1 2018-05-23 Luis Machado <luis.machado@linaro.org>
3 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4 <prefetch_dynamic_strides>: New const bool field.
5 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6 prefetch_dynamic_strides.
7 (exynosm1_prefetch_tune): Likewise.
8 (thunderxt88_prefetch_tune): Likewise.
9 (thunderx_prefetch_tune): Likewise.
10 (thunderx2t99_prefetch_tune): Likewise.
11 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13 (aarch64_override_options_internal): Update to set
14 PARAM_PREFETCH_DYNAMIC_STRIDES.
15 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19 prefetch-dynamic-strides setting.
21 2018-05-23 Luis Machado <luis.machado@linaro.org>
23 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
24 <minimum_stride>: New const int field.
25 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
26 minimum_stride field defaulting to -1.
27 (exynosm1_prefetch_tune): Likewise.
28 (thunderxt88_prefetch_tune): Likewise.
29 (thunderx_prefetch_tune): Likewise.
30 (thunderx2t99_prefetch_tune): Likewise.
31 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
32 <default_opt_level>: Set to 3.
33 (aarch64_override_options_internal): Update to set
34 PARAM_PREFETCH_MINIMUM_STRIDE.
35 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
36 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
37 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
38 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
39 stride is constant and is below the minimum stride threshold.
41 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
43 * config/arm/arm-cpus.in (mode26): Delete.
44 (armv4): Delete mode26 reference.
45 * config/arm/arm.c (arm_configure_build_target): Delete use of
48 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
50 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
52 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
53 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
55 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
56 (floatunsdidf2): Ditto.
58 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
60 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
61 (fixuns_trunc<mode>si2_avx512f): Ditto.
62 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
63 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
64 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
66 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
68 PR rtl-optimization/79985
69 * df-scan.c (df_insn_refs_collect): Remove special case for
70 global registers and asm statements.
72 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
74 * extend.texi (Global Register Variables): Rewrite the bullet list.
75 Note that the register is available for allocation. Note that access
76 via inline asm must use constraints. Add note about async-signal
77 handlers. Remove paragraph about automagic register selection.
79 2018-05-23 Richard Biener <rguenther@suse.de>
81 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
82 of fixed offset from memset VN.
84 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
86 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
88 (alloc_cand_and_find_basis): Initialize first_interp field.
89 (slsr_process_mul): Modify first_interp field.
90 (slsr_process_add): Likewise.
91 (slsr_process_cast): Modify first_interp field for each new
93 (slsr_process_copy): Likewise.
94 (dump_candidate): Dump first_interp field.
95 (replace_mult_candidate): Process all interpretations, not just
97 (replace_rhs_if_not_dup): Likewise.
98 (replace_one_candidate): Likewise.
100 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
102 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
104 (aarch64_needs_frame_chain): New function.
105 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
107 2018-05-23 Sudakshina Das <sudi.das@arm.com>
110 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
111 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
112 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
113 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
114 as true for strict-align.
115 (aarch64_can_inline_p): Perform checks even when callee has no
116 attributes to check for strict alignment.
117 * doc/extend.texi (AArch64 Function Attributes): Document
119 * doc/invoke.texi: (AArch64 Options): Likewise.
121 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
123 PR tree-optimization/85853
124 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
125 the handling of the root of the node to...
126 (vect_slp_analyze_node_operations_1): ...this new function,
127 and run the whole thing with the child nodes' def types
128 set according to their SLP node's def type.
130 2018-05-23 Richard Biener <rguenther@suse.de>
133 * tree-data-ref.c (create_runtime_alias_checks): Defer
134 and ignore overflow warnings.
136 2018-05-23 Yury Gribov <tetra2005@gmail.com>
138 PR tree-optimization/85822
139 * tree-vrp.c (is_masked_range_test): Fix handling of negative
142 2018-05-23 Richard Biener <rguenther@suse.de>
144 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
145 memset constants via native_interpret_expr.
147 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
150 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
152 (cgraph_node::create_alias): Likewise.
153 (cgraph_node::get_availability): Check ifunc_resolver instead
154 of looking up ifunc attribute.
155 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
156 * varasm.c (do_assemble_alias): Likewise.
157 (assemble_alias): Likewise.
158 (default_binds_local_p_3): Likewise.
159 * cgraph.h (cgraph_node): Add ifunc_resolver.
160 (cgraph_node::only_called_directly_or_aliased_p): Return false
162 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
164 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
165 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
166 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
167 instead of looking up ifunc attribute.
169 2018-05-22 Luis Machado <luis.machado@linaro.org>
171 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
173 2018-05-22 Martin Sebor <msebor@redhat.com>
176 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
177 only when expasion succeeds.
178 (expand_builtin_strcmp): Same.
179 (expand_builtin_strncmp): Same.
181 2018-05-22 Martin Sebor <msebor@redhat.com>
183 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
185 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
186 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
188 * config/aarch64/aarch64-ldpstp.md: Replace uses of
189 aarch64_mem_pair_operand with memory_operand and delete operand swapping
191 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
192 Add check for legitimate_address.
193 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
194 (aarch64_swap_ldrstr_operands): New.
195 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
198 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
199 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
201 * config/aarch64/aarch64.md: New patterns to generate stp
203 (store_pair_sw, store_pair_dw): New patterns to generate stp for
204 single words and double words.
205 (load_pair_sw, load_pair_dw): Likewise.
206 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
208 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
210 * config/aarch64/aarch64-ldpstp.md: Modify peephole
211 for different mode ldpstp and add peephole for merged zero stores.
213 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
215 (aarch64_gen_store_pair): Rename calls to match new patterns.
216 (aarch64_gen_load_pair): Rename calls to match new patterns.
217 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
218 (load_pair<DREG:mode><DREG2:mode>): ... This.
219 (store_pair<mode>): Rename to...
220 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
221 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
223 (V_INT_EQUIV): Handle SImode.
224 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
227 2018-05-22 Martin Sebor <msebor@redhat.com>
230 * calls.c (maybe_warn_nonstring_arg): Use string length to set
231 or ajust the presumed bound on an operation to avoid unnecessary
234 2018-05-22 Martin Sebor <msebor@redhat.com>
236 PR tree-optimization/85826
237 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
238 assuming that a DECL necesarily has a constant size.
240 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
243 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
245 2018-05-22 Richard Biener <rguenther@suse.de>
247 PR tree-optimization/85834
248 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
249 non-constant and non-zero memset arguments.
251 2018-05-22 Martin Liska <mliska@suse.cz>
254 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
256 2018-05-22 Richard Biener <rguenther@suse.de>
258 PR tree-optimization/85863
259 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
260 comparisons when vectype is specified.
261 (vectorizable_condition): Do not specify vectype for
262 vect_is_simple_cond when SLP vectorizing.
264 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
267 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
268 define __ibm128 as long double.
269 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
270 as a distinct type when IEEE 128-bit support is enabled.
271 (init_float128_ieee): Fix up conversions between IFmode and IEEE
272 128-bit types to use the correct functions.
273 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
274 convert between 128-bit floating point types that have different
275 modes but the same representation, instead of using gen_lowpart to
277 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
279 (IFKF_reg): New attributes to give the register constraints for
281 (extend<mode>tf2_internal): New insns to mark an explicit
282 conversion between 128-bit floating point types that have a
283 different mode but share the same representation.
285 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
287 PR tree-optimization/85814
288 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
289 a null return from get_strinfo when unsharing the next
290 strinfo in the chain.
292 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
295 * varasm.c (weak_finish): Clean up weak_decls.
297 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
299 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
300 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
301 UNSPEC_UADALP values.
302 * config/aarch64/iterators.md (ABAL): New int iterator.
305 (sur): Add mappings for the above.
306 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
308 (aarch64_<sur>abal<mode>_4): Likewise.
309 (aarch64_<sur>adalp<mode>_3): Likewise.
310 (<sur>sadv16qi): New define_expand.
312 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
314 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
315 (*movdf_internal): Ditto.
316 (*rcpsf2_sse): Ditto.
317 (*rsqrtsf2_sse): Ditto.
318 (*sqrt<mode>2_sse): Ditto.
320 2018-05-21 Tamar Christina <tamar.christina@arm.com>
322 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
324 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
325 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
326 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
327 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
329 * config/aarch64/arm_neon.h: Likewise.
330 * config/aarch64/iterators.md (VQ_I): New.
332 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
334 * config.gcc: Add arc/t-multilib-linux to tmake_file for
336 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
339 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/constraints.md (S): New constraint.
342 * config/nds32/nds32.md (call_internal): Use constraint S.
343 (call_value_internal): Likewise.
344 (sibcall_internal): Likewise.
345 (sibcall_value_internal): Likewise.
347 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
348 Chung-Ju Wu <jasonwucj@gmail.com>
350 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
353 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
354 Chung-Ju Wu <jasonwucj@gmail.com>
356 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
357 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
358 (nds32_rtx_costs_impl): Simplify.
359 (nds32_address_cost_impl): Simplify.
360 (nds32_init_rtx_costs): New function.
361 (nds32_rtx_costs_speed_prefer): Likewise.
362 (nds32_rtx_costs_size_prefer): Likewise.
363 (nds32_address_cost_speed_prefer): Likewise.
364 (nds32_address_cost_speed_fwprop): Likewise.
365 (nds32_address_cost_size_prefer): Likewise.
366 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
367 * config/nds32/nds32.c (nds32_option_override): Use
368 nds32_init_rtx_costs function.
370 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
372 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
373 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
374 (TARGET_PIPELINE_N8): Likewise.
375 (TARGET_PIPELINE_N10): Likewise.
376 (TARGET_PIPELINE_N13): Likewise.
377 (TARGET_PIPELINE_GRAYWOLF): Likewise.
379 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
381 * config/nds32/nds32-fpu.md: Update copyright year.
383 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
385 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
387 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
389 * config/nds32/nds32.c
390 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
391 * config/nds32/nds32.opt (minline-asm-r15): New option.
393 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
395 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
397 * config/nds32/nds32.md (abssi2): New pattern.
399 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
401 * config/i386/i386.md (rex64namesuffix): New mode attribute.
402 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
403 Merge insn pattern from sse_cvtsi2ss<round_name> and
404 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
405 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
406 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
407 using SWI48 mode iterator.
408 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
409 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
410 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
411 pattern from sse_cvttss2si<round_saeonly_name>
412 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
413 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
414 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
415 using SWI48 mode iterator.
416 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
417 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
418 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
419 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
420 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
421 using SWI48 mode iterator.
422 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
423 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
424 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
425 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
426 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
428 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
429 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
430 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
431 pattern from sse_cvttsd2si<round_saeonly_name>
432 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
434 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
436 * config/nds32/nds32-md-auxiliary.c
437 (nds32_valid_smw_lwm_base_p): Refine.
438 (nds32_output_smw_single_word): Refine.
439 (nds32_output_smw_double_word): New.
440 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
442 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
444 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
445 (nds32_output_stack_pop): Refine.
446 (nds32_expand_unaligned_load): Refine.
447 (nds32_expand_unaligned_store): Refine.
449 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
450 Chung-Ju Wu <jasonwucj@gmail.com>
452 * config/nds32/constants.md: Add TP_REGNUM constant.
453 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
454 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
456 * config/nds32/nds32-doubleword.md: Consider flag_pic.
457 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
458 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
459 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
460 and PIC code generation.
461 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
463 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
465 * config/nds32/nds32.md: Support TLS and PIC.
466 * config/nds32/nds32.c: Support TLS and PIC.
467 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
468 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
471 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
473 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
476 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
477 Chung-Ju Wu <jasonwucj@gmail.com>
479 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
480 * config/nds32/nds32-md-auxiliary.c
481 (symbolic_reference_mentioned_p): New.
482 (nds32_legitimize_ict_address): New.
483 (nds32_expand_ict_move): New.
484 (nds32_indirect_call_referenced_p): New.
485 (nds32_symbol_binds_local_p): Delete.
486 (nds32_long_call_p): Modify.
487 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
488 * config/nds32/nds32-protos.h
489 (symbolic_reference_mentioned_p): Declare.
490 (nds32_legitimize_ict_address): Declare.
491 (nds32_expand_ict_move): Declare.
492 (nds32_indirect_call_referenced_p): Declare.
493 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
494 (nds32_relax_group): Use nds32_ict_const_p as condition.
495 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
496 (nds32_asm_file_start): Output ict_model directive in asm code.
497 (nds32_legitimate_address_p): Consider indirect call.
498 (nds32_print_operand): Consider indirect call.
499 (nds32_print_operand_address): Consider indirect call.
500 (nds32_insert_attributes): Handle "indirect_call" attribute.
501 (TARGET_LEGITIMATE_ADDRESS_P): Define.
502 (TARGET_LEGITIMATE_CONSTANT_P): Define.
503 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
504 (TARGET_DELEGITIMIZE_ADDRESS): Define.
505 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
506 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
507 (TARGET_ICT_MODEL_SMALL): Define.
508 (TARGET_ICT_MODEL_LARGE): Define.
509 * config/nds32/nds32.md (movsi): Consider ict model.
510 (call, call_value): Consider ict model.
511 (sibcall, sibcall_value): Consider ict model.
512 * config/nds32/nds32.opt (mict-model): New option.
513 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
516 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
517 Monk Chiang <sh.chiang04@gmail.com>
518 Jim Wilson <jimw@sifive.com>
520 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
521 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
522 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
523 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
524 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
525 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
526 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
527 compute save_libcall_adjustment properly.
528 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
529 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
530 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
531 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
532 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
533 (ABI_SPEC): Handle mabi=ilp32e.
534 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
536 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
537 <-march>: Add rv32e as an example.
539 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
542 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
543 (intra_create_variable_infos): Handle C++ constructors.
545 2018-05-18 Martin Liska <mliska@suse.cz>
547 * passes.def: Remove a redundant pass.
549 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
552 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
554 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
556 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
558 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
560 (armv2, armv2a, armv3, armv3m): Delete architectures.
561 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
562 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
563 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
565 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
566 (*mulsidi3adddi): Likewise.
567 (mulsidi3): Likewise.
568 (*mulsidi3_nov6): Likewise.
569 (umulsidi3): Likewise.
570 (umulsidi3_nov6): Likewise.
571 (umaddsidi4): Likewise.
572 (*umulsidi3adddi): Likewise.
573 (smulsi3_highpart): Likewise.
574 (*smulsi3_highpart_nov6): Likewise.
575 (umulsi3_highpart): Likewise.
576 (*umulsi3_highpart_nov6): Likewise.
577 * config/arm/arm.h (arm_arch3m): Delete.
578 * config/arm/arm.c (arm_arch3m): Delete.
579 (arm_option_override_internal): Update armv3-related comment.
580 (arm_configure_build_target): Delete use of isa_bit_mode32.
581 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
582 (arm_rtx_costs_internal): Delete check of arm_arch3m.
583 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
586 * config/arm/arm-protos.h (arm_arch3m): Delete.
587 * config/arm/arm-tables.opt: Regenerate.
588 * config/arm/arm-tune.md: Likewise.
589 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
590 deleted architectures.
592 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
594 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
595 (armv5t, armv5te): New features.
596 (ARMv5, ARMv5e): Delete fgroups.
597 (ARMv5t, ARMv5te): Adjust for above changes.
599 (armv5, armv5e): Delete arches.
600 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
602 (*call_reg_arm): Likewise.
603 (*call_value_reg_armv5): Likewise.
604 (*call_value_reg_arm): Likewise.
605 (*call_symbol): Likewise.
606 (*call_value_symbol): Likewise.
607 (*sibcall_insn): Likewise.
608 (*sibcall_value_insn): Likewise.
610 (prefetch): Likewise.
611 (define_split and define_peephole2 dependent on arm_arch5):
613 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
615 (TARGET_ARM_QBIT): Likewise.
616 (TARGET_DSP_MULTIPLY): Likewise.
617 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
618 (arm_arch5, arm_arch5e): Delete.
619 (arm_arch5t, arm_arch5te): Declare.
620 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
621 (arm_arch5t): Declare.
622 (arm_option_reconfigure_globals): Update for the above.
623 (arm_options_perform_arch_sanity_checks): Update comment, replace
624 use of arm_arch5 with arm_arch5t.
625 (use_return_insn): Likewise.
626 (arm_emit_call_insn): Likewise.
627 (output_return_instruction): Likewise.
628 (arm_final_prescan_insn): Likewise.
629 (arm_coproc_builtin_available): Likewise.
630 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
631 arm_arch5e with arm_arch5t and arm_arch5te.
632 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
633 (arm_arch5t, arm_arch5te): Declare.
634 * config/arm/arm-tables.opt: Regenerate.
635 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
636 * config/arm/t-multilib: Likewise.
637 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
638 instead of arm_arch5.
639 (*call_reg_thumb1): Likewise.
640 (*call_value_reg_thumb1_v5): Likewise.
641 (*call_value_reg_thumb1): Likewise.
642 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
644 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
646 2018-05-18 Martin Liska <mliska@suse.cz>
648 PR gcov-profile/84846
649 * doc/gcov.texi: Document -t option of gcov tool.
651 2018-05-18 Martin Liska <mliska@suse.cz>
653 PR gcov-profile/84846
654 * gcov.c (print_usage): Add new -t option.
655 (process_args): Handle the option.
656 (generate_results): Use stdout as output when requested by
659 2018-05-18 Martin Liska <mliska@suse.cz>
661 PR gcov-profile/84846
662 * coverage.c (coverage_init): Write PWD to .gcno file.
663 * doc/gcov.texi: Document how working directory is printed.
664 * gcov-dump.c (dump_gcov_file): Print PWD.
665 * gcov.c (output_intermediate_file): Likewise.
666 (read_graph_file): Read PWD string.
667 (output_lines): Print PWD.
669 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
672 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
673 for retval and return false if all args to phi are zero.
675 2018-05-18 Richard Biener <rguenther@suse.de>
677 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
679 (evrp_dom_walker::before_dom_children): Call it.
681 2018-05-18 Richard Biener <rguenther@suse.de>
683 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
684 results when processing array refs with variable index.
686 2018-05-18 Toon Moene <toon@moene.org>
688 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
689 directly after that of -floop-interchange. Indicate that both
690 options are enabled by default when specifying -O3.
692 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
694 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
695 iterator. Delete separate integer-mode vec_set<mode> expander.
696 (aarch64_simd_vec_setv2di): Delete.
697 (vec_setv2di): Delete.
698 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
699 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
700 the "w, r" alternative.
702 2018-05-18 Martin Liska <mliska@suse.cz>
704 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
705 * tree-pass.h (make_pass_lower_switch_O0): New function.
706 * tree-switch-conversion.c (node_has_low_bound): Remove.
707 (node_has_high_bound): Likewise.
708 (node_is_bounded): Likewise.
709 (class pass_lower_switch): Make it a template type and create
711 (pass_lower_switch::execute): Add template argument.
712 (make_pass_lower_switch): New function.
713 (make_pass_lower_switch_O0): New function.
714 (do_jump_if_equal): Remove.
715 (emit_case_nodes): Simplify to just handle all 3 cases and leave
716 all the hard work to tree optimization passes.
718 2018-05-18 Martin Liska <mliska@suse.cz>
720 * dbgcnt.c (limit_low): Renamed from limit.
721 (limit_high): New variable.
722 (dbg_cnt_is_enabled): Check for upper limit.
723 (dbg_cnt): Adjust dumping.
724 (dbg_cnt_set_limit_by_index): Add new argument for high
726 (dbg_cnt_set_limit_by_name): Likewise.
727 (dbg_cnt_process_single_pair): Parse new format.
728 (dbg_cnt_process_opt): Use strtok.
729 (dbg_cnt_list_all_counters): Remove 'value' and add
731 * doc/invoke.texi: Document changes.
733 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
735 * doc/sourcebuild.texi (scalar_all_fma): Document.
736 * tree.def (FMA_EXPR): Delete.
737 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
738 * internal-fn.c (ternary_direct): New macro.
739 (expand_ternary_optab_fn): Likewise.
740 (direct_ternary_optab_supported_p): Likewise.
741 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
742 * builtins.c (fold_builtin_fma): Delete.
743 (fold_builtin_3): Don't call it.
744 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
745 * expr.c (expand_expr_real_2): Likewise.
746 * fold-const.c (operand_equal_p): Likewise.
747 (fold_ternary_loc): Likewise.
748 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
749 * gimple.c (DEFTREECODE): Likewise.
750 * gimplify.c (gimplify_expr): Likewise.
751 * optabs-tree.c (optab_for_tree_code): Likewise.
752 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
753 * tree-eh.c (operation_could_trap_p): Likewise.
754 (stmt_could_throw_1_p): Likewise.
755 * tree-inline.c (estimate_operator_cost): Likewise.
756 * tree-pretty-print.c (dump_generic_node): Likewise.
757 (op_code_prio): Likewise.
758 * tree-ssa-loop-im.c (stmt_cost): Likewise.
759 * tree-ssa-operands.c (get_expr_operands): Likewise.
760 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
761 * fold-const-call.h (fold_fma): Delete.
762 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
763 CFN_FNMA and CFN_FNMS.
765 * genmatch.c (combined_fn): New enum.
766 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
767 (commutative_op): New function.
768 (commutate): Use it. Handle more than 2 operands.
769 (dt_operand::gen_gimple_expr): Use commutative_op.
770 (parser::parse_expr): Allow :c to be used with non-binary
771 operators if the commutative operand is known.
772 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
773 CFN_FMS, CFN_FNMA and CFN_FNMS.
774 (backprop::process_assign_use): Remove FMA_EXPR handling.
775 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
776 (gen_hsa_fma): New function.
777 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
778 IFN_FNMA and IFN_FNMS.
779 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
780 * gimple-fold.h (follow_all_ssa_edges): Declare.
781 * gimple-fold.c (follow_all_ssa_edges): New function.
782 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
783 gimple_build interface and use follow_all_ssa_edges to fold the result.
784 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
785 instead of checking for optabs directly.
786 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
787 rather than FMA_EXPRs.
788 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
789 call to IFN_FMA instead of an FMA_EXPR.
791 2018-05-17 Jim Wilson <jimw@sifive.com>
793 * expr.c (do_tablejump): When converting index to Pmode, if we have a
794 sign extended promoted subreg, and the range does not have the sign bit
795 set, then do a sign extend.
797 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
798 test, check for sign extended subreg and/or constant operands, and
799 do a sign extend in that case.
801 2018-05-17 Steve Ellcey <sellcey@cavium.com>
803 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
804 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
806 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
807 Change logics_shift_reg to logics_shift_imm.
808 (thunderx2t99_fp_loadpair_basic): Delete.
809 (thunderx2t99_fp_storepair_basic): Delete.
810 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
811 (thunderx2t99_asimd_polynomial): Delete.
812 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
813 and neon_fp_mul_d_scalar_q.
814 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
815 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
816 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
817 (thunderx2t99_asimd_lut): Add missing tbl types.
818 (thunderx2t99_asimd_ext): Delete.
819 (thunderx2t99_asimd_load1_1_mult): Delete.
820 (thunderx2t99_asimd_load1_2_mult): Delete.
821 (thunderx2t99_asimd_load1_ldp): New.
822 (thunderx2t99_asimd_load1): New.
823 (thunderx2t99_asimd_load2): Add missing *load2* types.
824 (thunderx2t99_asimd_load3): New.
825 (thunderx2t99_asimd_load4): New.
826 (thunderx2t99_asimd_store1_1_mult): Delete.
827 (thunderx2t99_asimd_store1_2_mult): Delete.
828 (thunderx2t99_asimd_store2_mult): Delete.
829 (thunderx2t99_asimd_store2_onelane): Delete.
830 (thunderx2t99_asimd_store_stp): New.
831 (thunderx2t99_asimd_store1): New.
832 (thunderx2t99_asimd_store2): New.
833 (thunderx2t99_asimd_store3): New.
834 (thunderx2t99_asimd_store4): New.
836 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
838 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
839 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
841 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
842 Segher Boessenkool <segher@kernel.crashing.org>
845 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
848 2018-05-17 Richard Biener <rguenther@suse.de>
850 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
851 for pruning loop and prune defs feeding only already visited PHIs.
853 2018-05-17 Richard Biener <rguenther@suse.de>
855 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
857 2018-05-17 Bin Cheng <bin.cheng@arm.com>
858 Richard Biener <rguenther@suse.de>
860 PR tree-optimization/85793
861 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
862 for VMAT_ELEMENTWISE.
864 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
866 * internal-fn.h (lookup_internal_fn): Declare
867 * internal-fn.c (lookup_internal_fn): New function.
868 * gimple.c (gimple_build_call_from_tree): Handle calls to
870 * gimple-pretty-print.c (dump_gimple_call): Print "." before
871 internal function names.
872 * tree-pretty-print.c (dump_generic_node): Likewise.
873 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
875 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
877 * gimple-fold.h (gimple_build): Make the function forms take
878 combined_fn rather than built_in_function.
879 (gimple_simplify): Likewise.
880 * gimple-match-head.c (gimple_simplify): Likewise.
881 * gimple-fold.c (gimple_build): Likewise.
882 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
883 rather than gimple_build_call_internal.
884 (get_initial_defs_for_reduction): Likewise.
885 (vect_create_epilog_for_reduction): Likewise.
886 (vectorizable_live_operation): Likewise.
888 2018-05-17 Martin Liska <mliska@suse.cz>
890 * gimple-ssa-sprintf.c (format_directive): Do not use
891 space in between 'G_' and '('.
893 2018-05-17 Jakub Jelinek <jakub@redhat.com>
896 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
897 even if the mask is not all ones.
900 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
902 (ix86_gimple_fold_builtin): Likewise.
905 * config/i386/i386.c: Include tree-vector-builder.h.
906 (ix86_vector_shift_count): New function.
907 (ix86_fold_builtin): Fold shift builtins by scalar count.
908 (ix86_gimple_fold_builtin): Likewise.
910 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
911 _mm512_setzero): New intrinsics.
913 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
914 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
916 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
917 code generation for cases where splatting a value is not useful.
918 * simplify-rtx.c (simplify_ternary_operation): Simplify
919 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
920 mode to a vec_concat.
922 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
924 * config.gcc: Support "goldmont-plus".
925 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
927 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
928 PROCESSOR_GOLDMONT_PLUS.
929 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
930 (processor_target_table): Add "goldmont-plus".
931 (PTA_GOLDMONT_PLUS): Define.
932 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
933 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
934 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
935 (fold_builtin_cpu): Add "goldmont-plus".
936 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
937 (ix86_option_override_internal): Add "goldmont-plus".
938 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
939 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
940 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
941 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
943 2018-05-17 Richard Biener <rguenther@suse.de>
945 PR tree-optimization/85757
946 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
947 remove defs that only feed that PHI from further processing.
949 2018-05-16 Jim Wilson <jimw@sifive.com>
951 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
953 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
955 2018-05-16 Mark Wielaard <mark@klomp.org>
957 * dwarf2out.c (count_index_strings): New function.
958 (output_indirect_strings): Call count_index_strings and generate
959 header for dwarf_version >= 5.
961 2018-05-16 Mark Wielaard <mark@klomp.org>
963 * dwarf2out.c (dwarf_FORM): New function.
964 (set_indirect_string): Use dwarf_FORM.
965 (reset_indirect_string): Likewise.
966 (size_of_die): Likewise.
967 (value_format): Likewise.
968 (output_die): Likewise.
969 (add_skeleton_AT_string): Likewise.
970 (output_macinfo_op): Likewise.
971 (index_string): Likewise.
972 (output_index_string_offset): Likewise.
973 (output_index_string): Likewise.
974 (count_index_strings): Likewise.
976 2018-05-16 Carl Love <cel@us.ibm.com>
978 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
979 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
981 2018-05-16 Martin Jambor <mjambor@suse.cz>
983 * ipa-prop.c (ipa_free_all_edge_args): Remove.
984 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
986 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
988 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
989 (fnma<mode>4): Likewise.
990 (fms<mode>4): Likewise.
991 (fnms<mode>4): Likewise.
992 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
993 (aarch64_fnma<mode>4): Likewise.
994 (aarch64_fms<mode>4): Likewise.
995 (aarch64_fnms<mode>4): Likewise.
996 (aarch64_fnmadd<mode>4): Likewise.
998 2018-05-16 Jason Merrill <jason@redhat.com>
1000 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1002 2018-05-16 Richard Biener <rguenther@suse.de>
1004 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1005 (dump_stmt_cost): Declare.
1006 (add_stmt_cost): Dump cost we add.
1007 (add_stmt_costs): New function.
1008 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1010 (vect_analyze_stmt): Adjust prototype.
1011 (vectorizable_condition): Likewise.
1012 (vectorizable_live_operation): Likewise.
1013 (vectorizable_reduction): Likewise.
1014 (vectorizable_induction): Likewise.
1015 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1016 cost vector to pass to vectorizable_ and record afterwards.
1017 (vect_model_reduction_cost): Take cost vector argument and adjust.
1018 (vect_model_induction_cost): Likewise.
1019 (vectorizable_reduction): Likewise.
1020 (vectorizable_induction): Likewise.
1021 (vectorizable_live_operation): Likewise.
1022 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1023 SLP_TREE_NUMBER_OF_VEC_STMTS.
1024 (vect_analyze_slp_cost_1): Remove.
1025 (vect_analyze_slp_cost): Likewise.
1026 (vect_slp_analyze_node_operations): Take visited args and
1027 a target cost vector. Avoid processing already visited stmt sets.
1028 (vect_slp_analyze_operations): Use a local cost vector to gather
1029 costs and register those of non-discarded instances.
1030 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1031 (vect_schedule_slp_instance): Remove copying of
1032 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1034 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1035 adding cost. Record cost entry location.
1036 (vect_prologue_cost_for_slp_op): Function to compute cost of
1037 a constant or invariant generated for SLP vect in the prologue,
1038 split out from vect_analyze_slp_cost_1.
1039 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1040 (vect_model_promotion_demotion_cost): Likewise.
1041 (vect_model_store_cost): Likewise, make static.
1042 (vect_model_load_cost): Likewise.
1043 (vectorizable_bswap): Add cost vector arg and adjust.
1044 (vectorizable_call): Likewise.
1045 (vectorizable_simd_clone_call): Likewise.
1046 (vectorizable_conversion): Likewise.
1047 (vectorizable_assignment): Likewise.
1048 (vectorizable_shift): Likewise.
1049 (vectorizable_operation): Likewise.
1050 (vectorizable_store): Likewise.
1051 (vectorizable_load): Likewise.
1052 (vectorizable_condition): Likewise.
1053 (vectorizable_comparison): Likewise.
1054 (can_vectorize_live_stmts): Likewise.
1055 (vect_analyze_stmt): Likewise.
1056 (vect_transform_stmt): Adjust calls to vectorizable_*.
1057 * tree-vectorizer.c: Include gimple-pretty-print.h.
1058 (dump_stmt_cost): New function.
1060 2018-05-16 Richard Biener <rguenther@suse.de>
1062 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1063 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1064 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1065 (check_name): New callback.
1066 (dse_classify_store): Track cycles via a visited bitmap of PHI
1067 defs and simplify handling of in-loop and across loop dead stores
1068 and properly fail for loop-variant refs. Handle byte-tracking with
1069 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1072 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1074 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1075 (vect_get_mask_type_for_stmt): Likewise.
1076 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1078 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1079 to determine the statement's vector type and the vector type that
1080 should be used for calculating nunits. Deal with cases in which
1081 the type has to be deferred.
1082 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1083 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1084 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1085 (vect_determine_vf_for_stmt): New functions, split out from...
1086 (vect_determine_vectorization_factor): ...here.
1087 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1088 (vect_get_mask_type_for_stmt): New functions, split out from
1089 vect_determine_vectorization_factor.
1091 2018-05-16 Richard Biener <rguenther@suse.de>
1093 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1094 verify the [VEC_]COND_EXPR embedded comparison.
1096 2018-05-15 Martin Sebor <msebor@redhat.com>
1098 PR tree-optimization/85753
1099 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1100 RECORD_TYPE in addition to ARRAY_TYPE.
1102 2018-05-15 Martin Sebor <msebor@redhat.com>
1105 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1107 2018-05-15 Richard Biener <rguenther@suse.de>
1109 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1110 add by_clobber_p one. Change algorithm to collect all defs
1111 representing uses we need to walk and try reducing them to
1112 a single one before failing.
1113 (dse_dom_walker::dse_optimize_stmt): Adjust.
1115 2018-05-13 Mark Wielaard <mark@klomp.org>
1117 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1118 (size_of_loc_descr): Likewise.
1119 (output_loc_operands): Likewise.
1120 (output_loc_operands_raw): Likewise.
1121 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1122 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1123 (hash_loc_operands): Likewise.
1124 (compare_loc_operands): Likewise.
1126 2018-05-14 Mark Wielaard <mark@klomp.org>
1128 * dwarf2out.c (count_index_addrs): New function.
1129 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1131 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1133 PR tree-optimization/83648
1134 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1135 return value as malloc candidate.
1137 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1140 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1141 as true in call to suggest_attribute.
1143 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1145 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1146 -mreadonly-in-sdata.
1148 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1150 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1152 (aarch64_crypto_aesd_fused): Likewise.
1154 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1156 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1157 (movsi_aarch64): Likewise.
1158 (load_pairsi): Likewise.
1159 (load_pairdi): Likewise.
1160 (store_pairsi): Likewise.
1161 (store_pairdi): Likewise.
1162 (load_pairsf): Likewise.
1163 (load_pairdf): Likewise.
1164 (store_pairsf): Likewise.
1165 (store_pairdf): Likewise.
1166 (zero_extend): Likewise.
1167 (trunc): Swap alternatives.
1168 (fcvt_target): Add '?' to prefer w over r.
1170 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1173 * config/i386/i386.md: Disallow non-commutative arithmetics in
1174 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1175 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1176 in the peephole2 before it.
1178 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1180 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1181 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1182 (ix86_handle_option): Handle -mcldemote.
1183 * config.gcc: New header.
1184 * config/i386/cldemoteintrin.h: New file.
1185 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1186 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1188 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1189 OPTION_MASK_ISA_CLDEMOTE.
1190 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1191 (ix86_valid_target_attribute_inner_p): Ditto.
1192 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1193 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1194 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1195 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1196 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1198 * config/i386/i386.opt: Add -mcldemote.
1199 * config/i386/x86intrin.h: New header.
1200 * doc/invoke.texi: Add -mcldemote.
1202 2018-05-14 Richard Biener <rguenther@suse.de>
1204 * doc/match-and-simplify.texi: Adjust :s documentation.
1206 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1208 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1209 intended memcpy size.
1210 (REORDER_45): Likewise.
1212 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1214 * sort.cc: New file.
1215 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1216 * vec.c (qsort_chk): Use gcc_qsort.
1217 * Makefile.in (OBJS-libcommon): Add sort.o.
1218 (build/sort.o): New target. Use it...
1219 (BUILD_RTL): ... here, and...
1220 (build/gencfn-macros): ... here, and...
1221 (build/genmatch): ... here.
1223 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1224 Chung-Ju Wu <jasonwucj@gmail.com>
1226 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1227 * config/nds32/nds32-graywolf.md: New file.
1228 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1231 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1232 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1233 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1234 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1235 * config/nds32/pipelines.md: Include n15 settings.
1237 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1238 Chung-Ju Wu <jasonwucj@gmail.com>
1240 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1241 * config/nds32/nds32-n13.md: New file.
1242 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1243 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1245 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1246 * config/nds32/nds32.md (pipeline_model): Add n13.
1247 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1248 * config/nds32/pipelines.md: Include n13 settings.
1250 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1251 Chung-Ju Wu <jasonwucj@gmail.com>
1253 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1254 * config/nds32/nds32-n10.md: New file.
1255 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1256 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1258 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1259 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1260 * config/nds32/nds32.md (pipeline_model): Add n10.
1261 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1262 * config/nds32/pipelines.md: Include n10 settings.
1264 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1265 Kito Cheng <kito.cheng@gmail.com>
1266 Chung-Ju Wu <jasonwucj@gmail.com>
1268 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1269 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1270 Add enum values for DSP extension instructions.
1271 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1273 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1274 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1276 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1277 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1278 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1279 * config/nds32/nds32-intrinsic.md: Likewise.
1280 * config/nds32/nds32_intrinsic.h: Likewise.
1281 * config/nds32/nds32-md-auxiliary.c: Likewise.
1282 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1283 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1284 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1285 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1286 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1287 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1288 TYPE_DMAC in switch statement.
1289 * config/nds32/nds32.c: New checking and implementation for DSP
1290 extension instructions.
1291 * config/nds32/nds32.h: Likewise.
1292 * config/nds32/nds32.md: Likewise.
1293 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1294 * config/nds32/predicates.md: Implement new predicates for DSP
1297 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1299 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1300 Reformat alternatives and attributes so it is easier to identify
1301 which constraints/attributes go with which instruction.
1302 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1303 (mov<mode>_softfloat32, FMOVE64): Likewise.
1304 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1305 (mov<mode>_softfloat64, FMOVE64): Likewise.
1307 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1309 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1311 (Basic PowerPC Built-in Functions): The new name of the
1312 subsection previously known as "PowerPC Built-in Functions".
1313 (Basic PowerPC Built-in Functions Available on all Configurations):
1315 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1316 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1317 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1318 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1320 2018-05-11 Martin Jambor <mjambor@suse.cz>
1323 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1326 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1329 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1331 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1333 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1334 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1335 (ix86_handle_option): Handle -mwaitpkg.
1336 * config.gcc: New header.
1337 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1338 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1339 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1341 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1342 OPTION_MASK_ISA_WAITPKG.
1343 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1344 (ix86_option_override_internal): Add PTA_WAITPKG.
1345 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1346 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1347 IX86_BUILTIN_TPAUSE.
1348 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1349 __builtin_ia32_umwait and __builtin_ia32_tpause.
1350 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1351 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1352 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1353 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1354 UNSPECV_TPAUSE): New.
1355 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1356 * config/i386/i386.opt: Add -mwaitpkg.
1357 * config/i386/waitpkgintrin.h: New file.
1358 * config/i386/x86intrin.h: New header.
1359 * doc/invoke.texi: Add -mwaitpkg.
1361 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1364 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1366 (cortex-m0): Use armv6s-m isa.
1367 (cortex-m0plus): Likewise.
1368 (cortex-m1): Likewise.
1369 (cortex-m0.small-multiply): Likewise.
1370 (cortex-m0plus.small-multiply): Likewise.
1371 (cortex-m1.small-multiply): Likewise.
1373 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1374 Jakub Jelinek <jakub@redhat.com>
1376 PR tree-optimization/85692
1377 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1378 source permute as well.
1380 2018-05-11 Martin Liska <mliska@suse.cz>
1383 * doc/extend.texi: Document LLVM style format for no_sanitize
1386 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1388 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1389 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1390 (mode_supports_vsx_dform_quad): Likewise.
1391 (mode_supports_vmx_dform): Move these functions to be next to the
1392 other mode_supports functions.
1393 (mode_supports_dq_form): Likewise.
1394 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1395 mode_supports_dq_form.
1396 (reg_offset_addressing_ok_p): Likewise.
1397 (offsettable_ok_by_alignment): Likewise.
1398 (rs6000_legitimate_offset_address_p): Likewise.
1399 (legitimate_lo_sum_address_p): Likewise.
1400 (rs6000_legitimize_address): Likewise.
1401 (rs6000_legitimize_reload_address): Likewise.
1402 (rs6000_secondary_reload_inner): Likewise.
1403 (rs6000_preferred_reload_class): Likewise.
1404 (rs6000_output_move_128bit): Likewise.
1406 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1408 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1409 Generate SImode target register for null target.
1410 <case IX86_BUILTIN_XGETBV>: Ditto.
1411 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1412 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1414 2018-05-10 Carl Love <cel@us.ibm.com>
1416 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1417 dcbtt and dcbtstt if operands[2] is 0.
1419 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1422 * config/i386/sse.md (usadv64qi): New expander.
1424 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1426 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1427 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1428 -maltivec=be support.
1429 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1430 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1431 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1432 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1433 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1434 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1435 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1436 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1437 altivec_vsumsws): Adjust.
1438 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1439 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1440 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1442 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1443 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1444 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1445 (altivec_lve<VI_char>x): Delete expand.
1446 (*altivec_lve<VI_char>x_internal): Rename to...
1447 (altivec_lve<VI_char>x): ... this.
1448 (altivec_lvxl_<mode>): Delete expand.
1449 (*altivec_lvxl_<mode>_internal): Rename to ...
1450 (altivec_lvxl_<mode>): ... this.
1451 (altivec_stvxl_<mode>): Delete expand.
1452 (*altivec_stvxl_<mode>_internal): Rename to ...
1453 (altivec_stvxl_<mode>): ... this.
1454 (altivec_stve<VI_char>x): Delete expand.
1455 (*altivec_stve<VI_char>x_internal): Rename to ...
1456 (altivec_stve<VI_char>x): ... this.
1457 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1458 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1459 reduc_plus_scal_<mode>): Adjust.
1460 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1462 (rs6000_cpu_cpp_builtins): Adjust.
1463 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1464 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1465 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1466 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1467 -maltivec=be support.
1468 (rs6000_split_vec_extract_var): Adjust.
1469 (rs6000_split_v4si_init): Adjust.
1470 (swap_selector_for_mode): Delete.
1471 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1472 altivec_expand_stvex_be): Delete.
1473 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1474 -maltivec=be support.
1475 (rs6000_gimple_fold_builtin): Ditto.
1476 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1478 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1479 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1480 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1481 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1482 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1483 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1484 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1485 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1486 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1487 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1488 anonymous split): Adjust.
1489 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1490 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1492 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1494 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1495 when --with-gxx-include-dir is also specified.
1496 * configure: Regenerate.
1498 2018-05-09 Jim Wilson <jimw@sifive.com>
1501 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1502 * config/riscv/t-withmultilib: New.
1503 * config/riscv/withmultilib.h: New.
1504 * doc/install.texi: Document RISC-V --with-multilib-list support.
1506 2018-05-09 Richard Biener <rguenther@suse.de>
1508 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1510 (vect_bb_vectorization_profitable_p): Adjust. Compute
1511 actual scalar cost using the cost vector and the add_stmt_cost
1514 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1516 PR rtl-optimization/85645
1517 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1518 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1520 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1522 PR rtl-optimization/85645
1523 * shrink-wrap.c (spread_components): Return a boolean saying if
1524 anything was changed.
1525 (try_shrink_wrapping_separate): Iterate spread_components until
1526 nothing changes anymore.
1528 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1530 PR rtl-optimization/85645
1531 * regrename.c (build_def_use): Also kill the chains that include the
1532 destination of a REG_CFA_REGISTER note.
1534 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1536 PR rtl-optimization/85645
1537 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1538 insn that has a REG_CFA_REGISTER note.
1540 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1542 * cfgexpand.c (expand_clobber): New function.
1543 (expand_gimple_stmt_1): Use it.
1544 * tree-vect-stmts.c (vect_clobber_variable): New function,
1546 (vectorizable_simd_clone_call): ...here.
1547 (vectorizable_store): Emit a clobber either side of an
1548 IFN_STORE_LANES sequence.
1549 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1551 2018-05-09 Tom de Vries <tom@codesourcery.com>
1554 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1555 (define_insn "trap_if_false"): Add exit after trap.
1557 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1559 PR rtl-optimization/85638
1560 * bb-reorder.c: Include common/common-target.h.
1561 (create_forwarder_block): New function extracted from...
1562 (fix_up_crossing_landing_pad): ...here. Rename into...
1563 (dw2_fix_up_crossing_landing_pad): ...this.
1564 (sjlj_fix_up_crossing_landing_pad): New function.
1565 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1566 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1567 from both partitions and exit the loop after one iteration.
1569 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1572 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1574 (Basic PowerPC Built-in Functions): The new name of the
1575 subsection previously known as "PowerPC Built-in Functions".
1576 (Basic PowerPC Built-in Functions Available on all Configurations):
1578 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1580 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1582 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1584 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1587 2018-05-08 Jim Wilson <jimw@sifive.com>
1589 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1590 (LD_EMUL_SUFFIX): New.
1591 (LINK_SPEC): Use it.
1593 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1595 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1597 (Basic PowerPC Built-in Functions): The new name of the
1598 subsection previously known as "PowerPC Built-in Functions".
1599 (Basic PowerPC Built-in Functions Available on all Configurations):
1601 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1603 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1605 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1607 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1610 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1613 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1614 after cmpelim optimization.
1616 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1618 * config.gcc: Support "goldmont".
1619 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1620 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1622 * config/i386/i386.c (m_GOLDMONT): Define.
1623 (processor_target_table): Add "goldmont".
1624 (PTA_GOLDMONT): Define.
1625 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1626 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1627 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1628 (fold_builtin_cpu): Add "goldmont".
1629 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1630 (ix86_option_override_internal): Add "goldmont".
1631 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1632 (processor_type): Add PROCESSOR_GOLDMONT.
1633 * config/i386/i386.md: Add CPU "glm".
1634 * config/i386/glm.md: New file.
1635 * config/i386/x86-tune.def: Add m_GOLDMONT.
1636 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1638 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1641 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1643 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1644 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1645 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1648 * config/i386/i386.c (ix86_fold_builtin): Handle
1649 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1652 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1653 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1655 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1658 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1659 (check_arch): Likewise.
1660 (check_fpu): Return the result rather than printing it.
1661 (end arch): Fix operator precedence.
1662 (end cpu): Likewise.
1663 (END): Print the result from check_fpu.
1665 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1666 Alan Hayward <alan.hayward@arm.com>
1667 David Sherwood <david.sherwood@arm.com>
1669 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1670 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1671 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1672 (*fcmuo<mode>_and): New patterns.
1674 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1676 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1677 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1678 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1679 (cmp_op, sve_imm_con): New code attributes.
1680 (SVE_COND_INT_CMP, imm_con): Delete.
1681 (cmp_op): Remove above unspecs from int attribute.
1682 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1684 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1685 comparison-specific unspecs.
1686 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1687 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1688 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1689 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1690 (*vec_fcm<cmp_op><mode>): Rename to...
1691 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1692 (*vec_fcmuo<mode>): Rename to...
1693 (*fcmuo<mode>): ...this and adjust likewise.
1694 (*pred_fcm<cmp_op><mode>): New pattern.
1695 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1696 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1698 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1700 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1701 (aarch64_emit_sve_predicated_cond): New function.
1702 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1703 (aarch64_emit_unspec_cond_or): Replace with...
1704 (aarch64_emit_sve_or_conds): ...this new function. Use
1705 aarch64_emit_sve_ptrue_op for the individual comparisons and
1706 aarch64_emit_binop to OR them together.
1707 (aarch64_emit_inverted_unspec_cond): Replace with...
1708 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1709 aarch64_emit_sve_ptrue_op for the comparison and
1710 aarch64_emit_unop to invert the result.
1711 (aarch64_expand_sve_vec_cmp_float): Update after the above
1712 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1714 2018-05-07 Nathan Sidwell <nathan@acm.org>
1716 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1717 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1718 (Backwards Compatibility): Likewise.
1720 2018-05-07 Luis Machado <luis.machado@linaro.org>
1724 2018-05-07 Luis Machado <luis.machado@linaro.org>
1726 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1727 <prefetch_dynamic_strides>: New const bool field.
1728 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1729 prefetch_dynamic_strides.
1730 (exynosm1_prefetch_tune): Likewise.
1731 (thunderxt88_prefetch_tune): Likewise.
1732 (thunderx_prefetch_tune): Likewise.
1733 (thunderx2t99_prefetch_tune): Likewise.
1734 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1735 (aarch64_override_options_internal): Update to set
1736 PARAM_PREFETCH_DYNAMIC_STRIDES.
1737 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1738 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1739 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1740 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1741 prefetch-dynamic-strides setting.
1743 2018-05-07 Luis Machado <luis.machado@linaro.org>
1745 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1746 <minimum_stride>: New const int field.
1747 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1748 minimum_stride field.
1749 (exynosm1_prefetch_tune): Likewise.
1750 (thunderxt88_prefetch_tune): Likewise.
1751 (thunderx_prefetch_tune): Likewise.
1752 (thunderx2t99_prefetch_tune): Likewise.
1753 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1754 (aarch64_override_options_internal): Update to set
1755 PARAM_PREFETCH_MINIMUM_STRIDE.
1756 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1757 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1758 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1759 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1760 stride is constant and is below the minimum stride threshold.
1762 2018-05-07 Luis Machado <luis.machado@linaro.org>
1764 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1767 2018-05-07 Luis Machado <luis.machado@linaro.org>
1769 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1770 <prefetch_dynamic_strides>: New const bool field.
1771 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1772 prefetch_dynamic_strides.
1773 (exynosm1_prefetch_tune): Likewise.
1774 (thunderxt88_prefetch_tune): Likewise.
1775 (thunderx_prefetch_tune): Likewise.
1776 (thunderx2t99_prefetch_tune): Likewise.
1777 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1778 (aarch64_override_options_internal): Update to set
1779 PARAM_PREFETCH_DYNAMIC_STRIDES.
1780 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1781 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1782 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1783 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1784 prefetch-dynamic-strides setting.
1786 2018-05-07 Luis Machado <luis.machado@linaro.org>
1788 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1789 <minimum_stride>: New const int field.
1790 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1791 minimum_stride field.
1792 (exynosm1_prefetch_tune): Likewise.
1793 (thunderxt88_prefetch_tune): Likewise.
1794 (thunderx_prefetch_tune): Likewise.
1795 (thunderx2t99_prefetch_tune): Likewise.
1796 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1797 (aarch64_override_options_internal): Update to set
1798 PARAM_PREFETCH_MINIMUM_STRIDE.
1799 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1800 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1801 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1802 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1803 stride is constant and is below the minimum stride threshold.
1805 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1808 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1809 the type is addressable. Don't force op into register if it has
1812 2018-05-05 Roland McGrath <mcgrathr@google.com>
1815 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1816 any section for which we don't know a specific type it should have,
1817 regardless of name. Previously this was done only for the exact
1818 names ".init_array", ".fini_array", and ".preinit_array".
1819 (default_elf_asm_named_section): Add comment about
1820 relationship with default_section_type_flags and SECTION_NOTYPE.
1821 (get_section): Don't consider it a type conflict if one side has
1822 SECTION_NOTYPE and the other doesn't, as long as neither has the
1823 SECTION_BSS et al used in the default_section_type_flags logic.
1825 2018-05-05 Tom de Vries <tom@codesourcery.com>
1828 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1829 (workaround_barsyncs): New function.
1830 (nvptx_reorg): Use workaround_barsyncs.
1831 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1832 (define_expand "nvptx_membar_cta"): New define_expand.
1833 (define_insn "*nvptx_membar_cta"): New insn.
1835 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1837 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1838 To improve optimization opportunities.
1839 * builtin-types.def: The new needed builtin types for the above.
1841 2018-05-04 Richard Biener <rguenther@suse.de>
1843 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1844 * gimple-ssa-store-merging.c
1845 (imm_store_chain_info::output_merged_store): Remove redundant create,
1846 release split_store vector contents on failure.
1847 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1848 scalar stmt vector on cache hit.
1850 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1852 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1854 * config.gcc (powerpc-xilinx-eabi*): Remove.
1855 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1857 (fusion_addis_mem_combo_load): Ditto.
1858 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1860 (rs6000_cpu_cpp_builtins): Ditto.
1861 * config/rs6000/rs6000-linux.c
1862 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1863 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1864 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1866 (rs6000_setup_reg_addr_masks): Ditto.
1867 (rs6000_init_hard_regno_mode_ok): Ditto.
1868 (rs6000_option_override_internal): Ditto.
1869 (legitimate_lo_sum_address_p): Ditto.
1870 (rs6000_legitimize_address): Ditto.
1871 (rs6000_legitimize_reload_address): Ditto.
1872 (rs6000_legitimate_address_p): Ditto.
1873 (abi_v4_pass_in_fpr): Ditto.
1874 (setup_incoming_varargs): Ditto.
1875 (rs6000_gimplify_va_arg): Ditto.
1876 (rs6000_split_multireg_move): Ditto.
1877 (rs6000_savres_strategy): Ditto.
1878 (rs6000_emit_prologue_components): Ditto.
1879 (rs6000_emit_epilogue_components): Ditto.
1880 (rs6000_emit_prologue): Ditto.
1881 (rs6000_emit_epilogue): Ditto.
1882 (rs6000_elf_file_end): Ditto.
1883 (rs6000_function_value): Ditto.
1884 (rs6000_libcall_value): Ditto.
1885 * config/rs6000/rs6000.h: Ditto.
1886 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1887 (TARGET_MINMAX): ... this. New.
1888 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1889 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1890 (*movsi_internal1_single): Delete.
1891 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1892 mfpu=, mxilinx-fpu): Delete.
1893 * config/rs6000/singlefp.h: Delete.
1894 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1895 * config/rs6000/t-rs6000: Ditto.
1896 * config/rs6000/t-xilinx: Delete.
1897 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1898 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1899 (VStype_simple): Delete.
1900 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1901 * config/rs6000/xfpu.h: Delete.
1902 * config/rs6000/xfpu.md: Delete.
1903 * config/rs6000/xilinx.h: Delete.
1904 * config/rs6000/xilinx.opt: Delete.
1905 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1906 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1908 2018-05-04 Tom de Vries <tom@codesourcery.com>
1911 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1914 2018-05-04 Richard Biener <rguenther@suse.de>
1917 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1918 (expand_complex_div_wide): Likewise.
1919 (expand_complex_operations_1): Likewise.
1920 (expand_complex_libcall): Preserve EH info of the original stmt.
1921 (tree_lower_complex): Handle removed blocks.
1922 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1923 on complex multiplication and division libcall builtins.
1925 2018-05-04 Richard Biener <rguenther@suse.de>
1928 * fold-const.c (negate_expr_p): Restrict negation of operand
1929 zero of a division to when we know that can happen without
1931 (fold_negate_expr_1): Likewise.
1933 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1936 * real.h (real_nextafter): Declare.
1937 * real.c (real_nextafter): New function.
1938 * fold-const-call.c (fold_const_nextafter): New function.
1939 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1940 CASE_CFN_NEXTTOWARD.
1941 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1942 even when arg1_mode is different from arg0_mode.
1944 2018-05-03 Nathan Sidwell <nathan@acm.org>
1946 * doc/extend.texi (Deprecated Features): Remove
1948 (Backwards Compatibility): Likewise.
1949 * doc/invoke.texi (C++ Language Options): Likewise.
1950 (C++ Dialect Options): Likewise.
1952 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1955 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1956 _mm512_mask_mullox_epi64): New intrinsics.
1958 2018-05-03 Tom de Vries <tom@codesourcery.com>
1961 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1962 dump files): Add offload-tree.
1964 2018-05-03 Richard Biener <rguenther@suse.de>
1966 PR tree-optimization/85615
1967 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1968 to loops not nested in BBs loop father to avoid creating multi-entry
1971 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1973 PR tree-optimization/70291
1974 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1975 arguments. Change return type to tree. Emit libcall as a new
1976 statement rather than replacing existing one when inplace_p is true.
1977 (expand_complex_multiplication_components): New function.
1978 (expand_complex_multiplication): Expand floating-point complex
1979 multiplication using the above.
1980 (expand_complex_division): Rename inner_type parameter to type.
1981 Update expand_complex_libcall call-site.
1982 (expand_complex_operations_1): Update expand_complex_multiplication
1983 and expand_complex_division call-sites.
1985 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1988 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1989 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1990 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1991 the highest significant bit of the shift count mask is clear. In
1992 check whether and[sq]i3 is needed verify that all significant bits
1993 of the shift count other than the highest are set.
1995 2018-05-02 Tom de Vries <tom@codesourcery.com>
1998 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1999 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2000 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2001 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2002 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2003 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2004 __builtin_goacc_parlevel_size.
2006 2018-05-02 Richard Biener <rguenther@suse.de>
2008 PR tree-optimization/85597
2009 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2010 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2013 2018-05-02 Tom de Vries <tom@codesourcery.com>
2016 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2017 dump files): Add ltrans-tree.
2019 2018-05-02 Tom de Vries <tom@codesourcery.com>
2022 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2023 dump files): Add wpa-ipa.
2025 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2027 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2028 powerpc*-*-linux*paired* target.
2029 * config/rs6000/750cl.h: Delete.
2030 * config/rs6000/paired.h: Delete.
2031 * config/rs6000/paired.md: Delete.
2032 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2034 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2035 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2036 comment. Remove paired float support.
2037 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2038 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2040 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2041 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2043 * config/rs6000/rs6000.c: Remove paired float support.
2044 (paired_expand_vector_init, paired_expand_vector_move,
2045 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2046 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2047 paired_expand_builtin, paired_expand_predicate_builtin,
2048 paired_init_builtins): Delete.
2049 * config/rs6000/rs6000.h: Remove paired float support.
2050 * config/rs6000/rs6000.md: Remove paired float support.
2051 (move_from_CR_ov_bit): Delete.
2052 * config/rs6000/rs6000.opt (mpaired): Delete.
2053 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2054 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2056 2018-05-02 Richard Biener <rguenther@suse.de>
2059 * gimplify.c (gimplify_save_expr): When in SSA form allow
2060 SAVE_EXPRs to compute to SSA vars.
2062 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2065 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2066 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2067 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2068 clobber operands[2], instead use a new pseudo. Formatting fixes.
2070 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2072 PR tree-optimization/85586
2073 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2074 exit early for statements in the same group if the accesses are
2077 2018-05-02 Tom de Vries <tom@codesourcery.com>
2080 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2083 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2085 PR tree-optimization/85143
2086 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2088 2018-05-01 Tom de Vries <tom@codesourcery.com>
2091 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2092 not found" error message.
2094 2018-05-01 Tom de Vries <tom@codesourcery.com>
2097 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2098 * vec.c (test_ordered_remove_if): New function.
2099 (vec_c_tests): Call test_ordered_remove_if.
2100 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2101 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2102 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2103 VEC_ORDERED_REMOVE_IF.
2105 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2107 PR tree-optimization/82665
2108 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2109 pointer subtraction where arguments come from a memchr call.
2111 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2113 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2114 --push-state --as-needed and --pop-state instead of --as-needed and
2115 --no-as-needed if ld supports it.
2116 * configure: Regenerated.
2119 * doc/install.texi2html: Replace _002d with - and _002a with * in
2120 generated html files using sed.
2122 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2125 * gcc-rich-location.c (blank_line_before_p): New function.
2126 (use_new_line): New function.
2127 (gcc_rich_location::add_fixit_insert_formatted): New function.
2128 * gcc-rich-location.h
2129 (gcc_rich_location::add_fixit_insert_formatted): New function.
2131 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2133 * selftest.c (assert_streq): Rename "expected" and "actual" to
2134 "val1" and "val2". Extend NULL-handling to cover both inputs
2135 symmetrically, while still requiring both to be non-NULL for a pass.
2136 * selftest.h (assert_streq): Rename "expected" and "actual" to
2138 (ASSERT_EQ): Likewise.
2139 (ASSERT_EQ_AT): Likewise.
2140 (ASSERT_KNOWN_EQ): Likewise.
2141 (ASSERT_KNOWN_EQ_AT): Likewise.
2142 (ASSERT_NE): Likewise.
2143 (ASSERT_MAYBE_NE): Likewise.
2144 (ASSERT_MAYBE_NE_AT): Likewise.
2145 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2146 the assertion to pass.
2147 (ASSERT_STREQ_AT): Likewise.
2149 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2151 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2152 interaction with -pie.
2154 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2156 * selftest.h: Fix alphabetization of per-source-file selftest
2159 2018-04-30 Jason Merrill <jason@redhat.com>
2161 PR c++/61982 - dead stores to destroyed objects.
2162 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2165 2018-04-30 Jason Merrill <jason@redhat.com>
2167 * tree.c (build_clobber): New.
2168 * tree.h: Declare it.
2169 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2171 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2173 * diagnostic-show-locus.c (layout::layout): Update for
2174 location_get_source_line returning a char_span.
2175 (struct char_span): Move to input.h.
2176 (struct correction): Update for fields in char_span becoming
2178 (struct source_line): Update for location_get_source_line
2179 returning a char_span.
2180 (layout::print_line): Likewise.
2181 * edit-context.c (edited_file::print_content): Likewise.
2182 (edited_file::print_diff_hunk): Likewise.
2183 (edited_file::print_run_of_changed_lines): Likewise.
2184 (edited_file::get_num_lines): Likewise.
2185 (edited_line::edited_line): Likewise.
2186 * final.c (asm_show_source): Likewise.
2187 * input.c (location_get_source_line): Convert return type
2188 from const char * to char_span, losing the final "line_len"
2190 (dump_location_info): Update for the above.
2191 (get_substring_ranges_for_loc): Likewise. Use a char_span
2192 when handling the literal within the line.
2193 (test_reading_source_line): Update for location_get_source_line
2194 returning a char_span.
2195 * input.h (class char_span): Move here from
2196 diagnostic-show-locus.c, converting from a struct to a class.
2197 Make data members private.
2198 (char_span::operator bool): New.
2199 (char_span::length): New.
2200 (char_span::get_buffer): New.
2201 (char_span::operator[]): New.
2202 (char_span::subspan): Make const.
2203 (char_span::xstrdup): New.
2204 (location_get_source_line): Convert return type from const char *
2205 to char_span, losing the final "line_size" param.
2207 2018-04-30 Jan Hubicka <jh@suse.cz>
2209 * lto-wrapper.c (ltrans_priorities): New static var.
2210 (cmp_priority): New.
2211 (run_gcc): Read priorities and if doing parallel build order
2212 the Makefile by them.
2214 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2216 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2218 2018-04-30 Richard Biener <rguenther@suse.de>
2220 * tree-cfg.c (verify_address): Remove base argument, add
2221 flag whether to check TREE_ADDRESSABLE and do that.
2222 (verify_expr): Remove.
2223 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2224 (verify_gimple_assign_single): Likewise.
2225 (verify_gimple_switch): Likewise.
2226 (verify_expr_location_1): Dereference tp once. Add (disabled)
2227 piece from verify_expr.
2228 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2230 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2232 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2234 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2236 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2237 (small_data_pattern): Likewise.
2238 (arc_rewrite_small_data): Likewise.
2239 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2240 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2241 (get_symbol_alignment): New function.
2242 (legitimate_small_data_address_p): Likewise.
2243 (legitimate_scaled_address): Update, call
2244 legitimate_small_data_address_p.
2245 (output_sdata): New static variable.
2246 (arc_print_operand): Update how we handle small data operands.
2247 (arc_print_operand_address): Likewise.
2248 (arc_legitimate_address_p): Update, use
2249 legitimate_small_data_address_p.
2250 (arc_rewrite_small_data_p): Remove.
2251 (arc_rewrite_small_data_1): Likewise.
2252 (arc_rewrite_small_data): Likewise.
2253 (small_data_pattern): Likewise.
2254 (compact_sda_memory_operand): Update to use
2255 legitimate_small_data_address_p and get_symbol_alignment.
2256 (prepare_move_operands): Don't rewite sdata pattern.
2257 (prepare_extend_operands): Remove.
2258 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2260 (zero_extendqisi2): Likewise.
2261 (zero_extendhisi2): Likewise.
2262 (extendqihi2): Likewise.
2263 (extendqisi2): Likewise.
2264 (extendhisi2): Likewise.
2268 * config/arc/constraints.md (Usd): Change it to memory constraint.
2270 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2272 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2273 as source of std instructions.
2274 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2275 allow 6-bit constants as source for store instructions.
2276 (movdi_insn): Update instruction pattern to allow 6-bit constants
2277 as source for store instructions.
2279 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2281 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2283 2018-04-30 Nathan Sidwell <nathan@acm.org>
2284 Sandra Loosemore <sandra@codesourcery.com>
2286 * dumpfile.c (dump_open): Allow '-' for stdout.
2287 * doc/invoke.texi (Developer Options): Document dump filename
2288 determination early. Document stdin/stdout selection.
2290 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2292 Microblaze Target: PIC data text relative
2294 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2295 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2297 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2298 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2299 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2300 New addressing mode for data-text relative position indepenedent code.
2301 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2302 'ADDRESS_SYMBOLIC_TXT_REL'.
2303 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2304 (microblaze_legitimate_pic_operand): Exclude function calls from
2305 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2306 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2308 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2309 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2310 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2311 for 'address + offset'.
2312 (microblaze_expand_prologue): Add new function prologue call for
2314 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2315 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2316 table in case of TARGET_PIC_DATA_TEXT_REL.
2317 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2318 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2319 Add new macros 'UNSPEC_TEXT',
2320 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2321 + exclude function calls from 'UNSPEC_PLT' in case of data text
2323 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2324 new target hook for generating address diff vector tables in case of
2326 * doc/tm.texi : Regenerate.
2327 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2328 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2329 of addr diff vector generation.
2330 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2331 target hook definition.
2332 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2333 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2334 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2335 option for data text relative.
2337 2018-04-30 Richard Biener <rguenther@suse.de>
2339 * tree-chrec.h (evolution_function_is_constant_p): Remove
2341 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2343 2018-04-30 Richard Biener <rguenther@suse.de>
2346 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2348 2018-04-30 Richard Biener <rguenther@suse.de>
2350 PR tree-optimization/28364
2351 PR tree-optimization/85275
2352 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2353 copying first exit test.
2355 2018-04-28 Mark Wielaard <mark@klomp.org>
2357 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2359 (dwarf_AT): Handle DW_AT_addr_base.
2360 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2362 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2365 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2366 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2367 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2368 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2370 2018-04-28 Richard Biener <rguenther@suse.de>
2372 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2373 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2374 to reflect use. Only add interesting stmts.
2376 2018-04-27 Martin Jambor <mjambor@suse.cz>
2379 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2380 the jump function allows for passing through aggregate values.
2382 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2384 * input.h (in_system_header_at): Convert from macro to inline
2386 (from_macro_expansion_at): Likewise.
2387 (from_macro_definition_at): Likewise.
2389 2018-04-27 Jeff Law <law@redhat.com>
2391 * config.gcc: Mark tile* targets as deprecated/obsolete.
2393 2018-04-27 Richard Biener <rguenther@suse.de>
2395 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2398 2018-04-27 Richard Biener <rguenther@suse.de>
2400 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2402 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2404 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2405 with Yd constraint. Set "preferred_for_speed" attribute from
2406 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2408 (*movdi_internal): Ditto.
2409 (movti_interunit splitters): Remove
2410 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2411 (movdi_interunit splitters): Ditto.
2412 * config/i386/constraints.md (Ye): Remove.
2413 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2415 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2418 * config/aarch64/constraints.md (Usg): Limit to 31.
2421 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2423 PR tree-optimization/85529
2424 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2425 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2426 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2427 zero extension or masking of the MSB bit.
2428 (optimize_range_tests): Add FIRST_BB argument, pass it through
2429 to optimize_range_tests_var_bound.
2430 (maybe_optimize_range_tests, reassociate_bb): Adjust
2431 optimize_range_tests callers.
2433 2018-04-26 Richard Biener <rguenther@suse.de>
2434 Jakub Jelinek <jakub@redhat.com>
2436 * cgraph.h (symbol_table): Just declare debug method here.
2437 * symtab.c (symbol_table::debug): Define.
2439 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2441 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2443 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2445 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2446 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2447 (*movdi_internal): Substitute Yi and Yj constraint with x
2448 and Ym and Yn constraint with y constraint. Update "isa"
2449 attribute and set "preferred_for_speed" attribute from
2450 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2451 (*movsi_internal): Ditto.
2452 (*movdf_internal): Ditto.
2453 (*movsf_internal): Ditto.
2454 (*zero_extendsidi2): Ditto.
2455 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2456 (sse2_loadld): Ditto.
2457 (*vec_extract<ssevecmodelower>_0): Ditto.
2458 (*vec_extractv4si_0_zext_sse4): Ditto.
2459 (vec_concatv2di): Ditto.
2460 (*vec_dup<mode>): Ditto.
2461 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2462 * config/i386/constraints.md (Yi): Remove.
2467 2018-04-26 Nathan Sidwell <nathan@acm.org>
2469 * dumpfile.c (dump_open): New.
2470 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2471 (dump_finish): Detect stdio/stderr by value not name.
2473 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2475 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2477 2018-04-26 Tom de Vries <tom@codesourcery.com>
2480 * config/nvptx/nvptx.c (verify_neutering_jumps)
2481 (verify_neutering_labels): New function
2482 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2484 2018-04-26 Tom de Vries <tom@codesourcery.com>
2487 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2488 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2491 2018-04-26 Richard Biener <rguenther@suse.de>
2492 Tom de Vries <tom@codesourcery.com>
2495 * lto-streamer-out.c (output_function): Fixup loops if required to match
2496 discovery done in the reader.
2498 2018-04-26 Richard Biener <rguenther@suse.de>
2500 PR tree-optimization/85116
2501 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2502 have a loop exit from the single latch predecessor. Remove
2503 case of header with just condition.
2504 (ch_base::copy_headers): Exclude infinite loops from any
2506 (pass_ch::execute): Record exits.
2508 2018-04-26 Richard Biener <rguenther@suse.de>
2510 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2511 prologue cost vector and pass it to vect_get_load_cost.
2512 (vect_get_peeling_costs_all_drs): Likewise.
2513 (vect_peeling_hash_get_lowest_cost): Likewise.
2514 (vect_enhance_data_refs_alignment): Likewise.
2516 2018-04-26 Richard Biener <rguenther@suse.de>
2519 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2520 checking of integer<->pointer conversions.
2521 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2522 sign-/zero-extending pointer types.
2523 (expand_omp_for_static_chunk): Likewise.
2525 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2526 Jean Lee <xiaoyur347@gmail.com>
2528 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2529 (TARGET_ASAN_SHADOW_OFFSET): Define.
2530 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2531 true for -fsanitize=address.
2533 2018-04-25 Mark Wielaard <mark@klomp.org>
2535 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2538 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2540 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2541 than "alu", remove explicit "memory" and "imm_disp" attributes.
2542 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2545 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2546 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2547 gen_lowpart_no_emit.
2549 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2552 * config/i386/i386.c (ix86_expand_builtin): Change memory
2553 operand to XI, extend p0 to Pmode.
2554 * config/i386/i386.md: Change unspec volatile and operand
2555 1 mode to XI, change operand 0 mode to P.
2557 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2559 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2560 GET_MODE_MASK before any checking.
2561 (nds32_can_use_bset_p): Likewise.
2562 (nds32_can_use_btgl_p): Likewise.
2564 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2566 * config/nds32/nds32-doubleword.md: New define_split pattern for
2567 illegal register number.
2569 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2571 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2573 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2575 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2577 2018-04-25 Richard Biener <rguenther@suse.de>
2579 * lto-streamer.h (LTO_major_version): Bump to 8.
2581 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2583 * BASE-VER: Set to 9.0.0.
2585 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2587 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2588 in __abskf2 and __powikf2.
2590 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2593 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2594 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2595 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2596 Use the above on operand 2. Reindent.
2597 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2599 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2602 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2604 * config/i386/i386.opt (mcet): Removed.
2605 * doc/install.texi: Remove -mcet documentation.
2606 * doc/invoke.texi: Likewise.
2608 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2611 * doc/install.texi: Remove -mcet from bootstrap-cet.
2613 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2616 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2617 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2621 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2622 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2623 containing a CONST_VECTOR.
2625 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2627 * doc/install.texi: Update newlib dependency for nvptx.
2629 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2632 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2633 instead of INTVAL when shifting x left.
2635 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2637 PR tree-optimization/85478
2638 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2639 vect_grouped_store_supported for single element vectors.
2641 2018-04-24 Richard Biener <rguenther@suse.de>
2644 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2645 load cost increase to the case of non-constant step.
2647 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2650 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2651 destination if any_malformed_asm.
2653 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2656 * expr.c (store_field): In the bitfield case, if the value comes from
2657 a function call and is returned in registers by means of a PARALLEL,
2658 do not change the mode of the temporary unless BLKmode and VOIDmode.
2660 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2662 PR rtl-optimization/85423
2663 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2664 dependencies to debug insns when the previous insn is non-debug.
2666 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2668 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2669 enums into a single definition.
2670 (fls): Fix predicates and printing.
2673 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2675 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2676 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2677 and short u6 immediate.
2678 (check_if_valid_sleep_operand): Remove.
2679 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2681 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2683 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2684 flag_always_save_lp condition.
2685 * config/nds32/nds32.opt (malways-save-lp): New option.
2687 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2689 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2690 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2691 * config/nds32/nds32.h
2692 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2693 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2695 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2697 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2698 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2700 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2701 Chung-Ju Wu <jasonwucj@gmail.com>
2703 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2705 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2706 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2708 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2710 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2712 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2714 * config/nds32/nds32-protos.h (nds32_data_alignment,
2715 nds32_local_alignment): Declare.
2716 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2717 nds32_local_alignment): New functions.
2718 (TARGET_CONSTANT_ALIGNMENT): Define.
2719 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2721 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2723 * config/nds32/nds32.c
2724 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2725 (TARGET_MODES_TIEABLE_P): Likewise.
2727 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2729 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2732 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2733 Chung-Ju Wu <jasonwucj@gmail.com>
2735 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2736 for unaligned access.
2737 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2739 * config/nds32/nds32-intrinsic.md: Likewise.
2740 * config/nds32/nds32_intrinsic.h: Likewise.
2741 * config/nds32/nds32.h (nds32_builtins): Likewise.
2742 * config/nds32/nds32.opt (munaligned-access): New option.
2743 * config/nds32/nds32.c (nds32_asm_file_start): Display
2744 flag_unaligned_access status.
2746 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2748 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2749 -mno-relax is present.
2750 * config/riscv/linux.h (LINK_SPEC): Ditto.
2752 2018-04-20 Martin Sebor <msebor@redhat.com>
2755 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2757 (gimple_fold_builtin_stxcpy_chk): Same.
2758 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2760 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2763 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2764 __powikf2 when long double is IEEE 128-bit.
2766 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2768 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2769 step to make sure stack always aligned.
2771 2018-04-20 Carl Love <cel@us.ibm.com>
2774 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2775 size check for arg0.
2777 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2778 Tom de Vries <tom@codesourcery.com>
2781 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2782 Emit insns for calls too.
2783 (nvptx_find_par): Always look for worker-level predecessor insn.
2784 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2786 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2787 (nvptx_process_pars): Propagate frames for calls.
2789 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2792 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2794 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2795 (ix86_handle_option): Don't handle OPT_mibt.
2796 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2798 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2800 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2801 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2802 (ix86_target_macros): Define __CET__ with flag_cf_protection
2803 for -fcf-protection.
2804 * config/i386/i386.c (isa2_opts): Remove -mibt.
2805 * config/i386/i386.h (TARGET_IBT): Removed.
2806 (TARGET_IBT_P): Likewise.
2807 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2808 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2809 * config/i386/i386.opt (mcet): Update help message.
2812 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2813 -mcet as an alias for -mshstk.
2815 2018-04-20 Richard Biener <rguenther@suse.de>
2818 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2819 complexity by forcing a single use of the multiply operand.
2821 2018-04-20 Martin Jambor <mjambor@suse.cz>
2824 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2825 recursion dependency to only apply to non-clones.
2827 2018-04-20 Martin Jambor <mjambor@suse.cz>
2830 * ipa-cp.c (create_specialized_node): Check that clones of
2831 self-recursive edges exist during IPA-CP.
2833 2018-04-19 Toon Moene <toon@moene.org>
2835 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2838 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2840 PR tree-optimization/85467
2841 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2842 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2843 VECTOR_CST element to type.
2845 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2848 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2849 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2850 (builtin_longjmp): Likewise.
2851 (save_stack_nonlocal): New pattern.
2852 (restore_stack_nonlocal): Likewise.
2854 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2857 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2858 Replace ASM_OUTPUT_LABEL with fprintf.
2860 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2863 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2864 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2865 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2866 define __IBT__ and __SHSTK__ for -fcf-protection.
2867 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2869 (ix86_trampoline_init): Likewise.
2870 (x86_output_mi_thunk): Likewise.
2871 (ix86_notrack_prefixed_insn_p): Likewise.
2872 (ix86_option_override_internal): Don't disallow -fcf-protection.
2873 * config/i386/i386.md (rdssp<mode>): Also enable for
2875 (incssp<mode>): Likewise.
2876 (nop_endbr): Likewise.
2877 * config/i386/i386.opt (mcet): Change help message to built-in
2881 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2882 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2883 enable CET built-in functions.
2885 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2887 * common/config/i386/i386-common.c
2888 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2889 OPTION_MASK_ISA_MOVDIRI_UNSET,
2890 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2891 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2892 * config.gcc (movdirintrin.h): New header.
2893 * config/i386/cpuid.h (bit_MOVDIRI,
2894 bit_MOVDIR64B): New bits.
2895 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2897 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2898 (VOID, PVOID, PCVOID)): New function types.
2899 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2900 __builtin_ia32_directstoreu_u64,
2901 __builtin_ia32_movdir64b): New builtins.
2902 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2903 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2905 (ix86_valid_target_attribute_inner_p): Ditto.
2906 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2907 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2908 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2909 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2910 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2911 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2912 (movdiri<mode>, movdir64b_<mode>): New.
2913 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2914 * config/i386/immintrin.h: Include movdirintrin.h.
2915 * config/i386/movdirintrin.h: New file.
2916 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2918 2018-04-19 Richard Biener <rguenther@suse.de>
2921 * cfg.c (clear_bb_flags): When loop state says we have
2922 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2924 2018-04-19 Richard Biener <rguenther@suse.de>
2926 PR tree-optimization/84737
2927 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2928 copying restrict info.
2929 (vect_setup_realignment): Use it.
2930 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2931 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2932 the first DR to all generated stores.
2933 (vectorizable_load): Likewise for loads.
2935 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2937 PR tree-optimization/85446
2938 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2939 the integral and pointer types to have the same precision.
2941 * doc/install.texi: Document --disable-cet being the default and
2944 2018-04-18 Martin Liska <mliska@suse.cz>
2946 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2949 2018-04-18 Martin Liska <mliska@suse.cz>
2952 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2955 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2956 arguments if they are comparable.
2958 2018-04-18 Martin Liska <mliska@suse.cz>
2961 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2964 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2967 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2970 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2971 ENDBR after calling __morestack.
2973 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2976 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2977 by using gcc_base_ver to generate a gcc_driver_version, and use
2978 it when generating GCC_DRIVER_NAME.
2979 * configure: Regenerate.
2981 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2984 * config.gcc: Obsolete powerpc*-*-*spe*.
2986 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2989 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2990 (stabstr_D): Change type of unum from unsigned int to
2991 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2994 2018-04-17 Jim Wilson <jimw@sifive.com>
2997 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2998 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2999 Set arg_pointer_offset after using pretend_args_size.
3001 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3003 PR rtl-optimization/85431
3004 * dse.c (record_store): Ignore zero width stores.
3007 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3008 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3009 __builtin_stack_restore rather than after it.
3010 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3011 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3012 argument instead of virtual_dynamic_stack_rtx.
3014 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3016 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3018 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3019 Add note to error message to explain internal mapping of overloaded
3020 built-in function name to non-overloaded built-in function name.
3021 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3024 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3027 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3028 where the inputs overlap with the output.
3030 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3032 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3033 (=v, v) alternative and explicit "memory" attribute.
3034 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3035 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3037 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3038 "sselog1" type instead of "sselog".
3039 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3040 "sselog". Remove explicit "memory" attribute.
3041 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3042 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3044 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3045 "sselog1" type instead of "sselog". Remove explicit "memory"
3047 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3048 use "sselog1" type instead of "sselog". Remove explicit "memory"
3050 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3051 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3053 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3054 "sselog1" type instead of "sselog". Remove explicit "memory"
3056 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3057 use "sselog1" type instead of "sselog". Remove explicit "memory"
3061 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3064 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3067 2018-04-17 Martin Jambor <mjambor@suse.cz>
3070 * ipa-cp.c (create_specialized_node): Call
3071 expand_all_artificial_thunks if necessary.
3073 2018-04-17 Martin Liska <mliska@suse.cz>
3076 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3077 in message, remote space in between '_G' and '('.
3079 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3082 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3083 avx512f_vmcmp<mode>3<round_saeonly_name>,
3084 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3085 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3086 avx512f_rndscale<mode><round_saeonly_name>,
3087 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3088 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3089 Use %<iptr>2 instead of %2 for -masm=intel.
3090 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3091 avx512f_vcvttss2usi<round_saeonly_name>,
3092 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3094 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3095 avx512f_vcvttsd2usi<round_saeonly_name>,
3096 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3097 Use %q1 instead of %1 for -masm=intel.
3098 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3099 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3100 of %3 for -masm=intel.
3101 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3103 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3105 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3107 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3108 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3110 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3112 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3113 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3114 %g1 and one with %0 and %1.
3115 (avx512er_vmrcp28<mode><round_saeonly_name>,
3116 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3118 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3119 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3120 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3121 of %0 and %{%4%} for -masm=intel.
3122 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3123 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3124 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3125 order of %0 and %{%5%}%{z%} for -masm=intel.
3127 2018-04-17 Jan Hubicka <jh@suse.cz>
3130 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3132 2018-04-17 Martin Liska <mliska@suse.cz>
3135 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3136 for target_clone error message. Make default implementation
3137 clone to be a local declaration.
3138 (separate_attrs): Add new argument and check for an empty
3140 (expand_target_clones): Handle it.
3141 (ipa_target_clone): Make redirection just for target_clones
3144 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3145 Tom de Vries <tom@codesourcery.com>
3148 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3149 tiled basic blocks without omp continue statements.
3151 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3154 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3155 vec_extract expression as having side effects to make sure it gets
3158 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3161 * config/i386/i386.c (get_builtin_code_for_version): Check
3164 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3167 * gcc/config.gcc: Support "skylake".
3168 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3170 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3171 (processor_target_table): Add "skylake".
3172 (ix86_option_override_internal): Add "skylake".
3173 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3174 PROCESSOR_CANNONLAKE.
3175 (get_builtin_code_for_version): Fix priority for
3176 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3177 PROCESSOR_SKYLAKE-AVX512.
3178 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3179 (processor_type): Add PROCESSOR_SKYLAKE.
3181 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3182 Jason Merrill <jason@redhat.com>
3185 * convert.c (convert_to_integer_1): Use direct recursion for
3186 enumeral types and types with a precision less than the number
3187 of bits in their mode.
3189 2018-04-16 Julia Koval <julia.koval@intel.com>
3192 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3193 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3195 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3198 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3199 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3200 and -mno-direct-move.
3202 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3205 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3206 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3207 Ensure negative shifts result in {0}.
3209 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3211 PR rtl-optimization/79916
3212 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3213 regs (if any) to define how to gnerate SD moves when LRA is in
3216 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3218 PR rtl-optimization/85393
3219 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3220 * except.c (expand_dw2_landing_pad_for_region): Make static.
3221 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3222 a label and unconditional jump to old_bb, rather than
3223 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3226 PR rtl-optimization/85376
3227 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3228 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3229 instead of a specific value.
3231 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3232 Bin Cheng <bin.cheng@arm.com>
3234 PR tree-optimization/82965
3235 PR tree-optimization/83991
3236 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3237 by_profile_only parameter.
3238 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3239 information if the loop was predicted to iterate too many times.
3240 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3242 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3245 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3248 2018-04-13 Martin Liska <mliska@suse.cz>
3249 Jakub Jelinek <jakub@redhat.com>
3252 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3253 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3254 * builtins.c (expand_builtin_memory_copy_args): Use
3255 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3256 handle dest_addr == pc_rtx.
3258 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3261 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3262 asked to not generate direct moves.
3263 (fix_trunc<mode>si2_stfiwx): Similar.
3264 (fix_trunc<mode>si2_internal): Similar.
3266 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3269 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3270 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3271 lookup if dest in some wider mode is known to be const0_rtx and
3272 if so, record permanent equivalence for it to be ZERO_EXTEND of
3273 the narrower mode destination.
3275 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3277 * lto-streamer-out.c (output_function): Revert 259346.
3278 * omp-expand.c (expand_oacc_for): Likewise.
3280 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3282 PR rtl-optimization/85354
3283 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3284 * sel-sched.c (sel_global_init): ... here.
3286 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3289 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3290 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3291 mode for PE-COFF targets.
3292 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3293 (i386_pe_asm_lto_end): Likewise.
3294 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3295 (TARGET_ASM_LTO_END): Likewise.
3296 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3297 (i386_pe_asm_lto_start): New function.
3298 (i386_pe_asm_lto_end): Likewise.
3300 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3301 Richard Biener <rguenther@suse.de>
3304 * lto-streamer-out.c (output_function): Fix CFG loop state before
3306 * omp-expand.c (expand_oacc_for): Handle calls to internal
3307 functions like regular functions.
3309 2018-04-12 Richard Biener <rguenther@suse.de>
3312 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3313 for the early LTO debug to properly generate references to it
3314 during DIE emission. Do not re-use that for the skeleton for
3316 (dwarf2out_early_finish): Likewise.
3318 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3321 * config/i386/sse.md
3322 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3323 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3324 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3325 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3326 and output is a reg, avoid creating invalid lowpart subreg, but
3327 instead split into a 512-bit move. Don't split if not AVX512VL,
3328 input is xmm16+ reg and output is a mem.
3329 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3330 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3331 xmm16+ reg and output is a mem.
3333 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3335 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3336 also for flag_dwarf2_cfi_asm.
3338 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3340 PR rtl-optimization/85342
3341 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3342 a bool scalar var inside of the loop instead. Don't try to update
3343 recog_data.operand after failed apply_change_group.
3345 2018-04-12 Tom de Vries <tom@codesourcery.com>
3348 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3349 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3350 array with flexible array member as array without given dimension.
3351 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3352 argument for undefined param to true.
3354 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3357 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3358 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3359 from PowerPC section.
3360 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3361 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3362 help text that is too long.
3363 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3364 help text that is too long.
3365 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3366 help text that is too long.
3368 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3370 * config/alpha/alpha.md (stack_probe_internal): Rename
3371 from "probe_stack". Update all callers.
3373 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3375 PR rtl-optimization/84566
3376 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3377 sched_macro_fuse_insns.
3379 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3382 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3383 (compute_block_dependences): ... from here.
3385 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3387 PR tree-optimization/85331
3388 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3389 from int to HOST_WIDE_INT.
3391 2018-04-11 Martin Jambor <mjambor@suse.cz>
3394 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3395 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3396 not the same as the source val.
3397 (cgraph_edge_brings_value_p): New parameter.
3398 (gather_edges_for_value): Pass destination value to
3399 cgraph_edge_brings_value_p.
3400 (perhaps_add_new_callers): Likewise.
3401 (get_info_about_necessary_edges): Likewise and exclude values brought
3402 only by self-recursive edges.
3403 (create_specialized_node): Redirect only clones of self-calling edges.
3404 (+self_recursive_pass_through_p): New function.
3405 (find_more_scalar_values_for_callers_subset): Use it.
3406 (find_aggregate_values_for_callers_subset): Likewise.
3407 (known_aggs_to_agg_replacement_list): Removed.
3408 (decide_whether_version_node): Re-calculate known constants for all
3409 remaining context clones.
3411 2018-04-11 Richard Biener <rguenther@suse.de>
3414 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3415 from early DWARF output.
3416 (dwarf2out_early_finish): Output line info unconditionally into
3417 early DWARF and add reference to it.
3419 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3422 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3423 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3424 other than V2DFmode using iptr mode attribute.
3425 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3427 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3429 PR rtl-optimization/84659
3430 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3432 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3435 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3437 (output_loc_list): Pass address of a dummy size variable even in the
3438 locview handling loop.
3439 (index_location_lists): Add comment on why skip_loc_list_entry can't
3442 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3445 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3448 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3451 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3452 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3453 and -mstring-compare-inline-limit.
3455 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3458 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3459 for stack clash protection in a register whenever we need it to be in
3462 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3464 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3465 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3467 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3470 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3472 (mlong-double-): Ditto.
3473 * config/rs6000/sysv4.opt (msdata=): Ditto.
3474 (mtls-size=): Ditto.
3476 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3478 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3479 erroneous entries for
3480 "vector int vec_ldl (int, long int *)", and
3481 "vector unsigned int vec_ldl (int, unsigned long int *)".
3482 Add comments and entries for
3483 "vector bool char vec_ldl (int, bool char *)",
3484 "vector bool short vec_ldl (int, bool short *)",
3485 "vector bool int vec_ldl (int, bool int *)",
3486 "vector bool long long vec_ldl (int, bool long long *)",
3487 "vector pixel vec_ldl (int, pixel *)",
3488 "vector long long vec_ldl (int, long long *)",
3489 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3490 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3491 type tree bool_long_long_type_node and correct definition of
3492 bool_V2DI_type_node to make reference to this new type tree.
3493 (rs6000_mangle_type): Replace erroneous reference to
3494 bool_long_type_node with bool_long_long_type_node.
3495 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3496 comments to emphasize sign distinctions for char and int types and
3497 replace RS6000_BTI_bool_long constant with
3498 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3499 use of RS6000_BTI_pixel.
3500 (bool_long_type_node): Remove this macro definition.
3501 (bool_long_long_type_node): New macro definition
3503 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3505 PR rtl-optimization/85300
3506 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3507 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3508 simplify_unary_operation fails.
3510 2018-04-10 Martin Liska <mliska@suse.cz>
3512 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3513 cgraph_edge and ipa_ref.
3515 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3519 * config/i386/sse.md
3520 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3521 computation of the VEC_MERGE selector from mask.
3522 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3523 Fix decoding of the VEC_MERGE selector into mask.
3525 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3527 PR tree-optimization/85286
3528 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3530 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3532 * final.c (final_1): Set insn_last_address as well as
3533 insn_current_address.
3535 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3538 * explow.c (emit_stack_probe): Call validize_mem on memory location
3539 before passing it to gen_probe_stack. Create address operand and
3540 legitimize it for the probe_stack_address case.
3542 2018-04-09 Jan Hubicka <jh@suse.cz>
3545 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3546 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3547 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3549 (free_lang_data): Rebuild type inheritance graph.
3551 2018-04-09 Martin Sebor <msebor@redhat.com>
3553 * invoke.texi (-finline-small-functions): Mention other optimization
3555 (-findirect-inlining, -fpartial-inlining): Same.
3556 (-finline-functions-called-once): Same.
3557 (-freorder-blocks-and-partition): Same.
3559 2018-04-09 Jan Hubicka <jh@suse.cz>
3562 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3563 jumps; choose last target that matches the criteria (i.e.
3564 no partition changes for non-crossing jumps).
3565 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3566 support for redirecting crossing jumps to non-crossing.
3568 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3570 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3571 also for naked functions.
3573 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3575 * config/arc/arc.md (add_shift): New pattern.
3576 (add_shift2): Likewise.
3577 (sub_shift): Likewise.
3578 (sub_shift_cmp0_noout): Likewise.
3579 (compare_si_ashiftsi): Likewise.
3580 (xbfu_cmp0_noout): New combine pattern.
3581 (xbfu_cmp0"): Likewise.
3582 (movsi_set_cc_insn): Place the predicable variant first.
3583 (commutative_binary_cmp0_noout): Remove clobber.
3584 (commutative_binary_cmp0): New pattern.
3585 (noncommutative_binary_cmp0): Likewise.
3586 (noncommutative_binary_cmp0_noout): Likewise.
3587 (noncommutative_binary_comparison_result_used): Removed.
3588 (rsub_cmp0): New pattern.
3589 (rsub_cmp0_noout): Likewise.
3590 (extzvsi): Changed, keep only meaningful variants.
3591 (SQH, SEZ): New iterators.
3592 (SQH_postfix): New mode attribute.
3593 (SEZ_prefix): New code attribute.
3594 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3595 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3596 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3598 (noncommutative_operator): Check the availability of barrel
3601 2018-04-09 Richard Biener <rguenther@suse.de>
3603 PR tree-optimization/85284
3604 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3605 Only use the niter constraining form of simple_iv when the exit
3608 2018-04-09 Tom de Vries <tom@codesourcery.com>
3611 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3612 (define_expand "*memory_barrier"): New define_expand.
3613 (define_insn "memory_barrier"): New insn.
3615 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3617 PR rtl-optimization/80463
3618 PR rtl-optimization/83972
3619 PR rtl-optimization/83480
3621 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3622 correct producer for the insn.
3623 (tidy_control_flow): Fixup seqnos in case of debug insns.
3625 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3627 PR rtl-optimization/83913
3629 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3630 different sched-times when merging exprs.
3632 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3634 PR rtl-optimization/83962
3636 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3637 tidy_fallthru_edge and tidy_control_flow.
3639 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3641 PR rtl-optimization/83530
3643 * sel-sched.c (force_next_insn): New global variable.
3644 (remove_insn_for_debug): When force_next_insn is true, also leave only
3645 next insn in the ready list.
3646 (sel_sched_region): When the region wasn't scheduled, make another pass
3647 over it with force_next_insn set to 1.
3649 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3651 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3653 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3654 for interrupt control.
3655 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3656 functions for interrupt control.
3657 * config/nds32/nds32-intrinsic.md: Likewise.
3658 * config/nds32/nds32_intrinsic.h: Likewise.
3659 * config/nds32/nds32.h (nds32_builtins): Likewise.
3661 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3663 * config/nds32/nds32.c (nds32_init_machine_status,
3664 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3665 strict_aligned_p field.
3666 (nds32_expand_to_rtl_hook): New function.
3667 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3668 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3670 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3671 Chung-Ju Wu <jasonwucj@gmail.com>
3673 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3674 * config/nds32/nds32-n7.md: New file.
3675 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3676 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3678 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3679 * config/nds32/nds32.md (pipeline_model): Add n7.
3680 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3681 * config/nds32/pipelines.md: Include n7 settings.
3683 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3684 Chung-Ju Wu <jasonwucj@gmail.com>
3686 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3687 * config/nds32/nds32-e8.md: New file.
3688 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3689 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3691 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3692 * config/nds32/nds32.md (pipeline_model): Add e8.
3693 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3694 * config/nds32/pipelines.md: Include e8 settings.
3696 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3697 Chung-Ju Wu <jasonwucj@gmail.com>
3699 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3700 * config/nds32/nds32-n8.md: New file.
3701 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3702 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3704 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3705 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3706 * config/nds32/nds32.md (pipeline_model): Add n8.
3707 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3708 * config/nds32/pipelines.md: Include n8 settings.
3710 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3711 Chung-Ju Wu <jasonwucj@gmail.com>
3713 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3714 * config/nds32/nds32-n9-2r1w.md: New file.
3715 * config/nds32/nds32-n9-3r2w.md: New file.
3716 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3717 nds32_register_ports): New or modify for cpu n9.
3718 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3720 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3721 * config/nds32/nds32-utils.c: New file.
3722 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3723 TARGET_MUL_SLOW): Define.
3724 * config/nds32/nds32.md (pipeline_model): New attribute.
3725 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3726 New options that support cpu n9.
3727 * config/nds32/pipelines.md: Include n9 settings.
3728 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3730 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3732 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3733 information if necessary.
3734 (output_cond_branch_compare_zero): Likewise.
3735 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3736 (nds32_target_alignment): Refine for alignment.
3737 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3738 (FUNCTION_BOUNDARY): Modify.
3739 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3741 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3743 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3745 * config/nds32/constants.md (unspec_volatile_element): Add values for
3746 TLB operation and data prefetch.
3747 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3748 functions for TLB operation and data prefetch.
3749 * config/nds32/nds32-intrinsic.md: Likewise.
3750 * config/nds32/nds32_intrinsic.h: Likewise.
3751 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3752 (nds32_print_operand): Likewise.
3753 * config/nds32/nds32.h (nds32_builtins): Likewise.
3755 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3756 Andrew Pinski <pinsika@gcc.gnu.org>
3759 * match.pd: Use constant_boolean_node of correct type instead of
3760 boolean_true_node or boolean_false_node for simplifying
3761 pointer comparisons to zero.
3763 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3765 PR tree-optimization/80021
3766 * tree.c (verify_type_variant): Make error call in verify_variant_match
3767 translatable and remove final full stop.
3769 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3771 * config/nds32/constants.md (unspec_volatile_element): Add
3772 UNSPEC_VOLATILE_EH_RETURN.
3773 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3774 nds32_output_stack_pop): Support dwarf exception handling process.
3775 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3776 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3777 exception handling process.
3778 (nds32_compute_stack_frame): Likewise.
3779 (nds32_return_addr_rtx): Likewise.
3780 (nds32_initial_elimination_offset): Likewise.
3781 (nds32_expand_prologue): Likewise.
3782 (nds32_expand_epilogue): Likewise.
3783 (nds32_dynamic_chain_address): New function.
3784 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3786 (DYNAMIC_CHAIN_ADDRESS): Define.
3787 (EH_RETURN_DATA_REGNO): Define.
3788 (EH_RETURN_STACKADJ_RTX): Define.
3789 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3790 patterns for dwarf exception handling.
3792 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3794 * config/nds32/nds32.h: Clean up obsolete macros.
3796 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3798 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3799 Add enum values for particular instructions.
3800 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3801 particular intrinsic functions.
3802 * config/nds32/nds32-intrinsic.md: Likewise.
3803 * config/nds32/nds32_intrinsic.h: Likewise.
3804 * config/nds32/nds32.h (nds32_builtins): Likewise.
3805 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3806 (btst, ave): New patterns for particular instructions.
3808 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3810 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3811 Add enum values for atomic load/store and memory sync.
3812 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3814 * config/nds32/nds32-intrinsic.md: Likewise.
3815 * config/nds32/nds32_intrinsic.h: Likewise.
3816 * config/nds32/nds32.h (nds32_builtins): Likewise.
3818 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3820 PR tree-optimization/85257
3821 * fold-const.c (native_encode_vector): If not all elts could fit
3822 and off is -1, return 0 rather than offset.
3823 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3824 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3825 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3826 adjust buffer in native_interpret_expr call.
3828 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3830 * config/nds32/constants.md (unspec_volatile_element): Add cache
3831 control enum values.
3832 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3833 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3834 * config/nds32/nds32.c (nds32_cctl_names): New.
3835 (nds32_print_operand): Handle cache control register names.
3836 * config/nds32/nds32.h (nds32_builtins): New enum values.
3837 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3839 * config/nds32/nds32.md (type): Add mmu.
3840 * config/nds32/pipelines.md (simple_insn): Add mmu.
3842 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3844 * config/nds32/nds32.md (type): Remove call.
3845 * config/nds32/pipelines.md (simple_insn): Likewise.
3847 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3849 * config/nds32/constants.md (unspec_volatile_element): Add
3850 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3851 UNSPEC_VOLATILE_FMFCFG.
3852 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3853 description for fmfcfg and fmfcsr.
3854 (bdesc_1arg): Add fmtcsr.
3855 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3856 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3857 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3858 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3859 unspec_fmfcfg): New patterns.
3860 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3861 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3862 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3863 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3864 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3865 __nds32__fmfcfg): Define.
3867 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3869 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3870 intrinsic register names.
3871 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3872 intrinsic register enum values and macros.
3874 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3876 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3877 for load/store addressing form.
3878 (nds32_print_operand_address): Likewise.
3880 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3883 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3884 based on LABEL_REF. Remove useless assertion.
3885 (pic_address_needs_scratch): Fix formatting.
3886 (sparc_legitimize_pic_address): Minor tweaks.
3887 (sparc_delegitimize_address): Adjust assertion accordingly.
3888 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3889 into symbolic_operand.
3890 (movsi_high_pic_label_ref): Likewise.
3891 (movsi_lo_sum_pic_label_ref): Likewise.
3892 (movdi_pic_label_ref): Likewise.
3893 (movdi_high_pic_label_ref): Likewise.
3894 (movdi_lo_sum_pic_label_ref): Likewise.
3896 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3898 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3899 custom LIB_SPEC setup.
3901 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3902 Kito Cheng <kito.cheng@gmail.com>
3904 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3905 * config/riscv/freebsd.h: New.
3907 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3910 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3913 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3914 Kito Cheng <kito.cheng@gmail.com>
3916 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3917 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3918 * config/nds32/nds32-protos.h (nds32_output_call,
3919 nds32_output_return): Declare.
3920 * config/nds32/nds32.md: Refine all the call and return patterns.
3922 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3925 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3926 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3928 PR rtl-optimization/84872
3929 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3930 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3933 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3935 * expr.c (copy_blkmode_to_reg): Revert 254862.
3936 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3938 2018-04-06 Richard Biener <rguenther@suse.de>
3941 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3942 after seeing a component reference with an adjacent field. Treat
3943 refs to arrays at struct end of external decls similar to
3944 refs to unconstrained commons.
3946 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3949 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3950 look through SAVE_EXPRs with non-side-effects argument. Adjust
3952 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3955 2018-04-06 Richard Biener <rguenther@suse.de>
3958 * alias.c (find_base_term): New wrapper around find_base_term
3959 unwinding CSELIB_VAL_PTR changes.
3960 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3963 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3965 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3967 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3968 constant definitions.
3969 ("nop"): lr 0,0 -> nopr r0
3970 ("nop_lr0", "nop_lr1"): New insn definitions.
3972 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3974 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3975 NDS32_V3PUSH_AVAILABLE_P macro.
3977 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3978 Chung-Ju Wu <jasonwucj@gmail.com>
3980 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3981 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3982 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3983 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3984 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3985 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3986 * config/nds32/constraints.md: New constraints and checking for hard
3987 float configuration.
3988 * config/nds32/iterators.md: New mode iterator and attribute for hard
3989 float configuration.
3990 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3992 * config/nds32/nds32-fpu.md: New file.
3993 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3994 deal with hard float code generation.
3995 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3997 (abi_type, float_reg_number): New enum type.
3998 * config/nds32/nds32-predicates.c: New predicates for hard float.
3999 * config/nds32/nds32-protos.h: Declare functions for hard float.
4000 * config/nds32/nds32.c: Implementation for hard float configuration.
4001 * config/nds32/nds32.h: Definitions for hard float configuration.
4002 * config/nds32/nds32.md: Include hard float machine description and
4003 modify patterns for hard float configuration.
4004 * config/nds32/nds32.opt: New options for hard float configuration.
4005 * config/nds32/predicates.md: New predicates for hard float
4008 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4010 * common/config/nds32/nds32-common.c
4011 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4013 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4016 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4017 CONSTRUCTOR_ELT (ctor, ...)->value.
4019 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4022 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4024 2018-04-05 Tom de Vries <tom@codesourcery.com>
4027 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4030 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4031 Kito Cheng <kito.cheng@gmail.com>
4033 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4034 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4035 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4036 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4038 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4039 Kito Cheng <kito.cheng@gmail.com>
4041 * config/nds32/constraints.md (Ufe): New memory constraint.
4042 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4043 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4044 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4046 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4047 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4049 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4051 * config/nds32/nds32.md: Use optimize_size in the condition for
4052 alu-shift instructions.
4054 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4056 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4058 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4060 * config/nds32/nds32.md (negsi2): Refine pattern.
4062 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4063 Chung-Ju Wu <jasonwucj@gmail.com>
4065 * config/nds32/iterators.md (shift_rotate): New code iterator.
4066 (shift): New code attribute.
4067 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4068 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4069 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4070 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4071 bit-wise operations.
4072 (andsi3, *andsi3): Ditto.
4073 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4074 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4075 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4076 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4077 nds32_ior_operand, nds32_xor_operand): New predicates.
4079 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4081 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4082 (addsi3, subsi3): ... this.
4084 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4086 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4088 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4090 * config/nds32/nds32.md: Adjust indention.
4092 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4094 * config/nds32/nds32.md (feature): New attribute.
4096 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4098 * config/nds32/nds32.md (subtype): New attribute.
4100 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4103 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4104 expansion to perform a bitwise AND of the argument followed by a
4105 boolean negation of the result.
4107 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4109 PR rtl-optimization/84878
4110 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4111 the basic block. Assert the use reference is not artificial and that
4112 it has an associated insn.
4114 2018-04-04 Michael Matz <matz@suse.de>
4116 * builtins.c (compute_objsize): Pass correct operand
4117 to array_at_struct_end_p.
4119 2018-04-04 Richard Biener <rguenther@suse.de>
4122 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4123 from contexts for DINFO_LEVEL_TERSE and below.
4125 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4127 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4128 resiter_operand condition.
4129 * config/nds32/nds32.md (*move<mode>): Ditto.
4131 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4132 Monk Chiang <sh.chiang04@gmail.com>
4134 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4136 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4138 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4140 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4141 Kito Cheng <kito.cheng@gmail.com>
4143 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4144 nds32_cond_code_str, output_cond_branch,
4145 output_cond_branch_compare_zero, nds32_expand_cbranch,
4146 nds32_expand_cstore, nds32_expand_movcc,
4147 nds32_output_cbranchsi4_equality_zero,
4148 nds32_output_cbranchsi4_equality_reg,
4149 nds32_output_cbranchsi4_equality_reg_or_const_int,
4150 nds32_output_cbranchsi4_greater_less_zero: New functions.
4151 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4152 nds32_expand_cstore, nds32_expand_movcc,
4153 nds32_output_cbranchsi4_equality_zero,
4154 nds32_output_cbranchsi4_equality_reg,
4155 nds32_output_cbranchsi4_equality_reg_or_const_int,
4156 nds32_output_cbranchsi4_greater_less_zero): Declare.
4157 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4158 nds32_rimm11s_operand): New predicates.
4159 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4160 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4163 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4165 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4166 * config/nds32/nds32.md: Ditto.
4167 * config/nds32/pipelines.md: Ditto.
4169 2018-04-04 Richard Biener <rguenther@suse.de>
4171 PR tree-optimization/85168
4172 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4173 propagating abnormals.
4175 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4177 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4179 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4180 Kito Cheng <kito.cheng@gmail.com>
4182 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4183 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4184 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4185 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4186 * config/nds32/nds32.md (sibcall_internal): New.
4187 (sibcall_register): Remove.
4188 (sibcall_immediate): Remove.
4189 (sibcall_value_internal): New.
4190 (sibcall_value_register): Remove.
4191 (sibcall_value_immediate): Remove.
4192 * config/nds32/predicates.md (nds32_general_register_operand): New.
4193 (nds32_call_address_operand): New.
4195 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4197 PR rtl-optimization/85167
4198 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4199 bb_defs if *split_p, instead preinitialize it to NULL.
4201 PR tree-optimization/85156
4202 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4203 evaluating the argument multiple times.
4205 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4207 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4209 (_mm_cvtpd_ps): Likewise.
4210 (_mm_cvttpd_epi32): Likewise.
4211 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4212 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4213 vector, pixel, and bool following altivec.h include.
4215 2018-04-03 Martin Sebor <msebor@redhat.com>
4217 * doc/extend.texi (Common Function Attributes): Clarify.
4218 (const attribute): Likewise.
4219 (pure attribute): Likewise.
4221 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4224 * config/i386/i386.c (ix86_expand_vector_set): Use
4225 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4227 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4229 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4230 instructions when changing rounding bits to preserve precision bits
4231 in the x87 control word.
4233 2018-04-03 Martin Liska <mliska@suse.cz>
4235 PR tree-optimization/82491
4236 * rtl.h (strip_offset_and_add): Replace += suboffset with
4237 poly_uint64 () + suboffset.
4239 2018-03-29 Martin Liska <mliska@suse.cz>
4240 Martin Jambor <mjambor@suse.cz>
4243 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4244 param_type is not an integral or pointer type.
4246 2018-04-03 Richard Biener <rguenther@suse.de>
4248 * sese.h (recompute_all_dominators): Remove.
4250 2018-04-02 Martin Sebor <msebor@redhat.com>
4252 * doc/invoke.texi (-Wrestrict): Fix typos.
4254 2018-04-02 Jim Wilson <jimw@sifive.com>
4256 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4257 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4258 (<optab>di3, <optab>si3_extend): Likewise.
4259 (<optab>si3_mask, <optab>si3_mask_1): New.
4260 (<optab>di3_mask, <optab>di3_mask_1): New.
4261 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4262 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4263 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4265 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4267 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4270 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4272 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4273 (nds32_canonicalize_comparison): New function.
4275 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4276 Kito Cheng <kito.cheng@gmail.com>
4277 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4279 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4280 * config/nds32/constants.md (unspec_volatile_element): Add
4281 UNSPEC_VOLATILE_RELAX_GROUP.
4282 * config/nds32/nds32-relax-opt.c: New file.
4283 * config/nds32/nds32-predicates.c
4284 (nds32_symbol_load_store_p): New function.
4285 * config/nds32/nds32-protos.h
4286 (nds32_symbol_load_store_p): Declare function.
4287 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4288 * config/nds32/nds32.c
4289 (nds32_register_pass): New function to register pass.
4290 (nds32_register_passes): New function to register passes.
4291 * config/nds32/nds32.md (relax_group): New pattern.
4292 * config/nds32/nds32.opt (mrelax-hint): New option.
4293 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4295 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4297 * config/nds32/t-nds32: Modify files dependency.
4299 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4301 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4302 (PROFILE_HOOK): Define its implementation.
4304 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4306 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4307 type and 32-bit size.
4309 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4312 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4313 (V_128_256): New mode iterator.
4314 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4315 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4316 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4318 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4319 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4321 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4324 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4325 NaN inputs correctly.
4327 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4330 * config/rs6000/vsx.md (??r): New mode attribute.
4331 (*vsx_mov<mode>_64bit): Use it.
4332 (*vsx_mov<mode>_32bit): Likewise.
4334 2018-03-30 Martin Sebor <msebor@redhat.com>
4336 PR tree-optimization/84818
4337 * builtins.c (check_access): Use warning_n.
4339 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4342 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4344 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4347 2018-03-30 Julia Koval <julia.koval@intel.com>
4350 * x86-tune.def (movx, partial_reg_dependency): Enable for
4353 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4356 * lra-constraints.c (process_alt_operands): Move setting
4357 this_alternative_matches below.
4359 2018-03-29 Martin Liska <mliska@suse.cz>
4362 * doc/invoke.texi: Document how LTO works with debug info.
4363 Describe auto-load support of binutils. Mention 'x86-64'
4364 as valid option value of -march option.
4366 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4368 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4371 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4372 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4373 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4376 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4379 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4380 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4381 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4382 for RS6000_BTM_POWERPC64.
4383 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4384 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4385 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4390 2018-03-28 Carl Love <cel@us.ibm.com>
4393 2017-09-27 Carl Love <cel@us.ibm.com>
4395 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4396 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4397 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4400 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4402 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4403 instead of __vector bool.
4404 (_mm_max_pu8): Likewise.
4405 (_mm_min_pi16): Likewise.
4407 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4410 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4411 (DIVWEUO): Likewise.
4413 (DIVDEUO): Likewise.
4414 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4415 DIVWEUO and DIVDEUO.
4416 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4417 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4418 (div_extend): Likewise.
4419 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4421 (__builtin_divweuo): Likewise.
4422 (__builtin_divdeo): Likewise.
4423 (__builtin_divdeuo): Likewise.
4425 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4428 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4429 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4431 PR tree-optimization/82004
4432 * gimple-match-head.c (optimize_pow_to_exp): New function.
4433 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4434 Don't fold to exp if optimize_pow_to_exp is false.
4436 2018-03-28 Martin Liska <mliska@suse.cz>
4439 * calls.c (initialize_argument_information): Fix trailing space.
4440 * common.opt: Fix typo and provide better explanation for
4441 -fsanitize-coverage option.
4442 * config/i386/i386.opt: Fix typo.
4444 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4445 Martin Liska <mliska@suse.cz>
4448 * gimplify.c (asan_poison_variable): Don't do the check for
4449 gimplify_omp_ctxp here.
4450 (gimplify_decl_expr): Do it here.
4451 (gimplify_target_expr): Likewise.
4453 2018-03-28 Martin Liska <mliska@suse.cz>
4456 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4457 chkp_type_bounds_count if MPX is not enabled.
4459 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4461 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4463 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4466 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4467 function to create the function decl for complex long double
4468 multiply and divide for -mabi=ieeelongdouble.
4469 (init_float128_ieee): Call it.
4471 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4474 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4475 -fcf-protection=branch -mibt.
4476 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4478 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4481 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4483 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4486 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4487 extern array declarations.
4489 2018-03-27 Richard Biener <rguenther@suse.de>
4492 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4493 explicit single_use checks.
4495 2018-03-27 Richard Biener <rguenther@suse.de>
4497 PR tree-optimization/85082
4498 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4501 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4503 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4504 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4505 Turn on fasynchronous-unwind-tables and funwind-tables.
4507 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4510 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4511 (*bmi_blsr_<mode>_ccz): Ditto.
4513 2018-03-26 Tom de Vries <tom@codesourcery.com>
4515 PR tree-optimization/85063
4516 * omp-general.c (offloading_function_p): New function. Factor out
4518 * omp-offload.c (pass_omp_target_link::gate): ... here.
4519 * omp-general.h (offloading_function_p): Declare.
4520 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4521 with attribute omp declare target for offloading functions.
4523 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4525 PR tree-optimization/84005
4526 * tree-data-ref.h (get_base_for_alignment): Declare.
4527 * tree-data-ref.c (get_base_for_alignment_1): New function.
4528 (get_base_for_alignment): Likewise.
4529 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4530 get_base_for_alignment to find a suitable base object, instead
4531 of always using drb->base_address.
4533 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4536 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4537 known size by default.
4539 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4542 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4543 and non BLKmode operands.
4545 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4548 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4549 Clean up attributes.
4551 2018-03-23 Richard Biener <rguenther@suse.de>
4554 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4555 we are going to emit early debug for LTO.
4557 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4560 * function.c (match_asm_constraints_1): Don't optimize if input
4561 doesn't satisfy general_operand predicate for output's mode.
4564 * alias.c (write_dependence_p): Don't require for x_canonicalized
4565 non-VOIDmode if x has VOIDmode.
4568 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4569 just don't try to optimize it rather than assert it never happens.
4571 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4573 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4574 macro expansions for definition of ST_INTERNAL_<mode> and
4575 LD_INTERNAL_<mode> builtins.
4576 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4578 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4580 (altivec_expand_st_builtin): Likewise.
4581 (altivec_expand_builtin): Remove calls to deleted functions.
4582 (rs6000_address_for_altivec): Delete this function.
4583 * config/rs6000/vector.md: Remove expands for
4584 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4586 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4589 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4590 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4591 re-computing once computed.
4592 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4593 (arm_init_machine_status): Initialize
4594 machine->static_chain_stack_bytes.
4596 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4599 * doc/extend.texi: Add four new prototypes for vec_ld.
4600 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4601 definitions for more logical presentation.
4602 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4603 entries for V1TI variants of __builtin_altivec_ld builtin.
4604 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4605 handling of V1TI variant of LVX icode pattern.
4606 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4607 (rs6000_gimple_fold_builtin): Likewise.
4608 (altivec_init_builtins): Add code to define
4609 __builtin_altivec_lvx_v1ti function.
4611 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4614 * function.c (match_asm_constraints_1): Don't do the optimization
4615 if input isn't a REG, SUBREG, MEM or constant.
4617 2018-03-22 Tom de Vries <tom@codesourcery.com>
4619 PR tree-optimization/84956
4620 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4621 bb_has_abnormal_pred.
4623 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4626 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4627 DECL_INITIAL (decl) to decl at the end.
4628 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4631 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4633 * doc/extend.texi (__builtin_tgmath): Document when complex
4634 integer types are treated as _Complex _Float64.
4636 2018-03-21 Tom de Vries <tom@codesourcery.com>
4638 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4640 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4642 PR tree-optimization/84960
4643 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4644 if it is ENTRY block, move them into single succ of ENTRY in that case.
4646 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4648 PR tree-optimization/84811
4649 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4650 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4651 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4652 (known_subrange_p): Update accordingly. Cast each value involved
4653 in the size comparison, rather than casting the result of the
4656 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4658 PR tree-optimization/84982
4659 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4660 by flipping the least significant bit rather than all bits from
4661 bitpos to bitpos + bitsize - 1.
4663 2018-03-21 Nathan Sidwell <nathan@acm.org>
4665 * doc/extend.texi (Deprecated Features): Remove mention of
4666 long-deleted deprecations.
4668 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4671 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4672 * configure: Regenerate.
4674 2018-03-21 Tom de Vries <tom@codesourcery.com>
4676 PR tree-optimization/83126
4677 * tree-parloops.c (num_phis): New function.
4678 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4680 2018-03-21 Nathan Sidwell <nathan@acm.org>
4682 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4683 mention anon-struct/union members and trailing attributes.
4685 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4687 PR tree-optimization/84969
4688 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4689 builtin memset partitions if they set different rhs values.
4691 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4693 PR rtl-optimization/84989
4694 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4695 VEC_DUPLICATE with scalar result mode.
4697 2018-03-21 Martin Liska <mliska@suse.cz>
4700 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4701 not intended return statement.
4703 2018-03-21 Martin Liska <mliska@suse.cz>
4706 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4707 (chkp_find_bound_slots_1): Limit number of iterations.
4709 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4712 * Minor grammar fixes for x86 options.
4714 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4717 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4718 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4720 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4723 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4724 (altivec_lvx_<mode>_1op): Likewise.
4725 (altivec_stvx_<mode>_2op): Likewise.
4726 (altivec_stvx_<mode>_1op): Likewise.
4727 (altivec_lvx_<VM2:mode>): New define_expand.
4728 (altivec_stvx_<VM2:mode>): Likewise.
4729 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4730 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4731 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4732 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4733 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4734 (rs6000_gen_lvx): Likewise.
4735 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4736 (altivec_expand_stv_builtin): Likewise.
4737 (altivec_expand_builtin): Likewise.
4738 * config/rs6000/vector.md: Likewise.
4740 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4743 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4746 2018-03-20 Richard Biener <rguenther@suse.de>
4749 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4750 sign-conversions as zero, fall back to standard scalar_stmt
4753 2018-03-20 Martin Liska <mliska@suse.cz>
4756 * predict.c (rebuild_frequencies): Handle case when we have
4757 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4759 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4762 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4763 flag_section_anchors.
4764 * varasm.c (use_blocks_for_decl_p): Remove hack for
4765 dw2_force_const_mem.
4768 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4770 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4771 be created, use lowpart_subreg of operands[0] rather than operands[0]
4773 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4774 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4775 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4776 and n constraint instead of aarch64_shift_imm_di and Usd.
4777 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4778 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4780 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4783 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4784 to favor GPR over NEON registers.
4785 (<shift>di3_neon): Likewise.
4787 2018-03-20 Tom de Vries <tom@codesourcery.com>
4790 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4791 (nvptx_process_pars): Emit bar.sync asap and alap.
4793 2018-03-20 Tom de Vries <tom@codesourcery.com>
4796 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4797 seen_label if seen_label is already set.
4799 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4802 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4803 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4804 Use 1U instead of 1. Formatting fixes.
4807 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4808 instead of TREE_TYPE (s1) for the return value.
4810 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4812 PR tree-optimization/84946
4813 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4814 bitsize + bitsize in poly_uint64 rather than poly_int64.
4817 * dwarf2asm.c: Include fold-const.c.
4818 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4819 of decl rather than decl itself.
4821 PR rtl-optimization/84643
4822 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4824 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4827 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4828 calling assemble_variable.
4830 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4833 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4834 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4836 2018-03-19 Jim Wilson <jimw@sifive.com>
4839 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4840 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4841 (riscv_first_stack_step): Likewise.
4842 (riscv_option_override): Use STACK_BOUNDARY instead of
4844 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4846 (BIGGEST_ALIGNMENT): Set to 128.
4847 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4848 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4851 2018-03-19 Richard Biener <rguenther@suse.de>
4853 PR tree-optimization/84933
4854 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4855 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4857 2018-03-19 Richard Biener <rguenther@suse.de>
4859 PR tree-optimization/84859
4860 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4861 (cond_if_else_store_replacement): Perform sinking operation on
4862 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4863 Generalize what a BB with a single eligible store is.
4865 2018-03-19 Richard Biener <rguenther@suse.de>
4867 PR tree-optimization/84929
4868 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4869 chrec_is_positive against non-chrec arg.
4871 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4874 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4876 2018-03-18 Martin Liska <mliska@suse.cz>
4878 PR rtl-optimization/84635
4879 * regrename.c (build_def_use): Use matches_mode only when
4882 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4884 PR tree-optimization/84913
4885 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4886 vectorize chains of COND_EXPRs.
4888 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4890 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4892 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4894 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4896 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4898 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4900 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4901 Kito Cheng <kito.cheng@gmail.com>
4903 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4904 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4905 (nds32_adjust_reg_alloc_order): New function.
4906 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4908 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4910 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4911 nds32_print_operand, nds32_print_operand_address): Use
4912 HOST_WIDE_INT_PRINT_DEC instead.
4914 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4916 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4918 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4921 * config/i386/i386.c (initial_ix86_tune_features,
4922 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4924 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4925 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4926 rather than 1u << ix86_tune. Formatting fix.
4927 (ix86_option_override_internal): Change ix86_arch_mask from
4928 unsigned int to unsigned HOST_WIDE_INT, initialize to
4929 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4930 (ix86_function_specific_restore): Likewise.
4932 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4935 * postreload.c (reload_combine_recognize_pattern): Perform
4936 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4937 truncate_int_for_mode the result for the destination's mode.
4940 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4941 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4943 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4946 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4949 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4950 regno_allocno_class_array and sorted_pseudos.
4951 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4952 insns where regno is used.
4954 2018-03-16 Martin Liska <mliska@suse.cz>
4957 * multiple_target.c (create_dispatcher_calls): Redirect
4958 reference in the symbol table.
4960 2018-03-16 Martin Liska <mliska@suse.cz>
4963 * multiple_target.c (create_dispatcher_calls): Redirect also
4966 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4970 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4971 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4972 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4974 2018-03-16 Julia Koval <julia.koval@intel.com>
4976 * doc/invoke.texi (Skylake Server): Add CLWB.
4977 Cannonlake): Remove CLWB.
4979 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4981 PR tree-optimization/84841
4982 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4984 (FLOAT_ONE_CONST_TYPE): Define.
4985 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4986 (sort_by_operand_rank): Put entries with higher constant_type last
4987 rather than first to match comments.
4989 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4991 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4994 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4997 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4998 check and use address of target always.
5000 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5003 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5004 (indirect_thunk_bnd_needed): Likewise.
5005 (indirect_thunks_used): Likewise.
5006 (indirect_thunks_bnd_used): Likewise.
5007 (indirect_return_needed): New.
5008 (indirect_return_bnd_needed): Likewise.
5009 (output_indirect_thunk_function): Add a bool argument for
5011 (output_indirect_thunk_function): Don't generate alias for
5012 function return thunk.
5013 (ix86_code_end): Call output_indirect_thunk_function to generate
5014 function return thunks.
5015 (ix86_output_function_return): Set indirect_return_bnd_needed
5016 and indirect_return_needed instead of indirect_thunk_bnd_needed
5017 and indirect_thunk_needed.
5019 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5021 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5022 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5023 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5025 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5026 Paul Hua <paul.hua.gm@gmail.com>
5029 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5031 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5033 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5034 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5037 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5040 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5041 instead of GET_MODE_SIZE when comparing Units.
5043 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5046 * varasm.c (hash_section): Return an unchangeble hash value
5047 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5048 Return !aarch64_can_use_per_function_literal_pools_p ().
5050 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5053 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5054 rather than address of cmode as last argument to prepare_cmp_insn.
5056 2018-03-15 Julia Koval <julia.koval@intel.com>
5058 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5059 F_AVX512VNNI, F_AVX512BITALG): New.
5061 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5064 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5065 insn for floating-point loads and stores.
5067 2018-03-14 Carl Love <cel@us.ibm.com>
5069 * config/rs6000/rs6000-c.c: Add macro definitions for
5070 ALTIVEC_BUILTIN_VEC_PERMXOR.
5071 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5072 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5073 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5074 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5076 * config/doc/extend.texi: Add prototypes for vec_permxor.
5078 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5081 * diagnostic-show-locus.c (class layout_point): Convert m_line
5082 from int to linenum_type.
5083 (line_span::comparator): Use linenum "compare" function when
5084 comparing line numbers.
5085 (test_line_span): New function.
5086 (layout_range::contains_point): Convert param "row" from int to
5088 (layout_range::intersects_line_p): Likewise.
5089 (layout::will_show_line_p): Likewise.
5090 (layout::print_source_line): Likewise.
5091 (layout::should_print_annotation_line_p): Likewise.
5092 (layout::print_annotation_line): Likewise.
5093 (layout::print_leading_fixits): Likewise.
5094 (layout::annotation_line_showed_range_p): Likewise.
5095 (struct line_corrections): Likewise for field m_row.
5096 (line_corrections::line_corrections): Likewise for param "row".
5097 (layout::print_trailing_fixits): Likewise.
5098 (layout::get_state_at_point): Likewise.
5099 (layout::get_x_bound_for_row): Likewise.
5100 (layout::print_line): Likewise.
5101 (diagnostic_show_locus): Likewise for locals "last_line" and
5103 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5104 * input.c (selftest::test_linenum_comparisons): New function.
5105 (selftest::input_c_tests): Call it.
5106 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5107 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5108 * selftest.h (ASSERT_GT): New macro.
5109 (ASSERT_GT_AT): New macro.
5110 (ASSERT_LT): New macro.
5111 (ASSERT_LT_AT): New macro.
5113 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5115 PR rtl-optimization/84780
5116 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5118 2018-03-14 Martin Liska <mliska@suse.cz>
5120 * tree.c (record_node_allocation_statistics): Use
5121 get_stats_node_kind.
5122 (get_stats_node_kind): New function extracted from
5123 record_node_allocation_statistics.
5124 (free_node): Use get_stats_node_kind.
5126 2018-03-14 Richard Biener <rguenther@suse.de>
5128 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5129 that the value-set of ANTIC_IN doesn't grow.
5132 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5134 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5135 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5137 2018-03-14 Julia Koval <julia.koval@intel.com>
5139 * config.gcc (icelake-client, icelake-server): New.
5141 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5142 (initial_ix86_arch_features): Ditto.
5143 (PTA_SKYLAKE): Add SGX.
5144 (PTA_ICELAKE): Remove.
5145 (PTA_ICELAKE_CLIENT): New.
5146 (PTA_ICELAKE_SERVER): New.
5147 (ix86_option_override_internal): Split up icelake on icelake client and
5149 (get_builtin_code_for_version): Ditto.
5150 (fold_builtin_cpu): Ditto.
5151 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5152 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5153 * config/i386/i386.h (processor_type): Ditto.
5154 * doc/invoke.texi: Ditto.
5156 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5159 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5160 INTEGER_CST offset, add it together with bitpos / 8 and
5161 sign extend based on POINTER_SIZE.
5165 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5168 * config/i386/constraints.md (Yc): New register constraint.
5169 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5170 Use Yc constraint for alternative 2 of operand 0. Remove
5171 preferred_for_speed attribute.
5173 2018-03-14 Richard Biener <rguenther@suse.de>
5175 PR tree-optimization/84830
5176 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5177 with the old one to avoid oscillations.
5179 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5182 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5184 (assign_by_spills): Return a flag of reload assignment failure.
5185 Do not process the reload assignment failures. Do not spill other
5186 reload pseudos if they has the same reg class. Update n if
5188 (lra_assign): Add a return arg. Set up from the result of
5189 assign_by_spills call.
5190 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5191 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5192 usage_insns if it is not NULL.
5193 (spill_hard_reg_in_range): New function.
5194 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5195 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5196 function prototypes.
5197 (lra_assign): Change prototype.
5198 * lra.c (lra): Add code to deal with fails by splitting hard reg
5201 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5203 * config/riscv/riscv.opt (mrelax): New option.
5204 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5205 "norelax" when riscv_mrelax is disabled.
5206 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5208 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5211 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5212 reassociation for int modes.
5214 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5216 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5217 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5219 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5220 * config/aarch64/aarch64-sve.md
5221 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5222 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5223 (*extend<mode><Vwide>2): Rename to...
5224 (aarch64_sve_extend<mode><Vwide>2): ...this.
5225 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5226 renaming the old pattern to...
5227 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5229 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5230 define_expand, renaming the old pattern to...
5231 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5232 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5233 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5234 account when deciding which SVE instruction the optab should use.
5235 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5237 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5239 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5240 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5241 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5242 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5243 (tlsdesc_small_<mode>): Turn a define_expand and use
5244 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5245 (tlsdesc_small_advsimd_<mode>): ...this.
5246 (tlsdesc_small_sve_<mode>): New pattern.
5248 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5250 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5251 (UNSPEC_UMUL_HIGHPART): New constants.
5252 (MUL_HIGHPART): New int iteraor.
5253 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5254 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5256 (*<su>mul<mode>3_highpart): New define_insn.
5258 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5261 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5264 2018-03-13 Martin Liska <mliska@suse.cz>
5267 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5269 (sem_item_optimizer::~sem_item_optimizer): Release it.
5270 (sem_item_optimizer::merge_classes): Register variable aliases.
5271 (sem_item_optimizer::fixup_pt_set): New function.
5272 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5273 * ipa-icf.h: Declare new variables and functions.
5275 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5278 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5279 integer_pow2p@2 and test integer_pow2p in condition.
5280 (A < 0 ? C : 0): Similarly for @1.
5283 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5284 characters starting at p contain '\0' character, don't look beyond
5288 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5289 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5292 * reg-stack.c (change_stack): Change update_end var from int to
5293 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5294 also call set_block_for_insn on the newly added insns and rescan.
5297 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5298 on the last operand.
5301 * tree.c (stabilize_reference_1): Return save_expr (e) for
5302 STATEMENT_LIST even if it doesn't have side-effects.
5304 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5306 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5308 2018-03-12 Renlin Li <renlin.li@arm.com>
5310 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5311 aarch64_output_scalar_simd_mov_immediate.
5313 2018-03-12 Martin Sebor <msebor@redhat.com>
5315 PR tree-optimization/83456
5316 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5317 for perfectly overlapping calls to memcpy.
5318 (gimple_fold_builtin_memory_chk): Same.
5319 (gimple_fold_builtin_strcpy): Handle no-warning.
5320 (gimple_fold_builtin_stxcpy_chk): Same.
5321 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5323 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5325 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5326 parameter. Use it for SFmode.
5327 (rs6000_function_arg_advance_1): Adjust.
5328 (rs6000_function_arg): Adjust.
5329 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5331 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5333 PR rtl-optimization/84169
5334 PR rtl-optimization/84780
5335 * combine.c (can_combine_p): Check for a 2-insn combination whether
5336 the destination register is used between the two insns, too.
5338 2018-03-12 Richard Biener <rguenther@suse.de>
5340 PR tree-optimization/84803
5341 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5342 for refs DR analysis didn't process.
5344 2018-03-12 Richard Biener <rguenther@suse.de>
5346 PR tree-optimization/84777
5347 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5348 force-vectorize loops ignore whether we are optimizing for size.
5350 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5352 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5353 (TARGET_MD_ASM_ADJUST): Define.
5355 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5356 Kito Cheng <kito.cheng@gmail.com>
5357 Chung-Ju Wu <jasonwucj@gmail.com>
5359 * config/nds32/nds32.c (nds32_compute_stack_frame,
5360 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5361 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5362 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5363 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5364 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5365 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5366 * config/nds32/nds32.md (prologue, epilogue): Use macro
5367 NDS32_V3PUSH_AVAILABLE_P to do checking.
5369 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5372 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5373 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5374 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5375 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5376 addition of most attributes on !orig_type_die or the attribute not
5377 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5379 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5380 Chung-Ju Wu <jasonwucj@gmail.com>
5382 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5384 * config/nds32/nds32.opt (mvh): New option.
5386 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5387 Chung-Ju Wu <jasonwucj@gmail.com>
5389 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5391 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5392 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5395 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5396 Chung-Ju Wu <jasonwucj@gmail.com>
5398 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5400 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5401 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5403 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5404 Kito Cheng <kito.cheng@gmail.com>
5405 Chung-Ju Wu <jasonwucj@gmail.com>
5407 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5408 UNSPEC_FFMISM and UNSPEC_FLMISM.
5409 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5410 for ffb, ffmism and flmism.
5411 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5412 (unspec_ffmism): Ditto.
5413 (unspec_flmism): Ditto.
5414 (nds32_expand_builtin_impl): Check if string extension is available.
5415 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5416 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5418 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5421 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5424 * lra-assigns.c (assign_by_spills): Return a flag of reload
5425 assignment failure. Do not process the reload assignment
5426 failures. Do not spill other reload pseudos if they has the same
5428 (lra_assign): Add a return arg. Set up from the result of
5429 assign_by_spills call.
5430 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5431 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5432 usage_insns if it is not NULL.
5433 (spill_hard_reg_in_range): New function.
5434 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5435 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5436 function prototypes.
5437 (lra_assign): Change prototype.
5438 * lra.c (lra): Add code to deal with fails by splitting hard reg
5441 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5444 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5446 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5449 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5450 (dw_val_node): Add val_symbolic_view.
5451 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5452 (symview_upper_bound): New.
5453 (new_line_info_table): Initialize symviews_since_reset.
5454 (dwarf2out_source_line): Count symviews_since_reset and set
5455 symview_upper_bound.
5456 (dw_val_equal_p): Handle symview.
5457 (add_AT_symview): New.
5458 (print_dw_val): Handle symview.
5459 (attr_checksum, attr_checksum_ordered): Likewise.
5460 (same_dw_val_p, size_of_die): Likewise.
5461 (value_format, output_die): Likewise.
5462 (add_high_low_attributes): Use add_AT_symview for entry_view.
5463 (dwarf2out_finish): Reset symview_upper_bound, clear
5466 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5469 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5470 Add strict argument and use it.
5471 (rs6000_split_multireg_move): Update for new strict argument.
5472 (mem_operand_gpr): Disallow all non-offsettable addresses.
5473 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5475 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5478 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5479 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5480 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5483 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5484 decl, use remap_type if we want to use the type.
5486 2018-03-09 Martin Sebor <msebor@redhat.com>
5488 PR tree-optimization/84526
5489 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5491 (builtin_access::generic_overlap): Be prepared to handle non-array
5494 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5496 PR rtl-optimization/84682
5497 * lra-constraints.c (process_address_1): Check is_address flag
5498 for address constraints.
5499 (process_alt_operands): Likewise.
5500 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5501 preprocess_constraints.
5502 * recog.h (preprocess_constraints): Add oploc parameter.
5504 * recog.c (preprocess_constraints): Test address_operand for
5505 CT_ADDRESS constraints.
5507 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5510 * lra-assigns.c (assign_by_spills): Return a flag of reload
5511 assignment failure. Do not process the reload assignment
5512 failures. Do not spill other reload pseudos if they has the same
5514 (lra_assign): Add a return arg. Set up from the result of
5515 assign_by_spills call.
5516 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5517 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5518 usage_insns if it is not NULL.
5519 (spill_hard_reg_in_range): New function.
5520 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5521 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5522 function prototypes.
5523 (lra_assign): Change prototype.
5524 * lra.c (lra): Add code to deal with fails by splitting hard reg
5527 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5530 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5531 Accept complain bool parameter. Only emit errors if it is true.
5532 (arm_parse_cpu_option_name): Likewise.
5533 (arm_target_thumb_only): Adjust callers of the above.
5534 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5535 prototype to take a default true bool parameter.
5536 (arm_parse_arch_option_name): Likewise.
5538 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5539 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5543 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5544 * configure: Regenerate.
5545 * configure.ac ("linker --version-script option"): New.
5546 ("linker soname option"): New.
5548 2018-03-09 Richard Biener <rguenther@suse.de>
5550 PR tree-optimization/84775
5551 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5552 immediate uses of predicate stmts and mark them modified.
5555 PR tree-optimization/84178
5556 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5558 (version_loop_for_if_conversion): Delay update_ssa call.
5559 (tree_if_conversion): Delay update_ssa until after predicate
5562 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5565 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5566 when the function accesses prior frames.
5568 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5571 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5572 gen_llsym, otherwise call maybe_gen_llsym.
5575 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5576 has ',' character inside of it.
5578 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5581 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5582 as clobbering CC_REGNUM.
5584 2018-03-08 Richard Biener <rguenther@suse.de>
5587 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5588 (follow_copies_to_constant): Do not follow SSA names registered
5591 2018-03-08 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/84178
5594 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5596 (version_loop_for_if_conversion): Delay update_ssa call.
5597 (tree_if_conversion): Delay update_ssa until after predicate
5600 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5602 PR tree-optimization/84178
5603 * tree-if-conv.c (release_bb_predicate): Remove the
5604 the assertion that the stmts have NULL use_ops.
5605 Discard the statements, asserting that they haven't
5606 yet been added to a BB.
5608 2018-03-08 Richard Biener <rguenther@suse.de>
5610 PR tree-optimization/84746
5611 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5612 (phi_translate): Pass in destination ANTIC_OUT set.
5613 (phi_translate_1): Likewise. For a simplified result lookup
5614 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5615 (phi_translate_set): Adjust.
5616 (do_pre_regular_insertion): Likewise.
5617 (do_pre_partial_partial_insertion): Likewise.
5619 2018-03-08 Martin Liska <mliska@suse.cz>
5621 PR gcov-profile/84735
5622 * doc/gcov.texi: Document usage of profile files.
5623 * gcov-io.h: Document changes in the format.
5625 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5629 * dwarf2out.c (struct dw_line_info_table): Update comments for
5631 (FORCE_RESET_NEXT_VIEW): New.
5632 (FORCE_RESETTING_VIEW_P): New.
5633 (RESETTING_VIEW_P): Check for -1 too.
5634 (ZERO_VIEW_P): Likewise.
5635 (new_line_info_table): Force-reset next view.
5636 (dwarf2out_begin_function): Likewise.
5637 (dwarf2out_source_line): Simplify zero_view_p initialization.
5638 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5639 view directly. Omit view when omitting .loc at line 0.
5641 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5643 PR tree-optimization/84740
5644 * tree-switch-conversion.c (process_switch): Call build_constructors
5645 only if info.phi_count is non-zero.
5647 PR tree-optimization/84739
5648 * tree-tailcall.c (find_tail_calls): Check call arguments against
5649 DECL_ARGUMENTS (current_function_decl) rather than
5650 DECL_ARGUMENTS (func) when checking for tail recursion.
5652 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5654 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5655 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5656 Volker Reichelt's entry and add entries for people that perform
5657 GCC fuzzy testing and report numerous bugs.
5659 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5662 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5663 readonly data in sdata, if that is disabled.
5664 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5665 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5666 -mreadonly-in-sdata option.
5668 2018-03-07 Martin Sebor <msebor@redhat.com>
5670 PR tree-optimization/84468
5671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5672 basic block when looking for nul assignment.
5674 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5677 * except.h (output_function_exception_table): Adjust prototype.
5678 * except.c (output_function_exception_table): Remove FNNAME parameter
5679 and add SECTION parameter. Ouput one part of the table at a time.
5680 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5681 the first part of the exception table and emit unwind directives.
5682 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5683 (i386_pe_seh_cold_init): Likewise.
5684 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5685 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5686 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5687 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5688 * config/i386/winnt.c: Include except.h.
5689 (struct seh_frame_state): Add reg_offset, after_prologue and
5690 in_cold_section fields.
5691 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5692 (i386_pe_seh_cold_init): New function.
5693 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5694 to seh->in_cold_section.
5695 (seh_emit_push): Record the offset of the push.
5696 (seh_emit_save): Record the offet of the save.
5697 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5698 Test seh->after_prologue to disregard the epilogue.
5699 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5700 (i386_pe_end_cold_function): New function.
5702 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5705 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5706 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5709 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5711 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5712 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5715 * multiple_target.c: Include tree-inline.h and intl.h.
5716 (expand_target_clones): Diagnose and fail if node->definition and
5717 !tree_versionable_function_p (node->decl).
5719 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5721 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5723 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5724 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5725 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5727 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5730 * combine.c (try_combine): Use reg_or_subregno instead of handling
5731 just paradoxical SUBREGs and REGs.
5733 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5735 * config/arc/arc.c (arc_finalize_pic): Remove function.
5736 (arc_must_save_register): We use single base PIC register, remove
5737 checks to save/restore the PIC register.
5738 (arc_expand_prologue): Likewise.
5739 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5741 (arc_verify_short): Likewise.
5742 (arc_attr_type): Likewise.
5743 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5744 (walk_stores): Likewise.
5745 (arc_address_cost): Make it static.
5746 (arc_verify_short): Likewise.
5747 (branch_dest): Likewise.
5748 (arc_attr_type): Likewise.
5749 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5750 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5751 (arc_final_prescan_insn): Remove inserting the nops due to
5752 hardware hazards. It is done in reorg step.
5753 (insn_length_variant_t): Remove.
5754 (insn_length_parameters_t): Likewise.
5755 (arc_insn_length_parameters): Likewise.
5756 (arc_get_insn_variants): Likewise.
5757 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5759 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5762 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5765 PR tree-optimization/84687
5766 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5768 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5770 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5772 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5773 Rename to ppc_speculation_barrier.
5774 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5775 __builtin_ppc_speculation_barrier.
5777 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5780 * combine.c (combine_simplify_rtx): Don't try to simplify if
5781 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5784 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5786 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5787 to 32 bytes when compiling for POWER9.
5789 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5792 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5793 regparm >= 3 with no arg reg available also for calls with
5794 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5797 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5799 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5801 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5804 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5806 2018-03-05 Richard Biener <rguenther@suse.de>
5808 PR tree-optimization/84486
5809 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5810 When inserting a __builtin_assume_aligned call set the LHS
5811 SSA name alignment info accordingly.
5813 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5815 PR tree-optimization/84114
5816 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5817 Avoid reassociation of FLOAT_MODE addition.
5819 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5821 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5822 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5823 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5824 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5825 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5826 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5827 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5829 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5830 __builtin_ia32_wbinvd): New builtins.
5831 (SPECIAL_ARGS2): New.
5832 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5833 (SPECIAL_ARGS2): New.
5834 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5835 (ix86_valid_target_attribute_inner_p): Ditto.
5836 (ix86_init_mmx_sse_builtins): Add special_args2.
5837 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5838 TARGET_WBNOINVD_P): New.
5839 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5840 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5841 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5842 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5843 * config/i386/pconfigintrin.h: New file.
5844 * config/i386/wbnoinvdintrin.h: Ditto.
5845 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5846 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5848 2018-03-05 Richard Biener <rguenther@suse.de>
5850 PR tree-optimization/84670
5851 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5853 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5854 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5855 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5856 doesn't grow if all successors (recursively) were visited at least
5859 2018-03-05 Richard Biener <rguenther@suse.de>
5861 PR tree-optimization/84650
5862 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5863 if executed in the loop pipeline.
5865 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5867 * doc/configfiles.texi (Configuration Files): Move info about
5868 conditionalizing $target-protos.h to...
5869 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5870 differs from $target-protos.h.
5872 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5873 Chung-Ju Wu <jasonwucj@gmail.com>
5875 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5876 * config/nds32/nds32-multiple.md (setmemsi): Define.
5877 * config/nds32/nds32-memory-manipulation.c
5878 (nds32_gen_dup_4_byte_to_word_value): New.
5879 (emit_setmem_word_loop): New.
5880 (emit_setmem_byte_loop): New.
5881 (nds32_expand_setmem_loop): New.
5882 (nds32_expand_setmem_loop_v3m): New.
5883 (nds32_expand_setmem_unroll): New.
5884 (nds32_expand_setmem): New.
5886 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5887 Chung-Ju Wu <jasonwucj@gmail.com>
5889 * config/nds32/nds32-memory-manipulation.c
5890 (nds32_emit_load_store): New.
5891 (nds32_emit_post_inc_load_store): New.
5892 (nds32_emit_mem_move): New.
5893 (nds32_emit_mem_move_block): New.
5894 (nds32_expand_movmemsi_loop_unknown_size): New.
5895 (nds32_expand_movmemsi_loop_known_size): New.
5896 (nds32_expand_movmemsi_loop): New.
5897 (nds32_expand_movmemsi_unroll): New.
5898 (nds32_expand_movmemqi): Rename ...
5899 (nds32_expand_movmemsi): ... to this.
5900 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5901 (movmemsi): ... to this.
5902 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5903 (nds32_expand_movmemsi): ... to this.
5905 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5906 Monk Chiang <sh.chiang04@gmail.com>
5907 Chung-Ju Wu <jasonwucj@gmail.com>
5909 * config/nds32/nds32-protos.h
5910 (nds32_expand_load_multiple): New arguments.
5911 (nds32_expand_store_multiple): Ditto.
5912 (nds32_valid_multiple_load_store): Rename ...
5913 (nds32_valid_multiple_load_store_p): ... to this.
5914 * config/nds32/nds32-memory-manipulation.c
5915 (nds32_expand_load_multiple): Refine implementation.
5916 (nds32_expand_store_multiple): Ditto.
5917 * config/nds32/nds32-multiple.md
5918 (load_multiple): Update nds32_expand_load_multiple interface.
5919 (store_multiple): Update nds32_expand_store_multiple interface.
5920 * config/nds32/nds32-predicates.c
5921 (nds32_valid_multiple_load_store): Rename ...
5922 (nds32_valid_multiple_load_store_p): ... to this and refine
5924 * config/nds32/predicates.md
5925 (nds32_load_multiple_and_update_address_operation): New predicate.
5926 (nds32_store_multiple_and_update_address_operation): New predicate.
5928 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5929 Chung-Ju Wu <jasonwucj@gmail.com>
5931 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5932 (combo): New attribute.
5933 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5935 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5937 * config/nds32/nds32.opt: Change -mcmodel= default value.
5939 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5940 Monk Chiang <sh.chiang04@gmail.com>
5941 Chung-Ju Wu <jasonwucj@gmail.com>
5943 * config/nds32/constants.md (unspec_element): New enum.
5944 * config/nds32/constraints.md (Umw): New constraint.
5945 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5946 * config/nds32/nds32-intrinsic.md: Likewise.
5947 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5948 (nds32_valid_smw_lwm_base_p): New.
5949 (nds32_output_smw_single_word): New.
5950 (nds32_output_lmw_single_word): New.
5951 (nds32_expand_unaligned_load): New.
5952 (nds32_expand_unaligned_store): New.
5953 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5954 (nds32_output_smw_single_word): Declare.
5955 (nds32_output_lmw_single_word): Declare.
5956 (nds32_expand_unaligned_load): Declare.
5957 (nds32_expand_unaligned_store): Declare.
5958 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5959 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5960 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5961 NDS32_BUILTIN_UASTORE_DW.
5962 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5965 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5966 Kito Cheng <kito.cheng@gmail.com>
5967 Chung-Ju Wu <jasonwucj@gmail.com>
5969 * config/nds32/nds32-intrinsic.c
5970 (nds32_expand_builtin_null_ftype_reg): Delete.
5971 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5972 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5973 (nds32_read_argument): New.
5974 (nds32_legitimize_target): Ditto.
5975 (nds32_legitimize_argument): Ditto.
5976 (nds32_check_constant_argument): Ditto.
5977 (nds32_expand_unop_builtin): Ditto.
5978 (nds32_expand_unopimm_builtin): Ditto.
5979 (nds32_expand_binop_builtin): Ditto.
5980 (nds32_builtin_decl_impl): Ditto.
5981 (builtin_description): Ditto.
5982 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5983 (nds32_init_builtins_impl): Ditto.
5984 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5985 (nds32_builtin_decl): New.
5986 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5987 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5989 2018-03-02 Jeff Law <law@redhat.com>
5991 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5992 (redundant_insn, fill_simple_delay_slots): Likewise.
5993 (fill_slots_from_thread): Likewise.
5994 * resource.c (mark_referenced_resources): Likewise.
5995 (mark_set_resources, find_dead_or_set_registers): Likewise.
5997 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5999 * substring-locations.h (format_warning_va): Formatting fix for
6001 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6003 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6004 * substring-locations.c: Include intl.h.
6005 (format_warning_va): Turned into small wrapper around
6006 format_warning_n_va, renamed to ...
6007 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6008 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6010 (format_warning_at_substring_n): New function.
6011 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6012 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6013 format_warning_at_substring with just a shorter name instead of
6014 const function pointer.
6015 (fmtwarn_n): New function.
6016 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6017 appropriate, get rid of all the fmtstr temporaries, move conditionals
6018 with G_() wrapped string literals directly into fmtwarn arguments,
6019 cast dir.len to (int), formatting fixes.
6021 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6023 * doc/invoke.texi: Remove "Cilk Plus" references.
6025 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6026 Richard Biener <rguenther@suse.de>
6029 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6030 for error or warning attributes if CALL_FROM_THUNK_P is set.
6033 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6036 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6037 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6040 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6041 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6043 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6044 instead of -1U in last predictors element's probability member.
6046 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6049 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6050 arguments if they are comparable.
6052 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6054 PR tree-optimization/84634
6055 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6056 masks and masked_loop_p with a single loop_masks, making sure it's
6057 null for bb vectorization.
6059 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6061 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6062 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6063 loop->force_vectorize to check whether there is no alias.
6065 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6068 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6070 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6072 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6073 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6074 instead of a loop around prev_real_insn.
6075 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6079 * config/i386/i386.c (ix86_print_operand): Use conditional
6080 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6083 2018-03-02 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/84427
6086 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6087 (bitmap_set_subtract_values): Rewrite to handle multiple
6090 (prune_clobbered_mems): Likewise.
6091 (phi_translate): Take edge instead of pred/phiblock.
6092 (phi_translate_1): Likewise.
6093 (phi_translate_set): Likewise. Insert all translated
6094 exprs for a value into the set, keeping possibly multiple
6095 expressions per value.
6096 (compute_antic_aux): Adjust for phi_translate changes.
6097 When intersecting union the expressions and prune those
6098 not in the final value set, keeping possibly multiple
6099 expressions per value. Do not use value-insertion
6100 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6101 all expressions. Add verification that the value-sets
6102 only shrink during iteration.
6103 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6104 (do_pre_regular_insertion): Likewise.
6105 (do_pre_partial_partial_insertion): Likewise.
6107 2018-03-02 Richard Biener <rguenther@suse.de>
6110 * config/darwin.c (saved_debug_info_level): New static global.
6111 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6112 (darwin_asm_lto_end): Restore debug info generation settings.
6114 2018-03-01 Martin Liska <mliska@suse.cz>
6117 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6120 2018-03-01 Richard Biener <rguenther@suse.de>
6123 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6124 type annotation with LTO when debug was disabled at compile-time.
6126 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6128 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6130 (mips_final_postscan_insn): Likewise.
6132 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6134 PR rtl-optimization/84528
6135 * alias.c (init_alias_target): Add commentary.
6136 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6137 a unique base value if the frame pointer is not eliminated
6138 to the stack pointer.
6140 2018-03-01 Tom de Vries <tom@codesourcery.com>
6142 PR rtl-optimization/83327
6143 * lra-int.h (hard_regs_spilled_into): Declare.
6144 * lra.c (hard_regs_spilled_into): Define.
6145 (init_reg_info): Init hard_regs_spilled_into.
6146 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6147 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6148 (process_bb_lives): Handle hard_regs_spilled_into.
6149 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6150 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6152 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6154 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6155 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6156 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6157 * config/rs6000/aix72.h: New file.
6159 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6161 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6162 instead of warning_at with conditional singular and plural messages
6166 * stor-layout.c (update_alignment_for_field): For
6167 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6168 && !DECL_PACKED (field), do the alignment update, just use
6169 only desired_align instead of MAX (type_align, desired_align)
6171 (place_field): Don't do known_align < desired_align handling
6172 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6173 is non-NULL, instead do it after rli->prev_field handling and
6174 only if not within a bitfield word. For DECL_PACKED (field)
6175 use type_align of BITS_PER_UNIT.
6177 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6179 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6180 superfluous parentheses and trailing spaces.
6182 2018-02-28 Richard Biener <rguenther@suse.de>
6184 PR tree-optimization/84584
6185 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6186 SCoPs with fake exit edge.
6188 2018-02-28 Martin Liska <mliska@suse.cz>
6191 * timevar.c (timer::print): Fix format to properly print 100%
6194 2018-02-28 Richard Biener <rguenther@suse.de>
6197 * genmatch.c (capture_info::walk_match): Do not mark
6198 captured expressions without operands as expr_p given
6199 they act more like predicates and should be subject to
6200 "lost tail" side-effect preserving.
6202 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6204 PR rtl-optimization/81611
6205 * auto-inc-dec.c (attempt_change): Move dead note from
6206 mem_insn if it's the next use of regno
6207 (find_address): Take address use of reg holding
6208 non-incremented value. Add parm to limit search to the named
6210 (merge_in_block): Attempt to use a mem insn that is the next
6211 use of the original regno.
6213 2018-02-27 Martin Sebor <msebor@redhat.com>
6216 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6217 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6219 2018-02-27 Martin Sebor <msebor@redhat.com>
6221 PR translation/84207
6222 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6223 n argument to unsigned HOST_WIDE_INT.
6224 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6225 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6226 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6227 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6229 2018-02-27 Richard Biener <rguenther@suse.de>
6231 PR tree-optimization/84512
6232 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6233 Do not use the estimate returned from record_stmt_cost for
6234 the scalar iteration cost but sum properly using add_stmt_cost.
6236 2018-02-27 Richard Biener <rguenther@suse.de>
6238 PR tree-optimization/84466
6239 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6240 Adjust last change to less strictly validate use operands.
6242 2018-02-27 Martin Liska <mliska@suse.cz>
6244 PR gcov-profile/84548
6245 * gcov.c (process_file): Allow partial overlap and consider it
6246 also as group functions.
6247 (output_lines): Properly calculate range of lines for a group.
6249 2018-02-27 Martin Liska <mliska@suse.cz>
6251 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6252 'ggc' suffixes. Change first column width.
6253 (timer::print): Fix formatting of the column.
6255 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6257 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6258 preserve inline entry blocks for the sake of debug inline
6259 entry point markers alone.
6260 (remove_unused_locals): Suggest in comments a better place to
6261 force the preservation of inline entry blocks that are
6262 otherwise unused, but do not preserve them.
6264 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6266 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6268 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6271 * config/i386/constraints.md (Bs): Replace
6272 ix86_indirect_branch_register with
6273 TARGET_INDIRECT_BRANCH_REGISTER.
6275 * config/i386/i386.md (indirect_jump): Likewise.
6276 (tablejump): Likewise.
6277 (*sibcall_memory): Likewise.
6278 (*sibcall_value_memory): Likewise.
6279 Peepholes of indirect call and jump via memory: Likewise.
6280 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6281 (*sibcall_value_GOT_32): Likewise.
6282 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6283 (GOT_memory_operand): Likewise.
6284 (call_insn_operand): Likewise.
6285 (sibcall_insn_operand): Likewise.
6286 (GOT32_symbol_operand): Likewise.
6287 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6289 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6291 PR rtl-optimization/83496
6292 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6293 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6294 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6295 redundant insn, if any.
6296 (relax_delay_slots): Likewise.
6297 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6299 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6301 PR tree-optimization/83965
6302 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6303 that grouped statements are part of a reduction chain. Return
6304 true if the statement is not marked as a reduction itself but
6306 (vect_recog_dot_prod_pattern): Don't check whether the statement
6307 is part of a group here.
6308 (vect_recog_sad_pattern): Likewise.
6309 (vect_recog_widen_sum_pattern): Likewise.
6311 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6314 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6316 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6319 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6321 (ix86_output_indirect_function_return): New prototype.
6322 (ix86_split_simple_return_pop_internal): Likewise.
6323 * config/i386/i386.c (indirect_return_via_cx): New.
6324 (indirect_return_via_cx_bnd): Likewise.
6325 (indirect_thunk_name): Handle return va CX_REG.
6326 (output_indirect_thunk_function): Create alias for
6327 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6328 (ix86_output_indirect_jmp): Remove the bool argument.
6329 (ix86_output_indirect_function_return): New function.
6330 (ix86_split_simple_return_pop_internal): Likewise.
6331 * config/i386/i386.md (*indirect_jump): Don't pass false
6332 to ix86_output_indirect_jmp.
6333 (*tablejump_1): Likewise.
6334 (simple_return_pop_internal): Change it to define_insn_and_split.
6335 Call ix86_split_simple_return_pop_internal to split it for
6337 (simple_return_indirect_internal): Call
6338 ix86_output_indirect_function_return instead of
6339 ix86_output_indirect_jmp.
6341 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6344 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6345 memset and value initialization afterwards.
6347 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6349 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6351 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6354 * common/config/aarch64/aarch64-common.c
6355 (aarch_option_optimization_table[]): Switch
6356 off fomit-frame-pointer
6358 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6359 Chung-Ju Wu <jasonwucj@gmail.com>
6361 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6363 (store_multiple): Ditto.
6365 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6367 * config.gcc: Add --with-cpu support for nds32 target.
6368 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6369 * config/nds32/nds32.opt: Add -mcpu= option.
6371 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6373 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6374 isel=yes): Warn for these deprecated options.
6376 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6378 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6379 ISA_2_5_MASKS_EMBEDDED.
6381 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6383 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6384 p->max as pointers rather than using iterative_hash_expr.
6386 2018-02-23 Carl Love <cel@us.ibm.com>
6388 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6389 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6391 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6392 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6393 P8V_BUILTIN_VEC_VUNSIGNED2.
6395 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6398 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6399 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6401 * lra-constraints.c (curr_insn_transform): Set up
6402 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6404 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6405 * lra-eliminations.c (spill_pseudos): Ditto.
6406 (process_insn_for_elimination): Ditto.
6407 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6408 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6409 LRA_NON_CLOBBERED_ALT.
6411 2018-02-22 Martin Sebor <msebor@redhat.com>
6413 PR tree-optimization/84480
6414 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6415 to maybe_diag_stxncpy_trunc. Call it.
6416 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6417 from gimple_fold_builtin_strcpy. Print inlining stack.
6418 (handle_builtin_stxncpy): Print inlining stack.
6419 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6421 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6424 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6425 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6426 and -fcheck-pointer-bounds are used together.
6427 (indirect_thunk_prefix): New enum.
6428 (indirect_thunk_need_prefix): New function.
6429 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6430 "_nt" instead of "_bnd" for NOTRACK prefix.
6431 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6432 (output_indirect_thunk_function): Likewise.
6434 (ix86_code_end): Update output_indirect_thunk_function calls.
6435 (ix86_output_indirect_branch_via_reg): Replace
6436 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6437 (ix86_output_indirect_branch_via_push): Likewise.
6438 (ix86_output_function_return): Likewise.
6439 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6440 incompatible with -fcf-protection=branch and
6441 -fcheck-pointer-bounds.
6443 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6446 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6447 Change gcc_assert call to output_operand_lossage.
6449 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6451 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6453 2018-02-22 DJ Delorie <dj@redhat.com>
6454 Sebastian Perta <sebastian.perta@renesas.com>
6455 Oleg Endo <olegendo@gcc.gnu.org>
6457 * config/rx/rx.c (rx_rtx_costs): New function.
6458 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6460 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6462 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6464 2018-02-22 Martin Liska <mliska@suse.cz>
6467 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6468 Add "native" as a possible value.
6470 2018-02-22 Martin Liska <mliska@suse.cz>
6473 * config/i386/i386.c (ix86_option_override_internal):
6474 Add "native" as a possible value for -march and -mtune.
6476 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6479 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6480 to all type variants.
6482 PR tree-optimization/84503
6483 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6484 width as info->bitpos + info->bitsize - start.
6485 (merged_store_group::merge_overlapping): Simplify width computation.
6486 (check_no_overlap): New function.
6487 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6488 start + width and last_order of the group, fail if check_no_overlap
6490 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6491 to group if check_no_overlap fails.
6493 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6495 * config/rs6000/altivec.md: Delete contraint arguments to
6496 define_expand, define_split, and define_peephole2, and in
6497 define_insn_and_split if always unused.
6498 * config/rs6000/darwin.md: Ditto.
6499 * config/rs6000/dfp.md: Ditto.
6500 * config/rs6000/rs6000.md: Ditto.
6501 * config/rs6000/sync.md: Ditto.
6502 * config/rs6000/vector.md: Ditto.
6503 * config/rs6000/vsx.md: Ditto.
6505 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6507 * config/rs6000/altivec.md: Write output control strings as braced
6508 blocks instead of double-quoted strings.
6509 * config/rs6000/darwin.md: Ditto.
6510 * config/rs6000/rs6000.md: Ditto.
6511 * config/rs6000/vector.md: Ditto.
6512 * config/rs6000/vsx.md: Ditto.
6514 2018-02-21 Jason Merrill <jason@redhat.com>
6516 PR c++/84314 - ICE with templates and fastcall attribute.
6517 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6519 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6521 * ipa-cp.c (determine_versionability): Fix comment typos.
6523 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6526 * ipa-cp.c (determine_versionability): Do not version functions caling
6529 2018-02-21 Martin Liska <mliska@suse.cz>
6532 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6533 Add "native" as a possible value.
6534 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6535 the macro when native cpu detection is available.
6537 2018-02-21 Martin Liska <mliska@suse.cz>
6540 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6541 Add "native" as a possible value.
6542 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6543 when native cpu detection is available.
6545 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6546 Martin Sebor <msebor@redhat.com>
6548 PR tree-optimization/84478
6549 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6551 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6552 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6553 support which is conservatively correct, for 2 only stay conservative
6554 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6555 argument to the 2 argument get_range_strlen, adjust 6 arg
6556 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6558 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6559 (gimple_fold_builtin_strlen): Pass true as last argument to
6562 2018-02-20 Martin Sebor <msebor@redhat.com>
6565 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6566 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6567 (builtin_memref::builtin_memref): Factor out parts into
6568 set_base_and_offset and call it.
6570 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6573 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6574 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6575 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6576 search at the associated MODE_INT.
6578 2018-02-20 Jeff Law <law@redhat.com>
6581 PR tree-optimization/81592
6583 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6584 for range data rather than using global data.
6585 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6586 range data rather than using global data.
6587 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6588 pass it to children as needed.
6589 (struct directive::fmtresult): Similarly.
6590 (struct directive::set_width): Similarly.
6591 (struct directive::set_precision): Similarly.
6592 (format_integer, format_directive, parse_directive): Similarly.
6593 (format_none): Accept unnamed vr_values parameter.
6594 (format_percent, format_floating, format_character): Similarly.
6595 (format_string, format_plain): Similarly.
6596 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6597 the EVRP range analyzer for range data rather than using global data.
6598 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6599 gimple-ssa-evrp-analyze.h
6600 (class sprintf_dom_walker): Add after_dom_children member function.
6601 Add evrp_range_analyzer member.
6602 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6603 range analyzer as needed.
6604 (sprintf_dom_walker::after_dom_children): New member function.
6605 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6607 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6608 (evrp_range_analyzer::pop_to_marker): Likewise.
6610 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6612 PR tree-optimization/84419
6613 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6614 with the required type if its current type is compatible but
6617 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6620 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6621 after vectorization.
6623 2018-02-20 Martin Liska <mliska@suse.cz>
6626 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6627 possible values if we don't have a hint.
6629 2018-02-20 Martin Liska <mliska@suse.cz>
6633 * final.c (shorten_branches): Build align_tab array with one
6635 * opts.c (finish_options): Add alignment option limit check.
6636 (MAX_CODE_ALIGN): Likewise.
6637 (MAX_CODE_ALIGN_VALUE): Likewise.
6638 * doc/invoke.texi: Document maximum allowed option value for
6639 all -falign-* options.
6641 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6644 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6645 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6646 * var-tracking.c (emit_note_insn_var_location): Remove all references
6647 to NOTE_INSN_CALL_ARG_LOCATION.
6648 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6649 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6650 Use copy_rtx_if_shared.
6651 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6652 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6653 (dwarf2out_var_location): Remove handling of
6654 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6656 * final.c (final_scan_insn): Remove all references to
6657 NOTE_INSN_CALL_ARG_LOCATION.
6658 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6659 before dumping final insns.
6660 * except.c (emit_note_eh_region_end): Remove all references to
6661 NOTE_INSN_CALL_ARG_LOCATION.
6662 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6663 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6664 * config/arc/arc.c (hwloop_optimize): Likewise.
6665 * config/arm/arm.c (create_fix_barrier): Likewise.
6666 * config/s390/s390.c (s390_chunkify_start): Likewise.
6667 * config/sh/sh.c (find_barrier): Likewise.
6668 * config/i386/i386.c (rest_of_insert_endbranch,
6669 ix86_seh_fixup_eh_fallthru): Likewise.
6670 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6671 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6672 * config/frv/frv.c (frv_function_prologue): Likewise.
6673 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6675 (note_outside_basic_block_p): Remove all references to
6676 NOTE_INSN_CALL_ARG_LOCATION.
6677 * gengtype.c (adjust_field_rtx_def): Likewise.
6678 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6680 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6681 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6684 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6687 PR tree-optimization/84452
6688 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6689 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6692 2018-02-19 Martin Liska <mliska@suse.cz>
6695 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6697 2018-02-19 Martin Liska <mliska@suse.cz>
6698 Richard Sandiford <richard.sandiford@linaro.org>
6700 PR tree-optimization/82491
6701 * gimple-fold.c (get_base_constructor): Make earlier bail out
6704 2018-02-19 Carl Love <cel@us.ibm.com>
6706 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6707 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6709 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6710 P8V_BUILTIN_VEC_NEG.
6712 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6714 * config/rl78/rl78.md (movdf): New define expand.
6716 2018-02-19 Martin Liska <mliska@suse.cz>
6719 * doc/invoke.texi: Fix typo.
6720 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6722 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6724 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6725 handle rs6000_single_float and rs6000_double_float specially for
6728 2018-02-16 Jeff Law <law@redhat.com>
6730 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6733 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6736 * ipa-inline.c (inline_small_functions): Fix a typo.
6738 2018-02-16 Nathan Sidwell <nathan@acm.org>
6740 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6742 2018-02-16 Carl Love <cel@us.ibm.com>
6744 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6745 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6746 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6747 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6748 expansion to P8V_BUILTIN_VEC_FLOAT2.
6750 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6752 PR rtl-optimization/70023
6753 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6754 src_regno into account.
6756 2018-02-16 Carl Love <cel@us.ibm.com>
6758 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6759 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6760 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6761 * config/rs6000/rs6000.c: Remove case statements for
6762 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6763 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6764 and P9V_BUILTIN_VEC_VINSERT4B.
6765 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6766 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6767 * config/rs6000/vsx.md:
6768 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6771 2018-02-16 Carl Love <cel@us.ibm.com>
6773 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6775 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6777 * config/rs6000/rs6000-c.c: Add the definitions for
6778 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6779 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6780 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6781 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6782 definition for insert4b and define insn *insert3b_internal.
6783 * doc/extend.texi: Add documentation for vec_extract4b.
6785 2018-02-16 Nathan Sidwell <nathan@acm.org>
6787 * doc/extend.texi (Backwards Compatibility): Mention friend
6788 injection. Note for-scope is deprecated.
6789 * doc/invoke.texi (-ffriend-injection): Deprecate.
6791 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6793 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6794 that moved to I2, also allow destinations that are a paradoxical
6795 subreg (instead of a normal reg).
6797 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6800 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6803 2018-02-16 Richard Biener <rguenther@suse.de>
6805 PR tree-optimization/84037
6806 PR tree-optimization/84016
6808 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6809 Adjust vec_construct for the fact we need additional higher latency
6810 128bit inserts for AVX256 and AVX512 vector builds.
6811 (ix86_add_stmt_cost): Scale vector construction cost for
6814 2018-02-16 Richard Biener <rguenther@suse.de>
6816 PR tree-optimization/84417
6817 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6818 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6819 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6821 2018-02-16 Martin Liska <mliska@suse.cz>
6824 * internal-fn.def (ASAN_CHECK): Set proper flags.
6825 (ASAN_MARK): Likewise.
6827 2018-02-16 Julia Koval <julia.koval@intel.com>
6829 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6830 from PTA_CANNONLAKE.
6832 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6835 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6836 Use ++iter rather than iter++ for std::list iterators.
6837 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6838 defer deleting them until all nodes in the forest are processed. Do
6839 free even leaf nodes. Change to_process into auto_vec.
6842 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6843 * vec.h (vec_default_construct): Use memset instead of placement new
6844 if BROKEN_VALUE_INITIALIZATION is defined.
6845 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6846 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6849 PR rtl-optimization/83723
6850 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6851 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6852 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6854 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6856 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6858 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6860 PR rtl-optimization/81443
6861 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6862 from inner REGs to paradoxical SUBREGs.
6864 2018-02-16 Richard Biener <rguenther@suse.de>
6866 PR tree-optimization/84399
6867 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6868 For operands we can analyze at their definition make sure we can
6869 analyze them at each use as well.
6871 2018-02-16 Richard Biener <rguenther@suse.de>
6873 PR tree-optimization/84190
6874 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6875 volatile accesses if the decl isn't volatile.
6877 2018-02-15 Jason Merrill <jason@redhat.com>
6879 PR c++/84314 - ICE with templates and fastcall attribute.
6880 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6881 TYPE_CANONICAL on an existing type.
6883 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6885 PR tree-optimization/84383
6886 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6887 dstoff nor call operand_equal_p if dstbase is NULL.
6889 PR tree-optimization/84334
6890 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6891 also a CONSTANT_CLASS_P, punt.
6893 2018-02-14 Jim Wilson <jimw@sifive.com>
6895 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6896 first SMALL_OPERAND check. New local min_second_step. Move assert
6897 to where locals are set. Add TARGET_RVC support.
6898 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6900 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6902 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6904 2018-02-14 Martin Sebor <msebor@redhat.com>
6906 PR tree-optimization/83698
6907 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6908 arrays constrain the offset range to their bounds.
6909 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6910 (builtin_access::overlap): Avoid setting the size of overlap if it's
6912 (maybe_diag_overlap): Also consider arrays when deciding what values
6913 of offsets to include in diagnostics.
6915 2018-02-14 Martin Sebor <msebor@redhat.com>
6918 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6919 that correspond to the kind of a declaration.
6921 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6924 * config/pa/pa.md: Load address of PIC label using the linkage table
6925 if the label is nonlocal.
6927 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6929 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6930 warning message if user requests -maltivec=be.
6931 * doc/invoke.texi: Document deprecation of -maltivec=be.
6933 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6936 * config/rs6000/rs6000-c.c: Update definitions for
6937 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6938 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6940 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6943 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6944 add _get_ssp intrinsics. Remove argument from
6945 __builtin_ia32_rdssp[d|q].
6946 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6947 * config/i386/i386-builtin.def: Remove argument from
6948 __builtin_ia32_rdssp[d|q].
6949 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6950 ix86_expand_special_args_builtin for _rdssp[d|q].
6951 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6952 Clear register before usage.
6953 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6954 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6956 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6958 PR tree-optimization/84357
6959 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6960 operand 1 of an ARRAY_REF too.
6962 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6965 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6966 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6968 (set_of_reg): New struct.
6969 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6970 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6971 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6973 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6974 Split into bitclr, bitset, bitinvert patterns if appropriate.
6975 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6976 use rx_fuse_in_memory_bitop.
6977 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6978 to named insn, correct maximum insn length.
6980 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6983 * machmode.def: Define a complex mode for PARTIAL_INT.
6984 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6986 * doc/rtl.texi: Document CSPImode.
6987 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6989 (msp430_hard_regno_nregs_with_padding): Likewise.
6991 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6994 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6996 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6998 PR rtl-optimization/84169
6999 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7000 we generated a parallel as new i3 and we split that to new i2 and i3
7001 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7002 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7003 those to i2, not i1. Partially rewrite this scan code.
7005 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7008 * stor-layout.c (place_field): For variable length fields, adjust
7009 offset_align afterwards not just based on the field's alignment,
7010 but also on the size.
7013 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7014 of exps and logs in the use_exp2 case.
7016 2018-02-13 Jeff Law <law@redhat.com>
7018 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7021 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7024 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7028 * common.opt (gas-loc-support, gas-locview-support): New.
7029 (ginline-points, ginternal-reset-location-views): New.
7030 * doc/invoke.texi: Document them. Use @itemx where intended.
7031 (gvariable-location-views): Adjust.
7032 * target.def (reset_location_view): New.
7033 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7034 (TARGET_RESET_LOCATION_VIEW): New.
7035 * doc/tm.texi: Rebuilt.
7036 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7037 (dwarf2out_default_as_locview_support): New.
7038 (output_asm_line_debug_info): Use option variables.
7039 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7040 (output_loc_list): Likewise.
7041 (add_high_low_attributes): Check option variables.
7042 Don't output entry view attribute in strict mode.
7043 (gen_inlined_subroutine_die): Check option variables.
7044 (dwarf2out_inline_entry): Likewise.
7045 (init_sections_and_labels): Likewise.
7046 (dwarf2out_early_finish): Likewise.
7047 (maybe_reset_location_view): New, from...
7048 (dwarf2out_var_location): ... here. Call it.
7049 * debug.h (dwarf2out_default_as_loc_support): Declare.
7050 (dwarf2out_default_as_locview_support): Declare.
7051 * hooks.c (hook_int_rtx_insn_0): New.
7052 * hooks.h (hook_int_rtx_insn_0): Declare.
7053 * toplev.c (process_options): Take -gas-loc-support and
7054 -gas-locview-support from dwarf2out. Enable
7055 -gvariable-location-views by default only with locview
7056 assembler support. Enable -ginternal-reset-location-views by
7057 default only if the target defines the corresponding hook.
7058 Enable -ginline-points by default if location views are
7059 enabled; force it disabled if statement frontiers are
7061 * tree-inline.c (expand_call_inline): Check option variables.
7062 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7064 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7066 PR tree-optimization/84321
7067 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7068 handling. Also check whether the anti-range contains any values
7069 that satisfy the mask; switch to a VR_RANGE if not.
7071 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7074 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7076 2018-02-13 Martin Jambor <mjambor@suse.cz>
7079 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7080 of call statements, also set location of a load to a temporary.
7082 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7084 * config/rl78/rl78.c (add_vector_labels): New function.
7085 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7086 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7087 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7088 which checks that no arguments are passed.
7089 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7090 * doc/extend.texi: Documentation for the new attribute.
7092 2018-02-13 Andreas Schwab <schwab@suse.de>
7094 * config/riscv/linux.h (CPP_SPEC): Define.
7096 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7099 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7100 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7101 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7102 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7103 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7104 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7105 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7106 temporarily for AES and PCLMUL builtins.
7108 PR tree-optimization/84339
7109 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7110 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7114 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7115 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7116 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7118 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7120 * omp-simd-clone.h: New file.
7121 * omp-simd-clone.c: Include omp-simd-clone.h.
7122 (expand_simd_clones): No longer static.
7123 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7124 cgraph.h and omp-simd-clone.h.
7125 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7126 (vect_recog_widen_shift_pattern): Formatting fix.
7127 (vect_pattern_recog_1): Don't check optab for calls.
7130 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7131 operands[2] into a REG before using gen_lowpart on it.
7133 2018-02-12 Jeff Law <law@redhat.com>
7136 * config/sh/sh.c (find_barrier): Consider a sibling call
7139 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7140 successfully back substituting a reg.
7142 2018-02-12 Richard Biener <rguenther@suse.de>
7144 PR tree-optimization/84037
7145 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7146 parameter, move visited init to caller.
7147 (vect_slp_analyze_operations): Separate cost from validity
7148 check, initialize visited once for all instances.
7149 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7151 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7153 (vect_model_promotion_demotion_cost): Likewise.
7154 (vectorizable_bswap): Guard cost modeling with !slp_node
7155 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7157 (vectorizable_call): Likewise.
7158 (vectorizable_conversion): Likewise.
7159 (vectorizable_assignment): Likewise.
7160 (vectorizable_shift): Likewise.
7161 (vectorizable_operation): Likewise.
7162 (vectorizable_store): Likewise.
7163 (vectorizable_load): Likewise.
7164 (vectorizable_condition): Likewise.
7165 (vectorizable_comparison): Likewise.
7167 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7170 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7171 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7172 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7174 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7177 * params.def (inline-min-speedup): Increase from 8 to 15.
7178 (max-inline-insns-auto): Decrease from 40 to 30.
7179 * ipa-split.c (consider_split): Add some buffer for function to
7180 be considered inlining candidate.
7181 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7184 2018-02-12 Richard Biener <rguenther@suse.de>
7186 PR tree-optimization/84037
7187 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7188 matched stmts if we cannot swap the non-matched ones.
7190 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7192 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7193 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7194 _mm_maskz_scalef_round_ss): New intrinsics.
7195 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7196 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7197 __builtin_ia32_scalefss_round): Remove.
7198 (__builtin_ia32_scalefsd_mask_round,
7199 __builtin_ia32_scalefss_mask_round): New intrinsics.
7200 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7201 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7202 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7203 "<round_constraint>")): Changed to ...
7204 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7205 "<round_scalar_constraint>")): ... this.
7206 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7207 %0, %1, %2<round_op3>}"): Changed to ...
7208 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7209 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7210 %2<round_scalar_mask_op3>}"): ... this.
7211 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7213 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7215 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7216 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7217 (_mm_maskz_sqrt_round_ss): New intrinsics.
7218 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7219 (__builtin_ia32_sqrtsd_mask_round)
7220 (__builtin_ia32_sqrtss_mask_round): New builtins.
7221 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7222 (__builtin_ia32_sqrtss_round): Remove.
7223 (__builtin_ia32_sqrtsd_mask_round)
7224 (__builtin_ia32_sqrtss_mask_round): New builtins.
7225 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7226 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7227 ((match_operand:VF_128 1 "vector_operand"
7228 "xBm,<round_constraint>")): Changed to ...
7229 ((match_operand:VF_128 1 "vector_operand"
7230 "xBm,<round_scalar_constraint>")): ... this.
7231 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7232 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7233 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7234 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7235 %<iptr>1<round_scalar_mask_op3>}): ... this.
7236 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7237 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7239 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7242 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7243 Cast vec_cmpeq result to correct type.
7244 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7245 Cast vec_cmpgt result to correct type.
7247 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7249 * final.c (final_scan_insn_1): Renamed from...
7250 (final_scan_insn): ... this. New wrapper, to recover
7251 seen from the outermost call in recursive ones.
7252 * config/sparc/sparc.c (output_return): Drop seen from call.
7253 (output_sibcall): Likewise.
7254 * config/visium/visium.c (output_branch): Likewise.
7256 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7258 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7261 2018-02-10 Alan Modra <amodra@gmail.com>
7264 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7265 Specify LR as an input.
7267 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7270 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7271 remove_member_access_dummy_vars): New functions.
7272 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7273 lower_omp_1, execute_lower_omp): Use them.
7275 PR rtl-optimization/84308
7276 * shrink-wrap.c (spread_components): Release todo vector.
7278 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7280 PR rtl-optimization/57193
7281 * ira-color.c (struct allocno_color_data): Add member
7282 conflict_allocno_hard_prefs.
7283 (update_conflict_allocno_hard_prefs): New.
7284 (bucket_allocno_compare_func): Add a preference based on
7285 conflict_allocno_hard_prefs.
7286 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7287 (color_allocnos): Remove a dead code. Initiate
7288 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7290 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7293 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7294 constraint from =wa to wa. Avoid a subreg on the output operand,
7295 instead use a pseudo and subreg it in a move.
7296 (p9_xxbrd_<mode>): Changed to ...
7297 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7298 (p9_xxbrd_v2df): New expander.
7299 (p9_xxbrw_<mode>): Changed to ...
7300 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7301 (p9_xxbrw_v4sf): New expander.
7303 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7305 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7307 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7310 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7311 multiply in 32-bit mode.
7312 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7313 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7316 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7318 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7319 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7320 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7321 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7323 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7326 * dwarf2out.c (is_trivial_indirect_ref): New function.
7327 (dwarf2out_late_global_decl): Do not generate a location
7328 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7329 and that are not defined in the current unit.
7331 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7333 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7334 instead of a libcall for UNORDERED.
7336 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7339 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7340 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7342 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7345 * config/s390/s390.c (s390_set_current_function): Invoke
7346 s390_indirect_branch_settings also if fndecl didn't change.
7348 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7350 * config/rs6000/rs6000.md (blockage): Set length to zero.
7352 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7354 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7356 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7359 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7360 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7364 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7365 PARALLEL incoming that failed vt_get_decl_and_offset check.
7368 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7369 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7371 (get_variable_section): For decls in named .bss* sections pass true as
7372 second argument to bss_initializer_p.
7374 2018-02-09 Marek Polacek <polacek@redhat.com>
7375 Jakub Jelinek <jakub@redhat.com>
7378 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7379 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7380 Sync some changes from cxx_fold_indirect_ref.
7382 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7384 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7386 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7387 (BLOCK_INLINE_ENTRY_LABEL): New.
7388 (dwarf2out_var_location): Disregard inline entry markers.
7389 (inline_entry_data): New struct.
7390 (inline_entry_data_hasher): New hashtable type.
7391 (inline_entry_data_hasher::hash): New.
7392 (inline_entry_data_hasher::equal): New.
7393 (inline_entry_data_table): New variable.
7394 (add_high_low_attributes): Add DW_AT_entry_pc and
7395 DW_AT_GNU_entry_view attributes if a pending entry is found
7396 in inline_entry_data_table. Add old entry_pc attribute only
7397 if debug nonbinding markers are disabled.
7398 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7399 markers are enabled.
7400 (block_within_block_p, dwarf2out_inline_entry): New.
7401 (dwarf2out_finish): Check that no entries remained in
7402 inline_entry_data_table.
7403 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7404 (final_scan_insn, notice_source_line): Likewise.
7405 (rest_of_clean_state): Skip inline entry markers.
7406 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7408 * gimple.c (gimple_build_debug_inline_entry): New.
7409 * gimple.h (enum gimple_debug_subcode): Add
7410 GIMPLE_DEBUG_INLINE_ENTRY.
7411 (gimple_build_debug_inline_entry): Declare.
7412 (gimple_debug_inline_entry_p): New.
7413 (gimple_debug_nonbind_marker_p): Adjust.
7414 * insn-notes.def (INLINE_ENTRY): New.
7415 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7416 inline entry marker notes.
7417 (print_insn): Likewise.
7418 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7419 (INSN_DEBUG_MARKER_KIND): Likewise.
7420 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7421 * tree-inline.c (expand_call_inline): Build and insert
7422 debug_inline_entry stmt.
7423 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7424 inline entry blocks early, if nonbind markers are enabled.
7425 (dump_scope_block): Dump fragment info.
7426 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7427 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7428 (gimple_build_debug_inline_entry): New.
7429 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7430 Enable/disable inline entry points too.
7431 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7432 (DEBUG_INSN): Describe inline entry markers.
7434 * common.opt (gvariable-location-views): New.
7435 (gvariable-location-views=incompat5): New.
7436 * config.in: Rebuilt.
7437 * configure: Rebuilt.
7438 * configure.ac: Test assembler for view support.
7439 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7440 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7441 * dwarf2out.c (var_loc_view): New typedef.
7442 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7443 (dwarf2out_locviews_in_attribute): New.
7444 (dwarf2out_locviews_in_loclist): New.
7445 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7446 (enum dw_line_info_opcode): Add LI_adv_address.
7447 (struct dw_line_info_table): Add view.
7448 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7449 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7450 (zero_view_p): New variable.
7451 (ZERO_VIEW_P): New macro.
7452 (output_asm_line_debug_info): New.
7453 (struct var_loc_node): Add view.
7454 (add_AT_view_list, AT_loc_list): New.
7455 (add_var_loc_to_decl): Add view param. Test it against last.
7456 (new_loc_list): Add view params. Record them.
7457 (AT_loc_list_ptr): Handle loc and view lists.
7458 (view_list_to_loc_list_val_node): New.
7459 (print_dw_val): Handle dw_val_class_view_list.
7460 (size_of_die): Likewise.
7461 (value_format): Likewise.
7462 (loc_list_has_views): New.
7463 (gen_llsym): Set vl_symbol too.
7464 (maybe_gen_llsym, skip_loc_list_entry): New.
7465 (dwarf2out_maybe_output_loclist_view_pair): New.
7466 (output_loc_list): Output view list or entries too.
7467 (output_view_list_offset): New.
7468 (output_die): Handle dw_val_class_view_list.
7469 (output_dwarf_version): New.
7470 (output_compilation_unit_header): Use it.
7471 (output_skeleton_debug_sections): Likewise.
7472 (output_rnglists, output_line_info): Likewise.
7473 (output_pubnames, output_aranges): Update version comments.
7474 (output_one_line_info_table): Output view numbers in asm comments.
7475 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7476 Call maybe_gen_llsym.
7477 (loc_list_from_tree_1): Adjust.
7478 (add_AT_location_description): Create view list attribute if
7479 needed, check it's absent otherwise.
7480 (convert_cfa_to_fb_loc_list): Adjust.
7481 (maybe_emit_file): Call output_asm_line_debug_info for test.
7482 (dwarf2out_var_location): Reset views as needed. Precompute
7483 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7484 attribute. Set view.
7485 (new_line_info_table): Reset next view.
7486 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7487 (dwarf2out_source_line): Likewise. Output view resets and labels to
7488 the assembler, or select appropriate line info opcodes.
7489 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7490 (optimize_string_length): Catch it. Adjust.
7491 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7492 dw_val_class_view_list, and remove it if no longer needed.
7493 (hash_loc_list): Hash view numbers.
7494 (loc_list_hasher::equal): Compare them.
7495 (optimize_location_lists): Check whether a view list symbol is
7496 needed, and whether the locview attribute is present, and
7497 whether they match. Remove the locview attribute if no longer
7499 (index_location_lists): Call skip_loc_list_entry for test.
7500 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7501 Use output_dwarf_version.
7502 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7503 (struct dw_val_node): Add val_view_list.
7504 * final.c (SEEN_NEXT_VIEW): New.
7505 (set_next_view_needed): New.
7506 (clear_next_view_needed): New.
7507 (maybe_output_next_view): New.
7508 (final_start_function): Rename to...
7509 (final_start_function_1): ... this. Take pointer to FIRST,
7510 add SEEN parameter. Emit param bindings in the initial view.
7511 (final_start_function): Reintroduce SEEN-less interface.
7512 (final): Rename to...
7513 (final_1): ... this. Take SEEN parameter. Output final pending
7514 next view at the end.
7515 (final): Reintroduce seen-less interface.
7516 (final_scan_insn): Output pending next view before switching
7517 sections or ending a block. Mark the next view as needed when
7518 outputting variable locations. Notify debug backend of section
7519 changes, and of location view changes.
7520 (rest_of_handle_final): Adjust.
7521 * toplev.c (process_options): Autodetect value for debug variable
7522 location views option. Warn on incompat5 without -gdwarf-5.
7523 * doc/invoke.texi (gvariable-location-views): New.
7524 (gvariable-location-views=incompat5): New.
7525 (gno-variable-location-views): New.
7527 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7529 PR tree-optimization/84136
7530 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7531 that the result of find_edge is non-NULL.
7533 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7536 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7537 storing integer register in SImode. Fix cost of 256 and 512
7538 byte aligned SSE register store.
7540 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7542 * config/i386/i386.c (ix86_multiplication_cost): Fix
7543 multiplication cost for TARGET_AVX512DQ.
7545 2018-02-08 Marek Polacek <polacek@redhat.com>
7547 PR tree-optimization/84238
7548 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7551 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7553 PR tree-optimization/84265
7554 * tree-vect-stmts.c (vectorizable_store): Don't treat
7555 VMAT_CONTIGUOUS accesses as grouped.
7556 (vectorizable_load): Likewise.
7558 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7560 PR tree-optimization/81635
7561 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7562 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7563 (test_round_for_mask): New functions.
7564 (wide_int_cc_tests): Call test_round_for_mask.
7565 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7566 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7567 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7568 range returned by get_range_info.
7570 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7573 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7574 * symtab.c: Include builtins.h
7575 (symtab_node::output_to_lto_symbol_table_p): Move here
7576 from lto-streamer-out.c:output_symbol_p.
7577 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7578 (output_symbol_p): Move all logic to symtab.c
7579 (produce_symtab): Update.
7581 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7583 * config/s390/s390-opts.h (enum indirect_branch): Define.
7584 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7585 (s390_indirect_branch_via_thunk)
7586 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7587 (enum s390_indirect_branch_type): Define.
7588 * config/s390/s390.c (struct s390_frame_layout, struct
7589 machine_function): Remove.
7590 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7591 (indirect_branch_table_label_no, indirect_branch_table_name):
7593 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7594 (enum s390_indirect_branch_option): Define.
7595 (s390_return_addr_from_memory): New function.
7596 (s390_handle_string_attribute): New function.
7597 (s390_attribute_table): Add new attribute handler.
7598 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7599 (s390_indirect_branch_via_thunk): New function.
7600 (s390_indirect_branch_via_inline_thunk): New function.
7601 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7602 sibling call optimization for non z10 compiles.
7603 (s390_emit_call): Force indirect branch target to be a single
7604 register. Add r1 clobber for non-z10 compiles.
7605 (s390_emit_epilogue): Emit return jump via return_use expander.
7606 (s390_reorg): Handle JUMP_INSNs as execute targets.
7607 (s390_option_override_internal): Perform validity checks for the
7608 new command line options.
7609 (s390_indirect_branch_attrvalue): New function.
7610 (s390_indirect_branch_settings): New function.
7611 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7612 (s390_output_indirect_thunk_function): New function.
7613 (s390_code_end): Implement target hook.
7614 (s390_case_values_threshold): Implement target hook.
7615 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7617 * config/s390/s390.h (struct s390_frame_layout)
7618 (struct machine_function): Move here from s390.c.
7619 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7620 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7621 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7622 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7623 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7624 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7625 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7626 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7627 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7628 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7629 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7630 (mnemonic attribute): Add values which aren't recognized
7632 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7633 pattern for branch conversion. Fix mnemonic attribute.
7634 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7635 indirect branch via thunk if requested.
7636 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7637 ("*indirect_jump"): Disable for branch conversion using out of
7639 ("indirect_jump_via_thunk<mode>_z10")
7640 ("indirect_jump_via_thunk<mode>")
7641 ("indirect_jump_via_inlinethunk<mode>_z10")
7642 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7643 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7644 ("casesi_jump_via_inlinethunk<mode>_z10")
7645 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7646 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7647 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7648 ("*indirect2_jump"): Disable for branch conversion.
7649 ("casesi_jump"): Turn into expander and expand patterns for branch
7651 ("return_use"): New expander.
7652 ("*return"): Emit return via thunk and rename it to ...
7653 ("*return<mode>"): ... this one.
7654 * config/s390/s390.opt: Add new options and and enum for the
7657 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7659 * lra-constraints.c (match_reload): Unconditionally use
7660 gen_lowpart_SUBREG, rather than selecting between that
7661 and equivalent gen_rtx_SUBREG code.
7663 2018-02-08 Richard Biener <rguenther@suse.de>
7665 PR tree-optimization/84233
7666 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7667 changed flag instead of boguously re-using phi_inserted.
7669 2018-02-08 Martin Jambor <mjambor@suse.cz>
7671 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7672 static local variables.
7674 2018-02-08 Richard Biener <rguenther@suse.de>
7676 PR tree-optimization/84278
7677 * tree-vect-stmts.c (vectorizable_store): When looking for
7678 smaller vector types to perform grouped strided loads/stores
7679 make sure the mode is supported by the target.
7680 (vectorizable_load): Likewise.
7682 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7684 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7685 Increase LDP/STP opportunities by adding adjacent callee-saves.
7687 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7689 PR rtl-optimization/84068
7690 PR rtl-optimization/83459
7691 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7693 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7695 PR tree-optimization/84224
7696 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7697 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7700 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7703 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7704 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7705 count, remove one USE.
7707 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7709 * doc/install.texi (Configuration): Document the
7710 --with-long-double-format={ibm,ieee} PowerPC configuration
7714 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7715 Convert from define_expand to be define_insn_and_split. Rework
7716 float/double/_Float128 conversions to QI/HI/SImode to work with
7717 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7718 conversions to QI/HImode types did a store and then a load to
7719 truncate the value. For conversions to VSX registers, don't split
7720 the insn, instead emit the code directly. Use the code iterator
7721 any_fix to combine signed and unsigned conversions.
7722 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7723 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7724 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7725 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7726 (fix_<mode>di2_hw): Likewise.
7727 (fixuns_<mode>di2_hw): Likewise.
7728 (fix_<mode>si2_hw): Likewise.
7729 (fixuns_<mode>si2_hw): Likewise.
7730 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7731 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7732 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7733 fix<uns>_trunc<SFDF:mode>si2_p8.
7734 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7736 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7737 (fix<uns>_<mode>_mem): Likewise.
7738 (fctiw<u>z_<mode>_mem): Likewise.
7739 (fix<uns>_<mode>_mem): Likewise.
7740 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7741 the register allocator from doing a direct move to the GPRs to do
7742 a store, and instead use the ISA 3.0 store byte/half-word from
7743 vector register instruction. For IEEE 128-bit floating point,
7744 also optimize stores of 32-bit ints.
7745 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7747 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7749 * genextract.c (push_pathstr_operand): New function to support
7751 (walk_rtx): Call push_pathstr_operand.
7752 (print_path): Support [a-zA-Z].
7754 2018-02-07 Richard Biener <rguenther@suse.de>
7756 PR tree-optimization/84037
7757 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7758 (cse_and_gimplify_to_preheader): Declare.
7759 (vect_get_place_in_interleaving_chain): Likewise.
7760 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7762 (_loop_vec_info::~_loop_vec_info): Delete it.
7763 (cse_and_gimplify_to_preheader): New function.
7764 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7765 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7766 (vectorizable_load): Likewise. For grouped stores always base
7767 the IV on the first element.
7768 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7769 condition before gimplifying.
7771 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7773 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7774 *DIV_EXPR and *MOD_EXPR.
7776 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7779 * config/i386/i386.c (ix86_option_override_internal): Mask out
7780 the CF_SET bit when checking -fcf-protection.
7782 2018-02-07 Tom de Vries <tom@codesourcery.com>
7785 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7788 2018-02-07 Richard Biener <rguenther@suse.de>
7790 PR tree-optimization/84204
7791 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7794 PR tree-optimization/84205
7795 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7796 special-case isl_ast_op_zdiv_r.
7798 PR tree-optimization/84223
7799 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7800 Only add conditions from within the region.
7801 (gather_bbs::after_dom_children): Adjust.
7803 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7806 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7807 * config/avr/avr.md: Only post-reload split REG-REG moves if
7808 either register is GENERAL_REG_P.
7810 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7812 PR tree-optimization/84235
7813 * tree-ssa-scopedtables.c
7814 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7815 if the subtraction is performed in floating point type where NaNs are
7816 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7817 build 1. Formatting fix.
7819 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7822 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7823 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7824 and skip it regardless of bb boundaries. Use CALL_P macro,
7825 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7826 unnecessarily, formatting fix.
7828 2018-02-06 Michael Collison <michael.collison@arm.com>
7830 * config/arm/thumb2.md:
7831 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7832 (*thumb_mov_notscc): Ditto.
7834 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7837 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7838 unsigned_fix, not "s".
7840 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7842 * configure.ac (gcc_fn_eh_frame_ro): New function.
7843 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7844 correct .eh_frame permissions.
7845 * configure: Regenerate.
7847 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7849 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7852 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7854 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7855 Display warning message for -mno-speculate-indirect-jumps.
7857 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7859 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7861 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7863 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7865 PR tree-optimization/84225
7866 * tree-eh.c (find_trapping_overflow): Only call
7867 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7869 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7872 * config/i386/i386.c: Reimplement the check of possible options
7873 -mibt/-mshstk conbination. Change error messages.
7874 * doc/invoke.texi: Fix a typo: remove extra '='.
7876 2018-02-06 Marek Polacek <polacek@redhat.com>
7878 PR tree-optimization/84228
7879 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7881 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7884 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7885 emitted arch directives.
7886 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7887 __ARM_FEATURE_COPROC before changing architectures.
7889 2018-02-06 Richard Biener <rguenther@suse.de>
7891 * config/i386/i386.c (print_reg): Fix typo.
7892 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7894 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7896 * configure: Regenerate.
7898 2018-02-05 Martin Sebor <msebor@redhat.com>
7900 PR tree-optimization/83369
7901 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7904 2018-02-05 Martin Liska <mliska@suse.cz>
7906 * doc/invoke.texi: Cherry-pick upstream r323995.
7908 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7910 * ira.c (ira_init_register_move_cost): Adjust comment.
7912 2018-02-05 Martin Liska <mliska@suse.cz>
7914 PR gcov-profile/84137
7915 * doc/gcov.texi: Fix typo in documentation.
7917 2018-02-05 Martin Liska <mliska@suse.cz>
7919 PR gcov-profile/83879
7920 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7921 using dlopen functionality.
7923 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7925 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7926 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7927 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7928 _mm_maskz_range_round_ss): New intrinsics.
7929 (__builtin_ia32_rangesd128_round)
7930 (__builtin_ia32_rangess128_round): Remove.
7931 (__builtin_ia32_rangesd128_mask_round,
7932 __builtin_ia32_rangess128_mask_round): New builtins.
7933 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7934 __builtin_ia32_rangess128_round): Remove.
7935 (__builtin_ia32_rangesd128_mask_round,
7936 __builtin_ia32_rangess128_mask_round): New builtins.
7937 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7938 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7939 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7940 "<round_saeonly_constraint>")): Changed to ...
7941 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7942 "<round_saeonly_scalar_constraint>")): ... this.
7943 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7944 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7945 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7946 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7947 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7949 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7951 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7953 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7954 Remove all values except native, 8540 and 8548.
7956 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7958 * config/i386/i386.c (ix86_output_function_return): Pass
7959 INVALID_REGNUM, instead of -1, as invalid register number to
7960 indirect_thunk_name and output_indirect_thunk.
7962 2018-02-02 Julia Koval <julia.koval@intel.com>
7964 * config.gcc: Add -march=icelake.
7965 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7966 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7967 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7968 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7969 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7970 (processor_target_table): Add icelake.
7971 (ix86_option_override_internal): Handle new PTAs.
7972 (get_builtin_code_for_version): Handle icelake.
7973 (M_INTEL_COREI7_ICELAKE): New.
7974 (fold_builtin_cpu): Handle icelake.
7975 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7976 * doc/invoke.texi: Add -march=icelake.
7978 2018-02-02 Julia Koval <julia.koval@intel.com>
7980 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7981 to wide_int_bitmask.
7982 * wide-int-bitmask.h: New.
7984 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7987 * config/i386/i386.md: Replace Pmode with word_mode in
7988 builtin_setjmp_setup and builtin_longjmp to support x32.
7990 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7994 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7996 (rs6000_supported_cpu_names): New static variable.
7997 (linux_cpu_translation_table): Likewise.
7998 (elf_platform) <cpu>: Define new static variable and use it.
7999 Translate kernel AT_PLATFORM name to canonical name if needed.
8000 Error if platform name is unknown.
8002 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8005 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8007 2018-02-01 Jeff Law <law@redhat.com>
8010 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8011 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8012 the scratch if RELEASE_VIA_POP is false.
8013 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8014 If we have to save a temporary register, decrement SIZE appropriately.
8015 Pass new arguments to release_scratch_register_on_entry.
8016 (ix86_adjust_stack_and_probe): Likewise.
8017 (ix86_emit_probe_stack_range): Pass new arguments to
8018 release_scratch_register_on_entry.
8020 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8022 PR rtl-optimization/84157
8023 * combine.c (change_zero_ext): Use REG_P predicate in
8024 front of HARD_REGISTER_P predicate.
8026 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8028 * config/avr/avr.c (avr_option_override): Move disabling of
8029 -fdelete-null-pointer-checks to...
8030 * common/config/avr/avr-common.c (avr_option_optimization_table):
8033 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8035 PR tree-optimization/81635
8036 * tree-data-ref.c (split_constant_offset_1): For types that
8037 wrap on overflow, try to use range info to prove that wrapping
8040 2018-02-01 Renlin Li <renlin.li@arm.com>
8043 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8045 (aarch64_register_move_cost): Likewise.
8046 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8048 (REG_CLASS_NAMES): Likewise.
8049 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8050 TAILCALL_ADDR_REGS. Remove IP registers.
8051 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8053 2018-02-01 Richard Biener <rguenther@suse.de>
8055 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8056 for specifying RPO order and allow NULL for that.
8057 * domwalk.c (dom_walker::dom_walker): Likewise.
8058 (dom_walker::walk): Handle NULL RPO order.
8059 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8061 (rewrite_update_dom_walker): Likewise.
8062 (mark_def_dom_walker): Likewise.
8064 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8066 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8067 (aarch64_maybe_expand_sve_subreg_move): Declare.
8068 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8069 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8071 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8072 that are semantically a reverse operation.
8073 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8074 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8075 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8077 (aarch64_can_change_mode_class): For big-endian, forbid changes
8078 between two SVE modes if they have different element sizes.
8080 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8082 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8083 the TImode handling for big-endian targets.
8085 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8087 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8088 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8090 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8091 Remove BSWAP handing for big-endian targets and use the form of
8092 LD1RQ appropariate for the mode.
8094 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8096 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8097 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8100 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8103 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8104 check for operands that need to go through aarch64_sve_reload_be.
8106 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8108 PR tree-optimization/81661
8109 PR tree-optimization/84117
8110 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8111 * tree-eh.c: Include gimplify.h.
8112 (find_trapping_overflow, replace_trapping_overflow,
8113 rewrite_to_non_trapping_overflow): New functions.
8114 * tree-vect-loop.c: Include tree-eh.h.
8115 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8116 * tree-data-ref.c: Include tree-eh.h.
8117 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8119 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8121 PR rtl-optimization/84123
8122 * combine.c (change_zero_ext): Check if hard register satisfies
8123 can_change_dest_mode before calling gen_lowpart_SUBREG.
8125 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8128 * ira.c (ira_init_register_move_cost): Remove assert.
8130 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8132 PR rtl-optimization/84071
8133 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8134 * doc/tm.texi: Regenerate.
8136 2018-01-31 Richard Biener <rguenther@suse.de>
8138 PR tree-optimization/84132
8139 * tree-data-ref.c (analyze_miv_subscript): Properly
8140 check whether evolution_function_is_affine_multivariate_p
8141 before calling gcd_of_steps_may_divide_p.
8143 2018-01-31 Julia Koval <julia.koval@intel.com>
8146 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8147 * config/i386/i386.md (rdpid_rex64) New.
8148 (rdpid): Make 32bit only.
8150 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8153 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8154 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8156 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8159 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8161 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8163 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8165 PR rtl-optimization/84071
8166 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8167 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8169 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8171 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8172 (arc_attribute_table): Add 'aux' attribute.
8173 (arc_in_small_data_p): Consider aux like variables.
8174 (arc_is_aux_reg_p): New function.
8175 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8176 (arc_get_aux_arg): New function.
8177 (prepare_move_operands): Handle aux-register access.
8178 (arc_handle_aux_attribute): New function.
8179 * doc/extend.texi (ARC Variable attributes): Add subsection.
8181 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8183 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8184 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8185 (arc_attribute_table): Add 'uncached' attribute.
8186 (arc_print_operand): Print '.di' flag for uncached memory
8188 (arc_in_small_data_p): Do not consider for small data the uncached
8190 (arc_is_uncached_mem_p): New function.
8191 * config/arc/predicates.md (compact_store_memory_operand): Check
8192 for uncached memory accesses.
8193 (nonvol_nonimm_operand): Likewise.
8194 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8196 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8199 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8200 falign-loops=): Add Optimization flag.
8202 2018-01-30 Jeff Law <law@redhat.com>
8205 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8206 INT_REGISTERS_SAVED. Check it prior to calling
8207 get_scratch_register_on_entry.
8208 (ix86_adjust_stack_and_probe): Similarly.
8209 (ix86_emit_probe_stack_range): Similarly.
8210 (ix86_expand_prologue): Corresponding changes.
8212 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8215 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8216 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8218 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8221 * lra-constraints.c (curr_insn_transform): Process AND in the
8224 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8226 PR rtl-optimization/83986
8227 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8228 dependence against last_pending_memory_flush in addition to
8231 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8233 PR tree-optimization/81611
8234 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8237 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8240 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8243 2018-01-30 Richard Biener <rguenther@suse.de>
8244 Jakub Jelinek <jakub@redhat.com>
8246 PR tree-optimization/84111
8247 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8248 inner loops added during recursion, as they don't have up-to-date
8251 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8254 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8255 (can_inline_edge_by_limits_p): ... here.
8256 (can_early_inline_edge_p, check_callers,
8257 update_caller_keys, update_callee_keys, recursive_inlining,
8258 add_new_edges_to_heap, speculation_useful_p,
8259 inline_small_functions,
8260 inline_small_functions, flatten_function,
8261 inline_to_all_callers_1): Update.
8263 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8265 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8268 2018-01-30 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/83008
8271 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8272 invariant and constant vector uses in stmts when they need
8275 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8278 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8279 * configure: Regenerate.
8281 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8283 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8285 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8286 Use gen_rtx_REG rather than gen_lowpart.
8288 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8290 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8291 rather than 0 when creating partial subregs.
8293 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8295 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8298 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8301 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8302 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8303 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8304 flags. This restores the settings used before the 2017-07-24.
8305 Turning off pre increment/decrement/modify allows IVOPTS to
8306 optimize DF/SF loops where the index is an int.
8308 2018-01-29 Richard Biener <rguenther@suse.de>
8309 Kelvin Nilsen <kelvin@gcc.gnu.org>
8312 * tree-vect-stmts.c (vectorizable_call): Don't call
8313 targetm.vectorize_builtin_md_vectorized_function if callee is
8316 2018-01-22 Carl Love <cel@us.ibm.com>
8318 * doc/extend.tex: Fix typo in second arg in
8319 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8321 2018-01-29 Richard Biener <rguenther@suse.de>
8323 PR tree-optimization/84086
8324 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8325 (flush_ssaname_freelist): When SSA names were released reset
8326 the SCEV hash table.
8328 2018-01-29 Richard Biener <rguenther@suse.de>
8330 PR tree-optimization/84057
8331 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8332 removed paths when removing edges.
8334 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8336 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8337 -mfunction-return=@var{choice}.
8339 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8342 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8344 (layout::print_source_line): Likewise.
8345 (test_get_line_width_without_trailing_whitespace): Add test cases.
8347 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8350 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8353 2018-01-26 Jim Wilson <jimw@sifive.com>
8355 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8357 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8360 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8362 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8363 and CMP + SUB-immediate -> SUBS.
8365 2018-01-26 Martin Sebor <msebor@redhat.com>
8367 PR tree-optimization/83896
8368 * tree-ssa-strlen.c (get_string_len): Rename...
8369 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8370 Avoid assuming length is constant.
8371 (handle_char_store): Use HOST_WIDE_INT for string length.
8373 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8376 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8377 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8379 2018-01-26 Richard Biener <rguenther@suse.de>
8381 PR rtl-optimization/84003
8382 * dse.c (record_store): Only record redundant stores when
8383 the earlier store aliases at least all accesses the later one does.
8385 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8387 PR rtl-optimization/83985
8388 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8389 REG_CFA_RESTORE insns.
8390 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8391 REG_CFA_RESTORE insns here.
8394 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8395 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8397 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8399 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8400 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8401 (arc_init): Likewise.
8402 (arc_override_options): Likewise.
8403 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8405 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8407 * config/arc/arc.h (TARGET_DBNZ): Define.
8408 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8409 properly set the tune attribute.
8410 (dbnz): Use TARGET_DBNZ guard.
8411 * config/arc/arc.opt (mtune): Add core3 option.
8413 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8415 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8416 recognize new pic like addresses.
8417 (arc_delegitimize_address): Clean up.
8419 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8421 * config/arc/arc-arches.def: Option mrf16 valid for all
8423 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8424 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8425 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8426 * config/arc/arc-tables.opt: Regenerate.
8427 * config/arc/arc.c (arc_conditional_register_usage): Handle
8428 reduced register file case.
8429 (arc_file_start): Set must have build attributes.
8430 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8432 * config/arc/arc.opt (mrf16): Add new option.
8433 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8434 * config/arc/genmultilib.awk: Handle new mrf16 option.
8435 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8436 * config/arc/t-multilib: Regenerate.
8437 * doc/invoke.texi (ARC Options): Document mrf16 option.
8439 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8441 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8442 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8443 (arc_attribute_table): Add 'secure_call' attribute.
8444 (arc_print_operand): Print secure call operand.
8445 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8447 (arc_is_secure_call_p): New function. * config/arc/arc.md
8448 (call_i): Add support for sjli instruction.
8449 (call_value_i): Likewise.
8450 * config/arc/constraints.md (Csc): New constraint.
8452 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8453 John Eric Martin <John.Martin@emmicro-us.com>
8455 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8456 * config/arc/arc.c (_arc_jli_section): New struct.
8457 (arc_jli_section): New type.
8458 (rc_jli_sections): New static variable.
8459 (arc_handle_jli_attribute): New function.
8460 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8461 (arc_file_end): New function.
8462 (TARGET_ASM_FILE_END): Define.
8463 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8464 (arc_add_jli_section): New function.
8465 (jli_call_scan): Likewise.
8466 (arc_reorg): Call jli_call_scan.
8467 (arc_output_addsi): Remove 'S' from printing asm operand.
8468 (arc_is_jli_call_p): New function.
8469 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8471 (movhi_insn): Likewise.
8472 (movsi_insn): Likewise.
8473 (movsi_set_cc_insn): Likewise.
8474 (loadqi_update): Likewise.
8475 (load_zeroextendqisi_update): Likewise.
8476 (load_signextendqisi_update): Likewise.
8477 (loadhi_update): Likewise.
8478 (load_zeroextendhisi_update): Likewise.
8479 (load_signextendhisi_update): Likewise.
8480 (loadsi_update): Likewise.
8481 (loadsf_update): Likewise.
8482 (movsicc_insn): Likewise.
8483 (bset_insn): Likewise.
8484 (bxor_insn): Likewise.
8485 (bclr_insn): Likewise.
8486 (bmsk_insn): Likewise.
8487 (bicsi3_insn): Likewise.
8488 (cmpsi_cc_c_insn): Likewise.
8489 (movsi_ne): Likewise.
8490 (movsi_cond_exec): Likewise.
8491 (clrsbsi2): Likewise.
8502 (call_i): Remove 'S' asm letter, add jli instruction.
8503 (call_value_i): Likewise.
8504 * config/arc/arc.op (mjli-always): New option.
8505 * config/arc/constraints.md (Cji): New constraint.
8506 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8508 (subsf3_fpx): Likewise.
8509 (mulsf3_fpx): Likewise.
8510 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8512 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8514 * doc/invoke.texi (ARC): Document mjli-always option.
8516 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8518 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8519 avoid addition with 0 and use incw and decw where possible.
8521 2018-01-26 Richard Biener <rguenther@suse.de>
8523 PR tree-optimization/81082
8524 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8525 association if it requires casting to unsigned.
8526 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8527 from fold_plusminus_mult_expr to catch important cases late when
8528 range info is available.
8530 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8532 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8533 * configure.ac (hidden_linkonce): New test.
8534 * configure: Regenerate.
8535 * config.in: Regenerate.
8537 2018-01-26 Julia Koval <julia.koval@intel.com>
8539 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8540 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8541 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8542 _mm_mask_bitshuffle_epi64_mask): Fix type.
8543 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8544 USI_FTYPE_V4DI_V4DI_USI): Remove.
8545 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8546 __builtin_ia32_vpshufbitqmb256_mask,
8547 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8548 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8549 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8551 2018-01-26 Alan Modra <amodra@gmail.com>
8554 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8555 UNSPEC_VBPERMQ. Sort other unspecs.
8557 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8559 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8561 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8564 * predict.c (drop_profile): Do not push/pop cfun; update also
8566 (handle_missing_profiles): Fix logic looking for zero profiles.
8568 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8571 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8572 on functions with #pragma omp declare simd or functions with simd
8574 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8575 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8576 Remove trailing \n from warning_at calls.
8578 2018-01-25 Tom de Vries <tom@codesourcery.com>
8581 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8582 for neutered workers.
8584 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8587 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8588 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8590 2018-01-24 Jeff Law <law@redhat.com>
8593 * i386.c (get_probe_interval): Move to earlier point.
8594 (ix86_compute_frame_layout): If -fstack-clash-protection and
8595 the frame is larger than the probe interval, then use pushes
8596 to save registers rather than reg->mem moves.
8597 (ix86_expand_prologue): Remove conditional for int_registers_saved
8600 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8603 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8604 min/max for never referenced object.
8606 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8609 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8611 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8612 attributes from DECL_ATTRIBUTES (decl) without affecting
8613 DECL_ATTRIBUTES (current_function_decl).
8614 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8615 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8617 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8619 PR tree-optimization/83979
8620 * fold-const.c (fold_comparison): Use constant_boolean_node
8621 instead of boolean_{true,false}_node.
8623 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8625 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8628 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8630 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8631 Simplify the clause that sets the length attribute.
8632 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8633 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8634 clause that sets the length attribute.
8635 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8637 2018-01-24 Tom de Vries <tom@codesourcery.com>
8640 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8641 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8642 Add strict parameter.
8643 (prevent_branch_around_nothing): Insert dummy insn between branch to
8644 label and label with no ptx insn inbetween.
8645 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8647 2018-01-24 Tom de Vries <tom@codesourcery.com>
8650 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8651 for neutered threads in warp.
8652 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8654 2018-01-24 Richard Biener <rguenther@suse.de>
8656 PR tree-optimization/83176
8657 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8660 2018-01-24 Richard Biener <rguenther@suse.de>
8662 PR tree-optimization/82819
8663 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8664 code generating pluses that are no-ops in the target precision.
8666 2018-01-24 Richard Biener <rguenther@suse.de>
8669 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8671 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8673 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8674 to merge probabilities.
8675 * predict.c (probably_never_executed): Also mark as cold functions
8676 with global 0 profile and guessed local profile.
8677 * profile-count.c (profile_probability::combine_with_count): New
8679 * profile-count.h (profile_probability::operator*,
8680 profile_probability::operator*=, profile_probability::operator/,
8681 profile_probability::operator/=): Reduce precision to adjusted
8682 and set value to guessed on contradictory divisions.
8683 (profile_probability::combine_with_freq): Remove.
8684 (profile_probability::combine_wiht_count): Declare.
8685 (profile_count::force_nonzero):: Set to adjusted.
8686 (profile_count::probability_in):: Set quality to adjusted.
8687 * tree-ssa-tail-merge.c (replace_block_by): Use
8690 2018-01-23 Andrew Waterman <andrew@sifive.com>
8691 Jim Wilson <jimw@sifive.com>
8693 * config/riscv/riscv.c (riscv_stack_boundary): New.
8694 (riscv_option_override): Set riscv_stack_boundary. Handle
8695 riscv_preferred_stack_boundary_arg.
8696 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8697 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8698 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8699 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8700 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8701 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8703 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8706 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8707 of struct ix86_frame.
8708 (ix86_expand_epilogue): Likewise. Add a local variable for
8709 the reg_save_offset field in struct ix86_frame.
8711 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8713 PR tree-optimization/82604
8714 * tree-loop-distribution.c (enum partition_kind): New enum item
8715 PKIND_PARTIAL_MEMSET.
8716 (partition_builtin_p): Support above new enum item.
8717 (generate_code_for_partition): Ditto.
8718 (compute_access_range): Differentiate cases that equality can be
8719 proven at all loops, the innermost loops or no loops.
8720 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8721 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8722 (finalize_partitions, distribute_loop): Don't fuse partition of
8723 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8724 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8727 2018-01-23 Martin Liska <mliska@suse.cz>
8729 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8730 order to ignore the predictor.
8731 (PRED_POLYMORPHIC_CALL): Likewise.
8732 (PRED_RECURSIVE_CALL): Likewise.
8734 2018-01-23 Martin Liska <mliska@suse.cz>
8736 * tree-profile.c (tree_profiling): Print function header to
8737 aware reader which function we are working on.
8738 * value-prof.c (gimple_find_values_to_profile): Do not print
8739 not interesting value histograms.
8741 2018-01-23 Martin Liska <mliska@suse.cz>
8743 * profile-count.h (enum profile_quality): Add
8744 profile_uninitialized as the first value. Do not number values
8745 as they are zero based.
8746 (profile_count::verify): Update sanity check.
8747 (profile_probability::verify): Likewise.
8749 2018-01-23 Nathan Sidwell <nathan@acm.org>
8751 * doc/invoke.texi (ffor-scope): Deprecate.
8753 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8755 PR tree-optimization/83510
8756 * domwalk.c (set_all_edges_as_executable): New function.
8757 (dom_walker::dom_walker): Convert bool param
8758 "skip_unreachable_blocks" to enum reachability. Move setup of
8759 edge flags to set_all_edges_as_executable and only do it when
8760 reachability is REACHABLE_BLOCKS.
8761 * domwalk.h (enum dom_walker::reachability): New enum.
8762 (dom_walker::dom_walker): Convert bool param
8763 "skip_unreachable_blocks" to enum reachability.
8764 (set_all_edges_as_executable): New decl.
8765 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8766 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8768 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8769 but converting true to REACHABLE_BLOCKS.
8770 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8772 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8773 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8774 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8776 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8777 if check_all_array_refs will be called.
8779 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8781 * tree.c (selftest::test_location_wrappers): Add more test
8784 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8786 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8787 (selftest::test_bit_in_range): Likewise.
8789 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8792 * doc/sourcebuild.texi (vect_float): Say that the selector
8793 only describes the situation when -funsafe-math-optimizations is on.
8794 (vect_float_strict): Document.
8796 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8798 PR tree-optimization/83965
8799 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8800 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8801 instead of checking only for a reduction.
8802 (vect_recog_widen_sum_pattern): Likewise.
8804 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8806 * predict.c (probably_never_executed): Only use precise profile info.
8807 (compute_function_frequency): Skip after inlining hack since we now
8808 have quality checking.
8810 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8812 * profile-count.h (profile_probability::very_unlikely,
8813 profile_probability::unlikely, profile_probability::even): Set
8814 precision to guessed.
8816 2018-01-23 Richard Biener <rguenther@suse.de>
8818 PR tree-optimization/83963
8819 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8820 Properly terminate dominator walk when crossing the exit edge not
8821 when visiting its source block.
8823 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8826 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8827 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8829 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8831 PR tree-optimization/83957
8832 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8833 semicolon after for body surrounded by braces.
8835 PR tree-optimization/83081
8836 * profile-count.h (profile_probability::split): New method.
8837 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8838 Use profile_probability::split.
8839 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8840 when splitting a single conditional jump into 2.
8842 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8844 PR tree-optimization/69452
8845 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8848 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8850 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8851 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8852 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8854 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8856 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8857 * config/rl78/rl78.md (movdi): New define_expand.
8858 * config/rl78/rl78.c (rl78_split_movdi): New function.
8860 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8863 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8865 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8866 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8867 128-bit to produce an UNSPEC move to get the double word with the
8868 signbit and then a shift directly to do signbit.
8869 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8870 implementation with a new version that just does either a direct
8871 move or a regular move. Move memory interface to separate insns.
8872 Move insns so they are next to the expander.
8873 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8874 with signbit move. Split big and little endian case.
8875 (signbit<mode>2_dm_mem_le): Likewise.
8876 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8877 (signbit<mode>2_dm2): Likewise.
8879 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8881 * config/rl78/rl78.md (anddi3): New define_expand.
8883 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8885 * config/rl78/rl78.md (umindi3): New define_expand.
8887 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8889 * config/rl78/rl78.md (smindi3): New define_expand.
8891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8893 * config/rl78/rl78.md (smaxdi3): New define_expand.
8895 2018-01-22 Carl Love <cel@us.ibm.com>
8897 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8898 LVX_V1TI): Add macro expansion.
8899 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8900 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8901 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8902 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8903 Change check to determine if the instruction is a byte reversing
8904 entry. Fix typo in comment.
8905 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8906 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8907 Add def_builtin calls for new builtins.
8908 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8909 Add define_insn expansion.
8911 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8913 * config/rl78/rl78.md (umaxdi3): New define_expand.
8915 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8917 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8918 for non-QImode registers.
8920 2018-01-22 Richard Biener <rguenther@suse.de>
8922 PR tree-optimization/83963
8923 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8924 including the loop exit block.
8925 (scop_detection::merge_sese): Likewise.
8926 (scop_detection::add_scop): Do it here instead.
8928 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8930 * doc/sourcebuild.texi (arm_softfloat): Document.
8932 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8935 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8936 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8937 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8939 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8940 David Edelsohn <dje.gcc@gmail.com>
8943 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8944 Change "crset eq" to "crset 2".
8945 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8946 (*call_indirect_aix<mode>_nospec): Likewise.
8947 (*call_value_indirect_aix<mode>_nospec): Likewise.
8948 (*call_indirect_elfv2<mode>_nospec): Likewise.
8949 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8950 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8951 change assembly output from . to $.
8952 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8953 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8954 (*tablejump<mode>_internal1_nospec): Likewise.
8956 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8959 * config/sh/sh_optimize_sett_clrt.cc:
8960 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8962 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8964 PR tree-optimization/83940
8965 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8966 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8967 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8968 use it to pass back the definition type.
8969 (vect_check_store_rhs): Likewise rhs_dt_out.
8970 (vect_build_gather_load_calls): Add a mask_dt argument and use
8971 it instead of a call to vect_is_simple_use.
8972 (vectorizable_store): Update calls to vect_check_load_store_mask
8973 and vect_check_store_rhs. Use the dt returned by the latter instead
8974 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8975 instead of calls to vect_is_simple_use. Pass the scalar rather
8976 than the vector operand to vect_is_simple_use when handling
8977 second and subsequent copies of an rhs value.
8978 (vectorizable_load): Update calls to vect_check_load_store_mask
8979 and vect_build_gather_load_calls. Use the cached mask_dt and
8980 gs_info.offset_dt instead of calls to vect_is_simple_use.
8982 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8985 * tree-emutls.c: Include gimplify.h.
8986 (lower_emutls_2): New function.
8987 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8988 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8989 it before further processing.
8992 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8993 UINTVAL (trueop1) instead of INTVAL (op1).
8995 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8999 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9000 INCOMING_FRAME_SP_OFFSET if not defined.
9001 (scan_trace): Add ENTRY argument. If true and
9002 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9003 emit a note to adjust the CFA offset.
9004 (create_cfi_notes): Adjust scan_trace callers.
9005 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9006 INCOMING_FRAME_SP_OFFSET in the CIE.
9007 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9008 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9010 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9011 * doc/tm.texi: Regenerated.
9013 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9015 PR rtl-optimization/83147
9016 * lra-constraints.c (remove_inheritance_pseudos): Use
9017 lra_substitute_pseudo_within_insn.
9019 2018-01-19 Tom de Vries <tom@codesourcery.com>
9020 Cesar Philippidis <cesar@codesourcery.com>
9023 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9025 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9028 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9029 spaces for function labels.
9031 2018-01-19 Martin Liska <mliska@suse.cz>
9033 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9034 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9035 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9036 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9037 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9038 (PRED_CONST_RETURN): Change from 69 to 65.
9039 (PRED_NULL_RETURN): Change from 91 to 71.
9040 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9041 (PRED_LOOP_GUARD): Change from 66 to 73.
9043 2018-01-19 Martin Liska <mliska@suse.cz>
9045 * predict.c (predict_insn_def): Add new assert.
9046 (struct branch_predictor): Change type to signed integer.
9047 (test_prediction_value_range): Amend test to cover
9049 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9050 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9051 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9052 (PRED_LOOP_IV_COMPARE): Likewise.
9053 * predict.h (PROB_UNINITIALIZED): Define new constant.
9055 2018-01-19 Martin Liska <mliska@suse.cz>
9057 * predict.c (dump_prediction): Add new format for
9058 analyze_brprob.py script which is enabled with -details
9060 * profile-count.h (precise_p): New function.
9062 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9064 PR tree-optimization/83922
9065 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9066 there are no statements that need masking.
9067 (vect_active_double_reduction_p): New function.
9068 (vect_analyze_loop_operations): Use it when handling phis that
9069 are not in the loop header.
9071 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9073 PR tree-optimization/83914
9074 * tree-vect-loop.c (vectorizable_induction): Don't convert
9075 init_expr or apply the peeling adjustment for inductions
9076 that are nested within the vectorized loop.
9078 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9080 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9083 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9087 * function.h (gimplify_parameters): Add gimple_seq * argument.
9088 * function.c: Include gimple.h and options.h.
9089 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9090 for the added local temporaries if needed.
9091 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9092 if there are any parameter cleanups, wrap whole body into a
9093 try/finally with the cleanups.
9095 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9098 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9099 Use GET_MODE_CLASS for scalar floating point.
9101 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9105 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9106 Fix call of call_cgraph_insertion_hooks.
9108 2018-01-18 Martin Sebor <msebor@redhat.com>
9110 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9112 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9115 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9118 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9121 * common.opt: (-ffile-prefix-map): New option.
9122 * opts.c (common_handle_option): Defer it.
9123 * opts-global.c (handle_common_deferred_options): Handle it.
9124 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9125 * file-prefix-map.h: New file.
9126 (remap_debug_filename, add_debug_prefix_map): ...here.
9127 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9128 * final.c (debug_prefix_map, add_debug_prefix_map
9129 remap_debug_filename): Move to...
9130 * file-prefix-map.c: New file.
9131 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9132 generalize, get rid of alloca(), use strrchr() instead of strchr().
9133 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9134 Implement in terms of add_prefix_map().
9135 (remap_macro_filename, remap_debug_filename): Implement in term of
9137 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9138 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9139 * dbxout.c: Include file-prefix-map.h.
9140 * varasm.c: Likewise.
9141 * vmsdbgout.c: Likewise.
9142 * xcoffout.c: Likewise.
9143 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9144 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9145 * doc/invoke.texi (-ffile-prefix-map): Document.
9146 (-fdebug-prefix-map): Update description.
9148 2018-01-18 Martin Liska <mliska@suse.cz>
9150 * config/i386/i386.c (indirect_thunk_name): Document that also
9152 (output_indirect_thunk): Document why both instructions
9153 (pause and lfence) are generated.
9155 2018-01-18 Richard Biener <rguenther@suse.de>
9157 PR tree-optimization/83887
9158 * graphite-scop-detection.c
9159 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9160 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9161 (scop_detection::merge_sese): Re-implement with a flood-fill
9162 algorithm that properly finds a SESE region if it exists.
9164 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9167 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9168 pointer_diff optimizations use view_convert instead of convert.
9170 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9172 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9173 Generate different code for -mno-speculate-indirect-jumps.
9174 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9175 (*call_indirect_aix<mode>): Disable for
9176 -mno-speculate-indirect-jumps.
9177 (*call_indirect_aix<mode>_nospec): New define_insn.
9178 (*call_value_indirect_aix<mode>): Disable for
9179 -mno-speculate-indirect-jumps.
9180 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9181 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9182 -mno-speculate-indirect-jumps.
9183 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9185 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9187 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9188 long double type, set the flags for noting the default long double
9189 type, even if we don't pass or return a long double type.
9191 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9194 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9197 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9199 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9200 support for merge[hl].
9201 (fold_mergehl_helper): New helper function.
9202 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9203 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9204 (altivec_vmrglw_direct): Add xxmrglw insn.
9206 2018-01-17 Andrew Waterman <andrew@sifive.com>
9208 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9209 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9211 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9214 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9215 call the lto_location_cache before reading the
9216 DECL_SOURCE_LOCATION of the types.
9218 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9219 Richard Sandiford <richard.sandiford@linaro.org>
9221 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9222 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9223 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9224 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9225 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9227 * config/aarch64/constraints.md (aarch64_movti_operand):
9229 * config/aarch64/predicates.md (Uti): Add new constraint.
9231 2018-01-17 Carl Love <cel@us.ibm.com>
9233 * config/rs6000/vsx.md (define_expand xl_len_r,
9234 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9235 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9237 (define_expand, define_insn): Move the shift left from the
9238 define_insn to the define_expand for lxvl and stxvl instructions.
9239 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9240 and XL_LEN_R definitions to PURE.
9242 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9244 * config/i386/i386.c (indirect_thunk_name): Declare regno
9245 as unsigned int. Compare regno with INVALID_REGNUM.
9246 (output_indirect_thunk): Ditto.
9247 (output_indirect_thunk_function): Ditto.
9248 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9249 in the call to output_indirect_thunk_function.
9251 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9254 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9255 rather than the size of inner_type to determine the stack slot size
9256 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9258 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9261 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9264 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9266 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9267 endian Linux systems to optionally enable multilibs for selecting
9268 the long double type if the user configured an explicit type.
9269 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9270 have no long double multilibs if not defined.
9271 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9272 warn if the user used -mabi={ieee,ibm}longdouble and we built
9273 multilibs for long double.
9274 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9275 appropriate multilib option.
9276 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9278 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9279 for building long double multilibs.
9280 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9282 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9284 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9287 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9289 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9292 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9295 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9298 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9300 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9301 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9303 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9305 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9306 different rtl trees depending on TARGET_64BIT.
9307 (rs6000_gen_lvx): Likewise.
9309 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9311 * config/visium/visium.md (nop): Tweak comment.
9312 (hazard_nop): Likewise.
9314 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9316 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9317 -mspeculate-indirect-jumps.
9318 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9319 for -mno-speculate-indirect-jumps.
9320 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9321 (*call_value_indirect_elfv2<mode>): Disable for
9322 -mno-speculate-indirect-jumps.
9323 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9324 (indirect_jump): Emit different RTL for
9325 -mno-speculate-indirect-jumps.
9326 (*indirect_jump<mode>): Disable for
9327 -mno-speculate-indirect-jumps.
9328 (*indirect_jump<mode>_nospec): New define_insn.
9329 (tablejump): Emit different RTL for
9330 -mno-speculate-indirect-jumps.
9331 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9332 (tablejumpsi_nospec): New define_expand.
9333 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9334 (tablejumpdi_nospec): New define_expand.
9335 (*tablejump<mode>_internal1): Disable for
9336 -mno-speculate-indirect-jumps.
9337 (*tablejump<mode>_internal1_nospec): New define_insn.
9338 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9341 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9343 * caller-save.c (insert_save): Drop unnecessary parameter. All
9346 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9347 Richard Biener <rguenth@suse.de>
9350 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9351 return early, inline manually is_gimple_sizepos. Make sure if we
9352 call gimplify_expr we don't end up with a gimple constant.
9353 * tree.c (variably_modified_type_p): Don't return true for
9354 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9355 * gimplify.h (is_gimple_sizepos): Remove.
9357 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9359 PR tree-optimization/83857
9360 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9361 vectorizable_live_operation for pure SLP statements.
9362 (vectorizable_live_operation): Handle PHIs.
9364 2018-01-16 Richard Biener <rguenther@suse.de>
9366 PR tree-optimization/83867
9367 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9368 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9370 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9373 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9374 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9375 If off is not INTEGER_CST, issue a may not be aligned warning
9376 rather than isn't aligned. Use isn%'t rather than isn't.
9377 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9379 <case MULT_EXPR>: Improve the case when bottom and one of the
9380 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9381 operand, in that case check if the other operand is multiple of
9382 bottom divided by the INTEGER_CST operand.
9384 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9387 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9388 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9389 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9390 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9391 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9392 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9393 (pa_function_arg_size): New function.
9395 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9397 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9398 in a separate statement.
9400 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9402 PR tree-optimization/83847
9403 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9404 group gathers and scatters.
9406 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9408 PR rtl-optimization/86620
9409 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9411 PR rtl-optimization/83213
9412 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9413 to last if both are JUMP_INSNs.
9415 PR tree-optimization/83843
9416 * gimple-ssa-store-merging.c
9417 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9418 store_immediate_info for bswap/nop orig_stores.
9420 2018-01-15 Andrew Waterman <andrew@sifive.com>
9422 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9424 <UDIV>: Increase cost if !TARGET_DIV.
9426 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9428 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9429 (define_attr "cr_logical_3op"): New.
9430 (cceq_ior_compare): Adjust.
9431 (cceq_ior_compare_complement): Adjust.
9432 (*cceq_rev_compare): Adjust.
9433 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9434 (is_cracked_insn): Adjust.
9435 (insn_must_be_first_in_group): Adjust.
9436 * config/rs6000/40x.md: Adjust.
9437 * config/rs6000/440.md: Adjust.
9438 * config/rs6000/476.md: Adjust.
9439 * config/rs6000/601.md: Adjust.
9440 * config/rs6000/603.md: Adjust.
9441 * config/rs6000/6xx.md: Adjust.
9442 * config/rs6000/7450.md: Adjust.
9443 * config/rs6000/7xx.md: Adjust.
9444 * config/rs6000/8540.md: Adjust.
9445 * config/rs6000/cell.md: Adjust.
9446 * config/rs6000/e300c2c3.md: Adjust.
9447 * config/rs6000/e500mc.md: Adjust.
9448 * config/rs6000/e500mc64.md: Adjust.
9449 * config/rs6000/e5500.md: Adjust.
9450 * config/rs6000/e6500.md: Adjust.
9451 * config/rs6000/mpc.md: Adjust.
9452 * config/rs6000/power4.md: Adjust.
9453 * config/rs6000/power5.md: Adjust.
9454 * config/rs6000/power6.md: Adjust.
9455 * config/rs6000/power7.md: Adjust.
9456 * config/rs6000/power8.md: Adjust.
9457 * config/rs6000/power9.md: Adjust.
9458 * config/rs6000/rs64.md: Adjust.
9459 * config/rs6000/titan.md: Adjust.
9461 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9463 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9464 ix86_indirect_branch_register logic.
9466 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9468 * config/i386/constraints.md (Bs): Update
9469 ix86_indirect_branch_register check. Don't check
9470 ix86_indirect_branch_register with GOT_memory_operand.
9472 * config/i386/predicates.md (GOT_memory_operand): Don't check
9473 ix86_indirect_branch_register here.
9474 (GOT32_symbol_operand): Likewise.
9476 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9478 * config/i386/predicates.md (constant_call_address_operand):
9479 Rewrite ix86_indirect_branch_register logic.
9480 (sibcall_insn_operand): Likewise.
9482 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9484 * config/i386/constraints.md (Bs): Replace
9485 ix86_indirect_branch_thunk_register with
9486 ix86_indirect_branch_register.
9488 * config/i386/i386.md (indirect_jump): Likewise.
9489 (tablejump): Likewise.
9490 (*sibcall_memory): Likewise.
9491 (*sibcall_value_memory): Likewise.
9492 Peepholes of indirect call and jump via memory: Likewise.
9493 * config/i386/i386.opt: Likewise.
9494 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9495 (GOT_memory_operand): Likewise.
9496 (call_insn_operand): Likewise.
9497 (sibcall_insn_operand): Likewise.
9498 (GOT32_symbol_operand): Likewise.
9500 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9503 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9504 type rather than type addr's type points to.
9505 (expand_omp_atomic_mutex): Likewise.
9506 (expand_omp_atomic): Likewise.
9508 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9511 * config/i386/i386.c (output_indirect_thunk_function): Use
9512 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9513 for __x86_return_thunk.
9515 2018-01-15 Richard Biener <rguenther@suse.de>
9518 * expmed.c (extract_bit_field_1): Fix typo.
9520 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9523 * config/arm/iterators.md (VF): New mode iterator.
9524 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9525 Remove integer-related logic from pattern.
9526 (neon_vabd<mode>_3): Likewise.
9528 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9531 * common.opt (fstrict-overflow): No longer an alias.
9532 (fwrapv-pointer): New option.
9533 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9534 also for pointer types based on flag_wrapv_pointer.
9535 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9536 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9537 opts->x_flag_wrapv got set.
9538 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9539 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9540 POINTER_TYPE_OVERFLOW_UNDEFINED.
9541 * match.pd: Likewise in address comparison pattern.
9542 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9544 2018-01-15 Richard Biener <rguenther@suse.de>
9547 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9548 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9549 Reset type names to their identifier if their TYPE_DECL doesn't
9550 have linkage (and thus is used for ODR and devirt).
9551 (save_debug_info_for_decl): Remove.
9552 (save_debug_info_for_type): Likewise.
9553 (add_tree_to_fld_list): Adjust.
9554 * tree-pretty-print.c (dump_generic_node): Make dumping of
9555 type names more robust.
9557 2018-01-15 Richard Biener <rguenther@suse.de>
9559 * BASE-VER: Bump to 8.0.1.
9561 2018-01-14 Martin Sebor <msebor@redhat.com>
9564 * builtins.c (check_access): Avoid warning when the no-warning bit
9567 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9569 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9570 * ira-color (allocno_hard_regs_compare): Likewise.
9572 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9575 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9576 Use .pushsection/.popsection.
9578 2018-01-14 Martin Sebor <msebor@redhat.com>
9581 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9583 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9585 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9586 entry from extra_headers.
9587 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9588 extra_headers, make the list bitwise identical to the i?86-*-* one.
9590 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9592 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9593 -mcmodel=large with -mindirect-branch=thunk,
9594 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9595 -mfunction-return=thunk-extern.
9596 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9597 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9598 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9600 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9602 * config/i386/i386.c (print_reg): Print the name of the full
9603 integer register without '%'.
9604 (ix86_print_operand): Handle 'V'.
9605 * doc/extend.texi: Document 'V' modifier.
9607 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9609 * config/i386/constraints.md (Bs): Disallow memory operand for
9610 -mindirect-branch-register.
9612 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9613 (GOT_memory_operand): Likewise.
9614 (call_insn_operand): Likewise.
9615 (sibcall_insn_operand): Likewise.
9616 (GOT32_symbol_operand): Likewise.
9617 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9618 for -mindirect-branch-register.
9619 (tablejump): Likewise.
9620 (*sibcall_memory): Likewise.
9621 (*sibcall_value_memory): Likewise.
9622 Disallow peepholes of indirect call and jump via memory for
9623 -mindirect-branch-register.
9624 (*call_pop): Replace m with Bw.
9625 (*call_value_pop): Likewise.
9626 (*sibcall_pop_memory): Replace m with Bs.
9627 * config/i386/i386.opt (mindirect-branch-register): New option.
9628 * doc/invoke.texi: Document -mindirect-branch-register option.
9630 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9632 * config/i386/i386-protos.h (ix86_output_function_return): New.
9633 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9634 set function_return_type.
9635 (indirect_thunk_name): Add ret_p to indicate thunk for function
9637 (output_indirect_thunk_function): Pass false to
9638 indirect_thunk_name.
9639 (ix86_output_indirect_branch_via_reg): Likewise.
9640 (ix86_output_indirect_branch_via_push): Likewise.
9641 (output_indirect_thunk_function): Create alias for function
9642 return thunk if regno < 0.
9643 (ix86_output_function_return): New function.
9644 (ix86_handle_fndecl_attribute): Handle function_return.
9645 (ix86_attribute_table): Add function_return.
9646 * config/i386/i386.h (machine_function): Add
9647 function_return_type.
9648 * config/i386/i386.md (simple_return_internal): Use
9649 ix86_output_function_return.
9650 (simple_return_internal_long): Likewise.
9651 * config/i386/i386.opt (mfunction-return=): New option.
9652 (indirect_branch): Mention -mfunction-return=.
9653 * doc/extend.texi: Document function_return function attribute.
9654 * doc/invoke.texi: Document -mfunction-return= option.
9656 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9658 * config/i386/i386-opts.h (indirect_branch): New.
9659 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9660 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9661 with local indirect jump when converting indirect call and jump.
9662 (ix86_set_indirect_branch_type): New.
9663 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9664 (indirectlabelno): New.
9665 (indirect_thunk_needed): Likewise.
9666 (indirect_thunk_bnd_needed): Likewise.
9667 (indirect_thunks_used): Likewise.
9668 (indirect_thunks_bnd_used): Likewise.
9669 (INDIRECT_LABEL): Likewise.
9670 (indirect_thunk_name): Likewise.
9671 (output_indirect_thunk): Likewise.
9672 (output_indirect_thunk_function): Likewise.
9673 (ix86_output_indirect_branch_via_reg): Likewise.
9674 (ix86_output_indirect_branch_via_push): Likewise.
9675 (ix86_output_indirect_branch): Likewise.
9676 (ix86_output_indirect_jmp): Likewise.
9677 (ix86_code_end): Call output_indirect_thunk_function if needed.
9678 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9680 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9681 (ix86_attribute_table): Add indirect_branch.
9682 * config/i386/i386.h (machine_function): Add indirect_branch_type
9683 and has_local_indirect_jump.
9684 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9686 (tablejump): Likewise.
9687 (*indirect_jump): Use ix86_output_indirect_jmp.
9688 (*tablejump_1): Likewise.
9689 (simple_return_indirect_internal): Likewise.
9690 * config/i386/i386.opt (mindirect-branch=): New option.
9691 (indirect_branch): New.
9694 (thunk-inline): Likewise.
9695 (thunk-extern): Likewise.
9696 * doc/extend.texi: Document indirect_branch function attribute.
9697 * doc/invoke.texi: Document -mindirect-branch= option.
9699 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9702 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9704 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9706 * ipa-inline.c (want_inline_small_function_p): Return false if
9707 inlining has already failed with CIF_FINAL_ERROR.
9708 (update_caller_keys): Call want_inline_small_function_p before
9710 (update_callee_keys): Likewise.
9712 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9714 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9716 (rs6000_quadword_masked_address_p): Likewise.
9717 (quad_aligned_load_p): Likewise.
9718 (quad_aligned_store_p): Likewise.
9719 (const_load_sequence_p): Add comment to describe the outer-most loop.
9720 (mimic_memory_attributes_and_flags): New function.
9721 (rs6000_gen_stvx): Likewise.
9722 (replace_swapped_aligned_store): Likewise.
9723 (rs6000_gen_lvx): Likewise.
9724 (replace_swapped_aligned_load): Likewise.
9725 (replace_swapped_load_constant): Capitalize argument name in
9726 comment describing this function.
9727 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9728 and stores that access quad-word aligned addresses and replace
9729 with stvx or lvx instructions when appropriate.
9730 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9731 New function prototype.
9732 (rs6000_quadword_masked_address_p): Likewise.
9733 (rs6000_gen_lvx): Likewise.
9734 (rs6000_gen_stvx): Likewise.
9735 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9736 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9737 when memory address is aligned.
9738 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9739 this split to select lvx instruction when memory address is aligned.
9740 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9741 instruction when memory address is aligned.
9742 (*vsx_le_perm_load_v16qi): Likewise.
9743 (four unnamed splitters): Modify to select the stvx instruction
9744 when memory is aligned.
9746 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9748 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9749 which appears in the queue multiple times.
9751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9752 Alan Hayward <alan.hayward@arm.com>
9753 David Sherwood <david.sherwood@arm.com>
9755 * tree-vectorizer.h (vec_lower_bound): New structure.
9756 (_loop_vec_info): Add check_nonzero and lower_bounds.
9757 (LOOP_VINFO_CHECK_NONZERO): New macro.
9758 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9759 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9760 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9761 fields. Make seg_len the distance travelled, not including the
9763 (dr_direction_indicator): Declare.
9764 (dr_zero_step_indicator): Likewise.
9765 (dr_known_forward_stride_p): Likewise.
9766 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9768 (runtime_alias_check_p): Allow runtime alias checks with
9770 (operator ==): Compare access_size and align.
9771 (prune_runtime_alias_test_list): Rework for new distinction between
9772 the access_size and seg_len.
9773 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9775 (get_segment_min_max): New function.
9776 (create_intersect_range_checks): Use it.
9777 (dr_step_indicator): New function.
9778 (dr_direction_indicator): Likewise.
9779 (dr_zero_step_indicator): Likewise.
9780 (dr_known_forward_stride_p): Likewise.
9781 * tree-loop-distribution.c (data_ref_segment_size): Return
9782 DR_STEP * (niters - 1).
9783 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9785 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9786 (vect_preserves_scalar_order_p): New function, split out from...
9787 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9788 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9789 (vect_vfa_access_size): New function.
9790 (vect_vfa_align): Likewise.
9791 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9792 (dump_lower_bound): New function.
9793 (vect_check_lower_bound): Likewise.
9794 (vect_small_gap_p): Likewise.
9795 (vectorizable_with_step_bound_p): Likewise.
9796 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9797 depencies if the vectorization factor is 1. Convert the checks
9798 for nonzero steps into checks on the bounds of DR_STEP. Try using
9799 a bunds check for variable steps if the minimum required step is
9800 relatively small. Update calls to the dr_with_seg_len
9801 constructor and to vect_compile_time_alias.
9802 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9804 (vect_loop_versioning): Call it.
9805 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9807 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9810 Alan Hayward <alan.hayward@arm.com>
9811 David Sherwood <david.sherwood@arm.com>
9813 * doc/sourcebuild.texi (vect_scatter_store): Document.
9814 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9816 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9818 * genopinit.c (main): Add supports_vec_scatter_store and
9819 supports_vec_scatter_store_cached to target_optabs.
9820 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9821 IFN_MASK_SCATTER_STORE.
9822 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9824 * internal-fn.h (internal_store_fn_p): Declare.
9825 (internal_fn_stored_value_index): Likewise.
9826 * internal-fn.c (scatter_store_direct): New macro.
9827 (expand_scatter_store_optab_fn): New function.
9828 (direct_scatter_store_optab_supported_p): New macro.
9829 (internal_store_fn_p): New function.
9830 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9831 IFN_MASK_SCATTER_STORE.
9832 (internal_fn_mask_index): Likewise.
9833 (internal_fn_stored_value_index): New function.
9834 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9836 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9837 * optabs-query.c (supports_vec_scatter_store_p): New function.
9838 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9839 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9840 true for scatter stores.
9841 (vect_gather_scatter_fn_p): Handle scatter stores too.
9842 (vect_check_gather_scatter): Consider using scatter stores if
9843 supports_vec_scatter_store_p.
9844 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9846 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9847 internal_fn_stored_value_index.
9848 (check_load_store_masking): Handle scatter stores too.
9849 (vect_get_store_rhs): Make public.
9850 (vectorizable_call): Use internal_store_fn_p.
9851 (vectorizable_store): Handle scatter store internal functions.
9852 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9853 when deciding whether the end of the group has been reached.
9854 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9855 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9856 (mask_scatter_store<mode>): New insns.
9858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9859 Alan Hayward <alan.hayward@arm.com>
9860 David Sherwood <david.sherwood@arm.com>
9862 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9863 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9864 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9866 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9867 Use vect_truncate_gather_scatter_offset if we can't treat the
9868 operation as a normal gather load or scatter store.
9869 (get_group_load_store_type): Take the gather_scatter_info
9870 as argument. Try using a gather load or scatter store for
9871 single-element groups.
9872 (get_load_store_type): Update calls to get_group_load_store_type
9873 and vect_use_strided_gather_scatters_p.
9875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9876 Alan Hayward <alan.hayward@arm.com>
9877 David Sherwood <david.sherwood@arm.com>
9879 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9880 optional tree argument.
9881 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9883 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9884 but continue to use the current value as a fallback.
9885 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9886 to compare the updates.
9887 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9888 (get_load_store_type): Use it when handling a strided access.
9889 (vect_get_strided_load_store_ops): New function.
9890 (vect_get_data_ptr_increment): Likewise.
9891 (vectorizable_load): Handle strided gather loads. Always pass
9892 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9895 Alan Hayward <alan.hayward@arm.com>
9896 David Sherwood <david.sherwood@arm.com>
9898 * doc/md.texi (gather_load@var{m}): Document.
9899 (mask_gather_load@var{m}): Likewise.
9900 * genopinit.c (main): Add supports_vec_gather_load and
9901 supports_vec_gather_load_cached to target_optabs.
9902 * optabs-tree.c (init_tree_optimization_optabs): Use
9903 ggc_cleared_alloc to allocate target_optabs.
9904 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9905 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9907 * internal-fn.h (internal_load_fn_p): Declare.
9908 (internal_gather_scatter_fn_p): Likewise.
9909 (internal_fn_mask_index): Likewise.
9910 (internal_gather_scatter_fn_supported_p): Likewise.
9911 * internal-fn.c (gather_load_direct): New macro.
9912 (expand_gather_load_optab_fn): New function.
9913 (direct_gather_load_optab_supported_p): New macro.
9914 (direct_internal_fn_optab): New function.
9915 (internal_load_fn_p): Likewise.
9916 (internal_gather_scatter_fn_p): Likewise.
9917 (internal_fn_mask_index): Likewise.
9918 (internal_gather_scatter_fn_supported_p): Likewise.
9919 * optabs-query.c (supports_at_least_one_mode_p): New function.
9920 (supports_vec_gather_load_p): Likewise.
9921 * optabs-query.h (supports_vec_gather_load_p): Declare.
9922 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9923 and memory_type field.
9924 (NUM_PATTERNS): Bump to 15.
9925 * tree-vect-data-refs.c: Include internal-fn.h.
9926 (vect_gather_scatter_fn_p): New function.
9927 (vect_describe_gather_scatter_call): Likewise.
9928 (vect_check_gather_scatter): Try using internal functions for
9929 gather loads. Recognize existing calls to a gather load function.
9930 (vect_analyze_data_refs): Consider using gather loads if
9931 supports_vec_gather_load_p.
9932 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9933 (vect_get_gather_scatter_offset_type): Likewise.
9934 (vect_convert_mask_for_vectype): Likewise.
9935 (vect_add_conversion_to_patterm): Likewise.
9936 (vect_try_gather_scatter_pattern): Likewise.
9937 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9938 (vect_vect_recog_func_ptrs): Add it.
9939 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9940 internal_fn_mask_index and internal_gather_scatter_fn_p.
9941 (check_load_store_masking): Take the gather_scatter_info as an
9942 argument and handle gather loads.
9943 (vect_get_gather_scatter_ops): New function.
9944 (vectorizable_call): Check internal_load_fn_p.
9945 (vectorizable_load): Likewise. Handle gather load internal
9947 (vectorizable_store): Update call to check_load_store_masking.
9948 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9949 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9950 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9951 (aarch64_gather_scale_operand_d): New predicates.
9952 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9953 (mask_gather_load<mode>): New insns.
9955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9956 Alan Hayward <alan.hayward@arm.com>
9957 David Sherwood <david.sherwood@arm.com>
9959 * optabs.def (fold_left_plus_optab): New optab.
9960 * doc/md.texi (fold_left_plus_@var{m}): Document.
9961 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9962 * internal-fn.c (fold_left_direct): Define.
9963 (expand_fold_left_optab_fn): Likewise.
9964 (direct_fold_left_optab_supported_p): Likewise.
9965 * fold-const-call.c (fold_const_fold_left): New function.
9966 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9967 * tree-parloops.c (valid_reduction_p): New function.
9968 (gather_scalar_reductions): Use it.
9969 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9970 (vect_finish_replace_stmt): Declare.
9971 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9972 (needs_fold_left_reduction_p): New function, split out from...
9973 (vect_is_simple_reduction): ...here. Accept reductions that
9974 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9975 (vect_force_simple_reduction): Also store the reduction type in
9976 the assignment's STMT_VINFO_REDUC_TYPE.
9977 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9978 (merge_with_identity): New function.
9979 (vect_expand_fold_left): Likewise.
9980 (vectorize_fold_left_reduction): Likewise.
9981 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9982 scalar phi in place for it. Check for target support and reject
9983 cases that would reassociate the operation. Defer the transform
9984 phase to vectorize_fold_left_reduction.
9985 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9986 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9987 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9991 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9992 call to ifc_temp_var.
9994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9995 Alan Hayward <alan.hayward@arm.com>
9996 David Sherwood <david.sherwood@arm.com>
9998 * target.def (legitimize_address_displacement): Take the original
9999 offset as a poly_int.
10000 * targhooks.h (default_legitimize_address_displacement): Update
10002 * targhooks.c (default_legitimize_address_displacement): Likewise.
10003 * doc/tm.texi: Regenerate.
10004 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10005 as an argument, moving assert of ad->disp == ad->disp_term to...
10006 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10007 Try calling targetm.legitimize_address_displacement before expanding
10008 the address rather than afterwards, and adjust for the new interface.
10009 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10010 Match the new hook interface. Handle SVE addresses.
10011 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10012 new hook interface.
10014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10016 * Makefile.in (OBJS): Add early-remat.o.
10017 * target.def (select_early_remat_modes): New hook.
10018 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10019 * doc/tm.texi: Regenerate.
10020 * targhooks.h (default_select_early_remat_modes): Declare.
10021 * targhooks.c (default_select_early_remat_modes): New function.
10022 * timevar.def (TV_EARLY_REMAT): New timevar.
10023 * passes.def (pass_early_remat): New pass.
10024 * tree-pass.h (make_pass_early_remat): Declare.
10025 * early-remat.c: New file.
10026 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10028 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10031 Alan Hayward <alan.hayward@arm.com>
10032 David Sherwood <david.sherwood@arm.com>
10034 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10035 vfm1 with a bound_epilog parameter.
10036 (vect_do_peeling): Update calls accordingly, and move the prologue
10037 call earlier in the function. Treat the base bound_epilog as 0 for
10038 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10039 this base when peeling for gaps.
10040 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10041 with fully-masked loops.
10042 (vect_estimate_min_profitable_iters): Handle the single peeled
10043 iteration in that case.
10045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10046 Alan Hayward <alan.hayward@arm.com>
10047 David Sherwood <david.sherwood@arm.com>
10049 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10050 single-element interleaving even if the size is not a power of 2.
10051 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10052 accesses for single-element interleaving if the group size is
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
10060 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10061 * optabs.def (fold_extract_last_optab): New optab.
10062 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10063 * internal-fn.c (fold_extract_direct): New macro.
10064 (expand_fold_extract_optab_fn): Likewise.
10065 (direct_fold_extract_optab_supported_p): Likewise.
10066 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10067 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10068 EXTRACT_LAST_REDUCTION.
10069 (get_initial_def_for_reduction): Do not create an initial vector
10070 for EXTRACT_LAST_REDUCTION reductions.
10071 (vectorizable_reduction): Leave the scalar phi in place for
10072 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10073 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10074 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10075 transform phase to vectorizable_condition.
10076 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10078 (vect_finish_stmt_generation): ...here.
10079 (vect_finish_replace_stmt): New function.
10080 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10081 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10083 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * doc/md.texi (extract_last_@var{m}): Document.
10090 * optabs.def (extract_last_optab): New optab.
10091 * internal-fn.def (EXTRACT_LAST): New internal function.
10092 * internal-fn.c (cond_unary_direct): New macro.
10093 (expand_cond_unary_optab_fn): Likewise.
10094 (direct_cond_unary_optab_supported_p): Likewise.
10095 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10096 loops using EXTRACT_LAST.
10097 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10098 (extract_last_<mode>): ...this optab.
10099 (vec_extract<mode><Vel>): Update accordingly.
10101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10102 Alan Hayward <alan.hayward@arm.com>
10103 David Sherwood <david.sherwood@arm.com>
10105 * target.def (empty_mask_is_expensive): New hook.
10106 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10107 * doc/tm.texi: Regenerate.
10108 * targhooks.h (default_empty_mask_is_expensive): Declare.
10109 * targhooks.c (default_empty_mask_is_expensive): New function.
10110 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10111 if the target says that empty masks are expensive.
10112 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10114 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10117 Alan Hayward <alan.hayward@arm.com>
10118 David Sherwood <david.sherwood@arm.com>
10120 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10121 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10122 (vect_use_loop_mask_for_alignment_p): New function.
10123 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10124 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10125 niters_skip argument. Make sure that the first niters_skip elements
10126 of the first iteration are inactive.
10127 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10128 Update call to vect_set_loop_masks_directly.
10129 (get_misalign_in_elems): New function, split out from...
10130 (vect_gen_prolog_loop_niters): ...here.
10131 (vect_update_init_of_dr): Take a code argument that specifies whether
10132 the adjustment should be added or subtracted.
10133 (vect_update_init_of_drs): Likewise.
10134 (vect_prepare_for_masked_peels): New function.
10135 (vect_do_peeling): Skip prologue peeling if we're using a mask
10136 instead. Update call to vect_update_inits_of_drs.
10137 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10139 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10140 alignment. Do not include the number of peeled iterations in
10141 the minimum threshold in that case.
10142 (vectorizable_induction): Adjust the start value down by
10143 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10144 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10145 Take the number of skipped iterations into account when calculating
10147 * tree-vect-stmts.c (vect_gen_while_not): New function.
10149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10150 Alan Hayward <alan.hayward@arm.com>
10151 David Sherwood <david.sherwood@arm.com>
10153 * doc/sourcebuild.texi (vect_fully_masked): Document.
10154 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10155 default value to 0.
10156 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10158 (vect_analyze_loop_2): ...here. Don't check the vectorization
10159 factor against the number of loop iterations if the loop is
10162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10167 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10168 (dump_groups): Update accordingly.
10169 (iv_use::mem_type): New member variable.
10170 (address_p): New function.
10171 (record_use): Add a mem_type argument and initialize the new
10173 (record_group_use): Add a mem_type argument. Use address_p.
10174 Remove obsolete null checks of base_object. Update call to record_use.
10175 (find_interesting_uses_op): Update call to record_group_use.
10176 (find_interesting_uses_cond): Likewise.
10177 (find_interesting_uses_address): Likewise.
10178 (get_mem_type_for_internal_fn): New function.
10179 (find_address_like_use): Likewise.
10180 (find_interesting_uses_stmt): Try find_address_like_use before
10181 calling find_interesting_uses_op.
10182 (addr_offset_valid_p): Use the iv mem_type field as the type
10183 of the addressed memory.
10184 (add_autoinc_candidates): Likewise.
10185 (get_address_cost): Likewise.
10186 (split_small_address_groups_p): Use address_p.
10187 (split_address_groups): Likewise.
10188 (add_iv_candidate_for_use): Likewise.
10189 (autoinc_possible_for_pair): Likewise.
10190 (rewrite_groups): Likewise.
10191 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10192 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10193 (get_alias_ptr_type_for_ptr_address): New function.
10194 (rewrite_use_address): Rewrite address uses in calls that were
10195 identified by find_address_like_use.
10197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10198 Alan Hayward <alan.hayward@arm.com>
10199 David Sherwood <david.sherwood@arm.com>
10201 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10203 * gimple-expr.h (is_gimple_addressable: Likewise.
10204 * gimple-expr.c (is_gimple_address): Likewise.
10205 * internal-fn.c (expand_call_mem_ref): New function.
10206 (expand_mask_load_optab_fn): Use it.
10207 (expand_mask_store_optab_fn): Likewise.
10209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10210 Alan Hayward <alan.hayward@arm.com>
10211 David Sherwood <david.sherwood@arm.com>
10213 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10214 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10215 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10216 (cond_umax@var{mode}): Document.
10217 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10218 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10219 (cond_umin_optab, cond_umax_optab): New optabs.
10220 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10221 (COND_IOR, COND_XOR): New internal functions.
10222 * internal-fn.h (get_conditional_internal_fn): Declare.
10223 * internal-fn.c (cond_binary_direct): New macro.
10224 (expand_cond_binary_optab_fn): Likewise.
10225 (direct_cond_binary_optab_supported_p): Likewise.
10226 (get_conditional_internal_fn): New function.
10227 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10228 Cope with reduction statements that are vectorized as calls rather
10230 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10231 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10232 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10233 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10234 (UNSPEC_COND_EOR): New unspecs.
10235 (optab): Add mappings for them.
10236 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10237 (sve_int_op, sve_fp_op): New int attributes.
10239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10240 Alan Hayward <alan.hayward@arm.com>
10241 David Sherwood <david.sherwood@arm.com>
10243 * optabs.def (while_ult_optab): New optab.
10244 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10245 * internal-fn.def (WHILE_ULT): New internal function.
10246 * internal-fn.h (direct_internal_fn_supported_p): New override
10247 that takes two types as argument.
10248 * internal-fn.c (while_direct): New macro.
10249 (expand_while_optab_fn): New function.
10250 (convert_optab_supported_p): Likewise.
10251 (direct_while_optab_supported_p): New macro.
10252 * wide-int.h (wi::udiv_ceil): New function.
10253 * tree-vectorizer.h (rgroup_masks): New structure.
10254 (vec_loop_masks): New typedef.
10255 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10256 and fully_masked_p.
10257 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10258 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10259 (vect_max_vf): New function.
10260 (slpeel_make_loop_iterate_ntimes): Delete.
10261 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10262 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10263 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10264 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10265 internal-fn.h, stor-layout.h and optabs-query.h.
10266 (vect_set_loop_mask): New function.
10267 (add_preheader_seq): Likewise.
10268 (add_header_seq): Likewise.
10269 (interleave_supported_p): Likewise.
10270 (vect_maybe_permute_loop_masks): Likewise.
10271 (vect_set_loop_masks_directly): Likewise.
10272 (vect_set_loop_condition_masked): Likewise.
10273 (vect_set_loop_condition_unmasked): New function, split out from
10274 slpeel_make_loop_iterate_ntimes.
10275 (slpeel_make_loop_iterate_ntimes): Rename to..
10276 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10277 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10278 (vect_do_peeling): Update call accordingly.
10279 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10281 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10282 mask_compare_type, can_fully_mask_p and fully_masked_p.
10283 (release_vec_loop_masks): New function.
10284 (_loop_vec_info): Use it to free the loop masks.
10285 (can_produce_all_loop_masks_p): New function.
10286 (vect_get_max_nscalars_per_iter): Likewise.
10287 (vect_verify_full_masking): Likewise.
10288 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10289 retries, and free the mask rgroups before retrying. Check loop-wide
10290 reasons for disallowing fully-masked loops. Make the final decision
10291 about whether use a fully-masked loop or not.
10292 (vect_estimate_min_profitable_iters): Do not assume that peeling
10293 for the number of iterations will be needed for fully-masked loops.
10294 (vectorizable_reduction): Disable fully-masked loops.
10295 (vectorizable_live_operation): Likewise.
10296 (vect_halve_mask_nunits): New function.
10297 (vect_double_mask_nunits): Likewise.
10298 (vect_record_loop_mask): Likewise.
10299 (vect_get_loop_mask): Likewise.
10300 (vect_transform_loop): Handle the case in which the final loop
10301 iteration might handle a partial vector. Call vect_set_loop_condition
10302 instead of slpeel_make_loop_iterate_ntimes.
10303 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10304 (check_load_store_masking): New function.
10305 (prepare_load_store_mask): Likewise.
10306 (vectorizable_store): Handle fully-masked loops.
10307 (vectorizable_load): Likewise.
10308 (supportable_widening_operation): Use vect_halve_mask_nunits for
10310 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10311 (vect_gen_while): New function.
10312 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10313 (aarch64_uqdec<mode>): New insn.
10315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10316 Alan Hayward <alan.hayward@arm.com>
10317 David Sherwood <david.sherwood@arm.com>
10319 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10320 (reduc_xor_scal_optab): New optabs.
10321 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10322 (reduc_xor_scal_@var{m}): Document.
10323 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10324 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10325 internal functions.
10326 * fold-const-call.c (fold_const_call): Handle them.
10327 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10328 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10329 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10330 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10331 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10332 (UNSPEC_XORV): New unspecs.
10333 (optab): Add entries for them.
10334 (BITWISEV): New int iterator.
10335 (bit_reduc_op): New int attributes.
10337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10338 Alan Hayward <alan.hayward@arm.com>
10339 David Sherwood <david.sherwood@arm.com>
10341 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10342 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10343 * optabs.def (vec_shl_insert_optab): New optab.
10344 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10345 (duplicate_and_interleave): Likewise.
10346 * tree-vect-loop.c: Include internal-fn.h.
10347 (neutral_op_for_slp_reduction): New function, split out from
10348 get_initial_defs_for_reduction.
10349 (get_initial_def_for_reduction): Handle option 2 for variable-length
10350 vectors by loading the neutral value into a vector and then shifting
10351 the initial value into element 0.
10352 (get_initial_defs_for_reduction): Replace the code argument with
10353 the neutral value calculated by neutral_op_for_slp_reduction.
10354 Use gimple_build_vector for constant-length vectors.
10355 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10356 but the first group_size elements have a neutral value.
10357 Use duplicate_and_interleave otherwise.
10358 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10359 Update call to get_initial_defs_for_reduction. Handle SLP
10360 reductions for variable-length vectors by creating one vector
10361 result for each scalar result, with the elements associated
10362 with other scalar results stubbed out with the neutral value.
10363 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10364 Require IFN_VEC_SHL_INSERT for double reductions on
10365 variable-length vectors, or SLP reductions that have
10366 a neutral value. Require can_duplicate_and_interleave_p
10367 support for variable-length unchained SLP reductions if there
10368 is no neutral value, such as for MIN/MAX reductions. Also require
10369 the number of vector elements to be a multiple of the number of
10370 SLP statements when doing variable-length unchained SLP reductions.
10371 Update call to vect_create_epilog_for_reduction.
10372 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10373 and remove initial values.
10374 (duplicate_and_interleave): Make public.
10375 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10376 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10379 Alan Hayward <alan.hayward@arm.com>
10380 David Sherwood <david.sherwood@arm.com>
10382 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10383 (can_duplicate_and_interleave_p): New function.
10384 (vect_get_and_check_slp_defs): Take the vector of statements
10385 rather than just the current one. Remove excess parentheses.
10386 Restriction rejectinon of vect_constant_def and vect_external_def
10387 for variable-length vectors to boolean types, or types for which
10388 can_duplicate_and_interleave_p is false.
10389 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10390 (duplicate_and_interleave): New function.
10391 (vect_get_constant_vectors): Use gimple_build_vector for
10392 constant-length vectors and suitable variable-length constant
10393 vectors. Use duplicate_and_interleave for other variable-length
10394 vectors. Don't defer the update when inserting new statements.
10396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10397 Alan Hayward <alan.hayward@arm.com>
10398 David Sherwood <david.sherwood@arm.com>
10400 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10401 min_profitable_iters doesn't go negative.
10403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10404 Alan Hayward <alan.hayward@arm.com>
10405 David Sherwood <david.sherwood@arm.com>
10407 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10408 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10409 * optabs.def (vec_mask_load_lanes_optab): New optab.
10410 (vec_mask_store_lanes_optab): Likewise.
10411 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10412 (MASK_STORE_LANES): Likewise.
10413 * internal-fn.c (mask_load_lanes_direct): New macro.
10414 (mask_store_lanes_direct): Likewise.
10415 (expand_mask_load_optab_fn): Handle masked operations.
10416 (expand_mask_load_lanes_optab_fn): New macro.
10417 (expand_mask_store_optab_fn): Handle masked operations.
10418 (expand_mask_store_lanes_optab_fn): New macro.
10419 (direct_mask_load_lanes_optab_supported_p): Likewise.
10420 (direct_mask_store_lanes_optab_supported_p): Likewise.
10421 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10423 (vect_load_lanes_supported): Likewise.
10424 * tree-vect-data-refs.c (strip_conversion): New function.
10425 (can_group_stmts_p): Likewise.
10426 (vect_analyze_data_ref_accesses): Use it instead of checking
10427 for a pair of assignments.
10428 (vect_store_lanes_supported): Take a masked_p parameter.
10429 (vect_load_lanes_supported): Likewise.
10430 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10431 vect_store_lanes_supported and vect_load_lanes_supported.
10432 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10433 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10434 parameter. Don't allow gaps for masked accesses.
10435 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10436 and vect_load_lanes_supported.
10437 (get_load_store_type): Take a masked_p parameter and update
10438 call to get_group_load_store_type.
10439 (vectorizable_store): Update call to get_load_store_type.
10440 Handle IFN_MASK_STORE_LANES.
10441 (vectorizable_load): Update call to get_load_store_type.
10442 Handle IFN_MASK_LOAD_LANES.
10444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10445 Alan Hayward <alan.hayward@arm.com>
10446 David Sherwood <david.sherwood@arm.com>
10448 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10450 * config/aarch64/aarch64-protos.h
10451 (aarch64_sve_struct_memory_operand_p): Declare.
10452 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10453 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10454 (VPRED, vpred): Handle SVE structure modes.
10455 * config/aarch64/constraints.md (Utx): New constraint.
10456 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10457 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10458 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10459 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10460 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10461 structure modes. Split into pieces after RA.
10462 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10463 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10465 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10466 SVE structure modes.
10467 (aarch64_classify_address): Likewise.
10468 (sizetochar): Move earlier in file.
10469 (aarch64_print_operand): Handle SVE register lists.
10470 (aarch64_array_mode): New function.
10471 (aarch64_sve_struct_memory_operand_p): Likewise.
10472 (TARGET_ARRAY_MODE): Redefine.
10474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10475 Alan Hayward <alan.hayward@arm.com>
10476 David Sherwood <david.sherwood@arm.com>
10478 * target.def (array_mode): New target hook.
10479 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10480 * doc/tm.texi: Regenerate.
10481 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10482 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10483 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10484 targetm.array_mode.
10485 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10489 Alan Hayward <alan.hayward@arm.com>
10490 David Sherwood <david.sherwood@arm.com>
10492 * fold-const.c (fold_binary_loc): Check the argument types
10493 rather than the result type when testing for a vector operation.
10495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10497 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10498 * doc/tm.texi: Regenerate.
10500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10501 Alan Hayward <alan.hayward@arm.com>
10502 David Sherwood <david.sherwood@arm.com>
10504 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10505 (sve): Document new AArch64 extension.
10506 * doc/md.texi (w): Extend the description of the AArch64
10507 constraint to include SVE vectors.
10508 (Upl, Upa): Document new AArch64 predicate constraints.
10509 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10511 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10512 (msve-vector-bits=): New option.
10513 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10514 SVE when these are disabled.
10515 (sve): New extension.
10516 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10517 modes. Adjust their number of units based on aarch64_sve_vg.
10518 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10519 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10520 aarch64_addr_query_type.
10521 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10522 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10523 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10524 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10525 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10526 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10527 (aarch64_simd_imm_zero_p): Delete.
10528 (aarch64_check_zero_based_sve_index_immediate): Declare.
10529 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10530 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10531 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10532 (aarch64_sve_float_mul_immediate_p): Likewise.
10533 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10534 rather than an rtx.
10535 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10536 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10537 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10538 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10539 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10540 (aarch64_regmode_natural_size): Likewise.
10541 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10542 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10544 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10545 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10546 for VG and the SVE predicate registers.
10547 (V_ALIASES): Add a "z"-prefixed alias.
10548 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10549 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10550 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10551 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10552 (REG_CLASS_NAMES): Add entries for them.
10553 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10554 and the predicate registers.
10555 (aarch64_sve_vg): Declare.
10556 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10557 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10558 (REGMODE_NATURAL_SIZE): Define.
10559 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10561 * config/aarch64/aarch64.c: Include cfgrtl.h.
10562 (simd_immediate_info): Add a constructor for series vectors,
10563 and an associated step field.
10564 (aarch64_sve_vg): New variable.
10565 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10566 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10567 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10568 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10569 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10570 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10571 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10572 (aarch64_get_mask_mode): New functions.
10573 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10574 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10575 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10576 predicate modes and predicate registers. Explicitly restrict
10577 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10578 to store a vector mode if it is recognized by
10579 aarch64_classify_vector_mode.
10580 (aarch64_regmode_natural_size): New function.
10581 (aarch64_hard_regno_caller_save_mode): Return the original mode
10583 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10584 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10585 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10586 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10588 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10589 does not overlap dest if the function is frame-related. Handle
10591 (aarch64_split_add_offset): New function.
10592 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10593 them aarch64_add_offset.
10594 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10595 and update call to aarch64_sub_sp.
10596 (aarch64_add_cfa_expression): New function.
10597 (aarch64_expand_prologue): Pass extra temporary registers to the
10598 functions above. Handle the case in which we need to emit new
10599 DW_CFA_expressions for registers that were originally saved
10600 relative to the stack pointer, but now have to be expressed
10601 relative to the frame pointer.
10602 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10604 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10605 IP0 and IP1 values for SVE frames.
10606 (aarch64_expand_vec_series): New function.
10607 (aarch64_expand_sve_widened_duplicate): Likewise.
10608 (aarch64_expand_sve_const_vector): Likewise.
10609 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10610 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10611 into the register, rather than emitting a SET directly.
10612 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10613 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10614 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10615 (offset_9bit_signed_scaled_p): New functions.
10616 (aarch64_replicate_bitmask_imm): New function.
10617 (aarch64_bitmask_imm): Use it.
10618 (aarch64_cannot_force_const_mem): Reject expressions involving
10619 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10620 (aarch64_classify_index): Handle SVE indices, by requiring
10621 a plain register index with a scale that matches the element size.
10622 (aarch64_classify_address): Handle SVE addresses. Assert that
10623 the mode of the address is VOIDmode or an integer mode.
10624 Update call to aarch64_classify_symbol.
10625 (aarch64_classify_symbolic_expression): Update call to
10626 aarch64_classify_symbol.
10627 (aarch64_const_vec_all_in_range_p): New function.
10628 (aarch64_print_vector_float_operand): Likewise.
10629 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10630 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10631 and the FP immediates 1.0 and 0.5.
10632 (aarch64_print_address_internal): Handle SVE addresses.
10633 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10634 (aarch64_regno_regclass): Handle predicate registers.
10635 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10637 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10638 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10639 (aarch64_convert_sve_vector_bits): New function.
10640 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10641 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10642 rather than an rtx.
10643 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10644 Handle SVE vector and predicate modes. Accept VL-based constants
10645 that need only one temporary register, and VL offsets that require
10646 no temporary registers.
10647 (aarch64_conditional_register_usage): Mark the predicate registers
10648 as fixed if SVE isn't available.
10649 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10650 Return true for SVE vector and predicate modes.
10651 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10652 rather than an unsigned int. Handle SVE modes.
10653 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10655 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10657 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10658 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10659 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10660 (aarch64_sve_float_mul_immediate_p): New functions.
10661 (aarch64_sve_valid_immediate): New function.
10662 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10663 Explicitly reject structure modes. Check for INDEX constants.
10664 Handle PTRUE and PFALSE constants.
10665 (aarch64_check_zero_based_sve_index_immediate): New function.
10666 (aarch64_simd_imm_zero_p): Delete.
10667 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10668 vector modes. Accept constants in the range of CNT[BHWD].
10669 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10670 ask for an Advanced SIMD mode.
10671 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10672 (aarch64_simd_vector_alignment): Handle SVE predicates.
10673 (aarch64_vectorize_preferred_vector_alignment): New function.
10674 (aarch64_simd_vector_alignment_reachable): Use it instead of
10676 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10677 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10679 (MAX_VECT_LEN): Delete.
10680 (expand_vec_perm_d): Add a vec_flags field.
10681 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10682 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10683 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10685 (aarch64_evpc_rev): Rename to...
10686 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10687 (aarch64_evpc_rev_global): New function.
10688 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10689 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10691 (aarch64_evpc_sve_tbl): New function.
10692 (aarch64_expand_vec_perm_const_1): Update after rename of
10693 aarch64_evpc_rev. Handle SVE permutes too, trying
10694 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10695 than aarch64_evpc_tbl.
10696 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10697 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10698 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10699 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10700 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10701 (aarch64_expand_sve_vcond): New functions.
10702 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10703 of aarch64_vector_mode_p.
10704 (aarch64_dwarf_poly_indeterminate_value): New function.
10705 (aarch64_compute_pressure_classes): Likewise.
10706 (aarch64_can_change_mode_class): Likewise.
10707 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10708 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10709 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10710 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10711 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10712 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10713 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10714 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10716 (Dn, Dl, Dr): Accept const as well as const_vector.
10717 (Dz): Likewise. Compare against CONST0_RTX.
10718 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10719 of "vector" where appropriate.
10720 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10721 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10722 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10723 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10724 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10725 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10726 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10727 (v_int_equiv): Extend to SVE modes.
10728 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10730 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10731 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10732 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10733 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10734 (SVE_COND_FP_CMP): New int iterators.
10735 (perm_hilo): Handle the new unpack unspecs.
10736 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10738 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10739 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10740 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10741 (aarch64_equality_operator, aarch64_constant_vector_operand)
10742 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10743 (aarch64_sve_nonimmediate_operand): Likewise.
10744 (aarch64_sve_general_operand): Likewise.
10745 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10746 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10747 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10748 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10749 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10750 (aarch64_sve_float_arith_immediate): Likewise.
10751 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10752 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10753 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10754 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10755 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10756 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10757 (aarch64_sve_float_arith_operand): Likewise.
10758 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10759 (aarch64_sve_float_mul_operand): Likewise.
10760 (aarch64_sve_vec_perm_operand): Likewise.
10761 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10762 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10763 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10764 as well as const_vector.
10765 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10766 in file. Use CONST0_RTX and CONSTM1_RTX.
10767 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10768 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10769 Use aarch64_simd_imm_zero.
10770 * config/aarch64/aarch64-sve.md: New file.
10771 * config/aarch64/aarch64.md: Include it.
10772 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10773 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10774 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10775 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10776 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10777 (sve): New attribute.
10778 (enabled): Disable instructions with the sve attribute unless
10780 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10781 aarch64_expand_mov_immediate.
10782 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10783 CNT[BHSD] immediates.
10784 (movti): Split CONST_POLY_INT moves into two halves.
10785 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10786 Split additions that need a temporary here if the destination
10787 is the stack pointer.
10788 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10789 (*add<mode>3_poly_1): New instruction.
10790 (set_clobber_cc): New expander.
10792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10794 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10795 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10796 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10797 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10798 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10799 Change innermode from fixed_mode_size to machine_mode.
10800 (simplify_subreg): Update call accordingly. Handle a constant-sized
10801 subreg of a variable-length CONST_VECTOR.
10803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10808 (add_offset_to_base): New function, split out from...
10809 (create_mem_ref): ...here. When handling a scale other than 1,
10810 check first whether the address is valid without the offset.
10811 Add it into the base if so, leaving the index and scale as-is.
10813 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10816 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10817 fold_for_warn before checking if arg2 is INTEGER_CST.
10819 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10821 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10822 (store_multiple_operation): Delete.
10823 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10824 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10825 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10826 guarded by TARGET_STRING.
10827 (rs6000_output_load_multiple): Delete.
10828 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10829 OPTION_MASK_STRING / TARGET_STRING handling.
10830 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10831 (const rs6000_opt_masks) <"string">: Change mask to 0.
10832 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10833 (MASK_STRING): Delete.
10834 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10836 (load_multiple): Delete.
10843 (store_multiple): Delete.
10850 (movmemsi_8reg): Delete.
10851 (corresponding unnamed define_insn): Delete.
10852 (movmemsi_6reg): Delete.
10853 (corresponding unnamed define_insn): Delete.
10854 (movmemsi_4reg): Delete.
10855 (corresponding unnamed define_insn): Delete.
10856 (movmemsi_2reg): Delete.
10857 (corresponding unnamed define_insn): Delete.
10858 (movmemsi_1reg): Delete.
10859 (corresponding unnamed define_insn): Delete.
10860 * config/rs6000/rs6000.opt (mno-string): New.
10861 (mstring): Replace by deprecation warning stub.
10862 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10864 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10866 * regrename.c (regrename_do_replace): If replacing the same
10867 reg multiple times, try to reuse last created gen_raw_REG.
10870 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10871 main to workaround a bug in GDB.
10873 2018-01-12 Tom de Vries <tom@codesourcery.com>
10876 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10878 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10880 PR rtl-optimization/80481
10881 * ira-color.c (get_cap_member): New function.
10882 (allocnos_conflict_by_live_ranges_p): Use it.
10883 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10884 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10886 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10889 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10890 (*saddl_se_1): Ditto.
10891 (*ssubsi_1): Ditto.
10892 (*ssubl_se_1): Ditto.
10894 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10896 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10897 rather than wi::to_widest for DR_INITs.
10898 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10899 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10900 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10902 (vect_analyze_group_access_1): Note that here.
10904 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10906 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10907 polynomial type sizes.
10909 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10911 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10912 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10913 (gimple_add_tmp_var): Likewise.
10915 2018-01-12 Martin Liska <mliska@suse.cz>
10917 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10918 (gimple_alloc_sizes): Likewise.
10919 (dump_gimple_statistics): Use PRIu64 in printf format.
10920 * gimple.h: Change uint64_t to int.
10922 2018-01-12 Martin Liska <mliska@suse.cz>
10924 * tree-core.h: Use uint64_t instead of int.
10925 * tree.c (tree_node_counts): Likewise.
10926 (tree_node_sizes): Likewise.
10927 (dump_tree_statistics): Use PRIu64 in printf format.
10929 2018-01-12 Martin Liska <mliska@suse.cz>
10931 * Makefile.in: As qsort_chk is implemented in vec.c, add
10932 vec.o to linkage of gencfn-macros.
10933 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10934 passing the info to record_node_allocation_statistics.
10935 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10937 * ggc-common.c (struct ggc_usage): Add operator== and use
10938 it in operator< and compare function.
10939 * mem-stats.h (struct mem_usage): Likewise.
10940 * vec.c (struct vec_usage): Remove operator< and compare
10941 function. Can be simply inherited.
10943 2018-01-12 Martin Jambor <mjambor@suse.cz>
10946 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10947 * tree-ssa-math-opts.c: Include domwalk.h.
10948 (convert_mult_to_fma_1): New function.
10949 (fma_transformation_info): New type.
10950 (fma_deferring_state): Likewise.
10951 (cancel_fma_deferring): New function.
10952 (result_of_phi): Likewise.
10953 (last_fma_candidate_feeds_initial_phi): Likewise.
10954 (convert_mult_to_fma): Added deferring logic, split actual
10955 transformation to convert_mult_to_fma_1.
10956 (math_opts_dom_walker): New type.
10957 (math_opts_dom_walker::after_dom_children): New method, body moved
10958 here from pass_optimize_widening_mul::execute, added deferring logic
10960 (pass_optimize_widening_mul::execute): Moved most of code to
10961 math_opts_dom_walker::after_dom_children.
10962 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10963 * config/i386/i386.c (ix86_option_override_internal): Added
10964 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10966 2018-01-12 Richard Biener <rguenther@suse.de>
10969 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10970 inline instance vars.
10972 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10975 * config/rx/rx.c (rx_is_restricted_memory_address):
10976 Handle SUBREG case.
10978 2018-01-12 Richard Biener <rguenther@suse.de>
10980 PR tree-optimization/80846
10981 * target.def (split_reduction): New target hook.
10982 * targhooks.c (default_split_reduction): New function.
10983 * targhooks.h (default_split_reduction): Declare.
10984 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10985 target requests first reduce vectors by combining low and high
10987 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10988 (get_vectype_for_scalar_type_and_size): Export.
10989 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10990 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10991 * doc/tm.texi: Regenerate.
10992 * config/i386/i386.c (ix86_split_reduction): Implement
10993 TARGET_VECTORIZE_SPLIT_REDUCTION.
10995 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10998 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10999 in PIC mode except for TARGET_VXWORKS_RTP.
11000 * config/sparc/sparc.c: Include cfgrtl.h.
11001 (TARGET_INIT_PIC_REG): Define.
11002 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11003 (sparc_pic_register_p): New predicate.
11004 (sparc_legitimate_address_p): Use it.
11005 (sparc_legitimize_pic_address): Likewise.
11006 (sparc_delegitimize_address): Likewise.
11007 (sparc_mode_dependent_address_p): Likewise.
11008 (gen_load_pcrel_sym): Remove 4th parameter.
11009 (load_got_register): Adjust call to above. Remove obsolete stuff.
11010 (sparc_expand_prologue): Do not call load_got_register here.
11011 (sparc_flat_expand_prologue): Likewise.
11012 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11013 (sparc_use_pseudo_pic_reg): New function.
11014 (sparc_init_pic_reg): Likewise.
11015 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11016 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11018 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11020 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11021 Add item for branch_cost.
11023 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11025 PR rtl-optimization/83565
11026 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11027 not extend the result to a larger mode for rotate operations.
11028 (num_sign_bit_copies1): Likewise.
11030 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11033 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11035 Use values-Xc.o for -pedantic.
11036 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11038 2018-01-12 Martin Liska <mliska@suse.cz>
11041 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11043 (possible_polymorphic_call_targets): Use it.
11044 (ipa_devirt): Likewise.
11046 2018-01-12 Martin Liska <mliska@suse.cz>
11048 * profile-count.h (enum profile_quality): Use 0 as invalid
11049 enum value of profile_quality.
11051 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11053 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11054 -mext-string options.
11056 2018-01-12 Richard Biener <rguenther@suse.de>
11058 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11059 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11060 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11062 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11064 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11066 * configure.ac (--with-long-double-format): Add support for the
11067 configuration option to change the default long double format on
11069 * config.gcc (powerpc*-linux*-*): Likewise.
11070 * configure: Regenerate.
11071 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11072 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11073 used without modification.
11075 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11077 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11078 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11079 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11080 MISC_BUILTIN_SPEC_BARRIER.
11081 (rs6000_init_builtins): Likewise.
11082 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11084 (speculation_barrier): New define_insn.
11085 * doc/extend.texi: Document __builtin_speculation_barrier.
11087 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11090 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11091 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11092 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11094 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11095 integral modes instead of "ss" and "sd".
11096 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11097 vectors with 32-bit and 64-bit elements.
11098 (vecdupssescalarmodesuffix): New mode attribute.
11099 (vec_dup<mode>): Use it.
11101 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11104 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11105 frame if argument is passed on stack.
11107 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11110 * ree.c (combine_reaching_defs): Optimize also
11111 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11112 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11114 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11116 PR middle-end/83189
11117 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11119 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11121 PR middle-end/83718
11122 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11123 after they are computed.
11125 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11127 PR tree-optimization/83695
11128 * gimple-loop-linterchange.cc
11129 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11130 reset cached scev information after interchange.
11131 (pass_linterchange::execute): Remove call to scev_reset_htab.
11133 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11135 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11136 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11137 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11138 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11139 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11140 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11141 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11142 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11143 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11144 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11145 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11146 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11147 (V_lane_reg): Likewise.
11148 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11150 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11151 (vfmal_lane_low<mode>_intrinsic,
11152 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11153 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11154 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11155 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11156 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11157 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11159 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11161 * config/arm/arm-cpus.in (fp16fml): New feature.
11162 (ALL_SIMD): Add fp16fml.
11163 (armv8.2-a): Add fp16fml as an option.
11164 (armv8.3-a): Likewise.
11165 (armv8.4-a): Add fp16fml as part of fp16.
11166 * config/arm/arm.h (TARGET_FP16FML): Define.
11167 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11169 * config/arm/arm-modes.def (V2HF): Define.
11170 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11171 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11172 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11173 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11174 vfmsl_low, vfmsl_high): New set of builtins.
11175 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11176 (vfml_op): New code attribute.
11177 (VFMLHALVES): New int iterator.
11178 (VFML, VFMLSEL): New mode attributes.
11179 (V_reg): Define mapping for V2HF.
11180 (V_hi, V_lo): New mode attributes.
11181 (VF_constraint): Likewise.
11182 (vfml_half, vfml_half_selector): New int attributes.
11183 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11185 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11186 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11188 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11189 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11190 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11191 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11193 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11194 Document new effective target and option set.
11196 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11198 * config/arm/arm-cpus.in (armv8_4): New feature.
11199 (ARMv8_4a): New fgroup.
11200 (armv8.4-a): New arch.
11201 * config/arm/arm-tables.opt: Regenerate.
11202 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11203 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11204 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11205 Add matching rules for -march=armv8.4-a and extensions.
11206 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11208 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11211 * config/rx/rx.md (BW): New mode attribute.
11212 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11214 2018-01-11 Richard Biener <rguenther@suse.de>
11216 PR tree-optimization/83435
11217 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11218 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11219 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11221 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11222 Alan Hayward <alan.hayward@arm.com>
11223 David Sherwood <david.sherwood@arm.com>
11225 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11227 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11228 (aarch64_print_address_internal): Use it to check for a zero offset.
11230 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11231 Alan Hayward <alan.hayward@arm.com>
11232 David Sherwood <david.sherwood@arm.com>
11234 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11235 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11236 Return a poly_int64 rather than a HOST_WIDE_INT.
11237 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11238 rather than a HOST_WIDE_INT.
11239 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11240 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11241 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11242 final_offset from HOST_WIDE_INT to poly_int64.
11243 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11244 to_constant when getting the number of units in an Advanced SIMD
11246 (aarch64_builtin_vectorized_function): Check for a constant number
11248 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11250 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11251 attribute instead of GET_MODE_NUNITS.
11252 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11253 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11254 GET_MODE_SIZE for fixed-size registers.
11255 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11256 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11257 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11258 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11259 (aarch64_print_operand, aarch64_print_address_internal)
11260 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11261 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11262 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11263 Handle polynomial GET_MODE_SIZE.
11264 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11265 wider than SImode without modification.
11266 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11267 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11268 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11269 passing and returning SVE modes.
11270 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11271 rather than GEN_INT.
11272 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11273 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11274 (aarch64_allocate_and_probe_stack_space): Likewise.
11275 (aarch64_layout_frame): Cope with polynomial offsets.
11276 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11277 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11278 polynomial offsets.
11279 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11280 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11281 poly_int64 rather than a HOST_WIDE_INT.
11282 (aarch64_get_separate_components, aarch64_process_components)
11283 (aarch64_expand_prologue, aarch64_expand_epilogue)
11284 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11285 (aarch64_anchor_offset): New function, split out from...
11286 (aarch64_legitimize_address): ...here.
11287 (aarch64_builtin_vectorization_cost): Handle polynomial
11288 TYPE_VECTOR_SUBPARTS.
11289 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11291 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11292 number of elements from the PARALLEL rather than the mode.
11293 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11294 rather than GET_MODE_BITSIZE.
11295 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11296 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11297 (aarch64_expand_vec_perm_const_1): Handle polynomial
11298 d->perm.length () and d->perm elements.
11299 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11300 Apply to_constant to d->perm elements.
11301 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11302 polynomial CONST_VECTOR_NUNITS.
11303 (aarch64_move_pointer): Take amount as a poly_int64 rather
11305 (aarch64_progress_pointer): Avoid temporary variable.
11306 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11307 the mode attribute instead of GET_MODE.
11309 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11310 Alan Hayward <alan.hayward@arm.com>
11311 David Sherwood <david.sherwood@arm.com>
11313 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11314 x exists before using it.
11315 (aarch64_add_constant_internal): Rename to...
11316 (aarch64_add_offset_1): ...this. Replace regnum with separate
11317 src and dest rtxes. Handle the case in which they're different,
11318 including when the offset is zero. Replace scratchreg with an rtx.
11319 Use 2 additions if there is no spare register into which we can
11320 move a 16-bit constant.
11321 (aarch64_add_constant): Delete.
11322 (aarch64_add_offset): Replace reg with separate src and dest
11323 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11324 Use aarch64_add_offset_1.
11325 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11326 an rtx rather than an int. Take the delta as a poly_int64
11327 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11328 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11329 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11330 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11331 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11332 and aarch64_add_sp.
11333 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11334 aarch64_add_constant.
11336 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11338 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11339 Use scalar_float_mode.
11341 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11343 * config/aarch64/aarch64-simd.md
11344 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11345 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11346 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11347 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11348 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11349 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11350 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11351 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11352 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11353 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11355 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11358 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11359 targ_options->x_arm_arch_string is non NULL.
11361 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11363 * config/aarch64/aarch64.h
11364 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11366 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11369 * expmed.c (emit_store_flag_force): Swap if const op0
11370 and change VOIDmode to mode of op0.
11372 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11374 PR rtl-optimization/83761
11375 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11376 than bytes to mode_for_size.
11378 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11380 PR middle-end/83189
11381 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11382 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11385 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11387 PR middle-end/83575
11388 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11389 when in layout mode.
11390 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11391 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11394 2018-01-10 Michael Collison <michael.collison@arm.com>
11396 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11397 * config/aarch64/aarch64-option-extension.def: Add
11398 AARCH64_OPT_EXTENSION of 'fp16fml'.
11399 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11400 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11401 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11402 * config/aarch64/constraints.md (Ui7): New constraint.
11403 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11404 (VFMLA_SEL_W): Ditto.
11407 (VFMLA16_LOW): New int iterator.
11408 (VFMLA16_HIGH): Ditto.
11409 (UNSPEC_FMLAL): New unspec.
11410 (UNSPEC_FMLSL): Ditto.
11411 (UNSPEC_FMLAL2): Ditto.
11412 (UNSPEC_FMLSL2): Ditto.
11413 (f16mac): New code attribute.
11414 * config/aarch64/aarch64-simd-builtins.def
11415 (aarch64_fmlal_lowv2sf): Ditto.
11416 (aarch64_fmlsl_lowv2sf): Ditto.
11417 (aarch64_fmlalq_lowv4sf): Ditto.
11418 (aarch64_fmlslq_lowv4sf): Ditto.
11419 (aarch64_fmlal_highv2sf): Ditto.
11420 (aarch64_fmlsl_highv2sf): Ditto.
11421 (aarch64_fmlalq_highv4sf): Ditto.
11422 (aarch64_fmlslq_highv4sf): Ditto.
11423 (aarch64_fmlal_lane_lowv2sf): Ditto.
11424 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11425 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11426 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11427 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11428 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11429 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11430 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11431 (aarch64_fmlal_lane_highv2sf): Ditto.
11432 (aarch64_fmlsl_lane_highv2sf): Ditto.
11433 (aarch64_fmlal_laneq_highv2sf): Ditto.
11434 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11435 (aarch64_fmlalq_lane_highv4sf): Ditto.
11436 (aarch64_fmlsl_lane_highv4sf): Ditto.
11437 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11438 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11439 * config/aarch64/aarch64-simd.md:
11440 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11441 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11442 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11443 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11444 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11445 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11446 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11447 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11448 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11449 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11450 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11451 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11452 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11453 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11454 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11455 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11456 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11457 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11458 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11459 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11460 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11461 (vfmlsl_low_u32): Ditto.
11462 (vfmlalq_low_u32): Ditto.
11463 (vfmlslq_low_u32): Ditto.
11464 (vfmlal_high_u32): Ditto.
11465 (vfmlsl_high_u32): Ditto.
11466 (vfmlalq_high_u32): Ditto.
11467 (vfmlslq_high_u32): Ditto.
11468 (vfmlal_lane_low_u32): Ditto.
11469 (vfmlsl_lane_low_u32): Ditto.
11470 (vfmlal_laneq_low_u32): Ditto.
11471 (vfmlsl_laneq_low_u32): Ditto.
11472 (vfmlalq_lane_low_u32): Ditto.
11473 (vfmlslq_lane_low_u32): Ditto.
11474 (vfmlalq_laneq_low_u32): Ditto.
11475 (vfmlslq_laneq_low_u32): Ditto.
11476 (vfmlal_lane_high_u32): Ditto.
11477 (vfmlsl_lane_high_u32): Ditto.
11478 (vfmlal_laneq_high_u32): Ditto.
11479 (vfmlsl_laneq_high_u32): Ditto.
11480 (vfmlalq_lane_high_u32): Ditto.
11481 (vfmlslq_lane_high_u32): Ditto.
11482 (vfmlalq_laneq_high_u32): Ditto.
11483 (vfmlslq_laneq_high_u32): Ditto.
11484 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11485 (AARCH64_FL_FOR_ARCH8_4): New.
11486 (AARCH64_ISA_F16FML): New ISA flag.
11487 (TARGET_F16FML): New feature flag for fp16fml.
11488 (doc/invoke.texi): Document new fp16fml option.
11490 2018-01-10 Michael Collison <michael.collison@arm.com>
11492 * config/aarch64/aarch64-builtins.c:
11493 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11494 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11495 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11496 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11497 (AARCH64_ISA_SHA3): New ISA flag.
11498 (TARGET_SHA3): New feature flag for sha3.
11499 * config/aarch64/iterators.md (sha512_op): New int attribute.
11500 (CRYPTO_SHA512): New int iterator.
11501 (UNSPEC_SHA512H): New unspec.
11502 (UNSPEC_SHA512H2): Ditto.
11503 (UNSPEC_SHA512SU0): Ditto.
11504 (UNSPEC_SHA512SU1): Ditto.
11505 * config/aarch64/aarch64-simd-builtins.def
11506 (aarch64_crypto_sha512hqv2di): New builtin.
11507 (aarch64_crypto_sha512h2qv2di): Ditto.
11508 (aarch64_crypto_sha512su0qv2di): Ditto.
11509 (aarch64_crypto_sha512su1qv2di): Ditto.
11510 (aarch64_eor3qv8hi): Ditto.
11511 (aarch64_rax1qv2di): Ditto.
11512 (aarch64_xarqv2di): Ditto.
11513 (aarch64_bcaxqv8hi): Ditto.
11514 * config/aarch64/aarch64-simd.md:
11515 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11516 (aarch64_crypto_sha512su0qv2di): Ditto.
11517 (aarch64_crypto_sha512su1qv2di): Ditto.
11518 (aarch64_eor3qv8hi): Ditto.
11519 (aarch64_rax1qv2di): Ditto.
11520 (aarch64_xarqv2di): Ditto.
11521 (aarch64_bcaxqv8hi): Ditto.
11522 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11523 (vsha512h2q_u64): Ditto.
11524 (vsha512su0q_u64): Ditto.
11525 (vsha512su1q_u64): Ditto.
11526 (veor3q_u16): Ditto.
11527 (vrax1q_u64): Ditto.
11528 (vxarq_u64): Ditto.
11529 (vbcaxq_u16): Ditto.
11530 * config/arm/types.md (crypto_sha512): New type attribute.
11531 (crypto_sha3): Ditto.
11532 (doc/invoke.texi): Document new sha3 option.
11534 2018-01-10 Michael Collison <michael.collison@arm.com>
11536 * config/aarch64/aarch64-builtins.c:
11537 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11538 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11539 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11540 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11541 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11542 (AARCH64_ISA_SM4): New ISA flag.
11543 (TARGET_SM4): New feature flag for sm4.
11544 * config/aarch64/aarch64-simd-builtins.def
11545 (aarch64_sm3ss1qv4si): Ditto.
11546 (aarch64_sm3tt1aq4si): Ditto.
11547 (aarch64_sm3tt1bq4si): Ditto.
11548 (aarch64_sm3tt2aq4si): Ditto.
11549 (aarch64_sm3tt2bq4si): Ditto.
11550 (aarch64_sm3partw1qv4si): Ditto.
11551 (aarch64_sm3partw2qv4si): Ditto.
11552 (aarch64_sm4eqv4si): Ditto.
11553 (aarch64_sm4ekeyqv4si): Ditto.
11554 * config/aarch64/aarch64-simd.md:
11555 (aarch64_sm3ss1qv4si): Ditto.
11556 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11557 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11558 (aarch64_sm4eqv4si): Ditto.
11559 (aarch64_sm4ekeyqv4si): Ditto.
11560 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11561 (sm3part_op): Ditto.
11562 (CRYPTO_SM3TT): Ditto.
11563 (CRYPTO_SM3PART): Ditto.
11564 (UNSPEC_SM3SS1): New unspec.
11565 (UNSPEC_SM3TT1A): Ditto.
11566 (UNSPEC_SM3TT1B): Ditto.
11567 (UNSPEC_SM3TT2A): Ditto.
11568 (UNSPEC_SM3TT2B): Ditto.
11569 (UNSPEC_SM3PARTW1): Ditto.
11570 (UNSPEC_SM3PARTW2): Ditto.
11571 (UNSPEC_SM4E): Ditto.
11572 (UNSPEC_SM4EKEY): Ditto.
11573 * config/aarch64/constraints.md (Ui2): New constraint.
11574 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11575 * config/arm/types.md (crypto_sm3): New type attribute.
11576 (crypto_sm4): Ditto.
11577 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11578 (vsm3tt1aq_u32): Ditto.
11579 (vsm3tt1bq_u32): Ditto.
11580 (vsm3tt2aq_u32): Ditto.
11581 (vsm3tt2bq_u32): Ditto.
11582 (vsm3partw1q_u32): Ditto.
11583 (vsm3partw2q_u32): Ditto.
11584 (vsm4eq_u32): Ditto.
11585 (vsm4ekeyq_u32): Ditto.
11586 (doc/invoke.texi): Document new sm4 option.
11588 2018-01-10 Michael Collison <michael.collison@arm.com>
11590 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11591 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11592 (AARCH64_FL_FOR_ARCH8_4): New.
11593 (AARCH64_FL_V8_4): New flag.
11594 (doc/invoke.texi): Document new armv8.4-a option.
11596 2018-01-10 Michael Collison <michael.collison@arm.com>
11598 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11599 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11600 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11601 * config/aarch64/aarch64-option-extension.def: Add
11602 AARCH64_OPT_EXTENSION of 'sha2'.
11603 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11604 (crypto): Disable sha2 and aes if crypto disabled.
11605 (crypto): Enable aes and sha2 if enabled.
11606 (simd): Disable sha2 and aes if simd disabled.
11607 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11609 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11610 (TARGET_SHA2): New feature flag for sha2.
11611 (TARGET_AES): New feature flag for aes.
11612 * config/aarch64/aarch64-simd.md:
11613 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11614 conditional on TARGET_AES.
11615 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11616 (aarch64_crypto_sha1hsi): Make pattern conditional
11618 (aarch64_crypto_sha1hv4si): Ditto.
11619 (aarch64_be_crypto_sha1hv4si): Ditto.
11620 (aarch64_crypto_sha1su1v4si): Ditto.
11621 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11622 (aarch64_crypto_sha1su0v4si): Ditto.
11623 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11624 (aarch64_crypto_sha256su0v4si): Ditto.
11625 (aarch64_crypto_sha256su1v4si): Ditto.
11626 (doc/invoke.texi): Document new aes and sha2 options.
11628 2018-01-10 Martin Sebor <msebor@redhat.com>
11630 PR tree-optimization/83781
11631 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11634 2018-01-11 Martin Sebor <msebor@gmail.com>
11635 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11637 PR tree-optimization/83501
11638 PR tree-optimization/81703
11640 * tree-ssa-strlen.c (get_string_cst): Rename...
11641 (get_string_len): ...to this. Handle global constants.
11642 (handle_char_store): Adjust.
11644 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11645 Jim Wilson <jimw@sifive.com>
11647 * config/riscv/riscv-protos.h (riscv_output_return): New.
11648 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11649 (riscv_attribute_table, riscv_output_return),
11650 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11651 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11652 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11654 (riscv_expand_prologue): Add early return for naked function.
11655 (riscv_expand_epilogue): Likewise.
11656 (riscv_function_ok_for_sibcall): Return false for naked function.
11657 (riscv_set_current_function): New.
11658 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11659 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11660 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11661 * doc/extend.texi (RISC-V Function Attributes): New.
11663 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11665 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11666 check for 128-bit long double before checking TCmode.
11667 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11668 128-bit long doubles before checking TFmode or TCmode.
11669 (FLOAT128_IBM_P): Likewise.
11671 2018-01-10 Martin Sebor <msebor@redhat.com>
11673 PR tree-optimization/83671
11674 * builtins.c (c_strlen): Unconditionally return zero for the empty
11676 Use -Warray-bounds for warnings.
11677 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11678 for non-constant array indices with COMPONENT_REF, arrays of
11679 arrays, and pointers to arrays.
11680 (gimple_fold_builtin_strlen): Determine and set length range for
11681 non-constant character arrays.
11683 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11685 PR middle-end/81897
11686 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11689 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11691 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11693 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11696 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11697 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11698 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11699 indexed_or_indirect_operand predicate.
11700 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11701 (*vsx_le_perm_load_v8hi): Likewise.
11702 (*vsx_le_perm_load_v16qi): Likewise.
11703 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11704 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11705 (*vsx_le_perm_store_v8hi): Likewise.
11706 (*vsx_le_perm_store_v16qi): Likewise.
11707 (eight unnamed splitters): Likewise.
11709 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11711 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11712 * config/rs6000/emmintrin.h: Likewise.
11713 * config/rs6000/mmintrin.h: Likewise.
11714 * config/rs6000/xmmintrin.h: Likewise.
11716 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11719 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11721 * tree.c (tree_nop_conversion): Return true for location wrapper
11723 (maybe_wrap_with_location): New function.
11724 (selftest::check_strip_nops): New function.
11725 (selftest::test_location_wrappers): New function.
11726 (selftest::tree_c_tests): Call it.
11727 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11728 (maybe_wrap_with_location): New decl.
11729 (EXPR_LOCATION_WRAPPER_P): New macro.
11730 (location_wrapper_p): New inline function.
11731 (tree_strip_any_location_wrapper): New inline function.
11733 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11736 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11737 stack_realign_offset for the largest alignment of stack slot
11739 (ix86_find_max_used_stack_alignment): New function.
11740 (ix86_finalize_stack_frame_flags): Use it. Set
11741 max_used_stack_alignment if we don't realign stack.
11742 * config/i386/i386.h (machine_function): Add
11743 max_used_stack_alignment.
11745 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11747 * config/arm/arm.opt (-mbranch-cost): New option.
11748 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11751 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11754 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11755 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11757 2018-01-10 Richard Biener <rguenther@suse.de>
11760 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11761 early out so it also covers the case where we have a non-NULL
11764 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11766 PR tree-optimization/83753
11767 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11768 for non-strided grouped accesses if the number of elements is 1.
11770 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11773 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11774 * i386.h (TARGET_USE_GATHER): Define.
11775 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11777 2018-01-10 Martin Liska <mliska@suse.cz>
11780 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11781 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11783 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11784 CLEANUP_NO_PARTITIONING is not set.
11786 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11788 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11789 for vectors, as a partial revert of r254296.
11790 * rtl.h (const_vec_p): Delete.
11791 (const_vec_duplicate_p): Don't test for vector CONSTs.
11792 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11793 * expmed.c (make_tree): Likewise.
11796 * common.md (E, F): Use CONSTANT_P instead of checking for
11798 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11799 checking for CONST_VECTOR.
11801 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11803 PR middle-end/83575
11804 * predict.c (force_edge_cold): Handle in more sane way edges
11805 with no prediction.
11807 2018-01-09 Carl Love <cel@us.ibm.com>
11809 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11811 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11812 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11813 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11814 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11815 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11816 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11817 * config/rs6000/rs6000-protos.h: Add extern defition for
11818 rs6000_generate_float2_double_code.
11819 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11821 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11822 (float2_v2df): Add define_expand.
11824 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11827 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11828 op_mode in the force_to_mode call.
11830 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11832 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11833 instead of checking each element individually.
11834 (aarch64_evpc_uzp): Likewise.
11835 (aarch64_evpc_zip): Likewise.
11836 (aarch64_evpc_ext): Likewise.
11837 (aarch64_evpc_rev): Likewise.
11838 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11839 instead of checking each element individually. Return true without
11841 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11842 whether all selected elements come from the same input, instead of
11843 checking each element individually. Remove calls to gen_rtx_REG,
11844 start_sequence and end_sequence and instead assert that no rtl is
11847 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11849 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11850 order of HIGH and CONST checks.
11852 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11854 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11855 if the destination isn't an SSA_NAME.
11857 2018-01-09 Richard Biener <rguenther@suse.de>
11859 PR tree-optimization/83668
11860 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11862 (canonicalize_loop_form): ... here, renamed from ...
11863 (canonicalize_loop_closed_ssa_form): ... this and amended to
11864 swap successor edges for loop exit blocks to make us use
11865 the RPO order we need for initial schedule generation.
11867 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11869 PR tree-optimization/64811
11870 * match.pd: When optimizing comparisons with Inf, avoid
11871 introducing or losing exceptions from comparisons with NaN.
11873 2018-01-09 Martin Liska <mliska@suse.cz>
11876 * asan.c (shadow_mem_size): Add gcc_assert.
11878 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11880 Don't save registers in main().
11883 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11884 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11885 * config/avr/avr.c (avr_set_current_function): Don't error if
11886 naked, OS_task or OS_main are specified at the same time.
11887 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11889 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11891 * common/config/avr/avr-common.c (avr_option_optimization_table):
11892 Switch on -mmain-is-OS_task for optimizing compilations.
11894 2018-01-09 Richard Biener <rguenther@suse.de>
11896 PR tree-optimization/83572
11897 * graphite.c: Include cfganal.h.
11898 (graphite_transform_loops): Connect infinite loops to exit
11899 and remove fake edges at the end.
11901 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11903 * ipa-inline.c (edge_badness): Revert accidental checkin.
11905 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11908 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11909 symbols; not inline clones.
11911 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11914 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11915 hard registers. Formatting fixes.
11917 PR preprocessor/83722
11918 * gcc.c (try_generate_repro): Pass
11919 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11920 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11923 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11924 Kito Cheng <kito.cheng@gmail.com>
11926 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11927 (riscv_leaf_function_p): Delete.
11928 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11930 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11932 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11934 (do_ifelse): New function.
11935 (do_isel): New function.
11936 (do_sub3): New function.
11937 (do_add3): New function.
11938 (do_load_mask_compare): New function.
11939 (do_overlap_load_compare): New function.
11940 (expand_compare_loop): New function.
11941 (expand_block_compare): Call expand_compare_loop() when appropriate.
11942 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11943 option description.
11944 (-mblock-compare-inline-loop-limit): New option.
11946 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11949 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11950 Reverse order of second and third operands in first alternative.
11951 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11952 of first and second elements in UNSPEC_VPERMR vector.
11953 (altivec_expand_vec_perm_le): Likewise.
11955 2018-01-08 Jeff Law <law@redhat.com>
11957 PR rtl-optimizatin/81308
11958 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11959 (process_switch): If group_case_labels makes a change, then set
11961 (pass_convert_switch::execute): If a switch is converted, then
11962 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11964 PR rtl-optimization/81308
11965 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11968 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11970 PR target/83663 - Revert r255946
11971 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11972 generation for cases where splatting a value is not useful.
11973 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11974 across a vec_duplicate and a paradoxical subreg forming a vector
11975 mode to a vec_concat.
11977 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11979 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11980 -march=armv8.3-a variants.
11981 * config/arm/t-multilib: Likewise.
11982 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11984 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11986 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11988 (cceq_ior_compare_complement): Give it a name so I can use it, and
11989 change boolean_or_operator predicate to boolean_operator so it can
11990 be used to generate a crand.
11991 (eqne): New code iterator.
11992 (bd/bd_neg): New code_attrs.
11993 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11994 a single define_insn.
11995 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11996 decrement (bdnzt/bdnzf/bdzt/bdzf).
11997 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11998 with the new names of the branch decrement patterns, and added the
11999 names of the branch decrement conditional patterns.
12001 2018-01-08 Richard Biener <rguenther@suse.de>
12003 PR tree-optimization/83563
12004 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12007 2018-01-08 Richard Biener <rguenther@suse.de>
12009 PR middle-end/83713
12010 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12012 2018-01-08 Richard Biener <rguenther@suse.de>
12014 PR tree-optimization/83685
12015 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12016 references to abnormals.
12018 2018-01-08 Richard Biener <rguenther@suse.de>
12021 * dwarf2out.c (output_indirect_strings): Handle empty
12022 skeleton_debug_str_hash.
12023 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12025 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12027 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12028 (emit_store_direct): Likewise.
12029 (arc_trampoline_adjust_address): Likewise.
12030 (arc_asm_trampoline_template): New function.
12031 (arc_initialize_trampoline): Use asm_trampoline_template.
12032 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12033 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12034 * config/arc/arc.md (flush_icache): Delete pattern.
12036 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12038 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12039 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12042 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12045 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12046 by not USED_FOR_TARGET.
12047 (make_pass_resolve_sw_modes): Likewise.
12049 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12051 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12054 2018-01-08 Richard Biener <rguenther@suse.de>
12056 PR middle-end/83580
12057 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12059 2018-01-08 Richard Biener <rguenther@suse.de>
12061 PR middle-end/83517
12062 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12064 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12066 PR middle-end/81897
12067 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12068 basic blocks with a small number of successors.
12069 (convert_control_dep_chain_into_preds): Improve handling of
12071 (dump_predicates): Split apart into...
12072 (dump_pred_chain): ...here...
12073 (dump_pred_info): ...and here.
12074 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12075 (can_chain_union_be_invalidated_p): Improve check for invalidation
12077 (uninit_uses_cannot_happen): Avoid unnecessary if
12078 convert_control_dep_chain_into_preds yielded nothing.
12080 2018-01-06 Martin Sebor <msebor@redhat.com>
12082 PR tree-optimization/83640
12083 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12084 subtracting negative offset from size.
12085 (builtin_access::overlap): Adjust offset bounds of the access to fall
12086 within the size of the object if possible.
12088 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12090 PR rtl-optimization/83699
12091 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12092 extract_bit_field_as_subreg to cases in which the extracted
12093 value is also a vector.
12095 * lra-constraints.c (process_alt_operands): Test for the equivalence
12096 substitutions when detecting a possible reload cycle.
12098 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12101 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12102 by default if flag_selective_schedling{,2}. Formatting fixes.
12104 PR rtl-optimization/83682
12105 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12106 if it has non-VECTOR_MODE element mode.
12107 (vec_duplicate_p): Likewise.
12109 PR middle-end/83694
12110 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12111 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12113 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12116 * config/i386/i386-builtin.def
12117 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12118 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12119 Require also OPTION_MASK_ISA_AVX512F in addition to
12120 OPTION_MASK_ISA_GFNI.
12121 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12122 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12123 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12124 to OPTION_MASK_ISA_GFNI.
12125 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12126 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12127 OPTION_MASK_ISA_AVX512BW.
12128 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12129 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12130 addition to OPTION_MASK_ISA_GFNI.
12131 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12132 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12133 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12134 to OPTION_MASK_ISA_GFNI.
12135 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12136 a requirement for all ISAs rather than any of them with a few
12138 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12140 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12141 bitmasks to be enabled with 3 exceptions, instead of requiring any
12142 enabled ISA with lots of exceptions.
12143 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12144 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12145 Change avx512bw in isa attribute to avx512f.
12146 * config/i386/sgxintrin.h: Add license boilerplate.
12147 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12148 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12149 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12150 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12152 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12153 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12154 temporarily sse2 rather than sse if not enabled already.
12157 * config/i386/sse.md (VI248_VLBW): Rename to ...
12158 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12159 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12160 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12161 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12162 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12163 mode iterator instead of VI248_VLBW.
12165 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12167 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12168 (record_modified): Skip clobbers; add debug output.
12169 (param_change_prob): Use sreal frequencies.
12171 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12173 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12174 punt for user-aligned variables.
12176 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12178 * tree-chrec.c (chrec_contains_symbols): Return true for
12181 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12184 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12185 of (x|y) == x for BICS pattern.
12187 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12189 PR tree-optimization/83605
12190 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12191 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12194 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12196 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12197 * config/epiphany/rtems.h: New file.
12199 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12200 Uros Bizjak <ubizjak@gmail.com>
12203 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12204 QIreg_operand instead of register_operand predicate.
12205 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12206 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12207 comments instead of -fmitigate[-_]rop.
12209 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12212 * cgraphunit.c (symbol_table::compile): Switch to text_section
12213 before calling assembly_start debug hook.
12214 * run-rtl-passes.c (run_rtl_passes): Likewise.
12217 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12219 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12220 range_int_cst_p rather than !symbolic_range_p before calling
12221 extract_range_from_multiplicative_op_1.
12223 2018-01-04 Jeff Law <law@redhat.com>
12225 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12226 redundant test in assertion.
12228 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12230 * doc/rtl.texi: Document machine_mode wrapper classes.
12232 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12234 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12235 using tree_to_uhwi.
12237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12239 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12240 the VEC_PERM_EXPR fold to fail.
12242 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12245 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12246 to switched_sections.
12248 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12251 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12252 test for d.testing.
12254 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12257 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12258 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12260 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12263 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12264 is BLKmode and bitpos not zero or mode change is needed.
12266 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12269 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12272 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12275 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12276 instead of MULT rtx. Update all corresponding splitters.
12277 (*saddl_se): Ditto.
12278 (*ssub<modesuffix>): Ditto.
12279 (*ssubl_se): Ditto.
12280 (*cmp_sadd_di): Update split patterns.
12281 (*cmp_sadd_si): Ditto.
12282 (*cmp_sadd_sidi): Ditto.
12283 (*cmp_ssub_di): Ditto.
12284 (*cmp_ssub_si): Ditto.
12285 (*cmp_ssub_sidi): Ditto.
12286 * config/alpha/predicates.md (const23_operand): New predicate.
12287 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12288 Look for ASHIFT, not MULT inner operand.
12289 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12291 2018-01-04 Martin Liska <mliska@suse.cz>
12293 PR gcov-profile/83669
12294 * gcov.c (output_intermediate_file): Add version to intermediate
12296 * doc/gcov.texi: Document new field 'version' in intermediate
12297 file format. Fix location of '-k' option of gcov command.
12299 2018-01-04 Martin Liska <mliska@suse.cz>
12302 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12304 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12306 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12308 2018-01-03 Martin Sebor <msebor@redhat.com>
12310 PR tree-optimization/83655
12311 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12312 checking calls with invalid arguments.
12314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12316 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12317 (vectorizable_mask_load_store): Delete.
12318 (vectorizable_call): Return false for masked loads and stores.
12319 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12320 instead of gimple_assign_rhs1.
12321 (vectorizable_load): Handle IFN_MASK_LOAD.
12322 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12326 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12328 (vectorizable_mask_load_store): ...here.
12329 (vectorizable_load): ...and here.
12331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12333 * tree-vect-stmts.c (vect_build_all_ones_mask)
12334 (vect_build_zero_merge_argument): New functions, split out from...
12335 (vectorizable_load): ...here.
12337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12339 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12341 (vectorizable_mask_load_store): ...here.
12342 (vectorizable_store): ...and here.
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12346 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12348 (vectorizable_mask_load_store): ...here.
12350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12353 (vect_model_store_cost): Take a vec_load_store_type instead of a
12355 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12356 (vect_model_store_cost): Take a vec_load_store_type instead of a
12358 (vectorizable_mask_load_store): Update accordingly.
12359 (vectorizable_store): Likewise.
12360 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12364 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12365 IFN_MASK_LOAD calls here rather than...
12366 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12369 Alan Hayward <alan.hayward@arm.com>
12370 David Sherwood <david.sherwood@arm.com>
12372 * expmed.c (extract_bit_field_1): For vector extracts,
12373 fall back to extract_bit_field_as_subreg if vec_extract
12376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12377 Alan Hayward <alan.hayward@arm.com>
12378 David Sherwood <david.sherwood@arm.com>
12380 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12381 they are variable or constant sized.
12382 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12383 slots for constant-sized data.
12385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12386 Alan Hayward <alan.hayward@arm.com>
12387 David Sherwood <david.sherwood@arm.com>
12389 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12390 handling COND_EXPRs with boolean comparisons, try to find a better
12391 basis for the mask type than the boolean itself.
12393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12395 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12396 is calculated and how it can be overridden.
12397 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12398 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12400 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12404 Alan Hayward <alan.hayward@arm.com>
12405 David Sherwood <david.sherwood@arm.com>
12407 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12408 Remove the mode argument.
12409 (aarch64_simd_valid_immediate): Remove the mode and inverse
12411 * config/aarch64/iterators.md (bitsize): New iterator.
12412 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12413 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12414 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12415 aarch64_simd_valid_immediate.
12416 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12417 (aarch64_reg_or_bic_imm): Likewise.
12418 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12419 with an insn_type enum and msl with a modifier_type enum.
12420 Replace element_width with a scalar_mode. Change the shift
12421 to unsigned int. Add constructors for scalar_float_mode and
12422 scalar_int_mode elements.
12423 (aarch64_vect_float_const_representable_p): Delete.
12424 (aarch64_can_const_movi_rtx_p)
12425 (aarch64_simd_scalar_immediate_valid_for_move)
12426 (aarch64_simd_make_constant): Update call to
12427 aarch64_simd_valid_immediate.
12428 (aarch64_advsimd_valid_immediate_hs): New function.
12429 (aarch64_advsimd_valid_immediate): Likewise.
12430 (aarch64_simd_valid_immediate): Remove mode and inverse
12431 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12432 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12433 and aarch64_float_const_representable_p on the result.
12434 (aarch64_output_simd_mov_immediate): Remove mode argument.
12435 Update call to aarch64_simd_valid_immediate and use of
12436 simd_immediate_info.
12437 (aarch64_output_scalar_simd_mov_immediate): Update call
12440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12441 Alan Hayward <alan.hayward@arm.com>
12442 David Sherwood <david.sherwood@arm.com>
12444 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12445 (mode_nunits): Likewise CONST_MODE_NUNITS.
12446 * machmode.def (ADJUST_NUNITS): Document.
12447 * genmodes.c (mode_data::need_nunits_adj): New field.
12448 (blank_mode): Update accordingly.
12449 (adj_nunits): New variable.
12450 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12452 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12453 listed in adj_nunits.
12454 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12455 listed in adj_nunits. Don't emit case statements for such modes.
12456 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12457 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12458 nothing if adj_nunits is nonnull.
12459 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12460 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12461 (emit_mode_fbit): Update use of print_maybe_const_decl.
12462 (emit_move_size): Likewise. Treat the array as non-const
12464 (emit_mode_adjustments): Handle adj_nunits.
12466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12468 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12469 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12470 (VECTOR_MODES): Use it.
12471 (make_vector_modes): Take the prefix as an argument.
12473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12474 Alan Hayward <alan.hayward@arm.com>
12475 David Sherwood <david.sherwood@arm.com>
12477 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12478 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12479 for MODE_VECTOR_BOOL.
12480 * machmode.def (VECTOR_BOOL_MODE): Document.
12481 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12482 (make_vector_bool_mode): New function.
12483 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12485 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12486 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12488 * stor-layout.c (int_mode_for_mode): Likewise.
12489 * tree.c (build_vector_type_for_mode): Likewise.
12490 * varasm.c (output_constant_pool_2): Likewise.
12491 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12492 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12493 for MODE_VECTOR_BOOL.
12494 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12495 of mode class checks.
12496 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12497 instead of a list of mode class checks.
12498 (expand_vector_scalar_condition): Likewise.
12499 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12502 Alan Hayward <alan.hayward@arm.com>
12503 David Sherwood <david.sherwood@arm.com>
12505 * machmode.h (mode_size): Change from unsigned short to
12507 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12508 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12509 or if measurement_type is not polynomial.
12510 (fixed_size_mode::includes_p): Check for constant-sized modes.
12511 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12512 return a poly_uint16 rather than an unsigned short.
12513 (emit_mode_size): Change the type of mode_size from unsigned short
12514 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12515 (emit_mode_adjustments): Cope with polynomial vector sizes.
12516 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12518 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12520 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12521 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12522 * caller-save.c (setup_save_areas): Likewise.
12523 (replace_reg_with_saved_mem): Likewise.
12524 * calls.c (emit_library_call_value_1): Likewise.
12525 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12526 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12527 (gen_lowpart_for_combine): Likewise.
12528 * convert.c (convert_to_integer_1): Likewise.
12529 * cse.c (equiv_constant, cse_insn): Likewise.
12530 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12531 (cselib_subst_to_values): Likewise.
12532 * dce.c (word_dce_process_block): Likewise.
12533 * df-problems.c (df_word_lr_mark_ref): Likewise.
12534 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12535 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12536 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12537 (rtl_for_decl_location): Likewise.
12538 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12539 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12540 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12541 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12542 (expand_expr_real_1): Likewise.
12543 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12544 (pad_below): Likewise.
12545 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12546 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12547 * ira.c (get_subreg_tracking_sizes): Likewise.
12548 * ira-build.c (ira_create_allocno_objects): Likewise.
12549 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12550 (ira_sort_regnos_for_alter_reg): Likewise.
12551 * ira-costs.c (record_operand_costs): Likewise.
12552 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12553 (resolve_simple_move): Likewise.
12554 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12555 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12556 (lra_constraints): Likewise.
12557 (CONST_POOL_OK_P): Reject variable-sized modes.
12558 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12559 (add_pseudo_to_slot, lra_spill): Likewise.
12560 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12561 * optabs-query.c (get_best_extraction_insn): Likewise.
12562 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12563 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12564 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12565 * recog.c (offsettable_address_addr_space_p): Likewise.
12566 * regcprop.c (maybe_mode_change): Likewise.
12567 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12568 * regrename.c (build_def_use): Likewise.
12569 * regstat.c (dump_reg_info): Likewise.
12570 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12571 (find_reloads, find_reloads_subreg_address): Likewise.
12572 * reload1.c (eliminate_regs_1): Likewise.
12573 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12574 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12575 (simplify_binary_operation_1, simplify_subreg): Likewise.
12576 * targhooks.c (default_function_arg_padding): Likewise.
12577 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12578 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12579 (verify_gimple_assign_ternary): Likewise.
12580 * tree-inline.c (estimate_move_cost): Likewise.
12581 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12582 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12583 (get_address_cost_ainc): Likewise.
12584 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12585 (vect_supportable_dr_alignment): Likewise.
12586 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12587 (vectorizable_reduction): Likewise.
12588 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12589 (vectorizable_operation, vectorizable_load): Likewise.
12590 * tree.c (build_same_sized_truth_vector_type): Likewise.
12591 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12592 * var-tracking.c (emit_note_insn_var_location): Likewise.
12593 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12594 (ADDR_VEC_ALIGN): Likewise.
12596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12597 Alan Hayward <alan.hayward@arm.com>
12598 David Sherwood <david.sherwood@arm.com>
12600 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12602 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12603 or if measurement_type is polynomial.
12604 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12605 * combine.c (make_extraction): Likewise.
12606 * dse.c (find_shift_sequence): Likewise.
12607 * dwarf2out.c (mem_loc_descriptor): Likewise.
12608 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12609 (extract_bit_field, extract_low_bits): Likewise.
12610 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12611 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12612 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12613 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12614 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12615 * reload.c (find_reloads): Likewise.
12616 * reload1.c (alter_reg): Likewise.
12617 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12618 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12619 * tree-if-conv.c (predicate_mem_writes): Likewise.
12620 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12621 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12622 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12623 * valtrack.c (dead_debug_insert_temp): Likewise.
12624 * varasm.c (mergeable_constant_section): Likewise.
12625 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12628 Alan Hayward <alan.hayward@arm.com>
12629 David Sherwood <david.sherwood@arm.com>
12631 * expr.c (expand_assignment): Cope with polynomial mode sizes
12632 when assigning to a CONCAT.
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12635 Alan Hayward <alan.hayward@arm.com>
12636 David Sherwood <david.sherwood@arm.com>
12638 * machmode.h (mode_precision): Change from unsigned short to
12640 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12642 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12643 or if measurement_type is not polynomial.
12644 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12645 in which the mode is already known to be a scalar_int_mode.
12646 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12647 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12649 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12650 for GET_MODE_PRECISION.
12651 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12652 for GET_MODE_PRECISION.
12653 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12655 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12656 (expand_field_assignment, make_extraction): Likewise.
12657 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12658 (get_last_value): Likewise.
12659 * convert.c (convert_to_integer_1): Likewise.
12660 * cse.c (cse_insn): Likewise.
12661 * expr.c (expand_expr_real_1): Likewise.
12662 * lra-constraints.c (simplify_operand_subreg): Likewise.
12663 * optabs-query.c (can_atomic_load_p): Likewise.
12664 * optabs.c (expand_atomic_load): Likewise.
12665 (expand_atomic_store): Likewise.
12666 * ree.c (combine_reaching_defs): Likewise.
12667 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12668 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12669 * tree.h (type_has_mode_precision_p): Likewise.
12670 * ubsan.c (instrument_si_overflow): Likewise.
12672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12673 Alan Hayward <alan.hayward@arm.com>
12674 David Sherwood <david.sherwood@arm.com>
12676 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12677 polynomial numbers of units.
12678 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12679 (valid_vector_subparts_p): New function.
12680 (build_vector_type): Remove temporary shim and take the number
12681 of units as a poly_uint64 rather than an int.
12682 (build_opaque_vector_type): Take the number of units as a
12683 poly_uint64 rather than an int.
12684 * tree.c (build_vector_from_ctor): Handle polynomial
12685 TYPE_VECTOR_SUBPARTS.
12686 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12687 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12688 (build_vector_from_val): If the number of units is variable,
12689 use build_vec_duplicate_cst for constant operands and
12690 VEC_DUPLICATE_EXPR otherwise.
12691 (make_vector_type): Remove temporary is_constant ().
12692 (build_vector_type, build_opaque_vector_type): Take the number of
12693 units as a poly_uint64 rather than an int.
12694 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12696 * cfgexpand.c (expand_debug_expr): Likewise.
12697 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12698 (store_constructor, expand_expr_real_1): Likewise.
12699 (const_scalar_mask_from_tree): Likewise.
12700 * fold-const-call.c (fold_const_reduction): Likewise.
12701 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12702 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12703 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12704 (fold_relational_const): Likewise.
12705 (native_interpret_vector): Likewise. Change the size from an
12706 int to an unsigned int.
12707 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12708 TYPE_VECTOR_SUBPARTS.
12709 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12710 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12711 duplicating a non-constant operand into a variable-length vector.
12712 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12713 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12714 * ipa-icf.c (sem_variable::equals): Likewise.
12715 * match.pd: Likewise.
12716 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12717 * print-tree.c (print_node): Likewise.
12718 * stor-layout.c (layout_type): Likewise.
12719 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12720 * tree-cfg.c (verify_gimple_comparison): Likewise.
12721 (verify_gimple_assign_binary): Likewise.
12722 (verify_gimple_assign_ternary): Likewise.
12723 (verify_gimple_assign_single): Likewise.
12724 * tree-pretty-print.c (dump_generic_node): Likewise.
12725 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12726 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12727 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12728 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12729 (vect_shift_permute_load_chain): Likewise.
12730 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12731 (expand_vector_condition, optimize_vector_constructor): Likewise.
12732 (lower_vec_perm, get_compute_type): Likewise.
12733 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12734 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12735 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12736 (vect_recog_mask_conversion_pattern): Likewise.
12737 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12738 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12739 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12740 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12741 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12742 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12743 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12744 (supportable_widening_operation): Likewise.
12745 (supportable_narrowing_operation): Likewise.
12746 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12748 * varasm.c (output_constant): Likewise.
12750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12751 Alan Hayward <alan.hayward@arm.com>
12752 David Sherwood <david.sherwood@arm.com>
12754 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12755 so that both the length == 3 and length != 3 cases set up their
12756 own permute vectors. Add comments explaining why we know the
12757 number of elements is constant.
12758 (vect_permute_load_chain): 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 * machmode.h (mode_nunits): Change from unsigned char to
12766 (ONLY_FIXED_SIZE_MODES): New macro.
12767 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12768 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12769 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12771 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12772 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12773 or if measurement_type is not polynomial.
12774 * genmodes.c (ZERO_COEFFS): New macro.
12775 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12777 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12778 Use ZERO_COEFFS when emitting initializers.
12779 * data-streamer.h (bp_pack_poly_value): New function.
12780 (bp_unpack_poly_value): Likewise.
12781 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12782 for GET_MODE_NUNITS.
12783 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12784 for GET_MODE_NUNITS.
12785 * tree.c (make_vector_type): Remove temporary shim and make
12786 the real function take the number of units as a poly_uint64
12787 rather than an int.
12788 (build_vector_type_for_mode): Handle polynomial nunits.
12789 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12790 * emit-rtl.c (const_vec_series_p_1): Likewise.
12791 (gen_rtx_CONST_VECTOR): Likewise.
12792 * fold-const.c (test_vec_duplicate_folding): Likewise.
12793 * genrecog.c (validate_pattern): Likewise.
12794 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12795 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12796 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12797 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12798 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12799 * rtlanal.c (subreg_get_info): Likewise.
12800 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12801 (vect_grouped_load_supported): Likewise.
12802 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12803 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12804 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12805 (simplify_const_unary_operation, simplify_binary_operation_1)
12806 (simplify_const_binary_operation, simplify_ternary_operation)
12807 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12808 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12809 instead of CONST_VECTOR_NUNITS.
12810 * varasm.c (output_constant_pool_2): Likewise.
12811 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12812 explicit-encoded elements in the XVEC for variable-length vectors.
12814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12816 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12819 Alan Hayward <alan.hayward@arm.com>
12820 David Sherwood <david.sherwood@arm.com>
12822 * coretypes.h (fixed_size_mode): Declare.
12823 (fixed_size_mode_pod): New typedef.
12824 * builtins.h (target_builtins::x_apply_args_mode)
12825 (target_builtins::x_apply_result_mode): Change type to
12826 fixed_size_mode_pod.
12827 * builtins.c (apply_args_size, apply_result_size, result_vector)
12828 (expand_builtin_apply_args_1, expand_builtin_apply)
12829 (expand_builtin_return): Update accordingly.
12831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12833 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12834 * cselib.c (cselib_hash_rtx): Likewise.
12835 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12836 CONST_VECTOR encoding.
12838 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12839 Jeff Law <law@redhat.com>
12842 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12843 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12844 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12845 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12848 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12849 explicitly probe *sp in a noreturn function if there were any callee
12850 register saves or frame pointer is needed.
12852 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12855 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12856 BLKmode for ternary, binary or unary expressions.
12859 * var-tracking.c (delete_vta_debug_insn): New inline function.
12860 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12861 insns from get_insns () to NULL instead of each bb separately.
12862 Use delete_vta_debug_insn. No longer static.
12863 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12864 delete_vta_debug_insns callers.
12865 * rtl.h (delete_vta_debug_insns): Declare.
12866 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12867 instead of variable_tracking_main.
12869 2018-01-03 Martin Sebor <msebor@redhat.com>
12871 PR tree-optimization/83603
12872 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12873 arguments past the endof the argument list in functions declared
12874 without a prototype.
12875 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12876 Avoid checking when arguments are null.
12878 2018-01-03 Martin Sebor <msebor@redhat.com>
12881 * doc/extend.texi (attribute const): Fix a typo.
12882 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12883 issuing -Wsuggest-attribute for void functions.
12885 2018-01-03 Martin Sebor <msebor@redhat.com>
12887 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12888 offset_int::from instead of wide_int::to_shwi.
12889 (maybe_diag_overlap): Remove assertion.
12890 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12891 * gimple-ssa-sprintf.c (format_directive): Same.
12892 (parse_directive): Same.
12893 (sprintf_dom_walker::compute_format_length): Same.
12894 (try_substitute_return_value): Same.
12896 2018-01-03 Jeff Law <law@redhat.com>
12898 PR middle-end/83654
12899 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12900 non-constant residual for zero at runtime and avoid probing in
12901 that case. Reorganize code for trailing problem to mirror handling
12904 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12906 PR tree-optimization/83501
12907 * tree-ssa-strlen.c (get_string_cst): New.
12908 (handle_char_store): Call get_string_cst.
12910 2018-01-03 Martin Liska <mliska@suse.cz>
12912 PR tree-optimization/83593
12913 * tree-ssa-strlen.c: Include tree-cfg.h.
12914 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12915 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12916 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12918 (strlen_dom_walker::before_dom_children): Call
12919 gimple_purge_dead_eh_edges. Dump tranformation with details
12921 (strlen_dom_walker::before_dom_children): Update call by adding
12922 new argument cleanup_eh.
12923 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12925 2018-01-03 Martin Liska <mliska@suse.cz>
12928 * cif-code.def (VARIADIC_THUNK): New enum value.
12929 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12932 2018-01-03 Jan Beulich <jbeulich@suse.com>
12934 * sse.md (mov<mode>_internal): Tighten condition for when to use
12935 vmovdqu<ssescalarsize> for TI and OI modes.
12937 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12939 Update copyright years.
12941 2018-01-03 Martin Liska <mliska@suse.cz>
12944 * ipa-visibility.c (function_and_variable_visibility): Skip
12945 functions with noipa attribure.
12947 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12949 * gcc.c (process_command): Update copyright notice dates.
12950 * gcov-dump.c (print_version): Ditto.
12951 * gcov.c (print_version): Ditto.
12952 * gcov-tool.c (print_version): Ditto.
12953 * gengtype.c (create_file): Ditto.
12954 * doc/cpp.texi: Bump @copying's copyright year.
12955 * doc/cppinternals.texi: Ditto.
12956 * doc/gcc.texi: Ditto.
12957 * doc/gccint.texi: Ditto.
12958 * doc/gcov.texi: Ditto.
12959 * doc/install.texi: Ditto.
12960 * doc/invoke.texi: Ditto.
12962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12964 * vector-builder.h (vector_builder::m_full_nelts): Change from
12965 unsigned int to poly_uint64.
12966 (vector_builder::full_nelts): Update prototype accordingly.
12967 (vector_builder::new_vector): Likewise.
12968 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12969 (vector_builder::operator ==): Likewise.
12970 (vector_builder::finalize): Likewise.
12971 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12972 Take the number of elements as a poly_uint64 rather than an
12974 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12975 from unsigned int to poly_uint64.
12976 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12977 (vec_perm_indices::new_vector): Likewise.
12978 (vec_perm_indices::length): Likewise.
12979 (vec_perm_indices::nelts_per_input): Likewise.
12980 (vec_perm_indices::input_nelts): Likewise.
12981 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12982 number of elements per input as a poly_uint64 rather than an
12983 unsigned int. Use the original encoding for variable-length
12984 vectors, rather than clamping each individual element.
12985 For the second and subsequent elements in each pattern,
12986 clamp the step and base before clamping their sum.
12987 (vec_perm_indices::series_p): Handle polynomial element counts.
12988 (vec_perm_indices::all_in_range_p): Likewise.
12989 (vec_perm_indices_to_tree): Likewise.
12990 (vec_perm_indices_to_rtx): Likewise.
12991 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12992 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12993 (tree_vector_builder::new_binary_operation): Handle polynomial
12994 element counts. Return false if we need to know the number
12995 of elements at compile time.
12996 * fold-const.c (fold_vec_perm): Punt if the number of elements
12997 isn't known at compile time.
12999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13001 * vec-perm-indices.h (vec_perm_builder): Change element type
13002 from HOST_WIDE_INT to poly_int64.
13003 (vec_perm_indices::element_type): Update accordingly.
13004 (vec_perm_indices::clamp): Handle polynomial element_types.
13005 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13006 (vec_perm_indices::all_in_range_p): Likewise.
13007 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13009 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13010 polynomial vec_perm_indices element types.
13011 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13012 * fold-const.c (fold_vec_perm): Likewise.
13013 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13014 * tree-vect-generic.c (lower_vec_perm): Likewise.
13015 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13016 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13017 element type to HOST_WIDE_INT.
13019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13020 Alan Hayward <alan.hayward@arm.com>
13021 David Sherwood <david.sherwood@arm.com>
13023 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13024 rather than an int. Use plus_constant.
13025 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13026 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13029 Alan Hayward <alan.hayward@arm.com>
13030 David Sherwood <david.sherwood@arm.com>
13032 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13033 a HOST_WIDE_INT to a poly_int64.
13035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13036 Alan Hayward <alan.hayward@arm.com>
13037 David Sherwood <david.sherwood@arm.com>
13039 * calls.c (load_register_parameters): Cope with polynomial
13040 mode sizes. Require a constant size for BLKmode parameters
13041 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13042 forces a parameter to be padded at the lsb end in order to
13043 fill a complete number of words, require the parameter size
13044 to be ordered wrt UNITS_PER_WORD.
13046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13047 Alan Hayward <alan.hayward@arm.com>
13048 David Sherwood <david.sherwood@arm.com>
13050 * reload1.c (spill_stack_slot_width): Change element type
13051 from unsigned int to poly_uint64_pod.
13052 (alter_reg): Treat mode sizes as polynomial.
13054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13055 Alan Hayward <alan.hayward@arm.com>
13056 David Sherwood <david.sherwood@arm.com>
13058 * reload.c (complex_word_subreg_p): New function.
13059 (reload_inner_reg_of_subreg, push_reload): Use it.
13061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13062 Alan Hayward <alan.hayward@arm.com>
13063 David Sherwood <david.sherwood@arm.com>
13065 * lra-constraints.c (process_alt_operands): Reject matched
13066 operands whose sizes aren't ordered.
13067 (match_reload): Refer to this check here.
13069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13070 Alan Hayward <alan.hayward@arm.com>
13071 David Sherwood <david.sherwood@arm.com>
13073 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13074 that the mode size is in the set {1, 2, 4, 8, 16}.
13076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13077 Alan Hayward <alan.hayward@arm.com>
13078 David Sherwood <david.sherwood@arm.com>
13080 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13081 Use plus_constant instead of gen_rtx_PLUS.
13083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13084 Alan Hayward <alan.hayward@arm.com>
13085 David Sherwood <david.sherwood@arm.com>
13087 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13088 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13089 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13090 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13091 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13092 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13093 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13094 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13095 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13096 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13098 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13099 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13100 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13101 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13102 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13103 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13104 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13105 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13106 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13107 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13109 * expr.c (emit_move_resolve_push): Treat the input and result
13110 of PUSH_ROUNDING as a poly_int64.
13111 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13112 (emit_push_insn): Likewise.
13113 * lra-eliminations.c (mark_not_eliminable): Likewise.
13114 * recog.c (push_operand): Likewise.
13115 * reload1.c (elimination_effects): Likewise.
13116 * rtlanal.c (nonzero_bits1): Likewise.
13117 * calls.c (store_one_arg): Likewise. Require the padding to be
13118 known at compile time.
13120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13121 Alan Hayward <alan.hayward@arm.com>
13122 David Sherwood <david.sherwood@arm.com>
13124 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13125 Use plus_constant instead of gen_rtx_PLUS.
13127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13128 Alan Hayward <alan.hayward@arm.com>
13129 David Sherwood <david.sherwood@arm.com>
13131 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13132 rather than an int.
13134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13135 Alan Hayward <alan.hayward@arm.com>
13136 David Sherwood <david.sherwood@arm.com>
13138 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13139 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13140 via stack temporaries. Treat the mode size as polynomial too.
13142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13143 Alan Hayward <alan.hayward@arm.com>
13144 David Sherwood <david.sherwood@arm.com>
13146 * expr.c (expand_expr_real_2): When handling conversions involving
13147 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13148 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13149 as a poly_uint64 too.
13151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13152 Alan Hayward <alan.hayward@arm.com>
13153 David Sherwood <david.sherwood@arm.com>
13155 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13158 Alan Hayward <alan.hayward@arm.com>
13159 David Sherwood <david.sherwood@arm.com>
13161 * combine.c (can_change_dest_mode): Handle polynomial
13162 REGMODE_NATURAL_SIZE.
13163 * expmed.c (store_bit_field_1): Likewise.
13164 * expr.c (store_constructor): Likewise.
13165 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13166 and polynomial REGMODE_NATURAL_SIZE.
13167 (gen_lowpart_common): Likewise.
13168 * reginfo.c (record_subregs_of_mode): Likewise.
13169 * rtlanal.c (read_modify_subreg_p): Likewise.
13171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13172 Alan Hayward <alan.hayward@arm.com>
13173 David Sherwood <david.sherwood@arm.com>
13175 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13176 numbers of elements.
13178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13179 Alan Hayward <alan.hayward@arm.com>
13180 David Sherwood <david.sherwood@arm.com>
13182 * match.pd: Cope with polynomial numbers of vector elements.
13184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13185 Alan Hayward <alan.hayward@arm.com>
13186 David Sherwood <david.sherwood@arm.com>
13188 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13189 in a POINTER_PLUS_EXPR.
13191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13192 Alan Hayward <alan.hayward@arm.com>
13193 David Sherwood <david.sherwood@arm.com>
13195 * omp-simd-clone.c (simd_clone_subparts): New function.
13196 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13197 (ipa_simd_modify_function_body): Likewise.
13199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13200 Alan Hayward <alan.hayward@arm.com>
13201 David Sherwood <david.sherwood@arm.com>
13203 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13204 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13205 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13206 (expand_vector_condition, vector_element): Likewise.
13207 (subparts_gt): New function.
13208 (get_compute_type): Use subparts_gt.
13209 (count_type_subparts): Delete.
13210 (expand_vector_operations_1): Use subparts_gt instead of
13211 count_type_subparts.
13213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13214 Alan Hayward <alan.hayward@arm.com>
13215 David Sherwood <david.sherwood@arm.com>
13217 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13218 (vect_compile_time_alias): ...this new function. Do the calculation
13219 on poly_ints rather than trees.
13220 (vect_prune_runtime_alias_test_list): Update call accordingly.
13222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13223 Alan Hayward <alan.hayward@arm.com>
13224 David Sherwood <david.sherwood@arm.com>
13226 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13228 (vect_schedule_slp_instance): Likewise.
13230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13231 Alan Hayward <alan.hayward@arm.com>
13232 David Sherwood <david.sherwood@arm.com>
13234 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13235 constant and extern definitions for variable-length vectors.
13236 (vect_get_constant_vectors): Note that the number of units
13237 is known to be constant.
13239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13240 Alan Hayward <alan.hayward@arm.com>
13241 David Sherwood <david.sherwood@arm.com>
13243 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13244 of units as polynomial. Choose between WIDE and NARROW based
13247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13248 Alan Hayward <alan.hayward@arm.com>
13249 David Sherwood <david.sherwood@arm.com>
13251 * tree-vect-stmts.c (simd_clone_subparts): New function.
13252 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13255 Alan Hayward <alan.hayward@arm.com>
13256 David Sherwood <david.sherwood@arm.com>
13258 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13259 vectors as polynomial. Use build_index_vector for
13260 IFN_GOMP_SIMD_LANE.
13262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13263 Alan Hayward <alan.hayward@arm.com>
13264 David Sherwood <david.sherwood@arm.com>
13266 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13267 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13268 for variable-length vectors.
13269 (vectorizable_mask_load_store): Treat the number of units as
13270 polynomial, asserting that it is constant if the condition has
13271 already been enforced.
13272 (vectorizable_store, vectorizable_load): Likewise.
13274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13275 Alan Hayward <alan.hayward@arm.com>
13276 David Sherwood <david.sherwood@arm.com>
13278 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13279 of units as polynomial. Punt if we can't tell at compile time
13280 which vector contains the final result.
13282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13283 Alan Hayward <alan.hayward@arm.com>
13284 David Sherwood <david.sherwood@arm.com>
13286 * tree-vect-loop.c (vectorizable_induction): Treat the number
13287 of units as polynomial. Punt on SLP inductions. Use an integer
13288 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13289 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
13297 * tree.c (build_index_vector): New function.
13298 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13299 of units as polynomial, forcibly converting it to a constant if
13300 vectorizable_reduction has already enforced the condition.
13301 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13302 to create a {1,2,3,...} vector.
13303 (vectorizable_reduction): Treat the number of units as polynomial.
13304 Choose vectype_in based on the largest scalar element size rather
13305 than the smallest number of units. Enforce the restrictions
13308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13309 Alan Hayward <alan.hayward@arm.com>
13310 David Sherwood <david.sherwood@arm.com>
13312 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13313 number of units as polynomial.
13315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13316 Alan Hayward <alan.hayward@arm.com>
13317 David Sherwood <david.sherwood@arm.com>
13319 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13320 * target.def (autovectorize_vector_sizes): Return the vector sizes
13321 by pointer, using vector_sizes rather than a bitmask.
13322 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13323 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13324 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13326 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13327 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13328 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13329 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13330 * omp-general.c (omp_max_vf): Likewise.
13331 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13332 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13333 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13334 * tree-vect-slp.c (vect_slp_bb): Likewise.
13335 * doc/tm.texi: Regenerate.
13336 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13338 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13339 the vector size as a poly_uint64 rather than an unsigned int.
13340 (current_vector_size): Change from an unsigned int to a poly_uint64.
13341 (get_vectype_for_scalar_type): Update accordingly.
13342 * tree.h (build_truth_vector_type): Take the size and number of
13343 units as a poly_uint64 rather than an unsigned int.
13344 (build_vector_type): Add a temporary overload that takes
13345 the number of units as a poly_uint64 rather than an unsigned int.
13346 * tree.c (make_vector_type): Likewise.
13347 (build_truth_vector_type): Take the number of units as a poly_uint64
13348 rather than an unsigned int.
13350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13351 Alan Hayward <alan.hayward@arm.com>
13352 David Sherwood <david.sherwood@arm.com>
13354 * target.def (get_mask_mode): Take the number of units and length
13355 as poly_uint64s rather than unsigned ints.
13356 * targhooks.h (default_get_mask_mode): Update accordingly.
13357 * targhooks.c (default_get_mask_mode): Likewise.
13358 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13359 * doc/tm.texi: Regenerate.
13361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13362 Alan Hayward <alan.hayward@arm.com>
13363 David Sherwood <david.sherwood@arm.com>
13365 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13366 * omp-general.c (omp_max_vf): Likewise.
13367 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13368 (expand_omp_simd): Handle polynomial safelen.
13369 * omp-low.c (omplow_simd_context): Add a default constructor.
13370 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13371 (lower_rec_simd_input_clauses): Update accordingly.
13372 (lower_rec_input_clauses): Likewise.
13374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13375 Alan Hayward <alan.hayward@arm.com>
13376 David Sherwood <david.sherwood@arm.com>
13378 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13379 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13380 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13381 (vect_analyze_slp_cost): Likewise.
13382 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13383 (vect_model_load_cost): Likewise.
13385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13386 Alan Hayward <alan.hayward@arm.com>
13387 David Sherwood <david.sherwood@arm.com>
13389 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13390 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13391 from an unsigned int * to a poly_uint64_pod *.
13392 (calculate_unrolling_factor): New function.
13393 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
13400 from an unsigned int to a poly_uint64.
13401 (_loop_vec_info::slp_unrolling_factor): Likewise.
13402 (_loop_vec_info::vectorization_factor): Change from an int
13404 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13405 (vect_get_num_vectors): New function.
13406 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13407 (vect_get_num_copies): Use vect_get_num_vectors.
13408 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13409 to an unsigned int *.
13410 (vect_analyze_data_refs): Change min_vf from an int * to a
13412 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13413 than an unsigned HOST_WIDE_INT.
13414 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13415 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13416 to an unsigned int *.
13417 (vect_analyze_data_ref_dependences): Likewise.
13418 (vect_compute_data_ref_alignment): Handle polynomial vf.
13419 (vect_enhance_data_refs_alignment): Likewise.
13420 (vect_prune_runtime_alias_test_list): Likewise.
13421 (vect_shift_permute_load_chain): Likewise.
13422 (vect_supportable_dr_alignment): Likewise.
13423 (dependence_distance_ge_vf): Take the vectorization factor as a
13424 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13425 (vect_analyze_data_refs): Change min_vf from an int * to a
13427 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13428 vfm1 as a poly_uint64 rather than an int. Make the same change
13429 for the returned bound_scalar.
13430 (vect_gen_vector_loop_niters): Handle polynomial vf.
13431 (vect_do_peeling): Likewise. Update call to
13432 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13433 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13435 * tree-vect-loop.c (vect_determine_vectorization_factor)
13436 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13437 (vect_get_known_peeling_cost): Likewise.
13438 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13439 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13440 (vect_transform_loop): Likewise. Use the lowest possible VF when
13441 updating the upper bounds of the loop.
13442 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13443 rather than an int.
13444 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13445 polynomial unroll factors.
13446 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13447 (vect_make_slp_decision): Likewise.
13448 (vect_supported_load_permutation_p): Likewise, and polynomial
13450 (vect_analyze_slp_cost): Handle polynomial vf.
13451 (vect_slp_analyze_node_operations): Likewise.
13452 (vect_slp_analyze_bb_1): Likewise.
13453 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13454 than an unsigned HOST_WIDE_INT.
13455 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13456 (vectorizable_load): Handle polynomial vf.
13457 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13459 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13462 Alan Hayward <alan.hayward@arm.com>
13463 David Sherwood <david.sherwood@arm.com>
13465 * match.pd: Handle bit operations involving three constants
13466 and try to fold one pair.
13468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13470 * tree-vect-loop-manip.c: Include gimple-fold.h.
13471 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13472 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13473 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13474 Add a path that uses a step of VF instead of 1, but disable it
13476 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13477 and niters_no_overflow parameters. Update calls to
13478 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13479 Create a new SSA name if the latter choses to use a ste other
13480 than zero, and return it via niters_vector_mult_vf_var.
13481 * tree-vect-loop.c (vect_transform_loop): Update calls to
13482 vect_do_peeling, vect_gen_vector_loop_niters and
13483 slpeel_make_loop_iterate_ntimes.
13484 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13485 (vect_gen_vector_loop_niters): Update declarations after above changes.
13487 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13489 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13490 128-bit round to integer instructions.
13491 (ceil<mode>2): Likewise.
13492 (btrunc<mode>2): Likewise.
13493 (round<mode>2): Likewise.
13495 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13497 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13498 unaligned VSX load/store on P8/P9.
13499 (expand_block_clear): Allow the use of unaligned VSX
13500 load/store on P8/P9.
13502 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13504 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13506 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13507 swap associated with both a load and a store.
13509 2018-01-02 Andrew Waterman <andrew@sifive.com>
13511 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13512 * config/riscv/riscv.md (clear_cache): Use it.
13514 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13516 * web.c: Remove out-of-date comment.
13518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13520 * expr.c (fixup_args_size_notes): Check that any existing
13521 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13522 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13523 (emit_single_push_insn): ...here.
13525 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13527 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13528 (const_vector_encoded_nelts): New function.
13529 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13530 (const_vector_int_elt, const_vector_elt): Declare.
13531 * emit-rtl.c (const_vector_int_elt_1): New function.
13532 (const_vector_elt): Likewise.
13533 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13534 of CONST_VECTOR_ELT.
13536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13538 * expr.c: Include rtx-vector-builder.h.
13539 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13540 directly on the tree encoding.
13541 (const_vector_from_tree): Likewise.
13542 * optabs.c: Include rtx-vector-builder.h.
13543 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13544 sequence of "u" values.
13545 * vec-perm-indices.c: Include rtx-vector-builder.h.
13546 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13547 directly on the vec_perm_indices encoding.
13549 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13551 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13552 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13553 * rtx-vector-builder.h: New file.
13554 * rtx-vector-builder.c: Likewise.
13555 * rtl.h (rtx_def::u2): Add a const_vector field.
13556 (CONST_VECTOR_NPATTERNS): New macro.
13557 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13558 (CONST_VECTOR_DUPLICATE_P): Likewise.
13559 (CONST_VECTOR_STEPPED_P): Likewise.
13560 (CONST_VECTOR_ENCODED_ELT): Likewise.
13561 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13562 (unwrap_const_vec_duplicate): Likewise.
13563 (const_vec_series_p): Check for a non-duplicated vector encoding.
13564 Say that the function only returns true for integer vectors.
13565 * emit-rtl.c: Include rtx-vector-builder.h.
13566 (gen_const_vec_duplicate_1): Delete.
13567 (gen_const_vector): Call gen_const_vec_duplicate instead of
13568 gen_const_vec_duplicate_1.
13569 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13570 (gen_const_vec_duplicate): Use rtx_vector_builder.
13571 (gen_const_vec_series): Likewise.
13572 (gen_rtx_CONST_VECTOR): Likewise.
13573 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13574 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13575 Build a new vector rather than modifying a CONST_VECTOR in-place.
13576 (handle_special_swappables): Update call accordingly.
13577 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13578 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13579 Build a new vector rather than modifying a CONST_VECTOR in-place.
13580 (handle_special_swappables): Update call accordingly.
13582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13584 * simplify-rtx.c (simplify_const_binary_operation): Use
13585 CONST_VECTOR_ELT instead of XVECEXP.
13587 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13589 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13590 the selector elements to be different from the data elements
13591 if the selector is a VECTOR_CST.
13592 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13593 ssizetype for the selector.
13595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13597 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13598 before testing each element individually.
13599 * tree-vect-generic.c (lower_vec_perm): Likewise.
13601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13603 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13604 * selftest-run-tests.c (selftest::run_tests): Call it.
13605 * vector-builder.h (vector_builder::operator ==): New function.
13606 (vector_builder::operator !=): Likewise.
13607 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13608 (vec_perm_indices::all_from_input_p): New function.
13609 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13610 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13611 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13612 instead of reading the VECTOR_CST directly. Detect whether both
13613 vector inputs are the same before constructing the vec_perm_indices,
13614 and update the number of inputs argument accordingly. Use the
13615 utility functions added above. Only construct sel2 if we need to.
13617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13619 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13620 the broadcast of the low byte.
13621 (expand_mult_highpart): Use an explicit encoding for the permutes.
13622 * optabs-query.c (can_mult_highpart_p): Likewise.
13623 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13624 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13625 (vectorizable_bswap): Likewise.
13626 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13627 explicit encoding for the power-of-2 permutes.
13628 (vect_permute_store_chain): Likewise.
13629 (vect_grouped_load_supported): Likewise.
13630 (vect_permute_load_chain): Likewise.
13632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13634 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13635 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13636 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13637 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13638 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13639 (vect_gen_perm_mask_any): Likewise.
13641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13643 * int-vector-builder.h: New file.
13644 * vec-perm-indices.h: Include int-vector-builder.h.
13645 (vec_perm_indices): Redefine as an int_vector_builder.
13646 (auto_vec_perm_indices): Delete.
13647 (vec_perm_builder): Redefine as a stand-alone class.
13648 (vec_perm_indices::vec_perm_indices): New function.
13649 (vec_perm_indices::clamp): Likewise.
13650 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13651 (vec_perm_indices::new_vector): New function.
13652 (vec_perm_indices::new_expanded_vector): Update for new
13653 vec_perm_indices class.
13654 (vec_perm_indices::rotate_inputs): New function.
13655 (vec_perm_indices::all_in_range_p): Operate directly on the
13656 encoded form, without computing elided elements.
13657 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13658 encoding. Update for new vec_perm_indices class.
13659 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13660 the given vec_perm_builder.
13661 (expand_vec_perm_var): Update vec_perm_builder constructor.
13662 (expand_mult_highpart): Use vec_perm_builder instead of
13663 auto_vec_perm_indices.
13664 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13665 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13666 or double series encoding as appropriate.
13667 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13668 vec_perm_indices instead of auto_vec_perm_indices.
13669 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13670 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13671 (vect_permute_store_chain): Likewise.
13672 (vect_grouped_load_supported): Likewise.
13673 (vect_permute_load_chain): Likewise.
13674 (vect_shift_permute_load_chain): Likewise.
13675 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13676 (vect_transform_slp_perm_load): Likewise.
13677 (vect_schedule_slp_instance): Likewise.
13678 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13679 (vectorizable_mask_load_store): Likewise.
13680 (vectorizable_bswap): Likewise.
13681 (vectorizable_store): Likewise.
13682 (vectorizable_load): Likewise.
13683 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13684 vec_perm_indices instead of auto_vec_perm_indices. Use
13685 tree_to_vec_perm_builder to read the vector from a tree.
13686 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13687 vec_perm_builder instead of a vec_perm_indices.
13688 (have_whole_vector_shift): Use vec_perm_builder and
13689 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13690 truncation to calc_vec_perm_mask_for_shift.
13691 (vect_create_epilog_for_reduction): Likewise.
13692 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13693 from auto_vec_perm_indices to vec_perm_indices.
13694 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13695 instead of changing individual elements.
13696 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13697 the vector in d.perm.
13698 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13699 from auto_vec_perm_indices to vec_perm_indices.
13700 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13701 instead of changing individual elements.
13702 (arm_vectorize_vec_perm_const): Use new_vector to install
13703 the vector in d.perm.
13704 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13705 Update vec_perm_builder constructor.
13706 (rs6000_expand_interleave): Likewise.
13707 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13708 (rs6000_expand_interleave): Likewise.
13710 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13712 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13713 to qimode could truncate the indices.
13714 * optabs.c (expand_vec_perm_var): Likewise.
13716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13718 * Makefile.in (OBJS): Add vec-perm-indices.o.
13719 * vec-perm-indices.h: New file.
13720 * vec-perm-indices.c: Likewise.
13721 * target.h (vec_perm_indices): Replace with a forward class
13723 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13724 * optabs.h: Include vec-perm-indices.h.
13725 (expand_vec_perm): Delete.
13726 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13727 (expand_vec_perm_const): Declare.
13728 * target.def (vec_perm_const_ok): Replace with...
13729 (vec_perm_const): ...this new hook.
13730 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13731 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13732 * doc/tm.texi: Regenerate.
13733 * optabs.def (vec_perm_const): Delete.
13734 * doc/md.texi (vec_perm_const): Likewise.
13735 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13736 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13737 expand_vec_perm for constant permutation vectors. Assert that
13738 the mode of variable permutation vectors is the integer equivalent
13739 of the mode that is being permuted.
13740 * optabs-query.h (selector_fits_mode_p): Declare.
13741 * optabs-query.c: Include vec-perm-indices.h.
13742 (selector_fits_mode_p): New function.
13743 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13744 is defined, instead of checking whether the vec_perm_const_optab
13745 exists. Use targetm.vectorize.vec_perm_const instead of
13746 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13747 fit in the vector mode before using a variable permute.
13748 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13749 vec_perm_indices instead of an rtx.
13750 (expand_vec_perm): Replace with...
13751 (expand_vec_perm_const): ...this new function. Take the selector
13752 as a vec_perm_indices rather than an rtx. Also take the mode of
13753 the selector. Update call to shift_amt_for_vec_perm_mask.
13754 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13755 Use vec_perm_indices::new_expanded_vector to expand the original
13756 selector into bytes. Check whether the indices fit in the vector
13757 mode before using a variable permute.
13758 (expand_vec_perm_var): Make global.
13759 (expand_mult_highpart): Use expand_vec_perm_const.
13760 * fold-const.c: Includes vec-perm-indices.h.
13761 * tree-ssa-forwprop.c: Likewise.
13762 * tree-vect-data-refs.c: Likewise.
13763 * tree-vect-generic.c: Likewise.
13764 * tree-vect-loop.c: Likewise.
13765 * tree-vect-slp.c: Likewise.
13766 * tree-vect-stmts.c: Likewise.
13767 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13769 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13770 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13771 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13772 (aarch64_vectorize_vec_perm_const): ...this new function.
13773 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13774 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13775 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13776 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13777 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13778 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13779 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13781 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13782 check for NEON modes.
13783 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13784 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13785 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13786 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13788 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13789 the old VEC_PERM_CONST conditions.
13790 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13791 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13792 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13793 (ia64_vectorize_vec_perm_const_ok): Merge into...
13794 (ia64_vectorize_vec_perm_const): ...this new function.
13795 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13796 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13797 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13798 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13799 * config/mips/mips.c (mips_expand_vec_perm_const)
13800 (mips_vectorize_vec_perm_const_ok): Merge into...
13801 (mips_vectorize_vec_perm_const): ...this new function.
13802 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13803 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13804 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13805 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13806 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13807 (rs6000_expand_vec_perm_const): Delete.
13808 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13810 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13811 (altivec_expand_vec_perm_const_le): Take each operand individually.
13812 Operate on constant selectors rather than rtxes.
13813 (altivec_expand_vec_perm_const): Likewise. Update call to
13814 altivec_expand_vec_perm_const_le.
13815 (rs6000_expand_vec_perm_const): Delete.
13816 (rs6000_vectorize_vec_perm_const_ok): Delete.
13817 (rs6000_vectorize_vec_perm_const): New function.
13818 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13819 an element count and rtx array.
13820 (rs6000_expand_extract_even): Update call accordingly.
13821 (rs6000_expand_interleave): Likewise.
13822 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13823 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13824 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13825 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13826 (rs6000_expand_vec_perm_const): Delete.
13827 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13828 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13829 (altivec_expand_vec_perm_const_le): Take each operand individually.
13830 Operate on constant selectors rather than rtxes.
13831 (altivec_expand_vec_perm_const): Likewise. Update call to
13832 altivec_expand_vec_perm_const_le.
13833 (rs6000_expand_vec_perm_const): Delete.
13834 (rs6000_vectorize_vec_perm_const_ok): Delete.
13835 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13836 reference to the SPE evmerge intructions.
13837 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13838 an element count and rtx array.
13839 (rs6000_expand_extract_even): Update call accordingly.
13840 (rs6000_expand_interleave): Likewise.
13841 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13842 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13844 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13848 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13849 vector mode and that that mode matches the mode of the data
13851 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13852 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13853 directly using expand_vec_perm_1 when forcing selectors into
13855 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13859 * optabs-query.h (can_vec_perm_p): Delete.
13860 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13861 * optabs-query.c (can_vec_perm_p): Split into...
13862 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13863 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13864 particular selector is valid.
13865 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13866 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13867 (vect_grouped_load_supported): Likewise.
13868 (vect_shift_permute_load_chain): Likewise.
13869 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13870 (vect_transform_slp_perm_load): Likewise.
13871 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13872 (vectorizable_bswap): Likewise.
13873 (vect_gen_perm_mask_checked): Likewise.
13874 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13875 implementations of variable permutation vectors into account
13876 when deciding which selector to use.
13877 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13878 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13879 with a false third argument.
13880 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13881 to test whether the constant selector is valid and can_vec_perm_var_p
13882 to test whether a variable selector is valid.
13884 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13886 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13887 * optabs-query.c (can_vec_perm_p): Likewise.
13888 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13889 instead of vec_perm_indices.
13890 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13891 (vect_gen_perm_mask_checked): Likewise,
13892 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13893 (vect_gen_perm_mask_checked): Likewise,
13895 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13897 * optabs-query.h (qimode_for_vec_perm): Declare.
13898 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13899 (qimode_for_vec_perm): ...this new function.
13900 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13902 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13904 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13905 does not have a conditional at the top.
13907 2018-01-02 Richard Biener <rguenther@suse.de>
13909 * ipa-inline.c (big_speedup_p): Fix expression.
13911 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13914 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13917 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13921 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13922 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13923 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13924 cond_taken_branch_cost 3->4.
13926 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13928 PR tree-optimization/83581
13929 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13930 TODO_cleanup_cfg if any changes have been made.
13932 PR middle-end/83608
13933 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13934 convert_modes if target mode has the right side, but different mode
13937 PR middle-end/83609
13938 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13939 last argument when extracting from CONCAT. If either from_real or
13940 from_imag is NULL, use expansion through memory. If result is not
13941 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13942 the parts directly to inner mode, if even that fails, use expansion
13945 PR middle-end/83623
13946 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13947 check for bswap in mode rather than HImode and use that in expand_unop
13950 Copyright (C) 2018 Free Software Foundation, Inc.
13952 Copying and distribution of this file, with or without modification,
13953 are permitted in any medium without royalty provided the copyright
13954 notice and this notice are preserved.