subsection.
2782 (Basic PowerPC Built-in Functions): The new name of the
2783 subsection previously known as "PowerPC Built-in Functions".
2784 (Basic PowerPC Built-in Functions Available on all Configurations):
2785 New subsubsection.
2786 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2787 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2788 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2789 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2791 2018-05-11 Martin Jambor <mjambor@suse.cz>
2793 PR ipa/85655
2794 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2795 single const.
2797 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2799 PR target/85733
2800 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2802 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2804 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2805 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2806 (ix86_handle_option): Handle -mwaitpkg.
2807 * config.gcc: New header.
2808 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2809 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2810 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2811 function type.
2812 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2813 OPTION_MASK_ISA_WAITPKG.
2814 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2815 (ix86_option_override_internal): Add PTA_WAITPKG.
2816 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2817 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2818 IX86_BUILTIN_TPAUSE.
2819 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2820 __builtin_ia32_umwait and __builtin_ia32_tpause.
2821 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2822 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2823 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2824 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2825 UNSPECV_TPAUSE): New.
2826 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2827 * config/i386/i386.opt: Add -mwaitpkg.
2828 * config/i386/waitpkgintrin.h: New file.
2829 * config/i386/x86intrin.h: New header.
2830 * doc/invoke.texi: Add -mwaitpkg.
2832 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2834 PR target/85606
2835 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2836 equivalent.
2837 (cortex-m0): Use armv6s-m isa.
2838 (cortex-m0plus): Likewise.
2839 (cortex-m1): Likewise.
2840 (cortex-m0.small-multiply): Likewise.
2841 (cortex-m0plus.small-multiply): Likewise.
2842 (cortex-m1.small-multiply): Likewise.
2844 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2845 Jakub Jelinek <jakub@redhat.com>
2847 PR tree-optimization/85692
2848 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2849 source permute as well.
2851 2018-05-11 Martin Liska <mliska@suse.cz>
2853 PR sanitizer/85556
2854 * doc/extend.texi: Document LLVM style format for no_sanitize
2855 attribute.
2857 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2859 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2860 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2861 (mode_supports_vsx_dform_quad): Likewise.
2862 (mode_supports_vmx_dform): Move these functions to be next to the
2863 other mode_supports functions.
2864 (mode_supports_dq_form): Likewise.
2865 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2866 mode_supports_dq_form.
2867 (reg_offset_addressing_ok_p): Likewise.
2868 (offsettable_ok_by_alignment): Likewise.
2869 (rs6000_legitimate_offset_address_p): Likewise.
2870 (legitimate_lo_sum_address_p): Likewise.
2871 (rs6000_legitimize_address): Likewise.
2872 (rs6000_legitimize_reload_address): Likewise.
2873 (rs6000_secondary_reload_inner): Likewise.
2874 (rs6000_preferred_reload_class): Likewise.
2875 (rs6000_output_move_128bit): Likewise.
2877 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2879 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2880 Generate SImode target register for null target.
2881 <case IX86_BUILTIN_XGETBV>: Ditto.
2882 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2883 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2885 2018-05-10 Carl Love <cel@us.ibm.com>
2887 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2888 dcbtt and dcbtstt if operands[2] is 0.
2890 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2892 PR target/85693
2893 * config/i386/sse.md (usadv64qi): New expander.
2895 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2897 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2898 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2899 -maltivec=be support.
2900 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2901 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2902 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2903 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2904 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2905 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2906 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2907 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2908 altivec_vsumsws): Adjust.
2909 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2910 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2911 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2912 support.
2913 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2914 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2915 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2916 (altivec_lve<VI_char>x): Delete expand.
2917 (*altivec_lve<VI_char>x_internal): Rename to...
2918 (altivec_lve<VI_char>x): ... this.
2919 (altivec_lvxl_<mode>): Delete expand.
2920 (*altivec_lvxl_<mode>_internal): Rename to ...
2921 (altivec_lvxl_<mode>): ... this.
2922 (altivec_stvxl_<mode>): Delete expand.
2923 (*altivec_stvxl_<mode>_internal): Rename to ...
2924 (altivec_stvxl_<mode>): ... this.
2925 (altivec_stve<VI_char>x): Delete expand.
2926 (*altivec_stve<VI_char>x_internal): Rename to ...
2927 (altivec_stve<VI_char>x): ... this.
2928 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2929 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2930 reduc_plus_scal_<mode>): Adjust.
2931 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2932 comment.
2933 (rs6000_cpu_cpp_builtins): Adjust.
2934 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2935 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2936 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2937 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2938 -maltivec=be support.
2939 (rs6000_split_vec_extract_var): Adjust.
2940 (rs6000_split_v4si_init): Adjust.
2941 (swap_selector_for_mode): Delete.
2942 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2943 altivec_expand_stvex_be): Delete.
2944 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2945 -maltivec=be support.
2946 (rs6000_gimple_fold_builtin): Ditto.
2947 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2948 Adjust.
2949 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2950 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2951 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2952 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2953 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2954 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2955 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2956 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2957 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2958 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2959 anonymous split): Adjust.
2960 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2961 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2963 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2965 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2966 when --with-gxx-include-dir is also specified.
2967 * configure: Regenerate.
2969 2018-05-09 Jim Wilson <jimw@sifive.com>
2971 PR target/84797
2972 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2973 * config/riscv/t-withmultilib: New.
2974 * config/riscv/withmultilib.h: New.
2975 * doc/install.texi: Document RISC-V --with-multilib-list support.
2977 2018-05-09 Richard Biener <rguenther@suse.de>
2979 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2980 vector.
2981 (vect_bb_vectorization_profitable_p): Adjust. Compute
2982 actual scalar cost using the cost vector and the add_stmt_cost
2983 machinery.
2985 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2987 PR rtl-optimization/85645
2988 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2989 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2991 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2993 PR rtl-optimization/85645
2994 * shrink-wrap.c (spread_components): Return a boolean saying if
2995 anything was changed.
2996 (try_shrink_wrapping_separate): Iterate spread_components until
2997 nothing changes anymore.
2999 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3001 PR rtl-optimization/85645
3002 * regrename.c (build_def_use): Also kill the chains that include the
3003 destination of a REG_CFA_REGISTER note.
3005 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3007 PR rtl-optimization/85645
3008 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3009 insn that has a REG_CFA_REGISTER note.
3011 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3013 * cfgexpand.c (expand_clobber): New function.
3014 (expand_gimple_stmt_1): Use it.
3015 * tree-vect-stmts.c (vect_clobber_variable): New function,
3016 split out from...
3017 (vectorizable_simd_clone_call): ...here.
3018 (vectorizable_store): Emit a clobber either side of an
3019 IFN_STORE_LANES sequence.
3020 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3022 2018-05-09 Tom de Vries <tom@codesourcery.com>
3024 PR target/85626
3025 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3026 (define_insn "trap_if_false"): Add exit after trap.
3028 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3030 PR rtl-optimization/85638
3031 * bb-reorder.c: Include common/common-target.h.
3032 (create_forwarder_block): New function extracted from...
3033 (fix_up_crossing_landing_pad): ...here. Rename into...
3034 (dw2_fix_up_crossing_landing_pad): ...this.
3035 (sjlj_fix_up_crossing_landing_pad): New function.
3036 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3037 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3038 from both partitions and exit the loop after one iteration.
3040 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3042 Revert:
3043 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3044 subsection.
3045 (Basic PowerPC Built-in Functions): The new name of the
3046 subsection previously known as "PowerPC Built-in Functions".
3047 (Basic PowerPC Built-in Functions Available on all Configurations):
3048 New subsubsection.
3049 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3050 subsubsection.
3051 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3052 subsubsection.
3053 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3054 subsubsection.
3055 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3056 subsubsection.
3058 2018-05-08 Jim Wilson <jimw@sifive.com>
3060 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3061 (LD_EMUL_SUFFIX): New.
3062 (LINK_SPEC): Use it.
3064 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3066 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3067 subsection.
3068 (Basic PowerPC Built-in Functions): The new name of the
3069 subsection previously known as "PowerPC Built-in Functions".
3070 (Basic PowerPC Built-in Functions Available on all Configurations):
3071 New subsubsection.
3072 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3073 subsubsection.
3074 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3075 subsubsection.
3076 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3077 subsubsection.
3078 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3079 subsubsection.
3081 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3083 PR target/85683
3084 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3085 after cmpelim optimization.
3087 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3089 * config.gcc: Support "goldmont".
3090 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3092 PROCESSOR_GOLDMONT.
3093 * config/i386/i386.c (m_GOLDMONT): Define.
3094 (processor_target_table): Add "goldmont".
3095 (PTA_GOLDMONT): Define.
3096 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3097 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3098 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3099 (fold_builtin_cpu): Add "goldmont".
3100 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3101 (ix86_option_override_internal): Add "goldmont".
3102 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3103 (processor_type): Add PROCESSOR_GOLDMONT.
3104 * config/i386/i386.md: Add CPU "glm".
3105 * config/i386/glm.md: New file.
3106 * config/i386/x86-tune.def: Add m_GOLDMONT.
3107 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3109 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3111 PR target/85572
3112 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3113 E_V4DImode.
3114 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3115 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3116 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3118 PR target/85317
3119 * config/i386/i386.c (ix86_fold_builtin): Handle
3120 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3122 PR target/85480
3123 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3124 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3126 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3128 PR target/85658
3129 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3130 (check_arch): Likewise.
3131 (check_fpu): Return the result rather than printing it.
3132 (end arch): Fix operator precedence.
3133 (end cpu): Likewise.
3134 (END): Print the result from check_fpu.
3136 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3137 Alan Hayward <alan.hayward@arm.com>
3138 David Sherwood <david.sherwood@arm.com>
3140 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3141 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3142 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3143 (*fcmuo<mode>_and): New patterns.
3145 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3147 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3148 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3149 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3150 (cmp_op, sve_imm_con): New code attributes.
3151 (SVE_COND_INT_CMP, imm_con): Delete.
3152 (cmp_op): Remove above unspecs from int attribute.
3153 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3154 to...
3155 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3156 comparison-specific unspecs.
3157 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3158 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3159 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3160 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3161 (*vec_fcm<cmp_op><mode>): Rename to...
3162 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3163 (*vec_fcmuo<mode>): Rename to...
3164 (*fcmuo<mode>): ...this and adjust likewise.
3165 (*pred_fcm<cmp_op><mode>): New pattern.
3166 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3167 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3168 functions.
3169 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3170 and UNORDERED.
3171 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3172 (aarch64_emit_sve_predicated_cond): New function.
3173 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3174 (aarch64_emit_unspec_cond_or): Replace with...
3175 (aarch64_emit_sve_or_conds): ...this new function. Use
3176 aarch64_emit_sve_ptrue_op for the individual comparisons and
3177 aarch64_emit_binop to OR them together.
3178 (aarch64_emit_inverted_unspec_cond): Replace with...
3179 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3180 aarch64_emit_sve_ptrue_op for the comparison and
3181 aarch64_emit_unop to invert the result.
3182 (aarch64_expand_sve_vec_cmp_float): Update after the above
3183 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3185 2018-05-07 Nathan Sidwell <nathan@acm.org>
3187 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3188 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3189 (Backwards Compatibility): Likewise.
3191 2018-05-07 Luis Machado <luis.machado@linaro.org>
3193 PR bootstrap/85681
3194 Revert:
3195 2018-05-07 Luis Machado <luis.machado@linaro.org>
3197 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3198 <prefetch_dynamic_strides>: New const bool field.
3199 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3200 prefetch_dynamic_strides.
3201 (exynosm1_prefetch_tune): Likewise.
3202 (thunderxt88_prefetch_tune): Likewise.
3203 (thunderx_prefetch_tune): Likewise.
3204 (thunderx2t99_prefetch_tune): Likewise.
3205 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3206 to false.
3207 (aarch64_override_options_internal): Update to set
3208 PARAM_PREFETCH_DYNAMIC_STRIDES.
3209 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3210 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3211 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3212 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3213 prefetch-dynamic-strides setting.
3215 2018-05-07 Luis Machado <luis.machado@linaro.org>
3217 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3218 <minimum_stride>: New const int field.
3219 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3220 minimum_stride field.
3221 (exynosm1_prefetch_tune): Likewise.
3222 (thunderxt88_prefetch_tune): Likewise.
3223 (thunderx_prefetch_tune): Likewise.
3224 (thunderx2t99_prefetch_tune): Likewise.
3225 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3226 (aarch64_override_options_internal): Update to set
3227 PARAM_PREFETCH_MINIMUM_STRIDE.
3228 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3229 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3230 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3231 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3232 stride is constant and is below the minimum stride threshold.
3234 2018-05-07 Luis Machado <luis.machado@linaro.org>
3236 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3237 to 512.
3239 2018-05-07 Luis Machado <luis.machado@linaro.org>
3241 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3242 <prefetch_dynamic_strides>: New const bool field.
3243 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3244 prefetch_dynamic_strides.
3245 (exynosm1_prefetch_tune): Likewise.
3246 (thunderxt88_prefetch_tune): Likewise.
3247 (thunderx_prefetch_tune): Likewise.
3248 (thunderx2t99_prefetch_tune): Likewise.
3249 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3250 to false.
3251 (aarch64_override_options_internal): Update to set
3252 PARAM_PREFETCH_DYNAMIC_STRIDES.
3253 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3254 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3255 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3256 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3257 prefetch-dynamic-strides setting.
3259 2018-05-07 Luis Machado <luis.machado@linaro.org>
3261 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3262 <minimum_stride>: New const int field.
3263 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3264 minimum_stride field.
3265 (exynosm1_prefetch_tune): Likewise.
3266 (thunderxt88_prefetch_tune): Likewise.
3267 (thunderx_prefetch_tune): Likewise.
3268 (thunderx2t99_prefetch_tune): Likewise.
3269 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3270 (aarch64_override_options_internal): Update to set
3271 PARAM_PREFETCH_MINIMUM_STRIDE.
3272 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3273 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3274 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3275 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3276 stride is constant and is below the minimum stride threshold.
3278 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3280 PR c++/85659
3281 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3282 the type is addressable. Don't force op into register if it has
3283 BLKmode.
3285 2018-05-05 Roland McGrath <mcgrathr@google.com>
3287 PR other/77609
3288 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3289 any section for which we don't know a specific type it should have,
3290 regardless of name. Previously this was done only for the exact
3291 names ".init_array", ".fini_array", and ".preinit_array".
3292 (default_elf_asm_named_section): Add comment about
3293 relationship with default_section_type_flags and SECTION_NOTYPE.
3294 (get_section): Don't consider it a type conflict if one side has
3295 SECTION_NOTYPE and the other doesn't, as long as neither has the
3296 SECTION_BSS et al used in the default_section_type_flags logic.
3298 2018-05-05 Tom de Vries <tom@codesourcery.com>
3300 PR target/85653
3301 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3302 (workaround_barsyncs): New function.
3303 (nvptx_reorg): Use workaround_barsyncs.
3304 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3305 (define_expand "nvptx_membar_cta"): New define_expand.
3306 (define_insn "*nvptx_membar_cta"): New insn.
3308 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3310 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3311 To improve optimization opportunities.
3312 * builtin-types.def: The new needed builtin types for the above.
3314 2018-05-04 Richard Biener <rguenther@suse.de>
3316 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3317 * gimple-ssa-store-merging.c
3318 (imm_store_chain_info::output_merged_store): Remove redundant create,
3319 release split_store vector contents on failure.
3320 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3321 scalar stmt vector on cache hit.
3323 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3325 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3326 Xilinx FP support.
3327 * config.gcc (powerpc-xilinx-eabi*): Remove.
3328 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3329 support.
3330 (fusion_addis_mem_combo_load): Ditto.
3331 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3332 FP support.
3333 (rs6000_cpu_cpp_builtins): Ditto.
3334 * config/rs6000/rs6000-linux.c
3335 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3336 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3337 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3338 support.
3339 (rs6000_setup_reg_addr_masks): Ditto.
3340 (rs6000_init_hard_regno_mode_ok): Ditto.
3341 (rs6000_option_override_internal): Ditto.
3342 (legitimate_lo_sum_address_p): Ditto.
3343 (rs6000_legitimize_address): Ditto.
3344 (rs6000_legitimize_reload_address): Ditto.
3345 (rs6000_legitimate_address_p): Ditto.
3346 (abi_v4_pass_in_fpr): Ditto.
3347 (setup_incoming_varargs): Ditto.
3348 (rs6000_gimplify_va_arg): Ditto.
3349 (rs6000_split_multireg_move): Ditto.
3350 (rs6000_savres_strategy): Ditto.
3351 (rs6000_emit_prologue_components): Ditto.
3352 (rs6000_emit_epilogue_components): Ditto.
3353 (rs6000_emit_prologue): Ditto.
3354 (rs6000_emit_epilogue): Ditto.
3355 (rs6000_elf_file_end): Ditto.
3356 (rs6000_function_value): Ditto.
3357 (rs6000_libcall_value): Ditto.
3358 * config/rs6000/rs6000.h: Ditto.
3359 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3360 (TARGET_MINMAX): ... this. New.
3361 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3362 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3363 (*movsi_internal1_single): Delete.
3364 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3365 mfpu=, mxilinx-fpu): Delete.
3366 * config/rs6000/singlefp.h: Delete.
3367 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3368 * config/rs6000/t-rs6000: Ditto.
3369 * config/rs6000/t-xilinx: Delete.
3370 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3371 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3372 (VStype_simple): Delete.
3373 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3374 * config/rs6000/xfpu.h: Delete.
3375 * config/rs6000/xfpu.md: Delete.
3376 * config/rs6000/xilinx.h: Delete.
3377 * config/rs6000/xilinx.opt: Delete.
3378 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3379 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3381 2018-05-04 Tom de Vries <tom@codesourcery.com>
3383 PR libgomp/85639
3384 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3385 if ignore == 0.
3387 2018-05-04 Richard Biener <rguenther@suse.de>
3389 PR middle-end/85627
3390 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3391 (expand_complex_div_wide): Likewise.
3392 (expand_complex_operations_1): Likewise.
3393 (expand_complex_libcall): Preserve EH info of the original stmt.
3394 (tree_lower_complex): Handle removed blocks.
3395 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3396 on complex multiplication and division libcall builtins.
3398 2018-05-04 Richard Biener <rguenther@suse.de>
3400 PR middle-end/85574
3401 * fold-const.c (negate_expr_p): Restrict negation of operand
3402 zero of a division to when we know that can happen without
3403 overflow.
3404 (fold_negate_expr_1): Likewise.
3406 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3408 PR libstdc++/85466
3409 * real.h (real_nextafter): Declare.
3410 * real.c (real_nextafter): New function.
3411 * fold-const-call.c (fold_const_nextafter): New function.
3412 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3413 CASE_CFN_NEXTTOWARD.
3414 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3415 even when arg1_mode is different from arg0_mode.
3417 2018-05-03 Nathan Sidwell <nathan@acm.org>
3419 * doc/extend.texi (Deprecated Features): Remove
3420 -ffriend-injection.
3421 (Backwards Compatibility): Likewise.
3422 * doc/invoke.texi (C++ Language Options): Likewise.
3423 (C++ Dialect Options): Likewise.
3425 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3427 PR target/85530
3428 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3429 _mm512_mask_mullox_epi64): New intrinsics.
3431 2018-05-03 Tom de Vries <tom@codesourcery.com>
3433 PR testsuite/85106
3434 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3435 dump files): Add offload-tree.
3437 2018-05-03 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/85615
3440 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3441 to loops not nested in BBs loop father to avoid creating multi-entry
3442 loops.
3444 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3446 PR tree-optimization/70291
3447 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3448 arguments. Change return type to tree. Emit libcall as a new
3449 statement rather than replacing existing one when inplace_p is true.
3450 (expand_complex_multiplication_components): New function.
3451 (expand_complex_multiplication): Expand floating-point complex
3452 multiplication using the above.
3453 (expand_complex_division): Rename inner_type parameter to type.
3454 Update expand_complex_libcall call-site.
3455 (expand_complex_operations_1): Update expand_complex_multiplication
3456 and expand_complex_division call-sites.
3458 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3460 PR target/85582
3461 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3462 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3463 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3464 the highest significant bit of the shift count mask is clear. In
3465 check whether and[sq]i3 is needed verify that all significant bits
3466 of the shift count other than the highest are set.
3468 2018-05-02 Tom de Vries <tom@codesourcery.com>
3470 PR libgomp/82428
3471 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3472 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3473 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3474 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3475 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3476 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3477 __builtin_goacc_parlevel_size.
3479 2018-05-02 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/85597
3482 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3483 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3484 directly.
3486 2018-05-02 Tom de Vries <tom@codesourcery.com>
3488 PR testsuite/85106
3489 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3490 dump files): Add ltrans-tree.
3492 2018-05-02 Tom de Vries <tom@codesourcery.com>
3494 PR testsuite/85106
3495 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3496 dump files): Add wpa-ipa.
3498 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3500 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3501 powerpc*-*-linux*paired* target.
3502 * config/rs6000/750cl.h: Delete.
3503 * config/rs6000/paired.h: Delete.
3504 * config/rs6000/paired.md: Delete.
3505 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3506 float support.
3507 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3508 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3509 comment. Remove paired float support.
3510 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3511 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3512 VECTOR_PAIRED.
3513 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3514 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3515 declarations.
3516 * config/rs6000/rs6000.c: Remove paired float support.
3517 (paired_expand_vector_init, paired_expand_vector_move,
3518 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3519 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3520 paired_expand_builtin, paired_expand_predicate_builtin,
3521 paired_init_builtins): Delete.
3522 * config/rs6000/rs6000.h: Remove paired float support.
3523 * config/rs6000/rs6000.md: Remove paired float support.
3524 (move_from_CR_ov_bit): Delete.
3525 * config/rs6000/rs6000.opt (mpaired): Delete.
3526 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3527 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3529 2018-05-02 Richard Biener <rguenther@suse.de>
3531 PR middle-end/85567
3532 * gimplify.c (gimplify_save_expr): When in SSA form allow
3533 SAVE_EXPRs to compute to SSA vars.
3535 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3537 PR target/85582
3538 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3539 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3540 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3541 clobber operands[2], instead use a new pseudo. Formatting fixes.
3543 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3545 PR tree-optimization/85586
3546 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3547 exit early for statements in the same group if the accesses are
3548 not strided.
3550 2018-05-02 Tom de Vries <tom@codesourcery.com>
3552 PR lto/85451
3553 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3554 error message.
3556 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3558 PR tree-optimization/85143
3559 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3561 2018-05-01 Tom de Vries <tom@codesourcery.com>
3563 PR lto/85451
3564 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3565 not found" error message.
3567 2018-05-01 Tom de Vries <tom@codesourcery.com>
3569 PR other/83786
3570 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3571 * vec.c (test_ordered_remove_if): New function.
3572 (vec_c_tests): Call test_ordered_remove_if.
3573 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3574 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3575 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3576 VEC_ORDERED_REMOVE_IF.
3578 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3580 PR tree-optimization/82665
3581 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3582 pointer subtraction where arguments come from a memchr call.
3584 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3586 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3587 --push-state --as-needed and --pop-state instead of --as-needed and
3588 --no-as-needed if ld supports it.
3589 * configure: Regenerated.
3591 PR web/85578
3592 * doc/install.texi2html: Replace _002d with - and _002a with * in
3593 generated html files using sed.
3595 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3597 PR c++/85523
3598 * gcc-rich-location.c (blank_line_before_p): New function.
3599 (use_new_line): New function.
3600 (gcc_rich_location::add_fixit_insert_formatted): New function.
3601 * gcc-rich-location.h
3602 (gcc_rich_location::add_fixit_insert_formatted): New function.
3604 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3606 * selftest.c (assert_streq): Rename "expected" and "actual" to
3607 "val1" and "val2". Extend NULL-handling to cover both inputs
3608 symmetrically, while still requiring both to be non-NULL for a pass.
3609 * selftest.h (assert_streq): Rename "expected" and "actual" to
3610 "val1" and "val2".
3611 (ASSERT_EQ): Likewise.
3612 (ASSERT_EQ_AT): Likewise.
3613 (ASSERT_KNOWN_EQ): Likewise.
3614 (ASSERT_KNOWN_EQ_AT): Likewise.
3615 (ASSERT_NE): Likewise.
3616 (ASSERT_MAYBE_NE): Likewise.
3617 (ASSERT_MAYBE_NE_AT): Likewise.
3618 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3619 the assertion to pass.
3620 (ASSERT_STREQ_AT): Likewise.
3622 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3624 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3625 interaction with -pie.
3627 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3629 * selftest.h: Fix alphabetization of per-source-file selftest
3630 declarations.
3632 2018-04-30 Jason Merrill <jason@redhat.com>
3634 PR c++/61982 - dead stores to destroyed objects.
3635 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3636 of clobber.
3638 2018-04-30 Jason Merrill <jason@redhat.com>
3640 * tree.c (build_clobber): New.
3641 * tree.h: Declare it.
3642 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3644 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3646 * diagnostic-show-locus.c (layout::layout): Update for
3647 location_get_source_line returning a char_span.
3648 (struct char_span): Move to input.h.
3649 (struct correction): Update for fields in char_span becoming
3650 private.
3651 (struct source_line): Update for location_get_source_line
3652 returning a char_span.
3653 (layout::print_line): Likewise.
3654 * edit-context.c (edited_file::print_content): Likewise.
3655 (edited_file::print_diff_hunk): Likewise.
3656 (edited_file::print_run_of_changed_lines): Likewise.
3657 (edited_file::get_num_lines): Likewise.
3658 (edited_line::edited_line): Likewise.
3659 * final.c (asm_show_source): Likewise.
3660 * input.c (location_get_source_line): Convert return type
3661 from const char * to char_span, losing the final "line_len"
3662 param.
3663 (dump_location_info): Update for the above.
3664 (get_substring_ranges_for_loc): Likewise. Use a char_span
3665 when handling the literal within the line.
3666 (test_reading_source_line): Update for location_get_source_line
3667 returning a char_span.
3668 * input.h (class char_span): Move here from
3669 diagnostic-show-locus.c, converting from a struct to a class.
3670 Make data members private.
3671 (char_span::operator bool): New.
3672 (char_span::length): New.
3673 (char_span::get_buffer): New.
3674 (char_span::operator[]): New.
3675 (char_span::subspan): Make const.
3676 (char_span::xstrdup): New.
3677 (location_get_source_line): Convert return type from const char *
3678 to char_span, losing the final "line_size" param.
3680 2018-04-30 Jan Hubicka <jh@suse.cz>
3682 * lto-wrapper.c (ltrans_priorities): New static var.
3683 (cmp_priority): New.
3684 (run_gcc): Read priorities and if doing parallel build order
3685 the Makefile by them.
3687 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3689 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3691 2018-04-30 Richard Biener <rguenther@suse.de>
3693 * tree-cfg.c (verify_address): Remove base argument, add
3694 flag whether to check TREE_ADDRESSABLE and do that.
3695 (verify_expr): Remove.
3696 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3697 (verify_gimple_assign_single): Likewise.
3698 (verify_gimple_switch): Likewise.
3699 (verify_expr_location_1): Dereference tp once. Add (disabled)
3700 piece from verify_expr.
3701 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3703 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3705 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3707 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3709 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3710 (small_data_pattern): Likewise.
3711 (arc_rewrite_small_data): Likewise.
3712 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3713 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3714 (get_symbol_alignment): New function.
3715 (legitimate_small_data_address_p): Likewise.
3716 (legitimate_scaled_address): Update, call
3717 legitimate_small_data_address_p.
3718 (output_sdata): New static variable.
3719 (arc_print_operand): Update how we handle small data operands.
3720 (arc_print_operand_address): Likewise.
3721 (arc_legitimate_address_p): Update, use
3722 legitimate_small_data_address_p.
3723 (arc_rewrite_small_data_p): Remove.
3724 (arc_rewrite_small_data_1): Likewise.
3725 (arc_rewrite_small_data): Likewise.
3726 (small_data_pattern): Likewise.
3727 (compact_sda_memory_operand): Update to use
3728 legitimate_small_data_address_p and get_symbol_alignment.
3729 (prepare_move_operands): Don't rewite sdata pattern.
3730 (prepare_extend_operands): Remove.
3731 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3732 pattern.
3733 (zero_extendqisi2): Likewise.
3734 (zero_extendhisi2): Likewise.
3735 (extendqihi2): Likewise.
3736 (extendqisi2): Likewise.
3737 (extendhisi2): Likewise.
3738 (addsi3): Likewise.
3739 (subsi3): Likewise.
3740 (andsi3): Likewise.
3741 * config/arc/constraints.md (Usd): Change it to memory constraint.
3743 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3745 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3746 as source of std instructions.
3747 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3748 allow 6-bit constants as source for store instructions.
3749 (movdi_insn): Update instruction pattern to allow 6-bit constants
3750 as source for store instructions.
3752 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3754 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3756 2018-04-30 Nathan Sidwell <nathan@acm.org>
3757 Sandra Loosemore <sandra@codesourcery.com>
3759 * dumpfile.c (dump_open): Allow '-' for stdout.
3760 * doc/invoke.texi (Developer Options): Document dump filename
3761 determination early. Document stdin/stdout selection.
3763 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3765 Microblaze Target: PIC data text relative
3767 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3768 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3769 Add declaration.
3770 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3771 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3772 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3773 New addressing mode for data-text relative position indepenedent code.
3774 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3775 'ADDRESS_SYMBOLIC_TXT_REL'.
3776 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3777 (microblaze_legitimate_pic_operand): Exclude function calls from
3778 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3779 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3780 addresses cases.
3781 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3782 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3783 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3784 for 'address + offset'.
3785 (microblaze_expand_prologue): Add new function prologue call for
3786 'r20' assignation.
3787 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3788 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3789 table in case of TARGET_PIC_DATA_TEXT_REL.
3790 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3791 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3792 Add new macros 'UNSPEC_TEXT',
3793 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3794 + exclude function calls from 'UNSPEC_PLT' in case of data text
3795 relative mode.
3796 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3797 new target hook for generating address diff vector tables in case of
3798 flag_pic.
3799 * doc/tm.texi : Regenerate.
3800 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3801 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3802 of addr diff vector generation.
3803 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3804 target hook definition.
3805 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3806 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3807 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3808 option for data text relative.
3810 2018-04-30 Richard Biener <rguenther@suse.de>
3812 * tree-chrec.h (evolution_function_is_constant_p): Remove
3813 redundant check.
3814 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3816 2018-04-30 Richard Biener <rguenther@suse.de>
3818 PR bootstrap/85571
3819 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3821 2018-04-30 Richard Biener <rguenther@suse.de>
3823 PR tree-optimization/28364
3824 PR tree-optimization/85275
3825 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3826 copying first exit test.
3828 2018-04-28 Mark Wielaard <mark@klomp.org>
3830 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3831 dwarf_version >= 5.
3832 (dwarf_AT): Handle DW_AT_addr_base.
3833 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3835 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3837 PR target/84431
3838 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3839 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3840 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3841 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3843 2018-04-28 Richard Biener <rguenther@suse.de>
3845 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3846 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3847 to reflect use. Only add interesting stmts.
3849 2018-04-27 Martin Jambor <mjambor@suse.cz>
3851 PR ipa/85549
3852 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3853 the jump function allows for passing through aggregate values.
3855 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3857 * input.h (in_system_header_at): Convert from macro to inline
3858 function.
3859 (from_macro_expansion_at): Likewise.
3860 (from_macro_definition_at): Likewise.
3862 2018-04-27 Jeff Law <law@redhat.com>
3864 * config.gcc: Mark tile* targets as deprecated/obsolete.
3866 2018-04-27 Richard Biener <rguenther@suse.de>
3868 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3869 fix for ILP32.
3871 2018-04-27 Richard Biener <rguenther@suse.de>
3873 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3875 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3877 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3878 with Yd constraint. Set "preferred_for_speed" attribute from
3879 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3880 with Yd constraint.
3881 (*movdi_internal): Ditto.
3882 (movti_interunit splitters): Remove
3883 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3884 (movdi_interunit splitters): Ditto.
3885 * config/i386/constraints.md (Ye): Remove.
3886 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3888 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3890 PR target/85512
3891 * config/aarch64/constraints.md (Usg): Limit to 31.
3892 (Usj): Limit to 63.
3894 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3896 PR tree-optimization/85529
3897 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3898 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3899 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3900 zero extension or masking of the MSB bit.
3901 (optimize_range_tests): Add FIRST_BB argument, pass it through
3902 to optimize_range_tests_var_bound.
3903 (maybe_optimize_range_tests, reassociate_bb): Adjust
3904 optimize_range_tests callers.
3906 2018-04-26 Richard Biener <rguenther@suse.de>
3907 Jakub Jelinek <jakub@redhat.com>
3909 * cgraph.h (symbol_table): Just declare debug method here.
3910 * symtab.c (symbol_table::debug): Define.
3912 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3914 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3916 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3918 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3919 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3920 (*movdi_internal): Substitute Yi and Yj constraint with x
3921 and Ym and Yn constraint with y constraint. Update "isa"
3922 attribute and set "preferred_for_speed" attribute from
3923 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3924 (*movsi_internal): Ditto.
3925 (*movdf_internal): Ditto.
3926 (*movsf_internal): Ditto.
3927 (*zero_extendsidi2): Ditto.
3928 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3929 (sse2_loadld): Ditto.
3930 (*vec_extract<ssevecmodelower>_0): Ditto.
3931 (*vec_extractv4si_0_zext_sse4): Ditto.
3932 (vec_concatv2di): Ditto.
3933 (*vec_dup<mode>): Ditto.
3934 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3935 * config/i386/constraints.md (Yi): Remove.
3936 (Yj): Remove.
3937 (Ym): Remove.
3938 (Yn): Remove.
3940 2018-04-26 Nathan Sidwell <nathan@acm.org>
3942 * dumpfile.c (dump_open): New.
3943 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3944 (dump_finish): Detect stdio/stderr by value not name.
3946 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3948 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3950 2018-04-26 Tom de Vries <tom@codesourcery.com>
3952 PR target/84952
3953 * config/nvptx/nvptx.c (verify_neutering_jumps)
3954 (verify_neutering_labels): New function
3955 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3957 2018-04-26 Tom de Vries <tom@codesourcery.com>
3959 PR target/84025
3960 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3961 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3962 need neutering.
3964 2018-04-26 Richard Biener <rguenther@suse.de>
3965 Tom de Vries <tom@codesourcery.com>
3967 PR lto/85422
3968 * lto-streamer-out.c (output_function): Fixup loops if required to match
3969 discovery done in the reader.
3971 2018-04-26 Richard Biener <rguenther@suse.de>
3973 PR tree-optimization/85116
3974 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3975 have a loop exit from the single latch predecessor. Remove
3976 case of header with just condition.
3977 (ch_base::copy_headers): Exclude infinite loops from any
3978 processing.
3979 (pass_ch::execute): Record exits.
3981 2018-04-26 Richard Biener <rguenther@suse.de>
3983 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3984 prologue cost vector and pass it to vect_get_load_cost.
3985 (vect_get_peeling_costs_all_drs): Likewise.
3986 (vect_peeling_hash_get_lowest_cost): Likewise.
3987 (vect_enhance_data_refs_alignment): Likewise.
3989 2018-04-26 Richard Biener <rguenther@suse.de>
3991 PR middle-end/85450
3992 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3993 checking of integer<->pointer conversions.
3994 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3995 sign-/zero-extending pointer types.
3996 (expand_omp_for_static_chunk): Likewise.
3998 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3999 Jean Lee <xiaoyur347@gmail.com>
4001 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4002 (TARGET_ASAN_SHADOW_OFFSET): Define.
4003 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4004 true for -fsanitize=address.
4006 2018-04-25 Mark Wielaard <mark@klomp.org>
4008 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4009 shorter ones.
4011 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4013 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4014 than "alu", remove explicit "memory" and "imm_disp" attributes.
4015 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4017 PR middle-end/85414
4018 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4019 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4020 gen_lowpart_no_emit.
4022 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4024 PR target/85473
4025 * config/i386/i386.c (ix86_expand_builtin): Change memory
4026 operand to XI, extend p0 to Pmode.
4027 * config/i386/i386.md: Change unspec volatile and operand
4028 1 mode to XI, change operand 0 mode to P.
4030 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4032 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4033 GET_MODE_MASK before any checking.
4034 (nds32_can_use_bset_p): Likewise.
4035 (nds32_can_use_btgl_p): Likewise.
4037 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4039 * config/nds32/nds32-doubleword.md: New define_split pattern for
4040 illegal register number.
4042 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4044 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4046 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4048 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4050 2018-04-25 Richard Biener <rguenther@suse.de>
4052 * lto-streamer.h (LTO_major_version): Bump to 8.
4054 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4056 * BASE-VER: Set to 9.0.0.
4058 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4060 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4061 in __abskf2 and __powikf2.
4063 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4065 PR target/85512
4066 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4067 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4068 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4069 Use the above on operand 2. Reindent.
4070 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4072 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4074 PR target/85485
4075 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4076 handle OPT_mcet.
4077 * config/i386/i386.opt (mcet): Removed.
4078 * doc/install.texi: Remove -mcet documentation.
4079 * doc/invoke.texi: Likewise.
4081 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4083 PR target/85485
4084 * doc/install.texi: Remove -mcet from bootstrap-cet.
4086 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4088 PR target/85511
4089 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4090 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4091 if TARGET_64BIT.
4093 PR target/85503
4094 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4095 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4096 containing a CONST_VECTOR.
4098 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4100 * doc/install.texi: Update newlib dependency for nvptx.
4102 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4104 PR target/85508
4105 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4106 instead of INTVAL when shifting x left.
4108 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4110 PR tree-optimization/85478
4111 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4112 vect_grouped_store_supported for single element vectors.
4114 2018-04-24 Richard Biener <rguenther@suse.de>
4116 PR target/85491
4117 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4118 load cost increase to the case of non-constant step.
4120 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4122 PR target/84828
4123 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4124 destination if any_malformed_asm.
4126 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4128 PR middle-end/85496
4129 * expr.c (store_field): In the bitfield case, if the value comes from
4130 a function call and is returned in registers by means of a PARALLEL,
4131 do not change the mode of the temporary unless BLKmode and VOIDmode.
4133 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4135 PR rtl-optimization/85423
4136 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4137 dependencies to debug insns when the previous insn is non-debug.
4139 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4141 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4142 enums into a single definition.
4143 (fls): Fix predicates and printing.
4144 (seti): Likewise.
4146 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4148 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4149 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4150 and short u6 immediate.
4151 (check_if_valid_sleep_operand): Remove.
4152 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4154 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4156 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4157 flag_always_save_lp condition.
4158 * config/nds32/nds32.opt (malways-save-lp): New option.
4160 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4162 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4163 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4164 * config/nds32/nds32.h
4165 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4166 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4168 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4170 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4171 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4173 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4174 Chung-Ju Wu <jasonwucj@gmail.com>
4176 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4177 Declare.
4178 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4179 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4181 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4183 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4185 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4187 * config/nds32/nds32-protos.h (nds32_data_alignment,
4188 nds32_local_alignment): Declare.
4189 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4190 nds32_local_alignment): New functions.
4191 (TARGET_CONSTANT_ALIGNMENT): Define.
4192 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4194 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4196 * config/nds32/nds32.c
4197 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4198 (TARGET_MODES_TIEABLE_P): Likewise.
4200 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4202 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4203 level Ofast and Og.
4205 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4206 Chung-Ju Wu <jasonwucj@gmail.com>
4208 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4209 for unaligned access.
4210 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4211 unaligned access.
4212 * config/nds32/nds32-intrinsic.md: Likewise.
4213 * config/nds32/nds32_intrinsic.h: Likewise.
4214 * config/nds32/nds32.h (nds32_builtins): Likewise.
4215 * config/nds32/nds32.opt (munaligned-access): New option.
4216 * config/nds32/nds32.c (nds32_asm_file_start): Display
4217 flag_unaligned_access status.
4219 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4221 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4222 -mno-relax is present.
4223 * config/riscv/linux.h (LINK_SPEC): Ditto.
4225 2018-04-20 Martin Sebor <msebor@redhat.com>
4227 PR c/85365
4228 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4229 for null pointers.
4230 (gimple_fold_builtin_stxcpy_chk): Same.
4231 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4233 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4235 PR target/85456
4236 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4237 __powikf2 when long double is IEEE 128-bit.
4239 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4241 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4242 step to make sure stack always aligned.
4244 2018-04-20 Carl Love <cel@us.ibm.com>
4246 PR target/83402
4247 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4248 size check for arg0.
4250 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4251 Tom de Vries <tom@codesourcery.com>
4253 PR target/85445
4254 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4255 Emit insns for calls too.
4256 (nvptx_find_par): Always look for worker-level predecessor insn.
4257 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4258 calls.
4259 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4260 (nvptx_process_pars): Propagate frames for calls.
4262 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4264 PR target/85469
4265 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4266 Removed.
4267 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4268 (ix86_handle_option): Don't handle OPT_mibt.
4269 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4270 __SHSTK__.
4271 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4272 has_ibt and ibt.
4273 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4274 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4275 (ix86_target_macros): Define __CET__ with flag_cf_protection
4276 for -fcf-protection.
4277 * config/i386/i386.c (isa2_opts): Remove -mibt.
4278 * config/i386/i386.h (TARGET_IBT): Removed.
4279 (TARGET_IBT_P): Likewise.
4280 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4281 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4282 * config/i386/i386.opt (mcet): Update help message.
4283 (mshstk): Likewise.
4284 (mibt): Removed.
4285 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4286 -mcet as an alias for -mshstk.
4288 2018-04-20 Richard Biener <rguenther@suse.de>
4290 PR middle-end/85475
4291 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4292 complexity by forcing a single use of the multiply operand.
4294 2018-04-20 Martin Jambor <mjambor@suse.cz>
4296 ipa/85449
4297 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4298 recursion dependency to only apply to non-clones.
4300 2018-04-20 Martin Jambor <mjambor@suse.cz>
4302 ipa/85447
4303 * ipa-cp.c (create_specialized_node): Check that clones of
4304 self-recursive edges exist during IPA-CP.
4306 2018-04-19 Toon Moene <toon@moene.org>
4308 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4309 by -O3.
4311 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4313 PR tree-optimization/85467
4314 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4315 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4316 VECTOR_CST element to type.
4318 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4320 PR target/85397
4321 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4322 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4323 (builtin_longjmp): Likewise.
4324 (save_stack_nonlocal): New pattern.
4325 (restore_stack_nonlocal): Likewise.
4327 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4329 PR target/85404
4330 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4331 Replace ASM_OUTPUT_LABEL with fprintf.
4333 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4335 PR target/85417
4336 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4337 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4338 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4339 define __IBT__ and __SHSTK__ for -fcf-protection.
4340 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4341 TARGET_IBT.
4342 (ix86_trampoline_init): Likewise.
4343 (x86_output_mi_thunk): Likewise.
4344 (ix86_notrack_prefixed_insn_p): Likewise.
4345 (ix86_option_override_internal): Don't disallow -fcf-protection.
4346 * config/i386/i386.md (rdssp<mode>): Also enable for
4347 -fcf-protection.
4348 (incssp<mode>): Likewise.
4349 (nop_endbr): Likewise.
4350 * config/i386/i386.opt (mcet): Change help message to built-in
4351 functions only.
4352 (mibt): Likewise.
4353 (mshstk): Likewise.
4354 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4355 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4356 enable CET built-in functions.
4358 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4360 * common/config/i386/i386-common.c
4361 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4362 OPTION_MASK_ISA_MOVDIRI_UNSET,
4363 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4364 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4365 * config.gcc (movdirintrin.h): New header.
4366 * config/i386/cpuid.h (bit_MOVDIRI,
4367 bit_MOVDIR64B): New bits.
4368 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4369 and -mmvodir64b.
4370 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4371 (VOID, PVOID, PCVOID)): New function types.
4372 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4373 __builtin_ia32_directstoreu_u64,
4374 __builtin_ia32_movdir64b): New builtins.
4375 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4376 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4377 and -mmovdiri.
4378 (ix86_valid_target_attribute_inner_p): Ditto.
4379 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4380 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4381 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4382 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4383 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4384 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4385 (movdiri<mode>, movdir64b_<mode>): New.
4386 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4387 * config/i386/immintrin.h: Include movdirintrin.h.
4388 * config/i386/movdirintrin.h: New file.
4389 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4391 2018-04-19 Richard Biener <rguenther@suse.de>
4393 PR middle-end/85455
4394 * cfg.c (clear_bb_flags): When loop state says we have
4395 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4397 2018-04-19 Richard Biener <rguenther@suse.de>
4399 PR tree-optimization/84737
4400 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4401 copying restrict info.
4402 (vect_setup_realignment): Use it.
4403 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4404 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4405 the first DR to all generated stores.
4406 (vectorizable_load): Likewise for loads.
4408 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4410 PR tree-optimization/85446
4411 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4412 the integral and pointer types to have the same precision.
4414 * doc/install.texi: Document --disable-cet being the default and
4415 --enable-cet=auto.
4417 2018-04-18 Martin Liska <mliska@suse.cz>
4419 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4420 style.
4422 2018-04-18 Martin Liska <mliska@suse.cz>
4424 Revert
4425 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4427 PR ipa/83983
4428 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4429 arguments if they are comparable.
4431 2018-04-18 Martin Liska <mliska@suse.cz>
4433 Revert
4434 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4436 PR lto/84805
4437 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4438 incomplete types.
4440 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4442 PR target/85388
4443 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4444 ENDBR after calling __morestack.
4446 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4448 PR jit/85384
4449 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4450 by using gcc_base_ver to generate a gcc_driver_version, and use
4451 it when generating GCC_DRIVER_NAME.
4452 * configure: Regenerate.
4454 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4456 PR target/81084
4457 * config.gcc: Obsolete powerpc*-*-*spe*.
4459 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4461 PR debug/84637
4462 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4463 (stabstr_D): Change type of unum from unsigned int to
4464 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4465 type.
4467 2018-04-17 Jim Wilson <jimw@sifive.com>
4469 PR 84856
4470 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4471 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4472 Set arg_pointer_offset after using pretend_args_size.
4474 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4476 PR rtl-optimization/85431
4477 * dse.c (record_store): Ignore zero width stores.
4479 PR sanitizer/85230
4480 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4481 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4482 __builtin_stack_restore rather than after it.
4483 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4484 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4485 argument instead of virtual_dynamic_stack_rtx.
4487 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4489 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4490 New prototype.
4491 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4492 Add note to error message to explain internal mapping of overloaded
4493 built-in function name to non-overloaded built-in function name.
4494 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4495 function.
4497 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4499 PR target/85424
4500 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4501 where the inputs overlap with the output.
4503 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4505 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4506 (=v, v) alternative and explicit "memory" attribute.
4507 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4508 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4509 attributes.
4510 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4511 "sselog1" type instead of "sselog".
4512 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4513 "sselog". Remove explicit "memory" attribute.
4514 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4515 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4516 attributes.
4517 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4518 "sselog1" type instead of "sselog". Remove explicit "memory"
4519 attribute.
4520 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4521 use "sselog1" type instead of "sselog". Remove explicit "memory"
4522 attribute.
4523 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4524 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4525 attributes.
4526 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4527 "sselog1" type instead of "sselog". Remove explicit "memory"
4528 attribute.
4529 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4530 use "sselog1" type instead of "sselog". Remove explicit "memory"
4531 attribute.
4533 PR target/85430
4534 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4536 PR middle-end/85414
4537 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4538 on a SUBREG.
4540 2018-04-17 Martin Jambor <mjambor@suse.cz>
4542 PR ipa/85421
4543 * ipa-cp.c (create_specialized_node): Call
4544 expand_all_artificial_thunks if necessary.
4546 2018-04-17 Martin Liska <mliska@suse.cz>
4548 PR lto/85405
4549 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4550 in message, remote space in between '_G' and '('.
4552 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4554 PR target/85281
4555 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4556 avx512f_vmcmp<mode>3<round_saeonly_name>,
4557 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4558 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4559 avx512f_rndscale<mode><round_saeonly_name>,
4560 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4561 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4562 Use %<iptr>2 instead of %2 for -masm=intel.
4563 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4564 avx512f_vcvttss2usi<round_saeonly_name>,
4565 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4566 -masm=intel.
4567 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4568 avx512f_vcvttsd2usi<round_saeonly_name>,
4569 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4570 Use %q1 instead of %1 for -masm=intel.
4571 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4572 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4573 of %3 for -masm=intel.
4574 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4575 -masm=intel.
4576 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4577 -masm=intel.
4578 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4579 -masm=intel.
4580 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4581 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4582 %g1.
4583 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4584 -masm=intel.
4585 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4586 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4587 %g1 and one with %0 and %1.
4588 (avx512er_vmrcp28<mode><round_saeonly_name>,
4589 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4590 %1 for -masm=intel.
4591 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4592 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4593 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4594 of %0 and %{%4%} for -masm=intel.
4595 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4596 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4597 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4598 order of %0 and %{%5%}%{z%} for -masm=intel.
4600 2018-04-17 Jan Hubicka <jh@suse.cz>
4602 PR lto/85405
4603 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4605 2018-04-17 Martin Liska <mliska@suse.cz>
4607 PR ipa/85329
4608 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4609 for target_clone error message. Make default implementation
4610 clone to be a local declaration.
4611 (separate_attrs): Add new argument and check for an empty
4612 string.
4613 (expand_target_clones): Handle it.
4614 (ipa_target_clone): Make redirection just for target_clones
4615 functions.
4617 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4618 Tom de Vries <tom@codesourcery.com>
4620 PR middle-end/84955
4621 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4622 tiled basic blocks without omp continue statements.
4624 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4626 PR target/83660
4627 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4628 vec_extract expression as having side effects to make sure it gets
4629 a cleanup point.
4631 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4633 PR target/85403
4634 * config/i386/i386.c (get_builtin_code_for_version): Check
4635 error_mark_node.
4637 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4639 PR target/84331
4640 * gcc/config.gcc: Support "skylake".
4641 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4642 PROCESSOR_SKYLAKE.
4643 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4644 (processor_target_table): Add "skylake".
4645 (ix86_option_override_internal): Add "skylake".
4646 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4647 PROCESSOR_CANNONLAKE.
4648 (get_builtin_code_for_version): Fix priority for
4649 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4650 PROCESSOR_SKYLAKE-AVX512.
4651 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4652 (processor_type): Add PROCESSOR_SKYLAKE.
4654 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4655 Jason Merrill <jason@redhat.com>
4657 PR c++/85112
4658 * convert.c (convert_to_integer_1): Use direct recursion for
4659 enumeral types and types with a precision less than the number
4660 of bits in their mode.
4662 2018-04-16 Julia Koval <julia.koval@intel.com>
4664 PR target/84413
4665 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4666 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4668 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4670 PR target/85293
4671 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4672 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4673 and -mno-direct-move.
4675 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4677 PR target/83402
4678 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4679 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4680 Ensure negative shifts result in {0}.
4682 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4684 PR rtl-optimization/79916
4685 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4686 regs (if any) to define how to gnerate SD moves when LRA is in
4687 progress.
4689 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4691 PR rtl-optimization/85393
4692 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4693 * except.c (expand_dw2_landing_pad_for_region): Make static.
4694 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4695 a label and unconditional jump to old_bb, rather than
4696 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4697 basic block.
4699 PR rtl-optimization/85376
4700 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4701 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4702 instead of a specific value.
4704 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4705 Bin Cheng <bin.cheng@arm.com>
4707 PR tree-optimization/82965
4708 PR tree-optimization/83991
4709 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4710 by_profile_only parameter.
4711 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4712 information if the loop was predicted to iterate too many times.
4713 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4715 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4717 PR lto/71991
4718 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4719 always inline.
4721 2018-04-13 Martin Liska <mliska@suse.cz>
4722 Jakub Jelinek <jakub@redhat.com>
4724 PR middle-end/81657
4725 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4726 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4727 * builtins.c (expand_builtin_memory_copy_args): Use
4728 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4729 handle dest_addr == pc_rtx.
4731 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4733 PR target/85291
4734 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4735 asked to not generate direct moves.
4736 (fix_trunc<mode>si2_stfiwx): Similar.
4737 (fix_trunc<mode>si2_internal): Similar.
4739 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4741 PR debug/83157
4742 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4743 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4744 lookup if dest in some wider mode is known to be const0_rtx and
4745 if so, record permanent equivalence for it to be ZERO_EXTEND of
4746 the narrower mode destination.
4748 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4750 * lto-streamer-out.c (output_function): Revert 259346.
4751 * omp-expand.c (expand_oacc_for): Likewise.
4753 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4755 PR rtl-optimization/85354
4756 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4757 * sel-sched.c (sel_global_init): ... here.
4759 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4761 PR target/85238
4762 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4763 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4764 mode for PE-COFF targets.
4765 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4766 (i386_pe_asm_lto_end): Likewise.
4767 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4768 (TARGET_ASM_LTO_END): Likewise.
4769 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4770 (i386_pe_asm_lto_start): New function.
4771 (i386_pe_asm_lto_end): Likewise.
4773 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4774 Richard Biener <rguenther@suse.de>
4776 PR middle-end/84955
4777 * lto-streamer-out.c (output_function): Fix CFG loop state before
4778 streaming out.
4779 * omp-expand.c (expand_oacc_for): Handle calls to internal
4780 functions like regular functions.
4782 2018-04-12 Richard Biener <rguenther@suse.de>
4784 PR lto/85371
4785 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4786 for the early LTO debug to properly generate references to it
4787 during DIE emission. Do not re-use that for the skeleton for
4788 split-dwarf.
4789 (dwarf2out_early_finish): Likewise.
4791 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4793 PR target/85328
4794 * config/i386/sse.md
4795 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4796 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4797 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4798 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4799 and output is a reg, avoid creating invalid lowpart subreg, but
4800 instead split into a 512-bit move. Don't split if not AVX512VL,
4801 input is xmm16+ reg and output is a mem.
4802 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4803 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4804 xmm16+ reg and output is a mem.
4806 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4808 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4809 also for flag_dwarf2_cfi_asm.
4811 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4813 PR rtl-optimization/85342
4814 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4815 a bool scalar var inside of the loop instead. Don't try to update
4816 recog_data.operand after failed apply_change_group.
4818 2018-04-12 Tom de Vries <tom@codesourcery.com>
4820 PR target/85296
4821 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4822 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4823 array with flexible array member as array without given dimension.
4824 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4825 argument for undefined param to true.
4827 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4829 PR target/85321
4830 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4831 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4832 from PowerPC section.
4833 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4834 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4835 help text that is too long.
4836 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4837 help text that is too long.
4838 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4839 help text that is too long.
4841 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4843 * config/alpha/alpha.md (stack_probe_internal): Rename
4844 from "probe_stack". Update all callers.
4846 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4848 PR rtl-optimization/84566
4849 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4850 sched_macro_fuse_insns.
4852 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4854 PR target/84301
4855 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4856 (compute_block_dependences): ... from here.
4858 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4860 PR tree-optimization/85331
4861 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4862 from int to HOST_WIDE_INT.
4864 2018-04-11 Martin Jambor <mjambor@suse.cz>
4866 PR ipa/84149
4867 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4868 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4869 not the same as the source val.
4870 (cgraph_edge_brings_value_p): New parameter.
4871 (gather_edges_for_value): Pass destination value to
4872 cgraph_edge_brings_value_p.
4873 (perhaps_add_new_callers): Likewise.
4874 (get_info_about_necessary_edges): Likewise and exclude values brought
4875 only by self-recursive edges.
4876 (create_specialized_node): Redirect only clones of self-calling edges.
4877 (+self_recursive_pass_through_p): New function.
4878 (find_more_scalar_values_for_callers_subset): Use it.
4879 (find_aggregate_values_for_callers_subset): Likewise.
4880 (known_aggs_to_agg_replacement_list): Removed.
4881 (decide_whether_version_node): Re-calculate known constants for all
4882 remaining context clones.
4884 2018-04-11 Richard Biener <rguenther@suse.de>
4886 PR lto/85339
4887 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4888 from early DWARF output.
4889 (dwarf2out_early_finish): Output line info unconditionally into
4890 early DWARF and add reference to it.
4892 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4894 PR target/85281
4895 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4896 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4897 other than V2DFmode using iptr mode attribute.
4898 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4900 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4902 PR rtl-optimization/84659
4903 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4905 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4907 PR debug/85302
4908 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4909 SIZEP is NULL.
4910 (output_loc_list): Pass address of a dummy size variable even in the
4911 locview handling loop.
4912 (index_location_lists): Add comment on why skip_loc_list_entry can't
4913 call size_of_locs.
4915 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4917 PR target/85261
4918 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4919 into register.
4921 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4923 PR target/85321
4924 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4925 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4926 and -mstring-compare-inline-limit.
4928 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4930 PR target/85287
4931 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4932 for stack clash protection in a register whenever we need it to be in
4933 a register.
4935 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4937 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4938 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4940 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4942 PR target/85321
4943 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4944 the help text.
4945 (mlong-double-): Ditto.
4946 * config/rs6000/sysv4.opt (msdata=): Ditto.
4947 (mtls-size=): Ditto.
4949 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4951 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4952 erroneous entries for
4953 "vector int vec_ldl (int, long int *)", and
4954 "vector unsigned int vec_ldl (int, unsigned long int *)".
4955 Add comments and entries for
4956 "vector bool char vec_ldl (int, bool char *)",
4957 "vector bool short vec_ldl (int, bool short *)",
4958 "vector bool int vec_ldl (int, bool int *)",
4959 "vector bool long long vec_ldl (int, bool long long *)",
4960 "vector pixel vec_ldl (int, pixel *)",
4961 "vector long long vec_ldl (int, long long *)",
4962 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4963 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4964 type tree bool_long_long_type_node and correct definition of
4965 bool_V2DI_type_node to make reference to this new type tree.
4966 (rs6000_mangle_type): Replace erroneous reference to
4967 bool_long_type_node with bool_long_long_type_node.
4968 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4969 comments to emphasize sign distinctions for char and int types and
4970 replace RS6000_BTI_bool_long constant with
4971 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4972 use of RS6000_BTI_pixel.
4973 (bool_long_type_node): Remove this macro definition.
4974 (bool_long_long_type_node): New macro definition
4976 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4978 PR rtl-optimization/85300
4979 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4980 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4981 simplify_unary_operation fails.
4983 2018-04-10 Martin Liska <mliska@suse.cz>
4985 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4986 cgraph_edge and ipa_ref.
4988 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4990 PR target/85177
4991 PR target/85255
4992 * config/i386/sse.md
4993 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4994 computation of the VEC_MERGE selector from mask.
4995 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4996 Fix decoding of the VEC_MERGE selector into mask.
4998 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5000 PR tree-optimization/85286
5001 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5003 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5005 * final.c (final_1): Set insn_last_address as well as
5006 insn_current_address.
5008 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5010 PR target/85173
5011 * explow.c (emit_stack_probe): Call validize_mem on memory location
5012 before passing it to gen_probe_stack. Create address operand and
5013 legitimize it for the probe_stack_address case.
5015 2018-04-09 Jan Hubicka <jh@suse.cz>
5017 PR lto/85078
5018 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5019 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5020 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5021 walk basetypes.
5022 (free_lang_data): Rebuild type inheritance graph.
5024 2018-04-09 Martin Sebor <msebor@redhat.com>
5026 * invoke.texi (-finline-small-functions): Mention other optimization
5027 options.
5028 (-findirect-inlining, -fpartial-inlining): Same.
5029 (-finline-functions-called-once): Same.
5030 (-freorder-blocks-and-partition): Same.
5032 2018-04-09 Jan Hubicka <jh@suse.cz>
5034 PR rtl/84058
5035 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5036 jumps; choose last target that matches the criteria (i.e.
5037 no partition changes for non-crossing jumps).
5038 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5039 support for redirecting crossing jumps to non-crossing.
5041 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5043 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5044 also for naked functions.
5046 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5048 * config/arc/arc.md (add_shift): New pattern.
5049 (add_shift2): Likewise.
5050 (sub_shift): Likewise.
5051 (sub_shift_cmp0_noout): Likewise.
5052 (compare_si_ashiftsi): Likewise.
5053 (xbfu_cmp0_noout): New combine pattern.
5054 (xbfu_cmp0"): Likewise.
5055 (movsi_set_cc_insn): Place the predicable variant first.
5056 (commutative_binary_cmp0_noout): Remove clobber.
5057 (commutative_binary_cmp0): New pattern.
5058 (noncommutative_binary_cmp0): Likewise.
5059 (noncommutative_binary_cmp0_noout): Likewise.
5060 (noncommutative_binary_comparison_result_used): Removed.
5061 (rsub_cmp0): New pattern.
5062 (rsub_cmp0_noout): Likewise.
5063 (extzvsi): Changed, keep only meaningful variants.
5064 (SQH, SEZ): New iterators.
5065 (SQH_postfix): New mode attribute.
5066 (SEZ_prefix): New code attribute.
5067 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5068 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5069 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5070 of numerical value.
5071 (noncommutative_operator): Check the availability of barrel
5072 shifter option.
5074 2018-04-09 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/85284
5077 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5078 Only use the niter constraining form of simple_iv when the exit
5079 is always executed.
5081 2018-04-09 Tom de Vries <tom@codesourcery.com>
5083 PR target/84041
5084 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5085 (define_expand "*memory_barrier"): New define_expand.
5086 (define_insn "memory_barrier"): New insn.
5088 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5090 PR rtl-optimization/80463
5091 PR rtl-optimization/83972
5092 PR rtl-optimization/83480
5094 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5095 correct producer for the insn.
5096 (tidy_control_flow): Fixup seqnos in case of debug insns.
5098 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5100 PR rtl-optimization/83913
5102 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5103 different sched-times when merging exprs.
5105 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5107 PR rtl-optimization/83962
5109 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5110 tidy_fallthru_edge and tidy_control_flow.
5112 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5114 PR rtl-optimization/83530
5116 * sel-sched.c (force_next_insn): New global variable.
5117 (remove_insn_for_debug): When force_next_insn is true, also leave only
5118 next insn in the ready list.
5119 (sel_sched_region): When the region wasn't scheduled, make another pass
5120 over it with force_next_insn set to 1.
5122 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5124 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5125 into tm_file.
5126 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5127 for interrupt control.
5128 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5129 functions for interrupt control.
5130 * config/nds32/nds32-intrinsic.md: Likewise.
5131 * config/nds32/nds32_intrinsic.h: Likewise.
5132 * config/nds32/nds32.h (nds32_builtins): Likewise.
5134 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5136 * config/nds32/nds32.c (nds32_init_machine_status,
5137 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5138 strict_aligned_p field.
5139 (nds32_expand_to_rtl_hook): New function.
5140 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5141 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5143 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5144 Chung-Ju Wu <jasonwucj@gmail.com>
5146 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5147 * config/nds32/nds32-n7.md: New file.
5148 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5149 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5150 pipeline.
5151 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5152 * config/nds32/nds32.md (pipeline_model): Add n7.
5153 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5154 * config/nds32/pipelines.md: Include n7 settings.
5156 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5157 Chung-Ju Wu <jasonwucj@gmail.com>
5159 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5160 * config/nds32/nds32-e8.md: New file.
5161 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5162 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5163 pipeline.
5164 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5165 * config/nds32/nds32.md (pipeline_model): Add e8.
5166 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5167 * config/nds32/pipelines.md: Include e8 settings.
5169 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5170 Chung-Ju Wu <jasonwucj@gmail.com>
5172 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5173 * config/nds32/nds32-n8.md: New file.
5174 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5175 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5176 pipeline.
5177 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5178 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5179 * config/nds32/nds32.md (pipeline_model): Add n8.
5180 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5181 * config/nds32/pipelines.md: Include n8 settings.
5183 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5184 Chung-Ju Wu <jasonwucj@gmail.com>
5186 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5187 * config/nds32/nds32-n9-2r1w.md: New file.
5188 * config/nds32/nds32-n9-3r2w.md: New file.
5189 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5190 nds32_register_ports): New or modify for cpu n9.
5191 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5192 pipeline.
5193 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5194 * config/nds32/nds32-utils.c: New file.
5195 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5196 TARGET_MUL_SLOW): Define.
5197 * config/nds32/nds32.md (pipeline_model): New attribute.
5198 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5199 New options that support cpu n9.
5200 * config/nds32/pipelines.md: Include n9 settings.
5201 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5203 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5206 information if necessary.
5207 (output_cond_branch_compare_zero): Likewise.
5208 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5209 (nds32_target_alignment): Refine for alignment.
5210 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5211 (FUNCTION_BOUNDARY): Modify.
5212 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5213 align case.
5214 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5216 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5218 * config/nds32/constants.md (unspec_volatile_element): Add values for
5219 TLB operation and data prefetch.
5220 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5221 functions for TLB operation and data prefetch.
5222 * config/nds32/nds32-intrinsic.md: Likewise.
5223 * config/nds32/nds32_intrinsic.h: Likewise.
5224 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5225 (nds32_print_operand): Likewise.
5226 * config/nds32/nds32.h (nds32_builtins): Likewise.
5228 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5229 Andrew Pinski <pinsika@gcc.gnu.org>
5231 PR middle-end/82976
5232 * match.pd: Use constant_boolean_node of correct type instead of
5233 boolean_true_node or boolean_false_node for simplifying
5234 pointer comparisons to zero.
5236 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5238 PR tree-optimization/80021
5239 * tree.c (verify_type_variant): Make error call in verify_variant_match
5240 translatable and remove final full stop.
5242 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5244 * config/nds32/constants.md (unspec_volatile_element): Add
5245 UNSPEC_VOLATILE_EH_RETURN.
5246 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5247 nds32_output_stack_pop): Support dwarf exception handling process.
5248 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5249 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5250 exception handling process.
5251 (nds32_compute_stack_frame): Likewise.
5252 (nds32_return_addr_rtx): Likewise.
5253 (nds32_initial_elimination_offset): Likewise.
5254 (nds32_expand_prologue): Likewise.
5255 (nds32_expand_epilogue): Likewise.
5256 (nds32_dynamic_chain_address): New function.
5257 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5258 exception handling.
5259 (DYNAMIC_CHAIN_ADDRESS): Define.
5260 (EH_RETURN_DATA_REGNO): Define.
5261 (EH_RETURN_STACKADJ_RTX): Define.
5262 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5263 patterns for dwarf exception handling.
5265 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5267 * config/nds32/nds32.h: Clean up obsolete macros.
5269 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5271 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5272 Add enum values for particular instructions.
5273 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5274 particular intrinsic functions.
5275 * config/nds32/nds32-intrinsic.md: Likewise.
5276 * config/nds32/nds32_intrinsic.h: Likewise.
5277 * config/nds32/nds32.h (nds32_builtins): Likewise.
5278 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5279 (btst, ave): New patterns for particular instructions.
5281 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5283 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5284 Add enum values for atomic load/store and memory sync.
5285 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5286 and memory sync.
5287 * config/nds32/nds32-intrinsic.md: Likewise.
5288 * config/nds32/nds32_intrinsic.h: Likewise.
5289 * config/nds32/nds32.h (nds32_builtins): Likewise.
5291 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5293 PR tree-optimization/85257
5294 * fold-const.c (native_encode_vector): If not all elts could fit
5295 and off is -1, return 0 rather than offset.
5296 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5297 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5298 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5299 adjust buffer in native_interpret_expr call.
5301 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5303 * config/nds32/constants.md (unspec_volatile_element): Add cache
5304 control enum values.
5305 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5306 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5307 * config/nds32/nds32.c (nds32_cctl_names): New.
5308 (nds32_print_operand): Handle cache control register names.
5309 * config/nds32/nds32.h (nds32_builtins): New enum values.
5310 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5311 macros.
5312 * config/nds32/nds32.md (type): Add mmu.
5313 * config/nds32/pipelines.md (simple_insn): Add mmu.
5315 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5317 * config/nds32/nds32.md (type): Remove call.
5318 * config/nds32/pipelines.md (simple_insn): Likewise.
5320 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5322 * config/nds32/constants.md (unspec_volatile_element): Add
5323 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5324 UNSPEC_VOLATILE_FMFCFG.
5325 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5326 description for fmfcfg and fmfcsr.
5327 (bdesc_1arg): Add fmtcsr.
5328 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5329 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5330 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5331 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5332 unspec_fmfcfg): New patterns.
5333 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5334 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5335 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5336 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5337 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5338 __nds32__fmfcfg): Define.
5340 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5342 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5343 intrinsic register names.
5344 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5345 intrinsic register enum values and macros.
5347 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5349 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5350 for load/store addressing form.
5351 (nds32_print_operand_address): Likewise.
5353 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5355 PR target/85196
5356 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5357 based on LABEL_REF. Remove useless assertion.
5358 (pic_address_needs_scratch): Fix formatting.
5359 (sparc_legitimize_pic_address): Minor tweaks.
5360 (sparc_delegitimize_address): Adjust assertion accordingly.
5361 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5362 into symbolic_operand.
5363 (movsi_high_pic_label_ref): Likewise.
5364 (movsi_lo_sum_pic_label_ref): Likewise.
5365 (movdi_pic_label_ref): Likewise.
5366 (movdi_high_pic_label_ref): Likewise.
5367 (movdi_lo_sum_pic_label_ref): Likewise.
5369 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5371 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5372 custom LIB_SPEC setup.
5374 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5375 Kito Cheng <kito.cheng@gmail.com>
5377 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5378 * config/riscv/freebsd.h: New.
5380 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5382 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5383 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5384 file.
5386 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5387 Kito Cheng <kito.cheng@gmail.com>
5389 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5390 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5391 * config/nds32/nds32-protos.h (nds32_output_call,
5392 nds32_output_return): Declare.
5393 * config/nds32/nds32.md: Refine all the call and return patterns.
5395 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5397 PR debug/85252
5398 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5399 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5401 PR rtl-optimization/84872
5402 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5403 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5404 EDGE_CROSSING edge.
5406 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5408 * expr.c (copy_blkmode_to_reg): Revert 254862.
5409 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5411 2018-04-06 Richard Biener <rguenther@suse.de>
5413 PR middle-end/85244
5414 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5415 after seeing a component reference with an adjacent field. Treat
5416 refs to arrays at struct end of external decls similar to
5417 refs to unconstrained commons.
5419 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5421 PR sanitizer/85213
5422 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5423 look through SAVE_EXPRs with non-side-effects argument. Adjust
5424 recursive calls.
5425 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5426 save_p here.
5428 2018-04-06 Richard Biener <rguenther@suse.de>
5430 PR middle-end/85180
5431 * alias.c (find_base_term): New wrapper around find_base_term
5432 unwinding CSELIB_VAL_PTR changes.
5433 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5434 recursion.
5436 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5438 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5439 instructions.
5440 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5441 constant definitions.
5442 ("nop"): lr 0,0 -> nopr r0
5443 ("nop_lr0", "nop_lr1"): New insn definitions.
5445 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5447 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5448 NDS32_V3PUSH_AVAILABLE_P macro.
5450 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5451 Chung-Ju Wu <jasonwucj@gmail.com>
5453 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5454 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5455 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5456 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5457 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5458 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5459 * config/nds32/constraints.md: New constraints and checking for hard
5460 float configuration.
5461 * config/nds32/iterators.md: New mode iterator and attribute for hard
5462 float configuration.
5463 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5464 patterns.
5465 * config/nds32/nds32-fpu.md: New file.
5466 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5467 deal with hard float code generation.
5468 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5469 ARCH_V3S.
5470 (abi_type, float_reg_number): New enum type.
5471 * config/nds32/nds32-predicates.c: New predicates for hard float.
5472 * config/nds32/nds32-protos.h: Declare functions for hard float.
5473 * config/nds32/nds32.c: Implementation for hard float configuration.
5474 * config/nds32/nds32.h: Definitions for hard float configuration.
5475 * config/nds32/nds32.md: Include hard float machine description and
5476 modify patterns for hard float configuration.
5477 * config/nds32/nds32.opt: New options for hard float configuration.
5478 * config/nds32/predicates.md: New predicates for hard float
5479 configuration.
5481 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5483 * common/config/nds32/nds32-common.c
5484 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5486 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5488 PR middle-end/85195
5489 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5490 CONSTRUCTOR_ELT (ctor, ...)->value.
5492 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5494 PR target/85193
5495 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5497 2018-04-05 Tom de Vries <tom@codesourcery.com>
5499 PR target/85204
5500 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5501 cond jump.
5503 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5504 Kito Cheng <kito.cheng@gmail.com>
5506 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5507 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5508 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5509 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5511 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5512 Kito Cheng <kito.cheng@gmail.com>
5514 * config/nds32/constraints.md (Ufe): New memory constraint.
5515 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5516 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5517 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5518 operands.
5519 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5520 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5522 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5524 * config/nds32/nds32.md: Use optimize_size in the condition for
5525 alu-shift instructions.
5527 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5529 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5531 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5533 * config/nds32/nds32.md (negsi2): Refine pattern.
5535 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5536 Chung-Ju Wu <jasonwucj@gmail.com>
5538 * config/nds32/iterators.md (shift_rotate): New code iterator.
5539 (shift): New code attribute.
5540 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5541 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5542 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5543 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5544 bit-wise operations.
5545 (andsi3, *andsi3): Ditto.
5546 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5547 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5548 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5549 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5550 nds32_ior_operand, nds32_xor_operand): New predicates.
5552 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5554 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5555 (addsi3, subsi3): ... this.
5557 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5559 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5561 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5563 * config/nds32/nds32.md: Adjust indention.
5565 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5567 * config/nds32/nds32.md (feature): New attribute.
5569 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config/nds32/nds32.md (subtype): New attribute.
5573 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5575 PR target/85203
5576 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5577 expansion to perform a bitwise AND of the argument followed by a
5578 boolean negation of the result.
5580 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5582 PR rtl-optimization/84878
5583 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5584 the basic block. Assert the use reference is not artificial and that
5585 it has an associated insn.
5587 2018-04-04 Michael Matz <matz@suse.de>
5589 * builtins.c (compute_objsize): Pass correct operand
5590 to array_at_struct_end_p.
5592 2018-04-04 Richard Biener <rguenther@suse.de>
5594 PR lto/85176
5595 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5596 from contexts for DINFO_LEVEL_TERSE and below.
5598 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5600 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5601 resiter_operand condition.
5602 * config/nds32/nds32.md (*move<mode>): Ditto.
5604 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5605 Monk Chiang <sh.chiang04@gmail.com>
5607 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5609 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5611 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5613 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5614 Kito Cheng <kito.cheng@gmail.com>
5616 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5617 nds32_cond_code_str, output_cond_branch,
5618 output_cond_branch_compare_zero, nds32_expand_cbranch,
5619 nds32_expand_cstore, nds32_expand_movcc,
5620 nds32_output_cbranchsi4_equality_zero,
5621 nds32_output_cbranchsi4_equality_reg,
5622 nds32_output_cbranchsi4_equality_reg_or_const_int,
5623 nds32_output_cbranchsi4_greater_less_zero: New functions.
5624 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5625 nds32_expand_cstore, nds32_expand_movcc,
5626 nds32_output_cbranchsi4_equality_zero,
5627 nds32_output_cbranchsi4_equality_reg,
5628 nds32_output_cbranchsi4_equality_reg_or_const_int,
5629 nds32_output_cbranchsi4_greater_less_zero): Declare.
5630 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5631 nds32_rimm11s_operand): New predicates.
5632 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5633 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5634 patterns.
5636 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5638 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5639 * config/nds32/nds32.md: Ditto.
5640 * config/nds32/pipelines.md: Ditto.
5642 2018-04-04 Richard Biener <rguenther@suse.de>
5644 PR tree-optimization/85168
5645 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5646 propagating abnormals.
5648 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5650 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5652 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5653 Kito Cheng <kito.cheng@gmail.com>
5655 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5656 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5657 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5658 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5659 * config/nds32/nds32.md (sibcall_internal): New.
5660 (sibcall_register): Remove.
5661 (sibcall_immediate): Remove.
5662 (sibcall_value_internal): New.
5663 (sibcall_value_register): Remove.
5664 (sibcall_value_immediate): Remove.
5665 * config/nds32/predicates.md (nds32_general_register_operand): New.
5666 (nds32_call_address_operand): New.
5668 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5670 PR rtl-optimization/85167
5671 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5672 bb_defs if *split_p, instead preinitialize it to NULL.
5674 PR tree-optimization/85156
5675 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5676 evaluating the argument multiple times.
5678 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5680 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5681 than vector.
5682 (_mm_cvtpd_ps): Likewise.
5683 (_mm_cvttpd_epi32): Likewise.
5684 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5685 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5686 vector, pixel, and bool following altivec.h include.
5688 2018-04-03 Martin Sebor <msebor@redhat.com>
5690 * doc/extend.texi (Common Function Attributes): Clarify.
5691 (const attribute): Likewise.
5692 (pure attribute): Likewise.
5694 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5696 PR target/85169
5697 * config/i386/i386.c (ix86_expand_vector_set): Use
5698 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5700 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5702 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5703 instructions when changing rounding bits to preserve precision bits
5704 in the x87 control word.
5706 2018-04-03 Martin Liska <mliska@suse.cz>
5708 PR tree-optimization/82491
5709 * rtl.h (strip_offset_and_add): Replace += suboffset with
5710 poly_uint64 () + suboffset.
5712 2018-03-29 Martin Liska <mliska@suse.cz>
5713 Martin Jambor <mjambor@suse.cz>
5715 PR ipa/84947
5716 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5717 param_type is not an integral or pointer type.
5719 2018-04-03 Richard Biener <rguenther@suse.de>
5721 * sese.h (recompute_all_dominators): Remove.
5723 2018-04-02 Martin Sebor <msebor@redhat.com>
5725 * doc/invoke.texi (-Wrestrict): Fix typos.
5727 2018-04-02 Jim Wilson <jimw@sifive.com>
5729 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5730 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5731 (<optab>di3, <optab>si3_extend): Likewise.
5732 (<optab>si3_mask, <optab>si3_mask_1): New.
5733 (<optab>di3_mask, <optab>di3_mask_1): New.
5734 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5735 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5736 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5738 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5740 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5741 example.
5743 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5745 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5746 (nds32_canonicalize_comparison): New function.
5748 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5749 Kito Cheng <kito.cheng@gmail.com>
5750 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5752 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5753 * config/nds32/constants.md (unspec_volatile_element): Add
5754 UNSPEC_VOLATILE_RELAX_GROUP.
5755 * config/nds32/nds32-relax-opt.c: New file.
5756 * config/nds32/nds32-predicates.c
5757 (nds32_symbol_load_store_p): New function.
5758 * config/nds32/nds32-protos.h
5759 (nds32_symbol_load_store_p): Declare function.
5760 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5761 * config/nds32/nds32.c
5762 (nds32_register_pass): New function to register pass.
5763 (nds32_register_passes): New function to register passes.
5764 * config/nds32/nds32.md (relax_group): New pattern.
5765 * config/nds32/nds32.opt (mrelax-hint): New option.
5766 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5768 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5770 * config/nds32/t-nds32: Modify files dependency.
5772 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5774 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5775 (PROFILE_HOOK): Define its implementation.
5777 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5779 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5780 type and 32-bit size.
5782 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5784 PR middle-end/85090
5785 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5786 (V_128_256): New mode iterator.
5787 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5788 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5789 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5790 of V.
5791 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5792 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5794 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5796 PR target/83315
5797 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5798 NaN inputs correctly.
5800 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5802 PR target/80546
5803 * config/rs6000/vsx.md (??r): New mode attribute.
5804 (*vsx_mov<mode>_64bit): Use it.
5805 (*vsx_mov<mode>_32bit): Likewise.
5807 2018-03-30 Martin Sebor <msebor@redhat.com>
5809 PR tree-optimization/84818
5810 * builtins.c (check_access): Use warning_n.
5812 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5814 PR target/83822
5815 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5816 condition.
5817 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5818 condition.
5820 2018-03-30 Julia Koval <julia.koval@intel.com>
5822 PR target/84413
5823 * x86-tune.def (movx, partial_reg_dependency): Enable for
5824 m_SKYLAKE_AVX512.
5826 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5828 PR inline-asm/84985
5829 * lra-constraints.c (process_alt_operands): Move setting
5830 this_alternative_matches below.
5832 2018-03-29 Martin Liska <mliska@suse.cz>
5834 PR lto/84995.
5835 * doc/invoke.texi: Document how LTO works with debug info.
5836 Describe auto-load support of binutils. Mention 'x86-64'
5837 as valid option value of -march option.
5839 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5841 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5843 PR c/85094
5844 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5845 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5846 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5847 checking.
5849 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5851 PR target/84912
5852 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5853 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5854 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5855 for RS6000_BTM_POWERPC64.
5856 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5857 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5858 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5859 definition.
5860 (DIVDE): Use it.
5861 (DIVDEU): Likewise.
5863 2018-03-28 Carl Love <cel@us.ibm.com>
5865 Revert
5866 2017-09-27 Carl Love <cel@us.ibm.com>
5868 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5869 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5870 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5871 fctiw instruction.
5873 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5875 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5876 instead of __vector bool.
5877 (_mm_max_pu8): Likewise.
5878 (_mm_min_pi16): Likewise.
5880 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5882 PR target/84912
5883 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5884 (DIVWEUO): Likewise.
5885 (DIVDEO): Likewise.
5886 (DIVDEUO): Likewise.
5887 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5888 DIVWEUO and DIVDEUO.
5889 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5890 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5891 (div_extend): Likewise.
5892 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5893 builtin function.
5894 (__builtin_divweuo): Likewise.
5895 (__builtin_divdeo): Likewise.
5896 (__builtin_divdeuo): Likewise.
5898 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5900 PR target/85095
5901 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5902 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5904 PR tree-optimization/82004
5905 * gimple-match-head.c (optimize_pow_to_exp): New function.
5906 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5907 Don't fold to exp if optimize_pow_to_exp is false.
5909 2018-03-28 Martin Liska <mliska@suse.cz>
5911 PR other/84819
5912 * calls.c (initialize_argument_information): Fix trailing space.
5913 * common.opt: Fix typo and provide better explanation for
5914 -fsanitize-coverage option.
5915 * config/i386/i386.opt: Fix typo.
5917 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5918 Martin Liska <mliska@suse.cz>
5920 PR sanitizer/85081
5921 * gimplify.c (asan_poison_variable): Don't do the check for
5922 gimplify_omp_ctxp here.
5923 (gimplify_decl_expr): Do it here.
5924 (gimplify_target_expr): Likewise.
5926 2018-03-28 Martin Liska <mliska@suse.cz>
5928 PR target/84988
5929 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5930 chkp_type_bounds_count if MPX is not enabled.
5932 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5934 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5936 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5938 PR target/84914
5939 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5940 function to create the function decl for complex long double
5941 multiply and divide for -mabi=ieeelongdouble.
5942 (init_float128_ieee): Call it.
5944 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5946 PR target/85044
5947 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5948 -fcf-protection=branch -mibt.
5949 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5951 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5953 PR target/81863
5954 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5955 arm_word_relocations.
5957 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5959 PR target/85056
5960 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5961 extern array declarations.
5963 2018-03-27 Richard Biener <rguenther@suse.de>
5965 PR middle-end/84067
5966 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5967 explicit single_use checks.
5969 2018-03-27 Richard Biener <rguenther@suse.de>
5971 PR tree-optimization/85082
5972 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5973 Valueize the VUSE.
5975 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5977 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5978 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5979 Turn on fasynchronous-unwind-tables and funwind-tables.
5981 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5983 PR target/85073
5984 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5985 (*bmi_blsr_<mode>_ccz): Ditto.
5987 2018-03-26 Tom de Vries <tom@codesourcery.com>
5989 PR tree-optimization/85063
5990 * omp-general.c (offloading_function_p): New function. Factor out
5991 of ...
5992 * omp-offload.c (pass_omp_target_link::gate): ... here.
5993 * omp-general.h (offloading_function_p): Declare.
5994 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5995 with attribute omp declare target for offloading functions.
5997 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5999 PR tree-optimization/84005
6000 * tree-data-ref.h (get_base_for_alignment): Declare.
6001 * tree-data-ref.c (get_base_for_alignment_1): New function.
6002 (get_base_for_alignment): Likewise.
6003 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6004 get_base_for_alignment to find a suitable base object, instead
6005 of always using drb->base_address.
6007 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6009 PR inline-asm/85022
6010 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6011 known size by default.
6013 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6015 PR inline-asm/85030
6016 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6017 and non BLKmode operands.
6019 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6021 PR target/85026
6022 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6023 Clean up attributes.
6025 2018-03-23 Richard Biener <rguenther@suse.de>
6027 PR debug/85020
6028 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6029 we are going to emit early debug for LTO.
6031 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6033 PR inline-asm/85034
6034 * function.c (match_asm_constraints_1): Don't optimize if input
6035 doesn't satisfy general_operand predicate for output's mode.
6037 PR inline-asm/85022
6038 * alias.c (write_dependence_p): Don't require for x_canonicalized
6039 non-VOIDmode if x has VOIDmode.
6041 PR sanitizer/85029
6042 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6043 just don't try to optimize it rather than assert it never happens.
6045 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6047 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6048 macro expansions for definition of ST_INTERNAL_<mode> and
6049 LD_INTERNAL_<mode> builtins.
6050 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6051 Remove prototype.
6052 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6053 function.
6054 (altivec_expand_st_builtin): Likewise.
6055 (altivec_expand_builtin): Remove calls to deleted functions.
6056 (rs6000_address_for_altivec): Delete this function.
6057 * config/rs6000/vector.md: Remove expands for
6058 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6060 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6062 PR target/84826
6063 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6064 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6065 re-computing once computed.
6066 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6067 (arm_init_machine_status): Initialize
6068 machine->static_chain_stack_bytes.
6070 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6072 PR target/84760
6073 * doc/extend.texi: Add four new prototypes for vec_ld.
6074 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6075 definitions for more logical presentation.
6076 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6077 entries for V1TI variants of __builtin_altivec_ld builtin.
6078 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6079 handling of V1TI variant of LVX icode pattern.
6080 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6081 (rs6000_gimple_fold_builtin): Likewise.
6082 (altivec_init_builtins): Add code to define
6083 __builtin_altivec_lvx_v1ti function.
6085 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6087 PR inline-asm/84941
6088 * function.c (match_asm_constraints_1): Don't do the optimization
6089 if input isn't a REG, SUBREG, MEM or constant.
6091 2018-03-22 Tom de Vries <tom@codesourcery.com>
6093 PR tree-optimization/84956
6094 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6095 bb_has_abnormal_pred.
6097 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6099 PR sanitizer/85018
6100 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6101 DECL_INITIAL (decl) to decl at the end.
6102 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6103 adjust the comment.
6105 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6107 * doc/extend.texi (__builtin_tgmath): Document when complex
6108 integer types are treated as _Complex _Float64.
6110 2018-03-21 Tom de Vries <tom@codesourcery.com>
6112 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6114 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6116 PR tree-optimization/84960
6117 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6118 if it is ENTRY block, move them into single succ of ENTRY in that case.
6120 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6122 PR tree-optimization/84811
6123 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6124 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6125 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6126 (known_subrange_p): Update accordingly. Cast each value involved
6127 in the size comparison, rather than casting the result of the
6128 subtraction.
6130 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6132 PR tree-optimization/84982
6133 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6134 by flipping the least significant bit rather than all bits from
6135 bitpos to bitpos + bitsize - 1.
6137 2018-03-21 Nathan Sidwell <nathan@acm.org>
6139 * doc/extend.texi (Deprecated Features): Remove mention of
6140 long-deleted deprecations.
6142 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6144 PR jit/84288
6145 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6146 * configure: Regenerate.
6148 2018-03-21 Tom de Vries <tom@codesourcery.com>
6150 PR tree-optimization/83126
6151 * tree-parloops.c (num_phis): New function.
6152 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6154 2018-03-21 Nathan Sidwell <nathan@acm.org>
6156 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6157 mention anon-struct/union members and trailing attributes.
6159 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6161 PR tree-optimization/84969
6162 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6163 builtin memset partitions if they set different rhs values.
6165 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6167 PR rtl-optimization/84989
6168 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6169 VEC_DUPLICATE with scalar result mode.
6171 2018-03-21 Martin Liska <mliska@suse.cz>
6173 PR ipa/84963
6174 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6175 not intended return statement.
6177 2018-03-21 Martin Liska <mliska@suse.cz>
6179 PR target/84988
6180 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6181 (chkp_find_bound_slots_1): Limit number of iterations.
6183 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6185 PR target/84838
6186 * Minor grammar fixes for x86 options.
6188 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6190 PR debug/84875
6191 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6192 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6194 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6196 PR target/83789
6197 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6198 (altivec_lvx_<mode>_1op): Likewise.
6199 (altivec_stvx_<mode>_2op): Likewise.
6200 (altivec_stvx_<mode>_1op): Likewise.
6201 (altivec_lvx_<VM2:mode>): New define_expand.
6202 (altivec_stvx_<VM2:mode>): Likewise.
6203 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6204 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6205 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6206 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6207 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6208 (rs6000_gen_lvx): Likewise.
6209 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6210 (altivec_expand_stv_builtin): Likewise.
6211 (altivec_expand_builtin): Likewise.
6212 * config/rs6000/vector.md: Likewise.
6214 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6216 PR target/82518
6217 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6218 BYTES_BIG_ENDIAN.
6220 2018-03-20 Richard Biener <rguenther@suse.de>
6222 PR target/84986
6223 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6224 sign-conversions as zero, fall back to standard scalar_stmt
6225 cost for the rest.
6227 2018-03-20 Martin Liska <mliska@suse.cz>
6229 PR ipa/84825
6230 * predict.c (rebuild_frequencies): Handle case when we have
6231 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6233 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6235 PR target/84990
6236 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6237 flag_section_anchors.
6238 * varasm.c (use_blocks_for_decl_p): Remove hack for
6239 dw2_force_const_mem.
6241 PR target/84845
6242 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6243 to ...
6244 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6245 be created, use lowpart_subreg of operands[0] rather than operands[0]
6246 itself.
6247 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6248 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6249 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6250 and n constraint instead of aarch64_shift_imm_di and Usd.
6251 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6252 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6254 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6256 PR target/82989
6257 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6258 to favor GPR over NEON registers.
6259 (<shift>di3_neon): Likewise.
6261 2018-03-20 Tom de Vries <tom@codesourcery.com>
6263 PR target/84952
6264 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6265 (nvptx_process_pars): Emit bar.sync asap and alap.
6267 2018-03-20 Tom de Vries <tom@codesourcery.com>
6269 PR target/84954
6270 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6271 seen_label if seen_label is already set.
6273 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6275 PR target/84945
6276 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6277 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6278 Use 1U instead of 1. Formatting fixes.
6280 PR c/84953
6281 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6282 instead of TREE_TYPE (s1) for the return value.
6284 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6286 PR tree-optimization/84946
6287 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6288 bitsize + bitsize in poly_uint64 rather than poly_int64.
6290 PR sanitizer/78651
6291 * dwarf2asm.c: Include fold-const.c.
6292 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6293 of decl rather than decl itself.
6295 PR rtl-optimization/84643
6296 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6298 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6300 PR sanitizer/78651
6301 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6302 calling assemble_variable.
6304 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6306 PR target/81647
6307 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6308 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6310 2018-03-19 Jim Wilson <jimw@sifive.com>
6312 PR bootstrap/84856
6313 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6314 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6315 (riscv_first_stack_step): Likewise.
6316 (riscv_option_override): Use STACK_BOUNDARY instead of
6317 MIN_STACK_BOUNDARY.
6318 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6319 MIN_STACK_BOUNDARY.
6320 (BIGGEST_ALIGNMENT): Set to 128.
6321 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6322 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6323 STACK_BOUNDARY.
6325 2018-03-19 Richard Biener <rguenther@suse.de>
6327 PR tree-optimization/84933
6328 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6329 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6331 2018-03-19 Richard Biener <rguenther@suse.de>
6333 PR tree-optimization/84859
6334 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6335 (cond_if_else_store_replacement): Perform sinking operation on
6336 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6337 Generalize what a BB with a single eligible store is.
6339 2018-03-19 Richard Biener <rguenther@suse.de>
6341 PR tree-optimization/84929
6342 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6343 chrec_is_positive against non-chrec arg.
6345 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6347 PR target/84711
6348 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6350 2018-03-18 Martin Liska <mliska@suse.cz>
6352 PR rtl-optimization/84635
6353 * regrename.c (build_def_use): Use matches_mode only when
6354 matches >= 0.
6356 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6358 PR tree-optimization/84913
6359 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6360 vectorize chains of COND_EXPRs.
6362 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6364 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6366 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6368 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6370 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6372 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6374 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6375 Kito Cheng <kito.cheng@gmail.com>
6377 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6378 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6379 (nds32_adjust_reg_alloc_order): New function.
6380 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6382 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6384 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6385 nds32_print_operand, nds32_print_operand_address): Use
6386 HOST_WIDE_INT_PRINT_DEC instead.
6388 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6390 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6392 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6394 PR target/84902
6395 * config/i386/i386.c (initial_ix86_tune_features,
6396 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6397 unsigned long long.
6398 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6399 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6400 rather than 1u << ix86_tune. Formatting fix.
6401 (ix86_option_override_internal): Change ix86_arch_mask from
6402 unsigned int to unsigned HOST_WIDE_INT, initialize to
6403 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6404 (ix86_function_specific_restore): Likewise.
6406 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6408 PR target/84899
6409 * postreload.c (reload_combine_recognize_pattern): Perform
6410 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6411 truncate_int_for_mode the result for the destination's mode.
6413 PR c/84909
6414 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6415 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6416 Likewise.
6417 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6418 Likewise.
6420 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6422 PR target/84876
6423 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6424 regno_allocno_class_array and sorted_pseudos.
6425 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6426 insns where regno is used.
6428 2018-03-16 Martin Liska <mliska@suse.cz>
6430 PR ipa/84833
6431 * multiple_target.c (create_dispatcher_calls): Redirect
6432 reference in the symbol table.
6434 2018-03-16 Martin Liska <mliska@suse.cz>
6436 PR ipa/84722
6437 * multiple_target.c (create_dispatcher_calls): Redirect also
6438 an alias.
6440 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6442 PR c++/79937
6443 PR c++/82410
6444 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6445 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6446 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6448 2018-03-16 Julia Koval <julia.koval@intel.com>
6450 * doc/invoke.texi (Skylake Server): Add CLWB.
6451 Cannonlake): Remove CLWB.
6453 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6455 PR tree-optimization/84841
6456 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6457 1 << 3.
6458 (FLOAT_ONE_CONST_TYPE): Define.
6459 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6460 (sort_by_operand_rank): Put entries with higher constant_type last
6461 rather than first to match comments.
6463 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6465 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6466 split predicate.
6468 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6470 PR c++/79085
6471 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6472 check and use address of target always.
6474 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6476 PR target/84574
6477 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6478 (indirect_thunk_bnd_needed): Likewise.
6479 (indirect_thunks_used): Likewise.
6480 (indirect_thunks_bnd_used): Likewise.
6481 (indirect_return_needed): New.
6482 (indirect_return_bnd_needed): Likewise.
6483 (output_indirect_thunk_function): Add a bool argument for
6484 function return.
6485 (output_indirect_thunk_function): Don't generate alias for
6486 function return thunk.
6487 (ix86_code_end): Call output_indirect_thunk_function to generate
6488 function return thunks.
6489 (ix86_output_function_return): Set indirect_return_bnd_needed
6490 and indirect_return_needed instead of indirect_thunk_bnd_needed
6491 and indirect_thunk_needed.
6493 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6495 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6496 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6497 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6499 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6500 Paul Hua <paul.hua.gm@gmail.com>
6502 PR c/84852
6503 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6505 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6507 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6508 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6509 resp. SFmode cases.
6511 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6513 PR target/84711
6514 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6515 instead of GET_MODE_SIZE when comparing Units.
6517 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6519 PR target/68256
6520 * varasm.c (hash_section): Return an unchangeble hash value
6521 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6522 Return !aarch64_can_use_per_function_literal_pools_p ().
6524 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6526 PR target/84860
6527 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6528 rather than address of cmode as last argument to prepare_cmp_insn.
6530 2018-03-15 Julia Koval <julia.koval@intel.com>
6532 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6533 F_AVX512VNNI, F_AVX512BITALG): New.
6535 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6537 PR target/83451
6538 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6539 insn for floating-point loads and stores.
6541 2018-03-14 Carl Love <cel@us.ibm.com>
6543 * config/rs6000/rs6000-c.c: Add macro definitions for
6544 ALTIVEC_BUILTIN_VEC_PERMXOR.
6545 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6546 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6547 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6548 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6549 UNSPEC_VPERMXOR.
6550 * config/doc/extend.texi: Add prototypes for vec_permxor.
6552 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6554 PR c/84852
6555 * diagnostic-show-locus.c (class layout_point): Convert m_line
6556 from int to linenum_type.
6557 (line_span::comparator): Use linenum "compare" function when
6558 comparing line numbers.
6559 (test_line_span): New function.
6560 (layout_range::contains_point): Convert param "row" from int to
6561 linenum_type.
6562 (layout_range::intersects_line_p): Likewise.
6563 (layout::will_show_line_p): Likewise.
6564 (layout::print_source_line): Likewise.
6565 (layout::should_print_annotation_line_p): Likewise.
6566 (layout::print_annotation_line): Likewise.
6567 (layout::print_leading_fixits): Likewise.
6568 (layout::annotation_line_showed_range_p): Likewise.
6569 (struct line_corrections): Likewise for field m_row.
6570 (line_corrections::line_corrections): Likewise for param "row".
6571 (layout::print_trailing_fixits): Likewise.
6572 (layout::get_state_at_point): Likewise.
6573 (layout::get_x_bound_for_row): Likewise.
6574 (layout::print_line): Likewise.
6575 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6576 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6577 * input.c (selftest::test_linenum_comparisons): New function.
6578 (selftest::input_c_tests): Call it.
6579 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6580 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6581 * selftest.h (ASSERT_GT): New macro.
6582 (ASSERT_GT_AT): New macro.
6583 (ASSERT_LT): New macro.
6584 (ASSERT_LT_AT): New macro.
6586 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6588 PR rtl-optimization/84780
6589 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6591 2018-03-14 Martin Liska <mliska@suse.cz>
6593 * tree.c (record_node_allocation_statistics): Use
6594 get_stats_node_kind.
6595 (get_stats_node_kind): New function extracted from
6596 record_node_allocation_statistics.
6597 (free_node): Use get_stats_node_kind.
6599 2018-03-14 Richard Biener <rguenther@suse.de>
6601 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6602 that the value-set of ANTIC_IN doesn't grow.
6604 Revert
6605 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6606 member.
6607 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6608 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6610 2018-03-14 Julia Koval <julia.koval@intel.com>
6612 * config.gcc (icelake-client, icelake-server): New.
6613 (icelake): Remove.
6614 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6615 (initial_ix86_arch_features): Ditto.
6616 (PTA_SKYLAKE): Add SGX.
6617 (PTA_ICELAKE): Remove.
6618 (PTA_ICELAKE_CLIENT): New.
6619 (PTA_ICELAKE_SERVER): New.
6620 (ix86_option_override_internal): Split up icelake on icelake client and
6621 icelake server.
6622 (get_builtin_code_for_version): Ditto.
6623 (fold_builtin_cpu): Ditto.
6624 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6625 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6626 * config/i386/i386.h (processor_type): Ditto.
6627 * doc/invoke.texi: Ditto.
6629 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6631 PR sanitizer/83392
6632 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6633 INTEGER_CST offset, add it together with bitpos / 8 and
6634 sign extend based on POINTER_SIZE.
6636 PR target/84844
6637 Revert
6638 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6640 PR target/78090
6641 * config/i386/constraints.md (Yc): New register constraint.
6642 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6643 Use Yc constraint for alternative 2 of operand 0. Remove
6644 preferred_for_speed attribute.
6646 2018-03-14 Richard Biener <rguenther@suse.de>
6648 PR tree-optimization/84830
6649 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6650 with the old one to avoid oscillations.
6652 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6654 PR target/83712
6655 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6656 pseudos.
6657 (assign_by_spills): Return a flag of reload assignment failure.
6658 Do not process the reload assignment failures. Do not spill other
6659 reload pseudos if they has the same reg class. Update n if
6660 necessary.
6661 (lra_assign): Add a return arg. Set up from the result of
6662 assign_by_spills call.
6663 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6664 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6665 usage_insns if it is not NULL.
6666 (spill_hard_reg_in_range): New function.
6667 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6668 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6669 function prototypes.
6670 (lra_assign): Change prototype.
6671 * lra.c (lra): Add code to deal with fails by splitting hard reg
6672 live ranges.
6674 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6676 * config/riscv/riscv.opt (mrelax): New option.
6677 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6678 "norelax" when riscv_mrelax is disabled.
6679 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6681 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6683 PR target/84743
6684 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6685 reassociation for int modes.
6687 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6689 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6690 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6691 for big-endian.
6692 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6693 * config/aarch64/aarch64-sve.md
6694 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6695 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6696 (*extend<mode><Vwide>2): Rename to...
6697 (aarch64_sve_extend<mode><Vwide>2): ...this.
6698 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6699 renaming the old pattern to...
6700 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6701 unsigned packs.
6702 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6703 define_expand, renaming the old pattern to...
6704 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6705 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6706 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6707 account when deciding which SVE instruction the optab should use.
6708 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6710 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6712 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6713 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6714 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6715 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6716 (tlsdesc_small_<mode>): Turn a define_expand and use
6717 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6718 (tlsdesc_small_advsimd_<mode>): ...this.
6719 (tlsdesc_small_sve_<mode>): New pattern.
6721 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6723 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6724 (UNSPEC_UMUL_HIGHPART): New constants.
6725 (MUL_HIGHPART): New int iteraor.
6726 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6727 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6728 define_expand.
6729 (*<su>mul<mode>3_highpart): New define_insn.
6731 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6733 PR lto/84805
6734 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6735 incomplete types.
6737 2018-03-13 Martin Liska <mliska@suse.cz>
6739 PR ipa/84658.
6740 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6741 vector.
6742 (sem_item_optimizer::~sem_item_optimizer): Release it.
6743 (sem_item_optimizer::merge_classes): Register variable aliases.
6744 (sem_item_optimizer::fixup_pt_set): New function.
6745 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6746 * ipa-icf.h: Declare new variables and functions.
6748 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6750 PR middle-end/84834
6751 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6752 integer_pow2p@2 and test integer_pow2p in condition.
6753 (A < 0 ? C : 0): Similarly for @1.
6755 PR middle-end/84831
6756 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6757 characters starting at p contain '\0' character, don't look beyond
6758 that.
6760 PR target/84827
6761 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6762 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6764 PR target/84828
6765 * reg-stack.c (change_stack): Change update_end var from int to
6766 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6767 also call set_block_for_insn on the newly added insns and rescan.
6769 PR target/84786
6770 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6771 on the last operand.
6773 PR c++/84704
6774 * tree.c (stabilize_reference_1): Return save_expr (e) for
6775 STATEMENT_LIST even if it doesn't have side-effects.
6777 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6779 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6781 2018-03-12 Renlin Li <renlin.li@arm.com>
6783 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6784 aarch64_output_scalar_simd_mov_immediate.
6786 2018-03-12 Martin Sebor <msebor@redhat.com>
6788 PR tree-optimization/83456
6789 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6790 for perfectly overlapping calls to memcpy.
6791 (gimple_fold_builtin_memory_chk): Same.
6792 (gimple_fold_builtin_strcpy): Handle no-warning.
6793 (gimple_fold_builtin_stxcpy_chk): Same.
6794 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6796 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6798 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6799 parameter. Use it for SFmode.
6800 (rs6000_function_arg_advance_1): Adjust.
6801 (rs6000_function_arg): Adjust.
6802 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6804 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6806 PR rtl-optimization/84169
6807 PR rtl-optimization/84780
6808 * combine.c (can_combine_p): Check for a 2-insn combination whether
6809 the destination register is used between the two insns, too.
6811 2018-03-12 Richard Biener <rguenther@suse.de>
6813 PR tree-optimization/84803
6814 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6815 for refs DR analysis didn't process.
6817 2018-03-12 Richard Biener <rguenther@suse.de>
6819 PR tree-optimization/84777
6820 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6821 force-vectorize loops ignore whether we are optimizing for size.
6823 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6825 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6826 (TARGET_MD_ASM_ADJUST): Define.
6828 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6829 Kito Cheng <kito.cheng@gmail.com>
6830 Chung-Ju Wu <jasonwucj@gmail.com>
6832 * config/nds32/nds32.c (nds32_compute_stack_frame,
6833 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6834 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6835 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6836 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6837 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6838 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6839 * config/nds32/nds32.md (prologue, epilogue): Use macro
6840 NDS32_V3PUSH_AVAILABLE_P to do checking.
6842 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6844 PR debug/58150
6845 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6846 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6847 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6848 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6849 addition of most attributes on !orig_type_die or the attribute not
6850 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6852 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6853 Chung-Ju Wu <jasonwucj@gmail.com>
6855 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6856 __NDS32_VH__ macro.
6857 * config/nds32/nds32.opt (mvh): New option.
6859 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6860 Chung-Ju Wu <jasonwucj@gmail.com>
6862 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6863 function.
6864 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6865 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6866 definition.
6868 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6869 Chung-Ju Wu <jasonwucj@gmail.com>
6871 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6872 function.
6873 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6874 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6876 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6877 Kito Cheng <kito.cheng@gmail.com>
6878 Chung-Ju Wu <jasonwucj@gmail.com>
6880 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6881 UNSPEC_FFMISM and UNSPEC_FLMISM.
6882 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6883 for ffb, ffmism and flmism.
6884 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6885 (unspec_ffmism): Ditto.
6886 (unspec_flmism): Ditto.
6887 (nds32_expand_builtin_impl): Check if string extension is available.
6888 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6889 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6891 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6893 Reverting patch:
6894 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6896 PR target/83712
6897 * lra-assigns.c (assign_by_spills): Return a flag of reload
6898 assignment failure. Do not process the reload assignment
6899 failures. Do not spill other reload pseudos if they has the same
6900 reg class.
6901 (lra_assign): Add a return arg. Set up from the result of
6902 assign_by_spills call.
6903 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6904 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6905 usage_insns if it is not NULL.
6906 (spill_hard_reg_in_range): New function.
6907 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6908 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6909 function prototypes.
6910 (lra_assign): Change prototype.
6911 * lra.c (lra): Add code to deal with fails by splitting hard reg
6912 live ranges.
6914 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6916 PR target/84807
6917 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6919 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6921 PR debug/84620
6922 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6923 (dw_val_node): Add val_symbolic_view.
6924 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6925 (symview_upper_bound): New.
6926 (new_line_info_table): Initialize symviews_since_reset.
6927 (dwarf2out_source_line): Count symviews_since_reset and set
6928 symview_upper_bound.
6929 (dw_val_equal_p): Handle symview.
6930 (add_AT_symview): New.
6931 (print_dw_val): Handle symview.
6932 (attr_checksum, attr_checksum_ordered): Likewise.
6933 (same_dw_val_p, size_of_die): Likewise.
6934 (value_format, output_die): Likewise.
6935 (add_high_low_attributes): Use add_AT_symview for entry_view.
6936 (dwarf2out_finish): Reset symview_upper_bound, clear
6937 zero_view_p.
6939 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6941 PR target/83969
6942 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6943 Add strict argument and use it.
6944 (rs6000_split_multireg_move): Update for new strict argument.
6945 (mem_operand_gpr): Disallow all non-offsettable addresses.
6946 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6948 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6950 PR target/84772
6951 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6952 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6953 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6955 PR c++/84767
6956 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6957 decl, use remap_type if we want to use the type.
6959 2018-03-09 Martin Sebor <msebor@redhat.com>
6961 PR tree-optimization/84526
6962 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6963 Remove dead code.
6964 (builtin_access::generic_overlap): Be prepared to handle non-array
6965 base objects.
6967 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6969 PR rtl-optimization/84682
6970 * lra-constraints.c (process_address_1): Check is_address flag
6971 for address constraints.
6972 (process_alt_operands): Likewise.
6973 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6974 preprocess_constraints.
6975 * recog.h (preprocess_constraints): Add oploc parameter.
6976 Adjust callers.
6977 * recog.c (preprocess_constraints): Test address_operand for
6978 CT_ADDRESS constraints.
6980 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6982 PR target/83712
6983 * lra-assigns.c (assign_by_spills): Return a flag of reload
6984 assignment failure. Do not process the reload assignment
6985 failures. Do not spill other reload pseudos if they has the same
6986 reg class.
6987 (lra_assign): Add a return arg. Set up from the result of
6988 assign_by_spills call.
6989 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6990 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6991 usage_insns if it is not NULL.
6992 (spill_hard_reg_in_range): New function.
6993 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6994 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6995 function prototypes.
6996 (lra_assign): Change prototype.
6997 * lra.c (lra): Add code to deal with fails by splitting hard reg
6998 live ranges.
7000 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7002 PR target/83193
7003 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7004 Accept complain bool parameter. Only emit errors if it is true.
7005 (arm_parse_cpu_option_name): Likewise.
7006 (arm_target_thumb_only): Adjust callers of the above.
7007 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7008 prototype to take a default true bool parameter.
7009 (arm_parse_arch_option_name): Likewise.
7011 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7012 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7014 PR jit/64089
7015 PR jit/84288
7016 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7017 * configure: Regenerate.
7018 * configure.ac ("linker --version-script option"): New.
7019 ("linker soname option"): New.
7021 2018-03-09 Richard Biener <rguenther@suse.de>
7023 PR tree-optimization/84775
7024 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7025 immediate uses of predicate stmts and mark them modified.
7027 Revert
7028 PR tree-optimization/84178
7029 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7030 to caller.
7031 (version_loop_for_if_conversion): Delay update_ssa call.
7032 (tree_if_conversion): Delay update_ssa until after predicate
7033 insertion.
7035 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7037 PR target/84763
7038 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7039 when the function accesses prior frames.
7041 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7043 PR debug/84456
7044 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7045 gen_llsym, otherwise call maybe_gen_llsym.
7047 PR inline-asm/84742
7048 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7049 has ',' character inside of it.
7051 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7053 PR target/84748
7054 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7055 as clobbering CC_REGNUM.
7057 2018-03-08 Richard Biener <rguenther@suse.de>
7059 PR middle-end/84552
7060 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7061 (follow_copies_to_constant): Do not follow SSA names registered
7062 for update.
7064 2018-03-08 Richard Biener <rguenther@suse.de>
7066 PR tree-optimization/84178
7067 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7068 to caller.
7069 (version_loop_for_if_conversion): Delay update_ssa call.
7070 (tree_if_conversion): Delay update_ssa until after predicate
7071 insertion.
7073 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7075 PR tree-optimization/84178
7076 * tree-if-conv.c (release_bb_predicate): Remove the
7077 the assertion that the stmts have NULL use_ops.
7078 Discard the statements, asserting that they haven't
7079 yet been added to a BB.
7081 2018-03-08 Richard Biener <rguenther@suse.de>
7083 PR tree-optimization/84746
7084 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7085 (phi_translate): Pass in destination ANTIC_OUT set.
7086 (phi_translate_1): Likewise. For a simplified result lookup
7087 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7088 (phi_translate_set): Adjust.
7089 (do_pre_regular_insertion): Likewise.
7090 (do_pre_partial_partial_insertion): Likewise.
7092 2018-03-08 Martin Liska <mliska@suse.cz>
7094 PR gcov-profile/84735
7095 * doc/gcov.texi: Document usage of profile files.
7096 * gcov-io.h: Document changes in the format.
7098 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7100 PR debug/84404
7101 PR debug/84408
7102 * dwarf2out.c (struct dw_line_info_table): Update comments for
7103 view == -1.
7104 (FORCE_RESET_NEXT_VIEW): New.
7105 (FORCE_RESETTING_VIEW_P): New.
7106 (RESETTING_VIEW_P): Check for -1 too.
7107 (ZERO_VIEW_P): Likewise.
7108 (new_line_info_table): Force-reset next view.
7109 (dwarf2out_begin_function): Likewise.
7110 (dwarf2out_source_line): Simplify zero_view_p initialization.
7111 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7112 view directly. Omit view when omitting .loc at line 0.
7114 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7116 PR tree-optimization/84740
7117 * tree-switch-conversion.c (process_switch): Call build_constructors
7118 only if info.phi_count is non-zero.
7120 PR tree-optimization/84739
7121 * tree-tailcall.c (find_tail_calls): Check call arguments against
7122 DECL_ARGUMENTS (current_function_decl) rather than
7123 DECL_ARGUMENTS (func) when checking for tail recursion.
7125 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7127 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7128 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7129 Volker Reichelt's entry and add entries for people that perform
7130 GCC fuzzy testing and report numerous bugs.
7132 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7134 PR target/82411
7135 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7136 readonly data in sdata, if that is disabled.
7137 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7138 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7139 -mreadonly-in-sdata option.
7141 2018-03-07 Martin Sebor <msebor@redhat.com>
7143 PR tree-optimization/84468
7144 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7145 basic block when looking for nul assignment.
7147 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7149 PR target/84277
7150 * except.h (output_function_exception_table): Adjust prototype.
7151 * except.c (output_function_exception_table): Remove FNNAME parameter
7152 and add SECTION parameter. Ouput one part of the table at a time.
7153 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7154 the first part of the exception table and emit unwind directives.
7155 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7156 (i386_pe_seh_cold_init): Likewise.
7157 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7158 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7159 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7160 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7161 * config/i386/winnt.c: Include except.h.
7162 (struct seh_frame_state): Add reg_offset, after_prologue and
7163 in_cold_section fields.
7164 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7165 (i386_pe_seh_cold_init): New function.
7166 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7167 to seh->in_cold_section.
7168 (seh_emit_push): Record the offset of the push.
7169 (seh_emit_save): Record the offet of the save.
7170 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7171 Test seh->after_prologue to disregard the epilogue.
7172 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7173 (i386_pe_end_cold_function): New function.
7175 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7177 PR fortran/84565
7178 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7179 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7181 PR c++/84704
7182 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7183 on tmp_var.
7184 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7185 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7187 PR middle-end/84723
7188 * multiple_target.c: Include tree-inline.h and intl.h.
7189 (expand_target_clones): Diagnose and fail if node->definition and
7190 !tree_versionable_function_p (node->decl).
7192 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7194 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7195 sprint_ul.
7196 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7197 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7198 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7200 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7202 PR target/84710
7203 * combine.c (try_combine): Use reg_or_subregno instead of handling
7204 just paradoxical SUBREGs and REGs.
7206 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7208 * config/arc/arc.c (arc_finalize_pic): Remove function.
7209 (arc_must_save_register): We use single base PIC register, remove
7210 checks to save/restore the PIC register.
7211 (arc_expand_prologue): Likewise.
7212 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7213 Remove.
7214 (arc_verify_short): Likewise.
7215 (arc_attr_type): Likewise.
7216 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7217 (walk_stores): Likewise.
7218 (arc_address_cost): Make it static.
7219 (arc_verify_short): Likewise.
7220 (branch_dest): Likewise.
7221 (arc_attr_type): Likewise.
7222 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7223 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7224 (arc_final_prescan_insn): Remove inserting the nops due to
7225 hardware hazards. It is done in reorg step.
7226 (insn_length_variant_t): Remove.
7227 (insn_length_parameters_t): Likewise.
7228 (arc_insn_length_parameters): Likewise.
7229 (arc_get_insn_variants): Likewise.
7230 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7232 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7234 PR inline-asm/84683
7235 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7236 assertion failure.
7238 PR tree-optimization/84687
7239 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7240 on new_node->decl.
7241 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7243 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7245 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7246 Rename to ppc_speculation_barrier.
7247 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7248 __builtin_ppc_speculation_barrier.
7250 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7252 PR target/84700
7253 * combine.c (combine_simplify_rtx): Don't try to simplify if
7254 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7255 are equal to x.
7257 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7259 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7260 to 32 bytes when compiling for POWER9.
7262 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7264 PR target/84564
7265 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7266 regparm >= 3 with no arg reg available also for calls with
7267 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7269 PR target/84524
7270 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7271 orig,vex.
7272 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7274 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7276 PR target/84264
7277 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7279 2018-03-05 Richard Biener <rguenther@suse.de>
7281 PR tree-optimization/84486
7282 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7283 When inserting a __builtin_assume_aligned call set the LHS
7284 SSA name alignment info accordingly.
7286 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7288 PR tree-optimization/84114
7289 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7290 Avoid reassociation of FLOAT_MODE addition.
7292 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7294 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7295 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7296 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7297 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7298 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7299 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7300 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7301 and -mwbnoinvd.
7302 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7303 __builtin_ia32_wbinvd): New builtins.
7304 (SPECIAL_ARGS2): New.
7305 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7306 (SPECIAL_ARGS2): New.
7307 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7308 (ix86_valid_target_attribute_inner_p): Ditto.
7309 (ix86_init_mmx_sse_builtins): Add special_args2.
7310 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7311 TARGET_WBNOINVD_P): New.
7312 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7313 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7314 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7315 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7316 * config/i386/pconfigintrin.h: New file.
7317 * config/i386/wbnoinvdintrin.h: Ditto.
7318 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7319 wbnoinvdintrin.h.
7320 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7322 2018-03-05 Richard Biener <rguenther@suse.de>
7324 PR tree-optimization/84670
7325 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7326 member.
7327 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7328 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7329 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7330 doesn't grow if all successors (recursively) were visited at least
7331 once.
7333 2018-03-05 Richard Biener <rguenther@suse.de>
7335 PR tree-optimization/84650
7336 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7337 if executed in the loop pipeline.
7339 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7341 * doc/configfiles.texi (Configuration Files): Move info about
7342 conditionalizing $target-protos.h to...
7343 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7344 differs from $target-protos.h.
7346 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7347 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7350 * config/nds32/nds32-multiple.md (setmemsi): Define.
7351 * config/nds32/nds32-memory-manipulation.c
7352 (nds32_gen_dup_4_byte_to_word_value): New.
7353 (emit_setmem_word_loop): New.
7354 (emit_setmem_byte_loop): New.
7355 (nds32_expand_setmem_loop): New.
7356 (nds32_expand_setmem_loop_v3m): New.
7357 (nds32_expand_setmem_unroll): New.
7358 (nds32_expand_setmem): New.
7360 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7361 Chung-Ju Wu <jasonwucj@gmail.com>
7363 * config/nds32/nds32-memory-manipulation.c
7364 (nds32_emit_load_store): New.
7365 (nds32_emit_post_inc_load_store): New.
7366 (nds32_emit_mem_move): New.
7367 (nds32_emit_mem_move_block): New.
7368 (nds32_expand_movmemsi_loop_unknown_size): New.
7369 (nds32_expand_movmemsi_loop_known_size): New.
7370 (nds32_expand_movmemsi_loop): New.
7371 (nds32_expand_movmemsi_unroll): New.
7372 (nds32_expand_movmemqi): Rename ...
7373 (nds32_expand_movmemsi): ... to this.
7374 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7375 (movmemsi): ... to this.
7376 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7377 (nds32_expand_movmemsi): ... to this.
7379 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7380 Monk Chiang <sh.chiang04@gmail.com>
7381 Chung-Ju Wu <jasonwucj@gmail.com>
7383 * config/nds32/nds32-protos.h
7384 (nds32_expand_load_multiple): New arguments.
7385 (nds32_expand_store_multiple): Ditto.
7386 (nds32_valid_multiple_load_store): Rename ...
7387 (nds32_valid_multiple_load_store_p): ... to this.
7388 * config/nds32/nds32-memory-manipulation.c
7389 (nds32_expand_load_multiple): Refine implementation.
7390 (nds32_expand_store_multiple): Ditto.
7391 * config/nds32/nds32-multiple.md
7392 (load_multiple): Update nds32_expand_load_multiple interface.
7393 (store_multiple): Update nds32_expand_store_multiple interface.
7394 * config/nds32/nds32-predicates.c
7395 (nds32_valid_multiple_load_store): Rename ...
7396 (nds32_valid_multiple_load_store_p): ... to this and refine
7397 implementation.
7398 * config/nds32/predicates.md
7399 (nds32_load_multiple_and_update_address_operation): New predicate.
7400 (nds32_store_multiple_and_update_address_operation): New predicate.
7402 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7403 Chung-Ju Wu <jasonwucj@gmail.com>
7405 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7406 (combo): New attribute.
7407 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7409 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7411 * config/nds32/nds32.opt: Change -mcmodel= default value.
7413 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7414 Monk Chiang <sh.chiang04@gmail.com>
7415 Chung-Ju Wu <jasonwucj@gmail.com>
7417 * config/nds32/constants.md (unspec_element): New enum.
7418 * config/nds32/constraints.md (Umw): New constraint.
7419 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7420 * config/nds32/nds32-intrinsic.md: Likewise.
7421 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7422 (nds32_valid_smw_lwm_base_p): New.
7423 (nds32_output_smw_single_word): New.
7424 (nds32_output_lmw_single_word): New.
7425 (nds32_expand_unaligned_load): New.
7426 (nds32_expand_unaligned_store): New.
7427 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7428 (nds32_output_smw_single_word): Declare.
7429 (nds32_output_lmw_single_word): Declare.
7430 (nds32_expand_unaligned_load): Declare.
7431 (nds32_expand_unaligned_store): Declare.
7432 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7433 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7434 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7435 NDS32_BUILTIN_UASTORE_DW.
7436 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7437 predicate.
7439 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7440 Kito Cheng <kito.cheng@gmail.com>
7441 Chung-Ju Wu <jasonwucj@gmail.com>
7443 * config/nds32/nds32-intrinsic.c
7444 (nds32_expand_builtin_null_ftype_reg): Delete.
7445 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7446 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7447 (nds32_read_argument): New.
7448 (nds32_legitimize_target): Ditto.
7449 (nds32_legitimize_argument): Ditto.
7450 (nds32_check_constant_argument): Ditto.
7451 (nds32_expand_unop_builtin): Ditto.
7452 (nds32_expand_unopimm_builtin): Ditto.
7453 (nds32_expand_binop_builtin): Ditto.
7454 (nds32_builtin_decl_impl): Ditto.
7455 (builtin_description): Ditto.
7456 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7457 (nds32_init_builtins_impl): Ditto.
7458 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7459 (nds32_builtin_decl): New.
7460 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7461 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7463 2018-03-02 Jeff Law <law@redhat.com>
7465 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7466 (redundant_insn, fill_simple_delay_slots): Likewise.
7467 (fill_slots_from_thread): Likewise.
7468 * resource.c (mark_referenced_resources): Likewise.
7469 (mark_set_resources, find_dead_or_set_registers): Likewise.
7471 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7473 * substring-locations.h (format_warning_va): Formatting fix for
7474 ATTRIBUTE_GCC_DIAG.
7475 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7476 argument.
7477 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7478 * substring-locations.c: Include intl.h.
7479 (format_warning_va): Turned into small wrapper around
7480 format_warning_n_va, renamed to ...
7481 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7482 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7483 use ngettext.
7484 (format_warning_at_substring_n): New function.
7485 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7486 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7487 format_warning_at_substring with just a shorter name instead of
7488 const function pointer.
7489 (fmtwarn_n): New function.
7490 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7491 appropriate, get rid of all the fmtstr temporaries, move conditionals
7492 with G_() wrapped string literals directly into fmtwarn arguments,
7493 cast dir.len to (int), formatting fixes.
7495 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7497 * doc/invoke.texi: Remove "Cilk Plus" references.
7499 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7500 Richard Biener <rguenther@suse.de>
7502 PR ipa/84628
7503 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7504 for error or warning attributes if CALL_FROM_THUNK_P is set.
7505 Formatting fixes.
7507 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7509 PR target/56540
7510 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7511 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7513 PR target/56540
7514 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7515 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7517 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7518 instead of -1U in last predictors element's probability member.
7520 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7522 PR ipa/83983
7523 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7524 arguments if they are comparable.
7526 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7528 PR tree-optimization/84634
7529 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7530 masks and masked_loop_p with a single loop_masks, making sure it's
7531 null for bb vectorization.
7533 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7535 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7536 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7537 loop->force_vectorize to check whether there is no alias.
7539 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7541 PR target/84614
7542 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7543 prototypes.
7544 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7545 comments.
7546 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7547 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7548 instead of a loop around prev_real_insn.
7549 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7550 prev_real_insn.
7552 PR inline-asm/84625
7553 * config/i386/i386.c (ix86_print_operand): Use conditional
7554 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7555 zero vector.
7557 2018-03-02 Richard Biener <rguenther@suse.de>
7559 PR tree-optimization/84427
7560 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7561 (bitmap_set_subtract_values): Rewrite to handle multiple
7562 exprs per value.
7563 (clean): Likewise.
7564 (prune_clobbered_mems): Likewise.
7565 (phi_translate): Take edge instead of pred/phiblock.
7566 (phi_translate_1): Likewise.
7567 (phi_translate_set): Likewise. Insert all translated
7568 exprs for a value into the set, keeping possibly multiple
7569 expressions per value.
7570 (compute_antic_aux): Adjust for phi_translate changes.
7571 When intersecting union the expressions and prune those
7572 not in the final value set, keeping possibly multiple
7573 expressions per value. Do not use value-insertion
7574 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7575 all expressions. Add verification that the value-sets
7576 only shrink during iteration.
7577 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7578 (do_pre_regular_insertion): Likewise.
7579 (do_pre_partial_partial_insertion): Likewise.
7581 2018-03-02 Richard Biener <rguenther@suse.de>
7583 PR target/82005
7584 * config/darwin.c (saved_debug_info_level): New static global.
7585 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7586 (darwin_asm_lto_end): Restore debug info generation settings.
7588 2018-03-01 Martin Liska <mliska@suse.cz>
7590 PR sanitizer/82484
7591 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7592 volatile arguments.
7594 2018-03-01 Richard Biener <rguenther@suse.de>
7596 PR debug/84645
7597 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7598 type annotation with LTO when debug was disabled at compile-time.
7600 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7602 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7603 XINT with INTVAL.
7604 (mips_final_postscan_insn): Likewise.
7606 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7608 PR rtl-optimization/84528
7609 * alias.c (init_alias_target): Add commentary.
7610 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7611 a unique base value if the frame pointer is not eliminated
7612 to the stack pointer.
7614 2018-03-01 Tom de Vries <tom@codesourcery.com>
7616 PR rtl-optimization/83327
7617 * lra-int.h (hard_regs_spilled_into): Declare.
7618 * lra.c (hard_regs_spilled_into): Define.
7619 (init_reg_info): Init hard_regs_spilled_into.
7620 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7621 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7622 (process_bb_lives): Handle hard_regs_spilled_into.
7623 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7624 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7626 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7628 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7629 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7630 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7631 * config/rs6000/aix72.h: New file.
7633 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7635 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7636 instead of warning_at with conditional singular and plural messages
7637 where possible.
7639 PR target/52991
7640 * stor-layout.c (update_alignment_for_field): For
7641 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7642 && !DECL_PACKED (field), do the alignment update, just use
7643 only desired_align instead of MAX (type_align, desired_align)
7644 as the alignment.
7645 (place_field): Don't do known_align < desired_align handling
7646 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7647 is non-NULL, instead do it after rli->prev_field handling and
7648 only if not within a bitfield word. For DECL_PACKED (field)
7649 use type_align of BITS_PER_UNIT.
7651 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7653 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7654 superfluous parentheses and trailing spaces.
7656 2018-02-28 Richard Biener <rguenther@suse.de>
7658 PR tree-optimization/84584
7659 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7660 SCoPs with fake exit edge.
7662 2018-02-28 Martin Liska <mliska@suse.cz>
7664 PR testsuite/84597
7665 * timevar.c (timer::print): Fix format to properly print 100%
7666 values.
7668 2018-02-28 Richard Biener <rguenther@suse.de>
7670 PR middle-end/84607
7671 * genmatch.c (capture_info::walk_match): Do not mark
7672 captured expressions without operands as expr_p given
7673 they act more like predicates and should be subject to
7674 "lost tail" side-effect preserving.
7676 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7678 PR rtl-optimization/81611
7679 * auto-inc-dec.c (attempt_change): Move dead note from
7680 mem_insn if it's the next use of regno
7681 (find_address): Take address use of reg holding
7682 non-incremented value. Add parm to limit search to the named
7683 reg only.
7684 (merge_in_block): Attempt to use a mem insn that is the next
7685 use of the original regno.
7687 2018-02-27 Martin Sebor <msebor@redhat.com>
7689 PR c++/83871
7690 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7691 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7693 2018-02-27 Martin Sebor <msebor@redhat.com>
7695 PR translation/84207
7696 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7697 n argument to unsigned HOST_WIDE_INT.
7698 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7699 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7700 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7701 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7703 2018-02-27 Richard Biener <rguenther@suse.de>
7705 PR tree-optimization/84512
7706 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7707 Do not use the estimate returned from record_stmt_cost for
7708 the scalar iteration cost but sum properly using add_stmt_cost.
7710 2018-02-27 Richard Biener <rguenther@suse.de>
7712 PR tree-optimization/84466
7713 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7714 Adjust last change to less strictly validate use operands.
7716 2018-02-27 Martin Liska <mliska@suse.cz>
7718 PR gcov-profile/84548
7719 * gcov.c (process_file): Allow partial overlap and consider it
7720 also as group functions.
7721 (output_lines): Properly calculate range of lines for a group.
7723 2018-02-27 Martin Liska <mliska@suse.cz>
7725 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7726 'ggc' suffixes. Change first column width.
7727 (timer::print): Fix formatting of the column.
7729 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7731 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7732 preserve inline entry blocks for the sake of debug inline
7733 entry point markers alone.
7734 (remove_unused_locals): Suggest in comments a better place to
7735 force the preservation of inline entry blocks that are
7736 otherwise unused, but do not preserve them.
7738 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7740 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7742 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7744 PR target/84039
7745 * config/i386/constraints.md (Bs): Replace
7746 ix86_indirect_branch_register with
7747 TARGET_INDIRECT_BRANCH_REGISTER.
7748 (Bw): Likewise.
7749 * config/i386/i386.md (indirect_jump): Likewise.
7750 (tablejump): Likewise.
7751 (*sibcall_memory): Likewise.
7752 (*sibcall_value_memory): Likewise.
7753 Peepholes of indirect call and jump via memory: Likewise.
7754 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7755 (*sibcall_value_GOT_32): Likewise.
7756 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7757 (GOT_memory_operand): Likewise.
7758 (call_insn_operand): Likewise.
7759 (sibcall_insn_operand): Likewise.
7760 (GOT32_symbol_operand): Likewise.
7761 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7763 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7765 PR rtl-optimization/83496
7766 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7767 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7768 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7769 redundant insn, if any.
7770 (relax_delay_slots): Likewise.
7771 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7773 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7775 PR tree-optimization/83965
7776 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7777 that grouped statements are part of a reduction chain. Return
7778 true if the statement is not marked as a reduction itself but
7779 is part of a group.
7780 (vect_recog_dot_prod_pattern): Don't check whether the statement
7781 is part of a group here.
7782 (vect_recog_sad_pattern): Likewise.
7783 (vect_recog_widen_sum_pattern): Likewise.
7785 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7787 PR debug/84545
7788 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7790 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7792 PR target/84530
7793 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7794 the bool argument.
7795 (ix86_output_indirect_function_return): New prototype.
7796 (ix86_split_simple_return_pop_internal): Likewise.
7797 * config/i386/i386.c (indirect_return_via_cx): New.
7798 (indirect_return_via_cx_bnd): Likewise.
7799 (indirect_thunk_name): Handle return va CX_REG.
7800 (output_indirect_thunk_function): Create alias for
7801 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7802 (ix86_output_indirect_jmp): Remove the bool argument.
7803 (ix86_output_indirect_function_return): New function.
7804 (ix86_split_simple_return_pop_internal): Likewise.
7805 * config/i386/i386.md (*indirect_jump): Don't pass false
7806 to ix86_output_indirect_jmp.
7807 (*tablejump_1): Likewise.
7808 (simple_return_pop_internal): Change it to define_insn_and_split.
7809 Call ix86_split_simple_return_pop_internal to split it for
7810 -mfunction-return=.
7811 (simple_return_indirect_internal): Call
7812 ix86_output_indirect_function_return instead of
7813 ix86_output_indirect_jmp.
7815 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7817 PR bootstrap/84405
7818 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7819 memset and value initialization afterwards.
7821 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7823 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7825 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7827 PR target/84521
7828 * common/config/aarch64/aarch64-common.c
7829 (aarch_option_optimization_table[]): Switch
7830 off fomit-frame-pointer
7832 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7833 Chung-Ju Wu <jasonwucj@gmail.com>
7835 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7836 volatile memory.
7837 (store_multiple): Ditto.
7839 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7841 * config.gcc: Add --with-cpu support for nds32 target.
7842 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7843 * config/nds32/nds32.opt: Add -mcpu= option.
7845 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7847 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7848 isel=yes): Warn for these deprecated options.
7850 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7852 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7853 ISA_2_5_MASKS_EMBEDDED.
7855 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7857 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7858 p->max as pointers rather than using iterative_hash_expr.
7860 2018-02-23 Carl Love <cel@us.ibm.com>
7862 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7863 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7864 BU_P8V_OVERLOAD_2.
7865 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7866 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7867 P8V_BUILTIN_VEC_VUNSIGNED2.
7869 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7871 PR target/81572
7872 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7873 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7874 LRA_UNKNOWN_ALT.
7875 * lra-constraints.c (curr_insn_transform): Set up
7876 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7877 LRA_UNKNOWN_ALT.
7878 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7879 * lra-eliminations.c (spill_pseudos): Ditto.
7880 (process_insn_for_elimination): Ditto.
7881 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7882 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7883 LRA_NON_CLOBBERED_ALT.
7885 2018-02-22 Martin Sebor <msebor@redhat.com>
7887 PR tree-optimization/84480
7888 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7889 to maybe_diag_stxncpy_trunc. Call it.
7890 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7891 from gimple_fold_builtin_strcpy. Print inlining stack.
7892 (handle_builtin_stxncpy): Print inlining stack.
7893 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7895 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7897 PR target/84176
7898 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7899 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7900 and -fcheck-pointer-bounds are used together.
7901 (indirect_thunk_prefix): New enum.
7902 (indirect_thunk_need_prefix): New function.
7903 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7904 "_nt" instead of "_bnd" for NOTRACK prefix.
7905 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7906 (output_indirect_thunk_function): Likewise.
7907 (): Likewise.
7908 (ix86_code_end): Update output_indirect_thunk_function calls.
7909 (ix86_output_indirect_branch_via_reg): Replace
7910 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7911 (ix86_output_indirect_branch_via_push): Likewise.
7912 (ix86_output_function_return): Likewise.
7913 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7914 incompatible with -fcf-protection=branch and
7915 -fcheck-pointer-bounds.
7917 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7919 PR target/83335
7920 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7921 Change gcc_assert call to output_operand_lossage.
7923 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7925 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7927 2018-02-22 DJ Delorie <dj@redhat.com>
7928 Sebastian Perta <sebastian.perta@renesas.com>
7929 Oleg Endo <olegendo@gcc.gnu.org>
7931 * config/rx/rx.c (rx_rtx_costs): New function.
7932 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7934 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7936 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7938 2018-02-22 Martin Liska <mliska@suse.cz>
7940 PR driver/83193
7941 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7942 Add "native" as a possible value.
7944 2018-02-22 Martin Liska <mliska@suse.cz>
7946 PR driver/83193
7947 * config/i386/i386.c (ix86_option_override_internal):
7948 Add "native" as a possible value for -march and -mtune.
7950 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7952 PR target/84502
7953 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7954 to all type variants.
7956 PR tree-optimization/84503
7957 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7958 width as info->bitpos + info->bitsize - start.
7959 (merged_store_group::merge_overlapping): Simplify width computation.
7960 (check_no_overlap): New function.
7961 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7962 start + width and last_order of the group, fail if check_no_overlap
7963 fails.
7964 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7965 to group if check_no_overlap fails.
7967 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7969 * config/rs6000/altivec.md: Delete contraint arguments to
7970 define_expand, define_split, and define_peephole2, and in
7971 define_insn_and_split if always unused.
7972 * config/rs6000/darwin.md: Ditto.
7973 * config/rs6000/dfp.md: Ditto.
7974 * config/rs6000/rs6000.md: Ditto.
7975 * config/rs6000/sync.md: Ditto.
7976 * config/rs6000/vector.md: Ditto.
7977 * config/rs6000/vsx.md: Ditto.
7979 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7981 * config/rs6000/altivec.md: Write output control strings as braced
7982 blocks instead of double-quoted strings.
7983 * config/rs6000/darwin.md: Ditto.
7984 * config/rs6000/rs6000.md: Ditto.
7985 * config/rs6000/vector.md: Ditto.
7986 * config/rs6000/vsx.md: Ditto.
7988 2018-02-21 Jason Merrill <jason@redhat.com>
7990 PR c++/84314 - ICE with templates and fastcall attribute.
7991 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7993 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7995 * ipa-cp.c (determine_versionability): Fix comment typos.
7997 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7999 PR c/84229
8000 * ipa-cp.c (determine_versionability): Do not version functions caling
8001 va_arg_pack.
8003 2018-02-21 Martin Liska <mliska@suse.cz>
8005 PR driver/83193
8006 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8007 Add "native" as a possible value.
8008 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8009 the macro when native cpu detection is available.
8011 2018-02-21 Martin Liska <mliska@suse.cz>
8013 PR driver/83193
8014 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8015 Add "native" as a possible value.
8016 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8017 when native cpu detection is available.
8019 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8020 Martin Sebor <msebor@redhat.com>
8022 PR tree-optimization/84478
8023 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8024 false.
8025 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8026 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8027 support which is conservatively correct, for 2 only stay conservative
8028 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8029 argument to the 2 argument get_range_strlen, adjust 6 arg
8030 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8031 false.
8032 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8033 (gimple_fold_builtin_strlen): Pass true as last argument to
8034 get_range_strlen.
8036 2018-02-20 Martin Sebor <msebor@redhat.com>
8038 PR middle-end/84095
8039 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8040 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8041 (builtin_memref::builtin_memref): Factor out parts into
8042 set_base_and_offset and call it.
8044 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8046 PR middle-end/84406
8047 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8048 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8049 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8050 search at the associated MODE_INT.
8052 2018-02-20 Jeff Law <law@redhat.com>
8054 PR middle-end/82123
8055 PR tree-optimization/81592
8056 PR middle-end/79257
8057 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8058 for range data rather than using global data.
8059 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8060 range data rather than using global data.
8061 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8062 pass it to children as needed.
8063 (struct directive::fmtresult): Similarly.
8064 (struct directive::set_width): Similarly.
8065 (struct directive::set_precision): Similarly.
8066 (format_integer, format_directive, parse_directive): Similarly.
8067 (format_none): Accept unnamed vr_values parameter.
8068 (format_percent, format_floating, format_character): Similarly.
8069 (format_string, format_plain): Similarly.
8070 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8071 the EVRP range analyzer for range data rather than using global data.
8072 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8073 gimple-ssa-evrp-analyze.h
8074 (class sprintf_dom_walker): Add after_dom_children member function.
8075 Add evrp_range_analyzer member.
8076 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8077 range analyzer as needed.
8078 (sprintf_dom_walker::after_dom_children): New member function.
8079 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8080 if not optimizing.
8081 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8082 (evrp_range_analyzer::pop_to_marker): Likewise.
8084 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8086 PR tree-optimization/84419
8087 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8088 with the required type if its current type is compatible but
8089 different.
8091 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8093 PR middle-end/82004
8094 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8095 after vectorization.
8097 2018-02-20 Martin Liska <mliska@suse.cz>
8099 PR driver/83193
8100 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8101 possible values if we don't have a hint.
8103 2018-02-20 Martin Liska <mliska@suse.cz>
8105 PR c/84310
8106 PR target/79747
8107 * final.c (shorten_branches): Build align_tab array with one
8108 more element.
8109 * opts.c (finish_options): Add alignment option limit check.
8110 (MAX_CODE_ALIGN): Likewise.
8111 (MAX_CODE_ALIGN_VALUE): Likewise.
8112 * doc/invoke.texi: Document maximum allowed option value for
8113 all -falign-* options.
8115 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8117 PR target/84146
8118 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8119 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8120 * var-tracking.c (emit_note_insn_var_location): Remove all references
8121 to NOTE_INSN_CALL_ARG_LOCATION.
8122 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8123 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8124 Use copy_rtx_if_shared.
8125 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8126 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8127 (dwarf2out_var_location): Remove handling of
8128 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8129 on call_insn.
8130 * final.c (final_scan_insn): Remove all references to
8131 NOTE_INSN_CALL_ARG_LOCATION.
8132 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8133 before dumping final insns.
8134 * except.c (emit_note_eh_region_end): Remove all references to
8135 NOTE_INSN_CALL_ARG_LOCATION.
8136 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8137 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8138 * config/arc/arc.c (hwloop_optimize): Likewise.
8139 * config/arm/arm.c (create_fix_barrier): Likewise.
8140 * config/s390/s390.c (s390_chunkify_start): Likewise.
8141 * config/sh/sh.c (find_barrier): Likewise.
8142 * config/i386/i386.c (rest_of_insert_endbranch,
8143 ix86_seh_fixup_eh_fallthru): Likewise.
8144 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8145 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8146 * config/frv/frv.c (frv_function_prologue): Likewise.
8147 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8148 reg note.
8149 (note_outside_basic_block_p): Remove all references to
8150 NOTE_INSN_CALL_ARG_LOCATION.
8151 * gengtype.c (adjust_field_rtx_def): Likewise.
8152 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8153 Likewise.
8154 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8155 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8157 PR c++/84444
8158 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8159 is ADDR_EXPR.
8161 PR tree-optimization/84452
8162 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8163 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8164 is NULL.
8166 2018-02-19 Martin Liska <mliska@suse.cz>
8168 PR sanitizer/82183
8169 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8171 2018-02-19 Martin Liska <mliska@suse.cz>
8172 Richard Sandiford <richard.sandiford@linaro.org>
8174 PR tree-optimization/82491
8175 * gimple-fold.c (get_base_constructor): Make earlier bail out
8176 to prevent ubsan.
8178 2018-02-19 Carl Love <cel@us.ibm.com>
8180 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8181 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8182 BU_P8V_OVERLOAD_1.
8183 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8184 P8V_BUILTIN_VEC_NEG.
8186 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8188 * config/rl78/rl78.md (movdf): New define expand.
8190 2018-02-19 Martin Liska <mliska@suse.cz>
8192 PR other/80589
8193 * doc/invoke.texi: Fix typo.
8194 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8196 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8198 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8199 handle rs6000_single_float and rs6000_double_float specially for
8200 e500 family CPUs.
8202 2018-02-16 Jeff Law <law@redhat.com>
8204 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8205 (pop_regs): Use it.
8207 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8209 PR ipa/84425
8210 * ipa-inline.c (inline_small_functions): Fix a typo.
8212 2018-02-16 Nathan Sidwell <nathan@acm.org>
8214 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8216 2018-02-16 Carl Love <cel@us.ibm.com>
8218 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8219 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8220 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8221 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8222 expansion to P8V_BUILTIN_VEC_FLOAT2.
8224 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8226 PR rtl-optimization/70023
8227 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8228 src_regno into account.
8230 2018-02-16 Carl Love <cel@us.ibm.com>
8232 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8233 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8234 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8235 * config/rs6000/rs6000.c: Remove case statements for
8236 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8237 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8238 and P9V_BUILTIN_VEC_VINSERT4B.
8239 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8240 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8241 * config/rs6000/vsx.md:
8242 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8243 vec_insert4b.
8245 2018-02-16 Carl Love <cel@us.ibm.com>
8247 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8248 vec_insert4b.
8249 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8250 definitions.
8251 * config/rs6000/rs6000-c.c: Add the definitions for
8252 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8253 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8254 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8255 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8256 definition for insert4b and define insn *insert3b_internal.
8257 * doc/extend.texi: Add documentation for vec_extract4b.
8259 2018-02-16 Nathan Sidwell <nathan@acm.org>
8261 * doc/extend.texi (Backwards Compatibility): Mention friend
8262 injection. Note for-scope is deprecated.
8263 * doc/invoke.texi (-ffriend-injection): Deprecate.
8265 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8267 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8268 that moved to I2, also allow destinations that are a paradoxical
8269 subreg (instead of a normal reg).
8271 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8273 PR target/83831
8274 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8275 to QImode.
8277 2018-02-16 Richard Biener <rguenther@suse.de>
8279 PR tree-optimization/84037
8280 PR tree-optimization/84016
8281 PR target/82862
8282 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8283 Adjust vec_construct for the fact we need additional higher latency
8284 128bit inserts for AVX256 and AVX512 vector builds.
8285 (ix86_add_stmt_cost): Scale vector construction cost for
8286 elementwise loads.
8288 2018-02-16 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/84417
8291 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8292 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8293 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8295 2018-02-16 Martin Liska <mliska@suse.cz>
8297 PR sanitizer/84307
8298 * internal-fn.def (ASAN_CHECK): Set proper flags.
8299 (ASAN_MARK): Likewise.
8301 2018-02-16 Julia Koval <julia.koval@intel.com>
8303 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8304 from PTA_CANNONLAKE.
8306 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8308 PR target/84272
8309 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8310 Use ++iter rather than iter++ for std::list iterators.
8311 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8312 defer deleting them until all nodes in the forest are processed. Do
8313 free even leaf nodes. Change to_process into auto_vec.
8315 PR bootstrap/84405
8316 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8317 * vec.h (vec_default_construct): Use memset instead of placement new
8318 if BROKEN_VALUE_INITIALIZATION is defined.
8319 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8320 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8321 is defined.
8323 PR rtl-optimization/83723
8324 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8325 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8326 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8327 recursive calls.
8328 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8329 callers.
8330 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8332 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8334 PR rtl-optimization/81443
8335 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8336 from inner REGs to paradoxical SUBREGs.
8338 2018-02-16 Richard Biener <rguenther@suse.de>
8340 PR tree-optimization/84399
8341 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8342 For operands we can analyze at their definition make sure we can
8343 analyze them at each use as well.
8345 2018-02-16 Richard Biener <rguenther@suse.de>
8347 PR tree-optimization/84190
8348 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8349 volatile accesses if the decl isn't volatile.
8351 2018-02-15 Jason Merrill <jason@redhat.com>
8353 PR c++/84314 - ICE with templates and fastcall attribute.
8354 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8355 TYPE_CANONICAL on an existing type.
8357 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8359 PR tree-optimization/84383
8360 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8361 dstoff nor call operand_equal_p if dstbase is NULL.
8363 PR tree-optimization/84334
8364 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8365 also a CONSTANT_CLASS_P, punt.
8367 2018-02-14 Jim Wilson <jimw@sifive.com>
8369 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8370 first SMALL_OPERAND check. New local min_second_step. Move assert
8371 to where locals are set. Add TARGET_RVC support.
8372 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8374 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8376 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8378 2018-02-14 Martin Sebor <msebor@redhat.com>
8380 PR tree-optimization/83698
8381 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8382 arrays constrain the offset range to their bounds.
8383 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8384 (builtin_access::overlap): Avoid setting the size of overlap if it's
8385 already been set.
8386 (maybe_diag_overlap): Also consider arrays when deciding what values
8387 of offsets to include in diagnostics.
8389 2018-02-14 Martin Sebor <msebor@redhat.com>
8391 PR c/84108
8392 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8393 that correspond to the kind of a declaration.
8395 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8397 PR target/83984
8398 * config/pa/pa.md: Load address of PIC label using the linkage table
8399 if the label is nonlocal.
8401 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8403 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8404 warning message if user requests -maltivec=be.
8405 * doc/invoke.texi: Document deprecation of -maltivec=be.
8407 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8409 PR target/84220
8410 * config/rs6000/rs6000-c.c: Update definitions for
8411 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8412 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8414 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8416 PR target/84239
8417 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8418 add _get_ssp intrinsics. Remove argument from
8419 __builtin_ia32_rdssp[d|q].
8420 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8421 * config/i386/i386-builtin.def: Remove argument from
8422 __builtin_ia32_rdssp[d|q].
8423 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8424 ix86_expand_special_args_builtin for _rdssp[d|q].
8425 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8426 Clear register before usage.
8427 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8428 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8430 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8432 PR tree-optimization/84357
8433 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8434 operand 1 of an ARRAY_REF too.
8436 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8438 PR target/83831
8439 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8440 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8441 declarations.
8442 (set_of_reg): New struct.
8443 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8444 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8445 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8446 functions.
8447 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8448 Split into bitclr, bitset, bitinvert patterns if appropriate.
8449 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8450 use rx_fuse_in_memory_bitop.
8451 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8452 to named insn, correct maximum insn length.
8454 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8456 PR target/79242
8457 * machmode.def: Define a complex mode for PARTIAL_INT.
8458 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8459 MODE_PARTIAL_INT.
8460 * doc/rtl.texi: Document CSPImode.
8461 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8462 handling.
8463 (msp430_hard_regno_nregs_with_padding): Likewise.
8465 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8467 PR target/84279
8468 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8470 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8472 PR rtl-optimization/84169
8473 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8474 we generated a parallel as new i3 and we split that to new i2 and i3
8475 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8476 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8477 those to i2, not i1. Partially rewrite this scan code.
8479 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8481 PR c/82210
8482 * stor-layout.c (place_field): For variable length fields, adjust
8483 offset_align afterwards not just based on the field's alignment,
8484 but also on the size.
8486 PR middle-end/84309
8487 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8488 of exps and logs in the use_exp2 case.
8490 2018-02-13 Jeff Law <law@redhat.com>
8492 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8493 entry for "vector".
8495 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8496 ARGS as unused.
8498 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8500 PR debug/84342
8501 PR debug/84319
8502 * common.opt (gas-loc-support, gas-locview-support): New.
8503 (ginline-points, ginternal-reset-location-views): New.
8504 * doc/invoke.texi: Document them. Use @itemx where intended.
8505 (gvariable-location-views): Adjust.
8506 * target.def (reset_location_view): New.
8507 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8508 (TARGET_RESET_LOCATION_VIEW): New.
8509 * doc/tm.texi: Rebuilt.
8510 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8511 (dwarf2out_default_as_locview_support): New.
8512 (output_asm_line_debug_info): Use option variables.
8513 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8514 (output_loc_list): Likewise.
8515 (add_high_low_attributes): Check option variables.
8516 Don't output entry view attribute in strict mode.
8517 (gen_inlined_subroutine_die): Check option variables.
8518 (dwarf2out_inline_entry): Likewise.
8519 (init_sections_and_labels): Likewise.
8520 (dwarf2out_early_finish): Likewise.
8521 (maybe_reset_location_view): New, from...
8522 (dwarf2out_var_location): ... here. Call it.
8523 * debug.h (dwarf2out_default_as_loc_support): Declare.
8524 (dwarf2out_default_as_locview_support): Declare.
8525 * hooks.c (hook_int_rtx_insn_0): New.
8526 * hooks.h (hook_int_rtx_insn_0): Declare.
8527 * toplev.c (process_options): Take -gas-loc-support and
8528 -gas-locview-support from dwarf2out. Enable
8529 -gvariable-location-views by default only with locview
8530 assembler support. Enable -ginternal-reset-location-views by
8531 default only if the target defines the corresponding hook.
8532 Enable -ginline-points by default if location views are
8533 enabled; force it disabled if statement frontiers are
8534 disabled.
8535 * tree-inline.c (expand_call_inline): Check option variables.
8536 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8538 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8540 PR tree-optimization/84321
8541 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8542 handling. Also check whether the anti-range contains any values
8543 that satisfy the mask; switch to a VR_RANGE if not.
8545 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8547 PR sanitizer/84340
8548 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8550 2018-02-13 Martin Jambor <mjambor@suse.cz>
8552 PR c++/83990
8553 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8554 of call statements, also set location of a load to a temporary.
8556 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8558 * config/rl78/rl78.c (add_vector_labels): New function.
8559 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8560 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8561 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8562 which checks that no arguments are passed.
8563 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8564 * doc/extend.texi: Documentation for the new attribute.
8566 2018-02-13 Andreas Schwab <schwab@suse.de>
8568 * config/riscv/linux.h (CPP_SPEC): Define.
8570 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8572 PR target/84335
8573 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8574 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8575 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8576 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8577 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8578 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8579 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8580 temporarily for AES and PCLMUL builtins.
8582 PR tree-optimization/84339
8583 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8584 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8585 Formatting fixes.
8587 PR middle-end/84309
8588 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8589 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8590 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8591 inline function.
8592 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8593 inline function.
8594 * omp-simd-clone.h: New file.
8595 * omp-simd-clone.c: Include omp-simd-clone.h.
8596 (expand_simd_clones): No longer static.
8597 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8598 cgraph.h and omp-simd-clone.h.
8599 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8600 (vect_recog_widen_shift_pattern): Formatting fix.
8601 (vect_pattern_recog_1): Don't check optab for calls.
8603 PR target/84336
8604 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8605 operands[2] into a REG before using gen_lowpart on it.
8607 2018-02-12 Jeff Law <law@redhat.com>
8609 PR target/83760
8610 * config/sh/sh.c (find_barrier): Consider a sibling call
8611 a barrier as well.
8613 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8614 successfully back substituting a reg.
8616 2018-02-12 Richard Biener <rguenther@suse.de>
8618 PR tree-optimization/84037
8619 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8620 parameter, move visited init to caller.
8621 (vect_slp_analyze_operations): Separate cost from validity
8622 check, initialize visited once for all instances.
8623 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8624 for all instances.
8625 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8626 out an assert.
8627 (vect_model_promotion_demotion_cost): Likewise.
8628 (vectorizable_bswap): Guard cost modeling with !slp_node
8629 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8630 SLP stmts.
8631 (vectorizable_call): Likewise.
8632 (vectorizable_conversion): Likewise.
8633 (vectorizable_assignment): Likewise.
8634 (vectorizable_shift): Likewise.
8635 (vectorizable_operation): Likewise.
8636 (vectorizable_store): Likewise.
8637 (vectorizable_load): Likewise.
8638 (vectorizable_condition): Likewise.
8639 (vectorizable_comparison): Likewise.
8641 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8643 PR sanitizer/84307
8644 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8645 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8646 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8648 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8650 PR middle-end/83665
8651 * params.def (inline-min-speedup): Increase from 8 to 15.
8652 (max-inline-insns-auto): Decrease from 40 to 30.
8653 * ipa-split.c (consider_split): Add some buffer for function to
8654 be considered inlining candidate.
8655 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8656 default values.
8658 2018-02-12 Richard Biener <rguenther@suse.de>
8660 PR tree-optimization/84037
8661 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8662 matched stmts if we cannot swap the non-matched ones.
8664 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8666 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8667 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8668 _mm_maskz_scalef_round_ss): New intrinsics.
8669 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8670 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8671 __builtin_ia32_scalefss_round): Remove.
8672 (__builtin_ia32_scalefsd_mask_round,
8673 __builtin_ia32_scalefss_mask_round): New intrinsics.
8674 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8675 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8676 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8677 "<round_constraint>")): Changed to ...
8678 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8679 "<round_scalar_constraint>")): ... this.
8680 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8681 %0, %1, %2<round_op3>}"): Changed to ...
8682 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8683 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8684 %2<round_scalar_mask_op3>}"): ... this.
8685 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8687 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8689 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8690 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8691 (_mm_maskz_sqrt_round_ss): New intrinsics.
8692 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8693 (__builtin_ia32_sqrtsd_mask_round)
8694 (__builtin_ia32_sqrtss_mask_round): New builtins.
8695 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8696 (__builtin_ia32_sqrtss_round): Remove.
8697 (__builtin_ia32_sqrtsd_mask_round)
8698 (__builtin_ia32_sqrtss_mask_round): New builtins.
8699 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8700 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8701 ((match_operand:VF_128 1 "vector_operand"
8702 "xBm,<round_constraint>")): Changed to ...
8703 ((match_operand:VF_128 1 "vector_operand"
8704 "xBm,<round_scalar_constraint>")): ... this.
8705 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8706 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8707 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8708 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8709 %<iptr>1<round_scalar_mask_op3>}): ... this.
8710 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8711 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8713 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8715 PR target/84266
8716 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8717 Cast vec_cmpeq result to correct type.
8718 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8719 Cast vec_cmpgt result to correct type.
8721 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8723 * final.c (final_scan_insn_1): Renamed from...
8724 (final_scan_insn): ... this. New wrapper, to recover
8725 seen from the outermost call in recursive ones.
8726 * config/sparc/sparc.c (output_return): Drop seen from call.
8727 (output_sibcall): Likewise.
8728 * config/visium/visium.c (output_branch): Likewise.
8730 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8732 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8733 function label.
8735 2018-02-10 Alan Modra <amodra@gmail.com>
8737 PR target/84300
8738 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8739 Specify LR as an input.
8741 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8743 PR sanitizer/83987
8744 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8745 remove_member_access_dummy_vars): New functions.
8746 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8747 lower_omp_1, execute_lower_omp): Use them.
8749 PR rtl-optimization/84308
8750 * shrink-wrap.c (spread_components): Release todo vector.
8752 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8754 PR rtl-optimization/57193
8755 * ira-color.c (struct allocno_color_data): Add member
8756 conflict_allocno_hard_prefs.
8757 (update_conflict_allocno_hard_prefs): New.
8758 (bucket_allocno_compare_func): Add a preference based on
8759 conflict_allocno_hard_prefs.
8760 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8761 (color_allocnos): Remove a dead code. Initiate
8762 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8764 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8766 PR target/84226
8767 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8768 constraint from =wa to wa. Avoid a subreg on the output operand,
8769 instead use a pseudo and subreg it in a move.
8770 (p9_xxbrd_<mode>): Changed to ...
8771 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8772 (p9_xxbrd_v2df): New expander.
8773 (p9_xxbrw_<mode>): Changed to ...
8774 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8775 (p9_xxbrw_v4sf): New expander.
8777 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8779 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8781 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8783 PR target/83926
8784 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8785 multiply in 32-bit mode.
8786 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8787 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8788 mode.
8790 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8792 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8793 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8794 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8795 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8797 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8799 PR lto/84213
8800 * dwarf2out.c (is_trivial_indirect_ref): New function.
8801 (dwarf2out_late_global_decl): Do not generate a location
8802 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8803 and that are not defined in the current unit.
8805 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8807 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8808 instead of a libcall for UNORDERED.
8810 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8812 PR target/82641
8813 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8814 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8816 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8818 PR target/PR84295
8819 * config/s390/s390.c (s390_set_current_function): Invoke
8820 s390_indirect_branch_settings also if fndecl didn't change.
8822 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8824 * config/rs6000/rs6000.md (blockage): Set length to zero.
8826 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8828 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8830 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8832 PR sanitizer/84285
8833 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8834 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8835 -static-lib*san.
8837 PR debug/84252
8838 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8839 PARALLEL incoming that failed vt_get_decl_and_offset check.
8841 PR middle-end/84237
8842 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8843 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8844 TREE_READONLY bit.
8845 (get_variable_section): For decls in named .bss* sections pass true as
8846 second argument to bss_initializer_p.
8848 2018-02-09 Marek Polacek <polacek@redhat.com>
8849 Jakub Jelinek <jakub@redhat.com>
8851 PR c++/83659
8852 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8853 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8854 Sync some changes from cxx_fold_indirect_ref.
8856 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8858 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8859 markers.
8860 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8861 (BLOCK_INLINE_ENTRY_LABEL): New.
8862 (dwarf2out_var_location): Disregard inline entry markers.
8863 (inline_entry_data): New struct.
8864 (inline_entry_data_hasher): New hashtable type.
8865 (inline_entry_data_hasher::hash): New.
8866 (inline_entry_data_hasher::equal): New.
8867 (inline_entry_data_table): New variable.
8868 (add_high_low_attributes): Add DW_AT_entry_pc and
8869 DW_AT_GNU_entry_view attributes if a pending entry is found
8870 in inline_entry_data_table. Add old entry_pc attribute only
8871 if debug nonbinding markers are disabled.
8872 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8873 markers are enabled.
8874 (block_within_block_p, dwarf2out_inline_entry): New.
8875 (dwarf2out_finish): Check that no entries remained in
8876 inline_entry_data_table.
8877 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8878 (final_scan_insn, notice_source_line): Likewise.
8879 (rest_of_clean_state): Skip inline entry markers.
8880 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8881 markers.
8882 * gimple.c (gimple_build_debug_inline_entry): New.
8883 * gimple.h (enum gimple_debug_subcode): Add
8884 GIMPLE_DEBUG_INLINE_ENTRY.
8885 (gimple_build_debug_inline_entry): Declare.
8886 (gimple_debug_inline_entry_p): New.
8887 (gimple_debug_nonbind_marker_p): Adjust.
8888 * insn-notes.def (INLINE_ENTRY): New.
8889 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8890 inline entry marker notes.
8891 (print_insn): Likewise.
8892 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8893 (INSN_DEBUG_MARKER_KIND): Likewise.
8894 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8895 * tree-inline.c (expand_call_inline): Build and insert
8896 debug_inline_entry stmt.
8897 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8898 inline entry blocks early, if nonbind markers are enabled.
8899 (dump_scope_block): Dump fragment info.
8900 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8901 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8902 (gimple_build_debug_inline_entry): New.
8903 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8904 Enable/disable inline entry points too.
8905 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8906 (DEBUG_INSN): Describe inline entry markers.
8908 * common.opt (gvariable-location-views): New.
8909 (gvariable-location-views=incompat5): New.
8910 * config.in: Rebuilt.
8911 * configure: Rebuilt.
8912 * configure.ac: Test assembler for view support.
8913 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8914 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8915 * dwarf2out.c (var_loc_view): New typedef.
8916 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8917 (dwarf2out_locviews_in_attribute): New.
8918 (dwarf2out_locviews_in_loclist): New.
8919 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8920 (enum dw_line_info_opcode): Add LI_adv_address.
8921 (struct dw_line_info_table): Add view.
8922 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8923 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8924 (zero_view_p): New variable.
8925 (ZERO_VIEW_P): New macro.
8926 (output_asm_line_debug_info): New.
8927 (struct var_loc_node): Add view.
8928 (add_AT_view_list, AT_loc_list): New.
8929 (add_var_loc_to_decl): Add view param. Test it against last.
8930 (new_loc_list): Add view params. Record them.
8931 (AT_loc_list_ptr): Handle loc and view lists.
8932 (view_list_to_loc_list_val_node): New.
8933 (print_dw_val): Handle dw_val_class_view_list.
8934 (size_of_die): Likewise.
8935 (value_format): Likewise.
8936 (loc_list_has_views): New.
8937 (gen_llsym): Set vl_symbol too.
8938 (maybe_gen_llsym, skip_loc_list_entry): New.
8939 (dwarf2out_maybe_output_loclist_view_pair): New.
8940 (output_loc_list): Output view list or entries too.
8941 (output_view_list_offset): New.
8942 (output_die): Handle dw_val_class_view_list.
8943 (output_dwarf_version): New.
8944 (output_compilation_unit_header): Use it.
8945 (output_skeleton_debug_sections): Likewise.
8946 (output_rnglists, output_line_info): Likewise.
8947 (output_pubnames, output_aranges): Update version comments.
8948 (output_one_line_info_table): Output view numbers in asm comments.
8949 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8950 Call maybe_gen_llsym.
8951 (loc_list_from_tree_1): Adjust.
8952 (add_AT_location_description): Create view list attribute if
8953 needed, check it's absent otherwise.
8954 (convert_cfa_to_fb_loc_list): Adjust.
8955 (maybe_emit_file): Call output_asm_line_debug_info for test.
8956 (dwarf2out_var_location): Reset views as needed. Precompute
8957 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8958 attribute. Set view.
8959 (new_line_info_table): Reset next view.
8960 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8961 (dwarf2out_source_line): Likewise. Output view resets and labels to
8962 the assembler, or select appropriate line info opcodes.
8963 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8964 (optimize_string_length): Catch it. Adjust.
8965 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8966 dw_val_class_view_list, and remove it if no longer needed.
8967 (hash_loc_list): Hash view numbers.
8968 (loc_list_hasher::equal): Compare them.
8969 (optimize_location_lists): Check whether a view list symbol is
8970 needed, and whether the locview attribute is present, and
8971 whether they match. Remove the locview attribute if no longer
8972 needed.
8973 (index_location_lists): Call skip_loc_list_entry for test.
8974 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8975 Use output_dwarf_version.
8976 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8977 (struct dw_val_node): Add val_view_list.
8978 * final.c (SEEN_NEXT_VIEW): New.
8979 (set_next_view_needed): New.
8980 (clear_next_view_needed): New.
8981 (maybe_output_next_view): New.
8982 (final_start_function): Rename to...
8983 (final_start_function_1): ... this. Take pointer to FIRST,
8984 add SEEN parameter. Emit param bindings in the initial view.
8985 (final_start_function): Reintroduce SEEN-less interface.
8986 (final): Rename to...
8987 (final_1): ... this. Take SEEN parameter. Output final pending
8988 next view at the end.
8989 (final): Reintroduce seen-less interface.
8990 (final_scan_insn): Output pending next view before switching
8991 sections or ending a block. Mark the next view as needed when
8992 outputting variable locations. Notify debug backend of section
8993 changes, and of location view changes.
8994 (rest_of_handle_final): Adjust.
8995 * toplev.c (process_options): Autodetect value for debug variable
8996 location views option. Warn on incompat5 without -gdwarf-5.
8997 * doc/invoke.texi (gvariable-location-views): New.
8998 (gvariable-location-views=incompat5): New.
8999 (gno-variable-location-views): New.
9001 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9003 PR tree-optimization/84136
9004 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9005 that the result of find_edge is non-NULL.
9007 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9009 PR target/83008
9010 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9011 storing integer register in SImode. Fix cost of 256 and 512
9012 byte aligned SSE register store.
9014 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9016 * config/i386/i386.c (ix86_multiplication_cost): Fix
9017 multiplication cost for TARGET_AVX512DQ.
9019 2018-02-08 Marek Polacek <polacek@redhat.com>
9021 PR tree-optimization/84238
9022 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9023 get_range_strlen.
9025 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9027 PR tree-optimization/84265
9028 * tree-vect-stmts.c (vectorizable_store): Don't treat
9029 VMAT_CONTIGUOUS accesses as grouped.
9030 (vectorizable_load): Likewise.
9032 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9034 PR tree-optimization/81635
9035 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9036 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9037 (test_round_for_mask): New functions.
9038 (wide_int_cc_tests): Call test_round_for_mask.
9039 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9040 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9041 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9042 range returned by get_range_info.
9044 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9046 PR ipa/81360
9047 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9048 * symtab.c: Include builtins.h
9049 (symtab_node::output_to_lto_symbol_table_p): Move here
9050 from lto-streamer-out.c:output_symbol_p.
9051 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9052 (output_symbol_p): Move all logic to symtab.c
9053 (produce_symtab): Update.
9055 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9057 * config/s390/s390-opts.h (enum indirect_branch): Define.
9058 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9059 (s390_indirect_branch_via_thunk)
9060 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9061 (enum s390_indirect_branch_type): Define.
9062 * config/s390/s390.c (struct s390_frame_layout, struct
9063 machine_function): Remove.
9064 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9065 (indirect_branch_table_label_no, indirect_branch_table_name):
9066 Define variables.
9067 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9068 (enum s390_indirect_branch_option): Define.
9069 (s390_return_addr_from_memory): New function.
9070 (s390_handle_string_attribute): New function.
9071 (s390_attribute_table): Add new attribute handler.
9072 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9073 (s390_indirect_branch_via_thunk): New function.
9074 (s390_indirect_branch_via_inline_thunk): New function.
9075 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9076 sibling call optimization for non z10 compiles.
9077 (s390_emit_call): Force indirect branch target to be a single
9078 register. Add r1 clobber for non-z10 compiles.
9079 (s390_emit_epilogue): Emit return jump via return_use expander.
9080 (s390_reorg): Handle JUMP_INSNs as execute targets.
9081 (s390_option_override_internal): Perform validity checks for the
9082 new command line options.
9083 (s390_indirect_branch_attrvalue): New function.
9084 (s390_indirect_branch_settings): New function.
9085 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9086 (s390_output_indirect_thunk_function): New function.
9087 (s390_code_end): Implement target hook.
9088 (s390_case_values_threshold): Implement target hook.
9089 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9090 macros.
9091 * config/s390/s390.h (struct s390_frame_layout)
9092 (struct machine_function): Move here from s390.c.
9093 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9094 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9095 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9096 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9097 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9098 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9099 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9100 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9101 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9102 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9103 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9104 (mnemonic attribute): Add values which aren't recognized
9105 automatically.
9106 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9107 pattern for branch conversion. Fix mnemonic attribute.
9108 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9109 indirect branch via thunk if requested.
9110 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9111 ("*indirect_jump"): Disable for branch conversion using out of
9112 line thunks.
9113 ("indirect_jump_via_thunk<mode>_z10")
9114 ("indirect_jump_via_thunk<mode>")
9115 ("indirect_jump_via_inlinethunk<mode>_z10")
9116 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9117 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9118 ("casesi_jump_via_inlinethunk<mode>_z10")
9119 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9120 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9121 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9122 ("*indirect2_jump"): Disable for branch conversion.
9123 ("casesi_jump"): Turn into expander and expand patterns for branch
9124 conversion.
9125 ("return_use"): New expander.
9126 ("*return"): Emit return via thunk and rename it to ...
9127 ("*return<mode>"): ... this one.
9128 * config/s390/s390.opt: Add new options and and enum for the
9129 option values.
9131 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9133 * lra-constraints.c (match_reload): Unconditionally use
9134 gen_lowpart_SUBREG, rather than selecting between that
9135 and equivalent gen_rtx_SUBREG code.
9137 2018-02-08 Richard Biener <rguenther@suse.de>
9139 PR tree-optimization/84233
9140 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9141 changed flag instead of boguously re-using phi_inserted.
9143 2018-02-08 Martin Jambor <mjambor@suse.cz>
9145 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9146 static local variables.
9148 2018-02-08 Richard Biener <rguenther@suse.de>
9150 PR tree-optimization/84278
9151 * tree-vect-stmts.c (vectorizable_store): When looking for
9152 smaller vector types to perform grouped strided loads/stores
9153 make sure the mode is supported by the target.
9154 (vectorizable_load): Likewise.
9156 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9158 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9159 Increase LDP/STP opportunities by adding adjacent callee-saves.
9161 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9163 PR rtl-optimization/84068
9164 PR rtl-optimization/83459
9165 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9167 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9169 PR tree-optimization/84224
9170 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9171 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9172 non-zero arguments.
9174 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9176 PR target/84113
9177 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9178 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9179 count, remove one USE.
9181 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9183 * doc/install.texi (Configuration): Document the
9184 --with-long-double-format={ibm,ieee} PowerPC configuration
9185 options.
9187 PR target/84154
9188 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9189 Convert from define_expand to be define_insn_and_split. Rework
9190 float/double/_Float128 conversions to QI/HI/SImode to work with
9191 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9192 conversions to QI/HImode types did a store and then a load to
9193 truncate the value. For conversions to VSX registers, don't split
9194 the insn, instead emit the code directly. Use the code iterator
9195 any_fix to combine signed and unsigned conversions.
9196 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9197 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9198 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9199 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9200 (fix_<mode>di2_hw): Likewise.
9201 (fixuns_<mode>di2_hw): Likewise.
9202 (fix_<mode>si2_hw): Likewise.
9203 (fixuns_<mode>si2_hw): Likewise.
9204 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9205 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9206 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9207 fix<uns>_trunc<SFDF:mode>si2_p8.
9208 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9209 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9210 (fix<uns>_<mode>_mem): Likewise.
9211 (fctiw<u>z_<mode>_mem): Likewise.
9212 (fix<uns>_<mode>_mem): Likewise.
9213 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9214 the register allocator from doing a direct move to the GPRs to do
9215 a store, and instead use the ISA 3.0 store byte/half-word from
9216 vector register instruction. For IEEE 128-bit floating point,
9217 also optimize stores of 32-bit ints.
9218 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9220 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9222 * genextract.c (push_pathstr_operand): New function to support
9223 [a-zA-Z].
9224 (walk_rtx): Call push_pathstr_operand.
9225 (print_path): Support [a-zA-Z].
9227 2018-02-07 Richard Biener <rguenther@suse.de>
9229 PR tree-optimization/84037
9230 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9231 (cse_and_gimplify_to_preheader): Declare.
9232 (vect_get_place_in_interleaving_chain): Likewise.
9233 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9234 ivexpr_map.
9235 (_loop_vec_info::~_loop_vec_info): Delete it.
9236 (cse_and_gimplify_to_preheader): New function.
9237 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9238 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9239 (vectorizable_load): Likewise. For grouped stores always base
9240 the IV on the first element.
9241 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9242 condition before gimplifying.
9244 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9246 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9247 *DIV_EXPR and *MOD_EXPR.
9249 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9251 PR target/84248
9252 * config/i386/i386.c (ix86_option_override_internal): Mask out
9253 the CF_SET bit when checking -fcf-protection.
9255 2018-02-07 Tom de Vries <tom@codesourcery.com>
9257 PR libgomp/84217
9258 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9259 enough.
9261 2018-02-07 Richard Biener <rguenther@suse.de>
9263 PR tree-optimization/84204
9264 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9265 this place.
9267 PR tree-optimization/84205
9268 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9269 special-case isl_ast_op_zdiv_r.
9271 PR tree-optimization/84223
9272 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9273 Only add conditions from within the region.
9274 (gather_bbs::after_dom_children): Adjust.
9276 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9278 PR target/84209
9279 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9280 * config/avr/avr.md: Only post-reload split REG-REG moves if
9281 either register is GENERAL_REG_P.
9283 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9285 PR tree-optimization/84235
9286 * tree-ssa-scopedtables.c
9287 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9288 if the subtraction is performed in floating point type where NaNs are
9289 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9290 build 1. Formatting fix.
9292 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9294 PR target/84146
9295 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9296 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9297 and skip it regardless of bb boundaries. Use CALL_P macro,
9298 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9299 unnecessarily, formatting fix.
9301 2018-02-06 Michael Collison <michael.collison@arm.com>
9303 * config/arm/thumb2.md:
9304 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9305 (*thumb_mov_notscc): Ditto.
9307 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9309 PR target/84154
9310 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9311 unsigned_fix, not "s".
9313 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9315 * configure.ac (gcc_fn_eh_frame_ro): New function.
9316 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9317 correct .eh_frame permissions.
9318 * configure: Regenerate.
9320 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9322 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9323 irrelevant options.
9325 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9327 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9328 Display warning message for -mno-speculate-indirect-jumps.
9330 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9332 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9333 Undocumented.
9334 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9336 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9338 PR tree-optimization/84225
9339 * tree-eh.c (find_trapping_overflow): Only call
9340 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9342 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9344 PR target/84145
9345 * config/i386/i386.c: Reimplement the check of possible options
9346 -mibt/-mshstk conbination. Change error messages.
9347 * doc/invoke.texi: Fix a typo: remove extra '='.
9349 2018-02-06 Marek Polacek <polacek@redhat.com>
9351 PR tree-optimization/84228
9352 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9354 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9356 PR target/82641
9357 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9358 emitted arch directives.
9359 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9360 __ARM_FEATURE_COPROC before changing architectures.
9362 2018-02-06 Richard Biener <rguenther@suse.de>
9364 * config/i386/i386.c (print_reg): Fix typo.
9365 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9367 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9369 * configure: Regenerate.
9371 2018-02-05 Martin Sebor <msebor@redhat.com>
9373 PR tree-optimization/83369
9374 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9375 inlining context.
9377 2018-02-05 Martin Liska <mliska@suse.cz>
9379 * doc/invoke.texi: Cherry-pick upstream r323995.
9381 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9383 * ira.c (ira_init_register_move_cost): Adjust comment.
9385 2018-02-05 Martin Liska <mliska@suse.cz>
9387 PR gcov-profile/84137
9388 * doc/gcov.texi: Fix typo in documentation.
9390 2018-02-05 Martin Liska <mliska@suse.cz>
9392 PR gcov-profile/83879
9393 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9394 using dlopen functionality.
9396 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9398 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9399 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9400 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9401 _mm_maskz_range_round_ss): New intrinsics.
9402 (__builtin_ia32_rangesd128_round)
9403 (__builtin_ia32_rangess128_round): Remove.
9404 (__builtin_ia32_rangesd128_mask_round,
9405 __builtin_ia32_rangess128_mask_round): New builtins.
9406 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9407 __builtin_ia32_rangess128_round): Remove.
9408 (__builtin_ia32_rangesd128_mask_round,
9409 __builtin_ia32_rangess128_mask_round): New builtins.
9410 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9411 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9412 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9413 "<round_saeonly_constraint>")): Changed to ...
9414 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9415 "<round_saeonly_scalar_constraint>")): ... this.
9416 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9417 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9418 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9419 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9420 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9422 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9424 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9425 options.
9426 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9427 Remove all values except native, 8540 and 8548.
9429 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9431 * config/i386/i386.c (ix86_output_function_return): Pass
9432 INVALID_REGNUM, instead of -1, as invalid register number to
9433 indirect_thunk_name and output_indirect_thunk.
9435 2018-02-02 Julia Koval <julia.koval@intel.com>
9437 * config.gcc: Add -march=icelake.
9438 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9439 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9440 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9441 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9442 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9443 (processor_target_table): Add icelake.
9444 (ix86_option_override_internal): Handle new PTAs.
9445 (get_builtin_code_for_version): Handle icelake.
9446 (M_INTEL_COREI7_ICELAKE): New.
9447 (fold_builtin_cpu): Handle icelake.
9448 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9449 * doc/invoke.texi: Add -march=icelake.
9451 2018-02-02 Julia Koval <julia.koval@intel.com>
9453 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9454 to wide_int_bitmask.
9455 * wide-int-bitmask.h: New.
9457 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9459 PR target/84066
9460 * config/i386/i386.md: Replace Pmode with word_mode in
9461 builtin_setjmp_setup and builtin_longjmp to support x32.
9463 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9465 PR target/56010
9466 PR target/83743
9467 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9468 #include "opts.h".
9469 (rs6000_supported_cpu_names): New static variable.
9470 (linux_cpu_translation_table): Likewise.
9471 (elf_platform) <cpu>: Define new static variable and use it.
9472 Translate kernel AT_PLATFORM name to canonical name if needed.
9473 Error if platform name is unknown.
9475 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9477 PR target/84089
9478 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9480 2018-02-01 Jeff Law <law@redhat.com>
9482 PR target/84128
9483 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9484 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9485 the scratch if RELEASE_VIA_POP is false.
9486 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9487 If we have to save a temporary register, decrement SIZE appropriately.
9488 Pass new arguments to release_scratch_register_on_entry.
9489 (ix86_adjust_stack_and_probe): Likewise.
9490 (ix86_emit_probe_stack_range): Pass new arguments to
9491 release_scratch_register_on_entry.
9493 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9495 PR rtl-optimization/84157
9496 * combine.c (change_zero_ext): Use REG_P predicate in
9497 front of HARD_REGISTER_P predicate.
9499 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9501 * config/avr/avr.c (avr_option_override): Move disabling of
9502 -fdelete-null-pointer-checks to...
9503 * common/config/avr/avr-common.c (avr_option_optimization_table):
9504 ...here.
9506 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9508 PR tree-optimization/81635
9509 * tree-data-ref.c (split_constant_offset_1): For types that
9510 wrap on overflow, try to use range info to prove that wrapping
9511 cannot occur.
9513 2018-02-01 Renlin Li <renlin.li@arm.com>
9515 PR target/83370
9516 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9517 TAILCALL_ADDR_REGS.
9518 (aarch64_register_move_cost): Likewise.
9519 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9520 TAILCALL_ADDR_REGS.
9521 (REG_CLASS_NAMES): Likewise.
9522 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9523 TAILCALL_ADDR_REGS. Remove IP registers.
9524 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9526 2018-02-01 Richard Biener <rguenther@suse.de>
9528 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9529 for specifying RPO order and allow NULL for that.
9530 * domwalk.c (dom_walker::dom_walker): Likewise.
9531 (dom_walker::walk): Handle NULL RPO order.
9532 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9533 in RPO order.
9534 (rewrite_update_dom_walker): Likewise.
9535 (mark_def_dom_walker): Likewise.
9537 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9539 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9540 (aarch64_maybe_expand_sve_subreg_move): Declare.
9541 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9542 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9543 predicate.
9544 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9545 that are semantically a reverse operation.
9546 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9547 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9548 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9549 functions.
9550 (aarch64_can_change_mode_class): For big-endian, forbid changes
9551 between two SVE modes if they have different element sizes.
9553 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9555 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9556 the TImode handling for big-endian targets.
9558 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9560 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9561 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9562 not just bytes.
9563 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9564 Remove BSWAP handing for big-endian targets and use the form of
9565 LD1RQ appropariate for the mode.
9567 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9569 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9570 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9571 duplicated element.
9573 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9575 PR tearget/83845
9576 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9577 check for operands that need to go through aarch64_sve_reload_be.
9579 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9581 PR tree-optimization/81661
9582 PR tree-optimization/84117
9583 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9584 * tree-eh.c: Include gimplify.h.
9585 (find_trapping_overflow, replace_trapping_overflow,
9586 rewrite_to_non_trapping_overflow): New functions.
9587 * tree-vect-loop.c: Include tree-eh.h.
9588 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9589 * tree-data-ref.c: Include tree-eh.h.
9590 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9592 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9594 PR rtl-optimization/84123
9595 * combine.c (change_zero_ext): Check if hard register satisfies
9596 can_change_dest_mode before calling gen_lowpart_SUBREG.
9598 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9600 PR target/82444
9601 * ira.c (ira_init_register_move_cost): Remove assert.
9603 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9605 PR rtl-optimization/84071
9606 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9607 * doc/tm.texi: Regenerate.
9609 2018-01-31 Richard Biener <rguenther@suse.de>
9611 PR tree-optimization/84132
9612 * tree-data-ref.c (analyze_miv_subscript): Properly
9613 check whether evolution_function_is_affine_multivariate_p
9614 before calling gcd_of_steps_may_divide_p.
9616 2018-01-31 Julia Koval <julia.koval@intel.com>
9618 PR target/83618
9619 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9620 * config/i386/i386.md (rdpid_rex64) New.
9621 (rdpid): Make 32bit only.
9623 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9625 PR lto/84105
9626 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9627 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9629 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9631 Revert
9632 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9634 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9636 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9638 PR rtl-optimization/84071
9639 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9640 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9642 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9644 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9645 (arc_attribute_table): Add 'aux' attribute.
9646 (arc_in_small_data_p): Consider aux like variables.
9647 (arc_is_aux_reg_p): New function.
9648 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9649 (arc_get_aux_arg): New function.
9650 (prepare_move_operands): Handle aux-register access.
9651 (arc_handle_aux_attribute): New function.
9652 * doc/extend.texi (ARC Variable attributes): Add subsection.
9654 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9656 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9657 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9658 (arc_attribute_table): Add 'uncached' attribute.
9659 (arc_print_operand): Print '.di' flag for uncached memory
9660 accesses.
9661 (arc_in_small_data_p): Do not consider for small data the uncached
9662 types.
9663 (arc_is_uncached_mem_p): New function.
9664 * config/arc/predicates.md (compact_store_memory_operand): Check
9665 for uncached memory accesses.
9666 (nonvol_nonimm_operand): Likewise.
9667 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9669 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9671 PR c/84100
9672 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9673 falign-loops=): Add Optimization flag.
9675 2018-01-30 Jeff Law <law@redhat.com>
9677 PR target/84064
9678 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9679 INT_REGISTERS_SAVED. Check it prior to calling
9680 get_scratch_register_on_entry.
9681 (ix86_adjust_stack_and_probe): Similarly.
9682 (ix86_emit_probe_stack_range): Similarly.
9683 (ix86_expand_prologue): Corresponding changes.
9685 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9687 PR target/40411
9688 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9689 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9691 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9693 PR target/84112
9694 * lra-constraints.c (curr_insn_transform): Process AND in the
9695 address.
9697 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9699 PR rtl-optimization/83986
9700 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9701 dependence against last_pending_memory_flush in addition to
9702 pending_jump_insns.
9704 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9706 PR tree-optimization/81611
9707 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9708 copies.
9710 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9712 PR target/83758
9713 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9714 a reg rtx.
9716 2018-01-30 Richard Biener <rguenther@suse.de>
9717 Jakub Jelinek <jakub@redhat.com>
9719 PR tree-optimization/84111
9720 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9721 inner loops added during recursion, as they don't have up-to-date
9722 SSA form.
9724 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9726 PR ipa/81360
9727 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9728 (can_inline_edge_by_limits_p): ... here.
9729 (can_early_inline_edge_p, check_callers,
9730 update_caller_keys, update_callee_keys, recursive_inlining,
9731 add_new_edges_to_heap, speculation_useful_p,
9732 inline_small_functions,
9733 inline_small_functions, flatten_function,
9734 inline_to_all_callers_1): Update.
9736 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9738 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9739 zeros correctly.
9741 2018-01-30 Richard Biener <rguenther@suse.de>
9743 PR tree-optimization/83008
9744 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9745 invariant and constant vector uses in stmts when they need
9746 more than one stmt.
9748 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9750 PR bootstrap/84017
9751 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9752 * configure: Regenerate.
9754 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9756 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9757 pattern.
9758 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9759 Use gen_rtx_REG rather than gen_lowpart.
9761 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9763 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9764 rather than 0 when creating partial subregs.
9766 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9768 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9769 of usage.
9771 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9773 PR target/81550
9774 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9775 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9776 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9777 flags. This restores the settings used before the 2017-07-24.
9778 Turning off pre increment/decrement/modify allows IVOPTS to
9779 optimize DF/SF loops where the index is an int.
9781 2018-01-29 Richard Biener <rguenther@suse.de>
9782 Kelvin Nilsen <kelvin@gcc.gnu.org>
9784 PR bootstrap/80867
9785 * tree-vect-stmts.c (vectorizable_call): Don't call
9786 targetm.vectorize_builtin_md_vectorized_function if callee is
9787 NULL.
9789 2018-01-22 Carl Love <cel@us.ibm.com>
9791 * doc/extend.tex: Fix typo in second arg in
9792 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9794 2018-01-29 Richard Biener <rguenther@suse.de>
9796 PR tree-optimization/84086
9797 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9798 (flush_ssaname_freelist): When SSA names were released reset
9799 the SCEV hash table.
9801 2018-01-29 Richard Biener <rguenther@suse.de>
9803 PR tree-optimization/84057
9804 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9805 removed paths when removing edges.
9807 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9809 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9810 -mfunction-return=@var{choice}.
9812 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9814 PR diagnostic/84034
9815 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9816 Handle CR like TAB.
9817 (layout::print_source_line): Likewise.
9818 (test_get_line_width_without_trailing_whitespace): Add test cases.
9820 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9822 PR middle-end/84040
9823 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9824 debug insns.
9826 2018-01-26 Jim Wilson <jimw@sifive.com>
9828 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9830 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9831 specified.
9833 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9835 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9836 and CMP + SUB-immediate -> SUBS.
9838 2018-01-26 Martin Sebor <msebor@redhat.com>
9840 PR tree-optimization/83896
9841 * tree-ssa-strlen.c (get_string_len): Rename...
9842 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9843 Avoid assuming length is constant.
9844 (handle_char_store): Use HOST_WIDE_INT for string length.
9846 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9848 PR target/81763
9849 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9850 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9852 2018-01-26 Richard Biener <rguenther@suse.de>
9854 PR rtl-optimization/84003
9855 * dse.c (record_store): Only record redundant stores when
9856 the earlier store aliases at least all accesses the later one does.
9858 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9860 PR rtl-optimization/83985
9861 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9862 REG_CFA_RESTORE insns.
9863 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9864 REG_CFA_RESTORE insns here.
9866 PR c/83989
9867 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9868 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9870 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9872 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9873 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9874 (arc_init): Likewise.
9875 (arc_override_options): Likewise.
9876 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9877 value.
9878 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9879 support.
9880 * config/arc/arc.h (TARGET_DBNZ): Define.
9881 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9882 properly set the tune attribute.
9883 (dbnz): Use TARGET_DBNZ guard.
9884 * config/arc/arc.opt (mtune): Add core3 option.
9886 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9888 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9889 recognize new pic like addresses.
9890 (arc_delegitimize_address): Clean up.
9892 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9894 * config/arc/arc-arches.def: Option mrf16 valid for all
9895 architectures.
9896 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9897 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9898 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9899 * config/arc/arc-tables.opt: Regenerate.
9900 * config/arc/arc.c (arc_conditional_register_usage): Handle
9901 reduced register file case.
9902 (arc_file_start): Set must have build attributes.
9903 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9904 mrf16 option value.
9905 * config/arc/arc.opt (mrf16): Add new option.
9906 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9907 * config/arc/genmultilib.awk: Handle new mrf16 option.
9908 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9909 * config/arc/t-multilib: Regenerate.
9910 * doc/invoke.texi (ARC Options): Document mrf16 option.
9912 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9914 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9915 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9916 (arc_attribute_table): Add 'secure_call' attribute.
9917 (arc_print_operand): Print secure call operand.
9918 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9919 secure.
9920 (arc_is_secure_call_p): New function. * config/arc/arc.md
9921 (call_i): Add support for sjli instruction.
9922 (call_value_i): Likewise.
9923 * config/arc/constraints.md (Csc): New constraint.
9925 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9926 John Eric Martin <John.Martin@emmicro-us.com>
9928 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9929 * config/arc/arc.c (_arc_jli_section): New struct.
9930 (arc_jli_section): New type.
9931 (rc_jli_sections): New static variable.
9932 (arc_handle_jli_attribute): New function.
9933 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9934 (arc_file_end): New function.
9935 (TARGET_ASM_FILE_END): Define.
9936 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9937 (arc_add_jli_section): New function.
9938 (jli_call_scan): Likewise.
9939 (arc_reorg): Call jli_call_scan.
9940 (arc_output_addsi): Remove 'S' from printing asm operand.
9941 (arc_is_jli_call_p): New function.
9942 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9943 operand.
9944 (movhi_insn): Likewise.
9945 (movsi_insn): Likewise.
9946 (movsi_set_cc_insn): Likewise.
9947 (loadqi_update): Likewise.
9948 (load_zeroextendqisi_update): Likewise.
9949 (load_signextendqisi_update): Likewise.
9950 (loadhi_update): Likewise.
9951 (load_zeroextendhisi_update): Likewise.
9952 (load_signextendhisi_update): Likewise.
9953 (loadsi_update): Likewise.
9954 (loadsf_update): Likewise.
9955 (movsicc_insn): Likewise.
9956 (bset_insn): Likewise.
9957 (bxor_insn): Likewise.
9958 (bclr_insn): Likewise.
9959 (bmsk_insn): Likewise.
9960 (bicsi3_insn): Likewise.
9961 (cmpsi_cc_c_insn): Likewise.
9962 (movsi_ne): Likewise.
9963 (movsi_cond_exec): Likewise.
9964 (clrsbsi2): Likewise.
9965 (norm_f): Likewise.
9966 (normw): Likewise.
9967 (swap): Likewise.
9968 (divaw): Likewise.
9969 (flag): Likewise.
9970 (sr): Likewise.
9971 (kflag): Likewise.
9972 (ffs): Likewise.
9973 (ffs_f): Likewise.
9974 (fls): Likewise.
9975 (call_i): Remove 'S' asm letter, add jli instruction.
9976 (call_value_i): Likewise.
9977 * config/arc/arc.op (mjli-always): New option.
9978 * config/arc/constraints.md (Cji): New constraint.
9979 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9980 operand.
9981 (subsf3_fpx): Likewise.
9982 (mulsf3_fpx): Likewise.
9983 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9984 asm operand.
9985 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9986 function attrbutes.
9987 * doc/invoke.texi (ARC): Document mjli-always option.
9989 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9991 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9992 avoid addition with 0 and use incw and decw where possible.
9994 2018-01-26 Richard Biener <rguenther@suse.de>
9996 PR tree-optimization/81082
9997 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9998 association if it requires casting to unsigned.
9999 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10000 from fold_plusminus_mult_expr to catch important cases late when
10001 range info is available.
10003 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10005 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10006 * configure.ac (hidden_linkonce): New test.
10007 * configure: Regenerate.
10008 * config.in: Regenerate.
10010 2018-01-26 Julia Koval <julia.koval@intel.com>
10012 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10013 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10014 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10015 _mm_mask_bitshuffle_epi64_mask): Fix type.
10016 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10017 USI_FTYPE_V4DI_V4DI_USI): Remove.
10018 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10019 __builtin_ia32_vpshufbitqmb256_mask,
10020 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10021 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10022 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10024 2018-01-26 Alan Modra <amodra@gmail.com>
10026 PR target/84033
10027 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10028 UNSPEC_VBPERMQ. Sort other unspecs.
10030 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10032 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10034 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10036 PR middle-end/83055
10037 * predict.c (drop_profile): Do not push/pop cfun; update also
10038 node->count.
10039 (handle_missing_profiles): Fix logic looking for zero profiles.
10041 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10043 PR middle-end/83977
10044 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10045 on functions with #pragma omp declare simd or functions with simd
10046 attribute.
10047 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10048 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10049 Remove trailing \n from warning_at calls.
10051 2018-01-25 Tom de Vries <tom@codesourcery.com>
10053 PR target/84028
10054 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10055 for neutered workers.
10057 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10059 PR target/68467
10060 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10061 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10063 2018-01-24 Jeff Law <law@redhat.com>
10065 PR target/83994
10066 * i386.c (get_probe_interval): Move to earlier point.
10067 (ix86_compute_frame_layout): If -fstack-clash-protection and
10068 the frame is larger than the probe interval, then use pushes
10069 to save registers rather than reg->mem moves.
10070 (ix86_expand_prologue): Remove conditional for int_registers_saved
10071 assertion.
10073 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10075 PR target/84014
10076 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10077 min/max for never referenced object.
10079 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10081 PR middle-end/83977
10082 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10083 here.
10084 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10085 attributes from DECL_ATTRIBUTES (decl) without affecting
10086 DECL_ATTRIBUTES (current_function_decl).
10087 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10088 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10090 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10092 PR tree-optimization/83979
10093 * fold-const.c (fold_comparison): Use constant_boolean_node
10094 instead of boolean_{true,false}_node.
10096 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10098 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10099 with zero counts.
10101 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10103 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10104 Simplify the clause that sets the length attribute.
10105 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10106 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10107 clause that sets the length attribute.
10108 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10110 2018-01-24 Tom de Vries <tom@codesourcery.com>
10112 PR target/83589
10113 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10114 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10115 Add strict parameter.
10116 (prevent_branch_around_nothing): Insert dummy insn between branch to
10117 label and label with no ptx insn inbetween.
10118 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10120 2018-01-24 Tom de Vries <tom@codesourcery.com>
10122 PR target/81352
10123 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10124 for neutered threads in warp.
10125 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10127 2018-01-24 Richard Biener <rguenther@suse.de>
10129 PR tree-optimization/83176
10130 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10131 operands.
10133 2018-01-24 Richard Biener <rguenther@suse.de>
10135 PR tree-optimization/82819
10136 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10137 code generating pluses that are no-ops in the target precision.
10139 2018-01-24 Richard Biener <rguenther@suse.de>
10141 PR middle-end/84000
10142 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10144 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10146 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10147 to merge probabilities.
10148 * predict.c (probably_never_executed): Also mark as cold functions
10149 with global 0 profile and guessed local profile.
10150 * profile-count.c (profile_probability::combine_with_count): New
10151 member function.
10152 * profile-count.h (profile_probability::operator*,
10153 profile_probability::operator*=, profile_probability::operator/,
10154 profile_probability::operator/=): Reduce precision to adjusted
10155 and set value to guessed on contradictory divisions.
10156 (profile_probability::combine_with_freq): Remove.
10157 (profile_probability::combine_wiht_count): Declare.
10158 (profile_count::force_nonzero):: Set to adjusted.
10159 (profile_count::probability_in):: Set quality to adjusted.
10160 * tree-ssa-tail-merge.c (replace_block_by): Use
10161 combine_with_count.
10163 2018-01-23 Andrew Waterman <andrew@sifive.com>
10164 Jim Wilson <jimw@sifive.com>
10166 * config/riscv/riscv.c (riscv_stack_boundary): New.
10167 (riscv_option_override): Set riscv_stack_boundary. Handle
10168 riscv_preferred_stack_boundary_arg.
10169 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10170 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10171 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10172 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10173 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10174 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10176 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10178 PR target/83905
10179 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10180 of struct ix86_frame.
10181 (ix86_expand_epilogue): Likewise. Add a local variable for
10182 the reg_save_offset field in struct ix86_frame.
10184 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10186 PR tree-optimization/82604
10187 * tree-loop-distribution.c (enum partition_kind): New enum item
10188 PKIND_PARTIAL_MEMSET.
10189 (partition_builtin_p): Support above new enum item.
10190 (generate_code_for_partition): Ditto.
10191 (compute_access_range): Differentiate cases that equality can be
10192 proven at all loops, the innermost loops or no loops.
10193 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10194 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10195 (finalize_partitions, distribute_loop): Don't fuse partition of
10196 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10197 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10198 parloop is enabled.
10200 2018-01-23 Martin Liska <mliska@suse.cz>
10202 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10203 order to ignore the predictor.
10204 (PRED_POLYMORPHIC_CALL): Likewise.
10205 (PRED_RECURSIVE_CALL): Likewise.
10207 2018-01-23 Martin Liska <mliska@suse.cz>
10209 * tree-profile.c (tree_profiling): Print function header to
10210 aware reader which function we are working on.
10211 * value-prof.c (gimple_find_values_to_profile): Do not print
10212 not interesting value histograms.
10214 2018-01-23 Martin Liska <mliska@suse.cz>
10216 * profile-count.h (enum profile_quality): Add
10217 profile_uninitialized as the first value. Do not number values
10218 as they are zero based.
10219 (profile_count::verify): Update sanity check.
10220 (profile_probability::verify): Likewise.
10222 2018-01-23 Nathan Sidwell <nathan@acm.org>
10224 * doc/invoke.texi (ffor-scope): Deprecate.
10226 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10228 PR tree-optimization/83510
10229 * domwalk.c (set_all_edges_as_executable): New function.
10230 (dom_walker::dom_walker): Convert bool param
10231 "skip_unreachable_blocks" to enum reachability. Move setup of
10232 edge flags to set_all_edges_as_executable and only do it when
10233 reachability is REACHABLE_BLOCKS.
10234 * domwalk.h (enum dom_walker::reachability): New enum.
10235 (dom_walker::dom_walker): Convert bool param
10236 "skip_unreachable_blocks" to enum reachability.
10237 (set_all_edges_as_executable): New decl.
10238 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10239 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10240 "reachability".
10241 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10242 but converting true to REACHABLE_BLOCKS.
10243 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10244 * tree-vrp.c
10245 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10246 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10247 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10248 REACHABLE_BLOCKS.
10249 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10250 if check_all_array_refs will be called.
10252 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10254 * tree.c (selftest::test_location_wrappers): Add more test
10255 coverage.
10257 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10259 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10260 (selftest::test_bit_in_range): Likewise.
10262 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10264 PR testsuite/83888
10265 * doc/sourcebuild.texi (vect_float): Say that the selector
10266 only describes the situation when -funsafe-math-optimizations is on.
10267 (vect_float_strict): Document.
10269 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10271 PR tree-optimization/83965
10272 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10273 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10274 instead of checking only for a reduction.
10275 (vect_recog_widen_sum_pattern): Likewise.
10277 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10279 * predict.c (probably_never_executed): Only use precise profile info.
10280 (compute_function_frequency): Skip after inlining hack since we now
10281 have quality checking.
10283 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10285 * profile-count.h (profile_probability::very_unlikely,
10286 profile_probability::unlikely, profile_probability::even): Set
10287 precision to guessed.
10289 2018-01-23 Richard Biener <rguenther@suse.de>
10291 PR tree-optimization/83963
10292 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10293 Properly terminate dominator walk when crossing the exit edge not
10294 when visiting its source block.
10296 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10298 PR c++/83918
10299 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10300 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10302 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10304 PR tree-optimization/83957
10305 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10306 semicolon after for body surrounded by braces.
10308 PR tree-optimization/83081
10309 * profile-count.h (profile_probability::split): New method.
10310 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10311 Use profile_probability::split.
10312 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10313 when splitting a single conditional jump into 2.
10315 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10317 PR tree-optimization/69452
10318 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10319 decl.
10321 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10323 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10324 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10325 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10327 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10329 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10330 declaration.
10331 * config/rl78/rl78.md (movdi): New define_expand.
10332 * config/rl78/rl78.c (rl78_split_movdi): New function.
10334 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10336 PR target/83862
10337 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10338 no longer used.
10339 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10340 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10341 128-bit to produce an UNSPEC move to get the double word with the
10342 signbit and then a shift directly to do signbit.
10343 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10344 implementation with a new version that just does either a direct
10345 move or a regular move. Move memory interface to separate insns.
10346 Move insns so they are next to the expander.
10347 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10348 with signbit move. Split big and little endian case.
10349 (signbit<mode>2_dm_mem_le): Likewise.
10350 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10351 (signbit<mode>2_dm2): Likewise.
10353 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10355 * config/rl78/rl78.md (anddi3): New define_expand.
10357 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10359 * config/rl78/rl78.md (umindi3): New define_expand.
10361 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10363 * config/rl78/rl78.md (smindi3): New define_expand.
10365 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10367 * config/rl78/rl78.md (smaxdi3): New define_expand.
10369 2018-01-22 Carl Love <cel@us.ibm.com>
10371 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10372 LVX_V1TI): Add macro expansion.
10373 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10374 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10375 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10376 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10377 Change check to determine if the instruction is a byte reversing
10378 entry. Fix typo in comment.
10379 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10380 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10381 Add def_builtin calls for new builtins.
10382 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10383 Add define_insn expansion.
10385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10387 * config/rl78/rl78.md (umaxdi3): New define_expand.
10389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10391 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10392 for non-QImode registers.
10394 2018-01-22 Richard Biener <rguenther@suse.de>
10396 PR tree-optimization/83963
10397 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10398 including the loop exit block.
10399 (scop_detection::merge_sese): Likewise.
10400 (scop_detection::add_scop): Do it here instead.
10402 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10404 * doc/sourcebuild.texi (arm_softfloat): Document.
10406 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10408 PR gcc/77734
10409 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10410 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10411 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10413 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10414 David Edelsohn <dje.gcc@gmail.com>
10416 PR target/83946
10417 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10418 Change "crset eq" to "crset 2".
10419 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10420 (*call_indirect_aix<mode>_nospec): Likewise.
10421 (*call_value_indirect_aix<mode>_nospec): Likewise.
10422 (*call_indirect_elfv2<mode>_nospec): Likewise.
10423 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10424 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10425 change assembly output from . to $.
10426 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10427 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10428 (*tablejump<mode>_internal1_nospec): Likewise.
10430 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10432 PR target/80870
10433 * config/sh/sh_optimize_sett_clrt.cc:
10434 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10436 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10438 PR tree-optimization/83940
10439 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10440 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10441 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10442 use it to pass back the definition type.
10443 (vect_check_store_rhs): Likewise rhs_dt_out.
10444 (vect_build_gather_load_calls): Add a mask_dt argument and use
10445 it instead of a call to vect_is_simple_use.
10446 (vectorizable_store): Update calls to vect_check_load_store_mask
10447 and vect_check_store_rhs. Use the dt returned by the latter instead
10448 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10449 instead of calls to vect_is_simple_use. Pass the scalar rather
10450 than the vector operand to vect_is_simple_use when handling
10451 second and subsequent copies of an rhs value.
10452 (vectorizable_load): Update calls to vect_check_load_store_mask
10453 and vect_build_gather_load_calls. Use the cached mask_dt and
10454 gs_info.offset_dt instead of calls to vect_is_simple_use.
10456 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10458 PR middle-end/83945
10459 * tree-emutls.c: Include gimplify.h.
10460 (lower_emutls_2): New function.
10461 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10462 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10463 it before further processing.
10465 PR target/83930
10466 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10467 UINTVAL (trueop1) instead of INTVAL (op1).
10469 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10471 PR debug/81570
10472 PR debug/83728
10473 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10474 INCOMING_FRAME_SP_OFFSET if not defined.
10475 (scan_trace): Add ENTRY argument. If true and
10476 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10477 emit a note to adjust the CFA offset.
10478 (create_cfi_notes): Adjust scan_trace callers.
10479 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10480 INCOMING_FRAME_SP_OFFSET in the CIE.
10481 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10482 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10483 Likewise.
10484 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10485 * doc/tm.texi: Regenerated.
10487 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10489 PR rtl-optimization/83147
10490 * lra-constraints.c (remove_inheritance_pseudos): Use
10491 lra_substitute_pseudo_within_insn.
10493 2018-01-19 Tom de Vries <tom@codesourcery.com>
10494 Cesar Philippidis <cesar@codesourcery.com>
10496 PR target/83920
10497 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10499 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10501 PR target/83790
10502 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10503 spaces for function labels.
10505 2018-01-19 Martin Liska <mliska@suse.cz>
10507 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10508 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10509 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10510 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10511 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10512 (PRED_CONST_RETURN): Change from 69 to 65.
10513 (PRED_NULL_RETURN): Change from 91 to 71.
10514 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10515 (PRED_LOOP_GUARD): Change from 66 to 73.
10517 2018-01-19 Martin Liska <mliska@suse.cz>
10519 * predict.c (predict_insn_def): Add new assert.
10520 (struct branch_predictor): Change type to signed integer.
10521 (test_prediction_value_range): Amend test to cover
10522 PROB_UNINITIALIZED.
10523 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10524 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10525 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10526 (PRED_LOOP_IV_COMPARE): Likewise.
10527 * predict.h (PROB_UNINITIALIZED): Define new constant.
10529 2018-01-19 Martin Liska <mliska@suse.cz>
10531 * predict.c (dump_prediction): Add new format for
10532 analyze_brprob.py script which is enabled with -details
10533 suboption.
10534 * profile-count.h (precise_p): New function.
10536 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10538 PR tree-optimization/83922
10539 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10540 there are no statements that need masking.
10541 (vect_active_double_reduction_p): New function.
10542 (vect_analyze_loop_operations): Use it when handling phis that
10543 are not in the loop header.
10545 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10547 PR tree-optimization/83914
10548 * tree-vect-loop.c (vectorizable_induction): Don't convert
10549 init_expr or apply the peeling adjustment for inductions
10550 that are nested within the vectorized loop.
10552 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10554 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10555 instead of NEG.
10557 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10559 PR sanitizer/81715
10560 PR testsuite/83882
10561 * function.h (gimplify_parameters): Add gimple_seq * argument.
10562 * function.c: Include gimple.h and options.h.
10563 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10564 for the added local temporaries if needed.
10565 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10566 if there are any parameter cleanups, wrap whole body into a
10567 try/finally with the cleanups.
10569 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10571 PR target/82964
10572 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10573 Use GET_MODE_CLASS for scalar floating point.
10575 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10577 PR ipa/82256
10578 patch by PaX Team
10579 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10580 Fix call of call_cgraph_insertion_hooks.
10582 2018-01-18 Martin Sebor <msebor@redhat.com>
10584 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10586 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10588 PR ipa/83619
10589 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10590 frequencies.
10592 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10594 PR other/70268
10595 * common.opt: (-ffile-prefix-map): New option.
10596 * opts.c (common_handle_option): Defer it.
10597 * opts-global.c (handle_common_deferred_options): Handle it.
10598 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10599 * file-prefix-map.h: New file.
10600 (remap_debug_filename, add_debug_prefix_map): ...here.
10601 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10602 * final.c (debug_prefix_map, add_debug_prefix_map
10603 remap_debug_filename): Move to...
10604 * file-prefix-map.c: New file.
10605 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10606 generalize, get rid of alloca(), use strrchr() instead of strchr().
10607 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10608 Implement in terms of add_prefix_map().
10609 (remap_macro_filename, remap_debug_filename): Implement in term of
10610 remap_filename().
10611 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10612 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10613 * dbxout.c: Include file-prefix-map.h.
10614 * varasm.c: Likewise.
10615 * vmsdbgout.c: Likewise.
10616 * xcoffout.c: Likewise.
10617 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10618 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10619 * doc/invoke.texi (-ffile-prefix-map): Document.
10620 (-fdebug-prefix-map): Update description.
10622 2018-01-18 Martin Liska <mliska@suse.cz>
10624 * config/i386/i386.c (indirect_thunk_name): Document that also
10625 lfence is emitted.
10626 (output_indirect_thunk): Document why both instructions
10627 (pause and lfence) are generated.
10629 2018-01-18 Richard Biener <rguenther@suse.de>
10631 PR tree-optimization/83887
10632 * graphite-scop-detection.c
10633 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10634 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10635 (scop_detection::merge_sese): Re-implement with a flood-fill
10636 algorithm that properly finds a SESE region if it exists.
10638 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10640 PR c/61240
10641 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10642 pointer_diff optimizations use view_convert instead of convert.
10644 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10646 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10647 Generate different code for -mno-speculate-indirect-jumps.
10648 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10649 (*call_indirect_aix<mode>): Disable for
10650 -mno-speculate-indirect-jumps.
10651 (*call_indirect_aix<mode>_nospec): New define_insn.
10652 (*call_value_indirect_aix<mode>): Disable for
10653 -mno-speculate-indirect-jumps.
10654 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10655 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10656 -mno-speculate-indirect-jumps.
10657 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10659 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10661 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10662 long double type, set the flags for noting the default long double
10663 type, even if we don't pass or return a long double type.
10665 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10667 PR ipa/83051
10668 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10669 failure.
10671 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10673 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10674 support for merge[hl].
10675 (fold_mergehl_helper): New helper function.
10676 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10677 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10678 (altivec_vmrglw_direct): Add xxmrglw insn.
10680 2018-01-17 Andrew Waterman <andrew@sifive.com>
10682 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10683 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10685 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10687 PR lto/83121
10688 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10689 call the lto_location_cache before reading the
10690 DECL_SOURCE_LOCATION of the types.
10692 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10693 Richard Sandiford <richard.sandiford@linaro.org>
10695 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10696 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10697 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10698 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10699 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10700 Add declaration.
10701 * config/aarch64/constraints.md (aarch64_movti_operand):
10702 Limit immediates.
10703 * config/aarch64/predicates.md (Uti): Add new constraint.
10705 2018-01-17 Carl Love <cel@us.ibm.com>
10707 * config/rs6000/vsx.md (define_expand xl_len_r,
10708 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10709 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10710 lxvll.
10711 (define_expand, define_insn): Move the shift left from the
10712 define_insn to the define_expand for lxvl and stxvl instructions.
10713 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10714 and XL_LEN_R definitions to PURE.
10716 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10718 * config/i386/i386.c (indirect_thunk_name): Declare regno
10719 as unsigned int. Compare regno with INVALID_REGNUM.
10720 (output_indirect_thunk): Ditto.
10721 (output_indirect_thunk_function): Ditto.
10722 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10723 in the call to output_indirect_thunk_function.
10725 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10727 PR middle-end/83884
10728 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10729 rather than the size of inner_type to determine the stack slot size
10730 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10732 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10734 PR target/83546
10735 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10736 to PTA_SILVERMONT.
10738 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10740 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10741 endian Linux systems to optionally enable multilibs for selecting
10742 the long double type if the user configured an explicit type.
10743 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10744 have no long double multilibs if not defined.
10745 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10746 warn if the user used -mabi={ieee,ibm}longdouble and we built
10747 multilibs for long double.
10748 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10749 appropriate multilib option.
10750 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10751 multilib options.
10752 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10753 for building long double multilibs.
10754 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10756 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10758 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10759 copies.
10761 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10762 64 bits.
10763 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10764 128 bits.
10766 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10767 variables.
10769 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10770 return value.
10772 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10774 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10775 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10777 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10779 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10780 different rtl trees depending on TARGET_64BIT.
10781 (rs6000_gen_lvx): Likewise.
10783 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10785 * config/visium/visium.md (nop): Tweak comment.
10786 (hazard_nop): Likewise.
10788 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10790 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10791 -mspeculate-indirect-jumps.
10792 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10793 for -mno-speculate-indirect-jumps.
10794 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10795 (*call_value_indirect_elfv2<mode>): Disable for
10796 -mno-speculate-indirect-jumps.
10797 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10798 (indirect_jump): Emit different RTL for
10799 -mno-speculate-indirect-jumps.
10800 (*indirect_jump<mode>): Disable for
10801 -mno-speculate-indirect-jumps.
10802 (*indirect_jump<mode>_nospec): New define_insn.
10803 (tablejump): Emit different RTL for
10804 -mno-speculate-indirect-jumps.
10805 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10806 (tablejumpsi_nospec): New define_expand.
10807 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10808 (tablejumpdi_nospec): New define_expand.
10809 (*tablejump<mode>_internal1): Disable for
10810 -mno-speculate-indirect-jumps.
10811 (*tablejump<mode>_internal1_nospec): New define_insn.
10812 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10813 option.
10815 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10817 * caller-save.c (insert_save): Drop unnecessary parameter. All
10818 callers updated.
10820 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10821 Richard Biener <rguenth@suse.de>
10823 PR libgomp/83590
10824 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10825 return early, inline manually is_gimple_sizepos. Make sure if we
10826 call gimplify_expr we don't end up with a gimple constant.
10827 * tree.c (variably_modified_type_p): Don't return true for
10828 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10829 * gimplify.h (is_gimple_sizepos): Remove.
10831 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10833 PR tree-optimization/83857
10834 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10835 vectorizable_live_operation for pure SLP statements.
10836 (vectorizable_live_operation): Handle PHIs.
10838 2018-01-16 Richard Biener <rguenther@suse.de>
10840 PR tree-optimization/83867
10841 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10842 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10844 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10846 PR c/83844
10847 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10848 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10849 If off is not INTEGER_CST, issue a may not be aligned warning
10850 rather than isn't aligned. Use isn%'t rather than isn't.
10851 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10852 into MULT_EXPR.
10853 <case MULT_EXPR>: Improve the case when bottom and one of the
10854 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10855 operand, in that case check if the other operand is multiple of
10856 bottom divided by the INTEGER_CST operand.
10858 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10860 PR target/83858
10861 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10862 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10863 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10864 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10865 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10866 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10867 (pa_function_arg_size): New function.
10869 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10871 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10872 in a separate statement.
10874 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10876 PR tree-optimization/83847
10877 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10878 group gathers and scatters.
10880 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10882 PR rtl-optimization/86620
10883 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10885 PR rtl-optimization/83213
10886 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10887 to last if both are JUMP_INSNs.
10889 PR tree-optimization/83843
10890 * gimple-ssa-store-merging.c
10891 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10892 store_immediate_info for bswap/nop orig_stores.
10894 2018-01-15 Andrew Waterman <andrew@sifive.com>
10896 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10897 !TARGET_MUL.
10898 <UDIV>: Increase cost if !TARGET_DIV.
10900 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10902 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10903 (define_attr "cr_logical_3op"): New.
10904 (cceq_ior_compare): Adjust.
10905 (cceq_ior_compare_complement): Adjust.
10906 (*cceq_rev_compare): Adjust.
10907 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10908 (is_cracked_insn): Adjust.
10909 (insn_must_be_first_in_group): Adjust.
10910 * config/rs6000/40x.md: Adjust.
10911 * config/rs6000/440.md: Adjust.
10912 * config/rs6000/476.md: Adjust.
10913 * config/rs6000/601.md: Adjust.
10914 * config/rs6000/603.md: Adjust.
10915 * config/rs6000/6xx.md: Adjust.
10916 * config/rs6000/7450.md: Adjust.
10917 * config/rs6000/7xx.md: Adjust.
10918 * config/rs6000/8540.md: Adjust.
10919 * config/rs6000/cell.md: Adjust.
10920 * config/rs6000/e300c2c3.md: Adjust.
10921 * config/rs6000/e500mc.md: Adjust.
10922 * config/rs6000/e500mc64.md: Adjust.
10923 * config/rs6000/e5500.md: Adjust.
10924 * config/rs6000/e6500.md: Adjust.
10925 * config/rs6000/mpc.md: Adjust.
10926 * config/rs6000/power4.md: Adjust.
10927 * config/rs6000/power5.md: Adjust.
10928 * config/rs6000/power6.md: Adjust.
10929 * config/rs6000/power7.md: Adjust.
10930 * config/rs6000/power8.md: Adjust.
10931 * config/rs6000/power9.md: Adjust.
10932 * config/rs6000/rs64.md: Adjust.
10933 * config/rs6000/titan.md: Adjust.
10935 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10937 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10938 ix86_indirect_branch_register logic.
10940 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10942 * config/i386/constraints.md (Bs): Update
10943 ix86_indirect_branch_register check. Don't check
10944 ix86_indirect_branch_register with GOT_memory_operand.
10945 (Bw): Likewise.
10946 * config/i386/predicates.md (GOT_memory_operand): Don't check
10947 ix86_indirect_branch_register here.
10948 (GOT32_symbol_operand): Likewise.
10950 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10952 * config/i386/predicates.md (constant_call_address_operand):
10953 Rewrite ix86_indirect_branch_register logic.
10954 (sibcall_insn_operand): Likewise.
10956 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10958 * config/i386/constraints.md (Bs): Replace
10959 ix86_indirect_branch_thunk_register with
10960 ix86_indirect_branch_register.
10961 (Bw): Likewise.
10962 * config/i386/i386.md (indirect_jump): Likewise.
10963 (tablejump): Likewise.
10964 (*sibcall_memory): Likewise.
10965 (*sibcall_value_memory): Likewise.
10966 Peepholes of indirect call and jump via memory: Likewise.
10967 * config/i386/i386.opt: Likewise.
10968 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10969 (GOT_memory_operand): Likewise.
10970 (call_insn_operand): Likewise.
10971 (sibcall_insn_operand): Likewise.
10972 (GOT32_symbol_operand): Likewise.
10974 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10976 PR middle-end/83837
10977 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10978 type rather than type addr's type points to.
10979 (expand_omp_atomic_mutex): Likewise.
10980 (expand_omp_atomic): Likewise.
10982 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10984 PR target/83839
10985 * config/i386/i386.c (output_indirect_thunk_function): Use
10986 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10987 for __x86_return_thunk.
10989 2018-01-15 Richard Biener <rguenther@suse.de>
10991 PR middle-end/83850
10992 * expmed.c (extract_bit_field_1): Fix typo.
10994 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10996 PR target/83687
10997 * config/arm/iterators.md (VF): New mode iterator.
10998 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10999 Remove integer-related logic from pattern.
11000 (neon_vabd<mode>_3): Likewise.
11002 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11004 PR middle-end/82694
11005 * common.opt (fstrict-overflow): No longer an alias.
11006 (fwrapv-pointer): New option.
11007 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11008 also for pointer types based on flag_wrapv_pointer.
11009 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11010 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11011 opts->x_flag_wrapv got set.
11012 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11013 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11014 POINTER_TYPE_OVERFLOW_UNDEFINED.
11015 * match.pd: Likewise in address comparison pattern.
11016 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11018 2018-01-15 Richard Biener <rguenther@suse.de>
11020 PR lto/83804
11021 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11022 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11023 Reset type names to their identifier if their TYPE_DECL doesn't
11024 have linkage (and thus is used for ODR and devirt).
11025 (save_debug_info_for_decl): Remove.
11026 (save_debug_info_for_type): Likewise.
11027 (add_tree_to_fld_list): Adjust.
11028 * tree-pretty-print.c (dump_generic_node): Make dumping of
11029 type names more robust.
11031 2018-01-15 Richard Biener <rguenther@suse.de>
11033 * BASE-VER: Bump to 8.0.1.
11035 2018-01-14 Martin Sebor <msebor@redhat.com>
11037 PR other/83508
11038 * builtins.c (check_access): Avoid warning when the no-warning bit
11039 is set.
11041 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11043 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11044 * ira-color (allocno_hard_regs_compare): Likewise.
11046 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11048 PR target/83013
11049 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11050 Use .pushsection/.popsection.
11052 2018-01-14 Martin Sebor <msebor@redhat.com>
11054 PR c++/81327
11055 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11057 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11059 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11060 entry from extra_headers.
11061 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11062 extra_headers, make the list bitwise identical to the i?86-*-* one.
11064 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11066 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11067 -mcmodel=large with -mindirect-branch=thunk,
11068 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11069 -mfunction-return=thunk-extern.
11070 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11071 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11072 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11074 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11076 * config/i386/i386.c (print_reg): Print the name of the full
11077 integer register without '%'.
11078 (ix86_print_operand): Handle 'V'.
11079 * doc/extend.texi: Document 'V' modifier.
11081 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11083 * config/i386/constraints.md (Bs): Disallow memory operand for
11084 -mindirect-branch-register.
11085 (Bw): Likewise.
11086 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11087 (GOT_memory_operand): Likewise.
11088 (call_insn_operand): Likewise.
11089 (sibcall_insn_operand): Likewise.
11090 (GOT32_symbol_operand): Likewise.
11091 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11092 for -mindirect-branch-register.
11093 (tablejump): Likewise.
11094 (*sibcall_memory): Likewise.
11095 (*sibcall_value_memory): Likewise.
11096 Disallow peepholes of indirect call and jump via memory for
11097 -mindirect-branch-register.
11098 (*call_pop): Replace m with Bw.
11099 (*call_value_pop): Likewise.
11100 (*sibcall_pop_memory): Replace m with Bs.
11101 * config/i386/i386.opt (mindirect-branch-register): New option.
11102 * doc/invoke.texi: Document -mindirect-branch-register option.
11104 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11106 * config/i386/i386-protos.h (ix86_output_function_return): New.
11107 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11108 set function_return_type.
11109 (indirect_thunk_name): Add ret_p to indicate thunk for function
11110 return.
11111 (output_indirect_thunk_function): Pass false to
11112 indirect_thunk_name.
11113 (ix86_output_indirect_branch_via_reg): Likewise.
11114 (ix86_output_indirect_branch_via_push): Likewise.
11115 (output_indirect_thunk_function): Create alias for function
11116 return thunk if regno < 0.
11117 (ix86_output_function_return): New function.
11118 (ix86_handle_fndecl_attribute): Handle function_return.
11119 (ix86_attribute_table): Add function_return.
11120 * config/i386/i386.h (machine_function): Add
11121 function_return_type.
11122 * config/i386/i386.md (simple_return_internal): Use
11123 ix86_output_function_return.
11124 (simple_return_internal_long): Likewise.
11125 * config/i386/i386.opt (mfunction-return=): New option.
11126 (indirect_branch): Mention -mfunction-return=.
11127 * doc/extend.texi: Document function_return function attribute.
11128 * doc/invoke.texi: Document -mfunction-return= option.
11130 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11132 * config/i386/i386-opts.h (indirect_branch): New.
11133 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11134 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11135 with local indirect jump when converting indirect call and jump.
11136 (ix86_set_indirect_branch_type): New.
11137 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11138 (indirectlabelno): New.
11139 (indirect_thunk_needed): Likewise.
11140 (indirect_thunk_bnd_needed): Likewise.
11141 (indirect_thunks_used): Likewise.
11142 (indirect_thunks_bnd_used): Likewise.
11143 (INDIRECT_LABEL): Likewise.
11144 (indirect_thunk_name): Likewise.
11145 (output_indirect_thunk): Likewise.
11146 (output_indirect_thunk_function): Likewise.
11147 (ix86_output_indirect_branch_via_reg): Likewise.
11148 (ix86_output_indirect_branch_via_push): Likewise.
11149 (ix86_output_indirect_branch): Likewise.
11150 (ix86_output_indirect_jmp): Likewise.
11151 (ix86_code_end): Call output_indirect_thunk_function if needed.
11152 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11153 needed.
11154 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11155 (ix86_attribute_table): Add indirect_branch.
11156 * config/i386/i386.h (machine_function): Add indirect_branch_type
11157 and has_local_indirect_jump.
11158 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11159 to true.
11160 (tablejump): Likewise.
11161 (*indirect_jump): Use ix86_output_indirect_jmp.
11162 (*tablejump_1): Likewise.
11163 (simple_return_indirect_internal): Likewise.
11164 * config/i386/i386.opt (mindirect-branch=): New option.
11165 (indirect_branch): New.
11166 (keep): Likewise.
11167 (thunk): Likewise.
11168 (thunk-inline): Likewise.
11169 (thunk-extern): Likewise.
11170 * doc/extend.texi: Document indirect_branch function attribute.
11171 * doc/invoke.texi: Document -mindirect-branch= option.
11173 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11175 PR ipa/83051
11176 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11178 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11180 * ipa-inline.c (want_inline_small_function_p): Return false if
11181 inlining has already failed with CIF_FINAL_ERROR.
11182 (update_caller_keys): Call want_inline_small_function_p before
11183 can_inline_edge_p.
11184 (update_callee_keys): Likewise.
11186 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11188 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11189 New function.
11190 (rs6000_quadword_masked_address_p): Likewise.
11191 (quad_aligned_load_p): Likewise.
11192 (quad_aligned_store_p): Likewise.
11193 (const_load_sequence_p): Add comment to describe the outer-most loop.
11194 (mimic_memory_attributes_and_flags): New function.
11195 (rs6000_gen_stvx): Likewise.
11196 (replace_swapped_aligned_store): Likewise.
11197 (rs6000_gen_lvx): Likewise.
11198 (replace_swapped_aligned_load): Likewise.
11199 (replace_swapped_load_constant): Capitalize argument name in
11200 comment describing this function.
11201 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11202 and stores that access quad-word aligned addresses and replace
11203 with stvx or lvx instructions when appropriate.
11204 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11205 New function prototype.
11206 (rs6000_quadword_masked_address_p): Likewise.
11207 (rs6000_gen_lvx): Likewise.
11208 (rs6000_gen_stvx): Likewise.
11209 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11210 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11211 when memory address is aligned.
11212 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11213 this split to select lvx instruction when memory address is aligned.
11214 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11215 instruction when memory address is aligned.
11216 (*vsx_le_perm_load_v16qi): Likewise.
11217 (four unnamed splitters): Modify to select the stvx instruction
11218 when memory is aligned.
11220 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11222 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11223 which appears in the queue multiple times.
11225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11226 Alan Hayward <alan.hayward@arm.com>
11227 David Sherwood <david.sherwood@arm.com>
11229 * tree-vectorizer.h (vec_lower_bound): New structure.
11230 (_loop_vec_info): Add check_nonzero and lower_bounds.
11231 (LOOP_VINFO_CHECK_NONZERO): New macro.
11232 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11233 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11234 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11235 fields. Make seg_len the distance travelled, not including the
11236 access size.
11237 (dr_direction_indicator): Declare.
11238 (dr_zero_step_indicator): Likewise.
11239 (dr_known_forward_stride_p): Likewise.
11240 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11241 tree-ssanames.h.
11242 (runtime_alias_check_p): Allow runtime alias checks with
11243 variable strides.
11244 (operator ==): Compare access_size and align.
11245 (prune_runtime_alias_test_list): Rework for new distinction between
11246 the access_size and seg_len.
11247 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11248 segment lengths.
11249 (get_segment_min_max): New function.
11250 (create_intersect_range_checks): Use it.
11251 (dr_step_indicator): New function.
11252 (dr_direction_indicator): Likewise.
11253 (dr_zero_step_indicator): Likewise.
11254 (dr_known_forward_stride_p): Likewise.
11255 * tree-loop-distribution.c (data_ref_segment_size): Return
11256 DR_STEP * (niters - 1).
11257 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11258 constructor.
11259 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11260 (vect_preserves_scalar_order_p): New function, split out from...
11261 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11262 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11263 (vect_vfa_access_size): New function.
11264 (vect_vfa_align): Likewise.
11265 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11266 (dump_lower_bound): New function.
11267 (vect_check_lower_bound): Likewise.
11268 (vect_small_gap_p): Likewise.
11269 (vectorizable_with_step_bound_p): Likewise.
11270 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11271 depencies if the vectorization factor is 1. Convert the checks
11272 for nonzero steps into checks on the bounds of DR_STEP. Try using
11273 a bunds check for variable steps if the minimum required step is
11274 relatively small. Update calls to the dr_with_seg_len
11275 constructor and to vect_compile_time_alias.
11276 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11277 function.
11278 (vect_loop_versioning): Call it.
11279 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11280 when retrying.
11281 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11284 Alan Hayward <alan.hayward@arm.com>
11285 David Sherwood <david.sherwood@arm.com>
11287 * doc/sourcebuild.texi (vect_scatter_store): Document.
11288 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11289 optabs.
11290 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11291 Document.
11292 * genopinit.c (main): Add supports_vec_scatter_store and
11293 supports_vec_scatter_store_cached to target_optabs.
11294 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11295 IFN_MASK_SCATTER_STORE.
11296 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11297 functions.
11298 * internal-fn.h (internal_store_fn_p): Declare.
11299 (internal_fn_stored_value_index): Likewise.
11300 * internal-fn.c (scatter_store_direct): New macro.
11301 (expand_scatter_store_optab_fn): New function.
11302 (direct_scatter_store_optab_supported_p): New macro.
11303 (internal_store_fn_p): New function.
11304 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11305 IFN_MASK_SCATTER_STORE.
11306 (internal_fn_mask_index): Likewise.
11307 (internal_fn_stored_value_index): New function.
11308 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11309 for scatter stores.
11310 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11311 * optabs-query.c (supports_vec_scatter_store_p): New function.
11312 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11313 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11314 true for scatter stores.
11315 (vect_gather_scatter_fn_p): Handle scatter stores too.
11316 (vect_check_gather_scatter): Consider using scatter stores if
11317 supports_vec_scatter_store_p.
11318 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11319 scatter stores too.
11320 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11321 internal_fn_stored_value_index.
11322 (check_load_store_masking): Handle scatter stores too.
11323 (vect_get_store_rhs): Make public.
11324 (vectorizable_call): Use internal_store_fn_p.
11325 (vectorizable_store): Handle scatter store internal functions.
11326 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11327 when deciding whether the end of the group has been reached.
11328 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11329 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11330 (mask_scatter_store<mode>): New insns.
11332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11333 Alan Hayward <alan.hayward@arm.com>
11334 David Sherwood <david.sherwood@arm.com>
11336 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11337 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11338 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11339 function.
11340 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11341 Use vect_truncate_gather_scatter_offset if we can't treat the
11342 operation as a normal gather load or scatter store.
11343 (get_group_load_store_type): Take the gather_scatter_info
11344 as argument. Try using a gather load or scatter store for
11345 single-element groups.
11346 (get_load_store_type): Update calls to get_group_load_store_type
11347 and vect_use_strided_gather_scatters_p.
11349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11350 Alan Hayward <alan.hayward@arm.com>
11351 David Sherwood <david.sherwood@arm.com>
11353 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11354 optional tree argument.
11355 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11356 null target hooks.
11357 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11358 but continue to use the current value as a fallback.
11359 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11360 to compare the updates.
11361 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11362 (get_load_store_type): Use it when handling a strided access.
11363 (vect_get_strided_load_store_ops): New function.
11364 (vect_get_data_ptr_increment): Likewise.
11365 (vectorizable_load): Handle strided gather loads. Always pass
11366 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11369 Alan Hayward <alan.hayward@arm.com>
11370 David Sherwood <david.sherwood@arm.com>
11372 * doc/md.texi (gather_load@var{m}): Document.
11373 (mask_gather_load@var{m}): Likewise.
11374 * genopinit.c (main): Add supports_vec_gather_load and
11375 supports_vec_gather_load_cached to target_optabs.
11376 * optabs-tree.c (init_tree_optimization_optabs): Use
11377 ggc_cleared_alloc to allocate target_optabs.
11378 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11379 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11380 functions.
11381 * internal-fn.h (internal_load_fn_p): Declare.
11382 (internal_gather_scatter_fn_p): Likewise.
11383 (internal_fn_mask_index): Likewise.
11384 (internal_gather_scatter_fn_supported_p): Likewise.
11385 * internal-fn.c (gather_load_direct): New macro.
11386 (expand_gather_load_optab_fn): New function.
11387 (direct_gather_load_optab_supported_p): New macro.
11388 (direct_internal_fn_optab): New function.
11389 (internal_load_fn_p): Likewise.
11390 (internal_gather_scatter_fn_p): Likewise.
11391 (internal_fn_mask_index): Likewise.
11392 (internal_gather_scatter_fn_supported_p): Likewise.
11393 * optabs-query.c (supports_at_least_one_mode_p): New function.
11394 (supports_vec_gather_load_p): Likewise.
11395 * optabs-query.h (supports_vec_gather_load_p): Declare.
11396 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11397 and memory_type field.
11398 (NUM_PATTERNS): Bump to 15.
11399 * tree-vect-data-refs.c: Include internal-fn.h.
11400 (vect_gather_scatter_fn_p): New function.
11401 (vect_describe_gather_scatter_call): Likewise.
11402 (vect_check_gather_scatter): Try using internal functions for
11403 gather loads. Recognize existing calls to a gather load function.
11404 (vect_analyze_data_refs): Consider using gather loads if
11405 supports_vec_gather_load_p.
11406 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11407 (vect_get_gather_scatter_offset_type): Likewise.
11408 (vect_convert_mask_for_vectype): Likewise.
11409 (vect_add_conversion_to_patterm): Likewise.
11410 (vect_try_gather_scatter_pattern): Likewise.
11411 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11412 (vect_vect_recog_func_ptrs): Add it.
11413 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11414 internal_fn_mask_index and internal_gather_scatter_fn_p.
11415 (check_load_store_masking): Take the gather_scatter_info as an
11416 argument and handle gather loads.
11417 (vect_get_gather_scatter_ops): New function.
11418 (vectorizable_call): Check internal_load_fn_p.
11419 (vectorizable_load): Likewise. Handle gather load internal
11420 functions.
11421 (vectorizable_store): Update call to check_load_store_masking.
11422 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11423 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11424 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11425 (aarch64_gather_scale_operand_d): New predicates.
11426 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11427 (mask_gather_load<mode>): New insns.
11429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11430 Alan Hayward <alan.hayward@arm.com>
11431 David Sherwood <david.sherwood@arm.com>
11433 * optabs.def (fold_left_plus_optab): New optab.
11434 * doc/md.texi (fold_left_plus_@var{m}): Document.
11435 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11436 * internal-fn.c (fold_left_direct): Define.
11437 (expand_fold_left_optab_fn): Likewise.
11438 (direct_fold_left_optab_supported_p): Likewise.
11439 * fold-const-call.c (fold_const_fold_left): New function.
11440 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11441 * tree-parloops.c (valid_reduction_p): New function.
11442 (gather_scalar_reductions): Use it.
11443 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11444 (vect_finish_replace_stmt): Declare.
11445 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11446 (needs_fold_left_reduction_p): New function, split out from...
11447 (vect_is_simple_reduction): ...here. Accept reductions that
11448 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11449 (vect_force_simple_reduction): Also store the reduction type in
11450 the assignment's STMT_VINFO_REDUC_TYPE.
11451 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11452 (merge_with_identity): New function.
11453 (vect_expand_fold_left): Likewise.
11454 (vectorize_fold_left_reduction): Likewise.
11455 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11456 scalar phi in place for it. Check for target support and reject
11457 cases that would reassociate the operation. Defer the transform
11458 phase to vectorize_fold_left_reduction.
11459 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11460 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11461 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11465 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11466 call to ifc_temp_var.
11468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11469 Alan Hayward <alan.hayward@arm.com>
11470 David Sherwood <david.sherwood@arm.com>
11472 * target.def (legitimize_address_displacement): Take the original
11473 offset as a poly_int.
11474 * targhooks.h (default_legitimize_address_displacement): Update
11475 accordingly.
11476 * targhooks.c (default_legitimize_address_displacement): Likewise.
11477 * doc/tm.texi: Regenerate.
11478 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11479 as an argument, moving assert of ad->disp == ad->disp_term to...
11480 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11481 Try calling targetm.legitimize_address_displacement before expanding
11482 the address rather than afterwards, and adjust for the new interface.
11483 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11484 Match the new hook interface. Handle SVE addresses.
11485 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11486 new hook interface.
11488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11490 * Makefile.in (OBJS): Add early-remat.o.
11491 * target.def (select_early_remat_modes): New hook.
11492 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11493 * doc/tm.texi: Regenerate.
11494 * targhooks.h (default_select_early_remat_modes): Declare.
11495 * targhooks.c (default_select_early_remat_modes): New function.
11496 * timevar.def (TV_EARLY_REMAT): New timevar.
11497 * passes.def (pass_early_remat): New pass.
11498 * tree-pass.h (make_pass_early_remat): Declare.
11499 * early-remat.c: New file.
11500 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11501 function.
11502 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11505 Alan Hayward <alan.hayward@arm.com>
11506 David Sherwood <david.sherwood@arm.com>
11508 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11509 vfm1 with a bound_epilog parameter.
11510 (vect_do_peeling): Update calls accordingly, and move the prologue
11511 call earlier in the function. Treat the base bound_epilog as 0 for
11512 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11513 this base when peeling for gaps.
11514 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11515 with fully-masked loops.
11516 (vect_estimate_min_profitable_iters): Handle the single peeled
11517 iteration in that case.
11519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11520 Alan Hayward <alan.hayward@arm.com>
11521 David Sherwood <david.sherwood@arm.com>
11523 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11524 single-element interleaving even if the size is not a power of 2.
11525 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11526 accesses for single-element interleaving if the group size is
11527 not a power of 2.
11529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * doc/md.texi (fold_extract_last_@var{m}): Document.
11534 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11535 * optabs.def (fold_extract_last_optab): New optab.
11536 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11537 * internal-fn.c (fold_extract_direct): New macro.
11538 (expand_fold_extract_optab_fn): Likewise.
11539 (direct_fold_extract_optab_supported_p): Likewise.
11540 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11541 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11542 EXTRACT_LAST_REDUCTION.
11543 (get_initial_def_for_reduction): Do not create an initial vector
11544 for EXTRACT_LAST_REDUCTION reductions.
11545 (vectorizable_reduction): Leave the scalar phi in place for
11546 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11547 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11548 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11549 transform phase to vectorizable_condition.
11550 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11551 split out from...
11552 (vect_finish_stmt_generation): ...here.
11553 (vect_finish_replace_stmt): New function.
11554 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11555 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11556 pattern.
11557 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 * doc/md.texi (extract_last_@var{m}): Document.
11564 * optabs.def (extract_last_optab): New optab.
11565 * internal-fn.def (EXTRACT_LAST): New internal function.
11566 * internal-fn.c (cond_unary_direct): New macro.
11567 (expand_cond_unary_optab_fn): Likewise.
11568 (direct_cond_unary_optab_supported_p): Likewise.
11569 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11570 loops using EXTRACT_LAST.
11571 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11572 (extract_last_<mode>): ...this optab.
11573 (vec_extract<mode><Vel>): Update accordingly.
11575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11576 Alan Hayward <alan.hayward@arm.com>
11577 David Sherwood <david.sherwood@arm.com>
11579 * target.def (empty_mask_is_expensive): New hook.
11580 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11581 * doc/tm.texi: Regenerate.
11582 * targhooks.h (default_empty_mask_is_expensive): Declare.
11583 * targhooks.c (default_empty_mask_is_expensive): New function.
11584 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11585 if the target says that empty masks are expensive.
11586 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11587 New function.
11588 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11591 Alan Hayward <alan.hayward@arm.com>
11592 David Sherwood <david.sherwood@arm.com>
11594 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11595 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11596 (vect_use_loop_mask_for_alignment_p): New function.
11597 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11598 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11599 niters_skip argument. Make sure that the first niters_skip elements
11600 of the first iteration are inactive.
11601 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11602 Update call to vect_set_loop_masks_directly.
11603 (get_misalign_in_elems): New function, split out from...
11604 (vect_gen_prolog_loop_niters): ...here.
11605 (vect_update_init_of_dr): Take a code argument that specifies whether
11606 the adjustment should be added or subtracted.
11607 (vect_update_init_of_drs): Likewise.
11608 (vect_prepare_for_masked_peels): New function.
11609 (vect_do_peeling): Skip prologue peeling if we're using a mask
11610 instead. Update call to vect_update_inits_of_drs.
11611 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11612 mask_skip_niters.
11613 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11614 alignment. Do not include the number of peeled iterations in
11615 the minimum threshold in that case.
11616 (vectorizable_induction): Adjust the start value down by
11617 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11618 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11619 Take the number of skipped iterations into account when calculating
11620 the loop bounds.
11621 * tree-vect-stmts.c (vect_gen_while_not): New function.
11623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11624 Alan Hayward <alan.hayward@arm.com>
11625 David Sherwood <david.sherwood@arm.com>
11627 * doc/sourcebuild.texi (vect_fully_masked): Document.
11628 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11629 default value to 0.
11630 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11631 split out from...
11632 (vect_analyze_loop_2): ...here. Don't check the vectorization
11633 factor against the number of loop iterations if the loop is
11634 fully-masked.
11636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11637 Alan Hayward <alan.hayward@arm.com>
11638 David Sherwood <david.sherwood@arm.com>
11640 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11641 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11642 (dump_groups): Update accordingly.
11643 (iv_use::mem_type): New member variable.
11644 (address_p): New function.
11645 (record_use): Add a mem_type argument and initialize the new
11646 mem_type field.
11647 (record_group_use): Add a mem_type argument. Use address_p.
11648 Remove obsolete null checks of base_object. Update call to record_use.
11649 (find_interesting_uses_op): Update call to record_group_use.
11650 (find_interesting_uses_cond): Likewise.
11651 (find_interesting_uses_address): Likewise.
11652 (get_mem_type_for_internal_fn): New function.
11653 (find_address_like_use): Likewise.
11654 (find_interesting_uses_stmt): Try find_address_like_use before
11655 calling find_interesting_uses_op.
11656 (addr_offset_valid_p): Use the iv mem_type field as the type
11657 of the addressed memory.
11658 (add_autoinc_candidates): Likewise.
11659 (get_address_cost): Likewise.
11660 (split_small_address_groups_p): Use address_p.
11661 (split_address_groups): Likewise.
11662 (add_iv_candidate_for_use): Likewise.
11663 (autoinc_possible_for_pair): Likewise.
11664 (rewrite_groups): Likewise.
11665 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11666 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11667 (get_alias_ptr_type_for_ptr_address): New function.
11668 (rewrite_use_address): Rewrite address uses in calls that were
11669 identified by find_address_like_use.
11671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11672 Alan Hayward <alan.hayward@arm.com>
11673 David Sherwood <david.sherwood@arm.com>
11675 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11676 TARGET_MEM_REFs.
11677 * gimple-expr.h (is_gimple_addressable: Likewise.
11678 * gimple-expr.c (is_gimple_address): Likewise.
11679 * internal-fn.c (expand_call_mem_ref): New function.
11680 (expand_mask_load_optab_fn): Use it.
11681 (expand_mask_store_optab_fn): Likewise.
11683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11684 Alan Hayward <alan.hayward@arm.com>
11685 David Sherwood <david.sherwood@arm.com>
11687 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11688 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11689 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11690 (cond_umax@var{mode}): Document.
11691 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11692 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11693 (cond_umin_optab, cond_umax_optab): New optabs.
11694 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11695 (COND_IOR, COND_XOR): New internal functions.
11696 * internal-fn.h (get_conditional_internal_fn): Declare.
11697 * internal-fn.c (cond_binary_direct): New macro.
11698 (expand_cond_binary_optab_fn): Likewise.
11699 (direct_cond_binary_optab_supported_p): Likewise.
11700 (get_conditional_internal_fn): New function.
11701 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11702 Cope with reduction statements that are vectorized as calls rather
11703 than assignments.
11704 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11705 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11706 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11707 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11708 (UNSPEC_COND_EOR): New unspecs.
11709 (optab): Add mappings for them.
11710 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11711 (sve_int_op, sve_fp_op): New int attributes.
11713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11714 Alan Hayward <alan.hayward@arm.com>
11715 David Sherwood <david.sherwood@arm.com>
11717 * optabs.def (while_ult_optab): New optab.
11718 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11719 * internal-fn.def (WHILE_ULT): New internal function.
11720 * internal-fn.h (direct_internal_fn_supported_p): New override
11721 that takes two types as argument.
11722 * internal-fn.c (while_direct): New macro.
11723 (expand_while_optab_fn): New function.
11724 (convert_optab_supported_p): Likewise.
11725 (direct_while_optab_supported_p): New macro.
11726 * wide-int.h (wi::udiv_ceil): New function.
11727 * tree-vectorizer.h (rgroup_masks): New structure.
11728 (vec_loop_masks): New typedef.
11729 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11730 and fully_masked_p.
11731 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11732 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11733 (vect_max_vf): New function.
11734 (slpeel_make_loop_iterate_ntimes): Delete.
11735 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11736 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11737 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11738 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11739 internal-fn.h, stor-layout.h and optabs-query.h.
11740 (vect_set_loop_mask): New function.
11741 (add_preheader_seq): Likewise.
11742 (add_header_seq): Likewise.
11743 (interleave_supported_p): Likewise.
11744 (vect_maybe_permute_loop_masks): Likewise.
11745 (vect_set_loop_masks_directly): Likewise.
11746 (vect_set_loop_condition_masked): Likewise.
11747 (vect_set_loop_condition_unmasked): New function, split out from
11748 slpeel_make_loop_iterate_ntimes.
11749 (slpeel_make_loop_iterate_ntimes): Rename to..
11750 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11751 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11752 (vect_do_peeling): Update call accordingly.
11753 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11754 loops.
11755 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11756 mask_compare_type, can_fully_mask_p and fully_masked_p.
11757 (release_vec_loop_masks): New function.
11758 (_loop_vec_info): Use it to free the loop masks.
11759 (can_produce_all_loop_masks_p): New function.
11760 (vect_get_max_nscalars_per_iter): Likewise.
11761 (vect_verify_full_masking): Likewise.
11762 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11763 retries, and free the mask rgroups before retrying. Check loop-wide
11764 reasons for disallowing fully-masked loops. Make the final decision
11765 about whether use a fully-masked loop or not.
11766 (vect_estimate_min_profitable_iters): Do not assume that peeling
11767 for the number of iterations will be needed for fully-masked loops.
11768 (vectorizable_reduction): Disable fully-masked loops.
11769 (vectorizable_live_operation): Likewise.
11770 (vect_halve_mask_nunits): New function.
11771 (vect_double_mask_nunits): Likewise.
11772 (vect_record_loop_mask): Likewise.
11773 (vect_get_loop_mask): Likewise.
11774 (vect_transform_loop): Handle the case in which the final loop
11775 iteration might handle a partial vector. Call vect_set_loop_condition
11776 instead of slpeel_make_loop_iterate_ntimes.
11777 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11778 (check_load_store_masking): New function.
11779 (prepare_load_store_mask): Likewise.
11780 (vectorizable_store): Handle fully-masked loops.
11781 (vectorizable_load): Likewise.
11782 (supportable_widening_operation): Use vect_halve_mask_nunits for
11783 booleans.
11784 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11785 (vect_gen_while): New function.
11786 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11787 (aarch64_uqdec<mode>): New insn.
11789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11790 Alan Hayward <alan.hayward@arm.com>
11791 David Sherwood <david.sherwood@arm.com>
11793 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11794 (reduc_xor_scal_optab): New optabs.
11795 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11796 (reduc_xor_scal_@var{m}): Document.
11797 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11798 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11799 internal functions.
11800 * fold-const-call.c (fold_const_call): Handle them.
11801 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11802 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11803 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11804 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11805 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11806 (UNSPEC_XORV): New unspecs.
11807 (optab): Add entries for them.
11808 (BITWISEV): New int iterator.
11809 (bit_reduc_op): New int attributes.
11811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11812 Alan Hayward <alan.hayward@arm.com>
11813 David Sherwood <david.sherwood@arm.com>
11815 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11816 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11817 * optabs.def (vec_shl_insert_optab): New optab.
11818 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11819 (duplicate_and_interleave): Likewise.
11820 * tree-vect-loop.c: Include internal-fn.h.
11821 (neutral_op_for_slp_reduction): New function, split out from
11822 get_initial_defs_for_reduction.
11823 (get_initial_def_for_reduction): Handle option 2 for variable-length
11824 vectors by loading the neutral value into a vector and then shifting
11825 the initial value into element 0.
11826 (get_initial_defs_for_reduction): Replace the code argument with
11827 the neutral value calculated by neutral_op_for_slp_reduction.
11828 Use gimple_build_vector for constant-length vectors.
11829 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11830 but the first group_size elements have a neutral value.
11831 Use duplicate_and_interleave otherwise.
11832 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11833 Update call to get_initial_defs_for_reduction. Handle SLP
11834 reductions for variable-length vectors by creating one vector
11835 result for each scalar result, with the elements associated
11836 with other scalar results stubbed out with the neutral value.
11837 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11838 Require IFN_VEC_SHL_INSERT for double reductions on
11839 variable-length vectors, or SLP reductions that have
11840 a neutral value. Require can_duplicate_and_interleave_p
11841 support for variable-length unchained SLP reductions if there
11842 is no neutral value, such as for MIN/MAX reductions. Also require
11843 the number of vector elements to be a multiple of the number of
11844 SLP statements when doing variable-length unchained SLP reductions.
11845 Update call to vect_create_epilog_for_reduction.
11846 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11847 and remove initial values.
11848 (duplicate_and_interleave): Make public.
11849 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11850 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11853 Alan Hayward <alan.hayward@arm.com>
11854 David Sherwood <david.sherwood@arm.com>
11856 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11857 (can_duplicate_and_interleave_p): New function.
11858 (vect_get_and_check_slp_defs): Take the vector of statements
11859 rather than just the current one. Remove excess parentheses.
11860 Restriction rejectinon of vect_constant_def and vect_external_def
11861 for variable-length vectors to boolean types, or types for which
11862 can_duplicate_and_interleave_p is false.
11863 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11864 (duplicate_and_interleave): New function.
11865 (vect_get_constant_vectors): Use gimple_build_vector for
11866 constant-length vectors and suitable variable-length constant
11867 vectors. Use duplicate_and_interleave for other variable-length
11868 vectors. Don't defer the update when inserting new statements.
11870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11871 Alan Hayward <alan.hayward@arm.com>
11872 David Sherwood <david.sherwood@arm.com>
11874 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11875 min_profitable_iters doesn't go negative.
11877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11878 Alan Hayward <alan.hayward@arm.com>
11879 David Sherwood <david.sherwood@arm.com>
11881 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11882 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11883 * optabs.def (vec_mask_load_lanes_optab): New optab.
11884 (vec_mask_store_lanes_optab): Likewise.
11885 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11886 (MASK_STORE_LANES): Likewise.
11887 * internal-fn.c (mask_load_lanes_direct): New macro.
11888 (mask_store_lanes_direct): Likewise.
11889 (expand_mask_load_optab_fn): Handle masked operations.
11890 (expand_mask_load_lanes_optab_fn): New macro.
11891 (expand_mask_store_optab_fn): Handle masked operations.
11892 (expand_mask_store_lanes_optab_fn): New macro.
11893 (direct_mask_load_lanes_optab_supported_p): Likewise.
11894 (direct_mask_store_lanes_optab_supported_p): Likewise.
11895 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11896 parameter.
11897 (vect_load_lanes_supported): Likewise.
11898 * tree-vect-data-refs.c (strip_conversion): New function.
11899 (can_group_stmts_p): Likewise.
11900 (vect_analyze_data_ref_accesses): Use it instead of checking
11901 for a pair of assignments.
11902 (vect_store_lanes_supported): Take a masked_p parameter.
11903 (vect_load_lanes_supported): Likewise.
11904 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11905 vect_store_lanes_supported and vect_load_lanes_supported.
11906 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11907 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11908 parameter. Don't allow gaps for masked accesses.
11909 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11910 and vect_load_lanes_supported.
11911 (get_load_store_type): Take a masked_p parameter and update
11912 call to get_group_load_store_type.
11913 (vectorizable_store): Update call to get_load_store_type.
11914 Handle IFN_MASK_STORE_LANES.
11915 (vectorizable_load): Update call to get_load_store_type.
11916 Handle IFN_MASK_LOAD_LANES.
11918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11919 Alan Hayward <alan.hayward@arm.com>
11920 David Sherwood <david.sherwood@arm.com>
11922 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11923 modes for SVE.
11924 * config/aarch64/aarch64-protos.h
11925 (aarch64_sve_struct_memory_operand_p): Declare.
11926 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11927 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11928 (VPRED, vpred): Handle SVE structure modes.
11929 * config/aarch64/constraints.md (Utx): New constraint.
11930 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11931 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11932 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11933 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11934 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11935 structure modes. Split into pieces after RA.
11936 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11937 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11938 New patterns.
11939 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11940 SVE structure modes.
11941 (aarch64_classify_address): Likewise.
11942 (sizetochar): Move earlier in file.
11943 (aarch64_print_operand): Handle SVE register lists.
11944 (aarch64_array_mode): New function.
11945 (aarch64_sve_struct_memory_operand_p): Likewise.
11946 (TARGET_ARRAY_MODE): Redefine.
11948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11949 Alan Hayward <alan.hayward@arm.com>
11950 David Sherwood <david.sherwood@arm.com>
11952 * target.def (array_mode): New target hook.
11953 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11954 * doc/tm.texi: Regenerate.
11955 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11956 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11957 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11958 targetm.array_mode.
11959 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11960 type sizes.
11962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11963 Alan Hayward <alan.hayward@arm.com>
11964 David Sherwood <david.sherwood@arm.com>
11966 * fold-const.c (fold_binary_loc): Check the argument types
11967 rather than the result type when testing for a vector operation.
11969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11971 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11972 * doc/tm.texi: Regenerate.
11974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11975 Alan Hayward <alan.hayward@arm.com>
11976 David Sherwood <david.sherwood@arm.com>
11978 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11979 (sve): Document new AArch64 extension.
11980 * doc/md.texi (w): Extend the description of the AArch64
11981 constraint to include SVE vectors.
11982 (Upl, Upa): Document new AArch64 predicate constraints.
11983 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11984 enum.
11985 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11986 (msve-vector-bits=): New option.
11987 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11988 SVE when these are disabled.
11989 (sve): New extension.
11990 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11991 modes. Adjust their number of units based on aarch64_sve_vg.
11992 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11993 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11994 aarch64_addr_query_type.
11995 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11996 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11997 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11998 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11999 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12000 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12001 (aarch64_simd_imm_zero_p): Delete.
12002 (aarch64_check_zero_based_sve_index_immediate): Declare.
12003 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12004 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12005 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12006 (aarch64_sve_float_mul_immediate_p): Likewise.
12007 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12008 rather than an rtx.
12009 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12010 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12011 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12012 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12013 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12014 (aarch64_regmode_natural_size): Likewise.
12015 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12016 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12017 left one place.
12018 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12019 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12020 for VG and the SVE predicate registers.
12021 (V_ALIASES): Add a "z"-prefixed alias.
12022 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12023 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12024 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12025 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12026 (REG_CLASS_NAMES): Add entries for them.
12027 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12028 and the predicate registers.
12029 (aarch64_sve_vg): Declare.
12030 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12031 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12032 (REGMODE_NATURAL_SIZE): Define.
12033 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12034 SVE macros.
12035 * config/aarch64/aarch64.c: Include cfgrtl.h.
12036 (simd_immediate_info): Add a constructor for series vectors,
12037 and an associated step field.
12038 (aarch64_sve_vg): New variable.
12039 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12040 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12041 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12042 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12043 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12044 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12045 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12046 (aarch64_get_mask_mode): New functions.
12047 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12048 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12049 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12050 predicate modes and predicate registers. Explicitly restrict
12051 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12052 to store a vector mode if it is recognized by
12053 aarch64_classify_vector_mode.
12054 (aarch64_regmode_natural_size): New function.
12055 (aarch64_hard_regno_caller_save_mode): Return the original mode
12056 for predicates.
12057 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12058 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12059 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12060 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12061 functions.
12062 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12063 does not overlap dest if the function is frame-related. Handle
12064 SVE constants.
12065 (aarch64_split_add_offset): New function.
12066 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12067 them aarch64_add_offset.
12068 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12069 and update call to aarch64_sub_sp.
12070 (aarch64_add_cfa_expression): New function.
12071 (aarch64_expand_prologue): Pass extra temporary registers to the
12072 functions above. Handle the case in which we need to emit new
12073 DW_CFA_expressions for registers that were originally saved
12074 relative to the stack pointer, but now have to be expressed
12075 relative to the frame pointer.
12076 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12077 functions above.
12078 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12079 IP0 and IP1 values for SVE frames.
12080 (aarch64_expand_vec_series): New function.
12081 (aarch64_expand_sve_widened_duplicate): Likewise.
12082 (aarch64_expand_sve_const_vector): Likewise.
12083 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12084 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12085 into the register, rather than emitting a SET directly.
12086 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12087 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12088 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12089 (offset_9bit_signed_scaled_p): New functions.
12090 (aarch64_replicate_bitmask_imm): New function.
12091 (aarch64_bitmask_imm): Use it.
12092 (aarch64_cannot_force_const_mem): Reject expressions involving
12093 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12094 (aarch64_classify_index): Handle SVE indices, by requiring
12095 a plain register index with a scale that matches the element size.
12096 (aarch64_classify_address): Handle SVE addresses. Assert that
12097 the mode of the address is VOIDmode or an integer mode.
12098 Update call to aarch64_classify_symbol.
12099 (aarch64_classify_symbolic_expression): Update call to
12100 aarch64_classify_symbol.
12101 (aarch64_const_vec_all_in_range_p): New function.
12102 (aarch64_print_vector_float_operand): Likewise.
12103 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12104 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12105 and the FP immediates 1.0 and 0.5.
12106 (aarch64_print_address_internal): Handle SVE addresses.
12107 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12108 (aarch64_regno_regclass): Handle predicate registers.
12109 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12110 data modes.
12111 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12112 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12113 (aarch64_convert_sve_vector_bits): New function.
12114 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12115 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12116 rather than an rtx.
12117 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12118 Handle SVE vector and predicate modes. Accept VL-based constants
12119 that need only one temporary register, and VL offsets that require
12120 no temporary registers.
12121 (aarch64_conditional_register_usage): Mark the predicate registers
12122 as fixed if SVE isn't available.
12123 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12124 Return true for SVE vector and predicate modes.
12125 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12126 rather than an unsigned int. Handle SVE modes.
12127 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12128 SVE modes.
12129 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12130 if SVE is enabled.
12131 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12132 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12133 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12134 (aarch64_sve_float_mul_immediate_p): New functions.
12135 (aarch64_sve_valid_immediate): New function.
12136 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12137 Explicitly reject structure modes. Check for INDEX constants.
12138 Handle PTRUE and PFALSE constants.
12139 (aarch64_check_zero_based_sve_index_immediate): New function.
12140 (aarch64_simd_imm_zero_p): Delete.
12141 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12142 vector modes. Accept constants in the range of CNT[BHWD].
12143 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12144 ask for an Advanced SIMD mode.
12145 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12146 (aarch64_simd_vector_alignment): Handle SVE predicates.
12147 (aarch64_vectorize_preferred_vector_alignment): New function.
12148 (aarch64_simd_vector_alignment_reachable): Use it instead of
12149 the vector size.
12150 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12151 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12152 functions.
12153 (MAX_VECT_LEN): Delete.
12154 (expand_vec_perm_d): Add a vec_flags field.
12155 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12156 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12157 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12158 for SVE modes.
12159 (aarch64_evpc_rev): Rename to...
12160 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12161 (aarch64_evpc_rev_global): New function.
12162 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12163 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12164 MAX_VECT_LEN.
12165 (aarch64_evpc_sve_tbl): New function.
12166 (aarch64_expand_vec_perm_const_1): Update after rename of
12167 aarch64_evpc_rev. Handle SVE permutes too, trying
12168 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12169 than aarch64_evpc_tbl.
12170 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12171 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12172 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12173 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12174 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12175 (aarch64_expand_sve_vcond): New functions.
12176 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12177 of aarch64_vector_mode_p.
12178 (aarch64_dwarf_poly_indeterminate_value): New function.
12179 (aarch64_compute_pressure_classes): Likewise.
12180 (aarch64_can_change_mode_class): Likewise.
12181 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12182 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12183 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12184 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12185 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12186 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12187 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12188 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12189 constraints.
12190 (Dn, Dl, Dr): Accept const as well as const_vector.
12191 (Dz): Likewise. Compare against CONST0_RTX.
12192 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12193 of "vector" where appropriate.
12194 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12195 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12196 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12197 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12198 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12199 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12200 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12201 (v_int_equiv): Extend to SVE modes.
12202 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12203 mode attributes.
12204 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12205 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12206 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12207 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12208 (SVE_COND_FP_CMP): New int iterators.
12209 (perm_hilo): Handle the new unpack unspecs.
12210 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12211 attributes.
12212 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12213 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12214 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12215 (aarch64_equality_operator, aarch64_constant_vector_operand)
12216 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12217 (aarch64_sve_nonimmediate_operand): Likewise.
12218 (aarch64_sve_general_operand): Likewise.
12219 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12220 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12221 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12222 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12223 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12224 (aarch64_sve_float_arith_immediate): Likewise.
12225 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12226 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12227 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12228 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12229 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12230 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12231 (aarch64_sve_float_arith_operand): Likewise.
12232 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12233 (aarch64_sve_float_mul_operand): Likewise.
12234 (aarch64_sve_vec_perm_operand): Likewise.
12235 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12236 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12237 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12238 as well as const_vector.
12239 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12240 in file. Use CONST0_RTX and CONSTM1_RTX.
12241 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12242 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12243 Use aarch64_simd_imm_zero.
12244 * config/aarch64/aarch64-sve.md: New file.
12245 * config/aarch64/aarch64.md: Include it.
12246 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12247 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12248 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12249 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12250 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12251 (sve): New attribute.
12252 (enabled): Disable instructions with the sve attribute unless
12253 TARGET_SVE.
12254 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12255 aarch64_expand_mov_immediate.
12256 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12257 CNT[BHSD] immediates.
12258 (movti): Split CONST_POLY_INT moves into two halves.
12259 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12260 Split additions that need a temporary here if the destination
12261 is the stack pointer.
12262 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12263 (*add<mode>3_poly_1): New instruction.
12264 (set_clobber_cc): New expander.
12266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12268 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12269 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12270 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12271 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12272 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12273 Change innermode from fixed_mode_size to machine_mode.
12274 (simplify_subreg): Update call accordingly. Handle a constant-sized
12275 subreg of a variable-length CONST_VECTOR.
12277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12278 Alan Hayward <alan.hayward@arm.com>
12279 David Sherwood <david.sherwood@arm.com>
12281 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12282 (add_offset_to_base): New function, split out from...
12283 (create_mem_ref): ...here. When handling a scale other than 1,
12284 check first whether the address is valid without the offset.
12285 Add it into the base if so, leaving the index and scale as-is.
12287 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12289 PR c++/83778
12290 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12291 fold_for_warn before checking if arg2 is INTEGER_CST.
12293 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12295 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12296 (store_multiple_operation): Delete.
12297 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12298 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12299 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12300 guarded by TARGET_STRING.
12301 (rs6000_output_load_multiple): Delete.
12302 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12303 OPTION_MASK_STRING / TARGET_STRING handling.
12304 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12305 (const rs6000_opt_masks) <"string">: Change mask to 0.
12306 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12307 (MASK_STRING): Delete.
12308 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12309 parts. Simplify.
12310 (load_multiple): Delete.
12311 (*ldmsi8): Delete.
12312 (*ldmsi7): Delete.
12313 (*ldmsi6): Delete.
12314 (*ldmsi5): Delete.
12315 (*ldmsi4): Delete.
12316 (*ldmsi3): Delete.
12317 (store_multiple): Delete.
12318 (*stmsi8): Delete.
12319 (*stmsi7): Delete.
12320 (*stmsi6): Delete.
12321 (*stmsi5): Delete.
12322 (*stmsi4): Delete.
12323 (*stmsi3): Delete.
12324 (movmemsi_8reg): Delete.
12325 (corresponding unnamed define_insn): Delete.
12326 (movmemsi_6reg): Delete.
12327 (corresponding unnamed define_insn): Delete.
12328 (movmemsi_4reg): Delete.
12329 (corresponding unnamed define_insn): Delete.
12330 (movmemsi_2reg): Delete.
12331 (corresponding unnamed define_insn): Delete.
12332 (movmemsi_1reg): Delete.
12333 (corresponding unnamed define_insn): Delete.
12334 * config/rs6000/rs6000.opt (mno-string): New.
12335 (mstring): Replace by deprecation warning stub.
12336 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12338 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12340 * regrename.c (regrename_do_replace): If replacing the same
12341 reg multiple times, try to reuse last created gen_raw_REG.
12343 PR debug/81155
12344 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12345 main to workaround a bug in GDB.
12347 2018-01-12 Tom de Vries <tom@codesourcery.com>
12349 PR target/83737
12350 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12352 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12354 PR rtl-optimization/80481
12355 * ira-color.c (get_cap_member): New function.
12356 (allocnos_conflict_by_live_ranges_p): Use it.
12357 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12358 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12360 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12362 PR target/83628
12363 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12364 (*saddl_se_1): Ditto.
12365 (*ssubsi_1): Ditto.
12366 (*ssubl_se_1): Ditto.
12368 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12370 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12371 rather than wi::to_widest for DR_INITs.
12372 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12373 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12374 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12375 INTEGER_CSTs.
12376 (vect_analyze_group_access_1): Note that here.
12378 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12380 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12381 polynomial type sizes.
12383 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12385 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12386 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12387 (gimple_add_tmp_var): Likewise.
12389 2018-01-12 Martin Liska <mliska@suse.cz>
12391 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12392 (gimple_alloc_sizes): Likewise.
12393 (dump_gimple_statistics): Use PRIu64 in printf format.
12394 * gimple.h: Change uint64_t to int.
12396 2018-01-12 Martin Liska <mliska@suse.cz>
12398 * tree-core.h: Use uint64_t instead of int.
12399 * tree.c (tree_node_counts): Likewise.
12400 (tree_node_sizes): Likewise.
12401 (dump_tree_statistics): Use PRIu64 in printf format.
12403 2018-01-12 Martin Liska <mliska@suse.cz>
12405 * Makefile.in: As qsort_chk is implemented in vec.c, add
12406 vec.o to linkage of gencfn-macros.
12407 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12408 passing the info to record_node_allocation_statistics.
12409 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12410 and pass the info.
12411 * ggc-common.c (struct ggc_usage): Add operator== and use
12412 it in operator< and compare function.
12413 * mem-stats.h (struct mem_usage): Likewise.
12414 * vec.c (struct vec_usage): Remove operator< and compare
12415 function. Can be simply inherited.
12417 2018-01-12 Martin Jambor <mjambor@suse.cz>
12419 PR target/81616
12420 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12421 * tree-ssa-math-opts.c: Include domwalk.h.
12422 (convert_mult_to_fma_1): New function.
12423 (fma_transformation_info): New type.
12424 (fma_deferring_state): Likewise.
12425 (cancel_fma_deferring): New function.
12426 (result_of_phi): Likewise.
12427 (last_fma_candidate_feeds_initial_phi): Likewise.
12428 (convert_mult_to_fma): Added deferring logic, split actual
12429 transformation to convert_mult_to_fma_1.
12430 (math_opts_dom_walker): New type.
12431 (math_opts_dom_walker::after_dom_children): New method, body moved
12432 here from pass_optimize_widening_mul::execute, added deferring logic
12433 bits.
12434 (pass_optimize_widening_mul::execute): Moved most of code to
12435 math_opts_dom_walker::after_dom_children.
12436 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12437 * config/i386/i386.c (ix86_option_override_internal): Added
12438 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12440 2018-01-12 Richard Biener <rguenther@suse.de>
12442 PR debug/83157
12443 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12444 inline instance vars.
12446 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12448 PR target/81819
12449 * config/rx/rx.c (rx_is_restricted_memory_address):
12450 Handle SUBREG case.
12452 2018-01-12 Richard Biener <rguenther@suse.de>
12454 PR tree-optimization/80846
12455 * target.def (split_reduction): New target hook.
12456 * targhooks.c (default_split_reduction): New function.
12457 * targhooks.h (default_split_reduction): Declare.
12458 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12459 target requests first reduce vectors by combining low and high
12460 parts.
12461 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12462 (get_vectype_for_scalar_type_and_size): Export.
12463 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12464 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12465 * doc/tm.texi: Regenerate.
12466 * config/i386/i386.c (ix86_split_reduction): Implement
12467 TARGET_VECTORIZE_SPLIT_REDUCTION.
12469 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12471 PR target/83368
12472 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12473 in PIC mode except for TARGET_VXWORKS_RTP.
12474 * config/sparc/sparc.c: Include cfgrtl.h.
12475 (TARGET_INIT_PIC_REG): Define.
12476 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12477 (sparc_pic_register_p): New predicate.
12478 (sparc_legitimate_address_p): Use it.
12479 (sparc_legitimize_pic_address): Likewise.
12480 (sparc_delegitimize_address): Likewise.
12481 (sparc_mode_dependent_address_p): Likewise.
12482 (gen_load_pcrel_sym): Remove 4th parameter.
12483 (load_got_register): Adjust call to above. Remove obsolete stuff.
12484 (sparc_expand_prologue): Do not call load_got_register here.
12485 (sparc_flat_expand_prologue): Likewise.
12486 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12487 (sparc_use_pseudo_pic_reg): New function.
12488 (sparc_init_pic_reg): Likewise.
12489 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12490 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12492 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12494 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12495 Add item for branch_cost.
12497 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12499 PR rtl-optimization/83565
12500 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12501 not extend the result to a larger mode for rotate operations.
12502 (num_sign_bit_copies1): Likewise.
12504 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12506 PR target/40411
12507 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12508 -symbolic.
12509 Use values-Xc.o for -pedantic.
12510 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12512 2018-01-12 Martin Liska <mliska@suse.cz>
12514 PR ipa/83054
12515 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12516 New function.
12517 (possible_polymorphic_call_targets): Use it.
12518 (ipa_devirt): Likewise.
12520 2018-01-12 Martin Liska <mliska@suse.cz>
12522 * profile-count.h (enum profile_quality): Use 0 as invalid
12523 enum value of profile_quality.
12525 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12527 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12528 -mext-string options.
12530 2018-01-12 Richard Biener <rguenther@suse.de>
12532 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12533 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12534 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12535 Likewise.
12536 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12538 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12540 * configure.ac (--with-long-double-format): Add support for the
12541 configuration option to change the default long double format on
12542 PowerPC systems.
12543 * config.gcc (powerpc*-linux*-*): Likewise.
12544 * configure: Regenerate.
12545 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12546 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12547 used without modification.
12549 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12551 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12552 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12553 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12554 MISC_BUILTIN_SPEC_BARRIER.
12555 (rs6000_init_builtins): Likewise.
12556 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12557 enum value.
12558 (speculation_barrier): New define_insn.
12559 * doc/extend.texi: Document __builtin_speculation_barrier.
12561 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12563 PR target/83203
12564 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12565 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12566 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12567 iterators.
12568 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12569 integral modes instead of "ss" and "sd".
12570 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12571 vectors with 32-bit and 64-bit elements.
12572 (vecdupssescalarmodesuffix): New mode attribute.
12573 (vec_dup<mode>): Use it.
12575 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12577 PR target/83330
12578 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12579 frame if argument is passed on stack.
12581 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12583 PR target/82682
12584 * ree.c (combine_reaching_defs): Optimize also
12585 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12586 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12588 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12590 PR middle-end/83189
12591 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12593 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12595 PR middle-end/83718
12596 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12597 after they are computed.
12599 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12601 PR tree-optimization/83695
12602 * gimple-loop-linterchange.cc
12603 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12604 reset cached scev information after interchange.
12605 (pass_linterchange::execute): Remove call to scev_reset_htab.
12607 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12609 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12610 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12611 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12612 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12613 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12614 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12615 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12616 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12617 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12618 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12619 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12620 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12621 (V_lane_reg): Likewise.
12622 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12623 New define_expand.
12624 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12625 (vfmal_lane_low<mode>_intrinsic,
12626 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12627 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12628 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12629 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12630 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12631 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12633 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12635 * config/arm/arm-cpus.in (fp16fml): New feature.
12636 (ALL_SIMD): Add fp16fml.
12637 (armv8.2-a): Add fp16fml as an option.
12638 (armv8.3-a): Likewise.
12639 (armv8.4-a): Add fp16fml as part of fp16.
12640 * config/arm/arm.h (TARGET_FP16FML): Define.
12641 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12642 when appropriate.
12643 * config/arm/arm-modes.def (V2HF): Define.
12644 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12645 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12646 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12647 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12648 vfmsl_low, vfmsl_high): New set of builtins.
12649 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12650 (vfml_op): New code attribute.
12651 (VFMLHALVES): New int iterator.
12652 (VFML, VFMLSEL): New mode attributes.
12653 (V_reg): Define mapping for V2HF.
12654 (V_hi, V_lo): New mode attributes.
12655 (VF_constraint): Likewise.
12656 (vfml_half, vfml_half_selector): New int attributes.
12657 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12658 define_expand.
12659 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12660 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12661 New define_insn.
12662 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12663 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12664 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12665 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12666 documentation.
12667 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12668 Document new effective target and option set.
12670 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12672 * config/arm/arm-cpus.in (armv8_4): New feature.
12673 (ARMv8_4a): New fgroup.
12674 (armv8.4-a): New arch.
12675 * config/arm/arm-tables.opt: Regenerate.
12676 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12677 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12678 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12679 Add matching rules for -march=armv8.4-a and extensions.
12680 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12682 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12684 PR target/81821
12685 * config/rx/rx.md (BW): New mode attribute.
12686 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12688 2018-01-11 Richard Biener <rguenther@suse.de>
12690 PR tree-optimization/83435
12691 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12692 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12693 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12695 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12696 Alan Hayward <alan.hayward@arm.com>
12697 David Sherwood <david.sherwood@arm.com>
12699 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12700 field.
12701 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12702 (aarch64_print_address_internal): Use it to check for a zero offset.
12704 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12705 Alan Hayward <alan.hayward@arm.com>
12706 David Sherwood <david.sherwood@arm.com>
12708 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12709 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12710 Return a poly_int64 rather than a HOST_WIDE_INT.
12711 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12712 rather than a HOST_WIDE_INT.
12713 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12714 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12715 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12716 final_offset from HOST_WIDE_INT to poly_int64.
12717 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12718 to_constant when getting the number of units in an Advanced SIMD
12719 mode.
12720 (aarch64_builtin_vectorized_function): Check for a constant number
12721 of units.
12722 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12723 GET_MODE_SIZE.
12724 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12725 attribute instead of GET_MODE_NUNITS.
12726 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12727 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12728 GET_MODE_SIZE for fixed-size registers.
12729 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12730 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12731 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12732 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12733 (aarch64_print_operand, aarch64_print_address_internal)
12734 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12735 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12736 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12737 Handle polynomial GET_MODE_SIZE.
12738 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12739 wider than SImode without modification.
12740 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12741 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12742 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12743 passing and returning SVE modes.
12744 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12745 rather than GEN_INT.
12746 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12747 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12748 (aarch64_allocate_and_probe_stack_space): Likewise.
12749 (aarch64_layout_frame): Cope with polynomial offsets.
12750 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12751 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12752 polynomial offsets.
12753 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12754 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12755 poly_int64 rather than a HOST_WIDE_INT.
12756 (aarch64_get_separate_components, aarch64_process_components)
12757 (aarch64_expand_prologue, aarch64_expand_epilogue)
12758 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12759 (aarch64_anchor_offset): New function, split out from...
12760 (aarch64_legitimize_address): ...here.
12761 (aarch64_builtin_vectorization_cost): Handle polynomial
12762 TYPE_VECTOR_SUBPARTS.
12763 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12764 GET_MODE_NUNITS.
12765 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12766 number of elements from the PARALLEL rather than the mode.
12767 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12768 rather than GET_MODE_BITSIZE.
12769 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12770 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12771 (aarch64_expand_vec_perm_const_1): Handle polynomial
12772 d->perm.length () and d->perm elements.
12773 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12774 Apply to_constant to d->perm elements.
12775 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12776 polynomial CONST_VECTOR_NUNITS.
12777 (aarch64_move_pointer): Take amount as a poly_int64 rather
12778 than an int.
12779 (aarch64_progress_pointer): Avoid temporary variable.
12780 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12781 the mode attribute instead of GET_MODE.
12783 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12784 Alan Hayward <alan.hayward@arm.com>
12785 David Sherwood <david.sherwood@arm.com>
12787 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12788 x exists before using it.
12789 (aarch64_add_constant_internal): Rename to...
12790 (aarch64_add_offset_1): ...this. Replace regnum with separate
12791 src and dest rtxes. Handle the case in which they're different,
12792 including when the offset is zero. Replace scratchreg with an rtx.
12793 Use 2 additions if there is no spare register into which we can
12794 move a 16-bit constant.
12795 (aarch64_add_constant): Delete.
12796 (aarch64_add_offset): Replace reg with separate src and dest
12797 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12798 Use aarch64_add_offset_1.
12799 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12800 an rtx rather than an int. Take the delta as a poly_int64
12801 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12802 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12803 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12804 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12805 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12806 and aarch64_add_sp.
12807 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12808 aarch64_add_constant.
12810 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12812 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12813 Use scalar_float_mode.
12815 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12817 * config/aarch64/aarch64-simd.md
12818 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12819 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12820 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12821 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12822 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12823 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12824 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12825 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12826 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12827 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12829 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12831 PR target/83514
12832 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12833 targ_options->x_arm_arch_string is non NULL.
12835 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12837 * config/aarch64/aarch64.h
12838 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12840 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12842 PR target/82096
12843 * expmed.c (emit_store_flag_force): Swap if const op0
12844 and change VOIDmode to mode of op0.
12846 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12848 PR rtl-optimization/83761
12849 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12850 than bytes to mode_for_size.
12852 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12854 PR middle-end/83189
12855 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12856 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12857 profile.
12859 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12861 PR middle-end/83575
12862 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12863 when in layout mode.
12864 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12865 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12866 partition fixup.
12868 2018-01-10 Michael Collison <michael.collison@arm.com>
12870 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12871 * config/aarch64/aarch64-option-extension.def: Add
12872 AARCH64_OPT_EXTENSION of 'fp16fml'.
12873 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12874 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12875 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12876 * config/aarch64/constraints.md (Ui7): New constraint.
12877 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12878 (VFMLA_SEL_W): Ditto.
12879 (f16quad): Ditto.
12880 (f16mac1): Ditto.
12881 (VFMLA16_LOW): New int iterator.
12882 (VFMLA16_HIGH): Ditto.
12883 (UNSPEC_FMLAL): New unspec.
12884 (UNSPEC_FMLSL): Ditto.
12885 (UNSPEC_FMLAL2): Ditto.
12886 (UNSPEC_FMLSL2): Ditto.
12887 (f16mac): New code attribute.
12888 * config/aarch64/aarch64-simd-builtins.def
12889 (aarch64_fmlal_lowv2sf): Ditto.
12890 (aarch64_fmlsl_lowv2sf): Ditto.
12891 (aarch64_fmlalq_lowv4sf): Ditto.
12892 (aarch64_fmlslq_lowv4sf): Ditto.
12893 (aarch64_fmlal_highv2sf): Ditto.
12894 (aarch64_fmlsl_highv2sf): Ditto.
12895 (aarch64_fmlalq_highv4sf): Ditto.
12896 (aarch64_fmlslq_highv4sf): Ditto.
12897 (aarch64_fmlal_lane_lowv2sf): Ditto.
12898 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12899 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12900 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12901 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12902 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12903 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12904 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12905 (aarch64_fmlal_lane_highv2sf): Ditto.
12906 (aarch64_fmlsl_lane_highv2sf): Ditto.
12907 (aarch64_fmlal_laneq_highv2sf): Ditto.
12908 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12909 (aarch64_fmlalq_lane_highv4sf): Ditto.
12910 (aarch64_fmlsl_lane_highv4sf): Ditto.
12911 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12912 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12913 * config/aarch64/aarch64-simd.md:
12914 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12915 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12916 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12917 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12918 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12919 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12920 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12921 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12922 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12923 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12924 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12925 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12926 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12927 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12928 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12929 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12930 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12931 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12932 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12933 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12934 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12935 (vfmlsl_low_u32): Ditto.
12936 (vfmlalq_low_u32): Ditto.
12937 (vfmlslq_low_u32): Ditto.
12938 (vfmlal_high_u32): Ditto.
12939 (vfmlsl_high_u32): Ditto.
12940 (vfmlalq_high_u32): Ditto.
12941 (vfmlslq_high_u32): Ditto.
12942 (vfmlal_lane_low_u32): Ditto.
12943 (vfmlsl_lane_low_u32): Ditto.
12944 (vfmlal_laneq_low_u32): Ditto.
12945 (vfmlsl_laneq_low_u32): Ditto.
12946 (vfmlalq_lane_low_u32): Ditto.
12947 (vfmlslq_lane_low_u32): Ditto.
12948 (vfmlalq_laneq_low_u32): Ditto.
12949 (vfmlslq_laneq_low_u32): Ditto.
12950 (vfmlal_lane_high_u32): Ditto.
12951 (vfmlsl_lane_high_u32): Ditto.
12952 (vfmlal_laneq_high_u32): Ditto.
12953 (vfmlsl_laneq_high_u32): Ditto.
12954 (vfmlalq_lane_high_u32): Ditto.
12955 (vfmlslq_lane_high_u32): Ditto.
12956 (vfmlalq_laneq_high_u32): Ditto.
12957 (vfmlslq_laneq_high_u32): Ditto.
12958 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12959 (AARCH64_FL_FOR_ARCH8_4): New.
12960 (AARCH64_ISA_F16FML): New ISA flag.
12961 (TARGET_F16FML): New feature flag for fp16fml.
12962 (doc/invoke.texi): Document new fp16fml option.
12964 2018-01-10 Michael Collison <michael.collison@arm.com>
12966 * config/aarch64/aarch64-builtins.c:
12967 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12968 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12969 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12970 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12971 (AARCH64_ISA_SHA3): New ISA flag.
12972 (TARGET_SHA3): New feature flag for sha3.
12973 * config/aarch64/iterators.md (sha512_op): New int attribute.
12974 (CRYPTO_SHA512): New int iterator.
12975 (UNSPEC_SHA512H): New unspec.
12976 (UNSPEC_SHA512H2): Ditto.
12977 (UNSPEC_SHA512SU0): Ditto.
12978 (UNSPEC_SHA512SU1): Ditto.
12979 * config/aarch64/aarch64-simd-builtins.def
12980 (aarch64_crypto_sha512hqv2di): New builtin.
12981 (aarch64_crypto_sha512h2qv2di): Ditto.
12982 (aarch64_crypto_sha512su0qv2di): Ditto.
12983 (aarch64_crypto_sha512su1qv2di): Ditto.
12984 (aarch64_eor3qv8hi): Ditto.
12985 (aarch64_rax1qv2di): Ditto.
12986 (aarch64_xarqv2di): Ditto.
12987 (aarch64_bcaxqv8hi): Ditto.
12988 * config/aarch64/aarch64-simd.md:
12989 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12990 (aarch64_crypto_sha512su0qv2di): Ditto.
12991 (aarch64_crypto_sha512su1qv2di): Ditto.
12992 (aarch64_eor3qv8hi): Ditto.
12993 (aarch64_rax1qv2di): Ditto.
12994 (aarch64_xarqv2di): Ditto.
12995 (aarch64_bcaxqv8hi): Ditto.
12996 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12997 (vsha512h2q_u64): Ditto.
12998 (vsha512su0q_u64): Ditto.
12999 (vsha512su1q_u64): Ditto.
13000 (veor3q_u16): Ditto.
13001 (vrax1q_u64): Ditto.
13002 (vxarq_u64): Ditto.
13003 (vbcaxq_u16): Ditto.
13004 * config/arm/types.md (crypto_sha512): New type attribute.
13005 (crypto_sha3): Ditto.
13006 (doc/invoke.texi): Document new sha3 option.
13008 2018-01-10 Michael Collison <michael.collison@arm.com>
13010 * config/aarch64/aarch64-builtins.c:
13011 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13012 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13013 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13014 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13015 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13016 (AARCH64_ISA_SM4): New ISA flag.
13017 (TARGET_SM4): New feature flag for sm4.
13018 * config/aarch64/aarch64-simd-builtins.def
13019 (aarch64_sm3ss1qv4si): Ditto.
13020 (aarch64_sm3tt1aq4si): Ditto.
13021 (aarch64_sm3tt1bq4si): Ditto.
13022 (aarch64_sm3tt2aq4si): Ditto.
13023 (aarch64_sm3tt2bq4si): Ditto.
13024 (aarch64_sm3partw1qv4si): Ditto.
13025 (aarch64_sm3partw2qv4si): Ditto.
13026 (aarch64_sm4eqv4si): Ditto.
13027 (aarch64_sm4ekeyqv4si): Ditto.
13028 * config/aarch64/aarch64-simd.md:
13029 (aarch64_sm3ss1qv4si): Ditto.
13030 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13031 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13032 (aarch64_sm4eqv4si): Ditto.
13033 (aarch64_sm4ekeyqv4si): Ditto.
13034 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13035 (sm3part_op): Ditto.
13036 (CRYPTO_SM3TT): Ditto.
13037 (CRYPTO_SM3PART): Ditto.
13038 (UNSPEC_SM3SS1): New unspec.
13039 (UNSPEC_SM3TT1A): Ditto.
13040 (UNSPEC_SM3TT1B): Ditto.
13041 (UNSPEC_SM3TT2A): Ditto.
13042 (UNSPEC_SM3TT2B): Ditto.
13043 (UNSPEC_SM3PARTW1): Ditto.
13044 (UNSPEC_SM3PARTW2): Ditto.
13045 (UNSPEC_SM4E): Ditto.
13046 (UNSPEC_SM4EKEY): Ditto.
13047 * config/aarch64/constraints.md (Ui2): New constraint.
13048 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13049 * config/arm/types.md (crypto_sm3): New type attribute.
13050 (crypto_sm4): Ditto.
13051 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13052 (vsm3tt1aq_u32): Ditto.
13053 (vsm3tt1bq_u32): Ditto.
13054 (vsm3tt2aq_u32): Ditto.
13055 (vsm3tt2bq_u32): Ditto.
13056 (vsm3partw1q_u32): Ditto.
13057 (vsm3partw2q_u32): Ditto.
13058 (vsm4eq_u32): Ditto.
13059 (vsm4ekeyq_u32): Ditto.
13060 (doc/invoke.texi): Document new sm4 option.
13062 2018-01-10 Michael Collison <michael.collison@arm.com>
13064 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13065 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13066 (AARCH64_FL_FOR_ARCH8_4): New.
13067 (AARCH64_FL_V8_4): New flag.
13068 (doc/invoke.texi): Document new armv8.4-a option.
13070 2018-01-10 Michael Collison <michael.collison@arm.com>
13072 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13073 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13074 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13075 * config/aarch64/aarch64-option-extension.def: Add
13076 AARCH64_OPT_EXTENSION of 'sha2'.
13077 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13078 (crypto): Disable sha2 and aes if crypto disabled.
13079 (crypto): Enable aes and sha2 if enabled.
13080 (simd): Disable sha2 and aes if simd disabled.
13081 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13082 New flags.
13083 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13084 (TARGET_SHA2): New feature flag for sha2.
13085 (TARGET_AES): New feature flag for aes.
13086 * config/aarch64/aarch64-simd.md:
13087 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13088 conditional on TARGET_AES.
13089 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13090 (aarch64_crypto_sha1hsi): Make pattern conditional
13091 on TARGET_SHA2.
13092 (aarch64_crypto_sha1hv4si): Ditto.
13093 (aarch64_be_crypto_sha1hv4si): Ditto.
13094 (aarch64_crypto_sha1su1v4si): Ditto.
13095 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13096 (aarch64_crypto_sha1su0v4si): Ditto.
13097 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13098 (aarch64_crypto_sha256su0v4si): Ditto.
13099 (aarch64_crypto_sha256su1v4si): Ditto.
13100 (doc/invoke.texi): Document new aes and sha2 options.
13102 2018-01-10 Martin Sebor <msebor@redhat.com>
13104 PR tree-optimization/83781
13105 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13106 as string arrays.
13108 2018-01-11 Martin Sebor <msebor@gmail.com>
13109 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13111 PR tree-optimization/83501
13112 PR tree-optimization/81703
13114 * tree-ssa-strlen.c (get_string_cst): Rename...
13115 (get_string_len): ...to this. Handle global constants.
13116 (handle_char_store): Adjust.
13118 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13119 Jim Wilson <jimw@sifive.com>
13121 * config/riscv/riscv-protos.h (riscv_output_return): New.
13122 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13123 (riscv_attribute_table, riscv_output_return),
13124 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13125 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13126 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13127 function.
13128 (riscv_expand_prologue): Add early return for naked function.
13129 (riscv_expand_epilogue): Likewise.
13130 (riscv_function_ok_for_sibcall): Return false for naked function.
13131 (riscv_set_current_function): New.
13132 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13133 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13134 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13135 * doc/extend.texi (RISC-V Function Attributes): New.
13137 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13139 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13140 check for 128-bit long double before checking TCmode.
13141 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13142 128-bit long doubles before checking TFmode or TCmode.
13143 (FLOAT128_IBM_P): Likewise.
13145 2018-01-10 Martin Sebor <msebor@redhat.com>
13147 PR tree-optimization/83671
13148 * builtins.c (c_strlen): Unconditionally return zero for the empty
13149 string.
13150 Use -Warray-bounds for warnings.
13151 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13152 for non-constant array indices with COMPONENT_REF, arrays of
13153 arrays, and pointers to arrays.
13154 (gimple_fold_builtin_strlen): Determine and set length range for
13155 non-constant character arrays.
13157 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13159 PR middle-end/81897
13160 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13161 empty blocks.
13163 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13165 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13167 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13169 PR target/83399
13170 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13171 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13172 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13173 indexed_or_indirect_operand predicate.
13174 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13175 (*vsx_le_perm_load_v8hi): Likewise.
13176 (*vsx_le_perm_load_v16qi): Likewise.
13177 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13178 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13179 (*vsx_le_perm_store_v8hi): Likewise.
13180 (*vsx_le_perm_store_v16qi): Likewise.
13181 (eight unnamed splitters): Likewise.
13183 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13185 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13186 * config/rs6000/emmintrin.h: Likewise.
13187 * config/rs6000/mmintrin.h: Likewise.
13188 * config/rs6000/xmmintrin.h: Likewise.
13190 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13192 PR c++/43486
13193 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13194 "public_flag".
13195 * tree.c (tree_nop_conversion): Return true for location wrapper
13196 nodes.
13197 (maybe_wrap_with_location): New function.
13198 (selftest::check_strip_nops): New function.
13199 (selftest::test_location_wrappers): New function.
13200 (selftest::tree_c_tests): Call it.
13201 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13202 (maybe_wrap_with_location): New decl.
13203 (EXPR_LOCATION_WRAPPER_P): New macro.
13204 (location_wrapper_p): New inline function.
13205 (tree_strip_any_location_wrapper): New inline function.
13207 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13209 PR target/83735
13210 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13211 stack_realign_offset for the largest alignment of stack slot
13212 actually used.
13213 (ix86_find_max_used_stack_alignment): New function.
13214 (ix86_finalize_stack_frame_flags): Use it. Set
13215 max_used_stack_alignment if we don't realign stack.
13216 * config/i386/i386.h (machine_function): Add
13217 max_used_stack_alignment.
13219 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13221 * config/arm/arm.opt (-mbranch-cost): New option.
13222 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13223 account.
13225 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13227 PR target/83629
13228 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13229 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13231 2018-01-10 Richard Biener <rguenther@suse.de>
13233 PR debug/83765
13234 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13235 early out so it also covers the case where we have a non-NULL
13236 origin.
13238 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13240 PR tree-optimization/83753
13241 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13242 for non-strided grouped accesses if the number of elements is 1.
13244 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13246 PR target/81616
13247 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13248 * i386.h (TARGET_USE_GATHER): Define.
13249 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13251 2018-01-10 Martin Liska <mliska@suse.cz>
13253 PR bootstrap/82831
13254 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13255 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13256 partitioning.
13257 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13258 CLEANUP_NO_PARTITIONING is not set.
13260 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13262 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13263 for vectors, as a partial revert of r254296.
13264 * rtl.h (const_vec_p): Delete.
13265 (const_vec_duplicate_p): Don't test for vector CONSTs.
13266 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13267 * expmed.c (make_tree): Likewise.
13269 Revert:
13270 * common.md (E, F): Use CONSTANT_P instead of checking for
13271 CONST_VECTOR.
13272 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13273 checking for CONST_VECTOR.
13275 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13277 PR middle-end/83575
13278 * predict.c (force_edge_cold): Handle in more sane way edges
13279 with no prediction.
13281 2018-01-09 Carl Love <cel@us.ibm.com>
13283 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13284 V4SI, V4SF types.
13285 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13286 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13287 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13288 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13289 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13290 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13291 * config/rs6000/rs6000-protos.h: Add extern defition for
13292 rs6000_generate_float2_double_code.
13293 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13294 function.
13295 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13296 (float2_v2df): Add define_expand.
13298 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13300 PR target/83628
13301 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13302 op_mode in the force_to_mode call.
13304 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13306 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13307 instead of checking each element individually.
13308 (aarch64_evpc_uzp): Likewise.
13309 (aarch64_evpc_zip): Likewise.
13310 (aarch64_evpc_ext): Likewise.
13311 (aarch64_evpc_rev): Likewise.
13312 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13313 instead of checking each element individually. Return true without
13314 generating rtl if
13315 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13316 whether all selected elements come from the same input, instead of
13317 checking each element individually. Remove calls to gen_rtx_REG,
13318 start_sequence and end_sequence and instead assert that no rtl is
13319 generated.
13321 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13323 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13324 order of HIGH and CONST checks.
13326 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13328 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13329 if the destination isn't an SSA_NAME.
13331 2018-01-09 Richard Biener <rguenther@suse.de>
13333 PR tree-optimization/83668
13334 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13335 move prologue...
13336 (canonicalize_loop_form): ... here, renamed from ...
13337 (canonicalize_loop_closed_ssa_form): ... this and amended to
13338 swap successor edges for loop exit blocks to make us use
13339 the RPO order we need for initial schedule generation.
13341 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13343 PR tree-optimization/64811
13344 * match.pd: When optimizing comparisons with Inf, avoid
13345 introducing or losing exceptions from comparisons with NaN.
13347 2018-01-09 Martin Liska <mliska@suse.cz>
13349 PR sanitizer/82517
13350 * asan.c (shadow_mem_size): Add gcc_assert.
13352 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13354 Don't save registers in main().
13356 PR target/83738
13357 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13358 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13359 * config/avr/avr.c (avr_set_current_function): Don't error if
13360 naked, OS_task or OS_main are specified at the same time.
13361 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13362 OS_main.
13363 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13364 attribute.
13365 * common/config/avr/avr-common.c (avr_option_optimization_table):
13366 Switch on -mmain-is-OS_task for optimizing compilations.
13368 2018-01-09 Richard Biener <rguenther@suse.de>
13370 PR tree-optimization/83572
13371 * graphite.c: Include cfganal.h.
13372 (graphite_transform_loops): Connect infinite loops to exit
13373 and remove fake edges at the end.
13375 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13377 * ipa-inline.c (edge_badness): Revert accidental checkin.
13379 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13381 PR ipa/80763
13382 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13383 symbols; not inline clones.
13385 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13387 PR target/83507
13388 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13389 hard registers. Formatting fixes.
13391 PR preprocessor/83722
13392 * gcc.c (try_generate_repro): Pass
13393 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13394 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13395 do_report_bug.
13397 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13398 Kito Cheng <kito.cheng@gmail.com>
13400 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13401 (riscv_leaf_function_p): Delete.
13402 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13404 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13406 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13407 function.
13408 (do_ifelse): New function.
13409 (do_isel): New function.
13410 (do_sub3): New function.
13411 (do_add3): New function.
13412 (do_load_mask_compare): New function.
13413 (do_overlap_load_compare): New function.
13414 (expand_compare_loop): New function.
13415 (expand_block_compare): Call expand_compare_loop() when appropriate.
13416 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13417 option description.
13418 (-mblock-compare-inline-loop-limit): New option.
13420 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13422 PR target/83677
13423 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13424 Reverse order of second and third operands in first alternative.
13425 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13426 of first and second elements in UNSPEC_VPERMR vector.
13427 (altivec_expand_vec_perm_le): Likewise.
13429 2018-01-08 Jeff Law <law@redhat.com>
13431 PR rtl-optimizatin/81308
13432 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13433 (process_switch): If group_case_labels makes a change, then set
13434 cfg_altered.
13435 (pass_convert_switch::execute): If a switch is converted, then
13436 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13438 PR rtl-optimization/81308
13439 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13440 splitting insns.
13442 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13444 PR target/83663 - Revert r255946
13445 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13446 generation for cases where splatting a value is not useful.
13447 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13448 across a vec_duplicate and a paradoxical subreg forming a vector
13449 mode to a vec_concat.
13451 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13453 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13454 -march=armv8.3-a variants.
13455 * config/arm/t-multilib: Likewise.
13456 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13458 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13460 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13461 to generate rtl.
13462 (cceq_ior_compare_complement): Give it a name so I can use it, and
13463 change boolean_or_operator predicate to boolean_operator so it can
13464 be used to generate a crand.
13465 (eqne): New code iterator.
13466 (bd/bd_neg): New code_attrs.
13467 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13468 a single define_insn.
13469 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13470 decrement (bdnzt/bdnzf/bdzt/bdzf).
13471 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13472 with the new names of the branch decrement patterns, and added the
13473 names of the branch decrement conditional patterns.
13475 2018-01-08 Richard Biener <rguenther@suse.de>
13477 PR tree-optimization/83563
13478 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13479 cache.
13481 2018-01-08 Richard Biener <rguenther@suse.de>
13483 PR middle-end/83713
13484 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13486 2018-01-08 Richard Biener <rguenther@suse.de>
13488 PR tree-optimization/83685
13489 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13490 references to abnormals.
13492 2018-01-08 Richard Biener <rguenther@suse.de>
13494 PR lto/83719
13495 * dwarf2out.c (output_indirect_strings): Handle empty
13496 skeleton_debug_str_hash.
13497 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13499 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13501 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13502 (emit_store_direct): Likewise.
13503 (arc_trampoline_adjust_address): Likewise.
13504 (arc_asm_trampoline_template): New function.
13505 (arc_initialize_trampoline): Use asm_trampoline_template.
13506 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13507 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13508 * config/arc/arc.md (flush_icache): Delete pattern.
13510 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13512 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13513 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13514 munaligned-access.
13516 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13518 PR target/83681
13519 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13520 by not USED_FOR_TARGET.
13521 (make_pass_resolve_sw_modes): Likewise.
13523 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13525 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13526 USED_FOR_TARGET.
13528 2018-01-08 Richard Biener <rguenther@suse.de>
13530 PR middle-end/83580
13531 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13533 2018-01-08 Richard Biener <rguenther@suse.de>
13535 PR middle-end/83517
13536 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13538 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13540 PR middle-end/81897
13541 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13542 basic blocks with a small number of successors.
13543 (convert_control_dep_chain_into_preds): Improve handling of
13544 forwarder blocks.
13545 (dump_predicates): Split apart into...
13546 (dump_pred_chain): ...here...
13547 (dump_pred_info): ...and here.
13548 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13549 (can_chain_union_be_invalidated_p): Improve check for invalidation
13550 of paths.
13551 (uninit_uses_cannot_happen): Avoid unnecessary if
13552 convert_control_dep_chain_into_preds yielded nothing.
13554 2018-01-06 Martin Sebor <msebor@redhat.com>
13556 PR tree-optimization/83640
13557 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13558 subtracting negative offset from size.
13559 (builtin_access::overlap): Adjust offset bounds of the access to fall
13560 within the size of the object if possible.
13562 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13564 PR rtl-optimization/83699
13565 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13566 extract_bit_field_as_subreg to cases in which the extracted
13567 value is also a vector.
13569 * lra-constraints.c (process_alt_operands): Test for the equivalence
13570 substitutions when detecting a possible reload cycle.
13572 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13574 PR debug/83480
13575 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13576 by default if flag_selective_schedling{,2}. Formatting fixes.
13578 PR rtl-optimization/83682
13579 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13580 if it has non-VECTOR_MODE element mode.
13581 (vec_duplicate_p): Likewise.
13583 PR middle-end/83694
13584 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13585 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13587 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13589 PR target/83604
13590 * config/i386/i386-builtin.def
13591 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13592 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13593 Require also OPTION_MASK_ISA_AVX512F in addition to
13594 OPTION_MASK_ISA_GFNI.
13595 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13596 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13597 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13598 to OPTION_MASK_ISA_GFNI.
13599 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13600 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13601 OPTION_MASK_ISA_AVX512BW.
13602 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13603 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13604 addition to OPTION_MASK_ISA_GFNI.
13605 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13606 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13607 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13608 to OPTION_MASK_ISA_GFNI.
13609 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13610 a requirement for all ISAs rather than any of them with a few
13611 exceptions.
13612 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13613 processing.
13614 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13615 bitmasks to be enabled with 3 exceptions, instead of requiring any
13616 enabled ISA with lots of exceptions.
13617 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13618 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13619 Change avx512bw in isa attribute to avx512f.
13620 * config/i386/sgxintrin.h: Add license boilerplate.
13621 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13622 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13623 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13624 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13625 defined.
13626 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13627 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13628 temporarily sse2 rather than sse if not enabled already.
13630 PR target/83604
13631 * config/i386/sse.md (VI248_VLBW): Rename to ...
13632 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13633 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13634 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13635 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13636 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13637 mode iterator instead of VI248_VLBW.
13639 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13641 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13642 (record_modified): Skip clobbers; add debug output.
13643 (param_change_prob): Use sreal frequencies.
13645 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13647 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13648 punt for user-aligned variables.
13650 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13652 * tree-chrec.c (chrec_contains_symbols): Return true for
13653 POLY_INT_CST.
13655 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13657 PR target/82439
13658 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13659 of (x|y) == x for BICS pattern.
13661 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13663 PR tree-optimization/83605
13664 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13665 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13666 can throw.
13668 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13670 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13671 * config/epiphany/rtems.h: New file.
13673 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13674 Uros Bizjak <ubizjak@gmail.com>
13676 PR target/83554
13677 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13678 QIreg_operand instead of register_operand predicate.
13679 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13680 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13681 comments instead of -fmitigate[-_]rop.
13683 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13685 PR bootstrap/81926
13686 * cgraphunit.c (symbol_table::compile): Switch to text_section
13687 before calling assembly_start debug hook.
13688 * run-rtl-passes.c (run_rtl_passes): Likewise.
13689 Include output.h.
13691 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13693 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13694 range_int_cst_p rather than !symbolic_range_p before calling
13695 extract_range_from_multiplicative_op_1.
13697 2018-01-04 Jeff Law <law@redhat.com>
13699 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13700 redundant test in assertion.
13702 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13704 * doc/rtl.texi: Document machine_mode wrapper classes.
13706 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13708 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13709 using tree_to_uhwi.
13711 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13713 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13714 the VEC_PERM_EXPR fold to fail.
13716 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13718 PR debug/83585
13719 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13720 to switched_sections.
13722 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13724 PR target/83680
13725 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13726 test for d.testing.
13728 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13730 PR target/83387
13731 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13732 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13734 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13736 PR debug/83666
13737 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13738 is BLKmode and bitpos not zero or mode change is needed.
13740 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13742 PR target/83675
13743 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13744 TARGET_VIS2.
13746 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13748 PR target/83628
13749 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13750 instead of MULT rtx. Update all corresponding splitters.
13751 (*saddl_se): Ditto.
13752 (*ssub<modesuffix>): Ditto.
13753 (*ssubl_se): Ditto.
13754 (*cmp_sadd_di): Update split patterns.
13755 (*cmp_sadd_si): Ditto.
13756 (*cmp_sadd_sidi): Ditto.
13757 (*cmp_ssub_di): Ditto.
13758 (*cmp_ssub_si): Ditto.
13759 (*cmp_ssub_sidi): Ditto.
13760 * config/alpha/predicates.md (const23_operand): New predicate.
13761 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13762 Look for ASHIFT, not MULT inner operand.
13763 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13765 2018-01-04 Martin Liska <mliska@suse.cz>
13767 PR gcov-profile/83669
13768 * gcov.c (output_intermediate_file): Add version to intermediate
13769 gcov file.
13770 * doc/gcov.texi: Document new field 'version' in intermediate
13771 file format. Fix location of '-k' option of gcov command.
13773 2018-01-04 Martin Liska <mliska@suse.cz>
13775 PR ipa/82352
13776 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13778 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13780 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13782 2018-01-03 Martin Sebor <msebor@redhat.com>
13784 PR tree-optimization/83655
13785 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13786 checking calls with invalid arguments.
13788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13790 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13791 (vectorizable_mask_load_store): Delete.
13792 (vectorizable_call): Return false for masked loads and stores.
13793 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13794 instead of gimple_assign_rhs1.
13795 (vectorizable_load): Handle IFN_MASK_LOAD.
13796 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13800 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13801 split out from..,
13802 (vectorizable_mask_load_store): ...here.
13803 (vectorizable_load): ...and here.
13805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13807 * tree-vect-stmts.c (vect_build_all_ones_mask)
13808 (vect_build_zero_merge_argument): New functions, split out from...
13809 (vectorizable_load): ...here.
13811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13813 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13814 split out from...
13815 (vectorizable_mask_load_store): ...here.
13816 (vectorizable_store): ...and here.
13818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13820 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13821 split out from...
13822 (vectorizable_mask_load_store): ...here.
13824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13826 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13827 (vect_model_store_cost): Take a vec_load_store_type instead of a
13828 vect_def_type.
13829 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13830 (vect_model_store_cost): Take a vec_load_store_type instead of a
13831 vect_def_type.
13832 (vectorizable_mask_load_store): Update accordingly.
13833 (vectorizable_store): Likewise.
13834 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13838 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13839 IFN_MASK_LOAD calls here rather than...
13840 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13843 Alan Hayward <alan.hayward@arm.com>
13844 David Sherwood <david.sherwood@arm.com>
13846 * expmed.c (extract_bit_field_1): For vector extracts,
13847 fall back to extract_bit_field_as_subreg if vec_extract
13848 isn't available.
13850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13851 Alan Hayward <alan.hayward@arm.com>
13852 David Sherwood <david.sherwood@arm.com>
13854 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13855 they are variable or constant sized.
13856 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13857 slots for constant-sized data.
13859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13860 Alan Hayward <alan.hayward@arm.com>
13861 David Sherwood <david.sherwood@arm.com>
13863 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13864 handling COND_EXPRs with boolean comparisons, try to find a better
13865 basis for the mask type than the boolean itself.
13867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13869 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13870 is calculated and how it can be overridden.
13871 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13872 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13873 if defined.
13874 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13875 if nonzero.
13877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13878 Alan Hayward <alan.hayward@arm.com>
13879 David Sherwood <david.sherwood@arm.com>
13881 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13882 Remove the mode argument.
13883 (aarch64_simd_valid_immediate): Remove the mode and inverse
13884 arguments.
13885 * config/aarch64/iterators.md (bitsize): New iterator.
13886 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13887 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13888 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13889 aarch64_simd_valid_immediate.
13890 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13891 (aarch64_reg_or_bic_imm): Likewise.
13892 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13893 with an insn_type enum and msl with a modifier_type enum.
13894 Replace element_width with a scalar_mode. Change the shift
13895 to unsigned int. Add constructors for scalar_float_mode and
13896 scalar_int_mode elements.
13897 (aarch64_vect_float_const_representable_p): Delete.
13898 (aarch64_can_const_movi_rtx_p)
13899 (aarch64_simd_scalar_immediate_valid_for_move)
13900 (aarch64_simd_make_constant): Update call to
13901 aarch64_simd_valid_immediate.
13902 (aarch64_advsimd_valid_immediate_hs): New function.
13903 (aarch64_advsimd_valid_immediate): Likewise.
13904 (aarch64_simd_valid_immediate): Remove mode and inverse
13905 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13906 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13907 and aarch64_float_const_representable_p on the result.
13908 (aarch64_output_simd_mov_immediate): Remove mode argument.
13909 Update call to aarch64_simd_valid_immediate and use of
13910 simd_immediate_info.
13911 (aarch64_output_scalar_simd_mov_immediate): Update call
13912 accordingly.
13914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13915 Alan Hayward <alan.hayward@arm.com>
13916 David Sherwood <david.sherwood@arm.com>
13918 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13919 (mode_nunits): Likewise CONST_MODE_NUNITS.
13920 * machmode.def (ADJUST_NUNITS): Document.
13921 * genmodes.c (mode_data::need_nunits_adj): New field.
13922 (blank_mode): Update accordingly.
13923 (adj_nunits): New variable.
13924 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13925 parameter.
13926 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13927 listed in adj_nunits.
13928 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13929 listed in adj_nunits. Don't emit case statements for such modes.
13930 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13931 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13932 nothing if adj_nunits is nonnull.
13933 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13934 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13935 (emit_mode_fbit): Update use of print_maybe_const_decl.
13936 (emit_move_size): Likewise. Treat the array as non-const
13937 if adj_nunits.
13938 (emit_mode_adjustments): Handle adj_nunits.
13940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13942 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13943 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13944 (VECTOR_MODES): Use it.
13945 (make_vector_modes): Take the prefix as an argument.
13947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13948 Alan Hayward <alan.hayward@arm.com>
13949 David Sherwood <david.sherwood@arm.com>
13951 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13952 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13953 for MODE_VECTOR_BOOL.
13954 * machmode.def (VECTOR_BOOL_MODE): Document.
13955 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13956 (make_vector_bool_mode): New function.
13957 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13958 MODE_VECTOR_BOOL.
13959 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13960 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13961 Likewise.
13962 * stor-layout.c (int_mode_for_mode): Likewise.
13963 * tree.c (build_vector_type_for_mode): Likewise.
13964 * varasm.c (output_constant_pool_2): Likewise.
13965 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13966 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13967 for MODE_VECTOR_BOOL.
13968 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13969 of mode class checks.
13970 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13971 instead of a list of mode class checks.
13972 (expand_vector_scalar_condition): Likewise.
13973 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13976 Alan Hayward <alan.hayward@arm.com>
13977 David Sherwood <david.sherwood@arm.com>
13979 * machmode.h (mode_size): Change from unsigned short to
13980 poly_uint16_pod.
13981 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13982 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13983 or if measurement_type is not polynomial.
13984 (fixed_size_mode::includes_p): Check for constant-sized modes.
13985 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13986 return a poly_uint16 rather than an unsigned short.
13987 (emit_mode_size): Change the type of mode_size from unsigned short
13988 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13989 (emit_mode_adjustments): Cope with polynomial vector sizes.
13990 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13991 for GET_MODE_SIZE.
13992 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13993 for GET_MODE_SIZE.
13994 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13995 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13996 * caller-save.c (setup_save_areas): Likewise.
13997 (replace_reg_with_saved_mem): Likewise.
13998 * calls.c (emit_library_call_value_1): Likewise.
13999 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14000 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14001 (gen_lowpart_for_combine): Likewise.
14002 * convert.c (convert_to_integer_1): Likewise.
14003 * cse.c (equiv_constant, cse_insn): Likewise.
14004 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14005 (cselib_subst_to_values): Likewise.
14006 * dce.c (word_dce_process_block): Likewise.
14007 * df-problems.c (df_word_lr_mark_ref): Likewise.
14008 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14009 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14010 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14011 (rtl_for_decl_location): Likewise.
14012 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14013 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14014 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14015 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14016 (expand_expr_real_1): Likewise.
14017 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14018 (pad_below): Likewise.
14019 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14020 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14021 * ira.c (get_subreg_tracking_sizes): Likewise.
14022 * ira-build.c (ira_create_allocno_objects): Likewise.
14023 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14024 (ira_sort_regnos_for_alter_reg): Likewise.
14025 * ira-costs.c (record_operand_costs): Likewise.
14026 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14027 (resolve_simple_move): Likewise.
14028 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14029 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14030 (lra_constraints): Likewise.
14031 (CONST_POOL_OK_P): Reject variable-sized modes.
14032 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14033 (add_pseudo_to_slot, lra_spill): Likewise.
14034 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14035 * optabs-query.c (get_best_extraction_insn): Likewise.
14036 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14037 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14038 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14039 * recog.c (offsettable_address_addr_space_p): Likewise.
14040 * regcprop.c (maybe_mode_change): Likewise.
14041 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14042 * regrename.c (build_def_use): Likewise.
14043 * regstat.c (dump_reg_info): Likewise.
14044 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14045 (find_reloads, find_reloads_subreg_address): Likewise.
14046 * reload1.c (eliminate_regs_1): Likewise.
14047 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14048 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14049 (simplify_binary_operation_1, simplify_subreg): Likewise.
14050 * targhooks.c (default_function_arg_padding): Likewise.
14051 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14052 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14053 (verify_gimple_assign_ternary): Likewise.
14054 * tree-inline.c (estimate_move_cost): Likewise.
14055 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14056 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14057 (get_address_cost_ainc): Likewise.
14058 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14059 (vect_supportable_dr_alignment): Likewise.
14060 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14061 (vectorizable_reduction): Likewise.
14062 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14063 (vectorizable_operation, vectorizable_load): Likewise.
14064 * tree.c (build_same_sized_truth_vector_type): Likewise.
14065 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14066 * var-tracking.c (emit_note_insn_var_location): Likewise.
14067 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14068 (ADDR_VEC_ALIGN): Likewise.
14070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14071 Alan Hayward <alan.hayward@arm.com>
14072 David Sherwood <david.sherwood@arm.com>
14074 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14075 unsigned short.
14076 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14077 or if measurement_type is polynomial.
14078 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14079 * combine.c (make_extraction): Likewise.
14080 * dse.c (find_shift_sequence): Likewise.
14081 * dwarf2out.c (mem_loc_descriptor): Likewise.
14082 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14083 (extract_bit_field, extract_low_bits): Likewise.
14084 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14085 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14086 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14087 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14088 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14089 * reload.c (find_reloads): Likewise.
14090 * reload1.c (alter_reg): Likewise.
14091 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14092 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14093 * tree-if-conv.c (predicate_mem_writes): Likewise.
14094 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14095 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14096 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14097 * valtrack.c (dead_debug_insert_temp): Likewise.
14098 * varasm.c (mergeable_constant_section): Likewise.
14099 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14102 Alan Hayward <alan.hayward@arm.com>
14103 David Sherwood <david.sherwood@arm.com>
14105 * expr.c (expand_assignment): Cope with polynomial mode sizes
14106 when assigning to a CONCAT.
14108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14109 Alan Hayward <alan.hayward@arm.com>
14110 David Sherwood <david.sherwood@arm.com>
14112 * machmode.h (mode_precision): Change from unsigned short to
14113 poly_uint16_pod.
14114 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14115 short.
14116 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14117 or if measurement_type is not polynomial.
14118 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14119 in which the mode is already known to be a scalar_int_mode.
14120 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14121 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14122 initializer.
14123 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14124 for GET_MODE_PRECISION.
14125 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14126 for GET_MODE_PRECISION.
14127 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14128 as polynomial.
14129 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14130 (expand_field_assignment, make_extraction): Likewise.
14131 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14132 (get_last_value): Likewise.
14133 * convert.c (convert_to_integer_1): Likewise.
14134 * cse.c (cse_insn): Likewise.
14135 * expr.c (expand_expr_real_1): Likewise.
14136 * lra-constraints.c (simplify_operand_subreg): Likewise.
14137 * optabs-query.c (can_atomic_load_p): Likewise.
14138 * optabs.c (expand_atomic_load): Likewise.
14139 (expand_atomic_store): Likewise.
14140 * ree.c (combine_reaching_defs): Likewise.
14141 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14142 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14143 * tree.h (type_has_mode_precision_p): Likewise.
14144 * ubsan.c (instrument_si_overflow): Likewise.
14146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14147 Alan Hayward <alan.hayward@arm.com>
14148 David Sherwood <david.sherwood@arm.com>
14150 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14151 polynomial numbers of units.
14152 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14153 (valid_vector_subparts_p): New function.
14154 (build_vector_type): Remove temporary shim and take the number
14155 of units as a poly_uint64 rather than an int.
14156 (build_opaque_vector_type): Take the number of units as a
14157 poly_uint64 rather than an int.
14158 * tree.c (build_vector_from_ctor): Handle polynomial
14159 TYPE_VECTOR_SUBPARTS.
14160 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14161 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14162 (build_vector_from_val): If the number of units is variable,
14163 use build_vec_duplicate_cst for constant operands and
14164 VEC_DUPLICATE_EXPR otherwise.
14165 (make_vector_type): Remove temporary is_constant ().
14166 (build_vector_type, build_opaque_vector_type): Take the number of
14167 units as a poly_uint64 rather than an int.
14168 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14169 VECTOR_CST_NELTS.
14170 * cfgexpand.c (expand_debug_expr): Likewise.
14171 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14172 (store_constructor, expand_expr_real_1): Likewise.
14173 (const_scalar_mask_from_tree): Likewise.
14174 * fold-const-call.c (fold_const_reduction): Likewise.
14175 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14176 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14177 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14178 (fold_relational_const): Likewise.
14179 (native_interpret_vector): Likewise. Change the size from an
14180 int to an unsigned int.
14181 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14182 TYPE_VECTOR_SUBPARTS.
14183 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14184 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14185 duplicating a non-constant operand into a variable-length vector.
14186 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14187 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14188 * ipa-icf.c (sem_variable::equals): Likewise.
14189 * match.pd: Likewise.
14190 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14191 * print-tree.c (print_node): Likewise.
14192 * stor-layout.c (layout_type): Likewise.
14193 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14194 * tree-cfg.c (verify_gimple_comparison): Likewise.
14195 (verify_gimple_assign_binary): Likewise.
14196 (verify_gimple_assign_ternary): Likewise.
14197 (verify_gimple_assign_single): Likewise.
14198 * tree-pretty-print.c (dump_generic_node): Likewise.
14199 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14200 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14201 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14202 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14203 (vect_shift_permute_load_chain): Likewise.
14204 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14205 (expand_vector_condition, optimize_vector_constructor): Likewise.
14206 (lower_vec_perm, get_compute_type): Likewise.
14207 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14208 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14209 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14210 (vect_recog_mask_conversion_pattern): Likewise.
14211 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14212 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14213 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14214 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14215 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14216 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14217 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14218 (supportable_widening_operation): Likewise.
14219 (supportable_narrowing_operation): Likewise.
14220 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14221 Likewise.
14222 * varasm.c (output_constant): Likewise.
14224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14225 Alan Hayward <alan.hayward@arm.com>
14226 David Sherwood <david.sherwood@arm.com>
14228 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14229 so that both the length == 3 and length != 3 cases set up their
14230 own permute vectors. Add comments explaining why we know the
14231 number of elements is constant.
14232 (vect_permute_load_chain): Likewise.
14234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14235 Alan Hayward <alan.hayward@arm.com>
14236 David Sherwood <david.sherwood@arm.com>
14238 * machmode.h (mode_nunits): Change from unsigned char to
14239 poly_uint16_pod.
14240 (ONLY_FIXED_SIZE_MODES): New macro.
14241 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14242 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14243 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14244 New typedefs.
14245 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14246 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14247 or if measurement_type is not polynomial.
14248 * genmodes.c (ZERO_COEFFS): New macro.
14249 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14250 poly_uint16.
14251 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14252 Use ZERO_COEFFS when emitting initializers.
14253 * data-streamer.h (bp_pack_poly_value): New function.
14254 (bp_unpack_poly_value): Likewise.
14255 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14256 for GET_MODE_NUNITS.
14257 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14258 for GET_MODE_NUNITS.
14259 * tree.c (make_vector_type): Remove temporary shim and make
14260 the real function take the number of units as a poly_uint64
14261 rather than an int.
14262 (build_vector_type_for_mode): Handle polynomial nunits.
14263 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14264 * emit-rtl.c (const_vec_series_p_1): Likewise.
14265 (gen_rtx_CONST_VECTOR): Likewise.
14266 * fold-const.c (test_vec_duplicate_folding): Likewise.
14267 * genrecog.c (validate_pattern): Likewise.
14268 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14269 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14270 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14271 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14272 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14273 * rtlanal.c (subreg_get_info): Likewise.
14274 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14275 (vect_grouped_load_supported): Likewise.
14276 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14277 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14278 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14279 (simplify_const_unary_operation, simplify_binary_operation_1)
14280 (simplify_const_binary_operation, simplify_ternary_operation)
14281 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14282 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14283 instead of CONST_VECTOR_NUNITS.
14284 * varasm.c (output_constant_pool_2): Likewise.
14285 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14286 explicit-encoded elements in the XVEC for variable-length vectors.
14288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14290 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14293 Alan Hayward <alan.hayward@arm.com>
14294 David Sherwood <david.sherwood@arm.com>
14296 * coretypes.h (fixed_size_mode): Declare.
14297 (fixed_size_mode_pod): New typedef.
14298 * builtins.h (target_builtins::x_apply_args_mode)
14299 (target_builtins::x_apply_result_mode): Change type to
14300 fixed_size_mode_pod.
14301 * builtins.c (apply_args_size, apply_result_size, result_vector)
14302 (expand_builtin_apply_args_1, expand_builtin_apply)
14303 (expand_builtin_return): Update accordingly.
14305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14307 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14308 * cselib.c (cselib_hash_rtx): Likewise.
14309 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14310 CONST_VECTOR encoding.
14312 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14313 Jeff Law <law@redhat.com>
14315 PR target/83641
14316 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14317 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14318 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14319 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14321 PR target/83641
14322 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14323 explicitly probe *sp in a noreturn function if there were any callee
14324 register saves or frame pointer is needed.
14326 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14328 PR debug/83621
14329 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14330 BLKmode for ternary, binary or unary expressions.
14332 PR debug/83645
14333 * var-tracking.c (delete_vta_debug_insn): New inline function.
14334 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14335 insns from get_insns () to NULL instead of each bb separately.
14336 Use delete_vta_debug_insn. No longer static.
14337 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14338 delete_vta_debug_insns callers.
14339 * rtl.h (delete_vta_debug_insns): Declare.
14340 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14341 instead of variable_tracking_main.
14343 2018-01-03 Martin Sebor <msebor@redhat.com>
14345 PR tree-optimization/83603
14346 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14347 arguments past the endof the argument list in functions declared
14348 without a prototype.
14349 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14350 Avoid checking when arguments are null.
14352 2018-01-03 Martin Sebor <msebor@redhat.com>
14354 PR c/83559
14355 * doc/extend.texi (attribute const): Fix a typo.
14356 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14357 issuing -Wsuggest-attribute for void functions.
14359 2018-01-03 Martin Sebor <msebor@redhat.com>
14361 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14362 offset_int::from instead of wide_int::to_shwi.
14363 (maybe_diag_overlap): Remove assertion.
14364 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14365 * gimple-ssa-sprintf.c (format_directive): Same.
14366 (parse_directive): Same.
14367 (sprintf_dom_walker::compute_format_length): Same.
14368 (try_substitute_return_value): Same.
14370 2018-01-03 Jeff Law <law@redhat.com>
14372 PR middle-end/83654
14373 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14374 non-constant residual for zero at runtime and avoid probing in
14375 that case. Reorganize code for trailing problem to mirror handling
14376 of the residual.
14378 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14380 PR tree-optimization/83501
14381 * tree-ssa-strlen.c (get_string_cst): New.
14382 (handle_char_store): Call get_string_cst.
14384 2018-01-03 Martin Liska <mliska@suse.cz>
14386 PR tree-optimization/83593
14387 * tree-ssa-strlen.c: Include tree-cfg.h.
14388 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14389 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14390 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14391 to false.
14392 (strlen_dom_walker::before_dom_children): Call
14393 gimple_purge_dead_eh_edges. Dump tranformation with details
14394 dump flags.
14395 (strlen_dom_walker::before_dom_children): Update call by adding
14396 new argument cleanup_eh.
14397 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14399 2018-01-03 Martin Liska <mliska@suse.cz>
14401 PR ipa/83549
14402 * cif-code.def (VARIADIC_THUNK): New enum value.
14403 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14404 thunks.
14406 2018-01-03 Jan Beulich <jbeulich@suse.com>
14408 * sse.md (mov<mode>_internal): Tighten condition for when to use
14409 vmovdqu<ssescalarsize> for TI and OI modes.
14411 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14413 Update copyright years.
14415 2018-01-03 Martin Liska <mliska@suse.cz>
14417 PR ipa/83594
14418 * ipa-visibility.c (function_and_variable_visibility): Skip
14419 functions with noipa attribure.
14421 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14423 * gcc.c (process_command): Update copyright notice dates.
14424 * gcov-dump.c (print_version): Ditto.
14425 * gcov.c (print_version): Ditto.
14426 * gcov-tool.c (print_version): Ditto.
14427 * gengtype.c (create_file): Ditto.
14428 * doc/cpp.texi: Bump @copying's copyright year.
14429 * doc/cppinternals.texi: Ditto.
14430 * doc/gcc.texi: Ditto.
14431 * doc/gccint.texi: Ditto.
14432 * doc/gcov.texi: Ditto.
14433 * doc/install.texi: Ditto.
14434 * doc/invoke.texi: Ditto.
14436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14438 * vector-builder.h (vector_builder::m_full_nelts): Change from
14439 unsigned int to poly_uint64.
14440 (vector_builder::full_nelts): Update prototype accordingly.
14441 (vector_builder::new_vector): Likewise.
14442 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14443 (vector_builder::operator ==): Likewise.
14444 (vector_builder::finalize): Likewise.
14445 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14446 Take the number of elements as a poly_uint64 rather than an
14447 unsigned int.
14448 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14449 from unsigned int to poly_uint64.
14450 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14451 (vec_perm_indices::new_vector): Likewise.
14452 (vec_perm_indices::length): Likewise.
14453 (vec_perm_indices::nelts_per_input): Likewise.
14454 (vec_perm_indices::input_nelts): Likewise.
14455 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14456 number of elements per input as a poly_uint64 rather than an
14457 unsigned int. Use the original encoding for variable-length
14458 vectors, rather than clamping each individual element.
14459 For the second and subsequent elements in each pattern,
14460 clamp the step and base before clamping their sum.
14461 (vec_perm_indices::series_p): Handle polynomial element counts.
14462 (vec_perm_indices::all_in_range_p): Likewise.
14463 (vec_perm_indices_to_tree): Likewise.
14464 (vec_perm_indices_to_rtx): Likewise.
14465 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14466 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14467 (tree_vector_builder::new_binary_operation): Handle polynomial
14468 element counts. Return false if we need to know the number
14469 of elements at compile time.
14470 * fold-const.c (fold_vec_perm): Punt if the number of elements
14471 isn't known at compile time.
14473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14475 * vec-perm-indices.h (vec_perm_builder): Change element type
14476 from HOST_WIDE_INT to poly_int64.
14477 (vec_perm_indices::element_type): Update accordingly.
14478 (vec_perm_indices::clamp): Handle polynomial element_types.
14479 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14480 (vec_perm_indices::all_in_range_p): Likewise.
14481 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14482 than shwi trees.
14483 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14484 polynomial vec_perm_indices element types.
14485 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14486 * fold-const.c (fold_vec_perm): Likewise.
14487 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14488 * tree-vect-generic.c (lower_vec_perm): Likewise.
14489 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14490 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14491 element type to HOST_WIDE_INT.
14493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14494 Alan Hayward <alan.hayward@arm.com>
14495 David Sherwood <david.sherwood@arm.com>
14497 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14498 rather than an int. Use plus_constant.
14499 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14500 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14503 Alan Hayward <alan.hayward@arm.com>
14504 David Sherwood <david.sherwood@arm.com>
14506 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14507 a HOST_WIDE_INT to a poly_int64.
14509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14510 Alan Hayward <alan.hayward@arm.com>
14511 David Sherwood <david.sherwood@arm.com>
14513 * calls.c (load_register_parameters): Cope with polynomial
14514 mode sizes. Require a constant size for BLKmode parameters
14515 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14516 forces a parameter to be padded at the lsb end in order to
14517 fill a complete number of words, require the parameter size
14518 to be ordered wrt UNITS_PER_WORD.
14520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14521 Alan Hayward <alan.hayward@arm.com>
14522 David Sherwood <david.sherwood@arm.com>
14524 * reload1.c (spill_stack_slot_width): Change element type
14525 from unsigned int to poly_uint64_pod.
14526 (alter_reg): Treat mode sizes as polynomial.
14528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14529 Alan Hayward <alan.hayward@arm.com>
14530 David Sherwood <david.sherwood@arm.com>
14532 * reload.c (complex_word_subreg_p): New function.
14533 (reload_inner_reg_of_subreg, push_reload): Use it.
14535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14536 Alan Hayward <alan.hayward@arm.com>
14537 David Sherwood <david.sherwood@arm.com>
14539 * lra-constraints.c (process_alt_operands): Reject matched
14540 operands whose sizes aren't ordered.
14541 (match_reload): Refer to this check here.
14543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14544 Alan Hayward <alan.hayward@arm.com>
14545 David Sherwood <david.sherwood@arm.com>
14547 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14548 that the mode size is in the set {1, 2, 4, 8, 16}.
14550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14551 Alan Hayward <alan.hayward@arm.com>
14552 David Sherwood <david.sherwood@arm.com>
14554 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14555 Use plus_constant instead of gen_rtx_PLUS.
14557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14558 Alan Hayward <alan.hayward@arm.com>
14559 David Sherwood <david.sherwood@arm.com>
14561 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14562 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14563 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14564 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14565 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14566 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14567 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14568 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14569 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14570 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14571 a poly_int64.
14572 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14573 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14574 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14575 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14576 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14577 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14578 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14579 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14580 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14581 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14582 function.
14583 * expr.c (emit_move_resolve_push): Treat the input and result
14584 of PUSH_ROUNDING as a poly_int64.
14585 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14586 (emit_push_insn): Likewise.
14587 * lra-eliminations.c (mark_not_eliminable): Likewise.
14588 * recog.c (push_operand): Likewise.
14589 * reload1.c (elimination_effects): Likewise.
14590 * rtlanal.c (nonzero_bits1): Likewise.
14591 * calls.c (store_one_arg): Likewise. Require the padding to be
14592 known at compile time.
14594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14595 Alan Hayward <alan.hayward@arm.com>
14596 David Sherwood <david.sherwood@arm.com>
14598 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14599 Use plus_constant instead of gen_rtx_PLUS.
14601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14602 Alan Hayward <alan.hayward@arm.com>
14603 David Sherwood <david.sherwood@arm.com>
14605 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14606 rather than an int.
14608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14609 Alan Hayward <alan.hayward@arm.com>
14610 David Sherwood <david.sherwood@arm.com>
14612 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14613 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14614 via stack temporaries. Treat the mode size as polynomial too.
14616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14617 Alan Hayward <alan.hayward@arm.com>
14618 David Sherwood <david.sherwood@arm.com>
14620 * expr.c (expand_expr_real_2): When handling conversions involving
14621 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14622 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14623 as a poly_uint64 too.
14625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14626 Alan Hayward <alan.hayward@arm.com>
14627 David Sherwood <david.sherwood@arm.com>
14629 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14632 Alan Hayward <alan.hayward@arm.com>
14633 David Sherwood <david.sherwood@arm.com>
14635 * combine.c (can_change_dest_mode): Handle polynomial
14636 REGMODE_NATURAL_SIZE.
14637 * expmed.c (store_bit_field_1): Likewise.
14638 * expr.c (store_constructor): Likewise.
14639 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14640 and polynomial REGMODE_NATURAL_SIZE.
14641 (gen_lowpart_common): Likewise.
14642 * reginfo.c (record_subregs_of_mode): Likewise.
14643 * rtlanal.c (read_modify_subreg_p): Likewise.
14645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14646 Alan Hayward <alan.hayward@arm.com>
14647 David Sherwood <david.sherwood@arm.com>
14649 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14650 numbers of elements.
14652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14653 Alan Hayward <alan.hayward@arm.com>
14654 David Sherwood <david.sherwood@arm.com>
14656 * match.pd: Cope with polynomial numbers of vector elements.
14658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14659 Alan Hayward <alan.hayward@arm.com>
14660 David Sherwood <david.sherwood@arm.com>
14662 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14663 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
14670 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14671 (ipa_simd_modify_function_body): Likewise.
14673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14674 Alan Hayward <alan.hayward@arm.com>
14675 David Sherwood <david.sherwood@arm.com>
14677 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14678 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14679 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14680 (expand_vector_condition, vector_element): Likewise.
14681 (subparts_gt): New function.
14682 (get_compute_type): Use subparts_gt.
14683 (count_type_subparts): Delete.
14684 (expand_vector_operations_1): Use subparts_gt instead of
14685 count_type_subparts.
14687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14688 Alan Hayward <alan.hayward@arm.com>
14689 David Sherwood <david.sherwood@arm.com>
14691 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14692 (vect_compile_time_alias): ...this new function. Do the calculation
14693 on poly_ints rather than trees.
14694 (vect_prune_runtime_alias_test_list): Update call accordingly.
14696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14697 Alan Hayward <alan.hayward@arm.com>
14698 David Sherwood <david.sherwood@arm.com>
14700 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14701 numbers of units.
14702 (vect_schedule_slp_instance): Likewise.
14704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14705 Alan Hayward <alan.hayward@arm.com>
14706 David Sherwood <david.sherwood@arm.com>
14708 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14709 constant and extern definitions for variable-length vectors.
14710 (vect_get_constant_vectors): Note that the number of units
14711 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
14718 of units as polynomial. Choose between WIDE and NARROW based
14719 on multiple_p.
14721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14722 Alan Hayward <alan.hayward@arm.com>
14723 David Sherwood <david.sherwood@arm.com>
14725 * tree-vect-stmts.c (simd_clone_subparts): New function.
14726 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14729 Alan Hayward <alan.hayward@arm.com>
14730 David Sherwood <david.sherwood@arm.com>
14732 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14733 vectors as polynomial. Use build_index_vector for
14734 IFN_GOMP_SIMD_LANE.
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-stmts.c (get_load_store_type): Treat the number of
14741 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14742 for variable-length vectors.
14743 (vectorizable_mask_load_store): Treat the number of units as
14744 polynomial, asserting that it is constant if the condition has
14745 already been enforced.
14746 (vectorizable_store, vectorizable_load): Likewise.
14748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14749 Alan Hayward <alan.hayward@arm.com>
14750 David Sherwood <david.sherwood@arm.com>
14752 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14753 of units as polynomial. Punt if we can't tell at compile time
14754 which vector contains the final result.
14756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14757 Alan Hayward <alan.hayward@arm.com>
14758 David Sherwood <david.sherwood@arm.com>
14760 * tree-vect-loop.c (vectorizable_induction): Treat the number
14761 of units as polynomial. Punt on SLP inductions. Use an integer
14762 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14763 cast of such a series for variable-length floating-point
14764 reductions.
14766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14767 Alan Hayward <alan.hayward@arm.com>
14768 David Sherwood <david.sherwood@arm.com>
14770 * tree.h (build_index_vector): Declare.
14771 * tree.c (build_index_vector): New function.
14772 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14773 of units as polynomial, forcibly converting it to a constant if
14774 vectorizable_reduction has already enforced the condition.
14775 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14776 to create a {1,2,3,...} vector.
14777 (vectorizable_reduction): Treat the number of units as polynomial.
14778 Choose vectype_in based on the largest scalar element size rather
14779 than the smallest number of units. Enforce the restrictions
14780 relied on above.
14782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14783 Alan Hayward <alan.hayward@arm.com>
14784 David Sherwood <david.sherwood@arm.com>
14786 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14787 number of units as polynomial.
14789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14790 Alan Hayward <alan.hayward@arm.com>
14791 David Sherwood <david.sherwood@arm.com>
14793 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14794 * target.def (autovectorize_vector_sizes): Return the vector sizes
14795 by pointer, using vector_sizes rather than a bitmask.
14796 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14797 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14798 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14799 Likewise.
14800 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14801 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14802 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14803 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14804 * omp-general.c (omp_max_vf): Likewise.
14805 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14806 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14807 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14808 * tree-vect-slp.c (vect_slp_bb): Likewise.
14809 * doc/tm.texi: Regenerate.
14810 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14811 to a poly_uint64.
14812 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14813 the vector size as a poly_uint64 rather than an unsigned int.
14814 (current_vector_size): Change from an unsigned int to a poly_uint64.
14815 (get_vectype_for_scalar_type): Update accordingly.
14816 * tree.h (build_truth_vector_type): Take the size and number of
14817 units as a poly_uint64 rather than an unsigned int.
14818 (build_vector_type): Add a temporary overload that takes
14819 the number of units as a poly_uint64 rather than an unsigned int.
14820 * tree.c (make_vector_type): Likewise.
14821 (build_truth_vector_type): Take the number of units as a poly_uint64
14822 rather than an unsigned int.
14824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14825 Alan Hayward <alan.hayward@arm.com>
14826 David Sherwood <david.sherwood@arm.com>
14828 * target.def (get_mask_mode): Take the number of units and length
14829 as poly_uint64s rather than unsigned ints.
14830 * targhooks.h (default_get_mask_mode): Update accordingly.
14831 * targhooks.c (default_get_mask_mode): Likewise.
14832 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14833 * doc/tm.texi: Regenerate.
14835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14836 Alan Hayward <alan.hayward@arm.com>
14837 David Sherwood <david.sherwood@arm.com>
14839 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14840 * omp-general.c (omp_max_vf): Likewise.
14841 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14842 (expand_omp_simd): Handle polynomial safelen.
14843 * omp-low.c (omplow_simd_context): Add a default constructor.
14844 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14845 (lower_rec_simd_input_clauses): Update accordingly.
14846 (lower_rec_input_clauses): Likewise.
14848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14849 Alan Hayward <alan.hayward@arm.com>
14850 David Sherwood <david.sherwood@arm.com>
14852 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14853 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14854 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14855 (vect_analyze_slp_cost): Likewise.
14856 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14857 (vect_model_load_cost): Likewise.
14859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14860 Alan Hayward <alan.hayward@arm.com>
14861 David Sherwood <david.sherwood@arm.com>
14863 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14864 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14865 from an unsigned int * to a poly_uint64_pod *.
14866 (calculate_unrolling_factor): New function.
14867 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14870 Alan Hayward <alan.hayward@arm.com>
14871 David Sherwood <david.sherwood@arm.com>
14873 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14874 from an unsigned int to a poly_uint64.
14875 (_loop_vec_info::slp_unrolling_factor): Likewise.
14876 (_loop_vec_info::vectorization_factor): Change from an int
14877 to a poly_uint64.
14878 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14879 (vect_get_num_vectors): New function.
14880 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14881 (vect_get_num_copies): Use vect_get_num_vectors.
14882 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14883 to an unsigned int *.
14884 (vect_analyze_data_refs): Change min_vf from an int * to a
14885 poly_uint64 *.
14886 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14887 than an unsigned HOST_WIDE_INT.
14888 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14889 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14890 to an unsigned int *.
14891 (vect_analyze_data_ref_dependences): Likewise.
14892 (vect_compute_data_ref_alignment): Handle polynomial vf.
14893 (vect_enhance_data_refs_alignment): Likewise.
14894 (vect_prune_runtime_alias_test_list): Likewise.
14895 (vect_shift_permute_load_chain): Likewise.
14896 (vect_supportable_dr_alignment): Likewise.
14897 (dependence_distance_ge_vf): Take the vectorization factor as a
14898 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14899 (vect_analyze_data_refs): Change min_vf from an int * to a
14900 poly_uint64 *.
14901 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14902 vfm1 as a poly_uint64 rather than an int. Make the same change
14903 for the returned bound_scalar.
14904 (vect_gen_vector_loop_niters): Handle polynomial vf.
14905 (vect_do_peeling): Likewise. Update call to
14906 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14907 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14908 be constant.
14909 * tree-vect-loop.c (vect_determine_vectorization_factor)
14910 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14911 (vect_get_known_peeling_cost): Likewise.
14912 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14913 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14914 (vect_transform_loop): Likewise. Use the lowest possible VF when
14915 updating the upper bounds of the loop.
14916 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14917 rather than an int.
14918 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14919 polynomial unroll factors.
14920 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14921 (vect_make_slp_decision): Likewise.
14922 (vect_supported_load_permutation_p): Likewise, and polynomial
14923 vf too.
14924 (vect_analyze_slp_cost): Handle polynomial vf.
14925 (vect_slp_analyze_node_operations): Likewise.
14926 (vect_slp_analyze_bb_1): Likewise.
14927 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14928 than an unsigned HOST_WIDE_INT.
14929 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14930 (vectorizable_load): Handle polynomial vf.
14931 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14932 a poly_uint64.
14933 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14936 Alan Hayward <alan.hayward@arm.com>
14937 David Sherwood <david.sherwood@arm.com>
14939 * match.pd: Handle bit operations involving three constants
14940 and try to fold one pair.
14942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14944 * tree-vect-loop-manip.c: Include gimple-fold.h.
14945 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14946 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14947 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14948 Add a path that uses a step of VF instead of 1, but disable it
14949 for now.
14950 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14951 and niters_no_overflow parameters. Update calls to
14952 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14953 Create a new SSA name if the latter choses to use a ste other
14954 than zero, and return it via niters_vector_mult_vf_var.
14955 * tree-vect-loop.c (vect_transform_loop): Update calls to
14956 vect_do_peeling, vect_gen_vector_loop_niters and
14957 slpeel_make_loop_iterate_ntimes.
14958 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14959 (vect_gen_vector_loop_niters): Update declarations after above changes.
14961 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14963 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14964 128-bit round to integer instructions.
14965 (ceil<mode>2): Likewise.
14966 (btrunc<mode>2): Likewise.
14967 (round<mode>2): Likewise.
14969 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14971 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14972 unaligned VSX load/store on P8/P9.
14973 (expand_block_clear): Allow the use of unaligned VSX
14974 load/store on P8/P9.
14976 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14978 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14979 New function.
14980 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14981 swap associated with both a load and a store.
14983 2018-01-02 Andrew Waterman <andrew@sifive.com>
14985 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14986 * config/riscv/riscv.md (clear_cache): Use it.
14988 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14990 * web.c: Remove out-of-date comment.
14992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14994 * expr.c (fixup_args_size_notes): Check that any existing
14995 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14996 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14997 (emit_single_push_insn): ...here.
14999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15001 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15002 (const_vector_encoded_nelts): New function.
15003 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15004 (const_vector_int_elt, const_vector_elt): Declare.
15005 * emit-rtl.c (const_vector_int_elt_1): New function.
15006 (const_vector_elt): Likewise.
15007 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15008 of CONST_VECTOR_ELT.
15010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15012 * expr.c: Include rtx-vector-builder.h.
15013 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15014 directly on the tree encoding.
15015 (const_vector_from_tree): Likewise.
15016 * optabs.c: Include rtx-vector-builder.h.
15017 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15018 sequence of "u" values.
15019 * vec-perm-indices.c: Include rtx-vector-builder.h.
15020 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15021 directly on the vec_perm_indices encoding.
15023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15025 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15026 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15027 * rtx-vector-builder.h: New file.
15028 * rtx-vector-builder.c: Likewise.
15029 * rtl.h (rtx_def::u2): Add a const_vector field.
15030 (CONST_VECTOR_NPATTERNS): New macro.
15031 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15032 (CONST_VECTOR_DUPLICATE_P): Likewise.
15033 (CONST_VECTOR_STEPPED_P): Likewise.
15034 (CONST_VECTOR_ENCODED_ELT): Likewise.
15035 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15036 (unwrap_const_vec_duplicate): Likewise.
15037 (const_vec_series_p): Check for a non-duplicated vector encoding.
15038 Say that the function only returns true for integer vectors.
15039 * emit-rtl.c: Include rtx-vector-builder.h.
15040 (gen_const_vec_duplicate_1): Delete.
15041 (gen_const_vector): Call gen_const_vec_duplicate instead of
15042 gen_const_vec_duplicate_1.
15043 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15044 (gen_const_vec_duplicate): Use rtx_vector_builder.
15045 (gen_const_vec_series): Likewise.
15046 (gen_rtx_CONST_VECTOR): Likewise.
15047 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15048 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15049 Build a new vector rather than modifying a CONST_VECTOR in-place.
15050 (handle_special_swappables): Update call accordingly.
15051 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15052 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15053 Build a new vector rather than modifying a CONST_VECTOR in-place.
15054 (handle_special_swappables): Update call accordingly.
15056 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15058 * simplify-rtx.c (simplify_const_binary_operation): Use
15059 CONST_VECTOR_ELT instead of XVECEXP.
15061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15063 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15064 the selector elements to be different from the data elements
15065 if the selector is a VECTOR_CST.
15066 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15067 ssizetype for the selector.
15069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15071 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15072 before testing each element individually.
15073 * tree-vect-generic.c (lower_vec_perm): Likewise.
15075 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15077 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15078 * selftest-run-tests.c (selftest::run_tests): Call it.
15079 * vector-builder.h (vector_builder::operator ==): New function.
15080 (vector_builder::operator !=): Likewise.
15081 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15082 (vec_perm_indices::all_from_input_p): New function.
15083 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15084 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15085 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15086 instead of reading the VECTOR_CST directly. Detect whether both
15087 vector inputs are the same before constructing the vec_perm_indices,
15088 and update the number of inputs argument accordingly. Use the
15089 utility functions added above. Only construct sel2 if we need to.
15091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15093 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15094 the broadcast of the low byte.
15095 (expand_mult_highpart): Use an explicit encoding for the permutes.
15096 * optabs-query.c (can_mult_highpart_p): Likewise.
15097 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15098 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15099 (vectorizable_bswap): Likewise.
15100 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15101 explicit encoding for the power-of-2 permutes.
15102 (vect_permute_store_chain): Likewise.
15103 (vect_grouped_load_supported): Likewise.
15104 (vect_permute_load_chain): Likewise.
15106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15108 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15109 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15110 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15111 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15112 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15113 (vect_gen_perm_mask_any): Likewise.
15115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15117 * int-vector-builder.h: New file.
15118 * vec-perm-indices.h: Include int-vector-builder.h.
15119 (vec_perm_indices): Redefine as an int_vector_builder.
15120 (auto_vec_perm_indices): Delete.
15121 (vec_perm_builder): Redefine as a stand-alone class.
15122 (vec_perm_indices::vec_perm_indices): New function.
15123 (vec_perm_indices::clamp): Likewise.
15124 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15125 (vec_perm_indices::new_vector): New function.
15126 (vec_perm_indices::new_expanded_vector): Update for new
15127 vec_perm_indices class.
15128 (vec_perm_indices::rotate_inputs): New function.
15129 (vec_perm_indices::all_in_range_p): Operate directly on the
15130 encoded form, without computing elided elements.
15131 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15132 encoding. Update for new vec_perm_indices class.
15133 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15134 the given vec_perm_builder.
15135 (expand_vec_perm_var): Update vec_perm_builder constructor.
15136 (expand_mult_highpart): Use vec_perm_builder instead of
15137 auto_vec_perm_indices.
15138 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15139 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15140 or double series encoding as appropriate.
15141 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15142 vec_perm_indices instead of auto_vec_perm_indices.
15143 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15144 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15145 (vect_permute_store_chain): Likewise.
15146 (vect_grouped_load_supported): Likewise.
15147 (vect_permute_load_chain): Likewise.
15148 (vect_shift_permute_load_chain): Likewise.
15149 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15150 (vect_transform_slp_perm_load): Likewise.
15151 (vect_schedule_slp_instance): Likewise.
15152 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15153 (vectorizable_mask_load_store): Likewise.
15154 (vectorizable_bswap): Likewise.
15155 (vectorizable_store): Likewise.
15156 (vectorizable_load): Likewise.
15157 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15158 vec_perm_indices instead of auto_vec_perm_indices. Use
15159 tree_to_vec_perm_builder to read the vector from a tree.
15160 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15161 vec_perm_builder instead of a vec_perm_indices.
15162 (have_whole_vector_shift): Use vec_perm_builder and
15163 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15164 truncation to calc_vec_perm_mask_for_shift.
15165 (vect_create_epilog_for_reduction): Likewise.
15166 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15167 from auto_vec_perm_indices to vec_perm_indices.
15168 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15169 instead of changing individual elements.
15170 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15171 the vector in d.perm.
15172 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15173 from auto_vec_perm_indices to vec_perm_indices.
15174 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15175 instead of changing individual elements.
15176 (arm_vectorize_vec_perm_const): Use new_vector to install
15177 the vector in d.perm.
15178 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15179 Update vec_perm_builder constructor.
15180 (rs6000_expand_interleave): Likewise.
15181 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15182 (rs6000_expand_interleave): Likewise.
15184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15186 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15187 to qimode could truncate the indices.
15188 * optabs.c (expand_vec_perm_var): Likewise.
15190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15192 * Makefile.in (OBJS): Add vec-perm-indices.o.
15193 * vec-perm-indices.h: New file.
15194 * vec-perm-indices.c: Likewise.
15195 * target.h (vec_perm_indices): Replace with a forward class
15196 declaration.
15197 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15198 * optabs.h: Include vec-perm-indices.h.
15199 (expand_vec_perm): Delete.
15200 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15201 (expand_vec_perm_const): Declare.
15202 * target.def (vec_perm_const_ok): Replace with...
15203 (vec_perm_const): ...this new hook.
15204 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15205 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15206 * doc/tm.texi: Regenerate.
15207 * optabs.def (vec_perm_const): Delete.
15208 * doc/md.texi (vec_perm_const): Likewise.
15209 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15210 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15211 expand_vec_perm for constant permutation vectors. Assert that
15212 the mode of variable permutation vectors is the integer equivalent
15213 of the mode that is being permuted.
15214 * optabs-query.h (selector_fits_mode_p): Declare.
15215 * optabs-query.c: Include vec-perm-indices.h.
15216 (selector_fits_mode_p): New function.
15217 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15218 is defined, instead of checking whether the vec_perm_const_optab
15219 exists. Use targetm.vectorize.vec_perm_const instead of
15220 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15221 fit in the vector mode before using a variable permute.
15222 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15223 vec_perm_indices instead of an rtx.
15224 (expand_vec_perm): Replace with...
15225 (expand_vec_perm_const): ...this new function. Take the selector
15226 as a vec_perm_indices rather than an rtx. Also take the mode of
15227 the selector. Update call to shift_amt_for_vec_perm_mask.
15228 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15229 Use vec_perm_indices::new_expanded_vector to expand the original
15230 selector into bytes. Check whether the indices fit in the vector
15231 mode before using a variable permute.
15232 (expand_vec_perm_var): Make global.
15233 (expand_mult_highpart): Use expand_vec_perm_const.
15234 * fold-const.c: Includes vec-perm-indices.h.
15235 * tree-ssa-forwprop.c: Likewise.
15236 * tree-vect-data-refs.c: Likewise.
15237 * tree-vect-generic.c: Likewise.
15238 * tree-vect-loop.c: Likewise.
15239 * tree-vect-slp.c: Likewise.
15240 * tree-vect-stmts.c: Likewise.
15241 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15242 Delete.
15243 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15244 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15245 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15246 (aarch64_vectorize_vec_perm_const): ...this new function.
15247 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15248 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15249 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15250 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15251 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15252 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15253 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15254 into...
15255 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15256 check for NEON modes.
15257 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15258 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15259 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15260 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15261 into...
15262 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15263 the old VEC_PERM_CONST conditions.
15264 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15265 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15266 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15267 (ia64_vectorize_vec_perm_const_ok): Merge into...
15268 (ia64_vectorize_vec_perm_const): ...this new function.
15269 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15270 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15271 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15272 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15273 * config/mips/mips.c (mips_expand_vec_perm_const)
15274 (mips_vectorize_vec_perm_const_ok): Merge into...
15275 (mips_vectorize_vec_perm_const): ...this new function.
15276 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15277 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15278 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15279 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15280 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15281 (rs6000_expand_vec_perm_const): Delete.
15282 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15283 Delete.
15284 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15285 (altivec_expand_vec_perm_const_le): Take each operand individually.
15286 Operate on constant selectors rather than rtxes.
15287 (altivec_expand_vec_perm_const): Likewise. Update call to
15288 altivec_expand_vec_perm_const_le.
15289 (rs6000_expand_vec_perm_const): Delete.
15290 (rs6000_vectorize_vec_perm_const_ok): Delete.
15291 (rs6000_vectorize_vec_perm_const): New function.
15292 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15293 an element count and rtx array.
15294 (rs6000_expand_extract_even): Update call accordingly.
15295 (rs6000_expand_interleave): Likewise.
15296 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15297 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15298 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15299 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15300 (rs6000_expand_vec_perm_const): Delete.
15301 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15302 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15303 (altivec_expand_vec_perm_const_le): Take each operand individually.
15304 Operate on constant selectors rather than rtxes.
15305 (altivec_expand_vec_perm_const): Likewise. Update call to
15306 altivec_expand_vec_perm_const_le.
15307 (rs6000_expand_vec_perm_const): Delete.
15308 (rs6000_vectorize_vec_perm_const_ok): Delete.
15309 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15310 reference to the SPE evmerge intructions.
15311 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15312 an element count and rtx array.
15313 (rs6000_expand_extract_even): Update call accordingly.
15314 (rs6000_expand_interleave): Likewise.
15315 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15316 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15317 new function.
15318 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15320 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15322 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15323 vector mode and that that mode matches the mode of the data
15324 being permuted.
15325 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15326 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15327 directly using expand_vec_perm_1 when forcing selectors into
15328 registers.
15329 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15333 * optabs-query.h (can_vec_perm_p): Delete.
15334 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15335 * optabs-query.c (can_vec_perm_p): Split into...
15336 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15337 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15338 particular selector is valid.
15339 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15340 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15341 (vect_grouped_load_supported): Likewise.
15342 (vect_shift_permute_load_chain): Likewise.
15343 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15344 (vect_transform_slp_perm_load): Likewise.
15345 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15346 (vectorizable_bswap): Likewise.
15347 (vect_gen_perm_mask_checked): Likewise.
15348 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15349 implementations of variable permutation vectors into account
15350 when deciding which selector to use.
15351 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15352 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15353 with a false third argument.
15354 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15355 to test whether the constant selector is valid and can_vec_perm_var_p
15356 to test whether a variable selector is valid.
15358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15360 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15361 * optabs-query.c (can_vec_perm_p): Likewise.
15362 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15363 instead of vec_perm_indices.
15364 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15365 (vect_gen_perm_mask_checked): Likewise,
15366 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15367 (vect_gen_perm_mask_checked): Likewise,
15369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15371 * optabs-query.h (qimode_for_vec_perm): Declare.
15372 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15373 (qimode_for_vec_perm): ...this new function.
15374 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15376 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15378 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15379 does not have a conditional at the top.
15381 2018-01-02 Richard Biener <rguenther@suse.de>
15383 * ipa-inline.c (big_speedup_p): Fix expression.
15385 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15387 PR target/81616
15388 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15389 for generic 4->6.
15391 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15393 PR target/81616
15394 Generic tuning.
15395 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15396 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15397 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15398 cond_taken_branch_cost 3->4.
15400 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15402 PR tree-optimization/83581
15403 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15404 TODO_cleanup_cfg if any changes have been made.
15406 PR middle-end/83608
15407 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15408 convert_modes if target mode has the right side, but different mode
15409 class.
15411 PR middle-end/83609
15412 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15413 last argument when extracting from CONCAT. If either from_real or
15414 from_imag is NULL, use expansion through memory. If result is not
15415 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15416 the parts directly to inner mode, if even that fails, use expansion
15417 through memory.
15419 PR middle-end/83623
15420 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15421 check for bswap in mode rather than HImode and use that in expand_unop
15422 too.
15423 \f
15424 Copyright (C) 2018 Free Software Foundation, Inc.
15426 Copying and distribution of this file, with or without modification,
15427 are permitted in any medium without royalty provided the copyright
15428 notice and this notice are preserved.