1 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9 * doc/sourcebuild.texi (Endianness): New subsubsection.
11 2018-05-23 Luis Machado <luis.machado@linaro.org>
13 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14 <prefetch_dynamic_strides>: New const bool field.
15 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16 prefetch_dynamic_strides.
17 (exynosm1_prefetch_tune): Likewise.
18 (thunderxt88_prefetch_tune): Likewise.
19 (thunderx_prefetch_tune): Likewise.
20 (thunderx2t99_prefetch_tune): Likewise.
21 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
23 (aarch64_override_options_internal): Update to set
24 PARAM_PREFETCH_DYNAMIC_STRIDES.
25 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
26 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
27 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
28 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
29 prefetch-dynamic-strides setting.
31 2018-05-23 Luis Machado <luis.machado@linaro.org>
33 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
34 <minimum_stride>: New const int field.
35 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
36 minimum_stride field defaulting to -1.
37 (exynosm1_prefetch_tune): Likewise.
38 (thunderxt88_prefetch_tune): Likewise.
39 (thunderx_prefetch_tune): Likewise.
40 (thunderx2t99_prefetch_tune): Likewise.
41 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
42 <default_opt_level>: Set to 3.
43 (aarch64_override_options_internal): Update to set
44 PARAM_PREFETCH_MINIMUM_STRIDE.
45 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
46 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
47 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
48 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
49 stride is constant and is below the minimum stride threshold.
51 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
53 * config/arm/arm-cpus.in (mode26): Delete.
54 (armv4): Delete mode26 reference.
55 * config/arm/arm.c (arm_configure_build_target): Delete use of
58 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
60 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
62 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
63 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
65 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
66 (floatunsdidf2): Ditto.
68 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
70 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
71 (fixuns_trunc<mode>si2_avx512f): Ditto.
72 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
73 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
74 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
76 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
78 PR rtl-optimization/79985
79 * df-scan.c (df_insn_refs_collect): Remove special case for
80 global registers and asm statements.
82 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
84 * extend.texi (Global Register Variables): Rewrite the bullet list.
85 Note that the register is available for allocation. Note that access
86 via inline asm must use constraints. Add note about async-signal
87 handlers. Remove paragraph about automagic register selection.
89 2018-05-23 Richard Biener <rguenther@suse.de>
91 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
92 of fixed offset from memset VN.
94 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
96 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
98 (alloc_cand_and_find_basis): Initialize first_interp field.
99 (slsr_process_mul): Modify first_interp field.
100 (slsr_process_add): Likewise.
101 (slsr_process_cast): Modify first_interp field for each new
103 (slsr_process_copy): Likewise.
104 (dump_candidate): Dump first_interp field.
105 (replace_mult_candidate): Process all interpretations, not just
107 (replace_rhs_if_not_dup): Likewise.
108 (replace_one_candidate): Likewise.
110 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
112 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
114 (aarch64_needs_frame_chain): New function.
115 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
117 2018-05-23 Sudakshina Das <sudi.das@arm.com>
120 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
121 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
122 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
123 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
124 as true for strict-align.
125 (aarch64_can_inline_p): Perform checks even when callee has no
126 attributes to check for strict alignment.
127 * doc/extend.texi (AArch64 Function Attributes): Document
129 * doc/invoke.texi: (AArch64 Options): Likewise.
131 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
133 PR tree-optimization/85853
134 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
135 the handling of the root of the node to...
136 (vect_slp_analyze_node_operations_1): ...this new function,
137 and run the whole thing with the child nodes' def types
138 set according to their SLP node's def type.
140 2018-05-23 Richard Biener <rguenther@suse.de>
143 * tree-data-ref.c (create_runtime_alias_checks): Defer
144 and ignore overflow warnings.
146 2018-05-23 Yury Gribov <tetra2005@gmail.com>
148 PR tree-optimization/85822
149 * tree-vrp.c (is_masked_range_test): Fix handling of negative
152 2018-05-23 Richard Biener <rguenther@suse.de>
154 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
155 memset constants via native_interpret_expr.
157 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
160 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
162 (cgraph_node::create_alias): Likewise.
163 (cgraph_node::get_availability): Check ifunc_resolver instead
164 of looking up ifunc attribute.
165 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
166 * varasm.c (do_assemble_alias): Likewise.
167 (assemble_alias): Likewise.
168 (default_binds_local_p_3): Likewise.
169 * cgraph.h (cgraph_node): Add ifunc_resolver.
170 (cgraph_node::only_called_directly_or_aliased_p): Return false
172 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
174 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
175 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
176 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
177 instead of looking up ifunc attribute.
179 2018-05-22 Luis Machado <luis.machado@linaro.org>
181 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
183 2018-05-22 Martin Sebor <msebor@redhat.com>
186 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
187 only when expasion succeeds.
188 (expand_builtin_strcmp): Same.
189 (expand_builtin_strncmp): Same.
191 2018-05-22 Martin Sebor <msebor@redhat.com>
193 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
195 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
196 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
198 * config/aarch64/aarch64-ldpstp.md: Replace uses of
199 aarch64_mem_pair_operand with memory_operand and delete operand swapping
201 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
202 Add check for legitimate_address.
203 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
204 (aarch64_swap_ldrstr_operands): New.
205 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
208 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
209 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
211 * config/aarch64/aarch64.md: New patterns to generate stp
213 (store_pair_sw, store_pair_dw): New patterns to generate stp for
214 single words and double words.
215 (load_pair_sw, load_pair_dw): Likewise.
216 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
218 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
220 * config/aarch64/aarch64-ldpstp.md: Modify peephole
221 for different mode ldpstp and add peephole for merged zero stores.
223 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
225 (aarch64_gen_store_pair): Rename calls to match new patterns.
226 (aarch64_gen_load_pair): Rename calls to match new patterns.
227 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
228 (load_pair<DREG:mode><DREG2:mode>): ... This.
229 (store_pair<mode>): Rename to...
230 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
231 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
233 (V_INT_EQUIV): Handle SImode.
234 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
237 2018-05-22 Martin Sebor <msebor@redhat.com>
240 * calls.c (maybe_warn_nonstring_arg): Use string length to set
241 or ajust the presumed bound on an operation to avoid unnecessary
244 2018-05-22 Martin Sebor <msebor@redhat.com>
246 PR tree-optimization/85826
247 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
248 assuming that a DECL necesarily has a constant size.
250 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
253 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
255 2018-05-22 Richard Biener <rguenther@suse.de>
257 PR tree-optimization/85834
258 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
259 non-constant and non-zero memset arguments.
261 2018-05-22 Martin Liska <mliska@suse.cz>
264 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
266 2018-05-22 Richard Biener <rguenther@suse.de>
268 PR tree-optimization/85863
269 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
270 comparisons when vectype is specified.
271 (vectorizable_condition): Do not specify vectype for
272 vect_is_simple_cond when SLP vectorizing.
274 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
277 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
278 define __ibm128 as long double.
279 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
280 as a distinct type when IEEE 128-bit support is enabled.
281 (init_float128_ieee): Fix up conversions between IFmode and IEEE
282 128-bit types to use the correct functions.
283 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
284 convert between 128-bit floating point types that have different
285 modes but the same representation, instead of using gen_lowpart to
287 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
289 (IFKF_reg): New attributes to give the register constraints for
291 (extend<mode>tf2_internal): New insns to mark an explicit
292 conversion between 128-bit floating point types that have a
293 different mode but share the same representation.
295 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
297 PR tree-optimization/85814
298 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
299 a null return from get_strinfo when unsharing the next
300 strinfo in the chain.
302 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
305 * varasm.c (weak_finish): Clean up weak_decls.
307 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
309 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
310 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
311 UNSPEC_UADALP values.
312 * config/aarch64/iterators.md (ABAL): New int iterator.
315 (sur): Add mappings for the above.
316 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
318 (aarch64_<sur>abal<mode>_4): Likewise.
319 (aarch64_<sur>adalp<mode>_3): Likewise.
320 (<sur>sadv16qi): New define_expand.
322 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
324 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
325 (*movdf_internal): Ditto.
326 (*rcpsf2_sse): Ditto.
327 (*rsqrtsf2_sse): Ditto.
328 (*sqrt<mode>2_sse): Ditto.
330 2018-05-21 Tamar Christina <tamar.christina@arm.com>
332 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
334 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
335 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
336 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
337 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
339 * config/aarch64/arm_neon.h: Likewise.
340 * config/aarch64/iterators.md (VQ_I): New.
342 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
344 * config.gcc: Add arc/t-multilib-linux to tmake_file for
346 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
349 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
351 * config/nds32/constraints.md (S): New constraint.
352 * config/nds32/nds32.md (call_internal): Use constraint S.
353 (call_value_internal): Likewise.
354 (sibcall_internal): Likewise.
355 (sibcall_value_internal): Likewise.
357 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
358 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
363 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
364 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
367 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
368 (nds32_rtx_costs_impl): Simplify.
369 (nds32_address_cost_impl): Simplify.
370 (nds32_init_rtx_costs): New function.
371 (nds32_rtx_costs_speed_prefer): Likewise.
372 (nds32_rtx_costs_size_prefer): Likewise.
373 (nds32_address_cost_speed_prefer): Likewise.
374 (nds32_address_cost_speed_fwprop): Likewise.
375 (nds32_address_cost_size_prefer): Likewise.
376 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
377 * config/nds32/nds32.c (nds32_option_override): Use
378 nds32_init_rtx_costs function.
380 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
382 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
383 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
384 (TARGET_PIPELINE_N8): Likewise.
385 (TARGET_PIPELINE_N10): Likewise.
386 (TARGET_PIPELINE_N13): Likewise.
387 (TARGET_PIPELINE_GRAYWOLF): Likewise.
389 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
391 * config/nds32/nds32-fpu.md: Update copyright year.
393 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
395 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
397 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
399 * config/nds32/nds32.c
400 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
401 * config/nds32/nds32.opt (minline-asm-r15): New option.
403 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
405 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
407 * config/nds32/nds32.md (abssi2): New pattern.
409 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
411 * config/i386/i386.md (rex64namesuffix): New mode attribute.
412 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
413 Merge insn pattern from sse_cvtsi2ss<round_name> and
414 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
415 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
416 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
417 using SWI48 mode iterator.
418 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
419 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
420 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
421 pattern from sse_cvttss2si<round_saeonly_name>
422 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
423 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
424 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
425 using SWI48 mode iterator.
426 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
427 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
428 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
429 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
430 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
431 using SWI48 mode iterator.
432 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
433 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
434 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
435 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
436 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
438 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
439 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
440 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
441 pattern from sse_cvttsd2si<round_saeonly_name>
442 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
444 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
446 * config/nds32/nds32-md-auxiliary.c
447 (nds32_valid_smw_lwm_base_p): Refine.
448 (nds32_output_smw_single_word): Refine.
449 (nds32_output_smw_double_word): New.
450 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
452 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
454 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
455 (nds32_output_stack_pop): Refine.
456 (nds32_expand_unaligned_load): Refine.
457 (nds32_expand_unaligned_store): Refine.
459 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
460 Chung-Ju Wu <jasonwucj@gmail.com>
462 * config/nds32/constants.md: Add TP_REGNUM constant.
463 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
464 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
466 * config/nds32/nds32-doubleword.md: Consider flag_pic.
467 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
468 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
469 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
470 and PIC code generation.
471 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
473 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
475 * config/nds32/nds32.md: Support TLS and PIC.
476 * config/nds32/nds32.c: Support TLS and PIC.
477 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
478 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
481 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
483 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
486 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
487 Chung-Ju Wu <jasonwucj@gmail.com>
489 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
490 * config/nds32/nds32-md-auxiliary.c
491 (symbolic_reference_mentioned_p): New.
492 (nds32_legitimize_ict_address): New.
493 (nds32_expand_ict_move): New.
494 (nds32_indirect_call_referenced_p): New.
495 (nds32_symbol_binds_local_p): Delete.
496 (nds32_long_call_p): Modify.
497 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
498 * config/nds32/nds32-protos.h
499 (symbolic_reference_mentioned_p): Declare.
500 (nds32_legitimize_ict_address): Declare.
501 (nds32_expand_ict_move): Declare.
502 (nds32_indirect_call_referenced_p): Declare.
503 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
504 (nds32_relax_group): Use nds32_ict_const_p as condition.
505 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
506 (nds32_asm_file_start): Output ict_model directive in asm code.
507 (nds32_legitimate_address_p): Consider indirect call.
508 (nds32_print_operand): Consider indirect call.
509 (nds32_print_operand_address): Consider indirect call.
510 (nds32_insert_attributes): Handle "indirect_call" attribute.
511 (TARGET_LEGITIMATE_ADDRESS_P): Define.
512 (TARGET_LEGITIMATE_CONSTANT_P): Define.
513 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
514 (TARGET_DELEGITIMIZE_ADDRESS): Define.
515 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
516 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
517 (TARGET_ICT_MODEL_SMALL): Define.
518 (TARGET_ICT_MODEL_LARGE): Define.
519 * config/nds32/nds32.md (movsi): Consider ict model.
520 (call, call_value): Consider ict model.
521 (sibcall, sibcall_value): Consider ict model.
522 * config/nds32/nds32.opt (mict-model): New option.
523 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
526 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
527 Monk Chiang <sh.chiang04@gmail.com>
528 Jim Wilson <jimw@sifive.com>
530 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
531 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
532 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
533 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
534 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
535 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
536 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
537 compute save_libcall_adjustment properly.
538 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
539 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
540 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
541 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
542 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
543 (ABI_SPEC): Handle mabi=ilp32e.
544 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
546 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
547 <-march>: Add rv32e as an example.
549 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
552 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
553 (intra_create_variable_infos): Handle C++ constructors.
555 2018-05-18 Martin Liska <mliska@suse.cz>
557 * passes.def: Remove a redundant pass.
559 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
562 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
564 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
566 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
568 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
570 (armv2, armv2a, armv3, armv3m): Delete architectures.
571 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
572 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
573 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
575 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
576 (*mulsidi3adddi): Likewise.
577 (mulsidi3): Likewise.
578 (*mulsidi3_nov6): Likewise.
579 (umulsidi3): Likewise.
580 (umulsidi3_nov6): Likewise.
581 (umaddsidi4): Likewise.
582 (*umulsidi3adddi): Likewise.
583 (smulsi3_highpart): Likewise.
584 (*smulsi3_highpart_nov6): Likewise.
585 (umulsi3_highpart): Likewise.
586 (*umulsi3_highpart_nov6): Likewise.
587 * config/arm/arm.h (arm_arch3m): Delete.
588 * config/arm/arm.c (arm_arch3m): Delete.
589 (arm_option_override_internal): Update armv3-related comment.
590 (arm_configure_build_target): Delete use of isa_bit_mode32.
591 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
592 (arm_rtx_costs_internal): Delete check of arm_arch3m.
593 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
596 * config/arm/arm-protos.h (arm_arch3m): Delete.
597 * config/arm/arm-tables.opt: Regenerate.
598 * config/arm/arm-tune.md: Likewise.
599 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
600 deleted architectures.
602 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
604 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
605 (armv5t, armv5te): New features.
606 (ARMv5, ARMv5e): Delete fgroups.
607 (ARMv5t, ARMv5te): Adjust for above changes.
609 (armv5, armv5e): Delete arches.
610 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
612 (*call_reg_arm): Likewise.
613 (*call_value_reg_armv5): Likewise.
614 (*call_value_reg_arm): Likewise.
615 (*call_symbol): Likewise.
616 (*call_value_symbol): Likewise.
617 (*sibcall_insn): Likewise.
618 (*sibcall_value_insn): Likewise.
620 (prefetch): Likewise.
621 (define_split and define_peephole2 dependent on arm_arch5):
623 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
625 (TARGET_ARM_QBIT): Likewise.
626 (TARGET_DSP_MULTIPLY): Likewise.
627 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
628 (arm_arch5, arm_arch5e): Delete.
629 (arm_arch5t, arm_arch5te): Declare.
630 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
631 (arm_arch5t): Declare.
632 (arm_option_reconfigure_globals): Update for the above.
633 (arm_options_perform_arch_sanity_checks): Update comment, replace
634 use of arm_arch5 with arm_arch5t.
635 (use_return_insn): Likewise.
636 (arm_emit_call_insn): Likewise.
637 (output_return_instruction): Likewise.
638 (arm_final_prescan_insn): Likewise.
639 (arm_coproc_builtin_available): Likewise.
640 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
641 arm_arch5e with arm_arch5t and arm_arch5te.
642 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
643 (arm_arch5t, arm_arch5te): Declare.
644 * config/arm/arm-tables.opt: Regenerate.
645 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
646 * config/arm/t-multilib: Likewise.
647 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
648 instead of arm_arch5.
649 (*call_reg_thumb1): Likewise.
650 (*call_value_reg_thumb1_v5): Likewise.
651 (*call_value_reg_thumb1): Likewise.
652 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
654 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
656 2018-05-18 Martin Liska <mliska@suse.cz>
658 PR gcov-profile/84846
659 * doc/gcov.texi: Document -t option of gcov tool.
661 2018-05-18 Martin Liska <mliska@suse.cz>
663 PR gcov-profile/84846
664 * gcov.c (print_usage): Add new -t option.
665 (process_args): Handle the option.
666 (generate_results): Use stdout as output when requested by
669 2018-05-18 Martin Liska <mliska@suse.cz>
671 PR gcov-profile/84846
672 * coverage.c (coverage_init): Write PWD to .gcno file.
673 * doc/gcov.texi: Document how working directory is printed.
674 * gcov-dump.c (dump_gcov_file): Print PWD.
675 * gcov.c (output_intermediate_file): Likewise.
676 (read_graph_file): Read PWD string.
677 (output_lines): Print PWD.
679 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
682 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
683 for retval and return false if all args to phi are zero.
685 2018-05-18 Richard Biener <rguenther@suse.de>
687 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
689 (evrp_dom_walker::before_dom_children): Call it.
691 2018-05-18 Richard Biener <rguenther@suse.de>
693 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
694 results when processing array refs with variable index.
696 2018-05-18 Toon Moene <toon@moene.org>
698 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
699 directly after that of -floop-interchange. Indicate that both
700 options are enabled by default when specifying -O3.
702 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
704 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
705 iterator. Delete separate integer-mode vec_set<mode> expander.
706 (aarch64_simd_vec_setv2di): Delete.
707 (vec_setv2di): Delete.
708 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
709 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
710 the "w, r" alternative.
712 2018-05-18 Martin Liska <mliska@suse.cz>
714 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
715 * tree-pass.h (make_pass_lower_switch_O0): New function.
716 * tree-switch-conversion.c (node_has_low_bound): Remove.
717 (node_has_high_bound): Likewise.
718 (node_is_bounded): Likewise.
719 (class pass_lower_switch): Make it a template type and create
721 (pass_lower_switch::execute): Add template argument.
722 (make_pass_lower_switch): New function.
723 (make_pass_lower_switch_O0): New function.
724 (do_jump_if_equal): Remove.
725 (emit_case_nodes): Simplify to just handle all 3 cases and leave
726 all the hard work to tree optimization passes.
728 2018-05-18 Martin Liska <mliska@suse.cz>
730 * dbgcnt.c (limit_low): Renamed from limit.
731 (limit_high): New variable.
732 (dbg_cnt_is_enabled): Check for upper limit.
733 (dbg_cnt): Adjust dumping.
734 (dbg_cnt_set_limit_by_index): Add new argument for high
736 (dbg_cnt_set_limit_by_name): Likewise.
737 (dbg_cnt_process_single_pair): Parse new format.
738 (dbg_cnt_process_opt): Use strtok.
739 (dbg_cnt_list_all_counters): Remove 'value' and add
741 * doc/invoke.texi: Document changes.
743 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
745 * doc/sourcebuild.texi (scalar_all_fma): Document.
746 * tree.def (FMA_EXPR): Delete.
747 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
748 * internal-fn.c (ternary_direct): New macro.
749 (expand_ternary_optab_fn): Likewise.
750 (direct_ternary_optab_supported_p): Likewise.
751 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
752 * builtins.c (fold_builtin_fma): Delete.
753 (fold_builtin_3): Don't call it.
754 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
755 * expr.c (expand_expr_real_2): Likewise.
756 * fold-const.c (operand_equal_p): Likewise.
757 (fold_ternary_loc): Likewise.
758 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
759 * gimple.c (DEFTREECODE): Likewise.
760 * gimplify.c (gimplify_expr): Likewise.
761 * optabs-tree.c (optab_for_tree_code): Likewise.
762 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
763 * tree-eh.c (operation_could_trap_p): Likewise.
764 (stmt_could_throw_1_p): Likewise.
765 * tree-inline.c (estimate_operator_cost): Likewise.
766 * tree-pretty-print.c (dump_generic_node): Likewise.
767 (op_code_prio): Likewise.
768 * tree-ssa-loop-im.c (stmt_cost): Likewise.
769 * tree-ssa-operands.c (get_expr_operands): Likewise.
770 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
771 * fold-const-call.h (fold_fma): Delete.
772 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
773 CFN_FNMA and CFN_FNMS.
775 * genmatch.c (combined_fn): New enum.
776 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
777 (commutative_op): New function.
778 (commutate): Use it. Handle more than 2 operands.
779 (dt_operand::gen_gimple_expr): Use commutative_op.
780 (parser::parse_expr): Allow :c to be used with non-binary
781 operators if the commutative operand is known.
782 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
783 CFN_FMS, CFN_FNMA and CFN_FNMS.
784 (backprop::process_assign_use): Remove FMA_EXPR handling.
785 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
786 (gen_hsa_fma): New function.
787 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
788 IFN_FNMA and IFN_FNMS.
789 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
790 * gimple-fold.h (follow_all_ssa_edges): Declare.
791 * gimple-fold.c (follow_all_ssa_edges): New function.
792 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
793 gimple_build interface and use follow_all_ssa_edges to fold the result.
794 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
795 instead of checking for optabs directly.
796 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
797 rather than FMA_EXPRs.
798 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
799 call to IFN_FMA instead of an FMA_EXPR.
801 2018-05-17 Jim Wilson <jimw@sifive.com>
803 * expr.c (do_tablejump): When converting index to Pmode, if we have a
804 sign extended promoted subreg, and the range does not have the sign bit
805 set, then do a sign extend.
807 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
808 test, check for sign extended subreg and/or constant operands, and
809 do a sign extend in that case.
811 2018-05-17 Steve Ellcey <sellcey@cavium.com>
813 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
814 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
816 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
817 Change logics_shift_reg to logics_shift_imm.
818 (thunderx2t99_fp_loadpair_basic): Delete.
819 (thunderx2t99_fp_storepair_basic): Delete.
820 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
821 (thunderx2t99_asimd_polynomial): Delete.
822 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
823 and neon_fp_mul_d_scalar_q.
824 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
825 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
826 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
827 (thunderx2t99_asimd_lut): Add missing tbl types.
828 (thunderx2t99_asimd_ext): Delete.
829 (thunderx2t99_asimd_load1_1_mult): Delete.
830 (thunderx2t99_asimd_load1_2_mult): Delete.
831 (thunderx2t99_asimd_load1_ldp): New.
832 (thunderx2t99_asimd_load1): New.
833 (thunderx2t99_asimd_load2): Add missing *load2* types.
834 (thunderx2t99_asimd_load3): New.
835 (thunderx2t99_asimd_load4): New.
836 (thunderx2t99_asimd_store1_1_mult): Delete.
837 (thunderx2t99_asimd_store1_2_mult): Delete.
838 (thunderx2t99_asimd_store2_mult): Delete.
839 (thunderx2t99_asimd_store2_onelane): Delete.
840 (thunderx2t99_asimd_store_stp): New.
841 (thunderx2t99_asimd_store1): New.
842 (thunderx2t99_asimd_store2): New.
843 (thunderx2t99_asimd_store3): New.
844 (thunderx2t99_asimd_store4): New.
846 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
848 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
849 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
851 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
852 Segher Boessenkool <segher@kernel.crashing.org>
855 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
858 2018-05-17 Richard Biener <rguenther@suse.de>
860 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
861 for pruning loop and prune defs feeding only already visited PHIs.
863 2018-05-17 Richard Biener <rguenther@suse.de>
865 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
867 2018-05-17 Bin Cheng <bin.cheng@arm.com>
868 Richard Biener <rguenther@suse.de>
870 PR tree-optimization/85793
871 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
872 for VMAT_ELEMENTWISE.
874 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
876 * internal-fn.h (lookup_internal_fn): Declare
877 * internal-fn.c (lookup_internal_fn): New function.
878 * gimple.c (gimple_build_call_from_tree): Handle calls to
880 * gimple-pretty-print.c (dump_gimple_call): Print "." before
881 internal function names.
882 * tree-pretty-print.c (dump_generic_node): Likewise.
883 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
885 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
887 * gimple-fold.h (gimple_build): Make the function forms take
888 combined_fn rather than built_in_function.
889 (gimple_simplify): Likewise.
890 * gimple-match-head.c (gimple_simplify): Likewise.
891 * gimple-fold.c (gimple_build): Likewise.
892 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
893 rather than gimple_build_call_internal.
894 (get_initial_defs_for_reduction): Likewise.
895 (vect_create_epilog_for_reduction): Likewise.
896 (vectorizable_live_operation): Likewise.
898 2018-05-17 Martin Liska <mliska@suse.cz>
900 * gimple-ssa-sprintf.c (format_directive): Do not use
901 space in between 'G_' and '('.
903 2018-05-17 Jakub Jelinek <jakub@redhat.com>
906 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
907 even if the mask is not all ones.
910 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
912 (ix86_gimple_fold_builtin): Likewise.
915 * config/i386/i386.c: Include tree-vector-builder.h.
916 (ix86_vector_shift_count): New function.
917 (ix86_fold_builtin): Fold shift builtins by scalar count.
918 (ix86_gimple_fold_builtin): Likewise.
920 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
921 _mm512_setzero): New intrinsics.
923 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
924 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
926 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
927 code generation for cases where splatting a value is not useful.
928 * simplify-rtx.c (simplify_ternary_operation): Simplify
929 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
930 mode to a vec_concat.
932 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
934 * config.gcc: Support "goldmont-plus".
935 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
937 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
938 PROCESSOR_GOLDMONT_PLUS.
939 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
940 (processor_target_table): Add "goldmont-plus".
941 (PTA_GOLDMONT_PLUS): Define.
942 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
943 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
944 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
945 (fold_builtin_cpu): Add "goldmont-plus".
946 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
947 (ix86_option_override_internal): Add "goldmont-plus".
948 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
949 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
950 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
951 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
953 2018-05-17 Richard Biener <rguenther@suse.de>
955 PR tree-optimization/85757
956 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
957 remove defs that only feed that PHI from further processing.
959 2018-05-16 Jim Wilson <jimw@sifive.com>
961 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
963 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
965 2018-05-16 Mark Wielaard <mark@klomp.org>
967 * dwarf2out.c (count_index_strings): New function.
968 (output_indirect_strings): Call count_index_strings and generate
969 header for dwarf_version >= 5.
971 2018-05-16 Mark Wielaard <mark@klomp.org>
973 * dwarf2out.c (dwarf_FORM): New function.
974 (set_indirect_string): Use dwarf_FORM.
975 (reset_indirect_string): Likewise.
976 (size_of_die): Likewise.
977 (value_format): Likewise.
978 (output_die): Likewise.
979 (add_skeleton_AT_string): Likewise.
980 (output_macinfo_op): Likewise.
981 (index_string): Likewise.
982 (output_index_string_offset): Likewise.
983 (output_index_string): Likewise.
984 (count_index_strings): Likewise.
986 2018-05-16 Carl Love <cel@us.ibm.com>
988 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
989 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
991 2018-05-16 Martin Jambor <mjambor@suse.cz>
993 * ipa-prop.c (ipa_free_all_edge_args): Remove.
994 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
996 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
998 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
999 (fnma<mode>4): Likewise.
1000 (fms<mode>4): Likewise.
1001 (fnms<mode>4): Likewise.
1002 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1003 (aarch64_fnma<mode>4): Likewise.
1004 (aarch64_fms<mode>4): Likewise.
1005 (aarch64_fnms<mode>4): Likewise.
1006 (aarch64_fnmadd<mode>4): Likewise.
1008 2018-05-16 Jason Merrill <jason@redhat.com>
1010 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1012 2018-05-16 Richard Biener <rguenther@suse.de>
1014 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1015 (dump_stmt_cost): Declare.
1016 (add_stmt_cost): Dump cost we add.
1017 (add_stmt_costs): New function.
1018 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1020 (vect_analyze_stmt): Adjust prototype.
1021 (vectorizable_condition): Likewise.
1022 (vectorizable_live_operation): Likewise.
1023 (vectorizable_reduction): Likewise.
1024 (vectorizable_induction): Likewise.
1025 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1026 cost vector to pass to vectorizable_ and record afterwards.
1027 (vect_model_reduction_cost): Take cost vector argument and adjust.
1028 (vect_model_induction_cost): Likewise.
1029 (vectorizable_reduction): Likewise.
1030 (vectorizable_induction): Likewise.
1031 (vectorizable_live_operation): Likewise.
1032 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1033 SLP_TREE_NUMBER_OF_VEC_STMTS.
1034 (vect_analyze_slp_cost_1): Remove.
1035 (vect_analyze_slp_cost): Likewise.
1036 (vect_slp_analyze_node_operations): Take visited args and
1037 a target cost vector. Avoid processing already visited stmt sets.
1038 (vect_slp_analyze_operations): Use a local cost vector to gather
1039 costs and register those of non-discarded instances.
1040 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1041 (vect_schedule_slp_instance): Remove copying of
1042 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1044 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1045 adding cost. Record cost entry location.
1046 (vect_prologue_cost_for_slp_op): Function to compute cost of
1047 a constant or invariant generated for SLP vect in the prologue,
1048 split out from vect_analyze_slp_cost_1.
1049 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1050 (vect_model_promotion_demotion_cost): Likewise.
1051 (vect_model_store_cost): Likewise, make static.
1052 (vect_model_load_cost): Likewise.
1053 (vectorizable_bswap): Add cost vector arg and adjust.
1054 (vectorizable_call): Likewise.
1055 (vectorizable_simd_clone_call): Likewise.
1056 (vectorizable_conversion): Likewise.
1057 (vectorizable_assignment): Likewise.
1058 (vectorizable_shift): Likewise.
1059 (vectorizable_operation): Likewise.
1060 (vectorizable_store): Likewise.
1061 (vectorizable_load): Likewise.
1062 (vectorizable_condition): Likewise.
1063 (vectorizable_comparison): Likewise.
1064 (can_vectorize_live_stmts): Likewise.
1065 (vect_analyze_stmt): Likewise.
1066 (vect_transform_stmt): Adjust calls to vectorizable_*.
1067 * tree-vectorizer.c: Include gimple-pretty-print.h.
1068 (dump_stmt_cost): New function.
1070 2018-05-16 Richard Biener <rguenther@suse.de>
1072 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1073 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1074 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1075 (check_name): New callback.
1076 (dse_classify_store): Track cycles via a visited bitmap of PHI
1077 defs and simplify handling of in-loop and across loop dead stores
1078 and properly fail for loop-variant refs. Handle byte-tracking with
1079 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1082 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1084 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1085 (vect_get_mask_type_for_stmt): Likewise.
1086 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1088 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1089 to determine the statement's vector type and the vector type that
1090 should be used for calculating nunits. Deal with cases in which
1091 the type has to be deferred.
1092 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1093 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1094 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1095 (vect_determine_vf_for_stmt): New functions, split out from...
1096 (vect_determine_vectorization_factor): ...here.
1097 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1098 (vect_get_mask_type_for_stmt): New functions, split out from
1099 vect_determine_vectorization_factor.
1101 2018-05-16 Richard Biener <rguenther@suse.de>
1103 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1104 verify the [VEC_]COND_EXPR embedded comparison.
1106 2018-05-15 Martin Sebor <msebor@redhat.com>
1108 PR tree-optimization/85753
1109 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1110 RECORD_TYPE in addition to ARRAY_TYPE.
1112 2018-05-15 Martin Sebor <msebor@redhat.com>
1115 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1117 2018-05-15 Richard Biener <rguenther@suse.de>
1119 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1120 add by_clobber_p one. Change algorithm to collect all defs
1121 representing uses we need to walk and try reducing them to
1122 a single one before failing.
1123 (dse_dom_walker::dse_optimize_stmt): Adjust.
1125 2018-05-13 Mark Wielaard <mark@klomp.org>
1127 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1128 (size_of_loc_descr): Likewise.
1129 (output_loc_operands): Likewise.
1130 (output_loc_operands_raw): Likewise.
1131 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1132 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1133 (hash_loc_operands): Likewise.
1134 (compare_loc_operands): Likewise.
1136 2018-05-14 Mark Wielaard <mark@klomp.org>
1138 * dwarf2out.c (count_index_addrs): New function.
1139 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1141 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1143 PR tree-optimization/83648
1144 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1145 return value as malloc candidate.
1147 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1150 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1151 as true in call to suggest_attribute.
1153 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1155 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1156 -mreadonly-in-sdata.
1158 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1160 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1162 (aarch64_crypto_aesd_fused): Likewise.
1164 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1166 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1167 (movsi_aarch64): Likewise.
1168 (load_pairsi): Likewise.
1169 (load_pairdi): Likewise.
1170 (store_pairsi): Likewise.
1171 (store_pairdi): Likewise.
1172 (load_pairsf): Likewise.
1173 (load_pairdf): Likewise.
1174 (store_pairsf): Likewise.
1175 (store_pairdf): Likewise.
1176 (zero_extend): Likewise.
1177 (trunc): Swap alternatives.
1178 (fcvt_target): Add '?' to prefer w over r.
1180 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1183 * config/i386/i386.md: Disallow non-commutative arithmetics in
1184 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1185 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1186 in the peephole2 before it.
1188 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1190 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1191 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1192 (ix86_handle_option): Handle -mcldemote.
1193 * config.gcc: New header.
1194 * config/i386/cldemoteintrin.h: New file.
1195 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1196 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1198 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1199 OPTION_MASK_ISA_CLDEMOTE.
1200 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1201 (ix86_valid_target_attribute_inner_p): Ditto.
1202 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1203 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1204 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1205 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1206 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1208 * config/i386/i386.opt: Add -mcldemote.
1209 * config/i386/x86intrin.h: New header.
1210 * doc/invoke.texi: Add -mcldemote.
1212 2018-05-14 Richard Biener <rguenther@suse.de>
1214 * doc/match-and-simplify.texi: Adjust :s documentation.
1216 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1218 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1219 intended memcpy size.
1220 (REORDER_45): Likewise.
1222 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1224 * sort.cc: New file.
1225 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1226 * vec.c (qsort_chk): Use gcc_qsort.
1227 * Makefile.in (OBJS-libcommon): Add sort.o.
1228 (build/sort.o): New target. Use it...
1229 (BUILD_RTL): ... here, and...
1230 (build/gencfn-macros): ... here, and...
1231 (build/genmatch): ... here.
1233 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1234 Chung-Ju Wu <jasonwucj@gmail.com>
1236 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1237 * config/nds32/nds32-graywolf.md: New file.
1238 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1239 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1241 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1242 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1243 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1244 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1245 * config/nds32/pipelines.md: Include n15 settings.
1247 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1248 Chung-Ju Wu <jasonwucj@gmail.com>
1250 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1251 * config/nds32/nds32-n13.md: New file.
1252 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1253 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1255 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1256 * config/nds32/nds32.md (pipeline_model): Add n13.
1257 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1258 * config/nds32/pipelines.md: Include n13 settings.
1260 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1261 Chung-Ju Wu <jasonwucj@gmail.com>
1263 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1264 * config/nds32/nds32-n10.md: New file.
1265 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1266 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1268 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1269 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1270 * config/nds32/nds32.md (pipeline_model): Add n10.
1271 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1272 * config/nds32/pipelines.md: Include n10 settings.
1274 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1275 Kito Cheng <kito.cheng@gmail.com>
1276 Chung-Ju Wu <jasonwucj@gmail.com>
1278 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1279 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1280 Add enum values for DSP extension instructions.
1281 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1283 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1284 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1286 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1287 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1288 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1289 * config/nds32/nds32-intrinsic.md: Likewise.
1290 * config/nds32/nds32_intrinsic.h: Likewise.
1291 * config/nds32/nds32-md-auxiliary.c: Likewise.
1292 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1293 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1294 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1295 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1296 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1297 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1298 TYPE_DMAC in switch statement.
1299 * config/nds32/nds32.c: New checking and implementation for DSP
1300 extension instructions.
1301 * config/nds32/nds32.h: Likewise.
1302 * config/nds32/nds32.md: Likewise.
1303 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1304 * config/nds32/predicates.md: Implement new predicates for DSP
1307 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1309 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1310 Reformat alternatives and attributes so it is easier to identify
1311 which constraints/attributes go with which instruction.
1312 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1313 (mov<mode>_softfloat32, FMOVE64): Likewise.
1314 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1315 (mov<mode>_softfloat64, FMOVE64): Likewise.
1317 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1319 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1321 (Basic PowerPC Built-in Functions): The new name of the
1322 subsection previously known as "PowerPC Built-in Functions".
1323 (Basic PowerPC Built-in Functions Available on all Configurations):
1325 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1326 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1327 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1328 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1330 2018-05-11 Martin Jambor <mjambor@suse.cz>
1333 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1336 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1339 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1341 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1343 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1344 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1345 (ix86_handle_option): Handle -mwaitpkg.
1346 * config.gcc: New header.
1347 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1348 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1349 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1351 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1352 OPTION_MASK_ISA_WAITPKG.
1353 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1354 (ix86_option_override_internal): Add PTA_WAITPKG.
1355 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1356 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1357 IX86_BUILTIN_TPAUSE.
1358 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1359 __builtin_ia32_umwait and __builtin_ia32_tpause.
1360 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1361 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1362 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1363 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1364 UNSPECV_TPAUSE): New.
1365 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1366 * config/i386/i386.opt: Add -mwaitpkg.
1367 * config/i386/waitpkgintrin.h: New file.
1368 * config/i386/x86intrin.h: New header.
1369 * doc/invoke.texi: Add -mwaitpkg.
1371 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1374 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1376 (cortex-m0): Use armv6s-m isa.
1377 (cortex-m0plus): Likewise.
1378 (cortex-m1): Likewise.
1379 (cortex-m0.small-multiply): Likewise.
1380 (cortex-m0plus.small-multiply): Likewise.
1381 (cortex-m1.small-multiply): Likewise.
1383 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1384 Jakub Jelinek <jakub@redhat.com>
1386 PR tree-optimization/85692
1387 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1388 source permute as well.
1390 2018-05-11 Martin Liska <mliska@suse.cz>
1393 * doc/extend.texi: Document LLVM style format for no_sanitize
1396 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1398 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1399 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1400 (mode_supports_vsx_dform_quad): Likewise.
1401 (mode_supports_vmx_dform): Move these functions to be next to the
1402 other mode_supports functions.
1403 (mode_supports_dq_form): Likewise.
1404 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1405 mode_supports_dq_form.
1406 (reg_offset_addressing_ok_p): Likewise.
1407 (offsettable_ok_by_alignment): Likewise.
1408 (rs6000_legitimate_offset_address_p): Likewise.
1409 (legitimate_lo_sum_address_p): Likewise.
1410 (rs6000_legitimize_address): Likewise.
1411 (rs6000_legitimize_reload_address): Likewise.
1412 (rs6000_secondary_reload_inner): Likewise.
1413 (rs6000_preferred_reload_class): Likewise.
1414 (rs6000_output_move_128bit): Likewise.
1416 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1418 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1419 Generate SImode target register for null target.
1420 <case IX86_BUILTIN_XGETBV>: Ditto.
1421 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1422 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1424 2018-05-10 Carl Love <cel@us.ibm.com>
1426 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1427 dcbtt and dcbtstt if operands[2] is 0.
1429 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1432 * config/i386/sse.md (usadv64qi): New expander.
1434 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1436 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1437 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1438 -maltivec=be support.
1439 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1440 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1441 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1442 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1443 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1444 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1445 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1446 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1447 altivec_vsumsws): Adjust.
1448 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1449 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1450 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1452 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1453 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1454 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1455 (altivec_lve<VI_char>x): Delete expand.
1456 (*altivec_lve<VI_char>x_internal): Rename to...
1457 (altivec_lve<VI_char>x): ... this.
1458 (altivec_lvxl_<mode>): Delete expand.
1459 (*altivec_lvxl_<mode>_internal): Rename to ...
1460 (altivec_lvxl_<mode>): ... this.
1461 (altivec_stvxl_<mode>): Delete expand.
1462 (*altivec_stvxl_<mode>_internal): Rename to ...
1463 (altivec_stvxl_<mode>): ... this.
1464 (altivec_stve<VI_char>x): Delete expand.
1465 (*altivec_stve<VI_char>x_internal): Rename to ...
1466 (altivec_stve<VI_char>x): ... this.
1467 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1468 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1469 reduc_plus_scal_<mode>): Adjust.
1470 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1472 (rs6000_cpu_cpp_builtins): Adjust.
1473 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1474 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1475 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1476 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1477 -maltivec=be support.
1478 (rs6000_split_vec_extract_var): Adjust.
1479 (rs6000_split_v4si_init): Adjust.
1480 (swap_selector_for_mode): Delete.
1481 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1482 altivec_expand_stvex_be): Delete.
1483 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1484 -maltivec=be support.
1485 (rs6000_gimple_fold_builtin): Ditto.
1486 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1488 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1489 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1490 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1491 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1492 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1493 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1494 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1495 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1496 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1497 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1498 anonymous split): Adjust.
1499 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1500 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1502 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1504 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1505 when --with-gxx-include-dir is also specified.
1506 * configure: Regenerate.
1508 2018-05-09 Jim Wilson <jimw@sifive.com>
1511 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1512 * config/riscv/t-withmultilib: New.
1513 * config/riscv/withmultilib.h: New.
1514 * doc/install.texi: Document RISC-V --with-multilib-list support.
1516 2018-05-09 Richard Biener <rguenther@suse.de>
1518 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1520 (vect_bb_vectorization_profitable_p): Adjust. Compute
1521 actual scalar cost using the cost vector and the add_stmt_cost
1524 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1526 PR rtl-optimization/85645
1527 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1528 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1530 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1532 PR rtl-optimization/85645
1533 * shrink-wrap.c (spread_components): Return a boolean saying if
1534 anything was changed.
1535 (try_shrink_wrapping_separate): Iterate spread_components until
1536 nothing changes anymore.
1538 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1540 PR rtl-optimization/85645
1541 * regrename.c (build_def_use): Also kill the chains that include the
1542 destination of a REG_CFA_REGISTER note.
1544 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1546 PR rtl-optimization/85645
1547 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1548 insn that has a REG_CFA_REGISTER note.
1550 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1552 * cfgexpand.c (expand_clobber): New function.
1553 (expand_gimple_stmt_1): Use it.
1554 * tree-vect-stmts.c (vect_clobber_variable): New function,
1556 (vectorizable_simd_clone_call): ...here.
1557 (vectorizable_store): Emit a clobber either side of an
1558 IFN_STORE_LANES sequence.
1559 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1561 2018-05-09 Tom de Vries <tom@codesourcery.com>
1564 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1565 (define_insn "trap_if_false"): Add exit after trap.
1567 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1569 PR rtl-optimization/85638
1570 * bb-reorder.c: Include common/common-target.h.
1571 (create_forwarder_block): New function extracted from...
1572 (fix_up_crossing_landing_pad): ...here. Rename into...
1573 (dw2_fix_up_crossing_landing_pad): ...this.
1574 (sjlj_fix_up_crossing_landing_pad): New function.
1575 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1576 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1577 from both partitions and exit the loop after one iteration.
1579 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1582 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1584 (Basic PowerPC Built-in Functions): The new name of the
1585 subsection previously known as "PowerPC Built-in Functions".
1586 (Basic PowerPC Built-in Functions Available on all Configurations):
1588 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1590 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1592 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1594 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1597 2018-05-08 Jim Wilson <jimw@sifive.com>
1599 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1600 (LD_EMUL_SUFFIX): New.
1601 (LINK_SPEC): Use it.
1603 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1605 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1607 (Basic PowerPC Built-in Functions): The new name of the
1608 subsection previously known as "PowerPC Built-in Functions".
1609 (Basic PowerPC Built-in Functions Available on all Configurations):
1611 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1613 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1615 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1617 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1620 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1623 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1624 after cmpelim optimization.
1626 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1628 * config.gcc: Support "goldmont".
1629 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1630 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1632 * config/i386/i386.c (m_GOLDMONT): Define.
1633 (processor_target_table): Add "goldmont".
1634 (PTA_GOLDMONT): Define.
1635 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1636 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1637 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1638 (fold_builtin_cpu): Add "goldmont".
1639 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1640 (ix86_option_override_internal): Add "goldmont".
1641 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1642 (processor_type): Add PROCESSOR_GOLDMONT.
1643 * config/i386/i386.md: Add CPU "glm".
1644 * config/i386/glm.md: New file.
1645 * config/i386/x86-tune.def: Add m_GOLDMONT.
1646 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1648 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1651 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1653 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1654 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1655 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1658 * config/i386/i386.c (ix86_fold_builtin): Handle
1659 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1662 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1663 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1665 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1668 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1669 (check_arch): Likewise.
1670 (check_fpu): Return the result rather than printing it.
1671 (end arch): Fix operator precedence.
1672 (end cpu): Likewise.
1673 (END): Print the result from check_fpu.
1675 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1676 Alan Hayward <alan.hayward@arm.com>
1677 David Sherwood <david.sherwood@arm.com>
1679 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1680 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1681 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1682 (*fcmuo<mode>_and): New patterns.
1684 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1686 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1687 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1688 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1689 (cmp_op, sve_imm_con): New code attributes.
1690 (SVE_COND_INT_CMP, imm_con): Delete.
1691 (cmp_op): Remove above unspecs from int attribute.
1692 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1694 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1695 comparison-specific unspecs.
1696 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1697 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1698 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1699 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1700 (*vec_fcm<cmp_op><mode>): Rename to...
1701 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1702 (*vec_fcmuo<mode>): Rename to...
1703 (*fcmuo<mode>): ...this and adjust likewise.
1704 (*pred_fcm<cmp_op><mode>): New pattern.
1705 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1706 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1708 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1710 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1711 (aarch64_emit_sve_predicated_cond): New function.
1712 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1713 (aarch64_emit_unspec_cond_or): Replace with...
1714 (aarch64_emit_sve_or_conds): ...this new function. Use
1715 aarch64_emit_sve_ptrue_op for the individual comparisons and
1716 aarch64_emit_binop to OR them together.
1717 (aarch64_emit_inverted_unspec_cond): Replace with...
1718 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1719 aarch64_emit_sve_ptrue_op for the comparison and
1720 aarch64_emit_unop to invert the result.
1721 (aarch64_expand_sve_vec_cmp_float): Update after the above
1722 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1724 2018-05-07 Nathan Sidwell <nathan@acm.org>
1726 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1727 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1728 (Backwards Compatibility): Likewise.
1730 2018-05-07 Luis Machado <luis.machado@linaro.org>
1734 2018-05-07 Luis Machado <luis.machado@linaro.org>
1736 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1737 <prefetch_dynamic_strides>: New const bool field.
1738 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1739 prefetch_dynamic_strides.
1740 (exynosm1_prefetch_tune): Likewise.
1741 (thunderxt88_prefetch_tune): Likewise.
1742 (thunderx_prefetch_tune): Likewise.
1743 (thunderx2t99_prefetch_tune): Likewise.
1744 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1745 (aarch64_override_options_internal): Update to set
1746 PARAM_PREFETCH_DYNAMIC_STRIDES.
1747 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1748 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1749 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1750 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1751 prefetch-dynamic-strides setting.
1753 2018-05-07 Luis Machado <luis.machado@linaro.org>
1755 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1756 <minimum_stride>: New const int field.
1757 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1758 minimum_stride field.
1759 (exynosm1_prefetch_tune): Likewise.
1760 (thunderxt88_prefetch_tune): Likewise.
1761 (thunderx_prefetch_tune): Likewise.
1762 (thunderx2t99_prefetch_tune): Likewise.
1763 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1764 (aarch64_override_options_internal): Update to set
1765 PARAM_PREFETCH_MINIMUM_STRIDE.
1766 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1767 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1768 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1769 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1770 stride is constant and is below the minimum stride threshold.
1772 2018-05-07 Luis Machado <luis.machado@linaro.org>
1774 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1777 2018-05-07 Luis Machado <luis.machado@linaro.org>
1779 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1780 <prefetch_dynamic_strides>: New const bool field.
1781 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1782 prefetch_dynamic_strides.
1783 (exynosm1_prefetch_tune): Likewise.
1784 (thunderxt88_prefetch_tune): Likewise.
1785 (thunderx_prefetch_tune): Likewise.
1786 (thunderx2t99_prefetch_tune): Likewise.
1787 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1788 (aarch64_override_options_internal): Update to set
1789 PARAM_PREFETCH_DYNAMIC_STRIDES.
1790 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1791 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1792 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1793 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1794 prefetch-dynamic-strides setting.
1796 2018-05-07 Luis Machado <luis.machado@linaro.org>
1798 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1799 <minimum_stride>: New const int field.
1800 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1801 minimum_stride field.
1802 (exynosm1_prefetch_tune): Likewise.
1803 (thunderxt88_prefetch_tune): Likewise.
1804 (thunderx_prefetch_tune): Likewise.
1805 (thunderx2t99_prefetch_tune): Likewise.
1806 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1807 (aarch64_override_options_internal): Update to set
1808 PARAM_PREFETCH_MINIMUM_STRIDE.
1809 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1810 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1811 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1812 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1813 stride is constant and is below the minimum stride threshold.
1815 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1818 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1819 the type is addressable. Don't force op into register if it has
1822 2018-05-05 Roland McGrath <mcgrathr@google.com>
1825 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1826 any section for which we don't know a specific type it should have,
1827 regardless of name. Previously this was done only for the exact
1828 names ".init_array", ".fini_array", and ".preinit_array".
1829 (default_elf_asm_named_section): Add comment about
1830 relationship with default_section_type_flags and SECTION_NOTYPE.
1831 (get_section): Don't consider it a type conflict if one side has
1832 SECTION_NOTYPE and the other doesn't, as long as neither has the
1833 SECTION_BSS et al used in the default_section_type_flags logic.
1835 2018-05-05 Tom de Vries <tom@codesourcery.com>
1838 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1839 (workaround_barsyncs): New function.
1840 (nvptx_reorg): Use workaround_barsyncs.
1841 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1842 (define_expand "nvptx_membar_cta"): New define_expand.
1843 (define_insn "*nvptx_membar_cta"): New insn.
1845 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1847 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1848 To improve optimization opportunities.
1849 * builtin-types.def: The new needed builtin types for the above.
1851 2018-05-04 Richard Biener <rguenther@suse.de>
1853 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1854 * gimple-ssa-store-merging.c
1855 (imm_store_chain_info::output_merged_store): Remove redundant create,
1856 release split_store vector contents on failure.
1857 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1858 scalar stmt vector on cache hit.
1860 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1862 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1864 * config.gcc (powerpc-xilinx-eabi*): Remove.
1865 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1867 (fusion_addis_mem_combo_load): Ditto.
1868 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1870 (rs6000_cpu_cpp_builtins): Ditto.
1871 * config/rs6000/rs6000-linux.c
1872 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1873 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1874 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1876 (rs6000_setup_reg_addr_masks): Ditto.
1877 (rs6000_init_hard_regno_mode_ok): Ditto.
1878 (rs6000_option_override_internal): Ditto.
1879 (legitimate_lo_sum_address_p): Ditto.
1880 (rs6000_legitimize_address): Ditto.
1881 (rs6000_legitimize_reload_address): Ditto.
1882 (rs6000_legitimate_address_p): Ditto.
1883 (abi_v4_pass_in_fpr): Ditto.
1884 (setup_incoming_varargs): Ditto.
1885 (rs6000_gimplify_va_arg): Ditto.
1886 (rs6000_split_multireg_move): Ditto.
1887 (rs6000_savres_strategy): Ditto.
1888 (rs6000_emit_prologue_components): Ditto.
1889 (rs6000_emit_epilogue_components): Ditto.
1890 (rs6000_emit_prologue): Ditto.
1891 (rs6000_emit_epilogue): Ditto.
1892 (rs6000_elf_file_end): Ditto.
1893 (rs6000_function_value): Ditto.
1894 (rs6000_libcall_value): Ditto.
1895 * config/rs6000/rs6000.h: Ditto.
1896 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1897 (TARGET_MINMAX): ... this. New.
1898 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1899 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1900 (*movsi_internal1_single): Delete.
1901 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1902 mfpu=, mxilinx-fpu): Delete.
1903 * config/rs6000/singlefp.h: Delete.
1904 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1905 * config/rs6000/t-rs6000: Ditto.
1906 * config/rs6000/t-xilinx: Delete.
1907 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1908 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1909 (VStype_simple): Delete.
1910 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1911 * config/rs6000/xfpu.h: Delete.
1912 * config/rs6000/xfpu.md: Delete.
1913 * config/rs6000/xilinx.h: Delete.
1914 * config/rs6000/xilinx.opt: Delete.
1915 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1916 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1918 2018-05-04 Tom de Vries <tom@codesourcery.com>
1921 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1924 2018-05-04 Richard Biener <rguenther@suse.de>
1927 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1928 (expand_complex_div_wide): Likewise.
1929 (expand_complex_operations_1): Likewise.
1930 (expand_complex_libcall): Preserve EH info of the original stmt.
1931 (tree_lower_complex): Handle removed blocks.
1932 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1933 on complex multiplication and division libcall builtins.
1935 2018-05-04 Richard Biener <rguenther@suse.de>
1938 * fold-const.c (negate_expr_p): Restrict negation of operand
1939 zero of a division to when we know that can happen without
1941 (fold_negate_expr_1): Likewise.
1943 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1946 * real.h (real_nextafter): Declare.
1947 * real.c (real_nextafter): New function.
1948 * fold-const-call.c (fold_const_nextafter): New function.
1949 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1950 CASE_CFN_NEXTTOWARD.
1951 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1952 even when arg1_mode is different from arg0_mode.
1954 2018-05-03 Nathan Sidwell <nathan@acm.org>
1956 * doc/extend.texi (Deprecated Features): Remove
1958 (Backwards Compatibility): Likewise.
1959 * doc/invoke.texi (C++ Language Options): Likewise.
1960 (C++ Dialect Options): Likewise.
1962 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1965 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1966 _mm512_mask_mullox_epi64): New intrinsics.
1968 2018-05-03 Tom de Vries <tom@codesourcery.com>
1971 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1972 dump files): Add offload-tree.
1974 2018-05-03 Richard Biener <rguenther@suse.de>
1976 PR tree-optimization/85615
1977 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1978 to loops not nested in BBs loop father to avoid creating multi-entry
1981 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1983 PR tree-optimization/70291
1984 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1985 arguments. Change return type to tree. Emit libcall as a new
1986 statement rather than replacing existing one when inplace_p is true.
1987 (expand_complex_multiplication_components): New function.
1988 (expand_complex_multiplication): Expand floating-point complex
1989 multiplication using the above.
1990 (expand_complex_division): Rename inner_type parameter to type.
1991 Update expand_complex_libcall call-site.
1992 (expand_complex_operations_1): Update expand_complex_multiplication
1993 and expand_complex_division call-sites.
1995 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1998 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1999 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2000 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2001 the highest significant bit of the shift count mask is clear. In
2002 check whether and[sq]i3 is needed verify that all significant bits
2003 of the shift count other than the highest are set.
2005 2018-05-02 Tom de Vries <tom@codesourcery.com>
2008 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2009 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2010 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2011 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2012 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2013 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2014 __builtin_goacc_parlevel_size.
2016 2018-05-02 Richard Biener <rguenther@suse.de>
2018 PR tree-optimization/85597
2019 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2020 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2023 2018-05-02 Tom de Vries <tom@codesourcery.com>
2026 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2027 dump files): Add ltrans-tree.
2029 2018-05-02 Tom de Vries <tom@codesourcery.com>
2032 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2033 dump files): Add wpa-ipa.
2035 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2037 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2038 powerpc*-*-linux*paired* target.
2039 * config/rs6000/750cl.h: Delete.
2040 * config/rs6000/paired.h: Delete.
2041 * config/rs6000/paired.md: Delete.
2042 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2044 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2045 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2046 comment. Remove paired float support.
2047 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2048 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2050 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2051 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2053 * config/rs6000/rs6000.c: Remove paired float support.
2054 (paired_expand_vector_init, paired_expand_vector_move,
2055 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2056 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2057 paired_expand_builtin, paired_expand_predicate_builtin,
2058 paired_init_builtins): Delete.
2059 * config/rs6000/rs6000.h: Remove paired float support.
2060 * config/rs6000/rs6000.md: Remove paired float support.
2061 (move_from_CR_ov_bit): Delete.
2062 * config/rs6000/rs6000.opt (mpaired): Delete.
2063 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2064 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2066 2018-05-02 Richard Biener <rguenther@suse.de>
2069 * gimplify.c (gimplify_save_expr): When in SSA form allow
2070 SAVE_EXPRs to compute to SSA vars.
2072 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2075 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2076 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2077 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2078 clobber operands[2], instead use a new pseudo. Formatting fixes.
2080 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2082 PR tree-optimization/85586
2083 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2084 exit early for statements in the same group if the accesses are
2087 2018-05-02 Tom de Vries <tom@codesourcery.com>
2090 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2093 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2095 PR tree-optimization/85143
2096 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2098 2018-05-01 Tom de Vries <tom@codesourcery.com>
2101 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2102 not found" error message.
2104 2018-05-01 Tom de Vries <tom@codesourcery.com>
2107 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2108 * vec.c (test_ordered_remove_if): New function.
2109 (vec_c_tests): Call test_ordered_remove_if.
2110 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2111 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2112 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2113 VEC_ORDERED_REMOVE_IF.
2115 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2117 PR tree-optimization/82665
2118 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2119 pointer subtraction where arguments come from a memchr call.
2121 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2123 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2124 --push-state --as-needed and --pop-state instead of --as-needed and
2125 --no-as-needed if ld supports it.
2126 * configure: Regenerated.
2129 * doc/install.texi2html: Replace _002d with - and _002a with * in
2130 generated html files using sed.
2132 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2135 * gcc-rich-location.c (blank_line_before_p): New function.
2136 (use_new_line): New function.
2137 (gcc_rich_location::add_fixit_insert_formatted): New function.
2138 * gcc-rich-location.h
2139 (gcc_rich_location::add_fixit_insert_formatted): New function.
2141 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2143 * selftest.c (assert_streq): Rename "expected" and "actual" to
2144 "val1" and "val2". Extend NULL-handling to cover both inputs
2145 symmetrically, while still requiring both to be non-NULL for a pass.
2146 * selftest.h (assert_streq): Rename "expected" and "actual" to
2148 (ASSERT_EQ): Likewise.
2149 (ASSERT_EQ_AT): Likewise.
2150 (ASSERT_KNOWN_EQ): Likewise.
2151 (ASSERT_KNOWN_EQ_AT): Likewise.
2152 (ASSERT_NE): Likewise.
2153 (ASSERT_MAYBE_NE): Likewise.
2154 (ASSERT_MAYBE_NE_AT): Likewise.
2155 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2156 the assertion to pass.
2157 (ASSERT_STREQ_AT): Likewise.
2159 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2161 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2162 interaction with -pie.
2164 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2166 * selftest.h: Fix alphabetization of per-source-file selftest
2169 2018-04-30 Jason Merrill <jason@redhat.com>
2171 PR c++/61982 - dead stores to destroyed objects.
2172 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2175 2018-04-30 Jason Merrill <jason@redhat.com>
2177 * tree.c (build_clobber): New.
2178 * tree.h: Declare it.
2179 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2181 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2183 * diagnostic-show-locus.c (layout::layout): Update for
2184 location_get_source_line returning a char_span.
2185 (struct char_span): Move to input.h.
2186 (struct correction): Update for fields in char_span becoming
2188 (struct source_line): Update for location_get_source_line
2189 returning a char_span.
2190 (layout::print_line): Likewise.
2191 * edit-context.c (edited_file::print_content): Likewise.
2192 (edited_file::print_diff_hunk): Likewise.
2193 (edited_file::print_run_of_changed_lines): Likewise.
2194 (edited_file::get_num_lines): Likewise.
2195 (edited_line::edited_line): Likewise.
2196 * final.c (asm_show_source): Likewise.
2197 * input.c (location_get_source_line): Convert return type
2198 from const char * to char_span, losing the final "line_len"
2200 (dump_location_info): Update for the above.
2201 (get_substring_ranges_for_loc): Likewise. Use a char_span
2202 when handling the literal within the line.
2203 (test_reading_source_line): Update for location_get_source_line
2204 returning a char_span.
2205 * input.h (class char_span): Move here from
2206 diagnostic-show-locus.c, converting from a struct to a class.
2207 Make data members private.
2208 (char_span::operator bool): New.
2209 (char_span::length): New.
2210 (char_span::get_buffer): New.
2211 (char_span::operator[]): New.
2212 (char_span::subspan): Make const.
2213 (char_span::xstrdup): New.
2214 (location_get_source_line): Convert return type from const char *
2215 to char_span, losing the final "line_size" param.
2217 2018-04-30 Jan Hubicka <jh@suse.cz>
2219 * lto-wrapper.c (ltrans_priorities): New static var.
2220 (cmp_priority): New.
2221 (run_gcc): Read priorities and if doing parallel build order
2222 the Makefile by them.
2224 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2226 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2228 2018-04-30 Richard Biener <rguenther@suse.de>
2230 * tree-cfg.c (verify_address): Remove base argument, add
2231 flag whether to check TREE_ADDRESSABLE and do that.
2232 (verify_expr): Remove.
2233 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2234 (verify_gimple_assign_single): Likewise.
2235 (verify_gimple_switch): Likewise.
2236 (verify_expr_location_1): Dereference tp once. Add (disabled)
2237 piece from verify_expr.
2238 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2240 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2242 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2244 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2246 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2247 (small_data_pattern): Likewise.
2248 (arc_rewrite_small_data): Likewise.
2249 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2250 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2251 (get_symbol_alignment): New function.
2252 (legitimate_small_data_address_p): Likewise.
2253 (legitimate_scaled_address): Update, call
2254 legitimate_small_data_address_p.
2255 (output_sdata): New static variable.
2256 (arc_print_operand): Update how we handle small data operands.
2257 (arc_print_operand_address): Likewise.
2258 (arc_legitimate_address_p): Update, use
2259 legitimate_small_data_address_p.
2260 (arc_rewrite_small_data_p): Remove.
2261 (arc_rewrite_small_data_1): Likewise.
2262 (arc_rewrite_small_data): Likewise.
2263 (small_data_pattern): Likewise.
2264 (compact_sda_memory_operand): Update to use
2265 legitimate_small_data_address_p and get_symbol_alignment.
2266 (prepare_move_operands): Don't rewite sdata pattern.
2267 (prepare_extend_operands): Remove.
2268 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2270 (zero_extendqisi2): Likewise.
2271 (zero_extendhisi2): Likewise.
2272 (extendqihi2): Likewise.
2273 (extendqisi2): Likewise.
2274 (extendhisi2): Likewise.
2278 * config/arc/constraints.md (Usd): Change it to memory constraint.
2280 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2282 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2283 as source of std instructions.
2284 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2285 allow 6-bit constants as source for store instructions.
2286 (movdi_insn): Update instruction pattern to allow 6-bit constants
2287 as source for store instructions.
2289 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2291 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2293 2018-04-30 Nathan Sidwell <nathan@acm.org>
2294 Sandra Loosemore <sandra@codesourcery.com>
2296 * dumpfile.c (dump_open): Allow '-' for stdout.
2297 * doc/invoke.texi (Developer Options): Document dump filename
2298 determination early. Document stdin/stdout selection.
2300 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2302 Microblaze Target: PIC data text relative
2304 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2305 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2307 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2308 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2309 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2310 New addressing mode for data-text relative position indepenedent code.
2311 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2312 'ADDRESS_SYMBOLIC_TXT_REL'.
2313 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2314 (microblaze_legitimate_pic_operand): Exclude function calls from
2315 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2316 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2318 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2319 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2320 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2321 for 'address + offset'.
2322 (microblaze_expand_prologue): Add new function prologue call for
2324 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2325 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2326 table in case of TARGET_PIC_DATA_TEXT_REL.
2327 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2328 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2329 Add new macros 'UNSPEC_TEXT',
2330 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2331 + exclude function calls from 'UNSPEC_PLT' in case of data text
2333 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2334 new target hook for generating address diff vector tables in case of
2336 * doc/tm.texi : Regenerate.
2337 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2338 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2339 of addr diff vector generation.
2340 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2341 target hook definition.
2342 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2343 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2344 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2345 option for data text relative.
2347 2018-04-30 Richard Biener <rguenther@suse.de>
2349 * tree-chrec.h (evolution_function_is_constant_p): Remove
2351 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2353 2018-04-30 Richard Biener <rguenther@suse.de>
2356 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2358 2018-04-30 Richard Biener <rguenther@suse.de>
2360 PR tree-optimization/28364
2361 PR tree-optimization/85275
2362 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2363 copying first exit test.
2365 2018-04-28 Mark Wielaard <mark@klomp.org>
2367 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2369 (dwarf_AT): Handle DW_AT_addr_base.
2370 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2372 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2375 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2376 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2377 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2378 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2380 2018-04-28 Richard Biener <rguenther@suse.de>
2382 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2383 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2384 to reflect use. Only add interesting stmts.
2386 2018-04-27 Martin Jambor <mjambor@suse.cz>
2389 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2390 the jump function allows for passing through aggregate values.
2392 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2394 * input.h (in_system_header_at): Convert from macro to inline
2396 (from_macro_expansion_at): Likewise.
2397 (from_macro_definition_at): Likewise.
2399 2018-04-27 Jeff Law <law@redhat.com>
2401 * config.gcc: Mark tile* targets as deprecated/obsolete.
2403 2018-04-27 Richard Biener <rguenther@suse.de>
2405 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2408 2018-04-27 Richard Biener <rguenther@suse.de>
2410 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2412 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2414 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2415 with Yd constraint. Set "preferred_for_speed" attribute from
2416 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2418 (*movdi_internal): Ditto.
2419 (movti_interunit splitters): Remove
2420 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2421 (movdi_interunit splitters): Ditto.
2422 * config/i386/constraints.md (Ye): Remove.
2423 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2425 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2428 * config/aarch64/constraints.md (Usg): Limit to 31.
2431 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2433 PR tree-optimization/85529
2434 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2435 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2436 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2437 zero extension or masking of the MSB bit.
2438 (optimize_range_tests): Add FIRST_BB argument, pass it through
2439 to optimize_range_tests_var_bound.
2440 (maybe_optimize_range_tests, reassociate_bb): Adjust
2441 optimize_range_tests callers.
2443 2018-04-26 Richard Biener <rguenther@suse.de>
2444 Jakub Jelinek <jakub@redhat.com>
2446 * cgraph.h (symbol_table): Just declare debug method here.
2447 * symtab.c (symbol_table::debug): Define.
2449 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2451 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2453 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2455 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2456 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2457 (*movdi_internal): Substitute Yi and Yj constraint with x
2458 and Ym and Yn constraint with y constraint. Update "isa"
2459 attribute and set "preferred_for_speed" attribute from
2460 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2461 (*movsi_internal): Ditto.
2462 (*movdf_internal): Ditto.
2463 (*movsf_internal): Ditto.
2464 (*zero_extendsidi2): Ditto.
2465 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2466 (sse2_loadld): Ditto.
2467 (*vec_extract<ssevecmodelower>_0): Ditto.
2468 (*vec_extractv4si_0_zext_sse4): Ditto.
2469 (vec_concatv2di): Ditto.
2470 (*vec_dup<mode>): Ditto.
2471 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2472 * config/i386/constraints.md (Yi): Remove.
2477 2018-04-26 Nathan Sidwell <nathan@acm.org>
2479 * dumpfile.c (dump_open): New.
2480 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2481 (dump_finish): Detect stdio/stderr by value not name.
2483 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2485 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2487 2018-04-26 Tom de Vries <tom@codesourcery.com>
2490 * config/nvptx/nvptx.c (verify_neutering_jumps)
2491 (verify_neutering_labels): New function
2492 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2494 2018-04-26 Tom de Vries <tom@codesourcery.com>
2497 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2498 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2501 2018-04-26 Richard Biener <rguenther@suse.de>
2502 Tom de Vries <tom@codesourcery.com>
2505 * lto-streamer-out.c (output_function): Fixup loops if required to match
2506 discovery done in the reader.
2508 2018-04-26 Richard Biener <rguenther@suse.de>
2510 PR tree-optimization/85116
2511 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2512 have a loop exit from the single latch predecessor. Remove
2513 case of header with just condition.
2514 (ch_base::copy_headers): Exclude infinite loops from any
2516 (pass_ch::execute): Record exits.
2518 2018-04-26 Richard Biener <rguenther@suse.de>
2520 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2521 prologue cost vector and pass it to vect_get_load_cost.
2522 (vect_get_peeling_costs_all_drs): Likewise.
2523 (vect_peeling_hash_get_lowest_cost): Likewise.
2524 (vect_enhance_data_refs_alignment): Likewise.
2526 2018-04-26 Richard Biener <rguenther@suse.de>
2529 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2530 checking of integer<->pointer conversions.
2531 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2532 sign-/zero-extending pointer types.
2533 (expand_omp_for_static_chunk): Likewise.
2535 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2536 Jean Lee <xiaoyur347@gmail.com>
2538 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2539 (TARGET_ASAN_SHADOW_OFFSET): Define.
2540 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2541 true for -fsanitize=address.
2543 2018-04-25 Mark Wielaard <mark@klomp.org>
2545 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2548 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2550 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2551 than "alu", remove explicit "memory" and "imm_disp" attributes.
2552 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2555 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2556 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2557 gen_lowpart_no_emit.
2559 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2562 * config/i386/i386.c (ix86_expand_builtin): Change memory
2563 operand to XI, extend p0 to Pmode.
2564 * config/i386/i386.md: Change unspec volatile and operand
2565 1 mode to XI, change operand 0 mode to P.
2567 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2569 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2570 GET_MODE_MASK before any checking.
2571 (nds32_can_use_bset_p): Likewise.
2572 (nds32_can_use_btgl_p): Likewise.
2574 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2576 * config/nds32/nds32-doubleword.md: New define_split pattern for
2577 illegal register number.
2579 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2581 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2583 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2585 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2587 2018-04-25 Richard Biener <rguenther@suse.de>
2589 * lto-streamer.h (LTO_major_version): Bump to 8.
2591 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2593 * BASE-VER: Set to 9.0.0.
2595 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2597 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2598 in __abskf2 and __powikf2.
2600 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2603 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2604 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2605 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2606 Use the above on operand 2. Reindent.
2607 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2609 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2612 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2614 * config/i386/i386.opt (mcet): Removed.
2615 * doc/install.texi: Remove -mcet documentation.
2616 * doc/invoke.texi: Likewise.
2618 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2621 * doc/install.texi: Remove -mcet from bootstrap-cet.
2623 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2626 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2627 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2631 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2632 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2633 containing a CONST_VECTOR.
2635 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2637 * doc/install.texi: Update newlib dependency for nvptx.
2639 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2642 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2643 instead of INTVAL when shifting x left.
2645 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2647 PR tree-optimization/85478
2648 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2649 vect_grouped_store_supported for single element vectors.
2651 2018-04-24 Richard Biener <rguenther@suse.de>
2654 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2655 load cost increase to the case of non-constant step.
2657 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2660 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2661 destination if any_malformed_asm.
2663 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2666 * expr.c (store_field): In the bitfield case, if the value comes from
2667 a function call and is returned in registers by means of a PARALLEL,
2668 do not change the mode of the temporary unless BLKmode and VOIDmode.
2670 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2672 PR rtl-optimization/85423
2673 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2674 dependencies to debug insns when the previous insn is non-debug.
2676 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2678 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2679 enums into a single definition.
2680 (fls): Fix predicates and printing.
2683 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2685 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2686 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2687 and short u6 immediate.
2688 (check_if_valid_sleep_operand): Remove.
2689 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2691 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2693 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2694 flag_always_save_lp condition.
2695 * config/nds32/nds32.opt (malways-save-lp): New option.
2697 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2699 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2700 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2701 * config/nds32/nds32.h
2702 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2703 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2705 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2707 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2708 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2710 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2711 Chung-Ju Wu <jasonwucj@gmail.com>
2713 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2715 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2716 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2718 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2720 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2722 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2724 * config/nds32/nds32-protos.h (nds32_data_alignment,
2725 nds32_local_alignment): Declare.
2726 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2727 nds32_local_alignment): New functions.
2728 (TARGET_CONSTANT_ALIGNMENT): Define.
2729 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2731 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2733 * config/nds32/nds32.c
2734 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2735 (TARGET_MODES_TIEABLE_P): Likewise.
2737 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2739 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2742 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2743 Chung-Ju Wu <jasonwucj@gmail.com>
2745 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2746 for unaligned access.
2747 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2749 * config/nds32/nds32-intrinsic.md: Likewise.
2750 * config/nds32/nds32_intrinsic.h: Likewise.
2751 * config/nds32/nds32.h (nds32_builtins): Likewise.
2752 * config/nds32/nds32.opt (munaligned-access): New option.
2753 * config/nds32/nds32.c (nds32_asm_file_start): Display
2754 flag_unaligned_access status.
2756 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2758 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2759 -mno-relax is present.
2760 * config/riscv/linux.h (LINK_SPEC): Ditto.
2762 2018-04-20 Martin Sebor <msebor@redhat.com>
2765 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2767 (gimple_fold_builtin_stxcpy_chk): Same.
2768 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2770 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2773 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2774 __powikf2 when long double is IEEE 128-bit.
2776 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2778 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2779 step to make sure stack always aligned.
2781 2018-04-20 Carl Love <cel@us.ibm.com>
2784 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2785 size check for arg0.
2787 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2788 Tom de Vries <tom@codesourcery.com>
2791 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2792 Emit insns for calls too.
2793 (nvptx_find_par): Always look for worker-level predecessor insn.
2794 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2796 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2797 (nvptx_process_pars): Propagate frames for calls.
2799 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2802 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2804 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2805 (ix86_handle_option): Don't handle OPT_mibt.
2806 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2808 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2810 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2811 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2812 (ix86_target_macros): Define __CET__ with flag_cf_protection
2813 for -fcf-protection.
2814 * config/i386/i386.c (isa2_opts): Remove -mibt.
2815 * config/i386/i386.h (TARGET_IBT): Removed.
2816 (TARGET_IBT_P): Likewise.
2817 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2818 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2819 * config/i386/i386.opt (mcet): Update help message.
2822 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2823 -mcet as an alias for -mshstk.
2825 2018-04-20 Richard Biener <rguenther@suse.de>
2828 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2829 complexity by forcing a single use of the multiply operand.
2831 2018-04-20 Martin Jambor <mjambor@suse.cz>
2834 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2835 recursion dependency to only apply to non-clones.
2837 2018-04-20 Martin Jambor <mjambor@suse.cz>
2840 * ipa-cp.c (create_specialized_node): Check that clones of
2841 self-recursive edges exist during IPA-CP.
2843 2018-04-19 Toon Moene <toon@moene.org>
2845 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2848 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2850 PR tree-optimization/85467
2851 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2852 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2853 VECTOR_CST element to type.
2855 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2858 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2859 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2860 (builtin_longjmp): Likewise.
2861 (save_stack_nonlocal): New pattern.
2862 (restore_stack_nonlocal): Likewise.
2864 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2867 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2868 Replace ASM_OUTPUT_LABEL with fprintf.
2870 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2873 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2874 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2875 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2876 define __IBT__ and __SHSTK__ for -fcf-protection.
2877 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2879 (ix86_trampoline_init): Likewise.
2880 (x86_output_mi_thunk): Likewise.
2881 (ix86_notrack_prefixed_insn_p): Likewise.
2882 (ix86_option_override_internal): Don't disallow -fcf-protection.
2883 * config/i386/i386.md (rdssp<mode>): Also enable for
2885 (incssp<mode>): Likewise.
2886 (nop_endbr): Likewise.
2887 * config/i386/i386.opt (mcet): Change help message to built-in
2891 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2892 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2893 enable CET built-in functions.
2895 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2897 * common/config/i386/i386-common.c
2898 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2899 OPTION_MASK_ISA_MOVDIRI_UNSET,
2900 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2901 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2902 * config.gcc (movdirintrin.h): New header.
2903 * config/i386/cpuid.h (bit_MOVDIRI,
2904 bit_MOVDIR64B): New bits.
2905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2907 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2908 (VOID, PVOID, PCVOID)): New function types.
2909 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2910 __builtin_ia32_directstoreu_u64,
2911 __builtin_ia32_movdir64b): New builtins.
2912 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2913 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2915 (ix86_valid_target_attribute_inner_p): Ditto.
2916 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2917 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2918 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2919 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2920 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2921 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2922 (movdiri<mode>, movdir64b_<mode>): New.
2923 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2924 * config/i386/immintrin.h: Include movdirintrin.h.
2925 * config/i386/movdirintrin.h: New file.
2926 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2928 2018-04-19 Richard Biener <rguenther@suse.de>
2931 * cfg.c (clear_bb_flags): When loop state says we have
2932 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2934 2018-04-19 Richard Biener <rguenther@suse.de>
2936 PR tree-optimization/84737
2937 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2938 copying restrict info.
2939 (vect_setup_realignment): Use it.
2940 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2941 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2942 the first DR to all generated stores.
2943 (vectorizable_load): Likewise for loads.
2945 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2947 PR tree-optimization/85446
2948 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2949 the integral and pointer types to have the same precision.
2951 * doc/install.texi: Document --disable-cet being the default and
2954 2018-04-18 Martin Liska <mliska@suse.cz>
2956 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2959 2018-04-18 Martin Liska <mliska@suse.cz>
2962 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2965 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2966 arguments if they are comparable.
2968 2018-04-18 Martin Liska <mliska@suse.cz>
2971 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2974 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2977 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2980 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2981 ENDBR after calling __morestack.
2983 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2986 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2987 by using gcc_base_ver to generate a gcc_driver_version, and use
2988 it when generating GCC_DRIVER_NAME.
2989 * configure: Regenerate.
2991 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2994 * config.gcc: Obsolete powerpc*-*-*spe*.
2996 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2999 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3000 (stabstr_D): Change type of unum from unsigned int to
3001 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3004 2018-04-17 Jim Wilson <jimw@sifive.com>
3007 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3008 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3009 Set arg_pointer_offset after using pretend_args_size.
3011 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3013 PR rtl-optimization/85431
3014 * dse.c (record_store): Ignore zero width stores.
3017 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3018 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3019 __builtin_stack_restore rather than after it.
3020 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3021 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3022 argument instead of virtual_dynamic_stack_rtx.
3024 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3026 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3028 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3029 Add note to error message to explain internal mapping of overloaded
3030 built-in function name to non-overloaded built-in function name.
3031 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3034 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3037 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3038 where the inputs overlap with the output.
3040 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3042 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3043 (=v, v) alternative and explicit "memory" attribute.
3044 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3045 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3047 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3048 "sselog1" type instead of "sselog".
3049 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3050 "sselog". Remove explicit "memory" attribute.
3051 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3052 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3054 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3055 "sselog1" type instead of "sselog". Remove explicit "memory"
3057 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3058 use "sselog1" type instead of "sselog". Remove explicit "memory"
3060 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3061 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3063 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3064 "sselog1" type instead of "sselog". Remove explicit "memory"
3066 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3067 use "sselog1" type instead of "sselog". Remove explicit "memory"
3071 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3074 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3077 2018-04-17 Martin Jambor <mjambor@suse.cz>
3080 * ipa-cp.c (create_specialized_node): Call
3081 expand_all_artificial_thunks if necessary.
3083 2018-04-17 Martin Liska <mliska@suse.cz>
3086 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3087 in message, remote space in between '_G' and '('.
3089 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3092 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3093 avx512f_vmcmp<mode>3<round_saeonly_name>,
3094 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3095 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3096 avx512f_rndscale<mode><round_saeonly_name>,
3097 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3098 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3099 Use %<iptr>2 instead of %2 for -masm=intel.
3100 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3101 avx512f_vcvttss2usi<round_saeonly_name>,
3102 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3104 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3105 avx512f_vcvttsd2usi<round_saeonly_name>,
3106 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3107 Use %q1 instead of %1 for -masm=intel.
3108 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3109 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3110 of %3 for -masm=intel.
3111 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3113 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3115 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3117 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3118 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3120 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3122 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3123 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3124 %g1 and one with %0 and %1.
3125 (avx512er_vmrcp28<mode><round_saeonly_name>,
3126 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3128 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3129 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3130 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3131 of %0 and %{%4%} for -masm=intel.
3132 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3133 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3134 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3135 order of %0 and %{%5%}%{z%} for -masm=intel.
3137 2018-04-17 Jan Hubicka <jh@suse.cz>
3140 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3142 2018-04-17 Martin Liska <mliska@suse.cz>
3145 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3146 for target_clone error message. Make default implementation
3147 clone to be a local declaration.
3148 (separate_attrs): Add new argument and check for an empty
3150 (expand_target_clones): Handle it.
3151 (ipa_target_clone): Make redirection just for target_clones
3154 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3155 Tom de Vries <tom@codesourcery.com>
3158 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3159 tiled basic blocks without omp continue statements.
3161 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3164 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3165 vec_extract expression as having side effects to make sure it gets
3168 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3171 * config/i386/i386.c (get_builtin_code_for_version): Check
3174 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3177 * gcc/config.gcc: Support "skylake".
3178 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3180 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3181 (processor_target_table): Add "skylake".
3182 (ix86_option_override_internal): Add "skylake".
3183 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3184 PROCESSOR_CANNONLAKE.
3185 (get_builtin_code_for_version): Fix priority for
3186 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3187 PROCESSOR_SKYLAKE-AVX512.
3188 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3189 (processor_type): Add PROCESSOR_SKYLAKE.
3191 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3192 Jason Merrill <jason@redhat.com>
3195 * convert.c (convert_to_integer_1): Use direct recursion for
3196 enumeral types and types with a precision less than the number
3197 of bits in their mode.
3199 2018-04-16 Julia Koval <julia.koval@intel.com>
3202 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3203 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3205 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3208 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3209 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3210 and -mno-direct-move.
3212 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3215 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3216 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3217 Ensure negative shifts result in {0}.
3219 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3221 PR rtl-optimization/79916
3222 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3223 regs (if any) to define how to gnerate SD moves when LRA is in
3226 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3228 PR rtl-optimization/85393
3229 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3230 * except.c (expand_dw2_landing_pad_for_region): Make static.
3231 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3232 a label and unconditional jump to old_bb, rather than
3233 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3236 PR rtl-optimization/85376
3237 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3238 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3239 instead of a specific value.
3241 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3242 Bin Cheng <bin.cheng@arm.com>
3244 PR tree-optimization/82965
3245 PR tree-optimization/83991
3246 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3247 by_profile_only parameter.
3248 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3249 information if the loop was predicted to iterate too many times.
3250 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3252 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3255 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3258 2018-04-13 Martin Liska <mliska@suse.cz>
3259 Jakub Jelinek <jakub@redhat.com>
3262 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3263 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3264 * builtins.c (expand_builtin_memory_copy_args): Use
3265 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3266 handle dest_addr == pc_rtx.
3268 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3271 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3272 asked to not generate direct moves.
3273 (fix_trunc<mode>si2_stfiwx): Similar.
3274 (fix_trunc<mode>si2_internal): Similar.
3276 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3279 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3280 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3281 lookup if dest in some wider mode is known to be const0_rtx and
3282 if so, record permanent equivalence for it to be ZERO_EXTEND of
3283 the narrower mode destination.
3285 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3287 * lto-streamer-out.c (output_function): Revert 259346.
3288 * omp-expand.c (expand_oacc_for): Likewise.
3290 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3292 PR rtl-optimization/85354
3293 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3294 * sel-sched.c (sel_global_init): ... here.
3296 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3299 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3300 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3301 mode for PE-COFF targets.
3302 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3303 (i386_pe_asm_lto_end): Likewise.
3304 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3305 (TARGET_ASM_LTO_END): Likewise.
3306 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3307 (i386_pe_asm_lto_start): New function.
3308 (i386_pe_asm_lto_end): Likewise.
3310 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3311 Richard Biener <rguenther@suse.de>
3314 * lto-streamer-out.c (output_function): Fix CFG loop state before
3316 * omp-expand.c (expand_oacc_for): Handle calls to internal
3317 functions like regular functions.
3319 2018-04-12 Richard Biener <rguenther@suse.de>
3322 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3323 for the early LTO debug to properly generate references to it
3324 during DIE emission. Do not re-use that for the skeleton for
3326 (dwarf2out_early_finish): Likewise.
3328 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3331 * config/i386/sse.md
3332 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3333 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3334 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3335 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3336 and output is a reg, avoid creating invalid lowpart subreg, but
3337 instead split into a 512-bit move. Don't split if not AVX512VL,
3338 input is xmm16+ reg and output is a mem.
3339 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3340 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3341 xmm16+ reg and output is a mem.
3343 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3345 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3346 also for flag_dwarf2_cfi_asm.
3348 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3350 PR rtl-optimization/85342
3351 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3352 a bool scalar var inside of the loop instead. Don't try to update
3353 recog_data.operand after failed apply_change_group.
3355 2018-04-12 Tom de Vries <tom@codesourcery.com>
3358 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3359 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3360 array with flexible array member as array without given dimension.
3361 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3362 argument for undefined param to true.
3364 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3367 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3368 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3369 from PowerPC section.
3370 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3371 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3372 help text that is too long.
3373 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3374 help text that is too long.
3375 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3376 help text that is too long.
3378 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3380 * config/alpha/alpha.md (stack_probe_internal): Rename
3381 from "probe_stack". Update all callers.
3383 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3385 PR rtl-optimization/84566
3386 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3387 sched_macro_fuse_insns.
3389 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3392 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3393 (compute_block_dependences): ... from here.
3395 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3397 PR tree-optimization/85331
3398 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3399 from int to HOST_WIDE_INT.
3401 2018-04-11 Martin Jambor <mjambor@suse.cz>
3404 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3405 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3406 not the same as the source val.
3407 (cgraph_edge_brings_value_p): New parameter.
3408 (gather_edges_for_value): Pass destination value to
3409 cgraph_edge_brings_value_p.
3410 (perhaps_add_new_callers): Likewise.
3411 (get_info_about_necessary_edges): Likewise and exclude values brought
3412 only by self-recursive edges.
3413 (create_specialized_node): Redirect only clones of self-calling edges.
3414 (+self_recursive_pass_through_p): New function.
3415 (find_more_scalar_values_for_callers_subset): Use it.
3416 (find_aggregate_values_for_callers_subset): Likewise.
3417 (known_aggs_to_agg_replacement_list): Removed.
3418 (decide_whether_version_node): Re-calculate known constants for all
3419 remaining context clones.
3421 2018-04-11 Richard Biener <rguenther@suse.de>
3424 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3425 from early DWARF output.
3426 (dwarf2out_early_finish): Output line info unconditionally into
3427 early DWARF and add reference to it.
3429 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3432 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3433 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3434 other than V2DFmode using iptr mode attribute.
3435 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3437 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3439 PR rtl-optimization/84659
3440 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3442 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3445 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3447 (output_loc_list): Pass address of a dummy size variable even in the
3448 locview handling loop.
3449 (index_location_lists): Add comment on why skip_loc_list_entry can't
3452 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3455 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3458 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3461 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3462 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3463 and -mstring-compare-inline-limit.
3465 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3468 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3469 for stack clash protection in a register whenever we need it to be in
3472 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3474 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3475 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3477 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3480 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3482 (mlong-double-): Ditto.
3483 * config/rs6000/sysv4.opt (msdata=): Ditto.
3484 (mtls-size=): Ditto.
3486 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3488 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3489 erroneous entries for
3490 "vector int vec_ldl (int, long int *)", and
3491 "vector unsigned int vec_ldl (int, unsigned long int *)".
3492 Add comments and entries for
3493 "vector bool char vec_ldl (int, bool char *)",
3494 "vector bool short vec_ldl (int, bool short *)",
3495 "vector bool int vec_ldl (int, bool int *)",
3496 "vector bool long long vec_ldl (int, bool long long *)",
3497 "vector pixel vec_ldl (int, pixel *)",
3498 "vector long long vec_ldl (int, long long *)",
3499 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3500 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3501 type tree bool_long_long_type_node and correct definition of
3502 bool_V2DI_type_node to make reference to this new type tree.
3503 (rs6000_mangle_type): Replace erroneous reference to
3504 bool_long_type_node with bool_long_long_type_node.
3505 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3506 comments to emphasize sign distinctions for char and int types and
3507 replace RS6000_BTI_bool_long constant with
3508 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3509 use of RS6000_BTI_pixel.
3510 (bool_long_type_node): Remove this macro definition.
3511 (bool_long_long_type_node): New macro definition
3513 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3515 PR rtl-optimization/85300
3516 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3517 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3518 simplify_unary_operation fails.
3520 2018-04-10 Martin Liska <mliska@suse.cz>
3522 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3523 cgraph_edge and ipa_ref.
3525 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3529 * config/i386/sse.md
3530 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3531 computation of the VEC_MERGE selector from mask.
3532 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3533 Fix decoding of the VEC_MERGE selector into mask.
3535 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3537 PR tree-optimization/85286
3538 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3540 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3542 * final.c (final_1): Set insn_last_address as well as
3543 insn_current_address.
3545 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3548 * explow.c (emit_stack_probe): Call validize_mem on memory location
3549 before passing it to gen_probe_stack. Create address operand and
3550 legitimize it for the probe_stack_address case.
3552 2018-04-09 Jan Hubicka <jh@suse.cz>
3555 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3556 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3557 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3559 (free_lang_data): Rebuild type inheritance graph.
3561 2018-04-09 Martin Sebor <msebor@redhat.com>
3563 * invoke.texi (-finline-small-functions): Mention other optimization
3565 (-findirect-inlining, -fpartial-inlining): Same.
3566 (-finline-functions-called-once): Same.
3567 (-freorder-blocks-and-partition): Same.
3569 2018-04-09 Jan Hubicka <jh@suse.cz>
3572 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3573 jumps; choose last target that matches the criteria (i.e.
3574 no partition changes for non-crossing jumps).
3575 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3576 support for redirecting crossing jumps to non-crossing.
3578 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3580 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3581 also for naked functions.
3583 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3585 * config/arc/arc.md (add_shift): New pattern.
3586 (add_shift2): Likewise.
3587 (sub_shift): Likewise.
3588 (sub_shift_cmp0_noout): Likewise.
3589 (compare_si_ashiftsi): Likewise.
3590 (xbfu_cmp0_noout): New combine pattern.
3591 (xbfu_cmp0"): Likewise.
3592 (movsi_set_cc_insn): Place the predicable variant first.
3593 (commutative_binary_cmp0_noout): Remove clobber.
3594 (commutative_binary_cmp0): New pattern.
3595 (noncommutative_binary_cmp0): Likewise.
3596 (noncommutative_binary_cmp0_noout): Likewise.
3597 (noncommutative_binary_comparison_result_used): Removed.
3598 (rsub_cmp0): New pattern.
3599 (rsub_cmp0_noout): Likewise.
3600 (extzvsi): Changed, keep only meaningful variants.
3601 (SQH, SEZ): New iterators.
3602 (SQH_postfix): New mode attribute.
3603 (SEZ_prefix): New code attribute.
3604 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3605 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3606 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3608 (noncommutative_operator): Check the availability of barrel
3611 2018-04-09 Richard Biener <rguenther@suse.de>
3613 PR tree-optimization/85284
3614 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3615 Only use the niter constraining form of simple_iv when the exit
3618 2018-04-09 Tom de Vries <tom@codesourcery.com>
3621 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3622 (define_expand "*memory_barrier"): New define_expand.
3623 (define_insn "memory_barrier"): New insn.
3625 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3627 PR rtl-optimization/80463
3628 PR rtl-optimization/83972
3629 PR rtl-optimization/83480
3631 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3632 correct producer for the insn.
3633 (tidy_control_flow): Fixup seqnos in case of debug insns.
3635 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3637 PR rtl-optimization/83913
3639 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3640 different sched-times when merging exprs.
3642 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3644 PR rtl-optimization/83962
3646 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3647 tidy_fallthru_edge and tidy_control_flow.
3649 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3651 PR rtl-optimization/83530
3653 * sel-sched.c (force_next_insn): New global variable.
3654 (remove_insn_for_debug): When force_next_insn is true, also leave only
3655 next insn in the ready list.
3656 (sel_sched_region): When the region wasn't scheduled, make another pass
3657 over it with force_next_insn set to 1.
3659 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3661 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3663 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3664 for interrupt control.
3665 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3666 functions for interrupt control.
3667 * config/nds32/nds32-intrinsic.md: Likewise.
3668 * config/nds32/nds32_intrinsic.h: Likewise.
3669 * config/nds32/nds32.h (nds32_builtins): Likewise.
3671 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3673 * config/nds32/nds32.c (nds32_init_machine_status,
3674 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3675 strict_aligned_p field.
3676 (nds32_expand_to_rtl_hook): New function.
3677 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3678 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3680 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3681 Chung-Ju Wu <jasonwucj@gmail.com>
3683 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3684 * config/nds32/nds32-n7.md: New file.
3685 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3686 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3688 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3689 * config/nds32/nds32.md (pipeline_model): Add n7.
3690 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3691 * config/nds32/pipelines.md: Include n7 settings.
3693 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3694 Chung-Ju Wu <jasonwucj@gmail.com>
3696 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3697 * config/nds32/nds32-e8.md: New file.
3698 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3699 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3701 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3702 * config/nds32/nds32.md (pipeline_model): Add e8.
3703 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3704 * config/nds32/pipelines.md: Include e8 settings.
3706 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3707 Chung-Ju Wu <jasonwucj@gmail.com>
3709 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3710 * config/nds32/nds32-n8.md: New file.
3711 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3712 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3714 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3715 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3716 * config/nds32/nds32.md (pipeline_model): Add n8.
3717 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3718 * config/nds32/pipelines.md: Include n8 settings.
3720 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3721 Chung-Ju Wu <jasonwucj@gmail.com>
3723 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3724 * config/nds32/nds32-n9-2r1w.md: New file.
3725 * config/nds32/nds32-n9-3r2w.md: New file.
3726 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3727 nds32_register_ports): New or modify for cpu n9.
3728 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3730 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3731 * config/nds32/nds32-utils.c: New file.
3732 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3733 TARGET_MUL_SLOW): Define.
3734 * config/nds32/nds32.md (pipeline_model): New attribute.
3735 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3736 New options that support cpu n9.
3737 * config/nds32/pipelines.md: Include n9 settings.
3738 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3740 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3742 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3743 information if necessary.
3744 (output_cond_branch_compare_zero): Likewise.
3745 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3746 (nds32_target_alignment): Refine for alignment.
3747 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3748 (FUNCTION_BOUNDARY): Modify.
3749 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3751 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3753 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3755 * config/nds32/constants.md (unspec_volatile_element): Add values for
3756 TLB operation and data prefetch.
3757 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3758 functions for TLB operation and data prefetch.
3759 * config/nds32/nds32-intrinsic.md: Likewise.
3760 * config/nds32/nds32_intrinsic.h: Likewise.
3761 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3762 (nds32_print_operand): Likewise.
3763 * config/nds32/nds32.h (nds32_builtins): Likewise.
3765 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3766 Andrew Pinski <pinsika@gcc.gnu.org>
3769 * match.pd: Use constant_boolean_node of correct type instead of
3770 boolean_true_node or boolean_false_node for simplifying
3771 pointer comparisons to zero.
3773 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3775 PR tree-optimization/80021
3776 * tree.c (verify_type_variant): Make error call in verify_variant_match
3777 translatable and remove final full stop.
3779 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3781 * config/nds32/constants.md (unspec_volatile_element): Add
3782 UNSPEC_VOLATILE_EH_RETURN.
3783 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3784 nds32_output_stack_pop): Support dwarf exception handling process.
3785 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3786 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3787 exception handling process.
3788 (nds32_compute_stack_frame): Likewise.
3789 (nds32_return_addr_rtx): Likewise.
3790 (nds32_initial_elimination_offset): Likewise.
3791 (nds32_expand_prologue): Likewise.
3792 (nds32_expand_epilogue): Likewise.
3793 (nds32_dynamic_chain_address): New function.
3794 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3796 (DYNAMIC_CHAIN_ADDRESS): Define.
3797 (EH_RETURN_DATA_REGNO): Define.
3798 (EH_RETURN_STACKADJ_RTX): Define.
3799 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3800 patterns for dwarf exception handling.
3802 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3804 * config/nds32/nds32.h: Clean up obsolete macros.
3806 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3808 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3809 Add enum values for particular instructions.
3810 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3811 particular intrinsic functions.
3812 * config/nds32/nds32-intrinsic.md: Likewise.
3813 * config/nds32/nds32_intrinsic.h: Likewise.
3814 * config/nds32/nds32.h (nds32_builtins): Likewise.
3815 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3816 (btst, ave): New patterns for particular instructions.
3818 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3820 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3821 Add enum values for atomic load/store and memory sync.
3822 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3824 * config/nds32/nds32-intrinsic.md: Likewise.
3825 * config/nds32/nds32_intrinsic.h: Likewise.
3826 * config/nds32/nds32.h (nds32_builtins): Likewise.
3828 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3830 PR tree-optimization/85257
3831 * fold-const.c (native_encode_vector): If not all elts could fit
3832 and off is -1, return 0 rather than offset.
3833 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3834 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3835 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3836 adjust buffer in native_interpret_expr call.
3838 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3840 * config/nds32/constants.md (unspec_volatile_element): Add cache
3841 control enum values.
3842 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3843 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3844 * config/nds32/nds32.c (nds32_cctl_names): New.
3845 (nds32_print_operand): Handle cache control register names.
3846 * config/nds32/nds32.h (nds32_builtins): New enum values.
3847 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3849 * config/nds32/nds32.md (type): Add mmu.
3850 * config/nds32/pipelines.md (simple_insn): Add mmu.
3852 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32.md (type): Remove call.
3855 * config/nds32/pipelines.md (simple_insn): Likewise.
3857 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3859 * config/nds32/constants.md (unspec_volatile_element): Add
3860 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3861 UNSPEC_VOLATILE_FMFCFG.
3862 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3863 description for fmfcfg and fmfcsr.
3864 (bdesc_1arg): Add fmtcsr.
3865 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3866 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3867 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3868 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3869 unspec_fmfcfg): New patterns.
3870 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3871 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3872 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3873 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3874 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3875 __nds32__fmfcfg): Define.
3877 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3879 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3880 intrinsic register names.
3881 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3882 intrinsic register enum values and macros.
3884 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3886 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3887 for load/store addressing form.
3888 (nds32_print_operand_address): Likewise.
3890 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3893 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3894 based on LABEL_REF. Remove useless assertion.
3895 (pic_address_needs_scratch): Fix formatting.
3896 (sparc_legitimize_pic_address): Minor tweaks.
3897 (sparc_delegitimize_address): Adjust assertion accordingly.
3898 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3899 into symbolic_operand.
3900 (movsi_high_pic_label_ref): Likewise.
3901 (movsi_lo_sum_pic_label_ref): Likewise.
3902 (movdi_pic_label_ref): Likewise.
3903 (movdi_high_pic_label_ref): Likewise.
3904 (movdi_lo_sum_pic_label_ref): Likewise.
3906 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3908 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3909 custom LIB_SPEC setup.
3911 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3912 Kito Cheng <kito.cheng@gmail.com>
3914 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3915 * config/riscv/freebsd.h: New.
3917 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3919 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3920 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3923 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3924 Kito Cheng <kito.cheng@gmail.com>
3926 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3927 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3928 * config/nds32/nds32-protos.h (nds32_output_call,
3929 nds32_output_return): Declare.
3930 * config/nds32/nds32.md: Refine all the call and return patterns.
3932 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3935 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3936 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3938 PR rtl-optimization/84872
3939 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3940 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3943 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3945 * expr.c (copy_blkmode_to_reg): Revert 254862.
3946 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3948 2018-04-06 Richard Biener <rguenther@suse.de>
3951 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3952 after seeing a component reference with an adjacent field. Treat
3953 refs to arrays at struct end of external decls similar to
3954 refs to unconstrained commons.
3956 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3959 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3960 look through SAVE_EXPRs with non-side-effects argument. Adjust
3962 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3965 2018-04-06 Richard Biener <rguenther@suse.de>
3968 * alias.c (find_base_term): New wrapper around find_base_term
3969 unwinding CSELIB_VAL_PTR changes.
3970 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3973 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3975 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3977 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3978 constant definitions.
3979 ("nop"): lr 0,0 -> nopr r0
3980 ("nop_lr0", "nop_lr1"): New insn definitions.
3982 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3984 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3985 NDS32_V3PUSH_AVAILABLE_P macro.
3987 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3988 Chung-Ju Wu <jasonwucj@gmail.com>
3990 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3991 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3992 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3993 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3994 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3995 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3996 * config/nds32/constraints.md: New constraints and checking for hard
3997 float configuration.
3998 * config/nds32/iterators.md: New mode iterator and attribute for hard
3999 float configuration.
4000 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4002 * config/nds32/nds32-fpu.md: New file.
4003 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4004 deal with hard float code generation.
4005 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4007 (abi_type, float_reg_number): New enum type.
4008 * config/nds32/nds32-predicates.c: New predicates for hard float.
4009 * config/nds32/nds32-protos.h: Declare functions for hard float.
4010 * config/nds32/nds32.c: Implementation for hard float configuration.
4011 * config/nds32/nds32.h: Definitions for hard float configuration.
4012 * config/nds32/nds32.md: Include hard float machine description and
4013 modify patterns for hard float configuration.
4014 * config/nds32/nds32.opt: New options for hard float configuration.
4015 * config/nds32/predicates.md: New predicates for hard float
4018 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4020 * common/config/nds32/nds32-common.c
4021 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4023 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4026 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4027 CONSTRUCTOR_ELT (ctor, ...)->value.
4029 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4032 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4034 2018-04-05 Tom de Vries <tom@codesourcery.com>
4037 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4040 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4041 Kito Cheng <kito.cheng@gmail.com>
4043 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4044 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4045 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4046 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4048 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4049 Kito Cheng <kito.cheng@gmail.com>
4051 * config/nds32/constraints.md (Ufe): New memory constraint.
4052 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4053 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4054 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4056 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4057 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4059 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4061 * config/nds32/nds32.md: Use optimize_size in the condition for
4062 alu-shift instructions.
4064 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4066 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4068 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4070 * config/nds32/nds32.md (negsi2): Refine pattern.
4072 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4073 Chung-Ju Wu <jasonwucj@gmail.com>
4075 * config/nds32/iterators.md (shift_rotate): New code iterator.
4076 (shift): New code attribute.
4077 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4078 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4079 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4080 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4081 bit-wise operations.
4082 (andsi3, *andsi3): Ditto.
4083 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4084 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4085 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4086 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4087 nds32_ior_operand, nds32_xor_operand): New predicates.
4089 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4091 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4092 (addsi3, subsi3): ... this.
4094 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4096 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4098 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4100 * config/nds32/nds32.md: Adjust indention.
4102 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4104 * config/nds32/nds32.md (feature): New attribute.
4106 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4108 * config/nds32/nds32.md (subtype): New attribute.
4110 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4113 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4114 expansion to perform a bitwise AND of the argument followed by a
4115 boolean negation of the result.
4117 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4119 PR rtl-optimization/84878
4120 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4121 the basic block. Assert the use reference is not artificial and that
4122 it has an associated insn.
4124 2018-04-04 Michael Matz <matz@suse.de>
4126 * builtins.c (compute_objsize): Pass correct operand
4127 to array_at_struct_end_p.
4129 2018-04-04 Richard Biener <rguenther@suse.de>
4132 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4133 from contexts for DINFO_LEVEL_TERSE and below.
4135 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4137 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4138 resiter_operand condition.
4139 * config/nds32/nds32.md (*move<mode>): Ditto.
4141 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4142 Monk Chiang <sh.chiang04@gmail.com>
4144 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4146 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4148 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4150 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4151 Kito Cheng <kito.cheng@gmail.com>
4153 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4154 nds32_cond_code_str, output_cond_branch,
4155 output_cond_branch_compare_zero, nds32_expand_cbranch,
4156 nds32_expand_cstore, nds32_expand_movcc,
4157 nds32_output_cbranchsi4_equality_zero,
4158 nds32_output_cbranchsi4_equality_reg,
4159 nds32_output_cbranchsi4_equality_reg_or_const_int,
4160 nds32_output_cbranchsi4_greater_less_zero: New functions.
4161 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4162 nds32_expand_cstore, nds32_expand_movcc,
4163 nds32_output_cbranchsi4_equality_zero,
4164 nds32_output_cbranchsi4_equality_reg,
4165 nds32_output_cbranchsi4_equality_reg_or_const_int,
4166 nds32_output_cbranchsi4_greater_less_zero): Declare.
4167 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4168 nds32_rimm11s_operand): New predicates.
4169 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4170 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4173 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4175 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4176 * config/nds32/nds32.md: Ditto.
4177 * config/nds32/pipelines.md: Ditto.
4179 2018-04-04 Richard Biener <rguenther@suse.de>
4181 PR tree-optimization/85168
4182 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4183 propagating abnormals.
4185 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4187 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4189 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4190 Kito Cheng <kito.cheng@gmail.com>
4192 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4193 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4194 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4195 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4196 * config/nds32/nds32.md (sibcall_internal): New.
4197 (sibcall_register): Remove.
4198 (sibcall_immediate): Remove.
4199 (sibcall_value_internal): New.
4200 (sibcall_value_register): Remove.
4201 (sibcall_value_immediate): Remove.
4202 * config/nds32/predicates.md (nds32_general_register_operand): New.
4203 (nds32_call_address_operand): New.
4205 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4207 PR rtl-optimization/85167
4208 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4209 bb_defs if *split_p, instead preinitialize it to NULL.
4211 PR tree-optimization/85156
4212 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4213 evaluating the argument multiple times.
4215 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4217 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4219 (_mm_cvtpd_ps): Likewise.
4220 (_mm_cvttpd_epi32): Likewise.
4221 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4222 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4223 vector, pixel, and bool following altivec.h include.
4225 2018-04-03 Martin Sebor <msebor@redhat.com>
4227 * doc/extend.texi (Common Function Attributes): Clarify.
4228 (const attribute): Likewise.
4229 (pure attribute): Likewise.
4231 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4234 * config/i386/i386.c (ix86_expand_vector_set): Use
4235 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4237 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4239 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4240 instructions when changing rounding bits to preserve precision bits
4241 in the x87 control word.
4243 2018-04-03 Martin Liska <mliska@suse.cz>
4245 PR tree-optimization/82491
4246 * rtl.h (strip_offset_and_add): Replace += suboffset with
4247 poly_uint64 () + suboffset.
4249 2018-03-29 Martin Liska <mliska@suse.cz>
4250 Martin Jambor <mjambor@suse.cz>
4253 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4254 param_type is not an integral or pointer type.
4256 2018-04-03 Richard Biener <rguenther@suse.de>
4258 * sese.h (recompute_all_dominators): Remove.
4260 2018-04-02 Martin Sebor <msebor@redhat.com>
4262 * doc/invoke.texi (-Wrestrict): Fix typos.
4264 2018-04-02 Jim Wilson <jimw@sifive.com>
4266 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4267 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4268 (<optab>di3, <optab>si3_extend): Likewise.
4269 (<optab>si3_mask, <optab>si3_mask_1): New.
4270 (<optab>di3_mask, <optab>di3_mask_1): New.
4271 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4272 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4273 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4275 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4277 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4280 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4282 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4283 (nds32_canonicalize_comparison): New function.
4285 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4286 Kito Cheng <kito.cheng@gmail.com>
4287 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4289 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4290 * config/nds32/constants.md (unspec_volatile_element): Add
4291 UNSPEC_VOLATILE_RELAX_GROUP.
4292 * config/nds32/nds32-relax-opt.c: New file.
4293 * config/nds32/nds32-predicates.c
4294 (nds32_symbol_load_store_p): New function.
4295 * config/nds32/nds32-protos.h
4296 (nds32_symbol_load_store_p): Declare function.
4297 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4298 * config/nds32/nds32.c
4299 (nds32_register_pass): New function to register pass.
4300 (nds32_register_passes): New function to register passes.
4301 * config/nds32/nds32.md (relax_group): New pattern.
4302 * config/nds32/nds32.opt (mrelax-hint): New option.
4303 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4305 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4307 * config/nds32/t-nds32: Modify files dependency.
4309 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4311 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4312 (PROFILE_HOOK): Define its implementation.
4314 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4316 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4317 type and 32-bit size.
4319 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4322 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4323 (V_128_256): New mode iterator.
4324 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4325 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4326 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4328 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4329 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4331 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4334 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4335 NaN inputs correctly.
4337 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4340 * config/rs6000/vsx.md (??r): New mode attribute.
4341 (*vsx_mov<mode>_64bit): Use it.
4342 (*vsx_mov<mode>_32bit): Likewise.
4344 2018-03-30 Martin Sebor <msebor@redhat.com>
4346 PR tree-optimization/84818
4347 * builtins.c (check_access): Use warning_n.
4349 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4352 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4354 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4357 2018-03-30 Julia Koval <julia.koval@intel.com>
4360 * x86-tune.def (movx, partial_reg_dependency): Enable for
4363 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4366 * lra-constraints.c (process_alt_operands): Move setting
4367 this_alternative_matches below.
4369 2018-03-29 Martin Liska <mliska@suse.cz>
4372 * doc/invoke.texi: Document how LTO works with debug info.
4373 Describe auto-load support of binutils. Mention 'x86-64'
4374 as valid option value of -march option.
4376 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4378 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4381 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4382 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4383 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4386 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4389 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4390 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4391 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4392 for RS6000_BTM_POWERPC64.
4393 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4394 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4395 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4400 2018-03-28 Carl Love <cel@us.ibm.com>
4403 2017-09-27 Carl Love <cel@us.ibm.com>
4405 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4406 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4407 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4410 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4412 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4413 instead of __vector bool.
4414 (_mm_max_pu8): Likewise.
4415 (_mm_min_pi16): Likewise.
4417 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4420 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4421 (DIVWEUO): Likewise.
4423 (DIVDEUO): Likewise.
4424 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4425 DIVWEUO and DIVDEUO.
4426 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4427 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4428 (div_extend): Likewise.
4429 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4431 (__builtin_divweuo): Likewise.
4432 (__builtin_divdeo): Likewise.
4433 (__builtin_divdeuo): Likewise.
4435 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4438 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4439 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4441 PR tree-optimization/82004
4442 * gimple-match-head.c (optimize_pow_to_exp): New function.
4443 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4444 Don't fold to exp if optimize_pow_to_exp is false.
4446 2018-03-28 Martin Liska <mliska@suse.cz>
4449 * calls.c (initialize_argument_information): Fix trailing space.
4450 * common.opt: Fix typo and provide better explanation for
4451 -fsanitize-coverage option.
4452 * config/i386/i386.opt: Fix typo.
4454 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4455 Martin Liska <mliska@suse.cz>
4458 * gimplify.c (asan_poison_variable): Don't do the check for
4459 gimplify_omp_ctxp here.
4460 (gimplify_decl_expr): Do it here.
4461 (gimplify_target_expr): Likewise.
4463 2018-03-28 Martin Liska <mliska@suse.cz>
4466 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4467 chkp_type_bounds_count if MPX is not enabled.
4469 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4471 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4473 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4476 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4477 function to create the function decl for complex long double
4478 multiply and divide for -mabi=ieeelongdouble.
4479 (init_float128_ieee): Call it.
4481 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4484 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4485 -fcf-protection=branch -mibt.
4486 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4488 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4491 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4493 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4496 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4497 extern array declarations.
4499 2018-03-27 Richard Biener <rguenther@suse.de>
4502 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4503 explicit single_use checks.
4505 2018-03-27 Richard Biener <rguenther@suse.de>
4507 PR tree-optimization/85082
4508 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4511 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4513 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4514 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4515 Turn on fasynchronous-unwind-tables and funwind-tables.
4517 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4520 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4521 (*bmi_blsr_<mode>_ccz): Ditto.
4523 2018-03-26 Tom de Vries <tom@codesourcery.com>
4525 PR tree-optimization/85063
4526 * omp-general.c (offloading_function_p): New function. Factor out
4528 * omp-offload.c (pass_omp_target_link::gate): ... here.
4529 * omp-general.h (offloading_function_p): Declare.
4530 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4531 with attribute omp declare target for offloading functions.
4533 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4535 PR tree-optimization/84005
4536 * tree-data-ref.h (get_base_for_alignment): Declare.
4537 * tree-data-ref.c (get_base_for_alignment_1): New function.
4538 (get_base_for_alignment): Likewise.
4539 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4540 get_base_for_alignment to find a suitable base object, instead
4541 of always using drb->base_address.
4543 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4546 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4547 known size by default.
4549 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4552 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4553 and non BLKmode operands.
4555 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4558 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4559 Clean up attributes.
4561 2018-03-23 Richard Biener <rguenther@suse.de>
4564 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4565 we are going to emit early debug for LTO.
4567 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4570 * function.c (match_asm_constraints_1): Don't optimize if input
4571 doesn't satisfy general_operand predicate for output's mode.
4574 * alias.c (write_dependence_p): Don't require for x_canonicalized
4575 non-VOIDmode if x has VOIDmode.
4578 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4579 just don't try to optimize it rather than assert it never happens.
4581 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4583 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4584 macro expansions for definition of ST_INTERNAL_<mode> and
4585 LD_INTERNAL_<mode> builtins.
4586 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4588 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4590 (altivec_expand_st_builtin): Likewise.
4591 (altivec_expand_builtin): Remove calls to deleted functions.
4592 (rs6000_address_for_altivec): Delete this function.
4593 * config/rs6000/vector.md: Remove expands for
4594 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4596 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4599 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4600 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4601 re-computing once computed.
4602 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4603 (arm_init_machine_status): Initialize
4604 machine->static_chain_stack_bytes.
4606 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4609 * doc/extend.texi: Add four new prototypes for vec_ld.
4610 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4611 definitions for more logical presentation.
4612 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4613 entries for V1TI variants of __builtin_altivec_ld builtin.
4614 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4615 handling of V1TI variant of LVX icode pattern.
4616 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4617 (rs6000_gimple_fold_builtin): Likewise.
4618 (altivec_init_builtins): Add code to define
4619 __builtin_altivec_lvx_v1ti function.
4621 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4624 * function.c (match_asm_constraints_1): Don't do the optimization
4625 if input isn't a REG, SUBREG, MEM or constant.
4627 2018-03-22 Tom de Vries <tom@codesourcery.com>
4629 PR tree-optimization/84956
4630 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4631 bb_has_abnormal_pred.
4633 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4636 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4637 DECL_INITIAL (decl) to decl at the end.
4638 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4641 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4643 * doc/extend.texi (__builtin_tgmath): Document when complex
4644 integer types are treated as _Complex _Float64.
4646 2018-03-21 Tom de Vries <tom@codesourcery.com>
4648 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4650 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4652 PR tree-optimization/84960
4653 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4654 if it is ENTRY block, move them into single succ of ENTRY in that case.
4656 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4658 PR tree-optimization/84811
4659 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4660 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4661 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4662 (known_subrange_p): Update accordingly. Cast each value involved
4663 in the size comparison, rather than casting the result of the
4666 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4668 PR tree-optimization/84982
4669 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4670 by flipping the least significant bit rather than all bits from
4671 bitpos to bitpos + bitsize - 1.
4673 2018-03-21 Nathan Sidwell <nathan@acm.org>
4675 * doc/extend.texi (Deprecated Features): Remove mention of
4676 long-deleted deprecations.
4678 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4681 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4682 * configure: Regenerate.
4684 2018-03-21 Tom de Vries <tom@codesourcery.com>
4686 PR tree-optimization/83126
4687 * tree-parloops.c (num_phis): New function.
4688 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4690 2018-03-21 Nathan Sidwell <nathan@acm.org>
4692 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4693 mention anon-struct/union members and trailing attributes.
4695 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4697 PR tree-optimization/84969
4698 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4699 builtin memset partitions if they set different rhs values.
4701 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4703 PR rtl-optimization/84989
4704 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4705 VEC_DUPLICATE with scalar result mode.
4707 2018-03-21 Martin Liska <mliska@suse.cz>
4710 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4711 not intended return statement.
4713 2018-03-21 Martin Liska <mliska@suse.cz>
4716 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4717 (chkp_find_bound_slots_1): Limit number of iterations.
4719 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4722 * Minor grammar fixes for x86 options.
4724 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4727 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4728 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4730 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4733 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4734 (altivec_lvx_<mode>_1op): Likewise.
4735 (altivec_stvx_<mode>_2op): Likewise.
4736 (altivec_stvx_<mode>_1op): Likewise.
4737 (altivec_lvx_<VM2:mode>): New define_expand.
4738 (altivec_stvx_<VM2:mode>): Likewise.
4739 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4740 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4741 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4742 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4743 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4744 (rs6000_gen_lvx): Likewise.
4745 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4746 (altivec_expand_stv_builtin): Likewise.
4747 (altivec_expand_builtin): Likewise.
4748 * config/rs6000/vector.md: Likewise.
4750 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4753 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4756 2018-03-20 Richard Biener <rguenther@suse.de>
4759 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4760 sign-conversions as zero, fall back to standard scalar_stmt
4763 2018-03-20 Martin Liska <mliska@suse.cz>
4766 * predict.c (rebuild_frequencies): Handle case when we have
4767 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4769 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4772 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4773 flag_section_anchors.
4774 * varasm.c (use_blocks_for_decl_p): Remove hack for
4775 dw2_force_const_mem.
4778 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4780 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4781 be created, use lowpart_subreg of operands[0] rather than operands[0]
4783 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4784 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4785 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4786 and n constraint instead of aarch64_shift_imm_di and Usd.
4787 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4788 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4790 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4793 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4794 to favor GPR over NEON registers.
4795 (<shift>di3_neon): Likewise.
4797 2018-03-20 Tom de Vries <tom@codesourcery.com>
4800 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4801 (nvptx_process_pars): Emit bar.sync asap and alap.
4803 2018-03-20 Tom de Vries <tom@codesourcery.com>
4806 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4807 seen_label if seen_label is already set.
4809 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4812 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4813 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4814 Use 1U instead of 1. Formatting fixes.
4817 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4818 instead of TREE_TYPE (s1) for the return value.
4820 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4822 PR tree-optimization/84946
4823 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4824 bitsize + bitsize in poly_uint64 rather than poly_int64.
4827 * dwarf2asm.c: Include fold-const.c.
4828 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4829 of decl rather than decl itself.
4831 PR rtl-optimization/84643
4832 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4834 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4837 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4838 calling assemble_variable.
4840 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4843 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4844 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4846 2018-03-19 Jim Wilson <jimw@sifive.com>
4849 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4850 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4851 (riscv_first_stack_step): Likewise.
4852 (riscv_option_override): Use STACK_BOUNDARY instead of
4854 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4856 (BIGGEST_ALIGNMENT): Set to 128.
4857 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4858 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4861 2018-03-19 Richard Biener <rguenther@suse.de>
4863 PR tree-optimization/84933
4864 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4865 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4867 2018-03-19 Richard Biener <rguenther@suse.de>
4869 PR tree-optimization/84859
4870 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4871 (cond_if_else_store_replacement): Perform sinking operation on
4872 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4873 Generalize what a BB with a single eligible store is.
4875 2018-03-19 Richard Biener <rguenther@suse.de>
4877 PR tree-optimization/84929
4878 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4879 chrec_is_positive against non-chrec arg.
4881 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4884 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4886 2018-03-18 Martin Liska <mliska@suse.cz>
4888 PR rtl-optimization/84635
4889 * regrename.c (build_def_use): Use matches_mode only when
4892 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4894 PR tree-optimization/84913
4895 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4896 vectorize chains of COND_EXPRs.
4898 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4900 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4902 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4904 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4906 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4908 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4910 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4911 Kito Cheng <kito.cheng@gmail.com>
4913 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4914 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4915 (nds32_adjust_reg_alloc_order): New function.
4916 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4918 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4920 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4921 nds32_print_operand, nds32_print_operand_address): Use
4922 HOST_WIDE_INT_PRINT_DEC instead.
4924 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4926 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4928 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4931 * config/i386/i386.c (initial_ix86_tune_features,
4932 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4934 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4935 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4936 rather than 1u << ix86_tune. Formatting fix.
4937 (ix86_option_override_internal): Change ix86_arch_mask from
4938 unsigned int to unsigned HOST_WIDE_INT, initialize to
4939 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4940 (ix86_function_specific_restore): Likewise.
4942 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4945 * postreload.c (reload_combine_recognize_pattern): Perform
4946 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4947 truncate_int_for_mode the result for the destination's mode.
4950 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4951 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4953 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4956 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4959 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4960 regno_allocno_class_array and sorted_pseudos.
4961 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4962 insns where regno is used.
4964 2018-03-16 Martin Liska <mliska@suse.cz>
4967 * multiple_target.c (create_dispatcher_calls): Redirect
4968 reference in the symbol table.
4970 2018-03-16 Martin Liska <mliska@suse.cz>
4973 * multiple_target.c (create_dispatcher_calls): Redirect also
4976 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4980 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4981 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4982 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4984 2018-03-16 Julia Koval <julia.koval@intel.com>
4986 * doc/invoke.texi (Skylake Server): Add CLWB.
4987 Cannonlake): Remove CLWB.
4989 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4991 PR tree-optimization/84841
4992 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4994 (FLOAT_ONE_CONST_TYPE): Define.
4995 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4996 (sort_by_operand_rank): Put entries with higher constant_type last
4997 rather than first to match comments.
4999 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5001 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5004 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5007 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5008 check and use address of target always.
5010 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5013 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5014 (indirect_thunk_bnd_needed): Likewise.
5015 (indirect_thunks_used): Likewise.
5016 (indirect_thunks_bnd_used): Likewise.
5017 (indirect_return_needed): New.
5018 (indirect_return_bnd_needed): Likewise.
5019 (output_indirect_thunk_function): Add a bool argument for
5021 (output_indirect_thunk_function): Don't generate alias for
5022 function return thunk.
5023 (ix86_code_end): Call output_indirect_thunk_function to generate
5024 function return thunks.
5025 (ix86_output_function_return): Set indirect_return_bnd_needed
5026 and indirect_return_needed instead of indirect_thunk_bnd_needed
5027 and indirect_thunk_needed.
5029 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5031 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5032 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5033 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5035 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5036 Paul Hua <paul.hua.gm@gmail.com>
5039 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5041 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5043 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5044 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5047 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5050 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5051 instead of GET_MODE_SIZE when comparing Units.
5053 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5056 * varasm.c (hash_section): Return an unchangeble hash value
5057 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5058 Return !aarch64_can_use_per_function_literal_pools_p ().
5060 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5063 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5064 rather than address of cmode as last argument to prepare_cmp_insn.
5066 2018-03-15 Julia Koval <julia.koval@intel.com>
5068 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5069 F_AVX512VNNI, F_AVX512BITALG): New.
5071 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5074 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5075 insn for floating-point loads and stores.
5077 2018-03-14 Carl Love <cel@us.ibm.com>
5079 * config/rs6000/rs6000-c.c: Add macro definitions for
5080 ALTIVEC_BUILTIN_VEC_PERMXOR.
5081 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5082 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5083 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5084 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5086 * config/doc/extend.texi: Add prototypes for vec_permxor.
5088 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5091 * diagnostic-show-locus.c (class layout_point): Convert m_line
5092 from int to linenum_type.
5093 (line_span::comparator): Use linenum "compare" function when
5094 comparing line numbers.
5095 (test_line_span): New function.
5096 (layout_range::contains_point): Convert param "row" from int to
5098 (layout_range::intersects_line_p): Likewise.
5099 (layout::will_show_line_p): Likewise.
5100 (layout::print_source_line): Likewise.
5101 (layout::should_print_annotation_line_p): Likewise.
5102 (layout::print_annotation_line): Likewise.
5103 (layout::print_leading_fixits): Likewise.
5104 (layout::annotation_line_showed_range_p): Likewise.
5105 (struct line_corrections): Likewise for field m_row.
5106 (line_corrections::line_corrections): Likewise for param "row".
5107 (layout::print_trailing_fixits): Likewise.
5108 (layout::get_state_at_point): Likewise.
5109 (layout::get_x_bound_for_row): Likewise.
5110 (layout::print_line): Likewise.
5111 (diagnostic_show_locus): Likewise for locals "last_line" and
5113 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5114 * input.c (selftest::test_linenum_comparisons): New function.
5115 (selftest::input_c_tests): Call it.
5116 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5117 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5118 * selftest.h (ASSERT_GT): New macro.
5119 (ASSERT_GT_AT): New macro.
5120 (ASSERT_LT): New macro.
5121 (ASSERT_LT_AT): New macro.
5123 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5125 PR rtl-optimization/84780
5126 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5128 2018-03-14 Martin Liska <mliska@suse.cz>
5130 * tree.c (record_node_allocation_statistics): Use
5131 get_stats_node_kind.
5132 (get_stats_node_kind): New function extracted from
5133 record_node_allocation_statistics.
5134 (free_node): Use get_stats_node_kind.
5136 2018-03-14 Richard Biener <rguenther@suse.de>
5138 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5139 that the value-set of ANTIC_IN doesn't grow.
5142 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5144 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5145 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5147 2018-03-14 Julia Koval <julia.koval@intel.com>
5149 * config.gcc (icelake-client, icelake-server): New.
5151 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5152 (initial_ix86_arch_features): Ditto.
5153 (PTA_SKYLAKE): Add SGX.
5154 (PTA_ICELAKE): Remove.
5155 (PTA_ICELAKE_CLIENT): New.
5156 (PTA_ICELAKE_SERVER): New.
5157 (ix86_option_override_internal): Split up icelake on icelake client and
5159 (get_builtin_code_for_version): Ditto.
5160 (fold_builtin_cpu): Ditto.
5161 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5162 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5163 * config/i386/i386.h (processor_type): Ditto.
5164 * doc/invoke.texi: Ditto.
5166 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5169 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5170 INTEGER_CST offset, add it together with bitpos / 8 and
5171 sign extend based on POINTER_SIZE.
5175 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5178 * config/i386/constraints.md (Yc): New register constraint.
5179 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5180 Use Yc constraint for alternative 2 of operand 0. Remove
5181 preferred_for_speed attribute.
5183 2018-03-14 Richard Biener <rguenther@suse.de>
5185 PR tree-optimization/84830
5186 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5187 with the old one to avoid oscillations.
5189 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5192 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5194 (assign_by_spills): Return a flag of reload assignment failure.
5195 Do not process the reload assignment failures. Do not spill other
5196 reload pseudos if they has the same reg class. Update n if
5198 (lra_assign): Add a return arg. Set up from the result of
5199 assign_by_spills call.
5200 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5201 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5202 usage_insns if it is not NULL.
5203 (spill_hard_reg_in_range): New function.
5204 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5205 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5206 function prototypes.
5207 (lra_assign): Change prototype.
5208 * lra.c (lra): Add code to deal with fails by splitting hard reg
5211 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5213 * config/riscv/riscv.opt (mrelax): New option.
5214 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5215 "norelax" when riscv_mrelax is disabled.
5216 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5218 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5221 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5222 reassociation for int modes.
5224 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5226 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5227 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5229 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5230 * config/aarch64/aarch64-sve.md
5231 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5232 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5233 (*extend<mode><Vwide>2): Rename to...
5234 (aarch64_sve_extend<mode><Vwide>2): ...this.
5235 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5236 renaming the old pattern to...
5237 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5239 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5240 define_expand, renaming the old pattern to...
5241 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5242 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5243 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5244 account when deciding which SVE instruction the optab should use.
5245 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5247 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5249 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5250 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5251 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5252 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5253 (tlsdesc_small_<mode>): Turn a define_expand and use
5254 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5255 (tlsdesc_small_advsimd_<mode>): ...this.
5256 (tlsdesc_small_sve_<mode>): New pattern.
5258 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5260 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5261 (UNSPEC_UMUL_HIGHPART): New constants.
5262 (MUL_HIGHPART): New int iteraor.
5263 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5264 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5266 (*<su>mul<mode>3_highpart): New define_insn.
5268 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5271 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5274 2018-03-13 Martin Liska <mliska@suse.cz>
5277 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5279 (sem_item_optimizer::~sem_item_optimizer): Release it.
5280 (sem_item_optimizer::merge_classes): Register variable aliases.
5281 (sem_item_optimizer::fixup_pt_set): New function.
5282 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5283 * ipa-icf.h: Declare new variables and functions.
5285 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5288 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5289 integer_pow2p@2 and test integer_pow2p in condition.
5290 (A < 0 ? C : 0): Similarly for @1.
5293 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5294 characters starting at p contain '\0' character, don't look beyond
5298 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5299 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5302 * reg-stack.c (change_stack): Change update_end var from int to
5303 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5304 also call set_block_for_insn on the newly added insns and rescan.
5307 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5308 on the last operand.
5311 * tree.c (stabilize_reference_1): Return save_expr (e) for
5312 STATEMENT_LIST even if it doesn't have side-effects.
5314 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5316 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5318 2018-03-12 Renlin Li <renlin.li@arm.com>
5320 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5321 aarch64_output_scalar_simd_mov_immediate.
5323 2018-03-12 Martin Sebor <msebor@redhat.com>
5325 PR tree-optimization/83456
5326 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5327 for perfectly overlapping calls to memcpy.
5328 (gimple_fold_builtin_memory_chk): Same.
5329 (gimple_fold_builtin_strcpy): Handle no-warning.
5330 (gimple_fold_builtin_stxcpy_chk): Same.
5331 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5333 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5335 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5336 parameter. Use it for SFmode.
5337 (rs6000_function_arg_advance_1): Adjust.
5338 (rs6000_function_arg): Adjust.
5339 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5341 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5343 PR rtl-optimization/84169
5344 PR rtl-optimization/84780
5345 * combine.c (can_combine_p): Check for a 2-insn combination whether
5346 the destination register is used between the two insns, too.
5348 2018-03-12 Richard Biener <rguenther@suse.de>
5350 PR tree-optimization/84803
5351 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5352 for refs DR analysis didn't process.
5354 2018-03-12 Richard Biener <rguenther@suse.de>
5356 PR tree-optimization/84777
5357 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5358 force-vectorize loops ignore whether we are optimizing for size.
5360 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5362 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5363 (TARGET_MD_ASM_ADJUST): Define.
5365 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5366 Kito Cheng <kito.cheng@gmail.com>
5367 Chung-Ju Wu <jasonwucj@gmail.com>
5369 * config/nds32/nds32.c (nds32_compute_stack_frame,
5370 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5371 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5372 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5373 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5374 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5375 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5376 * config/nds32/nds32.md (prologue, epilogue): Use macro
5377 NDS32_V3PUSH_AVAILABLE_P to do checking.
5379 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5382 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5383 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5384 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5385 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5386 addition of most attributes on !orig_type_die or the attribute not
5387 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5389 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5390 Chung-Ju Wu <jasonwucj@gmail.com>
5392 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5394 * config/nds32/nds32.opt (mvh): New option.
5396 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5397 Chung-Ju Wu <jasonwucj@gmail.com>
5399 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5401 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5402 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5405 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5406 Chung-Ju Wu <jasonwucj@gmail.com>
5408 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5410 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5411 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5413 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5414 Kito Cheng <kito.cheng@gmail.com>
5415 Chung-Ju Wu <jasonwucj@gmail.com>
5417 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5418 UNSPEC_FFMISM and UNSPEC_FLMISM.
5419 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5420 for ffb, ffmism and flmism.
5421 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5422 (unspec_ffmism): Ditto.
5423 (unspec_flmism): Ditto.
5424 (nds32_expand_builtin_impl): Check if string extension is available.
5425 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5426 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5428 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5431 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5434 * lra-assigns.c (assign_by_spills): Return a flag of reload
5435 assignment failure. Do not process the reload assignment
5436 failures. Do not spill other reload pseudos if they has the same
5438 (lra_assign): Add a return arg. Set up from the result of
5439 assign_by_spills call.
5440 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5441 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5442 usage_insns if it is not NULL.
5443 (spill_hard_reg_in_range): New function.
5444 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5445 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5446 function prototypes.
5447 (lra_assign): Change prototype.
5448 * lra.c (lra): Add code to deal with fails by splitting hard reg
5451 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5454 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5456 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5459 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5460 (dw_val_node): Add val_symbolic_view.
5461 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5462 (symview_upper_bound): New.
5463 (new_line_info_table): Initialize symviews_since_reset.
5464 (dwarf2out_source_line): Count symviews_since_reset and set
5465 symview_upper_bound.
5466 (dw_val_equal_p): Handle symview.
5467 (add_AT_symview): New.
5468 (print_dw_val): Handle symview.
5469 (attr_checksum, attr_checksum_ordered): Likewise.
5470 (same_dw_val_p, size_of_die): Likewise.
5471 (value_format, output_die): Likewise.
5472 (add_high_low_attributes): Use add_AT_symview for entry_view.
5473 (dwarf2out_finish): Reset symview_upper_bound, clear
5476 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5479 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5480 Add strict argument and use it.
5481 (rs6000_split_multireg_move): Update for new strict argument.
5482 (mem_operand_gpr): Disallow all non-offsettable addresses.
5483 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5485 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5488 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5489 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5490 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5493 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5494 decl, use remap_type if we want to use the type.
5496 2018-03-09 Martin Sebor <msebor@redhat.com>
5498 PR tree-optimization/84526
5499 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5501 (builtin_access::generic_overlap): Be prepared to handle non-array
5504 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5506 PR rtl-optimization/84682
5507 * lra-constraints.c (process_address_1): Check is_address flag
5508 for address constraints.
5509 (process_alt_operands): Likewise.
5510 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5511 preprocess_constraints.
5512 * recog.h (preprocess_constraints): Add oploc parameter.
5514 * recog.c (preprocess_constraints): Test address_operand for
5515 CT_ADDRESS constraints.
5517 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5520 * lra-assigns.c (assign_by_spills): Return a flag of reload
5521 assignment failure. Do not process the reload assignment
5522 failures. Do not spill other reload pseudos if they has the same
5524 (lra_assign): Add a return arg. Set up from the result of
5525 assign_by_spills call.
5526 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5527 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5528 usage_insns if it is not NULL.
5529 (spill_hard_reg_in_range): New function.
5530 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5531 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5532 function prototypes.
5533 (lra_assign): Change prototype.
5534 * lra.c (lra): Add code to deal with fails by splitting hard reg
5537 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5540 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5541 Accept complain bool parameter. Only emit errors if it is true.
5542 (arm_parse_cpu_option_name): Likewise.
5543 (arm_target_thumb_only): Adjust callers of the above.
5544 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5545 prototype to take a default true bool parameter.
5546 (arm_parse_arch_option_name): Likewise.
5548 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5549 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5553 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5554 * configure: Regenerate.
5555 * configure.ac ("linker --version-script option"): New.
5556 ("linker soname option"): New.
5558 2018-03-09 Richard Biener <rguenther@suse.de>
5560 PR tree-optimization/84775
5561 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5562 immediate uses of predicate stmts and mark them modified.
5565 PR tree-optimization/84178
5566 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5568 (version_loop_for_if_conversion): Delay update_ssa call.
5569 (tree_if_conversion): Delay update_ssa until after predicate
5572 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5575 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5576 when the function accesses prior frames.
5578 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5581 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5582 gen_llsym, otherwise call maybe_gen_llsym.
5585 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5586 has ',' character inside of it.
5588 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5591 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5592 as clobbering CC_REGNUM.
5594 2018-03-08 Richard Biener <rguenther@suse.de>
5597 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5598 (follow_copies_to_constant): Do not follow SSA names registered
5601 2018-03-08 Richard Biener <rguenther@suse.de>
5603 PR tree-optimization/84178
5604 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5606 (version_loop_for_if_conversion): Delay update_ssa call.
5607 (tree_if_conversion): Delay update_ssa until after predicate
5610 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5612 PR tree-optimization/84178
5613 * tree-if-conv.c (release_bb_predicate): Remove the
5614 the assertion that the stmts have NULL use_ops.
5615 Discard the statements, asserting that they haven't
5616 yet been added to a BB.
5618 2018-03-08 Richard Biener <rguenther@suse.de>
5620 PR tree-optimization/84746
5621 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5622 (phi_translate): Pass in destination ANTIC_OUT set.
5623 (phi_translate_1): Likewise. For a simplified result lookup
5624 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5625 (phi_translate_set): Adjust.
5626 (do_pre_regular_insertion): Likewise.
5627 (do_pre_partial_partial_insertion): Likewise.
5629 2018-03-08 Martin Liska <mliska@suse.cz>
5631 PR gcov-profile/84735
5632 * doc/gcov.texi: Document usage of profile files.
5633 * gcov-io.h: Document changes in the format.
5635 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5639 * dwarf2out.c (struct dw_line_info_table): Update comments for
5641 (FORCE_RESET_NEXT_VIEW): New.
5642 (FORCE_RESETTING_VIEW_P): New.
5643 (RESETTING_VIEW_P): Check for -1 too.
5644 (ZERO_VIEW_P): Likewise.
5645 (new_line_info_table): Force-reset next view.
5646 (dwarf2out_begin_function): Likewise.
5647 (dwarf2out_source_line): Simplify zero_view_p initialization.
5648 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5649 view directly. Omit view when omitting .loc at line 0.
5651 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5653 PR tree-optimization/84740
5654 * tree-switch-conversion.c (process_switch): Call build_constructors
5655 only if info.phi_count is non-zero.
5657 PR tree-optimization/84739
5658 * tree-tailcall.c (find_tail_calls): Check call arguments against
5659 DECL_ARGUMENTS (current_function_decl) rather than
5660 DECL_ARGUMENTS (func) when checking for tail recursion.
5662 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5664 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5665 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5666 Volker Reichelt's entry and add entries for people that perform
5667 GCC fuzzy testing and report numerous bugs.
5669 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5673 readonly data in sdata, if that is disabled.
5674 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5675 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5676 -mreadonly-in-sdata option.
5678 2018-03-07 Martin Sebor <msebor@redhat.com>
5680 PR tree-optimization/84468
5681 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5682 basic block when looking for nul assignment.
5684 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5687 * except.h (output_function_exception_table): Adjust prototype.
5688 * except.c (output_function_exception_table): Remove FNNAME parameter
5689 and add SECTION parameter. Ouput one part of the table at a time.
5690 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5691 the first part of the exception table and emit unwind directives.
5692 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5693 (i386_pe_seh_cold_init): Likewise.
5694 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5695 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5696 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5697 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5698 * config/i386/winnt.c: Include except.h.
5699 (struct seh_frame_state): Add reg_offset, after_prologue and
5700 in_cold_section fields.
5701 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5702 (i386_pe_seh_cold_init): New function.
5703 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5704 to seh->in_cold_section.
5705 (seh_emit_push): Record the offset of the push.
5706 (seh_emit_save): Record the offet of the save.
5707 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5708 Test seh->after_prologue to disregard the epilogue.
5709 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5710 (i386_pe_end_cold_function): New function.
5712 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5715 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5716 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5719 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5721 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5722 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5725 * multiple_target.c: Include tree-inline.h and intl.h.
5726 (expand_target_clones): Diagnose and fail if node->definition and
5727 !tree_versionable_function_p (node->decl).
5729 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5731 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5733 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5734 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5735 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5737 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5740 * combine.c (try_combine): Use reg_or_subregno instead of handling
5741 just paradoxical SUBREGs and REGs.
5743 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5745 * config/arc/arc.c (arc_finalize_pic): Remove function.
5746 (arc_must_save_register): We use single base PIC register, remove
5747 checks to save/restore the PIC register.
5748 (arc_expand_prologue): Likewise.
5749 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5751 (arc_verify_short): Likewise.
5752 (arc_attr_type): Likewise.
5753 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5754 (walk_stores): Likewise.
5755 (arc_address_cost): Make it static.
5756 (arc_verify_short): Likewise.
5757 (branch_dest): Likewise.
5758 (arc_attr_type): Likewise.
5759 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5760 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5761 (arc_final_prescan_insn): Remove inserting the nops due to
5762 hardware hazards. It is done in reorg step.
5763 (insn_length_variant_t): Remove.
5764 (insn_length_parameters_t): Likewise.
5765 (arc_insn_length_parameters): Likewise.
5766 (arc_get_insn_variants): Likewise.
5767 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5769 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5772 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5775 PR tree-optimization/84687
5776 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5778 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5780 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5782 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5783 Rename to ppc_speculation_barrier.
5784 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5785 __builtin_ppc_speculation_barrier.
5787 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5790 * combine.c (combine_simplify_rtx): Don't try to simplify if
5791 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5794 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5796 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5797 to 32 bytes when compiling for POWER9.
5799 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5802 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5803 regparm >= 3 with no arg reg available also for calls with
5804 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5807 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5809 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5811 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5814 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5816 2018-03-05 Richard Biener <rguenther@suse.de>
5818 PR tree-optimization/84486
5819 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5820 When inserting a __builtin_assume_aligned call set the LHS
5821 SSA name alignment info accordingly.
5823 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5825 PR tree-optimization/84114
5826 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5827 Avoid reassociation of FLOAT_MODE addition.
5829 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5831 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5832 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5833 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5834 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5835 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5836 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5837 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5839 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5840 __builtin_ia32_wbinvd): New builtins.
5841 (SPECIAL_ARGS2): New.
5842 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5843 (SPECIAL_ARGS2): New.
5844 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5845 (ix86_valid_target_attribute_inner_p): Ditto.
5846 (ix86_init_mmx_sse_builtins): Add special_args2.
5847 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5848 TARGET_WBNOINVD_P): New.
5849 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5850 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5851 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5852 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5853 * config/i386/pconfigintrin.h: New file.
5854 * config/i386/wbnoinvdintrin.h: Ditto.
5855 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5856 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5858 2018-03-05 Richard Biener <rguenther@suse.de>
5860 PR tree-optimization/84670
5861 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5863 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5864 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5865 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5866 doesn't grow if all successors (recursively) were visited at least
5869 2018-03-05 Richard Biener <rguenther@suse.de>
5871 PR tree-optimization/84650
5872 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5873 if executed in the loop pipeline.
5875 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5877 * doc/configfiles.texi (Configuration Files): Move info about
5878 conditionalizing $target-protos.h to...
5879 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5880 differs from $target-protos.h.
5882 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5883 Chung-Ju Wu <jasonwucj@gmail.com>
5885 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5886 * config/nds32/nds32-multiple.md (setmemsi): Define.
5887 * config/nds32/nds32-memory-manipulation.c
5888 (nds32_gen_dup_4_byte_to_word_value): New.
5889 (emit_setmem_word_loop): New.
5890 (emit_setmem_byte_loop): New.
5891 (nds32_expand_setmem_loop): New.
5892 (nds32_expand_setmem_loop_v3m): New.
5893 (nds32_expand_setmem_unroll): New.
5894 (nds32_expand_setmem): New.
5896 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5897 Chung-Ju Wu <jasonwucj@gmail.com>
5899 * config/nds32/nds32-memory-manipulation.c
5900 (nds32_emit_load_store): New.
5901 (nds32_emit_post_inc_load_store): New.
5902 (nds32_emit_mem_move): New.
5903 (nds32_emit_mem_move_block): New.
5904 (nds32_expand_movmemsi_loop_unknown_size): New.
5905 (nds32_expand_movmemsi_loop_known_size): New.
5906 (nds32_expand_movmemsi_loop): New.
5907 (nds32_expand_movmemsi_unroll): New.
5908 (nds32_expand_movmemqi): Rename ...
5909 (nds32_expand_movmemsi): ... to this.
5910 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5911 (movmemsi): ... to this.
5912 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5913 (nds32_expand_movmemsi): ... to this.
5915 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5916 Monk Chiang <sh.chiang04@gmail.com>
5917 Chung-Ju Wu <jasonwucj@gmail.com>
5919 * config/nds32/nds32-protos.h
5920 (nds32_expand_load_multiple): New arguments.
5921 (nds32_expand_store_multiple): Ditto.
5922 (nds32_valid_multiple_load_store): Rename ...
5923 (nds32_valid_multiple_load_store_p): ... to this.
5924 * config/nds32/nds32-memory-manipulation.c
5925 (nds32_expand_load_multiple): Refine implementation.
5926 (nds32_expand_store_multiple): Ditto.
5927 * config/nds32/nds32-multiple.md
5928 (load_multiple): Update nds32_expand_load_multiple interface.
5929 (store_multiple): Update nds32_expand_store_multiple interface.
5930 * config/nds32/nds32-predicates.c
5931 (nds32_valid_multiple_load_store): Rename ...
5932 (nds32_valid_multiple_load_store_p): ... to this and refine
5934 * config/nds32/predicates.md
5935 (nds32_load_multiple_and_update_address_operation): New predicate.
5936 (nds32_store_multiple_and_update_address_operation): New predicate.
5938 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5939 Chung-Ju Wu <jasonwucj@gmail.com>
5941 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5942 (combo): New attribute.
5943 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5945 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5947 * config/nds32/nds32.opt: Change -mcmodel= default value.
5949 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5950 Monk Chiang <sh.chiang04@gmail.com>
5951 Chung-Ju Wu <jasonwucj@gmail.com>
5953 * config/nds32/constants.md (unspec_element): New enum.
5954 * config/nds32/constraints.md (Umw): New constraint.
5955 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5956 * config/nds32/nds32-intrinsic.md: Likewise.
5957 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5958 (nds32_valid_smw_lwm_base_p): New.
5959 (nds32_output_smw_single_word): New.
5960 (nds32_output_lmw_single_word): New.
5961 (nds32_expand_unaligned_load): New.
5962 (nds32_expand_unaligned_store): New.
5963 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5964 (nds32_output_smw_single_word): Declare.
5965 (nds32_output_lmw_single_word): Declare.
5966 (nds32_expand_unaligned_load): Declare.
5967 (nds32_expand_unaligned_store): Declare.
5968 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5969 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5970 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5971 NDS32_BUILTIN_UASTORE_DW.
5972 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5975 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5976 Kito Cheng <kito.cheng@gmail.com>
5977 Chung-Ju Wu <jasonwucj@gmail.com>
5979 * config/nds32/nds32-intrinsic.c
5980 (nds32_expand_builtin_null_ftype_reg): Delete.
5981 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5982 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5983 (nds32_read_argument): New.
5984 (nds32_legitimize_target): Ditto.
5985 (nds32_legitimize_argument): Ditto.
5986 (nds32_check_constant_argument): Ditto.
5987 (nds32_expand_unop_builtin): Ditto.
5988 (nds32_expand_unopimm_builtin): Ditto.
5989 (nds32_expand_binop_builtin): Ditto.
5990 (nds32_builtin_decl_impl): Ditto.
5991 (builtin_description): Ditto.
5992 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5993 (nds32_init_builtins_impl): Ditto.
5994 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5995 (nds32_builtin_decl): New.
5996 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5997 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5999 2018-03-02 Jeff Law <law@redhat.com>
6001 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6002 (redundant_insn, fill_simple_delay_slots): Likewise.
6003 (fill_slots_from_thread): Likewise.
6004 * resource.c (mark_referenced_resources): Likewise.
6005 (mark_set_resources, find_dead_or_set_registers): Likewise.
6007 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6009 * substring-locations.h (format_warning_va): Formatting fix for
6011 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6013 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6014 * substring-locations.c: Include intl.h.
6015 (format_warning_va): Turned into small wrapper around
6016 format_warning_n_va, renamed to ...
6017 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6018 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6020 (format_warning_at_substring_n): New function.
6021 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6022 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6023 format_warning_at_substring with just a shorter name instead of
6024 const function pointer.
6025 (fmtwarn_n): New function.
6026 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6027 appropriate, get rid of all the fmtstr temporaries, move conditionals
6028 with G_() wrapped string literals directly into fmtwarn arguments,
6029 cast dir.len to (int), formatting fixes.
6031 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6033 * doc/invoke.texi: Remove "Cilk Plus" references.
6035 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6036 Richard Biener <rguenther@suse.de>
6039 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6040 for error or warning attributes if CALL_FROM_THUNK_P is set.
6043 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6046 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6047 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6050 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6051 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6053 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6054 instead of -1U in last predictors element's probability member.
6056 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6059 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6060 arguments if they are comparable.
6062 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6064 PR tree-optimization/84634
6065 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6066 masks and masked_loop_p with a single loop_masks, making sure it's
6067 null for bb vectorization.
6069 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6071 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6072 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6073 loop->force_vectorize to check whether there is no alias.
6075 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6078 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6080 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6082 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6083 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6084 instead of a loop around prev_real_insn.
6085 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6089 * config/i386/i386.c (ix86_print_operand): Use conditional
6090 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6093 2018-03-02 Richard Biener <rguenther@suse.de>
6095 PR tree-optimization/84427
6096 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6097 (bitmap_set_subtract_values): Rewrite to handle multiple
6100 (prune_clobbered_mems): Likewise.
6101 (phi_translate): Take edge instead of pred/phiblock.
6102 (phi_translate_1): Likewise.
6103 (phi_translate_set): Likewise. Insert all translated
6104 exprs for a value into the set, keeping possibly multiple
6105 expressions per value.
6106 (compute_antic_aux): Adjust for phi_translate changes.
6107 When intersecting union the expressions and prune those
6108 not in the final value set, keeping possibly multiple
6109 expressions per value. Do not use value-insertion
6110 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6111 all expressions. Add verification that the value-sets
6112 only shrink during iteration.
6113 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6114 (do_pre_regular_insertion): Likewise.
6115 (do_pre_partial_partial_insertion): Likewise.
6117 2018-03-02 Richard Biener <rguenther@suse.de>
6120 * config/darwin.c (saved_debug_info_level): New static global.
6121 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6122 (darwin_asm_lto_end): Restore debug info generation settings.
6124 2018-03-01 Martin Liska <mliska@suse.cz>
6127 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6130 2018-03-01 Richard Biener <rguenther@suse.de>
6133 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6134 type annotation with LTO when debug was disabled at compile-time.
6136 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6138 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6140 (mips_final_postscan_insn): Likewise.
6142 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6144 PR rtl-optimization/84528
6145 * alias.c (init_alias_target): Add commentary.
6146 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6147 a unique base value if the frame pointer is not eliminated
6148 to the stack pointer.
6150 2018-03-01 Tom de Vries <tom@codesourcery.com>
6152 PR rtl-optimization/83327
6153 * lra-int.h (hard_regs_spilled_into): Declare.
6154 * lra.c (hard_regs_spilled_into): Define.
6155 (init_reg_info): Init hard_regs_spilled_into.
6156 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6157 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6158 (process_bb_lives): Handle hard_regs_spilled_into.
6159 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6160 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6162 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6164 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6165 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6166 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6167 * config/rs6000/aix72.h: New file.
6169 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6171 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6172 instead of warning_at with conditional singular and plural messages
6176 * stor-layout.c (update_alignment_for_field): For
6177 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6178 && !DECL_PACKED (field), do the alignment update, just use
6179 only desired_align instead of MAX (type_align, desired_align)
6181 (place_field): Don't do known_align < desired_align handling
6182 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6183 is non-NULL, instead do it after rli->prev_field handling and
6184 only if not within a bitfield word. For DECL_PACKED (field)
6185 use type_align of BITS_PER_UNIT.
6187 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6189 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6190 superfluous parentheses and trailing spaces.
6192 2018-02-28 Richard Biener <rguenther@suse.de>
6194 PR tree-optimization/84584
6195 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6196 SCoPs with fake exit edge.
6198 2018-02-28 Martin Liska <mliska@suse.cz>
6201 * timevar.c (timer::print): Fix format to properly print 100%
6204 2018-02-28 Richard Biener <rguenther@suse.de>
6207 * genmatch.c (capture_info::walk_match): Do not mark
6208 captured expressions without operands as expr_p given
6209 they act more like predicates and should be subject to
6210 "lost tail" side-effect preserving.
6212 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6214 PR rtl-optimization/81611
6215 * auto-inc-dec.c (attempt_change): Move dead note from
6216 mem_insn if it's the next use of regno
6217 (find_address): Take address use of reg holding
6218 non-incremented value. Add parm to limit search to the named
6220 (merge_in_block): Attempt to use a mem insn that is the next
6221 use of the original regno.
6223 2018-02-27 Martin Sebor <msebor@redhat.com>
6226 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6227 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6229 2018-02-27 Martin Sebor <msebor@redhat.com>
6231 PR translation/84207
6232 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6233 n argument to unsigned HOST_WIDE_INT.
6234 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6235 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6236 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6239 2018-02-27 Richard Biener <rguenther@suse.de>
6241 PR tree-optimization/84512
6242 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6243 Do not use the estimate returned from record_stmt_cost for
6244 the scalar iteration cost but sum properly using add_stmt_cost.
6246 2018-02-27 Richard Biener <rguenther@suse.de>
6248 PR tree-optimization/84466
6249 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6250 Adjust last change to less strictly validate use operands.
6252 2018-02-27 Martin Liska <mliska@suse.cz>
6254 PR gcov-profile/84548
6255 * gcov.c (process_file): Allow partial overlap and consider it
6256 also as group functions.
6257 (output_lines): Properly calculate range of lines for a group.
6259 2018-02-27 Martin Liska <mliska@suse.cz>
6261 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6262 'ggc' suffixes. Change first column width.
6263 (timer::print): Fix formatting of the column.
6265 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6267 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6268 preserve inline entry blocks for the sake of debug inline
6269 entry point markers alone.
6270 (remove_unused_locals): Suggest in comments a better place to
6271 force the preservation of inline entry blocks that are
6272 otherwise unused, but do not preserve them.
6274 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6276 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6278 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6281 * config/i386/constraints.md (Bs): Replace
6282 ix86_indirect_branch_register with
6283 TARGET_INDIRECT_BRANCH_REGISTER.
6285 * config/i386/i386.md (indirect_jump): Likewise.
6286 (tablejump): Likewise.
6287 (*sibcall_memory): Likewise.
6288 (*sibcall_value_memory): Likewise.
6289 Peepholes of indirect call and jump via memory: Likewise.
6290 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6291 (*sibcall_value_GOT_32): Likewise.
6292 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6293 (GOT_memory_operand): Likewise.
6294 (call_insn_operand): Likewise.
6295 (sibcall_insn_operand): Likewise.
6296 (GOT32_symbol_operand): Likewise.
6297 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6299 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6301 PR rtl-optimization/83496
6302 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6303 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6304 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6305 redundant insn, if any.
6306 (relax_delay_slots): Likewise.
6307 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6309 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6311 PR tree-optimization/83965
6312 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6313 that grouped statements are part of a reduction chain. Return
6314 true if the statement is not marked as a reduction itself but
6316 (vect_recog_dot_prod_pattern): Don't check whether the statement
6317 is part of a group here.
6318 (vect_recog_sad_pattern): Likewise.
6319 (vect_recog_widen_sum_pattern): Likewise.
6321 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6324 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6326 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6329 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6331 (ix86_output_indirect_function_return): New prototype.
6332 (ix86_split_simple_return_pop_internal): Likewise.
6333 * config/i386/i386.c (indirect_return_via_cx): New.
6334 (indirect_return_via_cx_bnd): Likewise.
6335 (indirect_thunk_name): Handle return va CX_REG.
6336 (output_indirect_thunk_function): Create alias for
6337 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6338 (ix86_output_indirect_jmp): Remove the bool argument.
6339 (ix86_output_indirect_function_return): New function.
6340 (ix86_split_simple_return_pop_internal): Likewise.
6341 * config/i386/i386.md (*indirect_jump): Don't pass false
6342 to ix86_output_indirect_jmp.
6343 (*tablejump_1): Likewise.
6344 (simple_return_pop_internal): Change it to define_insn_and_split.
6345 Call ix86_split_simple_return_pop_internal to split it for
6347 (simple_return_indirect_internal): Call
6348 ix86_output_indirect_function_return instead of
6349 ix86_output_indirect_jmp.
6351 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6354 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6355 memset and value initialization afterwards.
6357 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6359 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6361 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6364 * common/config/aarch64/aarch64-common.c
6365 (aarch_option_optimization_table[]): Switch
6366 off fomit-frame-pointer
6368 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6369 Chung-Ju Wu <jasonwucj@gmail.com>
6371 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6373 (store_multiple): Ditto.
6375 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6377 * config.gcc: Add --with-cpu support for nds32 target.
6378 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6379 * config/nds32/nds32.opt: Add -mcpu= option.
6381 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6383 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6384 isel=yes): Warn for these deprecated options.
6386 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6388 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6389 ISA_2_5_MASKS_EMBEDDED.
6391 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6393 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6394 p->max as pointers rather than using iterative_hash_expr.
6396 2018-02-23 Carl Love <cel@us.ibm.com>
6398 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6399 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6401 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6402 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6403 P8V_BUILTIN_VEC_VUNSIGNED2.
6405 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6408 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6409 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6411 * lra-constraints.c (curr_insn_transform): Set up
6412 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6414 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6415 * lra-eliminations.c (spill_pseudos): Ditto.
6416 (process_insn_for_elimination): Ditto.
6417 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6418 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6419 LRA_NON_CLOBBERED_ALT.
6421 2018-02-22 Martin Sebor <msebor@redhat.com>
6423 PR tree-optimization/84480
6424 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6425 to maybe_diag_stxncpy_trunc. Call it.
6426 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6427 from gimple_fold_builtin_strcpy. Print inlining stack.
6428 (handle_builtin_stxncpy): Print inlining stack.
6429 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6431 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6434 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6435 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6436 and -fcheck-pointer-bounds are used together.
6437 (indirect_thunk_prefix): New enum.
6438 (indirect_thunk_need_prefix): New function.
6439 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6440 "_nt" instead of "_bnd" for NOTRACK prefix.
6441 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6442 (output_indirect_thunk_function): Likewise.
6444 (ix86_code_end): Update output_indirect_thunk_function calls.
6445 (ix86_output_indirect_branch_via_reg): Replace
6446 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6447 (ix86_output_indirect_branch_via_push): Likewise.
6448 (ix86_output_function_return): Likewise.
6449 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6450 incompatible with -fcf-protection=branch and
6451 -fcheck-pointer-bounds.
6453 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6456 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6457 Change gcc_assert call to output_operand_lossage.
6459 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6461 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6463 2018-02-22 DJ Delorie <dj@redhat.com>
6464 Sebastian Perta <sebastian.perta@renesas.com>
6465 Oleg Endo <olegendo@gcc.gnu.org>
6467 * config/rx/rx.c (rx_rtx_costs): New function.
6468 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6470 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6472 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6474 2018-02-22 Martin Liska <mliska@suse.cz>
6477 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6478 Add "native" as a possible value.
6480 2018-02-22 Martin Liska <mliska@suse.cz>
6483 * config/i386/i386.c (ix86_option_override_internal):
6484 Add "native" as a possible value for -march and -mtune.
6486 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6489 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6490 to all type variants.
6492 PR tree-optimization/84503
6493 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6494 width as info->bitpos + info->bitsize - start.
6495 (merged_store_group::merge_overlapping): Simplify width computation.
6496 (check_no_overlap): New function.
6497 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6498 start + width and last_order of the group, fail if check_no_overlap
6500 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6501 to group if check_no_overlap fails.
6503 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6505 * config/rs6000/altivec.md: Delete contraint arguments to
6506 define_expand, define_split, and define_peephole2, and in
6507 define_insn_and_split if always unused.
6508 * config/rs6000/darwin.md: Ditto.
6509 * config/rs6000/dfp.md: Ditto.
6510 * config/rs6000/rs6000.md: Ditto.
6511 * config/rs6000/sync.md: Ditto.
6512 * config/rs6000/vector.md: Ditto.
6513 * config/rs6000/vsx.md: Ditto.
6515 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6517 * config/rs6000/altivec.md: Write output control strings as braced
6518 blocks instead of double-quoted strings.
6519 * config/rs6000/darwin.md: Ditto.
6520 * config/rs6000/rs6000.md: Ditto.
6521 * config/rs6000/vector.md: Ditto.
6522 * config/rs6000/vsx.md: Ditto.
6524 2018-02-21 Jason Merrill <jason@redhat.com>
6526 PR c++/84314 - ICE with templates and fastcall attribute.
6527 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6529 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6531 * ipa-cp.c (determine_versionability): Fix comment typos.
6533 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6536 * ipa-cp.c (determine_versionability): Do not version functions caling
6539 2018-02-21 Martin Liska <mliska@suse.cz>
6542 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6543 Add "native" as a possible value.
6544 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6545 the macro when native cpu detection is available.
6547 2018-02-21 Martin Liska <mliska@suse.cz>
6550 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6551 Add "native" as a possible value.
6552 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6553 when native cpu detection is available.
6555 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6556 Martin Sebor <msebor@redhat.com>
6558 PR tree-optimization/84478
6559 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6561 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6562 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6563 support which is conservatively correct, for 2 only stay conservative
6564 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6565 argument to the 2 argument get_range_strlen, adjust 6 arg
6566 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6568 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6569 (gimple_fold_builtin_strlen): Pass true as last argument to
6572 2018-02-20 Martin Sebor <msebor@redhat.com>
6575 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6576 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6577 (builtin_memref::builtin_memref): Factor out parts into
6578 set_base_and_offset and call it.
6580 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6583 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6584 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6585 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6586 search at the associated MODE_INT.
6588 2018-02-20 Jeff Law <law@redhat.com>
6591 PR tree-optimization/81592
6593 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6594 for range data rather than using global data.
6595 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6596 range data rather than using global data.
6597 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6598 pass it to children as needed.
6599 (struct directive::fmtresult): Similarly.
6600 (struct directive::set_width): Similarly.
6601 (struct directive::set_precision): Similarly.
6602 (format_integer, format_directive, parse_directive): Similarly.
6603 (format_none): Accept unnamed vr_values parameter.
6604 (format_percent, format_floating, format_character): Similarly.
6605 (format_string, format_plain): Similarly.
6606 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6607 the EVRP range analyzer for range data rather than using global data.
6608 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6609 gimple-ssa-evrp-analyze.h
6610 (class sprintf_dom_walker): Add after_dom_children member function.
6611 Add evrp_range_analyzer member.
6612 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6613 range analyzer as needed.
6614 (sprintf_dom_walker::after_dom_children): New member function.
6615 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6617 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6618 (evrp_range_analyzer::pop_to_marker): Likewise.
6620 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6622 PR tree-optimization/84419
6623 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6624 with the required type if its current type is compatible but
6627 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6630 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6631 after vectorization.
6633 2018-02-20 Martin Liska <mliska@suse.cz>
6636 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6637 possible values if we don't have a hint.
6639 2018-02-20 Martin Liska <mliska@suse.cz>
6643 * final.c (shorten_branches): Build align_tab array with one
6645 * opts.c (finish_options): Add alignment option limit check.
6646 (MAX_CODE_ALIGN): Likewise.
6647 (MAX_CODE_ALIGN_VALUE): Likewise.
6648 * doc/invoke.texi: Document maximum allowed option value for
6649 all -falign-* options.
6651 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6654 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6655 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6656 * var-tracking.c (emit_note_insn_var_location): Remove all references
6657 to NOTE_INSN_CALL_ARG_LOCATION.
6658 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6659 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6660 Use copy_rtx_if_shared.
6661 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6662 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6663 (dwarf2out_var_location): Remove handling of
6664 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6666 * final.c (final_scan_insn): Remove all references to
6667 NOTE_INSN_CALL_ARG_LOCATION.
6668 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6669 before dumping final insns.
6670 * except.c (emit_note_eh_region_end): Remove all references to
6671 NOTE_INSN_CALL_ARG_LOCATION.
6672 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6673 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6674 * config/arc/arc.c (hwloop_optimize): Likewise.
6675 * config/arm/arm.c (create_fix_barrier): Likewise.
6676 * config/s390/s390.c (s390_chunkify_start): Likewise.
6677 * config/sh/sh.c (find_barrier): Likewise.
6678 * config/i386/i386.c (rest_of_insert_endbranch,
6679 ix86_seh_fixup_eh_fallthru): Likewise.
6680 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6681 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6682 * config/frv/frv.c (frv_function_prologue): Likewise.
6683 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6685 (note_outside_basic_block_p): Remove all references to
6686 NOTE_INSN_CALL_ARG_LOCATION.
6687 * gengtype.c (adjust_field_rtx_def): Likewise.
6688 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6690 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6691 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6694 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6697 PR tree-optimization/84452
6698 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6699 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6702 2018-02-19 Martin Liska <mliska@suse.cz>
6705 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6707 2018-02-19 Martin Liska <mliska@suse.cz>
6708 Richard Sandiford <richard.sandiford@linaro.org>
6710 PR tree-optimization/82491
6711 * gimple-fold.c (get_base_constructor): Make earlier bail out
6714 2018-02-19 Carl Love <cel@us.ibm.com>
6716 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6717 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6719 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6720 P8V_BUILTIN_VEC_NEG.
6722 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6724 * config/rl78/rl78.md (movdf): New define expand.
6726 2018-02-19 Martin Liska <mliska@suse.cz>
6729 * doc/invoke.texi: Fix typo.
6730 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6732 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6734 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6735 handle rs6000_single_float and rs6000_double_float specially for
6738 2018-02-16 Jeff Law <law@redhat.com>
6740 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6743 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6746 * ipa-inline.c (inline_small_functions): Fix a typo.
6748 2018-02-16 Nathan Sidwell <nathan@acm.org>
6750 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6752 2018-02-16 Carl Love <cel@us.ibm.com>
6754 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6755 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6756 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6757 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6758 expansion to P8V_BUILTIN_VEC_FLOAT2.
6760 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6762 PR rtl-optimization/70023
6763 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6764 src_regno into account.
6766 2018-02-16 Carl Love <cel@us.ibm.com>
6768 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6769 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6770 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6771 * config/rs6000/rs6000.c: Remove case statements for
6772 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6773 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6774 and P9V_BUILTIN_VEC_VINSERT4B.
6775 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6776 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6777 * config/rs6000/vsx.md:
6778 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6781 2018-02-16 Carl Love <cel@us.ibm.com>
6783 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6785 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6787 * config/rs6000/rs6000-c.c: Add the definitions for
6788 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6789 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6790 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6791 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6792 definition for insert4b and define insn *insert3b_internal.
6793 * doc/extend.texi: Add documentation for vec_extract4b.
6795 2018-02-16 Nathan Sidwell <nathan@acm.org>
6797 * doc/extend.texi (Backwards Compatibility): Mention friend
6798 injection. Note for-scope is deprecated.
6799 * doc/invoke.texi (-ffriend-injection): Deprecate.
6801 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6803 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6804 that moved to I2, also allow destinations that are a paradoxical
6805 subreg (instead of a normal reg).
6807 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6810 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6813 2018-02-16 Richard Biener <rguenther@suse.de>
6815 PR tree-optimization/84037
6816 PR tree-optimization/84016
6818 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6819 Adjust vec_construct for the fact we need additional higher latency
6820 128bit inserts for AVX256 and AVX512 vector builds.
6821 (ix86_add_stmt_cost): Scale vector construction cost for
6824 2018-02-16 Richard Biener <rguenther@suse.de>
6826 PR tree-optimization/84417
6827 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6828 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6829 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6831 2018-02-16 Martin Liska <mliska@suse.cz>
6834 * internal-fn.def (ASAN_CHECK): Set proper flags.
6835 (ASAN_MARK): Likewise.
6837 2018-02-16 Julia Koval <julia.koval@intel.com>
6839 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6840 from PTA_CANNONLAKE.
6842 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6845 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6846 Use ++iter rather than iter++ for std::list iterators.
6847 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6848 defer deleting them until all nodes in the forest are processed. Do
6849 free even leaf nodes. Change to_process into auto_vec.
6852 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6853 * vec.h (vec_default_construct): Use memset instead of placement new
6854 if BROKEN_VALUE_INITIALIZATION is defined.
6855 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6856 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6859 PR rtl-optimization/83723
6860 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6861 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6862 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6864 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6866 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6868 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6870 PR rtl-optimization/81443
6871 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6872 from inner REGs to paradoxical SUBREGs.
6874 2018-02-16 Richard Biener <rguenther@suse.de>
6876 PR tree-optimization/84399
6877 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6878 For operands we can analyze at their definition make sure we can
6879 analyze them at each use as well.
6881 2018-02-16 Richard Biener <rguenther@suse.de>
6883 PR tree-optimization/84190
6884 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6885 volatile accesses if the decl isn't volatile.
6887 2018-02-15 Jason Merrill <jason@redhat.com>
6889 PR c++/84314 - ICE with templates and fastcall attribute.
6890 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6891 TYPE_CANONICAL on an existing type.
6893 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6895 PR tree-optimization/84383
6896 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6897 dstoff nor call operand_equal_p if dstbase is NULL.
6899 PR tree-optimization/84334
6900 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6901 also a CONSTANT_CLASS_P, punt.
6903 2018-02-14 Jim Wilson <jimw@sifive.com>
6905 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6906 first SMALL_OPERAND check. New local min_second_step. Move assert
6907 to where locals are set. Add TARGET_RVC support.
6908 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6910 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6912 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6914 2018-02-14 Martin Sebor <msebor@redhat.com>
6916 PR tree-optimization/83698
6917 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6918 arrays constrain the offset range to their bounds.
6919 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6920 (builtin_access::overlap): Avoid setting the size of overlap if it's
6922 (maybe_diag_overlap): Also consider arrays when deciding what values
6923 of offsets to include in diagnostics.
6925 2018-02-14 Martin Sebor <msebor@redhat.com>
6928 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6929 that correspond to the kind of a declaration.
6931 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6934 * config/pa/pa.md: Load address of PIC label using the linkage table
6935 if the label is nonlocal.
6937 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6939 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6940 warning message if user requests -maltivec=be.
6941 * doc/invoke.texi: Document deprecation of -maltivec=be.
6943 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6946 * config/rs6000/rs6000-c.c: Update definitions for
6947 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6948 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6950 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6953 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6954 add _get_ssp intrinsics. Remove argument from
6955 __builtin_ia32_rdssp[d|q].
6956 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6957 * config/i386/i386-builtin.def: Remove argument from
6958 __builtin_ia32_rdssp[d|q].
6959 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6960 ix86_expand_special_args_builtin for _rdssp[d|q].
6961 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6962 Clear register before usage.
6963 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6964 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6966 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6968 PR tree-optimization/84357
6969 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6970 operand 1 of an ARRAY_REF too.
6972 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6975 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6976 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6978 (set_of_reg): New struct.
6979 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6980 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6981 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6983 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6984 Split into bitclr, bitset, bitinvert patterns if appropriate.
6985 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6986 use rx_fuse_in_memory_bitop.
6987 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6988 to named insn, correct maximum insn length.
6990 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6993 * machmode.def: Define a complex mode for PARTIAL_INT.
6994 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6996 * doc/rtl.texi: Document CSPImode.
6997 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6999 (msp430_hard_regno_nregs_with_padding): Likewise.
7001 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7004 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7006 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7008 PR rtl-optimization/84169
7009 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7010 we generated a parallel as new i3 and we split that to new i2 and i3
7011 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7012 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7013 those to i2, not i1. Partially rewrite this scan code.
7015 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7018 * stor-layout.c (place_field): For variable length fields, adjust
7019 offset_align afterwards not just based on the field's alignment,
7020 but also on the size.
7023 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7024 of exps and logs in the use_exp2 case.
7026 2018-02-13 Jeff Law <law@redhat.com>
7028 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7031 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7034 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7038 * common.opt (gas-loc-support, gas-locview-support): New.
7039 (ginline-points, ginternal-reset-location-views): New.
7040 * doc/invoke.texi: Document them. Use @itemx where intended.
7041 (gvariable-location-views): Adjust.
7042 * target.def (reset_location_view): New.
7043 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7044 (TARGET_RESET_LOCATION_VIEW): New.
7045 * doc/tm.texi: Rebuilt.
7046 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7047 (dwarf2out_default_as_locview_support): New.
7048 (output_asm_line_debug_info): Use option variables.
7049 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7050 (output_loc_list): Likewise.
7051 (add_high_low_attributes): Check option variables.
7052 Don't output entry view attribute in strict mode.
7053 (gen_inlined_subroutine_die): Check option variables.
7054 (dwarf2out_inline_entry): Likewise.
7055 (init_sections_and_labels): Likewise.
7056 (dwarf2out_early_finish): Likewise.
7057 (maybe_reset_location_view): New, from...
7058 (dwarf2out_var_location): ... here. Call it.
7059 * debug.h (dwarf2out_default_as_loc_support): Declare.
7060 (dwarf2out_default_as_locview_support): Declare.
7061 * hooks.c (hook_int_rtx_insn_0): New.
7062 * hooks.h (hook_int_rtx_insn_0): Declare.
7063 * toplev.c (process_options): Take -gas-loc-support and
7064 -gas-locview-support from dwarf2out. Enable
7065 -gvariable-location-views by default only with locview
7066 assembler support. Enable -ginternal-reset-location-views by
7067 default only if the target defines the corresponding hook.
7068 Enable -ginline-points by default if location views are
7069 enabled; force it disabled if statement frontiers are
7071 * tree-inline.c (expand_call_inline): Check option variables.
7072 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7074 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7076 PR tree-optimization/84321
7077 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7078 handling. Also check whether the anti-range contains any values
7079 that satisfy the mask; switch to a VR_RANGE if not.
7081 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7084 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7086 2018-02-13 Martin Jambor <mjambor@suse.cz>
7089 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7090 of call statements, also set location of a load to a temporary.
7092 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7094 * config/rl78/rl78.c (add_vector_labels): New function.
7095 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7096 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7097 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7098 which checks that no arguments are passed.
7099 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7100 * doc/extend.texi: Documentation for the new attribute.
7102 2018-02-13 Andreas Schwab <schwab@suse.de>
7104 * config/riscv/linux.h (CPP_SPEC): Define.
7106 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7109 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7110 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7111 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7112 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7113 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7114 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7115 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7116 temporarily for AES and PCLMUL builtins.
7118 PR tree-optimization/84339
7119 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7120 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7124 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7125 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7126 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7128 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7130 * omp-simd-clone.h: New file.
7131 * omp-simd-clone.c: Include omp-simd-clone.h.
7132 (expand_simd_clones): No longer static.
7133 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7134 cgraph.h and omp-simd-clone.h.
7135 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7136 (vect_recog_widen_shift_pattern): Formatting fix.
7137 (vect_pattern_recog_1): Don't check optab for calls.
7140 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7141 operands[2] into a REG before using gen_lowpart on it.
7143 2018-02-12 Jeff Law <law@redhat.com>
7146 * config/sh/sh.c (find_barrier): Consider a sibling call
7149 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7150 successfully back substituting a reg.
7152 2018-02-12 Richard Biener <rguenther@suse.de>
7154 PR tree-optimization/84037
7155 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7156 parameter, move visited init to caller.
7157 (vect_slp_analyze_operations): Separate cost from validity
7158 check, initialize visited once for all instances.
7159 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7161 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7163 (vect_model_promotion_demotion_cost): Likewise.
7164 (vectorizable_bswap): Guard cost modeling with !slp_node
7165 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7167 (vectorizable_call): Likewise.
7168 (vectorizable_conversion): Likewise.
7169 (vectorizable_assignment): Likewise.
7170 (vectorizable_shift): Likewise.
7171 (vectorizable_operation): Likewise.
7172 (vectorizable_store): Likewise.
7173 (vectorizable_load): Likewise.
7174 (vectorizable_condition): Likewise.
7175 (vectorizable_comparison): Likewise.
7177 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7180 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7181 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7182 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7184 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7187 * params.def (inline-min-speedup): Increase from 8 to 15.
7188 (max-inline-insns-auto): Decrease from 40 to 30.
7189 * ipa-split.c (consider_split): Add some buffer for function to
7190 be considered inlining candidate.
7191 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7194 2018-02-12 Richard Biener <rguenther@suse.de>
7196 PR tree-optimization/84037
7197 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7198 matched stmts if we cannot swap the non-matched ones.
7200 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7202 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7203 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7204 _mm_maskz_scalef_round_ss): New intrinsics.
7205 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7206 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7207 __builtin_ia32_scalefss_round): Remove.
7208 (__builtin_ia32_scalefsd_mask_round,
7209 __builtin_ia32_scalefss_mask_round): New intrinsics.
7210 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7211 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7212 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7213 "<round_constraint>")): Changed to ...
7214 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7215 "<round_scalar_constraint>")): ... this.
7216 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7217 %0, %1, %2<round_op3>}"): Changed to ...
7218 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7219 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7220 %2<round_scalar_mask_op3>}"): ... this.
7221 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7223 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7225 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7226 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7227 (_mm_maskz_sqrt_round_ss): New intrinsics.
7228 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7229 (__builtin_ia32_sqrtsd_mask_round)
7230 (__builtin_ia32_sqrtss_mask_round): New builtins.
7231 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7232 (__builtin_ia32_sqrtss_round): Remove.
7233 (__builtin_ia32_sqrtsd_mask_round)
7234 (__builtin_ia32_sqrtss_mask_round): New builtins.
7235 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7236 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7237 ((match_operand:VF_128 1 "vector_operand"
7238 "xBm,<round_constraint>")): Changed to ...
7239 ((match_operand:VF_128 1 "vector_operand"
7240 "xBm,<round_scalar_constraint>")): ... this.
7241 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7242 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7243 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7244 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7245 %<iptr>1<round_scalar_mask_op3>}): ... this.
7246 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7247 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7249 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7252 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7253 Cast vec_cmpeq result to correct type.
7254 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7255 Cast vec_cmpgt result to correct type.
7257 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7259 * final.c (final_scan_insn_1): Renamed from...
7260 (final_scan_insn): ... this. New wrapper, to recover
7261 seen from the outermost call in recursive ones.
7262 * config/sparc/sparc.c (output_return): Drop seen from call.
7263 (output_sibcall): Likewise.
7264 * config/visium/visium.c (output_branch): Likewise.
7266 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7268 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7271 2018-02-10 Alan Modra <amodra@gmail.com>
7274 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7275 Specify LR as an input.
7277 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7280 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7281 remove_member_access_dummy_vars): New functions.
7282 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7283 lower_omp_1, execute_lower_omp): Use them.
7285 PR rtl-optimization/84308
7286 * shrink-wrap.c (spread_components): Release todo vector.
7288 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7290 PR rtl-optimization/57193
7291 * ira-color.c (struct allocno_color_data): Add member
7292 conflict_allocno_hard_prefs.
7293 (update_conflict_allocno_hard_prefs): New.
7294 (bucket_allocno_compare_func): Add a preference based on
7295 conflict_allocno_hard_prefs.
7296 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7297 (color_allocnos): Remove a dead code. Initiate
7298 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7300 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7303 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7304 constraint from =wa to wa. Avoid a subreg on the output operand,
7305 instead use a pseudo and subreg it in a move.
7306 (p9_xxbrd_<mode>): Changed to ...
7307 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7308 (p9_xxbrd_v2df): New expander.
7309 (p9_xxbrw_<mode>): Changed to ...
7310 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7311 (p9_xxbrw_v4sf): New expander.
7313 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7315 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7317 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7320 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7321 multiply in 32-bit mode.
7322 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7323 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7326 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7328 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7329 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7330 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7331 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7333 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7336 * dwarf2out.c (is_trivial_indirect_ref): New function.
7337 (dwarf2out_late_global_decl): Do not generate a location
7338 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7339 and that are not defined in the current unit.
7341 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7343 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7344 instead of a libcall for UNORDERED.
7346 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7349 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7350 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7352 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7355 * config/s390/s390.c (s390_set_current_function): Invoke
7356 s390_indirect_branch_settings also if fndecl didn't change.
7358 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7360 * config/rs6000/rs6000.md (blockage): Set length to zero.
7362 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7364 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7366 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7369 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7370 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7374 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7375 PARALLEL incoming that failed vt_get_decl_and_offset check.
7378 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7379 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7381 (get_variable_section): For decls in named .bss* sections pass true as
7382 second argument to bss_initializer_p.
7384 2018-02-09 Marek Polacek <polacek@redhat.com>
7385 Jakub Jelinek <jakub@redhat.com>
7388 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7389 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7390 Sync some changes from cxx_fold_indirect_ref.
7392 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7394 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7396 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7397 (BLOCK_INLINE_ENTRY_LABEL): New.
7398 (dwarf2out_var_location): Disregard inline entry markers.
7399 (inline_entry_data): New struct.
7400 (inline_entry_data_hasher): New hashtable type.
7401 (inline_entry_data_hasher::hash): New.
7402 (inline_entry_data_hasher::equal): New.
7403 (inline_entry_data_table): New variable.
7404 (add_high_low_attributes): Add DW_AT_entry_pc and
7405 DW_AT_GNU_entry_view attributes if a pending entry is found
7406 in inline_entry_data_table. Add old entry_pc attribute only
7407 if debug nonbinding markers are disabled.
7408 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7409 markers are enabled.
7410 (block_within_block_p, dwarf2out_inline_entry): New.
7411 (dwarf2out_finish): Check that no entries remained in
7412 inline_entry_data_table.
7413 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7414 (final_scan_insn, notice_source_line): Likewise.
7415 (rest_of_clean_state): Skip inline entry markers.
7416 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7418 * gimple.c (gimple_build_debug_inline_entry): New.
7419 * gimple.h (enum gimple_debug_subcode): Add
7420 GIMPLE_DEBUG_INLINE_ENTRY.
7421 (gimple_build_debug_inline_entry): Declare.
7422 (gimple_debug_inline_entry_p): New.
7423 (gimple_debug_nonbind_marker_p): Adjust.
7424 * insn-notes.def (INLINE_ENTRY): New.
7425 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7426 inline entry marker notes.
7427 (print_insn): Likewise.
7428 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7429 (INSN_DEBUG_MARKER_KIND): Likewise.
7430 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7431 * tree-inline.c (expand_call_inline): Build and insert
7432 debug_inline_entry stmt.
7433 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7434 inline entry blocks early, if nonbind markers are enabled.
7435 (dump_scope_block): Dump fragment info.
7436 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7437 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7438 (gimple_build_debug_inline_entry): New.
7439 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7440 Enable/disable inline entry points too.
7441 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7442 (DEBUG_INSN): Describe inline entry markers.
7444 * common.opt (gvariable-location-views): New.
7445 (gvariable-location-views=incompat5): New.
7446 * config.in: Rebuilt.
7447 * configure: Rebuilt.
7448 * configure.ac: Test assembler for view support.
7449 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7450 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7451 * dwarf2out.c (var_loc_view): New typedef.
7452 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7453 (dwarf2out_locviews_in_attribute): New.
7454 (dwarf2out_locviews_in_loclist): New.
7455 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7456 (enum dw_line_info_opcode): Add LI_adv_address.
7457 (struct dw_line_info_table): Add view.
7458 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7459 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7460 (zero_view_p): New variable.
7461 (ZERO_VIEW_P): New macro.
7462 (output_asm_line_debug_info): New.
7463 (struct var_loc_node): Add view.
7464 (add_AT_view_list, AT_loc_list): New.
7465 (add_var_loc_to_decl): Add view param. Test it against last.
7466 (new_loc_list): Add view params. Record them.
7467 (AT_loc_list_ptr): Handle loc and view lists.
7468 (view_list_to_loc_list_val_node): New.
7469 (print_dw_val): Handle dw_val_class_view_list.
7470 (size_of_die): Likewise.
7471 (value_format): Likewise.
7472 (loc_list_has_views): New.
7473 (gen_llsym): Set vl_symbol too.
7474 (maybe_gen_llsym, skip_loc_list_entry): New.
7475 (dwarf2out_maybe_output_loclist_view_pair): New.
7476 (output_loc_list): Output view list or entries too.
7477 (output_view_list_offset): New.
7478 (output_die): Handle dw_val_class_view_list.
7479 (output_dwarf_version): New.
7480 (output_compilation_unit_header): Use it.
7481 (output_skeleton_debug_sections): Likewise.
7482 (output_rnglists, output_line_info): Likewise.
7483 (output_pubnames, output_aranges): Update version comments.
7484 (output_one_line_info_table): Output view numbers in asm comments.
7485 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7486 Call maybe_gen_llsym.
7487 (loc_list_from_tree_1): Adjust.
7488 (add_AT_location_description): Create view list attribute if
7489 needed, check it's absent otherwise.
7490 (convert_cfa_to_fb_loc_list): Adjust.
7491 (maybe_emit_file): Call output_asm_line_debug_info for test.
7492 (dwarf2out_var_location): Reset views as needed. Precompute
7493 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7494 attribute. Set view.
7495 (new_line_info_table): Reset next view.
7496 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7497 (dwarf2out_source_line): Likewise. Output view resets and labels to
7498 the assembler, or select appropriate line info opcodes.
7499 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7500 (optimize_string_length): Catch it. Adjust.
7501 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7502 dw_val_class_view_list, and remove it if no longer needed.
7503 (hash_loc_list): Hash view numbers.
7504 (loc_list_hasher::equal): Compare them.
7505 (optimize_location_lists): Check whether a view list symbol is
7506 needed, and whether the locview attribute is present, and
7507 whether they match. Remove the locview attribute if no longer
7509 (index_location_lists): Call skip_loc_list_entry for test.
7510 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7511 Use output_dwarf_version.
7512 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7513 (struct dw_val_node): Add val_view_list.
7514 * final.c (SEEN_NEXT_VIEW): New.
7515 (set_next_view_needed): New.
7516 (clear_next_view_needed): New.
7517 (maybe_output_next_view): New.
7518 (final_start_function): Rename to...
7519 (final_start_function_1): ... this. Take pointer to FIRST,
7520 add SEEN parameter. Emit param bindings in the initial view.
7521 (final_start_function): Reintroduce SEEN-less interface.
7522 (final): Rename to...
7523 (final_1): ... this. Take SEEN parameter. Output final pending
7524 next view at the end.
7525 (final): Reintroduce seen-less interface.
7526 (final_scan_insn): Output pending next view before switching
7527 sections or ending a block. Mark the next view as needed when
7528 outputting variable locations. Notify debug backend of section
7529 changes, and of location view changes.
7530 (rest_of_handle_final): Adjust.
7531 * toplev.c (process_options): Autodetect value for debug variable
7532 location views option. Warn on incompat5 without -gdwarf-5.
7533 * doc/invoke.texi (gvariable-location-views): New.
7534 (gvariable-location-views=incompat5): New.
7535 (gno-variable-location-views): New.
7537 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7539 PR tree-optimization/84136
7540 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7541 that the result of find_edge is non-NULL.
7543 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7546 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7547 storing integer register in SImode. Fix cost of 256 and 512
7548 byte aligned SSE register store.
7550 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7552 * config/i386/i386.c (ix86_multiplication_cost): Fix
7553 multiplication cost for TARGET_AVX512DQ.
7555 2018-02-08 Marek Polacek <polacek@redhat.com>
7557 PR tree-optimization/84238
7558 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7561 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7563 PR tree-optimization/84265
7564 * tree-vect-stmts.c (vectorizable_store): Don't treat
7565 VMAT_CONTIGUOUS accesses as grouped.
7566 (vectorizable_load): Likewise.
7568 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7570 PR tree-optimization/81635
7571 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7572 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7573 (test_round_for_mask): New functions.
7574 (wide_int_cc_tests): Call test_round_for_mask.
7575 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7576 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7577 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7578 range returned by get_range_info.
7580 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7583 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7584 * symtab.c: Include builtins.h
7585 (symtab_node::output_to_lto_symbol_table_p): Move here
7586 from lto-streamer-out.c:output_symbol_p.
7587 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7588 (output_symbol_p): Move all logic to symtab.c
7589 (produce_symtab): Update.
7591 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7593 * config/s390/s390-opts.h (enum indirect_branch): Define.
7594 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7595 (s390_indirect_branch_via_thunk)
7596 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7597 (enum s390_indirect_branch_type): Define.
7598 * config/s390/s390.c (struct s390_frame_layout, struct
7599 machine_function): Remove.
7600 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7601 (indirect_branch_table_label_no, indirect_branch_table_name):
7603 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7604 (enum s390_indirect_branch_option): Define.
7605 (s390_return_addr_from_memory): New function.
7606 (s390_handle_string_attribute): New function.
7607 (s390_attribute_table): Add new attribute handler.
7608 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7609 (s390_indirect_branch_via_thunk): New function.
7610 (s390_indirect_branch_via_inline_thunk): New function.
7611 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7612 sibling call optimization for non z10 compiles.
7613 (s390_emit_call): Force indirect branch target to be a single
7614 register. Add r1 clobber for non-z10 compiles.
7615 (s390_emit_epilogue): Emit return jump via return_use expander.
7616 (s390_reorg): Handle JUMP_INSNs as execute targets.
7617 (s390_option_override_internal): Perform validity checks for the
7618 new command line options.
7619 (s390_indirect_branch_attrvalue): New function.
7620 (s390_indirect_branch_settings): New function.
7621 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7622 (s390_output_indirect_thunk_function): New function.
7623 (s390_code_end): Implement target hook.
7624 (s390_case_values_threshold): Implement target hook.
7625 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7627 * config/s390/s390.h (struct s390_frame_layout)
7628 (struct machine_function): Move here from s390.c.
7629 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7630 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7631 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7632 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7633 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7634 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7635 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7636 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7637 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7638 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7639 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7640 (mnemonic attribute): Add values which aren't recognized
7642 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7643 pattern for branch conversion. Fix mnemonic attribute.
7644 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7645 indirect branch via thunk if requested.
7646 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7647 ("*indirect_jump"): Disable for branch conversion using out of
7649 ("indirect_jump_via_thunk<mode>_z10")
7650 ("indirect_jump_via_thunk<mode>")
7651 ("indirect_jump_via_inlinethunk<mode>_z10")
7652 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7653 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7654 ("casesi_jump_via_inlinethunk<mode>_z10")
7655 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7656 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7657 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7658 ("*indirect2_jump"): Disable for branch conversion.
7659 ("casesi_jump"): Turn into expander and expand patterns for branch
7661 ("return_use"): New expander.
7662 ("*return"): Emit return via thunk and rename it to ...
7663 ("*return<mode>"): ... this one.
7664 * config/s390/s390.opt: Add new options and and enum for the
7667 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7669 * lra-constraints.c (match_reload): Unconditionally use
7670 gen_lowpart_SUBREG, rather than selecting between that
7671 and equivalent gen_rtx_SUBREG code.
7673 2018-02-08 Richard Biener <rguenther@suse.de>
7675 PR tree-optimization/84233
7676 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7677 changed flag instead of boguously re-using phi_inserted.
7679 2018-02-08 Martin Jambor <mjambor@suse.cz>
7681 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7682 static local variables.
7684 2018-02-08 Richard Biener <rguenther@suse.de>
7686 PR tree-optimization/84278
7687 * tree-vect-stmts.c (vectorizable_store): When looking for
7688 smaller vector types to perform grouped strided loads/stores
7689 make sure the mode is supported by the target.
7690 (vectorizable_load): Likewise.
7692 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7694 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7695 Increase LDP/STP opportunities by adding adjacent callee-saves.
7697 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7699 PR rtl-optimization/84068
7700 PR rtl-optimization/83459
7701 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7703 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7705 PR tree-optimization/84224
7706 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7707 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7710 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7713 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7714 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7715 count, remove one USE.
7717 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7719 * doc/install.texi (Configuration): Document the
7720 --with-long-double-format={ibm,ieee} PowerPC configuration
7724 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7725 Convert from define_expand to be define_insn_and_split. Rework
7726 float/double/_Float128 conversions to QI/HI/SImode to work with
7727 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7728 conversions to QI/HImode types did a store and then a load to
7729 truncate the value. For conversions to VSX registers, don't split
7730 the insn, instead emit the code directly. Use the code iterator
7731 any_fix to combine signed and unsigned conversions.
7732 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7733 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7734 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7735 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7736 (fix_<mode>di2_hw): Likewise.
7737 (fixuns_<mode>di2_hw): Likewise.
7738 (fix_<mode>si2_hw): Likewise.
7739 (fixuns_<mode>si2_hw): Likewise.
7740 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7741 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7742 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7743 fix<uns>_trunc<SFDF:mode>si2_p8.
7744 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7746 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7747 (fix<uns>_<mode>_mem): Likewise.
7748 (fctiw<u>z_<mode>_mem): Likewise.
7749 (fix<uns>_<mode>_mem): Likewise.
7750 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7751 the register allocator from doing a direct move to the GPRs to do
7752 a store, and instead use the ISA 3.0 store byte/half-word from
7753 vector register instruction. For IEEE 128-bit floating point,
7754 also optimize stores of 32-bit ints.
7755 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7757 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7759 * genextract.c (push_pathstr_operand): New function to support
7761 (walk_rtx): Call push_pathstr_operand.
7762 (print_path): Support [a-zA-Z].
7764 2018-02-07 Richard Biener <rguenther@suse.de>
7766 PR tree-optimization/84037
7767 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7768 (cse_and_gimplify_to_preheader): Declare.
7769 (vect_get_place_in_interleaving_chain): Likewise.
7770 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7772 (_loop_vec_info::~_loop_vec_info): Delete it.
7773 (cse_and_gimplify_to_preheader): New function.
7774 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7775 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7776 (vectorizable_load): Likewise. For grouped stores always base
7777 the IV on the first element.
7778 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7779 condition before gimplifying.
7781 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7783 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7784 *DIV_EXPR and *MOD_EXPR.
7786 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7789 * config/i386/i386.c (ix86_option_override_internal): Mask out
7790 the CF_SET bit when checking -fcf-protection.
7792 2018-02-07 Tom de Vries <tom@codesourcery.com>
7795 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7798 2018-02-07 Richard Biener <rguenther@suse.de>
7800 PR tree-optimization/84204
7801 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7804 PR tree-optimization/84205
7805 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7806 special-case isl_ast_op_zdiv_r.
7808 PR tree-optimization/84223
7809 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7810 Only add conditions from within the region.
7811 (gather_bbs::after_dom_children): Adjust.
7813 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7816 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7817 * config/avr/avr.md: Only post-reload split REG-REG moves if
7818 either register is GENERAL_REG_P.
7820 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7822 PR tree-optimization/84235
7823 * tree-ssa-scopedtables.c
7824 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7825 if the subtraction is performed in floating point type where NaNs are
7826 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7827 build 1. Formatting fix.
7829 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7832 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7833 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7834 and skip it regardless of bb boundaries. Use CALL_P macro,
7835 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7836 unnecessarily, formatting fix.
7838 2018-02-06 Michael Collison <michael.collison@arm.com>
7840 * config/arm/thumb2.md:
7841 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7842 (*thumb_mov_notscc): Ditto.
7844 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7847 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7848 unsigned_fix, not "s".
7850 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7852 * configure.ac (gcc_fn_eh_frame_ro): New function.
7853 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7854 correct .eh_frame permissions.
7855 * configure: Regenerate.
7857 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7859 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7862 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7864 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7865 Display warning message for -mno-speculate-indirect-jumps.
7867 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7869 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7871 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7873 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7875 PR tree-optimization/84225
7876 * tree-eh.c (find_trapping_overflow): Only call
7877 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7879 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7882 * config/i386/i386.c: Reimplement the check of possible options
7883 -mibt/-mshstk conbination. Change error messages.
7884 * doc/invoke.texi: Fix a typo: remove extra '='.
7886 2018-02-06 Marek Polacek <polacek@redhat.com>
7888 PR tree-optimization/84228
7889 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7891 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7894 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7895 emitted arch directives.
7896 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7897 __ARM_FEATURE_COPROC before changing architectures.
7899 2018-02-06 Richard Biener <rguenther@suse.de>
7901 * config/i386/i386.c (print_reg): Fix typo.
7902 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7904 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7906 * configure: Regenerate.
7908 2018-02-05 Martin Sebor <msebor@redhat.com>
7910 PR tree-optimization/83369
7911 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7914 2018-02-05 Martin Liska <mliska@suse.cz>
7916 * doc/invoke.texi: Cherry-pick upstream r323995.
7918 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7920 * ira.c (ira_init_register_move_cost): Adjust comment.
7922 2018-02-05 Martin Liska <mliska@suse.cz>
7924 PR gcov-profile/84137
7925 * doc/gcov.texi: Fix typo in documentation.
7927 2018-02-05 Martin Liska <mliska@suse.cz>
7929 PR gcov-profile/83879
7930 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7931 using dlopen functionality.
7933 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7935 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7936 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7937 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7938 _mm_maskz_range_round_ss): New intrinsics.
7939 (__builtin_ia32_rangesd128_round)
7940 (__builtin_ia32_rangess128_round): Remove.
7941 (__builtin_ia32_rangesd128_mask_round,
7942 __builtin_ia32_rangess128_mask_round): New builtins.
7943 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7944 __builtin_ia32_rangess128_round): Remove.
7945 (__builtin_ia32_rangesd128_mask_round,
7946 __builtin_ia32_rangess128_mask_round): New builtins.
7947 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7948 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7949 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7950 "<round_saeonly_constraint>")): Changed to ...
7951 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7952 "<round_saeonly_scalar_constraint>")): ... this.
7953 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7954 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7955 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7956 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7957 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7959 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7961 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7963 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7964 Remove all values except native, 8540 and 8548.
7966 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7968 * config/i386/i386.c (ix86_output_function_return): Pass
7969 INVALID_REGNUM, instead of -1, as invalid register number to
7970 indirect_thunk_name and output_indirect_thunk.
7972 2018-02-02 Julia Koval <julia.koval@intel.com>
7974 * config.gcc: Add -march=icelake.
7975 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7976 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7977 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7978 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7979 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7980 (processor_target_table): Add icelake.
7981 (ix86_option_override_internal): Handle new PTAs.
7982 (get_builtin_code_for_version): Handle icelake.
7983 (M_INTEL_COREI7_ICELAKE): New.
7984 (fold_builtin_cpu): Handle icelake.
7985 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7986 * doc/invoke.texi: Add -march=icelake.
7988 2018-02-02 Julia Koval <julia.koval@intel.com>
7990 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7991 to wide_int_bitmask.
7992 * wide-int-bitmask.h: New.
7994 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7997 * config/i386/i386.md: Replace Pmode with word_mode in
7998 builtin_setjmp_setup and builtin_longjmp to support x32.
8000 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8004 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8006 (rs6000_supported_cpu_names): New static variable.
8007 (linux_cpu_translation_table): Likewise.
8008 (elf_platform) <cpu>: Define new static variable and use it.
8009 Translate kernel AT_PLATFORM name to canonical name if needed.
8010 Error if platform name is unknown.
8012 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8015 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8017 2018-02-01 Jeff Law <law@redhat.com>
8020 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8021 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8022 the scratch if RELEASE_VIA_POP is false.
8023 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8024 If we have to save a temporary register, decrement SIZE appropriately.
8025 Pass new arguments to release_scratch_register_on_entry.
8026 (ix86_adjust_stack_and_probe): Likewise.
8027 (ix86_emit_probe_stack_range): Pass new arguments to
8028 release_scratch_register_on_entry.
8030 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8032 PR rtl-optimization/84157
8033 * combine.c (change_zero_ext): Use REG_P predicate in
8034 front of HARD_REGISTER_P predicate.
8036 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8038 * config/avr/avr.c (avr_option_override): Move disabling of
8039 -fdelete-null-pointer-checks to...
8040 * common/config/avr/avr-common.c (avr_option_optimization_table):
8043 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8045 PR tree-optimization/81635
8046 * tree-data-ref.c (split_constant_offset_1): For types that
8047 wrap on overflow, try to use range info to prove that wrapping
8050 2018-02-01 Renlin Li <renlin.li@arm.com>
8053 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8055 (aarch64_register_move_cost): Likewise.
8056 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8058 (REG_CLASS_NAMES): Likewise.
8059 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8060 TAILCALL_ADDR_REGS. Remove IP registers.
8061 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8063 2018-02-01 Richard Biener <rguenther@suse.de>
8065 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8066 for specifying RPO order and allow NULL for that.
8067 * domwalk.c (dom_walker::dom_walker): Likewise.
8068 (dom_walker::walk): Handle NULL RPO order.
8069 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8071 (rewrite_update_dom_walker): Likewise.
8072 (mark_def_dom_walker): Likewise.
8074 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8076 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8077 (aarch64_maybe_expand_sve_subreg_move): Declare.
8078 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8079 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8081 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8082 that are semantically a reverse operation.
8083 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8084 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8085 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8087 (aarch64_can_change_mode_class): For big-endian, forbid changes
8088 between two SVE modes if they have different element sizes.
8090 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8092 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8093 the TImode handling for big-endian targets.
8095 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8097 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8098 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8100 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8101 Remove BSWAP handing for big-endian targets and use the form of
8102 LD1RQ appropariate for the mode.
8104 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8106 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8107 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8110 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8113 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8114 check for operands that need to go through aarch64_sve_reload_be.
8116 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8118 PR tree-optimization/81661
8119 PR tree-optimization/84117
8120 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8121 * tree-eh.c: Include gimplify.h.
8122 (find_trapping_overflow, replace_trapping_overflow,
8123 rewrite_to_non_trapping_overflow): New functions.
8124 * tree-vect-loop.c: Include tree-eh.h.
8125 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8126 * tree-data-ref.c: Include tree-eh.h.
8127 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8129 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8131 PR rtl-optimization/84123
8132 * combine.c (change_zero_ext): Check if hard register satisfies
8133 can_change_dest_mode before calling gen_lowpart_SUBREG.
8135 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8138 * ira.c (ira_init_register_move_cost): Remove assert.
8140 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8142 PR rtl-optimization/84071
8143 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8144 * doc/tm.texi: Regenerate.
8146 2018-01-31 Richard Biener <rguenther@suse.de>
8148 PR tree-optimization/84132
8149 * tree-data-ref.c (analyze_miv_subscript): Properly
8150 check whether evolution_function_is_affine_multivariate_p
8151 before calling gcd_of_steps_may_divide_p.
8153 2018-01-31 Julia Koval <julia.koval@intel.com>
8156 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8157 * config/i386/i386.md (rdpid_rex64) New.
8158 (rdpid): Make 32bit only.
8160 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8163 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8164 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8166 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8169 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8171 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8173 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8175 PR rtl-optimization/84071
8176 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8177 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8179 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8181 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8182 (arc_attribute_table): Add 'aux' attribute.
8183 (arc_in_small_data_p): Consider aux like variables.
8184 (arc_is_aux_reg_p): New function.
8185 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8186 (arc_get_aux_arg): New function.
8187 (prepare_move_operands): Handle aux-register access.
8188 (arc_handle_aux_attribute): New function.
8189 * doc/extend.texi (ARC Variable attributes): Add subsection.
8191 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8193 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8194 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8195 (arc_attribute_table): Add 'uncached' attribute.
8196 (arc_print_operand): Print '.di' flag for uncached memory
8198 (arc_in_small_data_p): Do not consider for small data the uncached
8200 (arc_is_uncached_mem_p): New function.
8201 * config/arc/predicates.md (compact_store_memory_operand): Check
8202 for uncached memory accesses.
8203 (nonvol_nonimm_operand): Likewise.
8204 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8206 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8209 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8210 falign-loops=): Add Optimization flag.
8212 2018-01-30 Jeff Law <law@redhat.com>
8215 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8216 INT_REGISTERS_SAVED. Check it prior to calling
8217 get_scratch_register_on_entry.
8218 (ix86_adjust_stack_and_probe): Similarly.
8219 (ix86_emit_probe_stack_range): Similarly.
8220 (ix86_expand_prologue): Corresponding changes.
8222 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8225 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8226 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8228 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8231 * lra-constraints.c (curr_insn_transform): Process AND in the
8234 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8236 PR rtl-optimization/83986
8237 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8238 dependence against last_pending_memory_flush in addition to
8241 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8243 PR tree-optimization/81611
8244 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8247 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8250 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8253 2018-01-30 Richard Biener <rguenther@suse.de>
8254 Jakub Jelinek <jakub@redhat.com>
8256 PR tree-optimization/84111
8257 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8258 inner loops added during recursion, as they don't have up-to-date
8261 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8264 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8265 (can_inline_edge_by_limits_p): ... here.
8266 (can_early_inline_edge_p, check_callers,
8267 update_caller_keys, update_callee_keys, recursive_inlining,
8268 add_new_edges_to_heap, speculation_useful_p,
8269 inline_small_functions,
8270 inline_small_functions, flatten_function,
8271 inline_to_all_callers_1): Update.
8273 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8275 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8278 2018-01-30 Richard Biener <rguenther@suse.de>
8280 PR tree-optimization/83008
8281 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8282 invariant and constant vector uses in stmts when they need
8285 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8288 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8289 * configure: Regenerate.
8291 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8293 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8295 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8296 Use gen_rtx_REG rather than gen_lowpart.
8298 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8300 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8301 rather than 0 when creating partial subregs.
8303 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8305 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8308 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8311 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8312 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8313 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8314 flags. This restores the settings used before the 2017-07-24.
8315 Turning off pre increment/decrement/modify allows IVOPTS to
8316 optimize DF/SF loops where the index is an int.
8318 2018-01-29 Richard Biener <rguenther@suse.de>
8319 Kelvin Nilsen <kelvin@gcc.gnu.org>
8322 * tree-vect-stmts.c (vectorizable_call): Don't call
8323 targetm.vectorize_builtin_md_vectorized_function if callee is
8326 2018-01-22 Carl Love <cel@us.ibm.com>
8328 * doc/extend.tex: Fix typo in second arg in
8329 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8331 2018-01-29 Richard Biener <rguenther@suse.de>
8333 PR tree-optimization/84086
8334 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8335 (flush_ssaname_freelist): When SSA names were released reset
8336 the SCEV hash table.
8338 2018-01-29 Richard Biener <rguenther@suse.de>
8340 PR tree-optimization/84057
8341 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8342 removed paths when removing edges.
8344 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8346 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8347 -mfunction-return=@var{choice}.
8349 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8352 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8354 (layout::print_source_line): Likewise.
8355 (test_get_line_width_without_trailing_whitespace): Add test cases.
8357 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8360 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8363 2018-01-26 Jim Wilson <jimw@sifive.com>
8365 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8367 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8370 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8372 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8373 and CMP + SUB-immediate -> SUBS.
8375 2018-01-26 Martin Sebor <msebor@redhat.com>
8377 PR tree-optimization/83896
8378 * tree-ssa-strlen.c (get_string_len): Rename...
8379 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8380 Avoid assuming length is constant.
8381 (handle_char_store): Use HOST_WIDE_INT for string length.
8383 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8386 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8387 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8389 2018-01-26 Richard Biener <rguenther@suse.de>
8391 PR rtl-optimization/84003
8392 * dse.c (record_store): Only record redundant stores when
8393 the earlier store aliases at least all accesses the later one does.
8395 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8397 PR rtl-optimization/83985
8398 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8399 REG_CFA_RESTORE insns.
8400 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8401 REG_CFA_RESTORE insns here.
8404 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8405 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8407 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8409 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8410 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8411 (arc_init): Likewise.
8412 (arc_override_options): Likewise.
8413 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8415 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8417 * config/arc/arc.h (TARGET_DBNZ): Define.
8418 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8419 properly set the tune attribute.
8420 (dbnz): Use TARGET_DBNZ guard.
8421 * config/arc/arc.opt (mtune): Add core3 option.
8423 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8425 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8426 recognize new pic like addresses.
8427 (arc_delegitimize_address): Clean up.
8429 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8431 * config/arc/arc-arches.def: Option mrf16 valid for all
8433 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8434 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8435 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8436 * config/arc/arc-tables.opt: Regenerate.
8437 * config/arc/arc.c (arc_conditional_register_usage): Handle
8438 reduced register file case.
8439 (arc_file_start): Set must have build attributes.
8440 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8442 * config/arc/arc.opt (mrf16): Add new option.
8443 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8444 * config/arc/genmultilib.awk: Handle new mrf16 option.
8445 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8446 * config/arc/t-multilib: Regenerate.
8447 * doc/invoke.texi (ARC Options): Document mrf16 option.
8449 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8451 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8452 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8453 (arc_attribute_table): Add 'secure_call' attribute.
8454 (arc_print_operand): Print secure call operand.
8455 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8457 (arc_is_secure_call_p): New function. * config/arc/arc.md
8458 (call_i): Add support for sjli instruction.
8459 (call_value_i): Likewise.
8460 * config/arc/constraints.md (Csc): New constraint.
8462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8463 John Eric Martin <John.Martin@emmicro-us.com>
8465 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8466 * config/arc/arc.c (_arc_jli_section): New struct.
8467 (arc_jli_section): New type.
8468 (rc_jli_sections): New static variable.
8469 (arc_handle_jli_attribute): New function.
8470 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8471 (arc_file_end): New function.
8472 (TARGET_ASM_FILE_END): Define.
8473 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8474 (arc_add_jli_section): New function.
8475 (jli_call_scan): Likewise.
8476 (arc_reorg): Call jli_call_scan.
8477 (arc_output_addsi): Remove 'S' from printing asm operand.
8478 (arc_is_jli_call_p): New function.
8479 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8481 (movhi_insn): Likewise.
8482 (movsi_insn): Likewise.
8483 (movsi_set_cc_insn): Likewise.
8484 (loadqi_update): Likewise.
8485 (load_zeroextendqisi_update): Likewise.
8486 (load_signextendqisi_update): Likewise.
8487 (loadhi_update): Likewise.
8488 (load_zeroextendhisi_update): Likewise.
8489 (load_signextendhisi_update): Likewise.
8490 (loadsi_update): Likewise.
8491 (loadsf_update): Likewise.
8492 (movsicc_insn): Likewise.
8493 (bset_insn): Likewise.
8494 (bxor_insn): Likewise.
8495 (bclr_insn): Likewise.
8496 (bmsk_insn): Likewise.
8497 (bicsi3_insn): Likewise.
8498 (cmpsi_cc_c_insn): Likewise.
8499 (movsi_ne): Likewise.
8500 (movsi_cond_exec): Likewise.
8501 (clrsbsi2): Likewise.
8512 (call_i): Remove 'S' asm letter, add jli instruction.
8513 (call_value_i): Likewise.
8514 * config/arc/arc.op (mjli-always): New option.
8515 * config/arc/constraints.md (Cji): New constraint.
8516 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8518 (subsf3_fpx): Likewise.
8519 (mulsf3_fpx): Likewise.
8520 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8522 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8524 * doc/invoke.texi (ARC): Document mjli-always option.
8526 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8528 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8529 avoid addition with 0 and use incw and decw where possible.
8531 2018-01-26 Richard Biener <rguenther@suse.de>
8533 PR tree-optimization/81082
8534 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8535 association if it requires casting to unsigned.
8536 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8537 from fold_plusminus_mult_expr to catch important cases late when
8538 range info is available.
8540 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8542 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8543 * configure.ac (hidden_linkonce): New test.
8544 * configure: Regenerate.
8545 * config.in: Regenerate.
8547 2018-01-26 Julia Koval <julia.koval@intel.com>
8549 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8550 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8551 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8552 _mm_mask_bitshuffle_epi64_mask): Fix type.
8553 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8554 USI_FTYPE_V4DI_V4DI_USI): Remove.
8555 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8556 __builtin_ia32_vpshufbitqmb256_mask,
8557 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8558 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8559 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8561 2018-01-26 Alan Modra <amodra@gmail.com>
8564 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8565 UNSPEC_VBPERMQ. Sort other unspecs.
8567 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8569 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8571 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8574 * predict.c (drop_profile): Do not push/pop cfun; update also
8576 (handle_missing_profiles): Fix logic looking for zero profiles.
8578 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8581 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8582 on functions with #pragma omp declare simd or functions with simd
8584 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8585 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8586 Remove trailing \n from warning_at calls.
8588 2018-01-25 Tom de Vries <tom@codesourcery.com>
8591 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8592 for neutered workers.
8594 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8597 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8598 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8600 2018-01-24 Jeff Law <law@redhat.com>
8603 * i386.c (get_probe_interval): Move to earlier point.
8604 (ix86_compute_frame_layout): If -fstack-clash-protection and
8605 the frame is larger than the probe interval, then use pushes
8606 to save registers rather than reg->mem moves.
8607 (ix86_expand_prologue): Remove conditional for int_registers_saved
8610 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8613 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8614 min/max for never referenced object.
8616 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8619 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8621 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8622 attributes from DECL_ATTRIBUTES (decl) without affecting
8623 DECL_ATTRIBUTES (current_function_decl).
8624 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8625 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8627 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8629 PR tree-optimization/83979
8630 * fold-const.c (fold_comparison): Use constant_boolean_node
8631 instead of boolean_{true,false}_node.
8633 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8635 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8638 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8640 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8641 Simplify the clause that sets the length attribute.
8642 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8643 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8644 clause that sets the length attribute.
8645 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8647 2018-01-24 Tom de Vries <tom@codesourcery.com>
8650 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8651 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8652 Add strict parameter.
8653 (prevent_branch_around_nothing): Insert dummy insn between branch to
8654 label and label with no ptx insn inbetween.
8655 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8657 2018-01-24 Tom de Vries <tom@codesourcery.com>
8660 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8661 for neutered threads in warp.
8662 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8664 2018-01-24 Richard Biener <rguenther@suse.de>
8666 PR tree-optimization/83176
8667 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8670 2018-01-24 Richard Biener <rguenther@suse.de>
8672 PR tree-optimization/82819
8673 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8674 code generating pluses that are no-ops in the target precision.
8676 2018-01-24 Richard Biener <rguenther@suse.de>
8679 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8681 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8683 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8684 to merge probabilities.
8685 * predict.c (probably_never_executed): Also mark as cold functions
8686 with global 0 profile and guessed local profile.
8687 * profile-count.c (profile_probability::combine_with_count): New
8689 * profile-count.h (profile_probability::operator*,
8690 profile_probability::operator*=, profile_probability::operator/,
8691 profile_probability::operator/=): Reduce precision to adjusted
8692 and set value to guessed on contradictory divisions.
8693 (profile_probability::combine_with_freq): Remove.
8694 (profile_probability::combine_wiht_count): Declare.
8695 (profile_count::force_nonzero):: Set to adjusted.
8696 (profile_count::probability_in):: Set quality to adjusted.
8697 * tree-ssa-tail-merge.c (replace_block_by): Use
8700 2018-01-23 Andrew Waterman <andrew@sifive.com>
8701 Jim Wilson <jimw@sifive.com>
8703 * config/riscv/riscv.c (riscv_stack_boundary): New.
8704 (riscv_option_override): Set riscv_stack_boundary. Handle
8705 riscv_preferred_stack_boundary_arg.
8706 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8707 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8708 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8709 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8710 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8711 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8713 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8716 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8717 of struct ix86_frame.
8718 (ix86_expand_epilogue): Likewise. Add a local variable for
8719 the reg_save_offset field in struct ix86_frame.
8721 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8723 PR tree-optimization/82604
8724 * tree-loop-distribution.c (enum partition_kind): New enum item
8725 PKIND_PARTIAL_MEMSET.
8726 (partition_builtin_p): Support above new enum item.
8727 (generate_code_for_partition): Ditto.
8728 (compute_access_range): Differentiate cases that equality can be
8729 proven at all loops, the innermost loops or no loops.
8730 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8731 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8732 (finalize_partitions, distribute_loop): Don't fuse partition of
8733 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8734 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8737 2018-01-23 Martin Liska <mliska@suse.cz>
8739 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8740 order to ignore the predictor.
8741 (PRED_POLYMORPHIC_CALL): Likewise.
8742 (PRED_RECURSIVE_CALL): Likewise.
8744 2018-01-23 Martin Liska <mliska@suse.cz>
8746 * tree-profile.c (tree_profiling): Print function header to
8747 aware reader which function we are working on.
8748 * value-prof.c (gimple_find_values_to_profile): Do not print
8749 not interesting value histograms.
8751 2018-01-23 Martin Liska <mliska@suse.cz>
8753 * profile-count.h (enum profile_quality): Add
8754 profile_uninitialized as the first value. Do not number values
8755 as they are zero based.
8756 (profile_count::verify): Update sanity check.
8757 (profile_probability::verify): Likewise.
8759 2018-01-23 Nathan Sidwell <nathan@acm.org>
8761 * doc/invoke.texi (ffor-scope): Deprecate.
8763 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8765 PR tree-optimization/83510
8766 * domwalk.c (set_all_edges_as_executable): New function.
8767 (dom_walker::dom_walker): Convert bool param
8768 "skip_unreachable_blocks" to enum reachability. Move setup of
8769 edge flags to set_all_edges_as_executable and only do it when
8770 reachability is REACHABLE_BLOCKS.
8771 * domwalk.h (enum dom_walker::reachability): New enum.
8772 (dom_walker::dom_walker): Convert bool param
8773 "skip_unreachable_blocks" to enum reachability.
8774 (set_all_edges_as_executable): New decl.
8775 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8776 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8778 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8779 but converting true to REACHABLE_BLOCKS.
8780 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8782 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8783 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8784 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8786 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8787 if check_all_array_refs will be called.
8789 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8791 * tree.c (selftest::test_location_wrappers): Add more test
8794 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8796 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8797 (selftest::test_bit_in_range): Likewise.
8799 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8802 * doc/sourcebuild.texi (vect_float): Say that the selector
8803 only describes the situation when -funsafe-math-optimizations is on.
8804 (vect_float_strict): Document.
8806 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8808 PR tree-optimization/83965
8809 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8810 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8811 instead of checking only for a reduction.
8812 (vect_recog_widen_sum_pattern): Likewise.
8814 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8816 * predict.c (probably_never_executed): Only use precise profile info.
8817 (compute_function_frequency): Skip after inlining hack since we now
8818 have quality checking.
8820 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8822 * profile-count.h (profile_probability::very_unlikely,
8823 profile_probability::unlikely, profile_probability::even): Set
8824 precision to guessed.
8826 2018-01-23 Richard Biener <rguenther@suse.de>
8828 PR tree-optimization/83963
8829 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8830 Properly terminate dominator walk when crossing the exit edge not
8831 when visiting its source block.
8833 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8836 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8837 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8839 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8841 PR tree-optimization/83957
8842 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8843 semicolon after for body surrounded by braces.
8845 PR tree-optimization/83081
8846 * profile-count.h (profile_probability::split): New method.
8847 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8848 Use profile_probability::split.
8849 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8850 when splitting a single conditional jump into 2.
8852 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8854 PR tree-optimization/69452
8855 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8858 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8860 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8861 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8862 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8864 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8866 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8867 * config/rl78/rl78.md (movdi): New define_expand.
8868 * config/rl78/rl78.c (rl78_split_movdi): New function.
8870 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8873 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8875 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8876 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8877 128-bit to produce an UNSPEC move to get the double word with the
8878 signbit and then a shift directly to do signbit.
8879 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8880 implementation with a new version that just does either a direct
8881 move or a regular move. Move memory interface to separate insns.
8882 Move insns so they are next to the expander.
8883 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8884 with signbit move. Split big and little endian case.
8885 (signbit<mode>2_dm_mem_le): Likewise.
8886 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8887 (signbit<mode>2_dm2): Likewise.
8889 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8891 * config/rl78/rl78.md (anddi3): New define_expand.
8893 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8895 * config/rl78/rl78.md (umindi3): New define_expand.
8897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8899 * config/rl78/rl78.md (smindi3): New define_expand.
8901 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8903 * config/rl78/rl78.md (smaxdi3): New define_expand.
8905 2018-01-22 Carl Love <cel@us.ibm.com>
8907 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8908 LVX_V1TI): Add macro expansion.
8909 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8910 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8911 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8912 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8913 Change check to determine if the instruction is a byte reversing
8914 entry. Fix typo in comment.
8915 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8916 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8917 Add def_builtin calls for new builtins.
8918 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8919 Add define_insn expansion.
8921 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8923 * config/rl78/rl78.md (umaxdi3): New define_expand.
8925 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8927 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8928 for non-QImode registers.
8930 2018-01-22 Richard Biener <rguenther@suse.de>
8932 PR tree-optimization/83963
8933 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8934 including the loop exit block.
8935 (scop_detection::merge_sese): Likewise.
8936 (scop_detection::add_scop): Do it here instead.
8938 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8940 * doc/sourcebuild.texi (arm_softfloat): Document.
8942 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8945 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8946 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8947 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8949 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8950 David Edelsohn <dje.gcc@gmail.com>
8953 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8954 Change "crset eq" to "crset 2".
8955 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8956 (*call_indirect_aix<mode>_nospec): Likewise.
8957 (*call_value_indirect_aix<mode>_nospec): Likewise.
8958 (*call_indirect_elfv2<mode>_nospec): Likewise.
8959 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8960 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8961 change assembly output from . to $.
8962 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8963 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8964 (*tablejump<mode>_internal1_nospec): Likewise.
8966 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8969 * config/sh/sh_optimize_sett_clrt.cc:
8970 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8972 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8974 PR tree-optimization/83940
8975 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8976 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8977 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8978 use it to pass back the definition type.
8979 (vect_check_store_rhs): Likewise rhs_dt_out.
8980 (vect_build_gather_load_calls): Add a mask_dt argument and use
8981 it instead of a call to vect_is_simple_use.
8982 (vectorizable_store): Update calls to vect_check_load_store_mask
8983 and vect_check_store_rhs. Use the dt returned by the latter instead
8984 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8985 instead of calls to vect_is_simple_use. Pass the scalar rather
8986 than the vector operand to vect_is_simple_use when handling
8987 second and subsequent copies of an rhs value.
8988 (vectorizable_load): Update calls to vect_check_load_store_mask
8989 and vect_build_gather_load_calls. Use the cached mask_dt and
8990 gs_info.offset_dt instead of calls to vect_is_simple_use.
8992 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8995 * tree-emutls.c: Include gimplify.h.
8996 (lower_emutls_2): New function.
8997 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8998 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8999 it before further processing.
9002 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9003 UINTVAL (trueop1) instead of INTVAL (op1).
9005 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9009 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9010 INCOMING_FRAME_SP_OFFSET if not defined.
9011 (scan_trace): Add ENTRY argument. If true and
9012 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9013 emit a note to adjust the CFA offset.
9014 (create_cfi_notes): Adjust scan_trace callers.
9015 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9016 INCOMING_FRAME_SP_OFFSET in the CIE.
9017 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9018 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9020 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9021 * doc/tm.texi: Regenerated.
9023 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9025 PR rtl-optimization/83147
9026 * lra-constraints.c (remove_inheritance_pseudos): Use
9027 lra_substitute_pseudo_within_insn.
9029 2018-01-19 Tom de Vries <tom@codesourcery.com>
9030 Cesar Philippidis <cesar@codesourcery.com>
9033 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9035 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9038 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9039 spaces for function labels.
9041 2018-01-19 Martin Liska <mliska@suse.cz>
9043 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9044 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9045 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9046 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9047 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9048 (PRED_CONST_RETURN): Change from 69 to 65.
9049 (PRED_NULL_RETURN): Change from 91 to 71.
9050 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9051 (PRED_LOOP_GUARD): Change from 66 to 73.
9053 2018-01-19 Martin Liska <mliska@suse.cz>
9055 * predict.c (predict_insn_def): Add new assert.
9056 (struct branch_predictor): Change type to signed integer.
9057 (test_prediction_value_range): Amend test to cover
9059 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9060 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9061 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9062 (PRED_LOOP_IV_COMPARE): Likewise.
9063 * predict.h (PROB_UNINITIALIZED): Define new constant.
9065 2018-01-19 Martin Liska <mliska@suse.cz>
9067 * predict.c (dump_prediction): Add new format for
9068 analyze_brprob.py script which is enabled with -details
9070 * profile-count.h (precise_p): New function.
9072 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9074 PR tree-optimization/83922
9075 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9076 there are no statements that need masking.
9077 (vect_active_double_reduction_p): New function.
9078 (vect_analyze_loop_operations): Use it when handling phis that
9079 are not in the loop header.
9081 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9083 PR tree-optimization/83914
9084 * tree-vect-loop.c (vectorizable_induction): Don't convert
9085 init_expr or apply the peeling adjustment for inductions
9086 that are nested within the vectorized loop.
9088 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9090 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9093 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9097 * function.h (gimplify_parameters): Add gimple_seq * argument.
9098 * function.c: Include gimple.h and options.h.
9099 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9100 for the added local temporaries if needed.
9101 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9102 if there are any parameter cleanups, wrap whole body into a
9103 try/finally with the cleanups.
9105 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9108 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9109 Use GET_MODE_CLASS for scalar floating point.
9111 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9115 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9116 Fix call of call_cgraph_insertion_hooks.
9118 2018-01-18 Martin Sebor <msebor@redhat.com>
9120 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9122 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9125 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9128 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9131 * common.opt: (-ffile-prefix-map): New option.
9132 * opts.c (common_handle_option): Defer it.
9133 * opts-global.c (handle_common_deferred_options): Handle it.
9134 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9135 * file-prefix-map.h: New file.
9136 (remap_debug_filename, add_debug_prefix_map): ...here.
9137 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9138 * final.c (debug_prefix_map, add_debug_prefix_map
9139 remap_debug_filename): Move to...
9140 * file-prefix-map.c: New file.
9141 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9142 generalize, get rid of alloca(), use strrchr() instead of strchr().
9143 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9144 Implement in terms of add_prefix_map().
9145 (remap_macro_filename, remap_debug_filename): Implement in term of
9147 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9148 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9149 * dbxout.c: Include file-prefix-map.h.
9150 * varasm.c: Likewise.
9151 * vmsdbgout.c: Likewise.
9152 * xcoffout.c: Likewise.
9153 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9154 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9155 * doc/invoke.texi (-ffile-prefix-map): Document.
9156 (-fdebug-prefix-map): Update description.
9158 2018-01-18 Martin Liska <mliska@suse.cz>
9160 * config/i386/i386.c (indirect_thunk_name): Document that also
9162 (output_indirect_thunk): Document why both instructions
9163 (pause and lfence) are generated.
9165 2018-01-18 Richard Biener <rguenther@suse.de>
9167 PR tree-optimization/83887
9168 * graphite-scop-detection.c
9169 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9170 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9171 (scop_detection::merge_sese): Re-implement with a flood-fill
9172 algorithm that properly finds a SESE region if it exists.
9174 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9177 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9178 pointer_diff optimizations use view_convert instead of convert.
9180 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9182 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9183 Generate different code for -mno-speculate-indirect-jumps.
9184 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9185 (*call_indirect_aix<mode>): Disable for
9186 -mno-speculate-indirect-jumps.
9187 (*call_indirect_aix<mode>_nospec): New define_insn.
9188 (*call_value_indirect_aix<mode>): Disable for
9189 -mno-speculate-indirect-jumps.
9190 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9191 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9192 -mno-speculate-indirect-jumps.
9193 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9195 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9197 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9198 long double type, set the flags for noting the default long double
9199 type, even if we don't pass or return a long double type.
9201 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9204 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9207 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9209 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9210 support for merge[hl].
9211 (fold_mergehl_helper): New helper function.
9212 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9213 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9214 (altivec_vmrglw_direct): Add xxmrglw insn.
9216 2018-01-17 Andrew Waterman <andrew@sifive.com>
9218 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9219 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9221 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9224 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9225 call the lto_location_cache before reading the
9226 DECL_SOURCE_LOCATION of the types.
9228 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9229 Richard Sandiford <richard.sandiford@linaro.org>
9231 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9232 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9233 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9234 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9235 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9237 * config/aarch64/constraints.md (aarch64_movti_operand):
9239 * config/aarch64/predicates.md (Uti): Add new constraint.
9241 2018-01-17 Carl Love <cel@us.ibm.com>
9243 * config/rs6000/vsx.md (define_expand xl_len_r,
9244 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9245 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9247 (define_expand, define_insn): Move the shift left from the
9248 define_insn to the define_expand for lxvl and stxvl instructions.
9249 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9250 and XL_LEN_R definitions to PURE.
9252 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9254 * config/i386/i386.c (indirect_thunk_name): Declare regno
9255 as unsigned int. Compare regno with INVALID_REGNUM.
9256 (output_indirect_thunk): Ditto.
9257 (output_indirect_thunk_function): Ditto.
9258 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9259 in the call to output_indirect_thunk_function.
9261 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9264 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9265 rather than the size of inner_type to determine the stack slot size
9266 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9268 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9271 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9274 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9276 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9277 endian Linux systems to optionally enable multilibs for selecting
9278 the long double type if the user configured an explicit type.
9279 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9280 have no long double multilibs if not defined.
9281 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9282 warn if the user used -mabi={ieee,ibm}longdouble and we built
9283 multilibs for long double.
9284 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9285 appropriate multilib option.
9286 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9288 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9289 for building long double multilibs.
9290 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9292 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9294 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9297 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9299 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9302 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9305 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9308 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9310 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9311 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9313 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9315 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9316 different rtl trees depending on TARGET_64BIT.
9317 (rs6000_gen_lvx): Likewise.
9319 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9321 * config/visium/visium.md (nop): Tweak comment.
9322 (hazard_nop): Likewise.
9324 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9326 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9327 -mspeculate-indirect-jumps.
9328 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9329 for -mno-speculate-indirect-jumps.
9330 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9331 (*call_value_indirect_elfv2<mode>): Disable for
9332 -mno-speculate-indirect-jumps.
9333 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9334 (indirect_jump): Emit different RTL for
9335 -mno-speculate-indirect-jumps.
9336 (*indirect_jump<mode>): Disable for
9337 -mno-speculate-indirect-jumps.
9338 (*indirect_jump<mode>_nospec): New define_insn.
9339 (tablejump): Emit different RTL for
9340 -mno-speculate-indirect-jumps.
9341 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9342 (tablejumpsi_nospec): New define_expand.
9343 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9344 (tablejumpdi_nospec): New define_expand.
9345 (*tablejump<mode>_internal1): Disable for
9346 -mno-speculate-indirect-jumps.
9347 (*tablejump<mode>_internal1_nospec): New define_insn.
9348 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9351 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9353 * caller-save.c (insert_save): Drop unnecessary parameter. All
9356 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9357 Richard Biener <rguenth@suse.de>
9360 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9361 return early, inline manually is_gimple_sizepos. Make sure if we
9362 call gimplify_expr we don't end up with a gimple constant.
9363 * tree.c (variably_modified_type_p): Don't return true for
9364 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9365 * gimplify.h (is_gimple_sizepos): Remove.
9367 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9369 PR tree-optimization/83857
9370 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9371 vectorizable_live_operation for pure SLP statements.
9372 (vectorizable_live_operation): Handle PHIs.
9374 2018-01-16 Richard Biener <rguenther@suse.de>
9376 PR tree-optimization/83867
9377 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9378 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9380 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9383 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9384 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9385 If off is not INTEGER_CST, issue a may not be aligned warning
9386 rather than isn't aligned. Use isn%'t rather than isn't.
9387 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9389 <case MULT_EXPR>: Improve the case when bottom and one of the
9390 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9391 operand, in that case check if the other operand is multiple of
9392 bottom divided by the INTEGER_CST operand.
9394 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9397 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9398 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9399 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9400 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9401 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9402 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9403 (pa_function_arg_size): New function.
9405 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9407 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9408 in a separate statement.
9410 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9412 PR tree-optimization/83847
9413 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9414 group gathers and scatters.
9416 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9418 PR rtl-optimization/86620
9419 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9421 PR rtl-optimization/83213
9422 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9423 to last if both are JUMP_INSNs.
9425 PR tree-optimization/83843
9426 * gimple-ssa-store-merging.c
9427 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9428 store_immediate_info for bswap/nop orig_stores.
9430 2018-01-15 Andrew Waterman <andrew@sifive.com>
9432 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9434 <UDIV>: Increase cost if !TARGET_DIV.
9436 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9438 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9439 (define_attr "cr_logical_3op"): New.
9440 (cceq_ior_compare): Adjust.
9441 (cceq_ior_compare_complement): Adjust.
9442 (*cceq_rev_compare): Adjust.
9443 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9444 (is_cracked_insn): Adjust.
9445 (insn_must_be_first_in_group): Adjust.
9446 * config/rs6000/40x.md: Adjust.
9447 * config/rs6000/440.md: Adjust.
9448 * config/rs6000/476.md: Adjust.
9449 * config/rs6000/601.md: Adjust.
9450 * config/rs6000/603.md: Adjust.
9451 * config/rs6000/6xx.md: Adjust.
9452 * config/rs6000/7450.md: Adjust.
9453 * config/rs6000/7xx.md: Adjust.
9454 * config/rs6000/8540.md: Adjust.
9455 * config/rs6000/cell.md: Adjust.
9456 * config/rs6000/e300c2c3.md: Adjust.
9457 * config/rs6000/e500mc.md: Adjust.
9458 * config/rs6000/e500mc64.md: Adjust.
9459 * config/rs6000/e5500.md: Adjust.
9460 * config/rs6000/e6500.md: Adjust.
9461 * config/rs6000/mpc.md: Adjust.
9462 * config/rs6000/power4.md: Adjust.
9463 * config/rs6000/power5.md: Adjust.
9464 * config/rs6000/power6.md: Adjust.
9465 * config/rs6000/power7.md: Adjust.
9466 * config/rs6000/power8.md: Adjust.
9467 * config/rs6000/power9.md: Adjust.
9468 * config/rs6000/rs64.md: Adjust.
9469 * config/rs6000/titan.md: Adjust.
9471 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9473 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9474 ix86_indirect_branch_register logic.
9476 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9478 * config/i386/constraints.md (Bs): Update
9479 ix86_indirect_branch_register check. Don't check
9480 ix86_indirect_branch_register with GOT_memory_operand.
9482 * config/i386/predicates.md (GOT_memory_operand): Don't check
9483 ix86_indirect_branch_register here.
9484 (GOT32_symbol_operand): Likewise.
9486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9488 * config/i386/predicates.md (constant_call_address_operand):
9489 Rewrite ix86_indirect_branch_register logic.
9490 (sibcall_insn_operand): Likewise.
9492 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9494 * config/i386/constraints.md (Bs): Replace
9495 ix86_indirect_branch_thunk_register with
9496 ix86_indirect_branch_register.
9498 * config/i386/i386.md (indirect_jump): Likewise.
9499 (tablejump): Likewise.
9500 (*sibcall_memory): Likewise.
9501 (*sibcall_value_memory): Likewise.
9502 Peepholes of indirect call and jump via memory: Likewise.
9503 * config/i386/i386.opt: Likewise.
9504 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9505 (GOT_memory_operand): Likewise.
9506 (call_insn_operand): Likewise.
9507 (sibcall_insn_operand): Likewise.
9508 (GOT32_symbol_operand): Likewise.
9510 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9513 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9514 type rather than type addr's type points to.
9515 (expand_omp_atomic_mutex): Likewise.
9516 (expand_omp_atomic): Likewise.
9518 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9521 * config/i386/i386.c (output_indirect_thunk_function): Use
9522 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9523 for __x86_return_thunk.
9525 2018-01-15 Richard Biener <rguenther@suse.de>
9528 * expmed.c (extract_bit_field_1): Fix typo.
9530 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9533 * config/arm/iterators.md (VF): New mode iterator.
9534 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9535 Remove integer-related logic from pattern.
9536 (neon_vabd<mode>_3): Likewise.
9538 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9541 * common.opt (fstrict-overflow): No longer an alias.
9542 (fwrapv-pointer): New option.
9543 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9544 also for pointer types based on flag_wrapv_pointer.
9545 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9546 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9547 opts->x_flag_wrapv got set.
9548 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9549 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9550 POINTER_TYPE_OVERFLOW_UNDEFINED.
9551 * match.pd: Likewise in address comparison pattern.
9552 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9554 2018-01-15 Richard Biener <rguenther@suse.de>
9557 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9558 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9559 Reset type names to their identifier if their TYPE_DECL doesn't
9560 have linkage (and thus is used for ODR and devirt).
9561 (save_debug_info_for_decl): Remove.
9562 (save_debug_info_for_type): Likewise.
9563 (add_tree_to_fld_list): Adjust.
9564 * tree-pretty-print.c (dump_generic_node): Make dumping of
9565 type names more robust.
9567 2018-01-15 Richard Biener <rguenther@suse.de>
9569 * BASE-VER: Bump to 8.0.1.
9571 2018-01-14 Martin Sebor <msebor@redhat.com>
9574 * builtins.c (check_access): Avoid warning when the no-warning bit
9577 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9579 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9580 * ira-color (allocno_hard_regs_compare): Likewise.
9582 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9585 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9586 Use .pushsection/.popsection.
9588 2018-01-14 Martin Sebor <msebor@redhat.com>
9591 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9593 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9595 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9596 entry from extra_headers.
9597 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9598 extra_headers, make the list bitwise identical to the i?86-*-* one.
9600 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9602 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9603 -mcmodel=large with -mindirect-branch=thunk,
9604 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9605 -mfunction-return=thunk-extern.
9606 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9607 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9608 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9610 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9612 * config/i386/i386.c (print_reg): Print the name of the full
9613 integer register without '%'.
9614 (ix86_print_operand): Handle 'V'.
9615 * doc/extend.texi: Document 'V' modifier.
9617 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9619 * config/i386/constraints.md (Bs): Disallow memory operand for
9620 -mindirect-branch-register.
9622 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9623 (GOT_memory_operand): Likewise.
9624 (call_insn_operand): Likewise.
9625 (sibcall_insn_operand): Likewise.
9626 (GOT32_symbol_operand): Likewise.
9627 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9628 for -mindirect-branch-register.
9629 (tablejump): Likewise.
9630 (*sibcall_memory): Likewise.
9631 (*sibcall_value_memory): Likewise.
9632 Disallow peepholes of indirect call and jump via memory for
9633 -mindirect-branch-register.
9634 (*call_pop): Replace m with Bw.
9635 (*call_value_pop): Likewise.
9636 (*sibcall_pop_memory): Replace m with Bs.
9637 * config/i386/i386.opt (mindirect-branch-register): New option.
9638 * doc/invoke.texi: Document -mindirect-branch-register option.
9640 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9642 * config/i386/i386-protos.h (ix86_output_function_return): New.
9643 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9644 set function_return_type.
9645 (indirect_thunk_name): Add ret_p to indicate thunk for function
9647 (output_indirect_thunk_function): Pass false to
9648 indirect_thunk_name.
9649 (ix86_output_indirect_branch_via_reg): Likewise.
9650 (ix86_output_indirect_branch_via_push): Likewise.
9651 (output_indirect_thunk_function): Create alias for function
9652 return thunk if regno < 0.
9653 (ix86_output_function_return): New function.
9654 (ix86_handle_fndecl_attribute): Handle function_return.
9655 (ix86_attribute_table): Add function_return.
9656 * config/i386/i386.h (machine_function): Add
9657 function_return_type.
9658 * config/i386/i386.md (simple_return_internal): Use
9659 ix86_output_function_return.
9660 (simple_return_internal_long): Likewise.
9661 * config/i386/i386.opt (mfunction-return=): New option.
9662 (indirect_branch): Mention -mfunction-return=.
9663 * doc/extend.texi: Document function_return function attribute.
9664 * doc/invoke.texi: Document -mfunction-return= option.
9666 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9668 * config/i386/i386-opts.h (indirect_branch): New.
9669 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9670 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9671 with local indirect jump when converting indirect call and jump.
9672 (ix86_set_indirect_branch_type): New.
9673 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9674 (indirectlabelno): New.
9675 (indirect_thunk_needed): Likewise.
9676 (indirect_thunk_bnd_needed): Likewise.
9677 (indirect_thunks_used): Likewise.
9678 (indirect_thunks_bnd_used): Likewise.
9679 (INDIRECT_LABEL): Likewise.
9680 (indirect_thunk_name): Likewise.
9681 (output_indirect_thunk): Likewise.
9682 (output_indirect_thunk_function): Likewise.
9683 (ix86_output_indirect_branch_via_reg): Likewise.
9684 (ix86_output_indirect_branch_via_push): Likewise.
9685 (ix86_output_indirect_branch): Likewise.
9686 (ix86_output_indirect_jmp): Likewise.
9687 (ix86_code_end): Call output_indirect_thunk_function if needed.
9688 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9690 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9691 (ix86_attribute_table): Add indirect_branch.
9692 * config/i386/i386.h (machine_function): Add indirect_branch_type
9693 and has_local_indirect_jump.
9694 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9696 (tablejump): Likewise.
9697 (*indirect_jump): Use ix86_output_indirect_jmp.
9698 (*tablejump_1): Likewise.
9699 (simple_return_indirect_internal): Likewise.
9700 * config/i386/i386.opt (mindirect-branch=): New option.
9701 (indirect_branch): New.
9704 (thunk-inline): Likewise.
9705 (thunk-extern): Likewise.
9706 * doc/extend.texi: Document indirect_branch function attribute.
9707 * doc/invoke.texi: Document -mindirect-branch= option.
9709 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9712 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9714 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9716 * ipa-inline.c (want_inline_small_function_p): Return false if
9717 inlining has already failed with CIF_FINAL_ERROR.
9718 (update_caller_keys): Call want_inline_small_function_p before
9720 (update_callee_keys): Likewise.
9722 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9724 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9726 (rs6000_quadword_masked_address_p): Likewise.
9727 (quad_aligned_load_p): Likewise.
9728 (quad_aligned_store_p): Likewise.
9729 (const_load_sequence_p): Add comment to describe the outer-most loop.
9730 (mimic_memory_attributes_and_flags): New function.
9731 (rs6000_gen_stvx): Likewise.
9732 (replace_swapped_aligned_store): Likewise.
9733 (rs6000_gen_lvx): Likewise.
9734 (replace_swapped_aligned_load): Likewise.
9735 (replace_swapped_load_constant): Capitalize argument name in
9736 comment describing this function.
9737 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9738 and stores that access quad-word aligned addresses and replace
9739 with stvx or lvx instructions when appropriate.
9740 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9741 New function prototype.
9742 (rs6000_quadword_masked_address_p): Likewise.
9743 (rs6000_gen_lvx): Likewise.
9744 (rs6000_gen_stvx): Likewise.
9745 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9746 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9747 when memory address is aligned.
9748 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9749 this split to select lvx instruction when memory address is aligned.
9750 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9751 instruction when memory address is aligned.
9752 (*vsx_le_perm_load_v16qi): Likewise.
9753 (four unnamed splitters): Modify to select the stvx instruction
9754 when memory is aligned.
9756 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9758 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9759 which appears in the queue multiple times.
9761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9762 Alan Hayward <alan.hayward@arm.com>
9763 David Sherwood <david.sherwood@arm.com>
9765 * tree-vectorizer.h (vec_lower_bound): New structure.
9766 (_loop_vec_info): Add check_nonzero and lower_bounds.
9767 (LOOP_VINFO_CHECK_NONZERO): New macro.
9768 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9769 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9770 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9771 fields. Make seg_len the distance travelled, not including the
9773 (dr_direction_indicator): Declare.
9774 (dr_zero_step_indicator): Likewise.
9775 (dr_known_forward_stride_p): Likewise.
9776 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9778 (runtime_alias_check_p): Allow runtime alias checks with
9780 (operator ==): Compare access_size and align.
9781 (prune_runtime_alias_test_list): Rework for new distinction between
9782 the access_size and seg_len.
9783 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9785 (get_segment_min_max): New function.
9786 (create_intersect_range_checks): Use it.
9787 (dr_step_indicator): New function.
9788 (dr_direction_indicator): Likewise.
9789 (dr_zero_step_indicator): Likewise.
9790 (dr_known_forward_stride_p): Likewise.
9791 * tree-loop-distribution.c (data_ref_segment_size): Return
9792 DR_STEP * (niters - 1).
9793 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9795 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9796 (vect_preserves_scalar_order_p): New function, split out from...
9797 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9798 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9799 (vect_vfa_access_size): New function.
9800 (vect_vfa_align): Likewise.
9801 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9802 (dump_lower_bound): New function.
9803 (vect_check_lower_bound): Likewise.
9804 (vect_small_gap_p): Likewise.
9805 (vectorizable_with_step_bound_p): Likewise.
9806 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9807 depencies if the vectorization factor is 1. Convert the checks
9808 for nonzero steps into checks on the bounds of DR_STEP. Try using
9809 a bunds check for variable steps if the minimum required step is
9810 relatively small. Update calls to the dr_with_seg_len
9811 constructor and to vect_compile_time_alias.
9812 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9814 (vect_loop_versioning): Call it.
9815 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9817 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9820 Alan Hayward <alan.hayward@arm.com>
9821 David Sherwood <david.sherwood@arm.com>
9823 * doc/sourcebuild.texi (vect_scatter_store): Document.
9824 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9826 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9828 * genopinit.c (main): Add supports_vec_scatter_store and
9829 supports_vec_scatter_store_cached to target_optabs.
9830 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9831 IFN_MASK_SCATTER_STORE.
9832 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9834 * internal-fn.h (internal_store_fn_p): Declare.
9835 (internal_fn_stored_value_index): Likewise.
9836 * internal-fn.c (scatter_store_direct): New macro.
9837 (expand_scatter_store_optab_fn): New function.
9838 (direct_scatter_store_optab_supported_p): New macro.
9839 (internal_store_fn_p): New function.
9840 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9841 IFN_MASK_SCATTER_STORE.
9842 (internal_fn_mask_index): Likewise.
9843 (internal_fn_stored_value_index): New function.
9844 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9846 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9847 * optabs-query.c (supports_vec_scatter_store_p): New function.
9848 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9849 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9850 true for scatter stores.
9851 (vect_gather_scatter_fn_p): Handle scatter stores too.
9852 (vect_check_gather_scatter): Consider using scatter stores if
9853 supports_vec_scatter_store_p.
9854 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9856 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9857 internal_fn_stored_value_index.
9858 (check_load_store_masking): Handle scatter stores too.
9859 (vect_get_store_rhs): Make public.
9860 (vectorizable_call): Use internal_store_fn_p.
9861 (vectorizable_store): Handle scatter store internal functions.
9862 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9863 when deciding whether the end of the group has been reached.
9864 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9865 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9866 (mask_scatter_store<mode>): New insns.
9868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9869 Alan Hayward <alan.hayward@arm.com>
9870 David Sherwood <david.sherwood@arm.com>
9872 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9873 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9874 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9876 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9877 Use vect_truncate_gather_scatter_offset if we can't treat the
9878 operation as a normal gather load or scatter store.
9879 (get_group_load_store_type): Take the gather_scatter_info
9880 as argument. Try using a gather load or scatter store for
9881 single-element groups.
9882 (get_load_store_type): Update calls to get_group_load_store_type
9883 and vect_use_strided_gather_scatters_p.
9885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9886 Alan Hayward <alan.hayward@arm.com>
9887 David Sherwood <david.sherwood@arm.com>
9889 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9890 optional tree argument.
9891 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9893 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9894 but continue to use the current value as a fallback.
9895 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9896 to compare the updates.
9897 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9898 (get_load_store_type): Use it when handling a strided access.
9899 (vect_get_strided_load_store_ops): New function.
9900 (vect_get_data_ptr_increment): Likewise.
9901 (vectorizable_load): Handle strided gather loads. Always pass
9902 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9905 Alan Hayward <alan.hayward@arm.com>
9906 David Sherwood <david.sherwood@arm.com>
9908 * doc/md.texi (gather_load@var{m}): Document.
9909 (mask_gather_load@var{m}): Likewise.
9910 * genopinit.c (main): Add supports_vec_gather_load and
9911 supports_vec_gather_load_cached to target_optabs.
9912 * optabs-tree.c (init_tree_optimization_optabs): Use
9913 ggc_cleared_alloc to allocate target_optabs.
9914 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9915 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9917 * internal-fn.h (internal_load_fn_p): Declare.
9918 (internal_gather_scatter_fn_p): Likewise.
9919 (internal_fn_mask_index): Likewise.
9920 (internal_gather_scatter_fn_supported_p): Likewise.
9921 * internal-fn.c (gather_load_direct): New macro.
9922 (expand_gather_load_optab_fn): New function.
9923 (direct_gather_load_optab_supported_p): New macro.
9924 (direct_internal_fn_optab): New function.
9925 (internal_load_fn_p): Likewise.
9926 (internal_gather_scatter_fn_p): Likewise.
9927 (internal_fn_mask_index): Likewise.
9928 (internal_gather_scatter_fn_supported_p): Likewise.
9929 * optabs-query.c (supports_at_least_one_mode_p): New function.
9930 (supports_vec_gather_load_p): Likewise.
9931 * optabs-query.h (supports_vec_gather_load_p): Declare.
9932 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9933 and memory_type field.
9934 (NUM_PATTERNS): Bump to 15.
9935 * tree-vect-data-refs.c: Include internal-fn.h.
9936 (vect_gather_scatter_fn_p): New function.
9937 (vect_describe_gather_scatter_call): Likewise.
9938 (vect_check_gather_scatter): Try using internal functions for
9939 gather loads. Recognize existing calls to a gather load function.
9940 (vect_analyze_data_refs): Consider using gather loads if
9941 supports_vec_gather_load_p.
9942 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9943 (vect_get_gather_scatter_offset_type): Likewise.
9944 (vect_convert_mask_for_vectype): Likewise.
9945 (vect_add_conversion_to_patterm): Likewise.
9946 (vect_try_gather_scatter_pattern): Likewise.
9947 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9948 (vect_vect_recog_func_ptrs): Add it.
9949 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9950 internal_fn_mask_index and internal_gather_scatter_fn_p.
9951 (check_load_store_masking): Take the gather_scatter_info as an
9952 argument and handle gather loads.
9953 (vect_get_gather_scatter_ops): New function.
9954 (vectorizable_call): Check internal_load_fn_p.
9955 (vectorizable_load): Likewise. Handle gather load internal
9957 (vectorizable_store): Update call to check_load_store_masking.
9958 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9959 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9960 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9961 (aarch64_gather_scale_operand_d): New predicates.
9962 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9963 (mask_gather_load<mode>): New insns.
9965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9966 Alan Hayward <alan.hayward@arm.com>
9967 David Sherwood <david.sherwood@arm.com>
9969 * optabs.def (fold_left_plus_optab): New optab.
9970 * doc/md.texi (fold_left_plus_@var{m}): Document.
9971 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9972 * internal-fn.c (fold_left_direct): Define.
9973 (expand_fold_left_optab_fn): Likewise.
9974 (direct_fold_left_optab_supported_p): Likewise.
9975 * fold-const-call.c (fold_const_fold_left): New function.
9976 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9977 * tree-parloops.c (valid_reduction_p): New function.
9978 (gather_scalar_reductions): Use it.
9979 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9980 (vect_finish_replace_stmt): Declare.
9981 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9982 (needs_fold_left_reduction_p): New function, split out from...
9983 (vect_is_simple_reduction): ...here. Accept reductions that
9984 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9985 (vect_force_simple_reduction): Also store the reduction type in
9986 the assignment's STMT_VINFO_REDUC_TYPE.
9987 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9988 (merge_with_identity): New function.
9989 (vect_expand_fold_left): Likewise.
9990 (vectorize_fold_left_reduction): Likewise.
9991 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9992 scalar phi in place for it. Check for target support and reject
9993 cases that would reassociate the operation. Defer the transform
9994 phase to vectorize_fold_left_reduction.
9995 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9996 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9997 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10001 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10002 call to ifc_temp_var.
10004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10005 Alan Hayward <alan.hayward@arm.com>
10006 David Sherwood <david.sherwood@arm.com>
10008 * target.def (legitimize_address_displacement): Take the original
10009 offset as a poly_int.
10010 * targhooks.h (default_legitimize_address_displacement): Update
10012 * targhooks.c (default_legitimize_address_displacement): Likewise.
10013 * doc/tm.texi: Regenerate.
10014 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10015 as an argument, moving assert of ad->disp == ad->disp_term to...
10016 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10017 Try calling targetm.legitimize_address_displacement before expanding
10018 the address rather than afterwards, and adjust for the new interface.
10019 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10020 Match the new hook interface. Handle SVE addresses.
10021 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10022 new hook interface.
10024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10026 * Makefile.in (OBJS): Add early-remat.o.
10027 * target.def (select_early_remat_modes): New hook.
10028 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10029 * doc/tm.texi: Regenerate.
10030 * targhooks.h (default_select_early_remat_modes): Declare.
10031 * targhooks.c (default_select_early_remat_modes): New function.
10032 * timevar.def (TV_EARLY_REMAT): New timevar.
10033 * passes.def (pass_early_remat): New pass.
10034 * tree-pass.h (make_pass_early_remat): Declare.
10035 * early-remat.c: New file.
10036 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10038 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10041 Alan Hayward <alan.hayward@arm.com>
10042 David Sherwood <david.sherwood@arm.com>
10044 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10045 vfm1 with a bound_epilog parameter.
10046 (vect_do_peeling): Update calls accordingly, and move the prologue
10047 call earlier in the function. Treat the base bound_epilog as 0 for
10048 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10049 this base when peeling for gaps.
10050 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10051 with fully-masked loops.
10052 (vect_estimate_min_profitable_iters): Handle the single peeled
10053 iteration in that case.
10055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10056 Alan Hayward <alan.hayward@arm.com>
10057 David Sherwood <david.sherwood@arm.com>
10059 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10060 single-element interleaving even if the size is not a power of 2.
10061 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10062 accesses for single-element interleaving if the group size is
10065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10066 Alan Hayward <alan.hayward@arm.com>
10067 David Sherwood <david.sherwood@arm.com>
10069 * doc/md.texi (fold_extract_last_@var{m}): Document.
10070 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10071 * optabs.def (fold_extract_last_optab): New optab.
10072 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10073 * internal-fn.c (fold_extract_direct): New macro.
10074 (expand_fold_extract_optab_fn): Likewise.
10075 (direct_fold_extract_optab_supported_p): Likewise.
10076 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10077 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10078 EXTRACT_LAST_REDUCTION.
10079 (get_initial_def_for_reduction): Do not create an initial vector
10080 for EXTRACT_LAST_REDUCTION reductions.
10081 (vectorizable_reduction): Leave the scalar phi in place for
10082 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10083 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10084 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10085 transform phase to vectorizable_condition.
10086 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10088 (vect_finish_stmt_generation): ...here.
10089 (vect_finish_replace_stmt): New function.
10090 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10091 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10093 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10096 Alan Hayward <alan.hayward@arm.com>
10097 David Sherwood <david.sherwood@arm.com>
10099 * doc/md.texi (extract_last_@var{m}): Document.
10100 * optabs.def (extract_last_optab): New optab.
10101 * internal-fn.def (EXTRACT_LAST): New internal function.
10102 * internal-fn.c (cond_unary_direct): New macro.
10103 (expand_cond_unary_optab_fn): Likewise.
10104 (direct_cond_unary_optab_supported_p): Likewise.
10105 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10106 loops using EXTRACT_LAST.
10107 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10108 (extract_last_<mode>): ...this optab.
10109 (vec_extract<mode><Vel>): Update accordingly.
10111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10112 Alan Hayward <alan.hayward@arm.com>
10113 David Sherwood <david.sherwood@arm.com>
10115 * target.def (empty_mask_is_expensive): New hook.
10116 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10117 * doc/tm.texi: Regenerate.
10118 * targhooks.h (default_empty_mask_is_expensive): Declare.
10119 * targhooks.c (default_empty_mask_is_expensive): New function.
10120 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10121 if the target says that empty masks are expensive.
10122 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10124 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10127 Alan Hayward <alan.hayward@arm.com>
10128 David Sherwood <david.sherwood@arm.com>
10130 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10131 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10132 (vect_use_loop_mask_for_alignment_p): New function.
10133 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10134 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10135 niters_skip argument. Make sure that the first niters_skip elements
10136 of the first iteration are inactive.
10137 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10138 Update call to vect_set_loop_masks_directly.
10139 (get_misalign_in_elems): New function, split out from...
10140 (vect_gen_prolog_loop_niters): ...here.
10141 (vect_update_init_of_dr): Take a code argument that specifies whether
10142 the adjustment should be added or subtracted.
10143 (vect_update_init_of_drs): Likewise.
10144 (vect_prepare_for_masked_peels): New function.
10145 (vect_do_peeling): Skip prologue peeling if we're using a mask
10146 instead. Update call to vect_update_inits_of_drs.
10147 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10149 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10150 alignment. Do not include the number of peeled iterations in
10151 the minimum threshold in that case.
10152 (vectorizable_induction): Adjust the start value down by
10153 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10154 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10155 Take the number of skipped iterations into account when calculating
10157 * tree-vect-stmts.c (vect_gen_while_not): New function.
10159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10160 Alan Hayward <alan.hayward@arm.com>
10161 David Sherwood <david.sherwood@arm.com>
10163 * doc/sourcebuild.texi (vect_fully_masked): Document.
10164 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10165 default value to 0.
10166 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10168 (vect_analyze_loop_2): ...here. Don't check the vectorization
10169 factor against the number of loop iterations if the loop is
10172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10173 Alan Hayward <alan.hayward@arm.com>
10174 David Sherwood <david.sherwood@arm.com>
10176 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10177 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10178 (dump_groups): Update accordingly.
10179 (iv_use::mem_type): New member variable.
10180 (address_p): New function.
10181 (record_use): Add a mem_type argument and initialize the new
10183 (record_group_use): Add a mem_type argument. Use address_p.
10184 Remove obsolete null checks of base_object. Update call to record_use.
10185 (find_interesting_uses_op): Update call to record_group_use.
10186 (find_interesting_uses_cond): Likewise.
10187 (find_interesting_uses_address): Likewise.
10188 (get_mem_type_for_internal_fn): New function.
10189 (find_address_like_use): Likewise.
10190 (find_interesting_uses_stmt): Try find_address_like_use before
10191 calling find_interesting_uses_op.
10192 (addr_offset_valid_p): Use the iv mem_type field as the type
10193 of the addressed memory.
10194 (add_autoinc_candidates): Likewise.
10195 (get_address_cost): Likewise.
10196 (split_small_address_groups_p): Use address_p.
10197 (split_address_groups): Likewise.
10198 (add_iv_candidate_for_use): Likewise.
10199 (autoinc_possible_for_pair): Likewise.
10200 (rewrite_groups): Likewise.
10201 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10202 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10203 (get_alias_ptr_type_for_ptr_address): New function.
10204 (rewrite_use_address): Rewrite address uses in calls that were
10205 identified by find_address_like_use.
10207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10208 Alan Hayward <alan.hayward@arm.com>
10209 David Sherwood <david.sherwood@arm.com>
10211 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10213 * gimple-expr.h (is_gimple_addressable: Likewise.
10214 * gimple-expr.c (is_gimple_address): Likewise.
10215 * internal-fn.c (expand_call_mem_ref): New function.
10216 (expand_mask_load_optab_fn): Use it.
10217 (expand_mask_store_optab_fn): Likewise.
10219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10220 Alan Hayward <alan.hayward@arm.com>
10221 David Sherwood <david.sherwood@arm.com>
10223 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10224 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10225 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10226 (cond_umax@var{mode}): Document.
10227 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10228 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10229 (cond_umin_optab, cond_umax_optab): New optabs.
10230 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10231 (COND_IOR, COND_XOR): New internal functions.
10232 * internal-fn.h (get_conditional_internal_fn): Declare.
10233 * internal-fn.c (cond_binary_direct): New macro.
10234 (expand_cond_binary_optab_fn): Likewise.
10235 (direct_cond_binary_optab_supported_p): Likewise.
10236 (get_conditional_internal_fn): New function.
10237 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10238 Cope with reduction statements that are vectorized as calls rather
10240 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10241 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10242 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10243 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10244 (UNSPEC_COND_EOR): New unspecs.
10245 (optab): Add mappings for them.
10246 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10247 (sve_int_op, sve_fp_op): New int attributes.
10249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10250 Alan Hayward <alan.hayward@arm.com>
10251 David Sherwood <david.sherwood@arm.com>
10253 * optabs.def (while_ult_optab): New optab.
10254 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10255 * internal-fn.def (WHILE_ULT): New internal function.
10256 * internal-fn.h (direct_internal_fn_supported_p): New override
10257 that takes two types as argument.
10258 * internal-fn.c (while_direct): New macro.
10259 (expand_while_optab_fn): New function.
10260 (convert_optab_supported_p): Likewise.
10261 (direct_while_optab_supported_p): New macro.
10262 * wide-int.h (wi::udiv_ceil): New function.
10263 * tree-vectorizer.h (rgroup_masks): New structure.
10264 (vec_loop_masks): New typedef.
10265 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10266 and fully_masked_p.
10267 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10268 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10269 (vect_max_vf): New function.
10270 (slpeel_make_loop_iterate_ntimes): Delete.
10271 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10272 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10273 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10274 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10275 internal-fn.h, stor-layout.h and optabs-query.h.
10276 (vect_set_loop_mask): New function.
10277 (add_preheader_seq): Likewise.
10278 (add_header_seq): Likewise.
10279 (interleave_supported_p): Likewise.
10280 (vect_maybe_permute_loop_masks): Likewise.
10281 (vect_set_loop_masks_directly): Likewise.
10282 (vect_set_loop_condition_masked): Likewise.
10283 (vect_set_loop_condition_unmasked): New function, split out from
10284 slpeel_make_loop_iterate_ntimes.
10285 (slpeel_make_loop_iterate_ntimes): Rename to..
10286 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10287 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10288 (vect_do_peeling): Update call accordingly.
10289 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10291 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10292 mask_compare_type, can_fully_mask_p and fully_masked_p.
10293 (release_vec_loop_masks): New function.
10294 (_loop_vec_info): Use it to free the loop masks.
10295 (can_produce_all_loop_masks_p): New function.
10296 (vect_get_max_nscalars_per_iter): Likewise.
10297 (vect_verify_full_masking): Likewise.
10298 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10299 retries, and free the mask rgroups before retrying. Check loop-wide
10300 reasons for disallowing fully-masked loops. Make the final decision
10301 about whether use a fully-masked loop or not.
10302 (vect_estimate_min_profitable_iters): Do not assume that peeling
10303 for the number of iterations will be needed for fully-masked loops.
10304 (vectorizable_reduction): Disable fully-masked loops.
10305 (vectorizable_live_operation): Likewise.
10306 (vect_halve_mask_nunits): New function.
10307 (vect_double_mask_nunits): Likewise.
10308 (vect_record_loop_mask): Likewise.
10309 (vect_get_loop_mask): Likewise.
10310 (vect_transform_loop): Handle the case in which the final loop
10311 iteration might handle a partial vector. Call vect_set_loop_condition
10312 instead of slpeel_make_loop_iterate_ntimes.
10313 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10314 (check_load_store_masking): New function.
10315 (prepare_load_store_mask): Likewise.
10316 (vectorizable_store): Handle fully-masked loops.
10317 (vectorizable_load): Likewise.
10318 (supportable_widening_operation): Use vect_halve_mask_nunits for
10320 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10321 (vect_gen_while): New function.
10322 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10323 (aarch64_uqdec<mode>): New insn.
10325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10326 Alan Hayward <alan.hayward@arm.com>
10327 David Sherwood <david.sherwood@arm.com>
10329 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10330 (reduc_xor_scal_optab): New optabs.
10331 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10332 (reduc_xor_scal_@var{m}): Document.
10333 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10334 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10335 internal functions.
10336 * fold-const-call.c (fold_const_call): Handle them.
10337 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10338 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10339 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10340 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10341 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10342 (UNSPEC_XORV): New unspecs.
10343 (optab): Add entries for them.
10344 (BITWISEV): New int iterator.
10345 (bit_reduc_op): New int attributes.
10347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10348 Alan Hayward <alan.hayward@arm.com>
10349 David Sherwood <david.sherwood@arm.com>
10351 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10352 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10353 * optabs.def (vec_shl_insert_optab): New optab.
10354 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10355 (duplicate_and_interleave): Likewise.
10356 * tree-vect-loop.c: Include internal-fn.h.
10357 (neutral_op_for_slp_reduction): New function, split out from
10358 get_initial_defs_for_reduction.
10359 (get_initial_def_for_reduction): Handle option 2 for variable-length
10360 vectors by loading the neutral value into a vector and then shifting
10361 the initial value into element 0.
10362 (get_initial_defs_for_reduction): Replace the code argument with
10363 the neutral value calculated by neutral_op_for_slp_reduction.
10364 Use gimple_build_vector for constant-length vectors.
10365 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10366 but the first group_size elements have a neutral value.
10367 Use duplicate_and_interleave otherwise.
10368 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10369 Update call to get_initial_defs_for_reduction. Handle SLP
10370 reductions for variable-length vectors by creating one vector
10371 result for each scalar result, with the elements associated
10372 with other scalar results stubbed out with the neutral value.
10373 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10374 Require IFN_VEC_SHL_INSERT for double reductions on
10375 variable-length vectors, or SLP reductions that have
10376 a neutral value. Require can_duplicate_and_interleave_p
10377 support for variable-length unchained SLP reductions if there
10378 is no neutral value, such as for MIN/MAX reductions. Also require
10379 the number of vector elements to be a multiple of the number of
10380 SLP statements when doing variable-length unchained SLP reductions.
10381 Update call to vect_create_epilog_for_reduction.
10382 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10383 and remove initial values.
10384 (duplicate_and_interleave): Make public.
10385 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10386 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10389 Alan Hayward <alan.hayward@arm.com>
10390 David Sherwood <david.sherwood@arm.com>
10392 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10393 (can_duplicate_and_interleave_p): New function.
10394 (vect_get_and_check_slp_defs): Take the vector of statements
10395 rather than just the current one. Remove excess parentheses.
10396 Restriction rejectinon of vect_constant_def and vect_external_def
10397 for variable-length vectors to boolean types, or types for which
10398 can_duplicate_and_interleave_p is false.
10399 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10400 (duplicate_and_interleave): New function.
10401 (vect_get_constant_vectors): Use gimple_build_vector for
10402 constant-length vectors and suitable variable-length constant
10403 vectors. Use duplicate_and_interleave for other variable-length
10404 vectors. Don't defer the update when inserting new statements.
10406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10407 Alan Hayward <alan.hayward@arm.com>
10408 David Sherwood <david.sherwood@arm.com>
10410 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10411 min_profitable_iters doesn't go negative.
10413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10414 Alan Hayward <alan.hayward@arm.com>
10415 David Sherwood <david.sherwood@arm.com>
10417 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10418 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10419 * optabs.def (vec_mask_load_lanes_optab): New optab.
10420 (vec_mask_store_lanes_optab): Likewise.
10421 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10422 (MASK_STORE_LANES): Likewise.
10423 * internal-fn.c (mask_load_lanes_direct): New macro.
10424 (mask_store_lanes_direct): Likewise.
10425 (expand_mask_load_optab_fn): Handle masked operations.
10426 (expand_mask_load_lanes_optab_fn): New macro.
10427 (expand_mask_store_optab_fn): Handle masked operations.
10428 (expand_mask_store_lanes_optab_fn): New macro.
10429 (direct_mask_load_lanes_optab_supported_p): Likewise.
10430 (direct_mask_store_lanes_optab_supported_p): Likewise.
10431 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10433 (vect_load_lanes_supported): Likewise.
10434 * tree-vect-data-refs.c (strip_conversion): New function.
10435 (can_group_stmts_p): Likewise.
10436 (vect_analyze_data_ref_accesses): Use it instead of checking
10437 for a pair of assignments.
10438 (vect_store_lanes_supported): Take a masked_p parameter.
10439 (vect_load_lanes_supported): Likewise.
10440 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10441 vect_store_lanes_supported and vect_load_lanes_supported.
10442 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10443 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10444 parameter. Don't allow gaps for masked accesses.
10445 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10446 and vect_load_lanes_supported.
10447 (get_load_store_type): Take a masked_p parameter and update
10448 call to get_group_load_store_type.
10449 (vectorizable_store): Update call to get_load_store_type.
10450 Handle IFN_MASK_STORE_LANES.
10451 (vectorizable_load): Update call to get_load_store_type.
10452 Handle IFN_MASK_LOAD_LANES.
10454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10455 Alan Hayward <alan.hayward@arm.com>
10456 David Sherwood <david.sherwood@arm.com>
10458 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10460 * config/aarch64/aarch64-protos.h
10461 (aarch64_sve_struct_memory_operand_p): Declare.
10462 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10463 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10464 (VPRED, vpred): Handle SVE structure modes.
10465 * config/aarch64/constraints.md (Utx): New constraint.
10466 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10467 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10468 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10469 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10470 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10471 structure modes. Split into pieces after RA.
10472 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10473 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10475 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10476 SVE structure modes.
10477 (aarch64_classify_address): Likewise.
10478 (sizetochar): Move earlier in file.
10479 (aarch64_print_operand): Handle SVE register lists.
10480 (aarch64_array_mode): New function.
10481 (aarch64_sve_struct_memory_operand_p): Likewise.
10482 (TARGET_ARRAY_MODE): Redefine.
10484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10485 Alan Hayward <alan.hayward@arm.com>
10486 David Sherwood <david.sherwood@arm.com>
10488 * target.def (array_mode): New target hook.
10489 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10490 * doc/tm.texi: Regenerate.
10491 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10492 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10493 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10494 targetm.array_mode.
10495 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10499 Alan Hayward <alan.hayward@arm.com>
10500 David Sherwood <david.sherwood@arm.com>
10502 * fold-const.c (fold_binary_loc): Check the argument types
10503 rather than the result type when testing for a vector operation.
10505 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10507 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10508 * doc/tm.texi: Regenerate.
10510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10511 Alan Hayward <alan.hayward@arm.com>
10512 David Sherwood <david.sherwood@arm.com>
10514 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10515 (sve): Document new AArch64 extension.
10516 * doc/md.texi (w): Extend the description of the AArch64
10517 constraint to include SVE vectors.
10518 (Upl, Upa): Document new AArch64 predicate constraints.
10519 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10521 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10522 (msve-vector-bits=): New option.
10523 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10524 SVE when these are disabled.
10525 (sve): New extension.
10526 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10527 modes. Adjust their number of units based on aarch64_sve_vg.
10528 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10529 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10530 aarch64_addr_query_type.
10531 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10532 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10533 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10534 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10535 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10536 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10537 (aarch64_simd_imm_zero_p): Delete.
10538 (aarch64_check_zero_based_sve_index_immediate): Declare.
10539 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10540 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10541 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10542 (aarch64_sve_float_mul_immediate_p): Likewise.
10543 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10544 rather than an rtx.
10545 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10546 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10547 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10548 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10549 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10550 (aarch64_regmode_natural_size): Likewise.
10551 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10552 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10554 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10555 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10556 for VG and the SVE predicate registers.
10557 (V_ALIASES): Add a "z"-prefixed alias.
10558 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10559 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10560 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10561 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10562 (REG_CLASS_NAMES): Add entries for them.
10563 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10564 and the predicate registers.
10565 (aarch64_sve_vg): Declare.
10566 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10567 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10568 (REGMODE_NATURAL_SIZE): Define.
10569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10571 * config/aarch64/aarch64.c: Include cfgrtl.h.
10572 (simd_immediate_info): Add a constructor for series vectors,
10573 and an associated step field.
10574 (aarch64_sve_vg): New variable.
10575 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10576 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10577 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10578 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10579 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10580 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10581 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10582 (aarch64_get_mask_mode): New functions.
10583 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10584 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10585 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10586 predicate modes and predicate registers. Explicitly restrict
10587 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10588 to store a vector mode if it is recognized by
10589 aarch64_classify_vector_mode.
10590 (aarch64_regmode_natural_size): New function.
10591 (aarch64_hard_regno_caller_save_mode): Return the original mode
10593 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10594 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10595 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10596 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10598 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10599 does not overlap dest if the function is frame-related. Handle
10601 (aarch64_split_add_offset): New function.
10602 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10603 them aarch64_add_offset.
10604 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10605 and update call to aarch64_sub_sp.
10606 (aarch64_add_cfa_expression): New function.
10607 (aarch64_expand_prologue): Pass extra temporary registers to the
10608 functions above. Handle the case in which we need to emit new
10609 DW_CFA_expressions for registers that were originally saved
10610 relative to the stack pointer, but now have to be expressed
10611 relative to the frame pointer.
10612 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10614 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10615 IP0 and IP1 values for SVE frames.
10616 (aarch64_expand_vec_series): New function.
10617 (aarch64_expand_sve_widened_duplicate): Likewise.
10618 (aarch64_expand_sve_const_vector): Likewise.
10619 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10620 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10621 into the register, rather than emitting a SET directly.
10622 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10623 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10624 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10625 (offset_9bit_signed_scaled_p): New functions.
10626 (aarch64_replicate_bitmask_imm): New function.
10627 (aarch64_bitmask_imm): Use it.
10628 (aarch64_cannot_force_const_mem): Reject expressions involving
10629 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10630 (aarch64_classify_index): Handle SVE indices, by requiring
10631 a plain register index with a scale that matches the element size.
10632 (aarch64_classify_address): Handle SVE addresses. Assert that
10633 the mode of the address is VOIDmode or an integer mode.
10634 Update call to aarch64_classify_symbol.
10635 (aarch64_classify_symbolic_expression): Update call to
10636 aarch64_classify_symbol.
10637 (aarch64_const_vec_all_in_range_p): New function.
10638 (aarch64_print_vector_float_operand): Likewise.
10639 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10640 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10641 and the FP immediates 1.0 and 0.5.
10642 (aarch64_print_address_internal): Handle SVE addresses.
10643 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10644 (aarch64_regno_regclass): Handle predicate registers.
10645 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10647 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10648 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10649 (aarch64_convert_sve_vector_bits): New function.
10650 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10651 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10652 rather than an rtx.
10653 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10654 Handle SVE vector and predicate modes. Accept VL-based constants
10655 that need only one temporary register, and VL offsets that require
10656 no temporary registers.
10657 (aarch64_conditional_register_usage): Mark the predicate registers
10658 as fixed if SVE isn't available.
10659 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10660 Return true for SVE vector and predicate modes.
10661 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10662 rather than an unsigned int. Handle SVE modes.
10663 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10665 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10667 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10668 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10669 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10670 (aarch64_sve_float_mul_immediate_p): New functions.
10671 (aarch64_sve_valid_immediate): New function.
10672 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10673 Explicitly reject structure modes. Check for INDEX constants.
10674 Handle PTRUE and PFALSE constants.
10675 (aarch64_check_zero_based_sve_index_immediate): New function.
10676 (aarch64_simd_imm_zero_p): Delete.
10677 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10678 vector modes. Accept constants in the range of CNT[BHWD].
10679 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10680 ask for an Advanced SIMD mode.
10681 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10682 (aarch64_simd_vector_alignment): Handle SVE predicates.
10683 (aarch64_vectorize_preferred_vector_alignment): New function.
10684 (aarch64_simd_vector_alignment_reachable): Use it instead of
10686 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10687 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10689 (MAX_VECT_LEN): Delete.
10690 (expand_vec_perm_d): Add a vec_flags field.
10691 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10692 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10693 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10695 (aarch64_evpc_rev): Rename to...
10696 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10697 (aarch64_evpc_rev_global): New function.
10698 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10699 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10701 (aarch64_evpc_sve_tbl): New function.
10702 (aarch64_expand_vec_perm_const_1): Update after rename of
10703 aarch64_evpc_rev. Handle SVE permutes too, trying
10704 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10705 than aarch64_evpc_tbl.
10706 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10707 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10708 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10709 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10710 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10711 (aarch64_expand_sve_vcond): New functions.
10712 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10713 of aarch64_vector_mode_p.
10714 (aarch64_dwarf_poly_indeterminate_value): New function.
10715 (aarch64_compute_pressure_classes): Likewise.
10716 (aarch64_can_change_mode_class): Likewise.
10717 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10718 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10719 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10720 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10721 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10722 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10723 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10724 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10726 (Dn, Dl, Dr): Accept const as well as const_vector.
10727 (Dz): Likewise. Compare against CONST0_RTX.
10728 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10729 of "vector" where appropriate.
10730 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10731 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10732 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10733 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10734 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10735 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10736 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10737 (v_int_equiv): Extend to SVE modes.
10738 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10740 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10741 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10742 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10743 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10744 (SVE_COND_FP_CMP): New int iterators.
10745 (perm_hilo): Handle the new unpack unspecs.
10746 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10748 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10749 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10750 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10751 (aarch64_equality_operator, aarch64_constant_vector_operand)
10752 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10753 (aarch64_sve_nonimmediate_operand): Likewise.
10754 (aarch64_sve_general_operand): Likewise.
10755 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10756 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10757 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10758 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10759 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10760 (aarch64_sve_float_arith_immediate): Likewise.
10761 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10762 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10763 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10764 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10765 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10766 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10767 (aarch64_sve_float_arith_operand): Likewise.
10768 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10769 (aarch64_sve_float_mul_operand): Likewise.
10770 (aarch64_sve_vec_perm_operand): Likewise.
10771 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10772 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10773 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10774 as well as const_vector.
10775 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10776 in file. Use CONST0_RTX and CONSTM1_RTX.
10777 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10778 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10779 Use aarch64_simd_imm_zero.
10780 * config/aarch64/aarch64-sve.md: New file.
10781 * config/aarch64/aarch64.md: Include it.
10782 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10783 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10784 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10785 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10786 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10787 (sve): New attribute.
10788 (enabled): Disable instructions with the sve attribute unless
10790 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10791 aarch64_expand_mov_immediate.
10792 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10793 CNT[BHSD] immediates.
10794 (movti): Split CONST_POLY_INT moves into two halves.
10795 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10796 Split additions that need a temporary here if the destination
10797 is the stack pointer.
10798 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10799 (*add<mode>3_poly_1): New instruction.
10800 (set_clobber_cc): New expander.
10802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10804 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10805 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10806 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10807 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10808 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10809 Change innermode from fixed_mode_size to machine_mode.
10810 (simplify_subreg): Update call accordingly. Handle a constant-sized
10811 subreg of a variable-length CONST_VECTOR.
10813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10814 Alan Hayward <alan.hayward@arm.com>
10815 David Sherwood <david.sherwood@arm.com>
10817 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10818 (add_offset_to_base): New function, split out from...
10819 (create_mem_ref): ...here. When handling a scale other than 1,
10820 check first whether the address is valid without the offset.
10821 Add it into the base if so, leaving the index and scale as-is.
10823 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10826 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10827 fold_for_warn before checking if arg2 is INTEGER_CST.
10829 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10831 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10832 (store_multiple_operation): Delete.
10833 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10834 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10835 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10836 guarded by TARGET_STRING.
10837 (rs6000_output_load_multiple): Delete.
10838 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10839 OPTION_MASK_STRING / TARGET_STRING handling.
10840 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10841 (const rs6000_opt_masks) <"string">: Change mask to 0.
10842 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10843 (MASK_STRING): Delete.
10844 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10846 (load_multiple): Delete.
10853 (store_multiple): Delete.
10860 (movmemsi_8reg): Delete.
10861 (corresponding unnamed define_insn): Delete.
10862 (movmemsi_6reg): Delete.
10863 (corresponding unnamed define_insn): Delete.
10864 (movmemsi_4reg): Delete.
10865 (corresponding unnamed define_insn): Delete.
10866 (movmemsi_2reg): Delete.
10867 (corresponding unnamed define_insn): Delete.
10868 (movmemsi_1reg): Delete.
10869 (corresponding unnamed define_insn): Delete.
10870 * config/rs6000/rs6000.opt (mno-string): New.
10871 (mstring): Replace by deprecation warning stub.
10872 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10874 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10876 * regrename.c (regrename_do_replace): If replacing the same
10877 reg multiple times, try to reuse last created gen_raw_REG.
10880 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10881 main to workaround a bug in GDB.
10883 2018-01-12 Tom de Vries <tom@codesourcery.com>
10886 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10888 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10890 PR rtl-optimization/80481
10891 * ira-color.c (get_cap_member): New function.
10892 (allocnos_conflict_by_live_ranges_p): Use it.
10893 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10894 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10896 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10899 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10900 (*saddl_se_1): Ditto.
10901 (*ssubsi_1): Ditto.
10902 (*ssubl_se_1): Ditto.
10904 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10906 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10907 rather than wi::to_widest for DR_INITs.
10908 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10909 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10910 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10912 (vect_analyze_group_access_1): Note that here.
10914 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10916 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10917 polynomial type sizes.
10919 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10921 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10922 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10923 (gimple_add_tmp_var): Likewise.
10925 2018-01-12 Martin Liska <mliska@suse.cz>
10927 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10928 (gimple_alloc_sizes): Likewise.
10929 (dump_gimple_statistics): Use PRIu64 in printf format.
10930 * gimple.h: Change uint64_t to int.
10932 2018-01-12 Martin Liska <mliska@suse.cz>
10934 * tree-core.h: Use uint64_t instead of int.
10935 * tree.c (tree_node_counts): Likewise.
10936 (tree_node_sizes): Likewise.
10937 (dump_tree_statistics): Use PRIu64 in printf format.
10939 2018-01-12 Martin Liska <mliska@suse.cz>
10941 * Makefile.in: As qsort_chk is implemented in vec.c, add
10942 vec.o to linkage of gencfn-macros.
10943 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10944 passing the info to record_node_allocation_statistics.
10945 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10947 * ggc-common.c (struct ggc_usage): Add operator== and use
10948 it in operator< and compare function.
10949 * mem-stats.h (struct mem_usage): Likewise.
10950 * vec.c (struct vec_usage): Remove operator< and compare
10951 function. Can be simply inherited.
10953 2018-01-12 Martin Jambor <mjambor@suse.cz>
10956 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10957 * tree-ssa-math-opts.c: Include domwalk.h.
10958 (convert_mult_to_fma_1): New function.
10959 (fma_transformation_info): New type.
10960 (fma_deferring_state): Likewise.
10961 (cancel_fma_deferring): New function.
10962 (result_of_phi): Likewise.
10963 (last_fma_candidate_feeds_initial_phi): Likewise.
10964 (convert_mult_to_fma): Added deferring logic, split actual
10965 transformation to convert_mult_to_fma_1.
10966 (math_opts_dom_walker): New type.
10967 (math_opts_dom_walker::after_dom_children): New method, body moved
10968 here from pass_optimize_widening_mul::execute, added deferring logic
10970 (pass_optimize_widening_mul::execute): Moved most of code to
10971 math_opts_dom_walker::after_dom_children.
10972 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10973 * config/i386/i386.c (ix86_option_override_internal): Added
10974 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10976 2018-01-12 Richard Biener <rguenther@suse.de>
10979 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10980 inline instance vars.
10982 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10985 * config/rx/rx.c (rx_is_restricted_memory_address):
10986 Handle SUBREG case.
10988 2018-01-12 Richard Biener <rguenther@suse.de>
10990 PR tree-optimization/80846
10991 * target.def (split_reduction): New target hook.
10992 * targhooks.c (default_split_reduction): New function.
10993 * targhooks.h (default_split_reduction): Declare.
10994 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10995 target requests first reduce vectors by combining low and high
10997 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10998 (get_vectype_for_scalar_type_and_size): Export.
10999 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11000 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11001 * doc/tm.texi: Regenerate.
11002 * config/i386/i386.c (ix86_split_reduction): Implement
11003 TARGET_VECTORIZE_SPLIT_REDUCTION.
11005 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11008 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11009 in PIC mode except for TARGET_VXWORKS_RTP.
11010 * config/sparc/sparc.c: Include cfgrtl.h.
11011 (TARGET_INIT_PIC_REG): Define.
11012 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11013 (sparc_pic_register_p): New predicate.
11014 (sparc_legitimate_address_p): Use it.
11015 (sparc_legitimize_pic_address): Likewise.
11016 (sparc_delegitimize_address): Likewise.
11017 (sparc_mode_dependent_address_p): Likewise.
11018 (gen_load_pcrel_sym): Remove 4th parameter.
11019 (load_got_register): Adjust call to above. Remove obsolete stuff.
11020 (sparc_expand_prologue): Do not call load_got_register here.
11021 (sparc_flat_expand_prologue): Likewise.
11022 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11023 (sparc_use_pseudo_pic_reg): New function.
11024 (sparc_init_pic_reg): Likewise.
11025 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11026 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11028 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11030 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11031 Add item for branch_cost.
11033 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11035 PR rtl-optimization/83565
11036 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11037 not extend the result to a larger mode for rotate operations.
11038 (num_sign_bit_copies1): Likewise.
11040 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11043 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11045 Use values-Xc.o for -pedantic.
11046 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11048 2018-01-12 Martin Liska <mliska@suse.cz>
11051 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11053 (possible_polymorphic_call_targets): Use it.
11054 (ipa_devirt): Likewise.
11056 2018-01-12 Martin Liska <mliska@suse.cz>
11058 * profile-count.h (enum profile_quality): Use 0 as invalid
11059 enum value of profile_quality.
11061 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11063 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11064 -mext-string options.
11066 2018-01-12 Richard Biener <rguenther@suse.de>
11068 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11069 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11070 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11072 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11074 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11076 * configure.ac (--with-long-double-format): Add support for the
11077 configuration option to change the default long double format on
11079 * config.gcc (powerpc*-linux*-*): Likewise.
11080 * configure: Regenerate.
11081 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11082 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11083 used without modification.
11085 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11087 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11088 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11089 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11090 MISC_BUILTIN_SPEC_BARRIER.
11091 (rs6000_init_builtins): Likewise.
11092 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11094 (speculation_barrier): New define_insn.
11095 * doc/extend.texi: Document __builtin_speculation_barrier.
11097 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11100 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11101 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11102 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11104 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11105 integral modes instead of "ss" and "sd".
11106 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11107 vectors with 32-bit and 64-bit elements.
11108 (vecdupssescalarmodesuffix): New mode attribute.
11109 (vec_dup<mode>): Use it.
11111 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11114 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11115 frame if argument is passed on stack.
11117 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11120 * ree.c (combine_reaching_defs): Optimize also
11121 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11122 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11124 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11126 PR middle-end/83189
11127 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11129 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11131 PR middle-end/83718
11132 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11133 after they are computed.
11135 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11137 PR tree-optimization/83695
11138 * gimple-loop-linterchange.cc
11139 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11140 reset cached scev information after interchange.
11141 (pass_linterchange::execute): Remove call to scev_reset_htab.
11143 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11145 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11146 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11147 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11148 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11149 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11150 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11151 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11152 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11153 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11154 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11155 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11156 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11157 (V_lane_reg): Likewise.
11158 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11160 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11161 (vfmal_lane_low<mode>_intrinsic,
11162 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11163 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11164 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11165 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11166 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11167 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11169 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11171 * config/arm/arm-cpus.in (fp16fml): New feature.
11172 (ALL_SIMD): Add fp16fml.
11173 (armv8.2-a): Add fp16fml as an option.
11174 (armv8.3-a): Likewise.
11175 (armv8.4-a): Add fp16fml as part of fp16.
11176 * config/arm/arm.h (TARGET_FP16FML): Define.
11177 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11179 * config/arm/arm-modes.def (V2HF): Define.
11180 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11181 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11182 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11183 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11184 vfmsl_low, vfmsl_high): New set of builtins.
11185 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11186 (vfml_op): New code attribute.
11187 (VFMLHALVES): New int iterator.
11188 (VFML, VFMLSEL): New mode attributes.
11189 (V_reg): Define mapping for V2HF.
11190 (V_hi, V_lo): New mode attributes.
11191 (VF_constraint): Likewise.
11192 (vfml_half, vfml_half_selector): New int attributes.
11193 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11195 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11196 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11198 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11199 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11200 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11201 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11203 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11204 Document new effective target and option set.
11206 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11208 * config/arm/arm-cpus.in (armv8_4): New feature.
11209 (ARMv8_4a): New fgroup.
11210 (armv8.4-a): New arch.
11211 * config/arm/arm-tables.opt: Regenerate.
11212 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11213 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11214 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11215 Add matching rules for -march=armv8.4-a and extensions.
11216 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11218 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11221 * config/rx/rx.md (BW): New mode attribute.
11222 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11224 2018-01-11 Richard Biener <rguenther@suse.de>
11226 PR tree-optimization/83435
11227 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11228 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11229 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11231 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11232 Alan Hayward <alan.hayward@arm.com>
11233 David Sherwood <david.sherwood@arm.com>
11235 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11237 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11238 (aarch64_print_address_internal): Use it to check for a zero offset.
11240 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11241 Alan Hayward <alan.hayward@arm.com>
11242 David Sherwood <david.sherwood@arm.com>
11244 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11245 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11246 Return a poly_int64 rather than a HOST_WIDE_INT.
11247 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11248 rather than a HOST_WIDE_INT.
11249 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11250 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11251 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11252 final_offset from HOST_WIDE_INT to poly_int64.
11253 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11254 to_constant when getting the number of units in an Advanced SIMD
11256 (aarch64_builtin_vectorized_function): Check for a constant number
11258 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11260 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11261 attribute instead of GET_MODE_NUNITS.
11262 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11263 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11264 GET_MODE_SIZE for fixed-size registers.
11265 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11266 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11267 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11268 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11269 (aarch64_print_operand, aarch64_print_address_internal)
11270 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11271 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11272 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11273 Handle polynomial GET_MODE_SIZE.
11274 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11275 wider than SImode without modification.
11276 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11277 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11278 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11279 passing and returning SVE modes.
11280 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11281 rather than GEN_INT.
11282 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11283 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11284 (aarch64_allocate_and_probe_stack_space): Likewise.
11285 (aarch64_layout_frame): Cope with polynomial offsets.
11286 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11287 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11288 polynomial offsets.
11289 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11290 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11291 poly_int64 rather than a HOST_WIDE_INT.
11292 (aarch64_get_separate_components, aarch64_process_components)
11293 (aarch64_expand_prologue, aarch64_expand_epilogue)
11294 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11295 (aarch64_anchor_offset): New function, split out from...
11296 (aarch64_legitimize_address): ...here.
11297 (aarch64_builtin_vectorization_cost): Handle polynomial
11298 TYPE_VECTOR_SUBPARTS.
11299 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11301 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11302 number of elements from the PARALLEL rather than the mode.
11303 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11304 rather than GET_MODE_BITSIZE.
11305 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11306 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11307 (aarch64_expand_vec_perm_const_1): Handle polynomial
11308 d->perm.length () and d->perm elements.
11309 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11310 Apply to_constant to d->perm elements.
11311 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11312 polynomial CONST_VECTOR_NUNITS.
11313 (aarch64_move_pointer): Take amount as a poly_int64 rather
11315 (aarch64_progress_pointer): Avoid temporary variable.
11316 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11317 the mode attribute instead of GET_MODE.
11319 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11320 Alan Hayward <alan.hayward@arm.com>
11321 David Sherwood <david.sherwood@arm.com>
11323 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11324 x exists before using it.
11325 (aarch64_add_constant_internal): Rename to...
11326 (aarch64_add_offset_1): ...this. Replace regnum with separate
11327 src and dest rtxes. Handle the case in which they're different,
11328 including when the offset is zero. Replace scratchreg with an rtx.
11329 Use 2 additions if there is no spare register into which we can
11330 move a 16-bit constant.
11331 (aarch64_add_constant): Delete.
11332 (aarch64_add_offset): Replace reg with separate src and dest
11333 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11334 Use aarch64_add_offset_1.
11335 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11336 an rtx rather than an int. Take the delta as a poly_int64
11337 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11338 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11339 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11340 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11341 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11342 and aarch64_add_sp.
11343 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11344 aarch64_add_constant.
11346 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11348 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11349 Use scalar_float_mode.
11351 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11353 * config/aarch64/aarch64-simd.md
11354 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11355 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11356 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11357 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11358 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11359 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11360 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11361 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11362 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11363 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11365 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11368 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11369 targ_options->x_arm_arch_string is non NULL.
11371 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11373 * config/aarch64/aarch64.h
11374 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11376 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11379 * expmed.c (emit_store_flag_force): Swap if const op0
11380 and change VOIDmode to mode of op0.
11382 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11384 PR rtl-optimization/83761
11385 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11386 than bytes to mode_for_size.
11388 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11390 PR middle-end/83189
11391 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11392 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11395 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11397 PR middle-end/83575
11398 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11399 when in layout mode.
11400 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11401 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11404 2018-01-10 Michael Collison <michael.collison@arm.com>
11406 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11407 * config/aarch64/aarch64-option-extension.def: Add
11408 AARCH64_OPT_EXTENSION of 'fp16fml'.
11409 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11410 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11411 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11412 * config/aarch64/constraints.md (Ui7): New constraint.
11413 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11414 (VFMLA_SEL_W): Ditto.
11417 (VFMLA16_LOW): New int iterator.
11418 (VFMLA16_HIGH): Ditto.
11419 (UNSPEC_FMLAL): New unspec.
11420 (UNSPEC_FMLSL): Ditto.
11421 (UNSPEC_FMLAL2): Ditto.
11422 (UNSPEC_FMLSL2): Ditto.
11423 (f16mac): New code attribute.
11424 * config/aarch64/aarch64-simd-builtins.def
11425 (aarch64_fmlal_lowv2sf): Ditto.
11426 (aarch64_fmlsl_lowv2sf): Ditto.
11427 (aarch64_fmlalq_lowv4sf): Ditto.
11428 (aarch64_fmlslq_lowv4sf): Ditto.
11429 (aarch64_fmlal_highv2sf): Ditto.
11430 (aarch64_fmlsl_highv2sf): Ditto.
11431 (aarch64_fmlalq_highv4sf): Ditto.
11432 (aarch64_fmlslq_highv4sf): Ditto.
11433 (aarch64_fmlal_lane_lowv2sf): Ditto.
11434 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11435 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11436 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11437 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11438 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11439 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11440 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11441 (aarch64_fmlal_lane_highv2sf): Ditto.
11442 (aarch64_fmlsl_lane_highv2sf): Ditto.
11443 (aarch64_fmlal_laneq_highv2sf): Ditto.
11444 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11445 (aarch64_fmlalq_lane_highv4sf): Ditto.
11446 (aarch64_fmlsl_lane_highv4sf): Ditto.
11447 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11448 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11449 * config/aarch64/aarch64-simd.md:
11450 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11451 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11452 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11453 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11454 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11455 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11456 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11457 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11458 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11459 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11460 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11461 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11462 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11463 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11464 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11465 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11466 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11467 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11468 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11469 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11470 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11471 (vfmlsl_low_u32): Ditto.
11472 (vfmlalq_low_u32): Ditto.
11473 (vfmlslq_low_u32): Ditto.
11474 (vfmlal_high_u32): Ditto.
11475 (vfmlsl_high_u32): Ditto.
11476 (vfmlalq_high_u32): Ditto.
11477 (vfmlslq_high_u32): Ditto.
11478 (vfmlal_lane_low_u32): Ditto.
11479 (vfmlsl_lane_low_u32): Ditto.
11480 (vfmlal_laneq_low_u32): Ditto.
11481 (vfmlsl_laneq_low_u32): Ditto.
11482 (vfmlalq_lane_low_u32): Ditto.
11483 (vfmlslq_lane_low_u32): Ditto.
11484 (vfmlalq_laneq_low_u32): Ditto.
11485 (vfmlslq_laneq_low_u32): Ditto.
11486 (vfmlal_lane_high_u32): Ditto.
11487 (vfmlsl_lane_high_u32): Ditto.
11488 (vfmlal_laneq_high_u32): Ditto.
11489 (vfmlsl_laneq_high_u32): Ditto.
11490 (vfmlalq_lane_high_u32): Ditto.
11491 (vfmlslq_lane_high_u32): Ditto.
11492 (vfmlalq_laneq_high_u32): Ditto.
11493 (vfmlslq_laneq_high_u32): Ditto.
11494 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11495 (AARCH64_FL_FOR_ARCH8_4): New.
11496 (AARCH64_ISA_F16FML): New ISA flag.
11497 (TARGET_F16FML): New feature flag for fp16fml.
11498 (doc/invoke.texi): Document new fp16fml option.
11500 2018-01-10 Michael Collison <michael.collison@arm.com>
11502 * config/aarch64/aarch64-builtins.c:
11503 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11504 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11505 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11506 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11507 (AARCH64_ISA_SHA3): New ISA flag.
11508 (TARGET_SHA3): New feature flag for sha3.
11509 * config/aarch64/iterators.md (sha512_op): New int attribute.
11510 (CRYPTO_SHA512): New int iterator.
11511 (UNSPEC_SHA512H): New unspec.
11512 (UNSPEC_SHA512H2): Ditto.
11513 (UNSPEC_SHA512SU0): Ditto.
11514 (UNSPEC_SHA512SU1): Ditto.
11515 * config/aarch64/aarch64-simd-builtins.def
11516 (aarch64_crypto_sha512hqv2di): New builtin.
11517 (aarch64_crypto_sha512h2qv2di): Ditto.
11518 (aarch64_crypto_sha512su0qv2di): Ditto.
11519 (aarch64_crypto_sha512su1qv2di): Ditto.
11520 (aarch64_eor3qv8hi): Ditto.
11521 (aarch64_rax1qv2di): Ditto.
11522 (aarch64_xarqv2di): Ditto.
11523 (aarch64_bcaxqv8hi): Ditto.
11524 * config/aarch64/aarch64-simd.md:
11525 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11526 (aarch64_crypto_sha512su0qv2di): Ditto.
11527 (aarch64_crypto_sha512su1qv2di): Ditto.
11528 (aarch64_eor3qv8hi): Ditto.
11529 (aarch64_rax1qv2di): Ditto.
11530 (aarch64_xarqv2di): Ditto.
11531 (aarch64_bcaxqv8hi): Ditto.
11532 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11533 (vsha512h2q_u64): Ditto.
11534 (vsha512su0q_u64): Ditto.
11535 (vsha512su1q_u64): Ditto.
11536 (veor3q_u16): Ditto.
11537 (vrax1q_u64): Ditto.
11538 (vxarq_u64): Ditto.
11539 (vbcaxq_u16): Ditto.
11540 * config/arm/types.md (crypto_sha512): New type attribute.
11541 (crypto_sha3): Ditto.
11542 (doc/invoke.texi): Document new sha3 option.
11544 2018-01-10 Michael Collison <michael.collison@arm.com>
11546 * config/aarch64/aarch64-builtins.c:
11547 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11548 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11549 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11550 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11551 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11552 (AARCH64_ISA_SM4): New ISA flag.
11553 (TARGET_SM4): New feature flag for sm4.
11554 * config/aarch64/aarch64-simd-builtins.def
11555 (aarch64_sm3ss1qv4si): Ditto.
11556 (aarch64_sm3tt1aq4si): Ditto.
11557 (aarch64_sm3tt1bq4si): Ditto.
11558 (aarch64_sm3tt2aq4si): Ditto.
11559 (aarch64_sm3tt2bq4si): Ditto.
11560 (aarch64_sm3partw1qv4si): Ditto.
11561 (aarch64_sm3partw2qv4si): Ditto.
11562 (aarch64_sm4eqv4si): Ditto.
11563 (aarch64_sm4ekeyqv4si): Ditto.
11564 * config/aarch64/aarch64-simd.md:
11565 (aarch64_sm3ss1qv4si): Ditto.
11566 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11567 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11568 (aarch64_sm4eqv4si): Ditto.
11569 (aarch64_sm4ekeyqv4si): Ditto.
11570 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11571 (sm3part_op): Ditto.
11572 (CRYPTO_SM3TT): Ditto.
11573 (CRYPTO_SM3PART): Ditto.
11574 (UNSPEC_SM3SS1): New unspec.
11575 (UNSPEC_SM3TT1A): Ditto.
11576 (UNSPEC_SM3TT1B): Ditto.
11577 (UNSPEC_SM3TT2A): Ditto.
11578 (UNSPEC_SM3TT2B): Ditto.
11579 (UNSPEC_SM3PARTW1): Ditto.
11580 (UNSPEC_SM3PARTW2): Ditto.
11581 (UNSPEC_SM4E): Ditto.
11582 (UNSPEC_SM4EKEY): Ditto.
11583 * config/aarch64/constraints.md (Ui2): New constraint.
11584 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11585 * config/arm/types.md (crypto_sm3): New type attribute.
11586 (crypto_sm4): Ditto.
11587 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11588 (vsm3tt1aq_u32): Ditto.
11589 (vsm3tt1bq_u32): Ditto.
11590 (vsm3tt2aq_u32): Ditto.
11591 (vsm3tt2bq_u32): Ditto.
11592 (vsm3partw1q_u32): Ditto.
11593 (vsm3partw2q_u32): Ditto.
11594 (vsm4eq_u32): Ditto.
11595 (vsm4ekeyq_u32): Ditto.
11596 (doc/invoke.texi): Document new sm4 option.
11598 2018-01-10 Michael Collison <michael.collison@arm.com>
11600 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11601 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11602 (AARCH64_FL_FOR_ARCH8_4): New.
11603 (AARCH64_FL_V8_4): New flag.
11604 (doc/invoke.texi): Document new armv8.4-a option.
11606 2018-01-10 Michael Collison <michael.collison@arm.com>
11608 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11609 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11610 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11611 * config/aarch64/aarch64-option-extension.def: Add
11612 AARCH64_OPT_EXTENSION of 'sha2'.
11613 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11614 (crypto): Disable sha2 and aes if crypto disabled.
11615 (crypto): Enable aes and sha2 if enabled.
11616 (simd): Disable sha2 and aes if simd disabled.
11617 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11619 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11620 (TARGET_SHA2): New feature flag for sha2.
11621 (TARGET_AES): New feature flag for aes.
11622 * config/aarch64/aarch64-simd.md:
11623 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11624 conditional on TARGET_AES.
11625 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11626 (aarch64_crypto_sha1hsi): Make pattern conditional
11628 (aarch64_crypto_sha1hv4si): Ditto.
11629 (aarch64_be_crypto_sha1hv4si): Ditto.
11630 (aarch64_crypto_sha1su1v4si): Ditto.
11631 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11632 (aarch64_crypto_sha1su0v4si): Ditto.
11633 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11634 (aarch64_crypto_sha256su0v4si): Ditto.
11635 (aarch64_crypto_sha256su1v4si): Ditto.
11636 (doc/invoke.texi): Document new aes and sha2 options.
11638 2018-01-10 Martin Sebor <msebor@redhat.com>
11640 PR tree-optimization/83781
11641 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11644 2018-01-11 Martin Sebor <msebor@gmail.com>
11645 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11647 PR tree-optimization/83501
11648 PR tree-optimization/81703
11650 * tree-ssa-strlen.c (get_string_cst): Rename...
11651 (get_string_len): ...to this. Handle global constants.
11652 (handle_char_store): Adjust.
11654 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11655 Jim Wilson <jimw@sifive.com>
11657 * config/riscv/riscv-protos.h (riscv_output_return): New.
11658 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11659 (riscv_attribute_table, riscv_output_return),
11660 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11661 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11662 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11664 (riscv_expand_prologue): Add early return for naked function.
11665 (riscv_expand_epilogue): Likewise.
11666 (riscv_function_ok_for_sibcall): Return false for naked function.
11667 (riscv_set_current_function): New.
11668 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11669 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11670 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11671 * doc/extend.texi (RISC-V Function Attributes): New.
11673 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11675 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11676 check for 128-bit long double before checking TCmode.
11677 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11678 128-bit long doubles before checking TFmode or TCmode.
11679 (FLOAT128_IBM_P): Likewise.
11681 2018-01-10 Martin Sebor <msebor@redhat.com>
11683 PR tree-optimization/83671
11684 * builtins.c (c_strlen): Unconditionally return zero for the empty
11686 Use -Warray-bounds for warnings.
11687 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11688 for non-constant array indices with COMPONENT_REF, arrays of
11689 arrays, and pointers to arrays.
11690 (gimple_fold_builtin_strlen): Determine and set length range for
11691 non-constant character arrays.
11693 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11695 PR middle-end/81897
11696 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11699 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11701 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11703 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11706 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11707 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11708 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11709 indexed_or_indirect_operand predicate.
11710 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11711 (*vsx_le_perm_load_v8hi): Likewise.
11712 (*vsx_le_perm_load_v16qi): Likewise.
11713 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11714 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11715 (*vsx_le_perm_store_v8hi): Likewise.
11716 (*vsx_le_perm_store_v16qi): Likewise.
11717 (eight unnamed splitters): Likewise.
11719 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11721 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11722 * config/rs6000/emmintrin.h: Likewise.
11723 * config/rs6000/mmintrin.h: Likewise.
11724 * config/rs6000/xmmintrin.h: Likewise.
11726 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11729 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11731 * tree.c (tree_nop_conversion): Return true for location wrapper
11733 (maybe_wrap_with_location): New function.
11734 (selftest::check_strip_nops): New function.
11735 (selftest::test_location_wrappers): New function.
11736 (selftest::tree_c_tests): Call it.
11737 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11738 (maybe_wrap_with_location): New decl.
11739 (EXPR_LOCATION_WRAPPER_P): New macro.
11740 (location_wrapper_p): New inline function.
11741 (tree_strip_any_location_wrapper): New inline function.
11743 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11746 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11747 stack_realign_offset for the largest alignment of stack slot
11749 (ix86_find_max_used_stack_alignment): New function.
11750 (ix86_finalize_stack_frame_flags): Use it. Set
11751 max_used_stack_alignment if we don't realign stack.
11752 * config/i386/i386.h (machine_function): Add
11753 max_used_stack_alignment.
11755 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11757 * config/arm/arm.opt (-mbranch-cost): New option.
11758 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11761 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11764 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11765 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11767 2018-01-10 Richard Biener <rguenther@suse.de>
11770 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11771 early out so it also covers the case where we have a non-NULL
11774 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11776 PR tree-optimization/83753
11777 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11778 for non-strided grouped accesses if the number of elements is 1.
11780 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11783 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11784 * i386.h (TARGET_USE_GATHER): Define.
11785 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11787 2018-01-10 Martin Liska <mliska@suse.cz>
11790 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11791 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11793 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11794 CLEANUP_NO_PARTITIONING is not set.
11796 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11798 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11799 for vectors, as a partial revert of r254296.
11800 * rtl.h (const_vec_p): Delete.
11801 (const_vec_duplicate_p): Don't test for vector CONSTs.
11802 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11803 * expmed.c (make_tree): Likewise.
11806 * common.md (E, F): Use CONSTANT_P instead of checking for
11808 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11809 checking for CONST_VECTOR.
11811 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11813 PR middle-end/83575
11814 * predict.c (force_edge_cold): Handle in more sane way edges
11815 with no prediction.
11817 2018-01-09 Carl Love <cel@us.ibm.com>
11819 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11821 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11822 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11823 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11824 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11825 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11826 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11827 * config/rs6000/rs6000-protos.h: Add extern defition for
11828 rs6000_generate_float2_double_code.
11829 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11831 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11832 (float2_v2df): Add define_expand.
11834 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11837 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11838 op_mode in the force_to_mode call.
11840 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11842 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11843 instead of checking each element individually.
11844 (aarch64_evpc_uzp): Likewise.
11845 (aarch64_evpc_zip): Likewise.
11846 (aarch64_evpc_ext): Likewise.
11847 (aarch64_evpc_rev): Likewise.
11848 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11849 instead of checking each element individually. Return true without
11851 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11852 whether all selected elements come from the same input, instead of
11853 checking each element individually. Remove calls to gen_rtx_REG,
11854 start_sequence and end_sequence and instead assert that no rtl is
11857 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11859 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11860 order of HIGH and CONST checks.
11862 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11864 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11865 if the destination isn't an SSA_NAME.
11867 2018-01-09 Richard Biener <rguenther@suse.de>
11869 PR tree-optimization/83668
11870 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11872 (canonicalize_loop_form): ... here, renamed from ...
11873 (canonicalize_loop_closed_ssa_form): ... this and amended to
11874 swap successor edges for loop exit blocks to make us use
11875 the RPO order we need for initial schedule generation.
11877 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11879 PR tree-optimization/64811
11880 * match.pd: When optimizing comparisons with Inf, avoid
11881 introducing or losing exceptions from comparisons with NaN.
11883 2018-01-09 Martin Liska <mliska@suse.cz>
11886 * asan.c (shadow_mem_size): Add gcc_assert.
11888 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11890 Don't save registers in main().
11893 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11894 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11895 * config/avr/avr.c (avr_set_current_function): Don't error if
11896 naked, OS_task or OS_main are specified at the same time.
11897 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11899 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11901 * common/config/avr/avr-common.c (avr_option_optimization_table):
11902 Switch on -mmain-is-OS_task for optimizing compilations.
11904 2018-01-09 Richard Biener <rguenther@suse.de>
11906 PR tree-optimization/83572
11907 * graphite.c: Include cfganal.h.
11908 (graphite_transform_loops): Connect infinite loops to exit
11909 and remove fake edges at the end.
11911 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11913 * ipa-inline.c (edge_badness): Revert accidental checkin.
11915 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11918 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11919 symbols; not inline clones.
11921 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11924 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11925 hard registers. Formatting fixes.
11927 PR preprocessor/83722
11928 * gcc.c (try_generate_repro): Pass
11929 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11930 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11933 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11934 Kito Cheng <kito.cheng@gmail.com>
11936 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11937 (riscv_leaf_function_p): Delete.
11938 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11940 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11942 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11944 (do_ifelse): New function.
11945 (do_isel): New function.
11946 (do_sub3): New function.
11947 (do_add3): New function.
11948 (do_load_mask_compare): New function.
11949 (do_overlap_load_compare): New function.
11950 (expand_compare_loop): New function.
11951 (expand_block_compare): Call expand_compare_loop() when appropriate.
11952 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11953 option description.
11954 (-mblock-compare-inline-loop-limit): New option.
11956 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11959 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11960 Reverse order of second and third operands in first alternative.
11961 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11962 of first and second elements in UNSPEC_VPERMR vector.
11963 (altivec_expand_vec_perm_le): Likewise.
11965 2018-01-08 Jeff Law <law@redhat.com>
11967 PR rtl-optimizatin/81308
11968 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11969 (process_switch): If group_case_labels makes a change, then set
11971 (pass_convert_switch::execute): If a switch is converted, then
11972 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11974 PR rtl-optimization/81308
11975 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11978 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11980 PR target/83663 - Revert r255946
11981 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11982 generation for cases where splatting a value is not useful.
11983 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11984 across a vec_duplicate and a paradoxical subreg forming a vector
11985 mode to a vec_concat.
11987 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11989 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11990 -march=armv8.3-a variants.
11991 * config/arm/t-multilib: Likewise.
11992 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11994 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11996 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11998 (cceq_ior_compare_complement): Give it a name so I can use it, and
11999 change boolean_or_operator predicate to boolean_operator so it can
12000 be used to generate a crand.
12001 (eqne): New code iterator.
12002 (bd/bd_neg): New code_attrs.
12003 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12004 a single define_insn.
12005 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12006 decrement (bdnzt/bdnzf/bdzt/bdzf).
12007 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12008 with the new names of the branch decrement patterns, and added the
12009 names of the branch decrement conditional patterns.
12011 2018-01-08 Richard Biener <rguenther@suse.de>
12013 PR tree-optimization/83563
12014 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12017 2018-01-08 Richard Biener <rguenther@suse.de>
12019 PR middle-end/83713
12020 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12022 2018-01-08 Richard Biener <rguenther@suse.de>
12024 PR tree-optimization/83685
12025 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12026 references to abnormals.
12028 2018-01-08 Richard Biener <rguenther@suse.de>
12031 * dwarf2out.c (output_indirect_strings): Handle empty
12032 skeleton_debug_str_hash.
12033 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12035 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12037 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12038 (emit_store_direct): Likewise.
12039 (arc_trampoline_adjust_address): Likewise.
12040 (arc_asm_trampoline_template): New function.
12041 (arc_initialize_trampoline): Use asm_trampoline_template.
12042 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12043 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12044 * config/arc/arc.md (flush_icache): Delete pattern.
12046 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12048 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12049 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12052 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12055 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12056 by not USED_FOR_TARGET.
12057 (make_pass_resolve_sw_modes): Likewise.
12059 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12061 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12064 2018-01-08 Richard Biener <rguenther@suse.de>
12066 PR middle-end/83580
12067 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12069 2018-01-08 Richard Biener <rguenther@suse.de>
12071 PR middle-end/83517
12072 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12074 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12076 PR middle-end/81897
12077 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12078 basic blocks with a small number of successors.
12079 (convert_control_dep_chain_into_preds): Improve handling of
12081 (dump_predicates): Split apart into...
12082 (dump_pred_chain): ...here...
12083 (dump_pred_info): ...and here.
12084 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12085 (can_chain_union_be_invalidated_p): Improve check for invalidation
12087 (uninit_uses_cannot_happen): Avoid unnecessary if
12088 convert_control_dep_chain_into_preds yielded nothing.
12090 2018-01-06 Martin Sebor <msebor@redhat.com>
12092 PR tree-optimization/83640
12093 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12094 subtracting negative offset from size.
12095 (builtin_access::overlap): Adjust offset bounds of the access to fall
12096 within the size of the object if possible.
12098 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12100 PR rtl-optimization/83699
12101 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12102 extract_bit_field_as_subreg to cases in which the extracted
12103 value is also a vector.
12105 * lra-constraints.c (process_alt_operands): Test for the equivalence
12106 substitutions when detecting a possible reload cycle.
12108 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12111 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12112 by default if flag_selective_schedling{,2}. Formatting fixes.
12114 PR rtl-optimization/83682
12115 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12116 if it has non-VECTOR_MODE element mode.
12117 (vec_duplicate_p): Likewise.
12119 PR middle-end/83694
12120 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12121 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12123 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12126 * config/i386/i386-builtin.def
12127 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12128 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12129 Require also OPTION_MASK_ISA_AVX512F in addition to
12130 OPTION_MASK_ISA_GFNI.
12131 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12132 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12133 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12134 to OPTION_MASK_ISA_GFNI.
12135 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12136 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12137 OPTION_MASK_ISA_AVX512BW.
12138 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12139 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12140 addition to OPTION_MASK_ISA_GFNI.
12141 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12142 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12143 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12144 to OPTION_MASK_ISA_GFNI.
12145 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12146 a requirement for all ISAs rather than any of them with a few
12148 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12150 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12151 bitmasks to be enabled with 3 exceptions, instead of requiring any
12152 enabled ISA with lots of exceptions.
12153 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12154 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12155 Change avx512bw in isa attribute to avx512f.
12156 * config/i386/sgxintrin.h: Add license boilerplate.
12157 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12158 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12159 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12160 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12162 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12163 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12164 temporarily sse2 rather than sse if not enabled already.
12167 * config/i386/sse.md (VI248_VLBW): Rename to ...
12168 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12169 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12170 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12171 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12172 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12173 mode iterator instead of VI248_VLBW.
12175 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12177 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12178 (record_modified): Skip clobbers; add debug output.
12179 (param_change_prob): Use sreal frequencies.
12181 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12183 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12184 punt for user-aligned variables.
12186 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12188 * tree-chrec.c (chrec_contains_symbols): Return true for
12191 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12194 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12195 of (x|y) == x for BICS pattern.
12197 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12199 PR tree-optimization/83605
12200 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12201 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12204 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12206 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12207 * config/epiphany/rtems.h: New file.
12209 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12210 Uros Bizjak <ubizjak@gmail.com>
12213 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12214 QIreg_operand instead of register_operand predicate.
12215 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12216 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12217 comments instead of -fmitigate[-_]rop.
12219 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12222 * cgraphunit.c (symbol_table::compile): Switch to text_section
12223 before calling assembly_start debug hook.
12224 * run-rtl-passes.c (run_rtl_passes): Likewise.
12227 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12229 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12230 range_int_cst_p rather than !symbolic_range_p before calling
12231 extract_range_from_multiplicative_op_1.
12233 2018-01-04 Jeff Law <law@redhat.com>
12235 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12236 redundant test in assertion.
12238 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12240 * doc/rtl.texi: Document machine_mode wrapper classes.
12242 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12244 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12245 using tree_to_uhwi.
12247 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12249 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12250 the VEC_PERM_EXPR fold to fail.
12252 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12255 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12256 to switched_sections.
12258 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12261 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12262 test for d.testing.
12264 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12267 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12268 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12270 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12273 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12274 is BLKmode and bitpos not zero or mode change is needed.
12276 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12279 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12282 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12285 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12286 instead of MULT rtx. Update all corresponding splitters.
12287 (*saddl_se): Ditto.
12288 (*ssub<modesuffix>): Ditto.
12289 (*ssubl_se): Ditto.
12290 (*cmp_sadd_di): Update split patterns.
12291 (*cmp_sadd_si): Ditto.
12292 (*cmp_sadd_sidi): Ditto.
12293 (*cmp_ssub_di): Ditto.
12294 (*cmp_ssub_si): Ditto.
12295 (*cmp_ssub_sidi): Ditto.
12296 * config/alpha/predicates.md (const23_operand): New predicate.
12297 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12298 Look for ASHIFT, not MULT inner operand.
12299 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12301 2018-01-04 Martin Liska <mliska@suse.cz>
12303 PR gcov-profile/83669
12304 * gcov.c (output_intermediate_file): Add version to intermediate
12306 * doc/gcov.texi: Document new field 'version' in intermediate
12307 file format. Fix location of '-k' option of gcov command.
12309 2018-01-04 Martin Liska <mliska@suse.cz>
12312 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12314 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12316 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12318 2018-01-03 Martin Sebor <msebor@redhat.com>
12320 PR tree-optimization/83655
12321 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12322 checking calls with invalid arguments.
12324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12326 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12327 (vectorizable_mask_load_store): Delete.
12328 (vectorizable_call): Return false for masked loads and stores.
12329 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12330 instead of gimple_assign_rhs1.
12331 (vectorizable_load): Handle IFN_MASK_LOAD.
12332 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12336 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12338 (vectorizable_mask_load_store): ...here.
12339 (vectorizable_load): ...and here.
12341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12343 * tree-vect-stmts.c (vect_build_all_ones_mask)
12344 (vect_build_zero_merge_argument): New functions, split out from...
12345 (vectorizable_load): ...here.
12347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12349 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12351 (vectorizable_mask_load_store): ...here.
12352 (vectorizable_store): ...and here.
12354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12356 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12358 (vectorizable_mask_load_store): ...here.
12360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12363 (vect_model_store_cost): Take a vec_load_store_type instead of a
12365 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12366 (vect_model_store_cost): Take a vec_load_store_type instead of a
12368 (vectorizable_mask_load_store): Update accordingly.
12369 (vectorizable_store): Likewise.
12370 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12374 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12375 IFN_MASK_LOAD calls here rather than...
12376 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12379 Alan Hayward <alan.hayward@arm.com>
12380 David Sherwood <david.sherwood@arm.com>
12382 * expmed.c (extract_bit_field_1): For vector extracts,
12383 fall back to extract_bit_field_as_subreg if vec_extract
12386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12387 Alan Hayward <alan.hayward@arm.com>
12388 David Sherwood <david.sherwood@arm.com>
12390 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12391 they are variable or constant sized.
12392 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12393 slots for constant-sized data.
12395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12396 Alan Hayward <alan.hayward@arm.com>
12397 David Sherwood <david.sherwood@arm.com>
12399 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12400 handling COND_EXPRs with boolean comparisons, try to find a better
12401 basis for the mask type than the boolean itself.
12403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12405 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12406 is calculated and how it can be overridden.
12407 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12408 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12410 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12414 Alan Hayward <alan.hayward@arm.com>
12415 David Sherwood <david.sherwood@arm.com>
12417 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12418 Remove the mode argument.
12419 (aarch64_simd_valid_immediate): Remove the mode and inverse
12421 * config/aarch64/iterators.md (bitsize): New iterator.
12422 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12423 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12424 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12425 aarch64_simd_valid_immediate.
12426 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12427 (aarch64_reg_or_bic_imm): Likewise.
12428 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12429 with an insn_type enum and msl with a modifier_type enum.
12430 Replace element_width with a scalar_mode. Change the shift
12431 to unsigned int. Add constructors for scalar_float_mode and
12432 scalar_int_mode elements.
12433 (aarch64_vect_float_const_representable_p): Delete.
12434 (aarch64_can_const_movi_rtx_p)
12435 (aarch64_simd_scalar_immediate_valid_for_move)
12436 (aarch64_simd_make_constant): Update call to
12437 aarch64_simd_valid_immediate.
12438 (aarch64_advsimd_valid_immediate_hs): New function.
12439 (aarch64_advsimd_valid_immediate): Likewise.
12440 (aarch64_simd_valid_immediate): Remove mode and inverse
12441 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12442 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12443 and aarch64_float_const_representable_p on the result.
12444 (aarch64_output_simd_mov_immediate): Remove mode argument.
12445 Update call to aarch64_simd_valid_immediate and use of
12446 simd_immediate_info.
12447 (aarch64_output_scalar_simd_mov_immediate): Update call
12450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12451 Alan Hayward <alan.hayward@arm.com>
12452 David Sherwood <david.sherwood@arm.com>
12454 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12455 (mode_nunits): Likewise CONST_MODE_NUNITS.
12456 * machmode.def (ADJUST_NUNITS): Document.
12457 * genmodes.c (mode_data::need_nunits_adj): New field.
12458 (blank_mode): Update accordingly.
12459 (adj_nunits): New variable.
12460 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12462 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12463 listed in adj_nunits.
12464 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12465 listed in adj_nunits. Don't emit case statements for such modes.
12466 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12467 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12468 nothing if adj_nunits is nonnull.
12469 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12470 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12471 (emit_mode_fbit): Update use of print_maybe_const_decl.
12472 (emit_move_size): Likewise. Treat the array as non-const
12474 (emit_mode_adjustments): Handle adj_nunits.
12476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12478 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12479 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12480 (VECTOR_MODES): Use it.
12481 (make_vector_modes): Take the prefix as an argument.
12483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12484 Alan Hayward <alan.hayward@arm.com>
12485 David Sherwood <david.sherwood@arm.com>
12487 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12488 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12489 for MODE_VECTOR_BOOL.
12490 * machmode.def (VECTOR_BOOL_MODE): Document.
12491 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12492 (make_vector_bool_mode): New function.
12493 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12495 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12496 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12498 * stor-layout.c (int_mode_for_mode): Likewise.
12499 * tree.c (build_vector_type_for_mode): Likewise.
12500 * varasm.c (output_constant_pool_2): Likewise.
12501 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12502 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12503 for MODE_VECTOR_BOOL.
12504 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12505 of mode class checks.
12506 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12507 instead of a list of mode class checks.
12508 (expand_vector_scalar_condition): Likewise.
12509 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12512 Alan Hayward <alan.hayward@arm.com>
12513 David Sherwood <david.sherwood@arm.com>
12515 * machmode.h (mode_size): Change from unsigned short to
12517 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12518 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12519 or if measurement_type is not polynomial.
12520 (fixed_size_mode::includes_p): Check for constant-sized modes.
12521 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12522 return a poly_uint16 rather than an unsigned short.
12523 (emit_mode_size): Change the type of mode_size from unsigned short
12524 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12525 (emit_mode_adjustments): Cope with polynomial vector sizes.
12526 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12528 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12530 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12531 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12532 * caller-save.c (setup_save_areas): Likewise.
12533 (replace_reg_with_saved_mem): Likewise.
12534 * calls.c (emit_library_call_value_1): Likewise.
12535 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12536 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12537 (gen_lowpart_for_combine): Likewise.
12538 * convert.c (convert_to_integer_1): Likewise.
12539 * cse.c (equiv_constant, cse_insn): Likewise.
12540 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12541 (cselib_subst_to_values): Likewise.
12542 * dce.c (word_dce_process_block): Likewise.
12543 * df-problems.c (df_word_lr_mark_ref): Likewise.
12544 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12545 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12546 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12547 (rtl_for_decl_location): Likewise.
12548 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12549 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12550 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12551 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12552 (expand_expr_real_1): Likewise.
12553 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12554 (pad_below): Likewise.
12555 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12556 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12557 * ira.c (get_subreg_tracking_sizes): Likewise.
12558 * ira-build.c (ira_create_allocno_objects): Likewise.
12559 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12560 (ira_sort_regnos_for_alter_reg): Likewise.
12561 * ira-costs.c (record_operand_costs): Likewise.
12562 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12563 (resolve_simple_move): Likewise.
12564 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12565 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12566 (lra_constraints): Likewise.
12567 (CONST_POOL_OK_P): Reject variable-sized modes.
12568 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12569 (add_pseudo_to_slot, lra_spill): Likewise.
12570 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12571 * optabs-query.c (get_best_extraction_insn): Likewise.
12572 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12573 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12574 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12575 * recog.c (offsettable_address_addr_space_p): Likewise.
12576 * regcprop.c (maybe_mode_change): Likewise.
12577 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12578 * regrename.c (build_def_use): Likewise.
12579 * regstat.c (dump_reg_info): Likewise.
12580 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12581 (find_reloads, find_reloads_subreg_address): Likewise.
12582 * reload1.c (eliminate_regs_1): Likewise.
12583 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12584 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12585 (simplify_binary_operation_1, simplify_subreg): Likewise.
12586 * targhooks.c (default_function_arg_padding): Likewise.
12587 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12588 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12589 (verify_gimple_assign_ternary): Likewise.
12590 * tree-inline.c (estimate_move_cost): Likewise.
12591 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12592 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12593 (get_address_cost_ainc): Likewise.
12594 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12595 (vect_supportable_dr_alignment): Likewise.
12596 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12597 (vectorizable_reduction): Likewise.
12598 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12599 (vectorizable_operation, vectorizable_load): Likewise.
12600 * tree.c (build_same_sized_truth_vector_type): Likewise.
12601 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12602 * var-tracking.c (emit_note_insn_var_location): Likewise.
12603 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12604 (ADDR_VEC_ALIGN): Likewise.
12606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12607 Alan Hayward <alan.hayward@arm.com>
12608 David Sherwood <david.sherwood@arm.com>
12610 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12612 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12613 or if measurement_type is polynomial.
12614 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12615 * combine.c (make_extraction): Likewise.
12616 * dse.c (find_shift_sequence): Likewise.
12617 * dwarf2out.c (mem_loc_descriptor): Likewise.
12618 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12619 (extract_bit_field, extract_low_bits): Likewise.
12620 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12621 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12622 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12623 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12624 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12625 * reload.c (find_reloads): Likewise.
12626 * reload1.c (alter_reg): Likewise.
12627 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12628 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12629 * tree-if-conv.c (predicate_mem_writes): Likewise.
12630 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12631 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12632 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12633 * valtrack.c (dead_debug_insert_temp): Likewise.
12634 * varasm.c (mergeable_constant_section): Likewise.
12635 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12638 Alan Hayward <alan.hayward@arm.com>
12639 David Sherwood <david.sherwood@arm.com>
12641 * expr.c (expand_assignment): Cope with polynomial mode sizes
12642 when assigning to a CONCAT.
12644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12645 Alan Hayward <alan.hayward@arm.com>
12646 David Sherwood <david.sherwood@arm.com>
12648 * machmode.h (mode_precision): Change from unsigned short to
12650 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12652 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12653 or if measurement_type is not polynomial.
12654 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12655 in which the mode is already known to be a scalar_int_mode.
12656 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12657 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12659 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12660 for GET_MODE_PRECISION.
12661 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12662 for GET_MODE_PRECISION.
12663 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12665 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12666 (expand_field_assignment, make_extraction): Likewise.
12667 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12668 (get_last_value): Likewise.
12669 * convert.c (convert_to_integer_1): Likewise.
12670 * cse.c (cse_insn): Likewise.
12671 * expr.c (expand_expr_real_1): Likewise.
12672 * lra-constraints.c (simplify_operand_subreg): Likewise.
12673 * optabs-query.c (can_atomic_load_p): Likewise.
12674 * optabs.c (expand_atomic_load): Likewise.
12675 (expand_atomic_store): Likewise.
12676 * ree.c (combine_reaching_defs): Likewise.
12677 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12678 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12679 * tree.h (type_has_mode_precision_p): Likewise.
12680 * ubsan.c (instrument_si_overflow): Likewise.
12682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12687 polynomial numbers of units.
12688 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12689 (valid_vector_subparts_p): New function.
12690 (build_vector_type): Remove temporary shim and take the number
12691 of units as a poly_uint64 rather than an int.
12692 (build_opaque_vector_type): Take the number of units as a
12693 poly_uint64 rather than an int.
12694 * tree.c (build_vector_from_ctor): Handle polynomial
12695 TYPE_VECTOR_SUBPARTS.
12696 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12697 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12698 (build_vector_from_val): If the number of units is variable,
12699 use build_vec_duplicate_cst for constant operands and
12700 VEC_DUPLICATE_EXPR otherwise.
12701 (make_vector_type): Remove temporary is_constant ().
12702 (build_vector_type, build_opaque_vector_type): Take the number of
12703 units as a poly_uint64 rather than an int.
12704 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12706 * cfgexpand.c (expand_debug_expr): Likewise.
12707 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12708 (store_constructor, expand_expr_real_1): Likewise.
12709 (const_scalar_mask_from_tree): Likewise.
12710 * fold-const-call.c (fold_const_reduction): Likewise.
12711 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12712 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12713 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12714 (fold_relational_const): Likewise.
12715 (native_interpret_vector): Likewise. Change the size from an
12716 int to an unsigned int.
12717 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12718 TYPE_VECTOR_SUBPARTS.
12719 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12720 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12721 duplicating a non-constant operand into a variable-length vector.
12722 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12723 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12724 * ipa-icf.c (sem_variable::equals): Likewise.
12725 * match.pd: Likewise.
12726 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12727 * print-tree.c (print_node): Likewise.
12728 * stor-layout.c (layout_type): Likewise.
12729 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12730 * tree-cfg.c (verify_gimple_comparison): Likewise.
12731 (verify_gimple_assign_binary): Likewise.
12732 (verify_gimple_assign_ternary): Likewise.
12733 (verify_gimple_assign_single): Likewise.
12734 * tree-pretty-print.c (dump_generic_node): Likewise.
12735 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12736 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12737 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12738 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12739 (vect_shift_permute_load_chain): Likewise.
12740 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12741 (expand_vector_condition, optimize_vector_constructor): Likewise.
12742 (lower_vec_perm, get_compute_type): Likewise.
12743 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12744 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12745 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12746 (vect_recog_mask_conversion_pattern): Likewise.
12747 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12748 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12749 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12750 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12751 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12752 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12753 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12754 (supportable_widening_operation): Likewise.
12755 (supportable_narrowing_operation): Likewise.
12756 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12758 * varasm.c (output_constant): Likewise.
12760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12761 Alan Hayward <alan.hayward@arm.com>
12762 David Sherwood <david.sherwood@arm.com>
12764 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12765 so that both the length == 3 and length != 3 cases set up their
12766 own permute vectors. Add comments explaining why we know the
12767 number of elements is constant.
12768 (vect_permute_load_chain): Likewise.
12770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12771 Alan Hayward <alan.hayward@arm.com>
12772 David Sherwood <david.sherwood@arm.com>
12774 * machmode.h (mode_nunits): Change from unsigned char to
12776 (ONLY_FIXED_SIZE_MODES): New macro.
12777 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12778 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12779 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12781 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12782 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12783 or if measurement_type is not polynomial.
12784 * genmodes.c (ZERO_COEFFS): New macro.
12785 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12787 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12788 Use ZERO_COEFFS when emitting initializers.
12789 * data-streamer.h (bp_pack_poly_value): New function.
12790 (bp_unpack_poly_value): Likewise.
12791 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12792 for GET_MODE_NUNITS.
12793 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12794 for GET_MODE_NUNITS.
12795 * tree.c (make_vector_type): Remove temporary shim and make
12796 the real function take the number of units as a poly_uint64
12797 rather than an int.
12798 (build_vector_type_for_mode): Handle polynomial nunits.
12799 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12800 * emit-rtl.c (const_vec_series_p_1): Likewise.
12801 (gen_rtx_CONST_VECTOR): Likewise.
12802 * fold-const.c (test_vec_duplicate_folding): Likewise.
12803 * genrecog.c (validate_pattern): Likewise.
12804 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12805 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12806 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12807 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12808 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12809 * rtlanal.c (subreg_get_info): Likewise.
12810 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12811 (vect_grouped_load_supported): Likewise.
12812 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12813 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12814 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12815 (simplify_const_unary_operation, simplify_binary_operation_1)
12816 (simplify_const_binary_operation, simplify_ternary_operation)
12817 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12818 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12819 instead of CONST_VECTOR_NUNITS.
12820 * varasm.c (output_constant_pool_2): Likewise.
12821 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12822 explicit-encoded elements in the XVEC for variable-length vectors.
12824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12826 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12829 Alan Hayward <alan.hayward@arm.com>
12830 David Sherwood <david.sherwood@arm.com>
12832 * coretypes.h (fixed_size_mode): Declare.
12833 (fixed_size_mode_pod): New typedef.
12834 * builtins.h (target_builtins::x_apply_args_mode)
12835 (target_builtins::x_apply_result_mode): Change type to
12836 fixed_size_mode_pod.
12837 * builtins.c (apply_args_size, apply_result_size, result_vector)
12838 (expand_builtin_apply_args_1, expand_builtin_apply)
12839 (expand_builtin_return): Update accordingly.
12841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12843 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12844 * cselib.c (cselib_hash_rtx): Likewise.
12845 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12846 CONST_VECTOR encoding.
12848 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12849 Jeff Law <law@redhat.com>
12852 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12853 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12854 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12855 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12858 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12859 explicitly probe *sp in a noreturn function if there were any callee
12860 register saves or frame pointer is needed.
12862 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12865 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12866 BLKmode for ternary, binary or unary expressions.
12869 * var-tracking.c (delete_vta_debug_insn): New inline function.
12870 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12871 insns from get_insns () to NULL instead of each bb separately.
12872 Use delete_vta_debug_insn. No longer static.
12873 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12874 delete_vta_debug_insns callers.
12875 * rtl.h (delete_vta_debug_insns): Declare.
12876 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12877 instead of variable_tracking_main.
12879 2018-01-03 Martin Sebor <msebor@redhat.com>
12881 PR tree-optimization/83603
12882 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12883 arguments past the endof the argument list in functions declared
12884 without a prototype.
12885 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12886 Avoid checking when arguments are null.
12888 2018-01-03 Martin Sebor <msebor@redhat.com>
12891 * doc/extend.texi (attribute const): Fix a typo.
12892 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12893 issuing -Wsuggest-attribute for void functions.
12895 2018-01-03 Martin Sebor <msebor@redhat.com>
12897 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12898 offset_int::from instead of wide_int::to_shwi.
12899 (maybe_diag_overlap): Remove assertion.
12900 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12901 * gimple-ssa-sprintf.c (format_directive): Same.
12902 (parse_directive): Same.
12903 (sprintf_dom_walker::compute_format_length): Same.
12904 (try_substitute_return_value): Same.
12906 2018-01-03 Jeff Law <law@redhat.com>
12908 PR middle-end/83654
12909 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12910 non-constant residual for zero at runtime and avoid probing in
12911 that case. Reorganize code for trailing problem to mirror handling
12914 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12916 PR tree-optimization/83501
12917 * tree-ssa-strlen.c (get_string_cst): New.
12918 (handle_char_store): Call get_string_cst.
12920 2018-01-03 Martin Liska <mliska@suse.cz>
12922 PR tree-optimization/83593
12923 * tree-ssa-strlen.c: Include tree-cfg.h.
12924 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12925 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12926 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12928 (strlen_dom_walker::before_dom_children): Call
12929 gimple_purge_dead_eh_edges. Dump tranformation with details
12931 (strlen_dom_walker::before_dom_children): Update call by adding
12932 new argument cleanup_eh.
12933 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12935 2018-01-03 Martin Liska <mliska@suse.cz>
12938 * cif-code.def (VARIADIC_THUNK): New enum value.
12939 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12942 2018-01-03 Jan Beulich <jbeulich@suse.com>
12944 * sse.md (mov<mode>_internal): Tighten condition for when to use
12945 vmovdqu<ssescalarsize> for TI and OI modes.
12947 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12949 Update copyright years.
12951 2018-01-03 Martin Liska <mliska@suse.cz>
12954 * ipa-visibility.c (function_and_variable_visibility): Skip
12955 functions with noipa attribure.
12957 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12959 * gcc.c (process_command): Update copyright notice dates.
12960 * gcov-dump.c (print_version): Ditto.
12961 * gcov.c (print_version): Ditto.
12962 * gcov-tool.c (print_version): Ditto.
12963 * gengtype.c (create_file): Ditto.
12964 * doc/cpp.texi: Bump @copying's copyright year.
12965 * doc/cppinternals.texi: Ditto.
12966 * doc/gcc.texi: Ditto.
12967 * doc/gccint.texi: Ditto.
12968 * doc/gcov.texi: Ditto.
12969 * doc/install.texi: Ditto.
12970 * doc/invoke.texi: Ditto.
12972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12974 * vector-builder.h (vector_builder::m_full_nelts): Change from
12975 unsigned int to poly_uint64.
12976 (vector_builder::full_nelts): Update prototype accordingly.
12977 (vector_builder::new_vector): Likewise.
12978 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12979 (vector_builder::operator ==): Likewise.
12980 (vector_builder::finalize): Likewise.
12981 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12982 Take the number of elements as a poly_uint64 rather than an
12984 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12985 from unsigned int to poly_uint64.
12986 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12987 (vec_perm_indices::new_vector): Likewise.
12988 (vec_perm_indices::length): Likewise.
12989 (vec_perm_indices::nelts_per_input): Likewise.
12990 (vec_perm_indices::input_nelts): Likewise.
12991 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12992 number of elements per input as a poly_uint64 rather than an
12993 unsigned int. Use the original encoding for variable-length
12994 vectors, rather than clamping each individual element.
12995 For the second and subsequent elements in each pattern,
12996 clamp the step and base before clamping their sum.
12997 (vec_perm_indices::series_p): Handle polynomial element counts.
12998 (vec_perm_indices::all_in_range_p): Likewise.
12999 (vec_perm_indices_to_tree): Likewise.
13000 (vec_perm_indices_to_rtx): Likewise.
13001 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13002 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13003 (tree_vector_builder::new_binary_operation): Handle polynomial
13004 element counts. Return false if we need to know the number
13005 of elements at compile time.
13006 * fold-const.c (fold_vec_perm): Punt if the number of elements
13007 isn't known at compile time.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13011 * vec-perm-indices.h (vec_perm_builder): Change element type
13012 from HOST_WIDE_INT to poly_int64.
13013 (vec_perm_indices::element_type): Update accordingly.
13014 (vec_perm_indices::clamp): Handle polynomial element_types.
13015 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13016 (vec_perm_indices::all_in_range_p): Likewise.
13017 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13019 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13020 polynomial vec_perm_indices element types.
13021 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13022 * fold-const.c (fold_vec_perm): Likewise.
13023 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13024 * tree-vect-generic.c (lower_vec_perm): Likewise.
13025 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13026 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13027 element type to HOST_WIDE_INT.
13029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13030 Alan Hayward <alan.hayward@arm.com>
13031 David Sherwood <david.sherwood@arm.com>
13033 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13034 rather than an int. Use plus_constant.
13035 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13036 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13039 Alan Hayward <alan.hayward@arm.com>
13040 David Sherwood <david.sherwood@arm.com>
13042 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13043 a HOST_WIDE_INT to a poly_int64.
13045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13046 Alan Hayward <alan.hayward@arm.com>
13047 David Sherwood <david.sherwood@arm.com>
13049 * calls.c (load_register_parameters): Cope with polynomial
13050 mode sizes. Require a constant size for BLKmode parameters
13051 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13052 forces a parameter to be padded at the lsb end in order to
13053 fill a complete number of words, require the parameter size
13054 to be ordered wrt UNITS_PER_WORD.
13056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13057 Alan Hayward <alan.hayward@arm.com>
13058 David Sherwood <david.sherwood@arm.com>
13060 * reload1.c (spill_stack_slot_width): Change element type
13061 from unsigned int to poly_uint64_pod.
13062 (alter_reg): Treat mode sizes as polynomial.
13064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13065 Alan Hayward <alan.hayward@arm.com>
13066 David Sherwood <david.sherwood@arm.com>
13068 * reload.c (complex_word_subreg_p): New function.
13069 (reload_inner_reg_of_subreg, push_reload): Use it.
13071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13072 Alan Hayward <alan.hayward@arm.com>
13073 David Sherwood <david.sherwood@arm.com>
13075 * lra-constraints.c (process_alt_operands): Reject matched
13076 operands whose sizes aren't ordered.
13077 (match_reload): Refer to this check here.
13079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13080 Alan Hayward <alan.hayward@arm.com>
13081 David Sherwood <david.sherwood@arm.com>
13083 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13084 that the mode size is in the set {1, 2, 4, 8, 16}.
13086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13087 Alan Hayward <alan.hayward@arm.com>
13088 David Sherwood <david.sherwood@arm.com>
13090 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13091 Use plus_constant instead of gen_rtx_PLUS.
13093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13094 Alan Hayward <alan.hayward@arm.com>
13095 David Sherwood <david.sherwood@arm.com>
13097 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13098 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13099 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13100 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13101 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13102 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13103 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13104 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13105 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13106 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13108 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13109 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13110 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13111 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13112 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13113 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13114 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13115 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13116 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13117 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13119 * expr.c (emit_move_resolve_push): Treat the input and result
13120 of PUSH_ROUNDING as a poly_int64.
13121 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13122 (emit_push_insn): Likewise.
13123 * lra-eliminations.c (mark_not_eliminable): Likewise.
13124 * recog.c (push_operand): Likewise.
13125 * reload1.c (elimination_effects): Likewise.
13126 * rtlanal.c (nonzero_bits1): Likewise.
13127 * calls.c (store_one_arg): Likewise. Require the padding to be
13128 known at compile time.
13130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13131 Alan Hayward <alan.hayward@arm.com>
13132 David Sherwood <david.sherwood@arm.com>
13134 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13135 Use plus_constant instead of gen_rtx_PLUS.
13137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13138 Alan Hayward <alan.hayward@arm.com>
13139 David Sherwood <david.sherwood@arm.com>
13141 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13142 rather than an int.
13144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13149 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13150 via stack temporaries. Treat the mode size as polynomial too.
13152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13153 Alan Hayward <alan.hayward@arm.com>
13154 David Sherwood <david.sherwood@arm.com>
13156 * expr.c (expand_expr_real_2): When handling conversions involving
13157 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13158 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13159 as a poly_uint64 too.
13161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13162 Alan Hayward <alan.hayward@arm.com>
13163 David Sherwood <david.sherwood@arm.com>
13165 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13168 Alan Hayward <alan.hayward@arm.com>
13169 David Sherwood <david.sherwood@arm.com>
13171 * combine.c (can_change_dest_mode): Handle polynomial
13172 REGMODE_NATURAL_SIZE.
13173 * expmed.c (store_bit_field_1): Likewise.
13174 * expr.c (store_constructor): Likewise.
13175 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13176 and polynomial REGMODE_NATURAL_SIZE.
13177 (gen_lowpart_common): Likewise.
13178 * reginfo.c (record_subregs_of_mode): Likewise.
13179 * rtlanal.c (read_modify_subreg_p): Likewise.
13181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13182 Alan Hayward <alan.hayward@arm.com>
13183 David Sherwood <david.sherwood@arm.com>
13185 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13186 numbers of elements.
13188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13189 Alan Hayward <alan.hayward@arm.com>
13190 David Sherwood <david.sherwood@arm.com>
13192 * match.pd: Cope with polynomial numbers of vector elements.
13194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13195 Alan Hayward <alan.hayward@arm.com>
13196 David Sherwood <david.sherwood@arm.com>
13198 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13199 in a POINTER_PLUS_EXPR.
13201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13202 Alan Hayward <alan.hayward@arm.com>
13203 David Sherwood <david.sherwood@arm.com>
13205 * omp-simd-clone.c (simd_clone_subparts): New function.
13206 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13207 (ipa_simd_modify_function_body): Likewise.
13209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13210 Alan Hayward <alan.hayward@arm.com>
13211 David Sherwood <david.sherwood@arm.com>
13213 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13214 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13215 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13216 (expand_vector_condition, vector_element): Likewise.
13217 (subparts_gt): New function.
13218 (get_compute_type): Use subparts_gt.
13219 (count_type_subparts): Delete.
13220 (expand_vector_operations_1): Use subparts_gt instead of
13221 count_type_subparts.
13223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13224 Alan Hayward <alan.hayward@arm.com>
13225 David Sherwood <david.sherwood@arm.com>
13227 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13228 (vect_compile_time_alias): ...this new function. Do the calculation
13229 on poly_ints rather than trees.
13230 (vect_prune_runtime_alias_test_list): Update call accordingly.
13232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13233 Alan Hayward <alan.hayward@arm.com>
13234 David Sherwood <david.sherwood@arm.com>
13236 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13238 (vect_schedule_slp_instance): Likewise.
13240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13241 Alan Hayward <alan.hayward@arm.com>
13242 David Sherwood <david.sherwood@arm.com>
13244 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13245 constant and extern definitions for variable-length vectors.
13246 (vect_get_constant_vectors): Note that the number of units
13247 is known to be constant.
13249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13250 Alan Hayward <alan.hayward@arm.com>
13251 David Sherwood <david.sherwood@arm.com>
13253 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13254 of units as polynomial. Choose between WIDE and NARROW based
13257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13258 Alan Hayward <alan.hayward@arm.com>
13259 David Sherwood <david.sherwood@arm.com>
13261 * tree-vect-stmts.c (simd_clone_subparts): New function.
13262 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13265 Alan Hayward <alan.hayward@arm.com>
13266 David Sherwood <david.sherwood@arm.com>
13268 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13269 vectors as polynomial. Use build_index_vector for
13270 IFN_GOMP_SIMD_LANE.
13272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13273 Alan Hayward <alan.hayward@arm.com>
13274 David Sherwood <david.sherwood@arm.com>
13276 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13277 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13278 for variable-length vectors.
13279 (vectorizable_mask_load_store): Treat the number of units as
13280 polynomial, asserting that it is constant if the condition has
13281 already been enforced.
13282 (vectorizable_store, vectorizable_load): Likewise.
13284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13285 Alan Hayward <alan.hayward@arm.com>
13286 David Sherwood <david.sherwood@arm.com>
13288 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13289 of units as polynomial. Punt if we can't tell at compile time
13290 which vector contains the final result.
13292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13293 Alan Hayward <alan.hayward@arm.com>
13294 David Sherwood <david.sherwood@arm.com>
13296 * tree-vect-loop.c (vectorizable_induction): Treat the number
13297 of units as polynomial. Punt on SLP inductions. Use an integer
13298 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13299 cast of such a series for variable-length floating-point
13302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13303 Alan Hayward <alan.hayward@arm.com>
13304 David Sherwood <david.sherwood@arm.com>
13306 * tree.h (build_index_vector): Declare.
13307 * tree.c (build_index_vector): New function.
13308 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13309 of units as polynomial, forcibly converting it to a constant if
13310 vectorizable_reduction has already enforced the condition.
13311 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13312 to create a {1,2,3,...} vector.
13313 (vectorizable_reduction): Treat the number of units as polynomial.
13314 Choose vectype_in based on the largest scalar element size rather
13315 than the smallest number of units. Enforce the restrictions
13318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13319 Alan Hayward <alan.hayward@arm.com>
13320 David Sherwood <david.sherwood@arm.com>
13322 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13323 number of units as polynomial.
13325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13326 Alan Hayward <alan.hayward@arm.com>
13327 David Sherwood <david.sherwood@arm.com>
13329 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13330 * target.def (autovectorize_vector_sizes): Return the vector sizes
13331 by pointer, using vector_sizes rather than a bitmask.
13332 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13333 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13334 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13336 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13337 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13338 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13339 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13340 * omp-general.c (omp_max_vf): Likewise.
13341 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13342 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13343 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13344 * tree-vect-slp.c (vect_slp_bb): Likewise.
13345 * doc/tm.texi: Regenerate.
13346 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13348 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13349 the vector size as a poly_uint64 rather than an unsigned int.
13350 (current_vector_size): Change from an unsigned int to a poly_uint64.
13351 (get_vectype_for_scalar_type): Update accordingly.
13352 * tree.h (build_truth_vector_type): Take the size and number of
13353 units as a poly_uint64 rather than an unsigned int.
13354 (build_vector_type): Add a temporary overload that takes
13355 the number of units as a poly_uint64 rather than an unsigned int.
13356 * tree.c (make_vector_type): Likewise.
13357 (build_truth_vector_type): Take the number of units as a poly_uint64
13358 rather than an unsigned int.
13360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13361 Alan Hayward <alan.hayward@arm.com>
13362 David Sherwood <david.sherwood@arm.com>
13364 * target.def (get_mask_mode): Take the number of units and length
13365 as poly_uint64s rather than unsigned ints.
13366 * targhooks.h (default_get_mask_mode): Update accordingly.
13367 * targhooks.c (default_get_mask_mode): Likewise.
13368 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13369 * doc/tm.texi: Regenerate.
13371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13372 Alan Hayward <alan.hayward@arm.com>
13373 David Sherwood <david.sherwood@arm.com>
13375 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13376 * omp-general.c (omp_max_vf): Likewise.
13377 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13378 (expand_omp_simd): Handle polynomial safelen.
13379 * omp-low.c (omplow_simd_context): Add a default constructor.
13380 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13381 (lower_rec_simd_input_clauses): Update accordingly.
13382 (lower_rec_input_clauses): Likewise.
13384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13385 Alan Hayward <alan.hayward@arm.com>
13386 David Sherwood <david.sherwood@arm.com>
13388 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13389 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13390 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13391 (vect_analyze_slp_cost): Likewise.
13392 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13393 (vect_model_load_cost): Likewise.
13395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13396 Alan Hayward <alan.hayward@arm.com>
13397 David Sherwood <david.sherwood@arm.com>
13399 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13400 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13401 from an unsigned int * to a poly_uint64_pod *.
13402 (calculate_unrolling_factor): New function.
13403 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13406 Alan Hayward <alan.hayward@arm.com>
13407 David Sherwood <david.sherwood@arm.com>
13409 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13410 from an unsigned int to a poly_uint64.
13411 (_loop_vec_info::slp_unrolling_factor): Likewise.
13412 (_loop_vec_info::vectorization_factor): Change from an int
13414 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13415 (vect_get_num_vectors): New function.
13416 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13417 (vect_get_num_copies): Use vect_get_num_vectors.
13418 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13419 to an unsigned int *.
13420 (vect_analyze_data_refs): Change min_vf from an int * to a
13422 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13423 than an unsigned HOST_WIDE_INT.
13424 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13425 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13426 to an unsigned int *.
13427 (vect_analyze_data_ref_dependences): Likewise.
13428 (vect_compute_data_ref_alignment): Handle polynomial vf.
13429 (vect_enhance_data_refs_alignment): Likewise.
13430 (vect_prune_runtime_alias_test_list): Likewise.
13431 (vect_shift_permute_load_chain): Likewise.
13432 (vect_supportable_dr_alignment): Likewise.
13433 (dependence_distance_ge_vf): Take the vectorization factor as a
13434 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13435 (vect_analyze_data_refs): Change min_vf from an int * to a
13437 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13438 vfm1 as a poly_uint64 rather than an int. Make the same change
13439 for the returned bound_scalar.
13440 (vect_gen_vector_loop_niters): Handle polynomial vf.
13441 (vect_do_peeling): Likewise. Update call to
13442 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13443 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13445 * tree-vect-loop.c (vect_determine_vectorization_factor)
13446 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13447 (vect_get_known_peeling_cost): Likewise.
13448 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13449 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13450 (vect_transform_loop): Likewise. Use the lowest possible VF when
13451 updating the upper bounds of the loop.
13452 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13453 rather than an int.
13454 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13455 polynomial unroll factors.
13456 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13457 (vect_make_slp_decision): Likewise.
13458 (vect_supported_load_permutation_p): Likewise, and polynomial
13460 (vect_analyze_slp_cost): Handle polynomial vf.
13461 (vect_slp_analyze_node_operations): Likewise.
13462 (vect_slp_analyze_bb_1): Likewise.
13463 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13464 than an unsigned HOST_WIDE_INT.
13465 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13466 (vectorizable_load): Handle polynomial vf.
13467 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13469 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13472 Alan Hayward <alan.hayward@arm.com>
13473 David Sherwood <david.sherwood@arm.com>
13475 * match.pd: Handle bit operations involving three constants
13476 and try to fold one pair.
13478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13480 * tree-vect-loop-manip.c: Include gimple-fold.h.
13481 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13482 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13483 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13484 Add a path that uses a step of VF instead of 1, but disable it
13486 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13487 and niters_no_overflow parameters. Update calls to
13488 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13489 Create a new SSA name if the latter choses to use a ste other
13490 than zero, and return it via niters_vector_mult_vf_var.
13491 * tree-vect-loop.c (vect_transform_loop): Update calls to
13492 vect_do_peeling, vect_gen_vector_loop_niters and
13493 slpeel_make_loop_iterate_ntimes.
13494 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13495 (vect_gen_vector_loop_niters): Update declarations after above changes.
13497 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13499 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13500 128-bit round to integer instructions.
13501 (ceil<mode>2): Likewise.
13502 (btrunc<mode>2): Likewise.
13503 (round<mode>2): Likewise.
13505 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13507 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13508 unaligned VSX load/store on P8/P9.
13509 (expand_block_clear): Allow the use of unaligned VSX
13510 load/store on P8/P9.
13512 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13514 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13516 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13517 swap associated with both a load and a store.
13519 2018-01-02 Andrew Waterman <andrew@sifive.com>
13521 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13522 * config/riscv/riscv.md (clear_cache): Use it.
13524 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13526 * web.c: Remove out-of-date comment.
13528 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13530 * expr.c (fixup_args_size_notes): Check that any existing
13531 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13532 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13533 (emit_single_push_insn): ...here.
13535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13537 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13538 (const_vector_encoded_nelts): New function.
13539 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13540 (const_vector_int_elt, const_vector_elt): Declare.
13541 * emit-rtl.c (const_vector_int_elt_1): New function.
13542 (const_vector_elt): Likewise.
13543 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13544 of CONST_VECTOR_ELT.
13546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13548 * expr.c: Include rtx-vector-builder.h.
13549 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13550 directly on the tree encoding.
13551 (const_vector_from_tree): Likewise.
13552 * optabs.c: Include rtx-vector-builder.h.
13553 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13554 sequence of "u" values.
13555 * vec-perm-indices.c: Include rtx-vector-builder.h.
13556 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13557 directly on the vec_perm_indices encoding.
13559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13561 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13562 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13563 * rtx-vector-builder.h: New file.
13564 * rtx-vector-builder.c: Likewise.
13565 * rtl.h (rtx_def::u2): Add a const_vector field.
13566 (CONST_VECTOR_NPATTERNS): New macro.
13567 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13568 (CONST_VECTOR_DUPLICATE_P): Likewise.
13569 (CONST_VECTOR_STEPPED_P): Likewise.
13570 (CONST_VECTOR_ENCODED_ELT): Likewise.
13571 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13572 (unwrap_const_vec_duplicate): Likewise.
13573 (const_vec_series_p): Check for a non-duplicated vector encoding.
13574 Say that the function only returns true for integer vectors.
13575 * emit-rtl.c: Include rtx-vector-builder.h.
13576 (gen_const_vec_duplicate_1): Delete.
13577 (gen_const_vector): Call gen_const_vec_duplicate instead of
13578 gen_const_vec_duplicate_1.
13579 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13580 (gen_const_vec_duplicate): Use rtx_vector_builder.
13581 (gen_const_vec_series): Likewise.
13582 (gen_rtx_CONST_VECTOR): Likewise.
13583 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13584 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13585 Build a new vector rather than modifying a CONST_VECTOR in-place.
13586 (handle_special_swappables): Update call accordingly.
13587 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13588 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13589 Build a new vector rather than modifying a CONST_VECTOR in-place.
13590 (handle_special_swappables): Update call accordingly.
13592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13594 * simplify-rtx.c (simplify_const_binary_operation): Use
13595 CONST_VECTOR_ELT instead of XVECEXP.
13597 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13599 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13600 the selector elements to be different from the data elements
13601 if the selector is a VECTOR_CST.
13602 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13603 ssizetype for the selector.
13605 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13607 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13608 before testing each element individually.
13609 * tree-vect-generic.c (lower_vec_perm): Likewise.
13611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13613 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13614 * selftest-run-tests.c (selftest::run_tests): Call it.
13615 * vector-builder.h (vector_builder::operator ==): New function.
13616 (vector_builder::operator !=): Likewise.
13617 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13618 (vec_perm_indices::all_from_input_p): New function.
13619 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13620 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13621 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13622 instead of reading the VECTOR_CST directly. Detect whether both
13623 vector inputs are the same before constructing the vec_perm_indices,
13624 and update the number of inputs argument accordingly. Use the
13625 utility functions added above. Only construct sel2 if we need to.
13627 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13629 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13630 the broadcast of the low byte.
13631 (expand_mult_highpart): Use an explicit encoding for the permutes.
13632 * optabs-query.c (can_mult_highpart_p): Likewise.
13633 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13634 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13635 (vectorizable_bswap): Likewise.
13636 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13637 explicit encoding for the power-of-2 permutes.
13638 (vect_permute_store_chain): Likewise.
13639 (vect_grouped_load_supported): Likewise.
13640 (vect_permute_load_chain): Likewise.
13642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13644 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13645 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13646 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13647 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13648 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13649 (vect_gen_perm_mask_any): Likewise.
13651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13653 * int-vector-builder.h: New file.
13654 * vec-perm-indices.h: Include int-vector-builder.h.
13655 (vec_perm_indices): Redefine as an int_vector_builder.
13656 (auto_vec_perm_indices): Delete.
13657 (vec_perm_builder): Redefine as a stand-alone class.
13658 (vec_perm_indices::vec_perm_indices): New function.
13659 (vec_perm_indices::clamp): Likewise.
13660 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13661 (vec_perm_indices::new_vector): New function.
13662 (vec_perm_indices::new_expanded_vector): Update for new
13663 vec_perm_indices class.
13664 (vec_perm_indices::rotate_inputs): New function.
13665 (vec_perm_indices::all_in_range_p): Operate directly on the
13666 encoded form, without computing elided elements.
13667 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13668 encoding. Update for new vec_perm_indices class.
13669 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13670 the given vec_perm_builder.
13671 (expand_vec_perm_var): Update vec_perm_builder constructor.
13672 (expand_mult_highpart): Use vec_perm_builder instead of
13673 auto_vec_perm_indices.
13674 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13675 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13676 or double series encoding as appropriate.
13677 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13678 vec_perm_indices instead of auto_vec_perm_indices.
13679 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13680 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13681 (vect_permute_store_chain): Likewise.
13682 (vect_grouped_load_supported): Likewise.
13683 (vect_permute_load_chain): Likewise.
13684 (vect_shift_permute_load_chain): Likewise.
13685 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13686 (vect_transform_slp_perm_load): Likewise.
13687 (vect_schedule_slp_instance): Likewise.
13688 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13689 (vectorizable_mask_load_store): Likewise.
13690 (vectorizable_bswap): Likewise.
13691 (vectorizable_store): Likewise.
13692 (vectorizable_load): Likewise.
13693 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13694 vec_perm_indices instead of auto_vec_perm_indices. Use
13695 tree_to_vec_perm_builder to read the vector from a tree.
13696 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13697 vec_perm_builder instead of a vec_perm_indices.
13698 (have_whole_vector_shift): Use vec_perm_builder and
13699 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13700 truncation to calc_vec_perm_mask_for_shift.
13701 (vect_create_epilog_for_reduction): Likewise.
13702 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13703 from auto_vec_perm_indices to vec_perm_indices.
13704 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13705 instead of changing individual elements.
13706 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13707 the vector in d.perm.
13708 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13709 from auto_vec_perm_indices to vec_perm_indices.
13710 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13711 instead of changing individual elements.
13712 (arm_vectorize_vec_perm_const): Use new_vector to install
13713 the vector in d.perm.
13714 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13715 Update vec_perm_builder constructor.
13716 (rs6000_expand_interleave): Likewise.
13717 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13718 (rs6000_expand_interleave): Likewise.
13720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13722 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13723 to qimode could truncate the indices.
13724 * optabs.c (expand_vec_perm_var): Likewise.
13726 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13728 * Makefile.in (OBJS): Add vec-perm-indices.o.
13729 * vec-perm-indices.h: New file.
13730 * vec-perm-indices.c: Likewise.
13731 * target.h (vec_perm_indices): Replace with a forward class
13733 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13734 * optabs.h: Include vec-perm-indices.h.
13735 (expand_vec_perm): Delete.
13736 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13737 (expand_vec_perm_const): Declare.
13738 * target.def (vec_perm_const_ok): Replace with...
13739 (vec_perm_const): ...this new hook.
13740 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13741 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13742 * doc/tm.texi: Regenerate.
13743 * optabs.def (vec_perm_const): Delete.
13744 * doc/md.texi (vec_perm_const): Likewise.
13745 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13746 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13747 expand_vec_perm for constant permutation vectors. Assert that
13748 the mode of variable permutation vectors is the integer equivalent
13749 of the mode that is being permuted.
13750 * optabs-query.h (selector_fits_mode_p): Declare.
13751 * optabs-query.c: Include vec-perm-indices.h.
13752 (selector_fits_mode_p): New function.
13753 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13754 is defined, instead of checking whether the vec_perm_const_optab
13755 exists. Use targetm.vectorize.vec_perm_const instead of
13756 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13757 fit in the vector mode before using a variable permute.
13758 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13759 vec_perm_indices instead of an rtx.
13760 (expand_vec_perm): Replace with...
13761 (expand_vec_perm_const): ...this new function. Take the selector
13762 as a vec_perm_indices rather than an rtx. Also take the mode of
13763 the selector. Update call to shift_amt_for_vec_perm_mask.
13764 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13765 Use vec_perm_indices::new_expanded_vector to expand the original
13766 selector into bytes. Check whether the indices fit in the vector
13767 mode before using a variable permute.
13768 (expand_vec_perm_var): Make global.
13769 (expand_mult_highpart): Use expand_vec_perm_const.
13770 * fold-const.c: Includes vec-perm-indices.h.
13771 * tree-ssa-forwprop.c: Likewise.
13772 * tree-vect-data-refs.c: Likewise.
13773 * tree-vect-generic.c: Likewise.
13774 * tree-vect-loop.c: Likewise.
13775 * tree-vect-slp.c: Likewise.
13776 * tree-vect-stmts.c: Likewise.
13777 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13779 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13780 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13781 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13782 (aarch64_vectorize_vec_perm_const): ...this new function.
13783 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13784 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13785 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13786 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13787 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13788 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13789 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13791 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13792 check for NEON modes.
13793 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13794 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13795 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13796 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13798 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13799 the old VEC_PERM_CONST conditions.
13800 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13801 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13802 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13803 (ia64_vectorize_vec_perm_const_ok): Merge into...
13804 (ia64_vectorize_vec_perm_const): ...this new function.
13805 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13806 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13807 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13808 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13809 * config/mips/mips.c (mips_expand_vec_perm_const)
13810 (mips_vectorize_vec_perm_const_ok): Merge into...
13811 (mips_vectorize_vec_perm_const): ...this new function.
13812 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13813 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13814 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13815 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13816 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13817 (rs6000_expand_vec_perm_const): Delete.
13818 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13820 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13821 (altivec_expand_vec_perm_const_le): Take each operand individually.
13822 Operate on constant selectors rather than rtxes.
13823 (altivec_expand_vec_perm_const): Likewise. Update call to
13824 altivec_expand_vec_perm_const_le.
13825 (rs6000_expand_vec_perm_const): Delete.
13826 (rs6000_vectorize_vec_perm_const_ok): Delete.
13827 (rs6000_vectorize_vec_perm_const): New function.
13828 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13829 an element count and rtx array.
13830 (rs6000_expand_extract_even): Update call accordingly.
13831 (rs6000_expand_interleave): Likewise.
13832 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13833 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13834 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13835 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13836 (rs6000_expand_vec_perm_const): Delete.
13837 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13838 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13839 (altivec_expand_vec_perm_const_le): Take each operand individually.
13840 Operate on constant selectors rather than rtxes.
13841 (altivec_expand_vec_perm_const): Likewise. Update call to
13842 altivec_expand_vec_perm_const_le.
13843 (rs6000_expand_vec_perm_const): Delete.
13844 (rs6000_vectorize_vec_perm_const_ok): Delete.
13845 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13846 reference to the SPE evmerge intructions.
13847 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13848 an element count and rtx array.
13849 (rs6000_expand_extract_even): Update call accordingly.
13850 (rs6000_expand_interleave): Likewise.
13851 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13852 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13854 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13858 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13859 vector mode and that that mode matches the mode of the data
13861 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13862 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13863 directly using expand_vec_perm_1 when forcing selectors into
13865 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13869 * optabs-query.h (can_vec_perm_p): Delete.
13870 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13871 * optabs-query.c (can_vec_perm_p): Split into...
13872 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13873 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13874 particular selector is valid.
13875 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13876 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13877 (vect_grouped_load_supported): Likewise.
13878 (vect_shift_permute_load_chain): Likewise.
13879 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13880 (vect_transform_slp_perm_load): Likewise.
13881 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13882 (vectorizable_bswap): Likewise.
13883 (vect_gen_perm_mask_checked): Likewise.
13884 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13885 implementations of variable permutation vectors into account
13886 when deciding which selector to use.
13887 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13888 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13889 with a false third argument.
13890 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13891 to test whether the constant selector is valid and can_vec_perm_var_p
13892 to test whether a variable selector is valid.
13894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13896 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13897 * optabs-query.c (can_vec_perm_p): Likewise.
13898 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13899 instead of vec_perm_indices.
13900 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13901 (vect_gen_perm_mask_checked): Likewise,
13902 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13903 (vect_gen_perm_mask_checked): Likewise,
13905 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13907 * optabs-query.h (qimode_for_vec_perm): Declare.
13908 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13909 (qimode_for_vec_perm): ...this new function.
13910 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13912 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13914 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13915 does not have a conditional at the top.
13917 2018-01-02 Richard Biener <rguenther@suse.de>
13919 * ipa-inline.c (big_speedup_p): Fix expression.
13921 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13924 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13927 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13931 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13932 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13933 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13934 cond_taken_branch_cost 3->4.
13936 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13938 PR tree-optimization/83581
13939 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13940 TODO_cleanup_cfg if any changes have been made.
13942 PR middle-end/83608
13943 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13944 convert_modes if target mode has the right side, but different mode
13947 PR middle-end/83609
13948 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13949 last argument when extracting from CONCAT. If either from_real or
13950 from_imag is NULL, use expansion through memory. If result is not
13951 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13952 the parts directly to inner mode, if even that fails, use expansion
13955 PR middle-end/83623
13956 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13957 check for bswap in mode rather than HImode and use that in expand_unop
13960 Copyright (C) 2018 Free Software Foundation, Inc.
13962 Copying and distribution of this file, with or without modification,
13963 are permitted in any medium without royalty provided the copyright
13964 notice and this notice are preserved.